American Journal of Epidemiology

Wilcox, Allen J.; Szklo, Moyses; Comstock, George W.; Fox, Jaclyn

doi: N/Apmid: N/A

Wilcox, Allen J.;Szklo, Moyses;Comstock, George W.;Fox, Jaclyn

doi: 10.1093/aje/153.8.721pmid: N/A

AJE, American Journal of Epidemiology The American Journal of Epidemiology (AJE) is the most highly cited journal in the field of public health. It is a staple of health libraries throughout the United States and Europe. Beyond the developed world, however, the picture is sharply different. In the poorest countries, where epidemiologists deal with some of the most difficult public health problems, AJE is nearly inaccessible. Among the 200 countries in the world, there are 63 listed by the World Bank as “low income.” These hold more than half of the world's population, yet they receive 5 percent of AJE's institutional subscriptions. AJE is pleased to announce a partnership to bridge this barrier. In alliance with our publisher (Oxford University Press), our owner (Johns Hopkins School of Public Health), and our affiliated scientific society (Society for Epidemiologic Research), AJE is launching a program to make electronic subscriptions available at no cost to nonprofit institutions in the poor countries of the world. Citizens of developed nations may not fully realize how widely dispersed electronic communication has become in the less developed nations. In settings where paper is scarce and postal delivery uncertain, access to the Internet flourishes. AJE started in November 2000 to publish its issues on the World Wide Web through HighWire Press, an Internet service. In February 2001, AJE began a pilot project to make this electronic, full-text subscription free to institutions in five of the world's poorest countries—Ethiopia, Nigeria, Indonesia, Pakistan, and Nicaragua. These subscriptions are dubbed “Jubilee Subscriptions,” in concert with the international Jubilee movement to forgive Third-World debt. Electronic subscriptions are awarded for a 2-year period and will be renewable. All nonprofit organizations in these countries are eligible, including government health agencies, nonprofit agencies, and public universities. Our next step will be to expand eligibility to the remaining 58 low-income countries. The idea behind this program is neither original nor unique. The International Society for Environmental Epidemiology already provides deeply discounted membership (with a subscription to Epidemiology) to anyone living outside the 32 wealthiest nations. An increasing number of scientific journals provide online access to their current issues. The more radical step of unlimited free electronic subscriptions was proposed in 1997 by Ron LaPorte (1), who suggested that Internet servers could, in principle, make scientific journals easily available to the underdeveloped nations of the world. As AJE proceeds to make this leap, we recognize that increased access to the scientific literature is only a small step in addressing the public health problems of the developing world. However, at the very least, Jubilee subscriptions will extend the community of AJE's readership. We hope this can help to relieve the burden of professional isolation that too often comes with work in developing countries. We welcome this closer connection to colleagues who are tackling some of the world's toughest public health challenges. We also recognize that this connection is not a one-way street. Perspectives from the developing world (through the publication of research reports, commentaries, and correspondence) can help to foster in the developed world a more resilient and comprehensive understanding of the role of epidemiology. We ask our readers' assistance in making the availability of Jubilee subscriptions widely known. E-mail your colleagues in Ethiopia, Nigeria, Indonesia, Pakistan, and Nicaragua to let them know about the Jubilee pilot program. A full description is available through the AJE website (http://www.aje.oupjournals.org/). As the Jubilee program is expanded to more countries, we will ask your help in spreading the word. Finally, as other public health journals move into electronic publication, we urge them to consider creating their own Jubilee subscriptions. If a core of public health journals make this step, an even wider circle of medical journals might consider it as well. It is seldom that we have the opportunity to do so much good for so little cost. References 1. LaPorte RE. Internet server with targeted access would cure information deficiency in developing countries. The Global Health Network. (Letter). BMJ  1997; 314: 980. Google Scholar

Thomas, David B.; Ray, Roberta M.; Koetsawang, Amorn; Kiviat, Nancy; Kuypers, Jane; Qin, Qin; Ashley, Rhoda L.; Koetsawang, Suporn

doi: N/Apmid: N/A

Personal interviews, tests for antibodies to herpes simplex virus type 2, Treponema pallidum, and hepatitis B, tests for hepatitis B surface antigen (HBsAg), and polymerase chain reaction-based assays for human papillomavirus (HPV) DNA in cervical scrapings were obtained from 190 women with squamous cell and 42 women with adenomatous cervical carcinoma and from 291 hospitalized controls diagnosed in Bangkok, Thailand, between September 1991 and September 1993. Risk was strongly associated with oncogenic HPV types, with types 16 and 18 predominating in squamous and adenomatous lesions, respectively. The 126 cases with HPV-16 and the 42 cases with HPV-18 were compared with 250 controls with no evidence of any HPV. The risk of both viral tumor types increased with decreasing age at first intercourse in this predominantly monogamous population, which may be explained by more visits to prostitutes by the husbands of cases with early than late age at first intercourse. HPV-16 tumors were weakly associated with HBsAg carrier state and smoking. The risk of tumors of both viral types increased with parity and use of oral contraceptives but not with injectable progestogens. Factors that may predispose to persistent, oncogenic HPV-16 or -18 infection may include estrogens or progestins in the presence of estrogens, immunosuppression, and smoking, but other factors related to low socioeconomic status are also involved.

Thomas, David B.;Ray, Roberta M.;Koetsawang, Amorn;Kiviat, Nancy;Kuypers, Jane;Qin, Qin;Ashley, Rhoda L.;Koetsawang, Suporn

doi: 10.1093/aje/153.8.723pmid: 11296143

Abstract Personal interviews, tests for antibodies to herpes simplex virus type 2, Treponema pallidum, and hepatitis B, tests for hepatitis B surface antigen (HBsAg), and polymerase chain reaction-based assays for human papillomavirus (HPV) DNA in cervical scrapings were obtained from 190 women with squamous cell and 42 women with adenomatous cervical carcinoma and from 291 hospitalized controls diagnosed in Bangkok, Thailand, between September 1991 and September 1993. Risk was strongly associated with oncogenic HPV types, with types 16 and 18 predominating in squamous and adenomatous lesions, respectively. The 126 cases with HPV-16 and the 42 cases with HPV-18 were compared with 250 controls with no evidence of any HPV. The risk of both viral tumor types increased with decreasing age at first intercourse in this predominantly monogamous population, which may be explained by more visits to prostitutes by the husbands of cases with early than late age at first intercourse. HPV-16 tumors were weakly associated with HBsAg carrier state and smoking. The risk of tumors of both viral types increased with parity and use of oral contraceptives but not with injectable progestogens. Factors that may predispose to persistent, oncogenic HPV-16 or -18 infection may include estrogens or progestins in the presence of estrogens, immunosuppression, and smoking, but other factors related to low socioeconomic status are also involved. adenocarcinoma, carcinoma, adenosquamous, carcinoma, squamous cell, cervix neoplasms, papillomavirus, human HBsAg, hepatitis B surface antigen, HBV, hepatitis B virus, HIV, human immunodeficiency virus, HPV, human papillomavirus, HSV, herpes simplex virus, IUD, intrauterine device In 1995 a working group of the International Agency for Research on Cancer concluded that human papillomavirus (HPV) types 16 and 18 are carcinogenic to humans, that types 31 and 33 are probably carcinogenic to humans, and that some other types are possibly carcinogenic to humans (1). Epidemiologic evidence that these viruses are causes of invasive cervical carcinoma includes the following: the sexual mode of transmission (1–5), which provides an explanation for the sexual risk factors for cervical cancer; case-control studies showing strong associations of cervical cancer with specific HPV types based on demonstration of HPV DNA in cervical epithelial cells (1, 6–8) and serologic tests for HPV-16 antibodies (9); and historical prospective studies showing increased risks of subsequent invasive disease in women with antibodies to HPV-16 (10–12) and HPV-18 (10) and in women with evidence of HPV DNA types 16, 18, and others in cervical smears (13, 14). In addition, a clear mechanism for the carcinogenic effect of the viruses has been elucidated; the viral genes E6 and E7 from oncogenic types of HPV bind to and inactivate the protein products of the tumor suppressor genes p53 and Rb, respectively (15, 16). Although cervical carcinoma may arise in the absence of HPV (17), the preponderance of evidence suggests that oncogenic types of these viruses are necessary causes of cervical carcinoma (18, 19). However, many more women are infected by oncogenic types of HPV sometime during their lives than ultimately develop cervical carcinoma (1), suggesting an important role for cofactors. Case-control studies have consistently shown, in HPV-positive women, associations of various measures of socioeconomic status and use of oral contraceptives with risk of invasive cervical carcinoma (6–8, 20), including both squamous cell (6, 7) and adenomatous types (6, 7). Other cofactors that have been investigated have only inconsistently been associated with risk of invasive disease, possibly because of variations in the sensitivity of the polymerase chain reaction-based assays used to detect oncogenic HPV DNA in normal exfoliated cells of controls, variations in expression of HPV in controls due to variation in the prevalence of factors that alter this expression, and differences in the proportion of positive assays that represent persistent and transient infections (2, 21). In this first of three papers, we present results of an investigation of possible differences in the epidemiologic features of HPV-16- and HPV-18-associated carcinomas. To maximize the chances of identifying variables associated with any phase of the carcinogenic process, cases were compared with controls with no evidence of HPV in cervical scrapings. In the second paper (22), we present results of an attempt to identify cofactors associated with progression from in situ to invasive disease. The third paper (23) addresses the role of male sexual behavior in the transmission of oncogenic types of HPV, the role of commercial sex workers as reservoirs of these viruses, risk factors for HPV infection in this heavily exposed group, and the role of the human immunodeficiency virus (HIV) and other variables as possible cofactors with HPV in the early stages of the carcinogenic process. MATERIALS AND METHODS Cases eligible for these studies were women who were admitted to public wards of Siriraj Hospital in Bangkok, Thailand, with a new histologically confirmed diagnosis of either in situ or invasive cervical carcinoma between September 1991 and September 1993, who were born in 1930 or later, and who resided for at least the past year in Thailand. Older women were not included because they would have been unlikely to have used steroid contraceptives, a possible cofactor of interest. An attempt was made to select two controls for each woman with invasive disease from one otolaryngology and two general surgery wards. Wards were visited on a rotating basis to select the first two women who were admitted during the previous 24 hours, who were in the same 5-year age group, and who resided in the same region of the country as the corresponding case. Women admitted for treatment of conditions that have been associated with the use of steroid contraceptives, including circulatory or cardiovascular diseases, diabetes, chronic renal disease, benign breast disease, a previously diagnosed cancer, chronic liver disease, or any obstetric or gynecologic condition, were not selected, although women with a history of these conditions were eligible. In most instances, if an eligible woman refused to participate, the selection process was continued until two controls were found for each case. For the first 50 cases of invasive cervical cancer, one control woman admitted for a hysterectomy for a noncancerous condition (hysterectomy control) was also selected and matched to the corresponding case on 5-year age group and region of residence. All cases and controls were interviewed in the hospital to obtain information on sexual and contraceptive practices, reproductive history, prior cervical smears, use of tobacco and alcohol, and indices of socioeconomic status and use of medical services. As detailed elsewhere (23), attempts were also made to interview the husbands of currently married cases and controls. All interviewed women were requested to donate a 15-ml blood specimen. The resultant serum was stored in four aliquots at −70°C; two aliquots were retained locally, and two were shipped to Seattle, Washington, on dry ice. Cervical scrapings for HPV DNA assays were obtained from the hysterectomy controls and the cases prior to surgery or radiotherapy by the patient's surgeon. The cervix, including the cervical os, was scraped with a Teflon (E. I. du Pont de Nemours and Company, Wilmington, Delaware)-coated swab. The end of the swab was broken off into a vial containing 2 ml of specimen transport medium (Digene Diagnostics, Inc., Beltsville, Maryland) and frozen at −70°C. Tumor tissue specimens were also obtained from each case at the time of surgery (hysterectomy or conization) or by punch biopsy prior to radiation treatment for women with inoperable disease. Normal cervical tissue was similarly obtained at surgery from the control women having hysterectomies. All scrapings and tissue specimens were frozen at −70°C and shipped to Seattle on dry ice. The same surgeons who treated the cases took cervical smears and cervical scrapings for HPV DNA assays from the control women while they were hospitalized. The cervical smears were read locally, and any woman found to have a high-grade cervical intraepithelial lesion was considered ineligible for inclusion in this study as a control. Histologic slides from the blocks that were used to make the diagnosis in the cases and comparable sections from two blocks from the hysterectomy controls were stained with hematoxylin and eosin and read by a collaborating pathologist in Bangkok who also provided information on the source of the specimen, tumor size, and stage. Invasive carcinomas were coded according to the World Health Organization's “Histological Typing of Female Genital Tract Tumours” (24) as squamous cell carcinoma, adenocarcinoma, or adenosquamous carcinoma. The enzyme-linked immunosorbent assay procedure was used to test for antibodies to HIV, and those that tested positive were confirmed by Western blot (25). All samples were assayed for anti-hepatitis B core antibodies, anti-hepatitis B surface antibody, and hepatitis B surface antigen (HBsAg) (26). The standard venereal disease research laboratory test for syphilis was performed on all samples, and those that tested positive were confirmed using the microhemagglutination assay for antibodies to Treponema pallidum (27). Antibodies to herpes simplex virus types 1 and 2 (HSV-1 and HSV-2) were confirmed by Western blot assay (28, 29). The samples for HPV DNA assays were digested with proteinase K, and the DNA was precipitated by ethanol and suspended in Tris-ethylenediaminetetraacetic acid (EDTA) buffer. The DNA from 20 μl of the original sample was analyzed in duplicate by polymerase chain reaction amplification using two sets of primers. Consensus primers MY09 and MY11 target an approximately 450-base pair region of the HPV L1 gene, and PC04 and GH20 target a 268-base pair region of the human β-globin gene (30). The presence of the β-globin fragment on ethidium bromide-stained gels after electrophoresis of the polymerase chain reaction amplicons was used to assess whether the samples contained adequate DNA and whether polymerase chain reaction inhibitors were present. Samples negative for the β- globin fragment were considered unsatisfactory. After amplification, 3 μl of each amplicon were spotted onto nylon filters and hybridized with a biotin-labeled generic HPV probe and with five mixtures of biotin-labeled, type-specific oligonucleotide probes for HPV types 6 and 11, type 16, type 18, type 31, 33, 35, and 39, and type 45. Samples hybridizing with the generic probe but not with any of the type-specific probes were considered positive for unclassified HPV. Each polymerase chain reaction run included the following controls: DNA from 400 SiHa cells, an HPV type 16-positive cell line; DNA from K562 cells, an HPV-negative human cell line; K562 cells in specimen transport medium, handled concurrently with the patient samples; and a reagent blank (no DNA). All negative controls were negative for HPV DNA by polymerase chain reaction. For this report, women with invasive disease were compared with the hospital controls, excluding the hysterectomy controls. Odds ratios, as estimates of relative risks, were calculated using unconditional logistic regression models (31). RESULTS Adenocarcinomas and adenosquamous carcinomas have been combined into a single group, hereafter referred to as “adenomatous.” Slides from 141 cases diagnosed in Bangkok as invasive cervical carcinoma were read again in Seattle. Of 118 cases originally diagnosed as squamous cell, 116 (92.8 percent) were so diagnosed in Seattle, and of 23 cases diagnosed in Bangkok as adenomatous, 19 (73.1 percent) were similarly confirmed. Because of this satisfactory level of concordance between the two pathologists, slides from some cases were not read in Seattle, and for the purposes of this report, the diagnosis made in Bangkok was used to classify tumors as to histologic type. Of the 420 eligible cases identified, 338 (80.5 percent) were interviewed. Cervical scrapings were obtained from 337 (99.7 percent) of those interviewed, 216 of whom had invasive squamous and 46 of whom had invasive adenomatous cervical carcinoma. The remaining 75 cases had carcinoma in situ. The cervical samples from 191 and 42 of the invasive squamous and adenomatous cases, respectively, contained adequate amounts of DNA for assay. One woman who claimed no history of sexual intercourse was omitted from the analysis, leaving 190 women with invasive squamous cell and 42 women with adenomatous cervical carcinoma in the analyses for this report. Of the 614 women selected as controls, 490 (79.8 percent) were interviewed. Cervical scrapings were obtained from 306 (62.4 percent) of these 490 women, 298 (97.4 percent) of which yielded adequate samples of DNA. Three additional women with no history of sexual intercourse, one woman with a subtotal hysterectomy, and three with ineligible diagnoses were subsequently excluded, leaving 291 controls in the analyses for this report. The age distributions of the cases and controls are shown in table 1. TABLE 1. Age distributions of invasive squamous cell and adenomatous cervical carcinoma cases, cases with HPV-16* and HPV-18* DNA, and controls, Bangkok, Thailand, 1991–1993† Age (years)  Squamous   Adenomatous   HPV-16   HPV-18   Controls   No.  %  No.  %  No.  %  No.  %  No.  %  ≤30  3  1.6  1  2.4  2  1.6  1  2.4  13  4.5  31–35  23  12.1  8  19.0  21  16.7  3  7.1  44  15.1  36–40  36  18.9  8  19.0  25  19.8  10  23.8  56  19.2  41–45  33  17.4  14  33.3  26  20.6  9  21.4  60  20.6  46–50  30  15.8  4  9.5  13  10.3  9  21.4  36  12.4  >50  65  34.2  7  16.7  39  31.0  10  23.8  82  28.2  Total  190  100.0  42  99.9  126  100.0  42  99.9  291  100.0  Age (years)  Squamous   Adenomatous   HPV-16   HPV-18   Controls   No.  %  No.  %  No.  %  No.  %  No.  %  ≤30  3  1.6  1  2.4  2  1.6  1  2.4  13  4.5  31–35  23  12.1  8  19.0  21  16.7  3  7.1  44  15.1  36–40  36  18.9  8  19.0  25  19.8  10  23.8  56  19.2  41–45  33  17.4  14  33.3  26  20.6  9  21.4  60  20.6  46–50  30  15.8  4  9.5  13  10.3  9  21.4  36  12.4  >50  65  34.2  7  16.7  39  31.0  10  23.8  82  28.2  Total  190  100.0  42  99.9  126  100.0  42  99.9  291  100.0  * HPV-16 and HPV-18, human papillomavirus types 16 and 18, respectively. † The mean and median ages (years) are the following: squamous cell carcinoma, 45.5 and 45.5; adenomatous cervical carcinoma, 42.4 and 42.5; HPV-16, 44.3 and 43.0; HPV-18, 44.3 and 44.0; controls, 43.9 and 43.0, respectively. View Large The 306 controls with cervical scrapings and the 184 interviewed control women without cervical scrapings did not differ appreciably by age or number of sexual partners. A higher proportion of women with scrapings had used oral contraceptives (43.4 percent vs. 38.0 percent), injectable contraceptives (20.9 percent vs. 16.3 percent), and an intrauterine device (IUD) (19.9 percent vs. 13.0 percent) and had had at least one cervical smear (34.0 percent vs. 20.6 percent), which would result in spuriosly low odds ratios in relation to these factors. Fewer women with scrapings had their first sexual experience before age 18 (24.8 percent vs. 34.2 percent), which would result in an overestimate of the strength of the association with this variable. Results of HPV DNA assays on a sample of the tissue specimens and scrapings from the same individuals were compared. Of the 30 paired samples from cases, seven were positive for HPV DNA type 16 or 18 in scrapings but negative in the corresponding tumor tissue, whereas none that tested negative in scrapings were positive in tumor tissue. Of 45 tested hysterectomy controls with normal cervixes, only one tested positive for oncogenic HPV DNA (type 16) and only in the scraping. Results of assays of exfoliated cells were therefore the more sensitive indicators of the presence of oncogenic HPV DNA and were used in the analyses for this report. No study subjects were positive for HPV type 6 or type 11. As shown in table 2, 79 percent of the scrapings from the women with squamous cell carcinomas and 76 percent of the scrapings from the women with adenomatous carcinomas contained known oncogenic HPV DNA (types 16, 18, 31, 33, 35, 39, or 45), compared with only 3 percent of the scrapings from the controls. Additional samples contained untyped HPV DNA, and those with strong reactions in the generic probe are shown in the sixth line of the table. If weakly reactive tests in the generic probe are considered positive, then the numbers (and percentages) of women with squamous tumors, adenomatous tumors, and no tumors (controls) with any type of HPV DNA are 169 (89 percent), 35 (83.3 percent), and 41 (14.1 percent), respectively. TABLE 2. Age-adjusted odds ratios of invasive squamous cell and adenomatous cervical carcinoma in relation to specific types of human papillomaviruses, Bangkok, Thailand, 1991–1993 HPV type†  Positive cases   Positive controls (n = 291)   Squamous   Adenomatous   Squamous (n = 190)   Adenomatous (n = 42)   No.  %  No.  %  No.  %  OR*  95% CI*  OR  95% CI  16  114  60.0  14  33.3  6  2.1  83  39, 232  24  8.7, 76  18  26  13.7  18  42.9  0  0  ∞  22, ∞  ∞  165, ∞  31/33/35/39  25  13.2  3  7.1  3  1.0  14  4.9, 61  7.1  1.2, 44  45  0  0  1  2.4  0  0          Any oncogenic type‡  150  78.9  32  76.2  9  3.1  155  72, 385  106  41, 317  Untyped§  14  7.4  1  2.4  11  3.8  2.0  0.9, 4.7  0.6  0.03, 3.4  Any type  164  86.3  33  78.6  20  6.9  97  52, 193  49  21, 125  HPV type†  Positive cases   Positive controls (n = 291)   Squamous   Adenomatous   Squamous (n = 190)   Adenomatous (n = 42)   No.  %  No.  %  No.  %  OR*  95% CI*  OR  95% CI  16  114  60.0  14  33.3  6  2.1  83  39, 232  24  8.7, 76  18  26  13.7  18  42.9  0  0  ∞  22, ∞  ∞  165, ∞  31/33/35/39  25  13.2  3  7.1  3  1.0  14  4.9, 61  7.1  1.2, 44  45  0  0  1  2.4  0  0          Any oncogenic type‡  150  78.9  32  76.2  9  3.1  155  72, 385  106  41, 317  Untyped§  14  7.4  1  2.4  11  3.8  2.0  0.9, 4.7  0.6  0.03, 3.4  Any type  164  86.3  33  78.6  20  6.9  97  52, 193  49  21, 125  * HPV, human papillomavirus; OR, odds ratio in relation to women without the specific HPV type; CI, confidence interval. † No study subjects tested positive for HPV type 6 or 11. ‡ Includes types 16/18/31/33/35/39/45. § Weakly reactive tests considered negative. View Large HPV-16 is most common in squamous cell tumors, and HPV-18 is the most frequent type in adenomatous carcinomas. These differences are unlikely due to chance (p < 0.001). The two tumor types did not differ significantly in the proportions with HPV types 31/33/35/39 (p = 0.36), untyped HPV DNA (p = 0.34), any HPV type (p = 0.17), or any oncogenic type (p = 0.52). The odds ratios for the oncogenic types are all very large with lower 95 percent confidence limits well above unity. The presence of untyped HPV DNA is not significantly associated with the risk of either squamous or adenomatous carcinoma. No other risk factor considered in this investigation was found to significantly distinguish adenomatous and squamous cell carcinomas (not shown), so the two histologic types were combined for the purpose of comparing risk factors for tumors with HPV types 16 and 18. Two cases were infected by both types 16 and 18 and have been eliminated from these analyses, leaving 126 cases with type 16 and 42 cases with type 18. Of the 126 cases with HPV-16, 113 had squamous and 13 had adenomatous carcinomas, and of the 42 tumors with HPV-18, 25 were squamous and 17 were adenomatous carcinomas. The age distributions of the women with HPV-16- and HPV-18-associated tumors are shown in table 1. These women were compared with the 250 controls with no evidence of any types of HPV; that is, the 41 controls with either any type of HPV DNA or a weakly reactive test in the generic probe were omitted from the analyses. The age distribution of these 250 controls did not differ appreciably from that of all controls shown in table 1. As shown in table 3, the risk of both a carcinoma with HPV-16 and a carcinoma with HPV-18 was weakly associated with having more than one sexual partner. The risk also tended to increase with decreasing age at first intercourse. However, of the cases with HPV-16 whose husbands were interviewed, 10 (58.5 percent) of 17 women who first had sexual intercourse before age 19 had husbands who had visited prostitutes over 80 times, compared with eight (36.4 percent) of 22 women who became sexually active at a later age (chi-square = 1.4; p = 0.24). The relation of the risk of HPV-16-associated tumors to age at first intercourse was not altered by controlling for months since the last cervical smear, number of pregnancies, use of oral contraceptives, and use of an IUD (not shown). Similar adjustments could not be made in the analyses of the HPV-18-associated tumors because of the small numbers of cases. The odds ratios adjusted only for age are therefore presented in this and subsequent tables to provide comparable estimates for both tumor types. With the possible exception of an increase in risk of a carcinoma with HPV-16 in women with HSV-2 antibodies, no significant associations were observed with serologic evidence of prior infection with HSV-1 (not shown), genital herpes simplex, syphilis, or hepatitis B, and the associations with the sexual variables shown were generally similar for carcinomas with HPV-16 and -18. Only one case and no controls were HIV positive. Although quite possibly due to chance, the odds ratio of tumors with HPV-16 is higher in women with HBsAg than in women who have anti-HBV antibodies. TABLE 3. Age-adjusted odds ratios of HPV-16*- and HPV-18*-associated invasive cervical carcinoma in relation to sexual factors and infectious agents, Bangkok, Thailand, 1991–1993 Factor/agent  Cases†   Controls (no HPV*) (no.)  HPV-16   HPV-18   HPV-16 (no.)  HPV-18 (no.)  OR*  95% CI*  OR  95% CI  No. of sexual partners                 1  116  39  237  1.0  Reference  1.0  Reference   >1  10  3  13  1.6  0.7, 3.7  1.5  0.4, 5.5  Age (years) at first intercourse                ≤16  23  1  21  1.4  2.0, 9.7  0.7  0.1, 6.3   17–18  30  13  39  3.1  1.6, 6.3  4.6  1.6, 13.1   19–20  30  13  57  2.1  1.1, 4.1  2.9  1.0, 8.2   21–23  24  9  57  1.7  0.8, 3.3  2.1  0.7, 6.4  ≥24  19  6  76  1.0  Reference  1.0  Reference  HSV-2* antibodies‡                 Negative  55  24  117  1.0  Reference  1.0  Reference   Positive  69  17  111  1.6  1.0, 2.5  0.9  0.5, 1.7  Syphilis serology‡                 Negative  116  39  229  1.0  Reference  1.0  Reference   Positive  8  3  12  1.4  0.6, 3.5  1.5  0.4, 5.8  HBsAg*,‡                 Negative  113  41  227  1.0  Reference  1.0  Reference   Positive  11  1  12  1.9  0.8, 4.3  0.5  0.1, 3.6  Anti-HBs antibodies*,‡                 Negative  72  24  145  1.0  Reference  1.0  Reference   Positive  52  18  94  1.1  0.7, 1.8  1.0  0.5, 2.1  Anti-HBc antibodies*,‡                 Negative  44  19  103  1.0  Reference  1.0  Reference   Positive  80  23  136  1.4  0.9, 2.2  0.9  0.4, 1.7  Factor/agent  Cases†   Controls (no HPV*) (no.)  HPV-16   HPV-18   HPV-16 (no.)  HPV-18 (no.)  OR*  95% CI*  OR  95% CI  No. of sexual partners                 1  116  39  237  1.0  Reference  1.0  Reference   >1  10  3  13  1.6  0.7, 3.7  1.5  0.4, 5.5  Age (years) at first intercourse                ≤16  23  1  21  1.4  2.0, 9.7  0.7  0.1, 6.3   17–18  30  13  39  3.1  1.6, 6.3  4.6  1.6, 13.1   19–20  30  13  57  2.1  1.1, 4.1  2.9  1.0, 8.2   21–23  24  9  57  1.7  0.8, 3.3  2.1  0.7, 6.4  ≥24  19  6  76  1.0  Reference  1.0  Reference  HSV-2* antibodies‡                 Negative  55  24  117  1.0  Reference  1.0  Reference   Positive  69  17  111  1.6  1.0, 2.5  0.9  0.5, 1.7  Syphilis serology‡                 Negative  116  39  229  1.0  Reference  1.0  Reference   Positive  8  3  12  1.4  0.6, 3.5  1.5  0.4, 5.8  HBsAg*,‡                 Negative  113  41  227  1.0  Reference  1.0  Reference   Positive  11  1  12  1.9  0.8, 4.3  0.5  0.1, 3.6  Anti-HBs antibodies*,‡                 Negative  72  24  145  1.0  Reference  1.0  Reference   Positive  52  18  94  1.1  0.7, 1.8  1.0  0.5, 2.1  Anti-HBc antibodies*,‡                 Negative  44  19  103  1.0  Reference  1.0  Reference   Positive  80  23  136  1.4  0.9, 2.2  0.9  0.4, 1.7  * HPV-16 and HPV-18, human papillomavirus types 16 and 18, respectively; HPV, human papillomavirus; OR, odds ratio; CI, confidence interval; HSV-2, herpes simplex virus type 2; HBsAg, hepatitis B surface antigen; anti-HBs antibodies, anti-hepatitis B surface antibodies; anti-HBc antibodies, anti-hepatitis B core antibodies. † Two cases with both HPV-16 and HPV-18 were omitted from analyses. ‡ Study subjects with no serologic test results were omitted from analyses. View Large The risk of both types of carcinomas increased with the number of pregnancies (table 4). This association with HPV-16 tumors remained after adjustment for months since last cervical smear, age at first intercourse, use of oral contraceptives, and use of an IUD. Weak positive associations with both types were found with ever having had a spontaneous abortion and having used oral contraceptives, and a slight reduction in the risk of both types of tumors in users of depot-medroxyprogesterone acetate for contraception was seen, although all of these possible associations could readily have occurred by chance. No trends of increasing risk with duration of use of oral contraceptives were observed (not shown). The weak association between HPV-16-associated tumors and oral contraceptive use was not further attenuated by controlling for months since the last cervical smear, age at first sexual intercourse, and number of pregnancies. None of the odds ratios for any of the hormonal variables shown differed significantly for tumors associated with HPV types 16 and 18. TABLE 4. Age-adjusted odds ratios of HPV-16*- and HPV-18*-associated invasive cervical carcinoma in relation to hormonal factors, Bangkok, Thailand, 1991–1993 Factor  Cases   Controls (no HPV*) (no.)  HPV-16   HPV-18   HPV-16 (no.)  HPV-18 (no.)  OR*  95% CI*  OR  95% CI  No. of pregnancies                 0  1  0  9  0.3  0.04, 2.7†       1–2  31  7  97  1.0  Reference  1.0  Reference‡   3–4  49  24  81  2.1  1.1, 3.6  4.1  1.7, 10.3   5–8  40  10  56  2.8  1.5, 5.6  2.7  0.9, 8.2  ≥9  5  1  7  3.0  0.8, 11.1  2.5  0.2, 26.2  Ever had stillbirth§                 No  119  41  232  1.0  Reference  1.0  Reference   Yes  6  1  9  1.3  0.5, 3.8  0.5  0.1, 4.5  Ever had spontaneous abortion§                 No  99  29  200  1.0  Reference  1.0  Reference   Yes  26  13  41  1.3  0.7, 2.2  2.2  1.1, 4.7  Ever had induced abortion§                 No  118  41  234  1.0  Reference  1.0  Reference   Yes  7  1  7  2.0  0.7, 6.0  0.8  0.1, 7.1  Ever used oral contraceptives                 No  62  20  136  1.0  Reference  1.0  Reference   Yes  64  22  114  1.3  0.8, 2.0  1.4  0.7, 2.7  Ever used DMPA*                 No  103  35  195  1.0  Reference  1.0  Reference   Yes  23  7  55  0.8  0.5, 1.4  0.8  0.9, 1.9  Factor  Cases   Controls (no HPV*) (no.)  HPV-16   HPV-18   HPV-16 (no.)  HPV-18 (no.)  OR*  95% CI*  OR  95% CI  No. of pregnancies                 0  1  0  9  0.3  0.04, 2.7†       1–2  31  7  97  1.0  Reference  1.0  Reference‡   3–4  49  24  81  2.1  1.1, 3.6  4.1  1.7, 10.3   5–8  40  10  56  2.8  1.5, 5.6  2.7  0.9, 8.2  ≥9  5  1  7  3.0  0.8, 11.1  2.5  0.2, 26.2  Ever had stillbirth§                 No  119  41  232  1.0  Reference  1.0  Reference   Yes  6  1  9  1.3  0.5, 3.8  0.5  0.1, 4.5  Ever had spontaneous abortion§                 No  99  29  200  1.0  Reference  1.0  Reference   Yes  26  13  41  1.3  0.7, 2.2  2.2  1.1, 4.7  Ever had induced abortion§                 No  118  41  234  1.0  Reference  1.0  Reference   Yes  7  1  7  2.0  0.7, 6.0  0.8  0.1, 7.1  Ever used oral contraceptives                 No  62  20  136  1.0  Reference  1.0  Reference   Yes  64  22  114  1.3  0.8, 2.0  1.4  0.7, 2.7  Ever used DMPA*                 No  103  35  195  1.0  Reference  1.0  Reference   Yes  23  7  55  0.8  0.5, 1.4  0.8  0.9, 1.9  * HPV-16 and HPV-18, human papillomavirus types 16 and 18, respectively; HPV, human papillomavirus; OR, odds ratio; CI, confidence interval; DMPA, depot-medroxyprogesterone acetate. † p value of test for trend < 0.01. ‡ p value of test for trend = 0.19. § Analyses restricted to parous women. View Large The risks of both HPV-16 and HPV-18 tumor types were reduced in users of an IUD and in women who had had cervical smears, had ever had a chest radiographic examination, and had attended school (table 5). The reduced risk in relation to cervical smears was greatest in women whose last smear was within the past year. These relations with the HPV-16 tumor type were not altered by controlling for the other variables in the table or by the use of oral contraceptives, number of pregnancies, or age at first intercourse. No significant associations with tubal ligation or alcohol consumption were observed. Cigarette smoking was weakly associated with HPV type 16 tumors only, but the number of smokers in this study population was too small to allow control for other variables or for the absence of smokers among women with HPV-18 tumors to be very meaningful. With this possible exception, the tumors with HPV-16 and HPV-18 did not differ appreciably by any of the factors shown. TABLE 5. Age-adjusted odds ratios of HPV-16*- and HPV-18*-associated invasive cervical carcinoma in relation to nonsexual and nonhormonal factors, Bangkok, Thailand, 1991–1993 Factor  Cases   Controls (no HPV*) (no.)  HPV-16   HPV-18   HPV-16 (no.)  HPV-18 (no.)  OR*  95% CI*  OR  95% CI  Ever used an IUD*                 No  115  38  199  1.0  Reference  1.0  Reference   Yes  11  4  51  0.4  0.2, 0.7  0.4  0.1, 1.2  Tubal ligation                 No  90  25  164  1.0  Reference  1.0  Reference   Yes  36  17  86  0.8  0.5, 1.2  1.2  0.6, 2.3  Months since last cervical smear                 No cervical smear  112  35  165  1.0  Reference  1.0  Reference   1–12  5  1  44  0.2  0.1, 0.4  0.1  0.01, 0.7   >12  4  2  21  0.3  0.1, 0.8  0.4  0.1, 1.8   Unknown months  5  4  20  0.4  0.1, 1.0  0.9  0.3, 2.7  Ever smoked ≥100 cigarettes                 No  116  42  238  1.0  Reference  1.0  Reference   Yes†  9  0  12  1.6  0.6, 3.8      Ever drank alcoholic beverages                 No  61  17  125  1.0  Reference  1.0  Reference   Yes  65  25  125  1.1  0.7, 1.6  1.5  0.8, 2.9  Ever had chest radiographic examination                 No  99  36  171  1.0  Reference  1.0  Reference   Yes  27  6  79  0.6  0.4, 1.0  0.4  0.1, 0.9  Ever attended school                 No  21  6  22  1.0  Reference  1.0  Reference   Yes  105  36  228  0.5  0.3, 0.9  0.5  0.2, 1.5  Factor  Cases   Controls (no HPV*) (no.)  HPV-16   HPV-18   HPV-16 (no.)  HPV-18 (no.)  OR*  95% CI*  OR  95% CI  Ever used an IUD*                 No  115  38  199  1.0  Reference  1.0  Reference   Yes  11  4  51  0.4  0.2, 0.7  0.4  0.1, 1.2  Tubal ligation                 No  90  25  164  1.0  Reference  1.0  Reference   Yes  36  17  86  0.8  0.5, 1.2  1.2  0.6, 2.3  Months since last cervical smear                 No cervical smear  112  35  165  1.0  Reference  1.0  Reference   1–12  5  1  44  0.2  0.1, 0.4  0.1  0.01, 0.7   >12  4  2  21  0.3  0.1, 0.8  0.4  0.1, 1.8   Unknown months  5  4  20  0.4  0.1, 1.0  0.9  0.3, 2.7  Ever smoked ≥100 cigarettes                 No  116  42  238  1.0  Reference  1.0  Reference   Yes†  9  0  12  1.6  0.6, 3.8      Ever drank alcoholic beverages                 No  61  17  125  1.0  Reference  1.0  Reference   Yes  65  25  125  1.1  0.7, 1.6  1.5  0.8, 2.9  Ever had chest radiographic examination                 No  99  36  171  1.0  Reference  1.0  Reference   Yes  27  6  79  0.6  0.4, 1.0  0.4  0.1, 0.9  Ever attended school                 No  21  6  22  1.0  Reference  1.0  Reference   Yes  105  36  228  0.5  0.3, 0.9  0.5  0.2, 1.5  * HPV-16 and HPV-18, human papillomavirus types 16 and 18, respectively; HPV, human papillomavirus; OR, odds ratio; CI, confidence interval; IUD, intrauterine device. † One case with unknown smoking history was excluded. View Large DISCUSSION As with previous studies (18), oncogenic types of HPV DNA were demonstrated in a high proportion of the cases, with HPV-16 and HPV-18 predominating in squamous cell and adenomatous carcinomas, respectively, and with other oncogenic types (i.e., 31, 33, 35, 39, and 45) occurring in smaller and similar proportions of the tumors of both histologic types. Use of additional probes would not have altered these results, since only 15 cases and 11 controls were found to have untyped HPV DNA, and some of these are probably nononcogenic types since they occurred nearly as frequently in controls as in cases. The high proportion of tumors with HPV DNA supports the hypothesis that HPV infection is necessary for the development of cervical carcinoma. In accordance with prior investigations (1, 6, 8), the odds ratios for both squamous and adenomatous carcinomas, in relation to HPV types 16, 18, and 31/33/35/39, were very large. Moreover, in accordance with prior studies (32–36), the epidemiologic features of squamous and adenomatous carcinomas were similar, justifying our decision to combine the two types in the analyses. Although having more than one sexual partner was a risk factor for both HPV-16 and HPV-18 tumor types, the association was not strong and few women were not monogamous. These observations are consistent with the demonstrated importance of the husband's behavior in the development of cervical cancer in this population (23, 37). Although the risk of both tumor types was increased in women with an early age at first sexual intercourse, such women tended to be married to men who frequently visited prostitutes, which provides an alternative to the hypothesis that HPV infection at an early age is more likely to lead to persistence of infection and the development of cervical carcinoma than infection later in life. Although the odds ratio of 1.9 for tumors with HPV-16 in women who are positive for HBsAg could readily be due to chance, it is higher than that in women who developed HBV antibodies, suggesting a role for relative immunoincompetence in the genesis of persistent HPV-16 infection and cervical carcinoma. An association was also observed between the prevalence of HPV-16 and HBsAg in prostitutes in Bangkok (23). The absence of appreciable associations of either HPV-16 or HPV-18 tumor type with antibodies to HSV-2, syphilis, and hepatitis B suggests that these serologic tests are not good indices of behavior that leads to acquisition or persistence of HPV infection and that these organisms do not act as cofactors in the genesis of cervical cancer. Consistent with these conclusions, associations with various sexual practices and sexually transmitted agents have only inconsistently been observed among studies based on HPV-positive cases and controls (6, 8). The increase in risk of both HPV-16 and HPV-18 tumor types with number of pregnancies observed in this study is consistent with evidence from some (7, 8), although not other (6, 20), recent investigations and suggests that hormonal changes associated with pregnancy may enhance the likelihood that an infection with either HPV-16 or HPV-18 will lead to cervical cancer. Similarly, most previous studies (38), including all those that have compared cases and controls with oncogenic HPV types (6–8, 20), have shown the risk of cervical carcinoma to be weakly associated with the use of oral contraceptives, and the modest associations observed in this study suggest that the use of oral contraceptives may enhance the likelihood that infection with HPV-16 and HPV-18 will result in carcinogenesis. The absence of associations with the progestational contraceptive depot-medroxyprogesterone acetate in this study and a prior study (39) suggests that the possible effect of oral contraceptives and pregnancies on risk is due to the action of either estrogens or progestins in the presence of estrogens on the cervical epithelium. A role for progestins in the presence of estrogens is suggested by prior observations of a stronger association of adenomatous cervical carcinomas with combined oral contraceptives of high than low progestin potency (40) and enhancement of in vitro oncogenic transformation by HPV-16 in the presence of progestins from oral contraceptives (41). The consistently observed relation of cervical cancer to some measure of socioeconomic status among studies in which cases were compared with HPV-infected controls (6–8, 20) suggests that other unidentified cofactors associated with social class remain to be identified. The association of risk with absence of schooling and prior chest radiographic examinations in this study provides evidence that such cofactors interact with both HPV-16 and HPV-18. The low level of smoking in the women in this study suggests that other factors related to socioeconomic status must be operating. Although smoking does not play an important role in the genesis of cervical cancer in Thai women, a group at high risk of cervical cancer (42), the association of smoking with HPV-16, the viral type that predominates in squamous cell carcinomas, but not with HPV-18, which is the main virus type in adenomatous tumors, although possibly due to chance, is consistent with observations by others (32, 33, 43) that smoking is related to squamous, but not adenomatous, carcinomas. As expected, cervical smear screening was shown in this study to reduce the risk of both HPV-16 and HPV-18 tumor types. The reduced risk in women who had used an IUD likely represents an effect of screening at the time of IUD insertion or removal, a phenomenon also reported from China (44). The similarity in risk factors for cervical carcinomas with HPV-16 and HPV-18 DNA provides strong evidence that the same cofactors operate to enhance the carcinogenicity of these two viral types. The preponderance of HPV-16 in squamous cell carcinomas and the more frequent occurrence of HPV-18 in adenomatous carcinomas are thus a reflection of inherent differences in the viruses themselves. HPV-16 is the more common of the two viruses in exfoliated cells from both cervical and penile scrapings (1), perhaps because it is better adapted to survival in squamous epithelial tissue than is HPV-18. HPV-18 may be more dependent on reaching the glandular epithelium of the endocervical canal for survival. Correspondence to Dr. David B. Thomas, Program in Epidemiology, Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue North, Seattle, WA 98109 (e-mail: [email protected]). This study was supported by grant CA49044 from the US National Institutes of Health. The assistance of Drs. Anna Marie Beckmann and Larry Corey is gratefully acknowledged. REFERENCES 1. International Agency for Research on Cancer. Human papillomaviruses. IARC Monogr Eval Carcinog Risks Hum  1995; 64: 1–409. Google Scholar 2. Ho GY, Bierman R, Beardsley L, et al. Natural history of cervicovaginal papillomavirus infection in young women. N Engl J Med  1998; 338: 423–8. Google Scholar 3. Svare EI, Kjaer SK, Worm AM, et al. Risk factors for HPV infection in women from sexually transmitted disease clinics: comparison between two areas with different cervical cancer incidence. Int J Cancer  1998; 75: 1–8. Google Scholar 4. Young TK, McNicol P, Beauvais J. Factors associated with human papillomavirus infection detected by polymerase chain reaction among urban Canadian aboriginal and non-aboriginal women. Sex Transm Dis  1997; 24: 293–8. Google Scholar 5. Kjaer SK, van den Brule AJ, Bock JE, et al. Determinants for genital human papillomavirus (HPV) infection in 1000 randomly chosen young Danish women with normal Pap smear: are there different risk profiles for oncogenic and nononcogenic HPV types? Cancer Epidemiol Biomarkers Prev  1997; 6: 799–805. Google Scholar 6. Chichareon S, Herrero R, Muñoz N, et al. Risk factors for cervical cancer in Thailand: a case-control study. J Natl Cancer Inst  1998; 90: 50–7. Google Scholar 7. Ngelangel C, Muñoz N, Bosch FX, et al. Causes of cervical cancer in the Philippines: a case-control study. J Natl Cancer Inst  1998; 90: 43–9. Google Scholar 8. Chaouk N, Bosch FX, Muñoz N, et al. The viral origin of cervical cancer in Rabat, Morocco. Int J Cancer  1998; 75: 546–54. Google Scholar 9. Ferrera A, Baay MF, Herbrink P, et al. A sero-epidemiological study of the relationship between sexually transmitted agents and cervical cancer in Honduras. Int J Cancer  1997; 73: 781–5. Google Scholar 10. Dillner J, Lehtinen M, Bjorge T, et al. Prospective seroepidemiologic study of human papillomavirus infection as a risk factor for invasive cervical cancer. J Natl Cancer Inst  1997; 89: 1293–9. Google Scholar 11. Lehtinen M, Dillner J, Knekt P, et al. Serologically diagnosed infection with human papillomavirus type 16 and risk for subsequent development of cervical carcinoma: nested case-control study. BMJ  1996; 312: 537–9. Google Scholar 12. Shah KV, Viscidi RP, Alberg AJ, et al. Antibodies to human papillomavirus 16 and subsequent in situ or invasive cancer of the cervix. Cancer Epidemiol Biomarkers Prev  1997; 6: 233–7. Google Scholar 13. Cavalcanti SM, Deus FC, Zardo LG, et al. Human papillomavirus infection and cervical cancer in Brazil: a retrospective study. Mem Inst Oswaldo Cruz  1996; 91: 433–40. Google Scholar 14. Chua KL, Hjerpe A. Persistence of human papillomavirus (HPV) infections preceding cervical carcinoma. Cancer  1996; 77: 121–7. Google Scholar 15. Lowy DR, Schiller JT. Human papillomavirus biology. J Natl Cancer Inst Monogr  1996; 21: 141–3. Google Scholar 16. Pillai MR, Halabi S, McKalip A, et al. The presence of human papillomavirus-16/-18 E6, p53, and Bcl-2 protein in cervico-vaginal smears from patients with invasive cervical cancer. Cancer Epidemiol Biomarkers Prev  1996; 5: 329–35. Google Scholar 17. Scheffner M, Munger K, Byrne JC, et al. The state of the p53 and retinoblastoma genes in human cervical carcinoma cell lines. Proc Natl Acad Sci U S A  1991; 88: 5523–7. Google Scholar 18. Bosch FX, Manos MM, Muñoz N, et al. Prevalence of human papillomavirus in cervical cancer: a worldwide perspective. J Natl Cancer Inst  1995; 87: 796–802. Google Scholar 19. Franco EL, Rohan TE, Villa LL. Epidemiologic evidence and human papillomavirus infection as a necessary cause of cervical cancer. J Natl Cancer Inst  1999; 91: 506–11. Google Scholar 20. Bosch FX, Muñoz N, de Sanjosé S, et al. Risk factors for cervical cancer in Colombia and Spain. Int J Cancer  1992; 52: 750–8. Google Scholar 21. Wheeler CM, Greer CE, Becker TM, et al. Short-term fluctuations in the detection of cervical human papillomavirus DNA. Obstet Gynecol  1996; 88: 261–8. Google Scholar 22. Thomas DB, Qin Q, Kuypers J, et al. Human papillomaviruses and cervical cancer in Bangkok. II. Risk factors for in situ and invasive squamous cell cervical carcinomas. Am J Epidemiol  2001; 153: 732–9. Google Scholar 23. Thomas DB, Ray RM, Kuypers J, et al. Human papillomaviruses and cervical cancer in Bangkok. III. The role of husbands and commercial sex workers. Am J Epidemiol  2001; 153: 740–8. Google Scholar 24. Histological typing of female genital tract tumours. In: Poulsen HE, Taylor CW, Sobin LH, eds. International histological classification of tumours. Geneva, Switzerland: World Health Organization, 1975. Google Scholar 25. Sarngadharan MG, Popovic M, Bruch L, et al. Antibodies reactive with human T-lymphotropic retroviruses (HTLV-III) in the serum of patients with AIDS. Science  1984; 224: 506–8. Google Scholar 26. Hollinger FB, Dienstag JL. Hepatitis B and D viruses. In: Murray PR, Baron EJ, Pfaller MA, et al, eds. Manual of clinical microbiology. 7th ed. Washington, DC: American Society for Microbiology, 1999:1025–42. Google Scholar 27. Larsen SA, Hunter EF, Kraus SJ, eds. A manual of tests for syphilis. 8th ed. Washington, DC: American Public Health Association, 1990:1–165. Google Scholar 28. Ashley RL, Militoni J. Use of densitometric analysis for interpreting HSV serologies based on Western blot. J Virol Methods  1987; 18: 159–68. Google Scholar 29. Ashley RL, Militoni J, Lee F, et al. Comparison of Western blot (Immunoblot) and glycoprotein G-specific immunodot enzyme assay for detecting antibodies to herpes simplex virus types 1 and 2 in human sera. J Clin Microbiol  1988; 26: 662–7. Google Scholar 30. Bauer HM, Ting Y, Greer CE, et al. Genital human papillomavirus infection in female university students as determined by a PCR-based method. JAMA  1991; 265: 472–7. Google Scholar 31. Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol I. The analysis of case-control studies. Lyon, France: International Agency for Research on Cancer, 1980:325–38. (IARC scientific publication no. 32). Google Scholar 32. Honore LH, Koch M, Brown LB. Comparison of oral contraceptive use in women with adenocarcinoma and squamous cell carcinoma of the uterine cervix. Gynecol Obstet Invest  1991; 32: 98–101. Google Scholar 33. Hopkins MP, Morley GW. A comparison of adenocarcinoma and squamous cell carcinoma of the cervix. Obstet Gynecol  1991; 77: 912–17. Google Scholar 34. Brinton LA, Herrero R, Reeves WC, et al. Risk factors for cervical cancer by histology. Gynecol Oncol  1993; 51: 301–6. Google Scholar 35. Brinton LA, Tashima KT, Lehman HF, et al. Epidemiology of cervical cancer by cell type. Cancer Res  1987; 47: 1706–11. Google Scholar 36. Bjorge T, Kravdal O. Reproductive variables and risk of uterine cervical cancer in Norwegian registry data. Cancer Causes Control  1996; 7: 351–7. Google Scholar 37. Thomas DB, Ray RM, Pardthaisong T, et al. Prostitution, condom use, and invasive squamous cell cervical cancer in Thailand. Am J Epidemiol  1996; 143: 779–86. Google Scholar 38. International Agency for Research on Cancer. Hormonal contraception and post-menopausal hormonal therapy. IARC Monogr Eval Carcinog Risks Hum  1999; 72: 1–660. Google Scholar 39. The WHO Collaborative Study of Neoplasia and Steroid Contraceptives. Depot-medroxyprogesterone acetate (DMPA) and risk of invasive squamous cell cervical cancer. Contraception  1992; 45: 299–312. Google Scholar 40. Thomas DB, Ray RM, WHO Collaborative Study of Neoplasia and Steroid Contraceptives. Oral contraceptives and invasive adenocarcinomas and adenosquamous carcinomas of the uterine cervix. Am J Epidemiol  1996; 144: 281–9. Google Scholar 41. Pater A, Bayatpur M, Pater MM. Oncogenic transformation by human papilloma viruses type 16 desoxyribonucleic acid in the presence of progesterone or progestins from oral contraceptives. Am J Obstet Gynecol  1990; 162: 1099–103. Google Scholar 42. International Agency for Research on Cancer. Cancer incidence in five continents. Vol VII. Lyon, France: International Agency for Research on Cancer, 1997:434–7. (IARC scientific publication no. 143). Google Scholar 43. Ursin G, Pike MC, Preston-Martin S, et al. Sexual, reproductive, and other risk factors for adenocarcinoma of the cervix: results from a population-based case-control study (California, United States). Cancer Causes Control  1996; 7: 391–401. Google Scholar 44. Li HQ, Thomas DB, Jin SK, et al. Tubal sterilization and use of an IUD and risk of cervical cancer. J Womens Health Gend Based Med  2000; 9: 303–10. Google Scholar

Thomas, David B.; Qin, Qin; Kuypers, Jane; Kiviat, Nancy; Ashley, Rhoda L.; Koetsawang, Amorn; Ray, Roberta M.; Koetsawang, Suporn

doi: N/Apmid: N/A

To identify risk factors for progression of intraepithelial cervical lesions, 190 women with invasive cervical cancer were compared with 75 women with in situ disease diagnosed in Bangkok, Thailand, between September 1991 and September 1993. Polymerase chain reaction-based assays for type-specific human papillomavirus (HPV) DNA in cervical scrapings revealed oncogenic types in 79% of invasive and 57% of intraepithelial tumors. Types 16 and 18, but not types 31/33/35/39, were more common in invasive than intraepithelial tumors, and untyped HPV DNA was found more commonly in the in situ lesions, suggesting that in situ disease is four times more likely to become invasive if due to type 16 or 18 than to other causes, and that tumors with only untyped HPV are not at increased risk of progression. After controlling for HPV type, the risk of developing invasive diseases, compared with the risk of developing intraepithelial lesions, was not related to any of a large number of sexual and hormonal factors considered or to smoking, suggesting that any cofactors these variables represent act before the development of in situ carcinoma. Two indices of socioeconomic status were associated with a reduced risk of only invasive disease, suggesting the existence of unknown protective factors that operate after intraepithelial lesions develop.

Thomas, David B.;Qin, Qin;Kuypers, Jane;Kiviat, Nancy;Ashley, Rhoda L.;Koetsawang, Amorn;Ray, Roberta M.;Koetsawang, Suporn

doi: 10.1093/aje/153.8.732pmid: 11296144

Abstract To identify risk factors for progression of intraepithelial cervical lesions, 190 women with invasive cervical cancer were compared with 75 women with in situ disease diagnosed in Bangkok, Thailand, between September 1991 and September 1993. Polymerase chain reaction-based assays for type-specific human papillomavirus (HPV) DNA in cervical scrapings revealed oncogenic types in 79% of invasive and 57% of intraepithelial tumors. Types 16 and 18, but not types 31/33/35/39, were more common in invasive than intraepithelial tumors, and untyped HPV DNA was found more commonly in the in situ lesions, suggesting that in situ disease is four times more likely to become invasive if due to type 16 or 18 than to other causes, and that tumors with only untyped HPV are not at increased risk of progression. After controlling for HPV type, the risk of developing invasive diseases, compared with the risk of developing intraepithelial lesions, was not related to any of a large number of sexual and hormonal factors considered or to smoking, suggesting that any cofactors these variables represent act before the development of in situ carcinoma. Two indices of socioeconomic status were associated with a reduced risk of only invasive disease, suggesting the existence of unknown protective factors that operate after intraepithelial lesions develop. carcinoma, squamous cell, cervical intraepithelial neoplasia, cervix neoplasms, papillomavirus, human CIN-III, cervical intraepithelial neoplasia grade 3, HPV, human papillomavirus Some types of human papillomaviruses (HPVs) are carcinogenic for the cervix uteri (1). Although they are quite likely necessary causes of cervical carcinoma, they are not sufficient causes. Most infected women do not develop invasive cervical cancer, and cofactors must therefore play a role in allowing these infections to become persistent and carcinogenic. Cervical carcinoma in situ (cervical intra-epithelial neoplasia grade 3 (CIN-III)) is an intermediate morphologic lesion on the pathway from normal cervical epithelium to invasive disease, and cofactors for invasive cervical carcinoma could operate either prior to the development of carcinoma in situ or afterward. Those acting before the development of the intraepithelial lesion would be observed in relation to both in situ and invasive lesions, whereas those acting to enhance the probability that an in situ lesion becomes invasive would be observed only in relation to invasive disease. The identification of cofactors for preinvasive disease is best attempted in the context of prospective studies of HPV-infected women. Concurrent prospective studies of untreated women with carcinoma in situ to determine factors associated with progression to invasive diseases would be unethical, and studies of treated women would be uninformative because treatment removes or destroys the in situ lesion. A few historical cohort studies of HPV-infected subjects that included both in situ and invasive cervical carcinoma as endpoints (1–4) and case-control studies of invasive disease (1, 5–9) have identified a variety of possible cofactors with varying degrees of consistency that could act either before or after the development of in situ disease. These include the following: early age at first sexual intercourse, which may reflect particular vulnerability of the immature cervix to HPV-induced carcinogenesis; hormonal factors such as high parity and use of oral contraceptives; exposures to chemical agents in cigarette smoke; sexually transmitted agents other than HPV; deficiencies in certain micronutrients; an immunocompromised state resulting from human immunodeficiency virus (HIV) infection or immunosuppresion for prevention of organ transplant rejection; and various measures of socioeconomic status, suggesting the presence of additional unrecognized cofactors associated with social class. Most of these factors, however, have been observed in relation to both in situ and invasive disease (1) and, in studies in which the epidemiologic features of these two stages of cervical carcinoma can be compared, few consistent differences in risk factors have been observed (5, 10–15). These observations suggest that most putative cofactors are operative before the development of carcinoma in situ and that factors associated with progression to invasion have yet to be identified. This is a report of analysis performed to compare risk factors for in situ and invasive disease in order to identify variables associated only with the latter condition, after controlling for the presence of oncogenic HPV types in the tumor tissue. If women with in situ disease are considered the controls and if women with invasive disease are considered the cases in a standard case-control analysis, then elevated odds ratios in relation to a factor represent the risk of invasive disease relative to the risk of in situ disease in women with the factor (16). Because it is reasonable to assume that most or all invasive cervical carcinomas have passed through an in situ stage, an elevated odds ratio would imply that the factor enhances probability of progression from in situ to invasion. To our knowledge, results of direct comparisons of women with in situ and invasive disease have been reported only once previously (10). MATERIALS AND METHODS Women who were admitted to the public wards of Siriraj Hospital in Bangkok, Thailand, with a new histologically confirmed diagnosis of either CIN-III or invasive cervical carcinoma between September 1991 and September 1993, who were born after 1929, and who resided in Thailand for at least the past year were eligible as cases for this study. Although CIN-III includes both severe dysplasia and carcinoma in situ, the older terms, carcinoma in situ and in situ disease, are used interchangeably with CIN-III in this report. Two controls for each woman with invasive disease were selected from otolaryngology and general surgery wards of Siriraj Hospital, as previously described (17). The cases of carcinoma in situ were women who had had an abnormal cervical smear during a visit to either the family planning or gynecologic clinic associated with Siriraj Hospital, and an attempt was made to select two controls for each case from the same clinic from which the case came. Cervical smears are taken from many of the women who attend these clinics, and the women selected as controls were the next two women with appointments in the clinic after the appointment for the case, in the same 5-year age group and from the same region of the country as the case, who had returned to the clinic to learn the results of their cervical smears, and whose cervical smears revealed no suspicion of neoplastic change. If a woman refused to participate, the selection process continued until two controls per case were selected. All cases and hospitalized controls were interviewed while hospitalized; the controls selected for the cases of carcinoma in situ were interviewed in the clinics on the day they were selected. Information was obtained on sexual and reproductive history, prior cervical smears, use of tobacco and alcohol, and indices of prior use of medical resources and socioeconomic status. Dates of all prior cervical smears were obtained, and those taken within the past year were not considered prior screening cervical smears. The presenting symptom of each case was also ascertained and recorded as none (detected by routine cytology only), vaginal bleeding, abnormal vaginal discharge, or other. A 15-ml blood specimen was obtained from all interviewed women. Four aliquots of serum were stored at −70°C, two of which were retained in Bangkok and two of which were shipped to Seattle, Washington, on dry ice. Cervical scrapings for HPV DNA assays were obtained from cases by the patient's surgeon prior to treatment. Cervical smears and cervical scrapings were taken by these same surgeons from the hospital controls while they were hospitalized and from the clinic controls in the clinic on the same day they were selected and interviewed. The cervix, including the cervical os, was scraped with a Teflon (E. I. du Pont de Nemours and Company, Wilmington, Delaware)-coated swab, the end of which was then broken off into a vial containing 2 ml of specimen transport medium (Digene Diagnostics, Inc., Beltsville, Maryland). The specimens were stored at −70°C and periodically shipped on dry ice to Seattle. Histologic slides from the blocks that were used to make the diagnosis in the cases were read by a single collaborating pathologist in Bangkok. Information was recorded on source of specimen, tumor size, stage, and histologic diagnosis. A pathologist in Seattle reread the slides from 174 of the cases included in this report. Of 23 in situ cases, only one (4.3 percent) was considered invasive by the reference pathologist, and only eight (5.3 percent) of 151 cases considered invasive in Bangkok were read in Seattle as in situ with no evidence of microinvasion. The diagnosis made in Bangkok was thus used in the analysis in this report. As detailed elsewhere (17), serologic assays were performed for antibodies to human immunodeficiency virus, hepatitis B, Treponema pallidum, and herpes simplex virus types 1 and 2, as well as for hepatitis B surface antigen. Cervical scrapings were assayed for evidence of any HPV DNA and for type-specific DNA of types 6 and 11, type 16, type 18, types 31/33/35/39, and type 45 using polymerase chain reaction-based technology (17). Samples that hybridized with a generic probe but not with any of the type-specific probes were considered positive for untyped HPV. Cases and controls were compared, and odds ratios as estimates of relative risks were calculated using unconditional logistic regression (18) when comparing invasive cases with their unmatched controls and using conditional logistic regression (18) when comparing in situ cases with their matched controls. In addition, to directly compare risk factors for invasive and in situ disease, we performed unconditional logistic regression analyses using the women with invasive disease as “cases” and women with carcinoma in situ as “controls.” Because the purpose of these analyses was to identify any factors that distinguish invasive from in situ disease, and because no such factors were identified other than HPV status, no attempts were made to adjust odds ratios for factors other than age and (for the case-case comparisons) HPV status. RESULTS A total of 420 eligible cases were identified, of whom 338 (80.5 percent) were interviewed. Cervical scrapings were obtained from 337 (99.7 percent) of those interviewed, 216 and 75 of whom had invasive and in situ squamous cell carcinoma, respectively. Adequate amounts of DNA for HPV DNA assays were found in cervical scrapings from 191 of the women with invasive disease and all 75 of the women with carcinoma in situ. One woman with invasive disease who claimed to be a virgin was omitted, leaving 190 cases of invasive cervical carcinoma in the analyses. Controls were not found for 10 of the women with carcinoma in situ, leaving 65 of these cases for inclusion in the analyses in which cases and controls were compared; all 75 cases were included in the analyses in which women with invasive and in situ disease were compared. Of the 614 women selected as controls for the cases of invasive disease, 490 (79.8 percent) were interviewed. Cervical scrapings were obtained from 306 (62.4 percent) of these 490 women, 298 (97.4 percent) of which yielded adequate DNA samples. Three women with no history of sexual intercourse, one with a subtotal hysterectomy, and three with ineligible diagnoses were excluded, leaving 291 controls for analysis. Of the 161 women selected as controls for the in situ cases, 160 (99.4 percent) were interviewed, 144 of whom were subsequently confirmed as meeting the eligibility criteria for inclusion in the study. Cervical scrapings were obtained from 133 (92.4 percent) of these 144 women, 128 of which contained adequate amounts of DNA. Four of the 128 women from whom these samples were taken had been matched to excluded cases and were omitted, leaving 124 women as controls for the in situ cases in the analyses. As shown in table 1, women with invasive disease were slightly older than their controls. The women with carcinoma in situ and their controls were comparable in age and younger than the women with invasive carcinoma and their controls. The 10 additional cases of carcinoma in situ used in the comparisons of in situ and invasive disease are not appreciably different in age from the 65 cases shown in the table. All odds ratio estimates presented were adjusted for age using the age categories shown in the table. TABLE 1. Age distribution of invasive and in situ squamous cell cervical carcinomas and their corresponding controls, Bangkok, Thailand, 1991–1993* Age (years)  Invasive   In situ   Cases   Controls   Cases   Controls   No.  %  No.  %  No.  %  No.  %  <30  3  1.6  13  4.5  16  24.6  29  23.4  30–34  23  12.1  44  15.1  11  16.9  24  19.3  35–39  36  18.9  56  19.2  15  23.1  32  25.8  40–44  33  17.4  60  20.6  12  18.5  17  13.7  45–49  30  15.8  36  12.4  4  6.1  10  8.1  ≥50  65  34.2  82  28.2  7  10.8  12  9.7  Total  190  100.0  291  100.0  65  100.0  124  100.0  Age (years)  Invasive   In situ   Cases   Controls   Cases   Controls   No.  %  No.  %  No.  %  No.  %  <30  3  1.6  13  4.5  16  24.6  29  23.4  30–34  23  12.1  44  15.1  11  16.9  24  19.3  35–39  36  18.9  56  19.2  15  23.1  32  25.8  40–44  33  17.4  60  20.6  12  18.5  17  13.7  45–49  30  15.8  36  12.4  4  6.1  10  8.1  ≥50  65  34.2  82  28.2  7  10.8  12  9.7  Total  190  100.0  291  100.0  65  100.0  124  100.0  * The mean and median ages (years) are the following: invasive cases, 45.5 and 45.5; invasive controls, 43.9 and 43.0; in situ cases, 37.8 and 38.0; and in situ controls, 37.2 and 37.0, respectively. View Large Vaginal bleeding was the presenting symptom in 124 (65.3 percent) of the 190 women with invasive disease and in 47 (62.7 percent) of the 75 women with carcinoma in situ; 63 (33.2 percent) of the women with invasive carcinoma and 23 (30.7 percent) of those with carcinoma in situ reported no specific symptoms. Oncogenic types of HPV were strongly associated with both invasive and in situ disease (table 2). Type 16 DNA was the most frequently identified type in both invasive and in situ tumors. Both types 16 and 18 were found in a higher proportion of invasive than in situ carcinomas. Types 31/33/35/39 occurred with nearly equal frequency in both tumor types, and untyped HPV DNA was found in a higher proportion of the in situ than invasive lesions. Few women in either control group had any type of HPV in their cervical scrapings, although the frequency of a positive test was slightly higher in the controls for the in situ cases. The estimates of the odds ratio of both invasive and in situ carcinomas are large, with those in relation to types 16 and 18, any oncogenic type, and any HPV type being stronger for invasive disease; those in relation to types 31/33/35/39 being similar for both tumor types; but those in relation to untyped HPV being greater for in situ disease. As shown in table 3, after controlling for age and other HPV types, the odds ratio of invasive disease is four times greater than the risk of in situ disease in women with HPV types 16 and 18, twice as high in women with HPV types 31/33/35/39 (although possibly due to chance), and also significantly higher in relation to any oncogenic type and any type of HPV. The risk is not greater for invasive than in situ disease in women with untyped HPV DNA. TABLE 2. Age-adjusted odds ratios of invasive and in situ squamous cell cervical carcinoma in relation to specific types of human papillomaviruses, Bangkok, Thailand, 1991–1993 HPV* type†  Positive subjects           Invasive   In situ   Invasive  In situ  Cases (n = 190)   Controls (n = 291)   Cases (n = 65)   Controls (n = 124)           No.  %  No.  %  No.  %  No.  %  OR*  95% CI*  OR  95% CI  16  114  60.0  6  2.1  25  38.4  6  4.8  83  39, 232  11  3.9, 33  18  26  13.7  0  0  5  7.7  1  0.8  ∞  22, ∞  10  1.2, 86  31/33/35/39  25  13.2  3  1.0  9  13.8  1  0.8  14  4.9, 61  17  2.2, 135  Any oncogenic type‡  150  78.9  9  3.1  37  56.9  7  5.6  155  72, 384  16  5.9, 47  Untyped§  14  7.4  11  3.8  12  18.5  6  4.8  2.0  0.9, 4.7  4.9  1.6, 15  Any type§  164  86.3  20  6.9  50  76.9  13  10.4  97  52, 193  21  7.4, 57  HPV* type†  Positive subjects           Invasive   In situ   Invasive  In situ  Cases (n = 190)   Controls (n = 291)   Cases (n = 65)   Controls (n = 124)           No.  %  No.  %  No.  %  No.  %  OR*  95% CI*  OR  95% CI  16  114  60.0  6  2.1  25  38.4  6  4.8  83  39, 232  11  3.9, 33  18  26  13.7  0  0  5  7.7  1  0.8  ∞  22, ∞  10  1.2, 86  31/33/35/39  25  13.2  3  1.0  9  13.8  1  0.8  14  4.9, 61  17  2.2, 135  Any oncogenic type‡  150  78.9  9  3.1  37  56.9  7  5.6  155  72, 384  16  5.9, 47  Untyped§  14  7.4  11  3.8  12  18.5  6  4.8  2.0  0.9, 4.7  4.9  1.6, 15  Any type§  164  86.3  20  6.9  50  76.9  13  10.4  97  52, 193  21  7.4, 57  * HPV, human papillomavirus; OR, odds ratio in relation to women without the specific HPV type; CI, confidence interval. † One woman with carcinoma in situ had HPV type 6/11. No cases or controls had type 45. ‡ Includes types 16, 18, and 31/33/35/39. § Weakly reactive tests were considered negative. View Large TABLE 3. Risks of invasive squamous cell cervical carcinoma relative to risks of in situ squamous cell cervical carcinoma in relation to specific types of human papillomaviruses, Bangkok, Thailand, 1991–1993 HPV* type  No. of cases   OR*,†  95% CI*  Invasive  In situ  No HPV  26  18  1.0  Reference  16  114  29  4.2  1.8, 9.5  18  26  5  4.0  1.2, 13  31/33/35/39  25  11  2.1  0.8, 5.9  Any oncogenic type  150  42  3.5  1.6, 7.7  Untyped  14  14  0.7  0.3, 2.1  Any type  164  57  2.7  1.3, 5.2  Total cases  190  75      HPV* type  No. of cases   OR*,†  95% CI*  Invasive  In situ  No HPV  26  18  1.0  Reference  16  114  29  4.2  1.8, 9.5  18  26  5  4.0  1.2, 13  31/33/35/39  25  11  2.1  0.8, 5.9  Any oncogenic type  150  42  3.5  1.6, 7.7  Untyped  14  14  0.7  0.3, 2.1  Any type  164  57  2.7  1.3, 5.2  Total cases  190  75      * HPV, human papillomavirus; OR, odds ratio; CI, confidence interval. † Adjusted for age and all other types of HPV shown in the table. View Large In tables 4 through 6, estimated risks of invasive disease relative to risks of in situ disease are adjusted for age and the presence of oncogenic HPV, other HPV, or no HPV (three categories). These estimates did not differ appreciably from estimates adjusted only for age. No other factors were identified that appeared to confound the estimates presented. TABLE 4. Age-adjusted odds ratios of invasive and in situ squamous cell cervical carcinoma in relation to sexual factors and infectious agents other than human papillomaviruses, Bangkok, Thailand, 1991–1993 Factor/agent  Invasive   In situ   Invasive   In situ   Invasive/in situ†   Cases (no.)  Controls (no.)  Cases‡ (no.)  Controls (no.)  OR§  95% CI§  OR  95% CI  OR  95% CI  No. of sexual partners                       1  175  272  60/67  113  1.0  Reference  1.0  Reference  1.0  Reference   >1  15  19  5/8  11  1.2  0.6, 2.5  0.9  0.3, 2.9  0.8  0.3, 2.1  Age (years) at first intercourse                      ≤16  30  22  10/13  6  4.4  2.2, 9.1*  6.9  1.6, 30*  1.4  0.3, 3.9   17–18  45  47  19/21  21  2.8  1.6, 5.1  2.6  1.1, 6.3  1.0  0.4, 2.6   19–20  51  69  13/16  26  2.0  1.2, 3.5  1.5  0.6, 4.0  1.0  0.4, 2.6   21–23  33  66  11/12  35  1.4  0.8, 2.6  0.9  0.4, 2.4  1.0  0.3, 2.8  ≥24  31  87  12/13  76  1.0  Reference  1.0  Reference  1.0  Reference  HSV-2§ antibodies¶                       Negative  79  133  28/33  61  1.0  Reference  1.0  Reference  1.0  Reference   Positive  108  135  37/42  54  1.4  1.0, 2.0  1.5  0.8, 2.9  1.3  0.7, 2.4  Syphilis serology¶                       Negative  175  267  61/69  118  1.0  Reference  1.0  Reference  1.0  Reference   Positive  13  14  4/4  2  1.5  0.7, 3.2  3.6  0.7, 2.0  1.4  0.4, 5.1  HBsAg§,¶                       Negative  177  256  62/70  112  1.0  Reference  1.0  Reference  1.0  Reference   Positive  11  14  3/3  6  1.2  0.5, 2.8  0.9  0.2, 3.7  1.4  0.3, 6.1  Anti-HBs antibodies§,¶                       Negative  109  167  35/38  84  1.0  Reference  1.0  Reference  1.0  Reference   Positive  79  112  30/35  34  1.0  0.7, 1.5  2.3  1.2, 4.4  0.6  0.3, 1.1  Anti-HBc antibodies§,¶                       Negative  77  116  32/36  73  1.0  Reference  1.0  Reference  1.0  Reference   Positive  111  163  33/37  45  1.0  0.7, 1.4  1.9  1.0, 3.7  1.0  0.6, 1.9  Factor/agent  Invasive   In situ   Invasive   In situ   Invasive/in situ†   Cases (no.)  Controls (no.)  Cases‡ (no.)  Controls (no.)  OR§  95% CI§  OR  95% CI  OR  95% CI  No. of sexual partners                       1  175  272  60/67  113  1.0  Reference  1.0  Reference  1.0  Reference   >1  15  19  5/8  11  1.2  0.6, 2.5  0.9  0.3, 2.9  0.8  0.3, 2.1  Age (years) at first intercourse                      ≤16  30  22  10/13  6  4.4  2.2, 9.1*  6.9  1.6, 30*  1.4  0.3, 3.9   17–18  45  47  19/21  21  2.8  1.6, 5.1  2.6  1.1, 6.3  1.0  0.4, 2.6   19–20  51  69  13/16  26  2.0  1.2, 3.5  1.5  0.6, 4.0  1.0  0.4, 2.6   21–23  33  66  11/12  35  1.4  0.8, 2.6  0.9  0.4, 2.4  1.0  0.3, 2.8  ≥24  31  87  12/13  76  1.0  Reference  1.0  Reference  1.0  Reference  HSV-2§ antibodies¶                       Negative  79  133  28/33  61  1.0  Reference  1.0  Reference  1.0  Reference   Positive  108  135  37/42  54  1.4  1.0, 2.0  1.5  0.8, 2.9  1.3  0.7, 2.4  Syphilis serology¶                       Negative  175  267  61/69  118  1.0  Reference  1.0  Reference  1.0  Reference   Positive  13  14  4/4  2  1.5  0.7, 3.2  3.6  0.7, 2.0  1.4  0.4, 5.1  HBsAg§,¶                       Negative  177  256  62/70  112  1.0  Reference  1.0  Reference  1.0  Reference   Positive  11  14  3/3  6  1.2  0.5, 2.8  0.9  0.2, 3.7  1.4  0.3, 6.1  Anti-HBs antibodies§,¶                       Negative  109  167  35/38  84  1.0  Reference  1.0  Reference  1.0  Reference   Positive  79  112  30/35  34  1.0  0.7, 1.5  2.3  1.2, 4.4  0.6  0.3, 1.1  Anti-HBc antibodies§,¶                       Negative  77  116  32/36  73  1.0  Reference  1.0  Reference  1.0  Reference   Positive  111  163  33/37  45  1.0  0.7, 1.4  1.9  1.0, 3.7  1.0  0.6, 1.9  * p value of test for trend < 0.01. † Risks of invasive disease relative to risk of in situ disease, adjusted for age and presence or absence of oncogenic human papillomavirus (HPV) type or other and unknown HPV type or no HPV (three categories). ‡ In situ cases included in comparisons of cases versus controls/in comparisons of invasive versus in situ cases. § OR, odds ratio; CI, confidence interval; HSV-2, herpes simplex virus type 2; HBsAg, hepatitis B surface antigen; anti-HBs antibodies, anti-hepatitis B surface antibodies; anti-HBc antibodies, anti-hepatitis B core antibodies. ¶ Study subjects with no serologic tests were omitted from analyses. View Large TABLE 5. Age-adjusted odds ratios of invasive and in situ squamous cell cervical carcinoma in relation to hormonal factors, Bangkok, Thailand, 1991–1993 Factor  Invasive   In situ   Invasive   In situ   Invasive/in situ†   Cases (no.)  Controls (no.)  Cases‡ (no.)  Controls (no.)  OR§  95% CI§  OR  95% CI  OR  95% CI  No. of pregnancies                       0  2  9  4/6  17  0.6  0.1, 2.4*  0.5  0.1, 1.8*  0.3  0.1, 2.0   1–2  41  114  28/30  68  1.0  Reference  1.0  Reference  1.0  Reference   3–4  72  95  22/26  33  2.2  1.3, 3.6  1.9  0.9, 4.3  1.2  0.6, 2.5   5–8  67  66  11/13  6  3.0  1.0, 5.3  8.6  1.7, 4.2  1.9  0.8, 4.5   ≥9  8  7  0  0  3.4  1.7, 11          Ever had stillbirth¶                       No  180  272  59/66  106  1.0  Reference  1.0  Reference  1.0  Reference   Yes  2  10  2/3  1  1.1  0.4, 2.8  3.6  0.3, 4.0  0.6  0.1, 2.7  Ever had spontaneous abortion¶                       No  146  235  51/58  90  1.0  Reference  1.0  Reference  1.0  Reference   Yes  42  47  10/11  17  1.4  0.9, 2.8  1.0  0.4, 2.4  1.5  0.7, 3.3  Ever had induced abortion¶                       No  179  271  57/65  98  1.0  Reference  1.0  Reference  1.0  Reference   Yes  9  11  4/4  9  1.3  0.5, 3.3  0.7  0.2, 2.5  1.2  0.3, 4.5  Ever used oral contraceptives                       No  98  162  25/26  63  1.0  Reference  1.0  Reference  1.0  Reference   Yes  92  129  40/49  61  1.3  0.9, 2.0  1.7  0.9, 3.2  0.6  0.3, 1.2  Ever used DMPA§                       No  161  229  46/55  95  1.0  Reference  1.0  Reference  1.0  Reference   Yes  29  62  19/20  29  0.8  0.5, 1.3  1.4  1.3, 1.5  0.7  0.4, 1.6  Factor  Invasive   In situ   Invasive   In situ   Invasive/in situ†   Cases (no.)  Controls (no.)  Cases‡ (no.)  Controls (no.)  OR§  95% CI§  OR  95% CI  OR  95% CI  No. of pregnancies                       0  2  9  4/6  17  0.6  0.1, 2.4*  0.5  0.1, 1.8*  0.3  0.1, 2.0   1–2  41  114  28/30  68  1.0  Reference  1.0  Reference  1.0  Reference   3–4  72  95  22/26  33  2.2  1.3, 3.6  1.9  0.9, 4.3  1.2  0.6, 2.5   5–8  67  66  11/13  6  3.0  1.0, 5.3  8.6  1.7, 4.2  1.9  0.8, 4.5   ≥9  8  7  0  0  3.4  1.7, 11          Ever had stillbirth¶                       No  180  272  59/66  106  1.0  Reference  1.0  Reference  1.0  Reference   Yes  2  10  2/3  1  1.1  0.4, 2.8  3.6  0.3, 4.0  0.6  0.1, 2.7  Ever had spontaneous abortion¶                       No  146  235  51/58  90  1.0  Reference  1.0  Reference  1.0  Reference   Yes  42  47  10/11  17  1.4  0.9, 2.8  1.0  0.4, 2.4  1.5  0.7, 3.3  Ever had induced abortion¶                       No  179  271  57/65  98  1.0  Reference  1.0  Reference  1.0  Reference   Yes  9  11  4/4  9  1.3  0.5, 3.3  0.7  0.2, 2.5  1.2  0.3, 4.5  Ever used oral contraceptives                       No  98  162  25/26  63  1.0  Reference  1.0  Reference  1.0  Reference   Yes  92  129  40/49  61  1.3  0.9, 2.0  1.7  0.9, 3.2  0.6  0.3, 1.2  Ever used DMPA§                       No  161  229  46/55  95  1.0  Reference  1.0  Reference  1.0  Reference   Yes  29  62  19/20  29  0.8  0.5, 1.3  1.4  1.3, 1.5  0.7  0.4, 1.6  * p value of test for trend < 0.01. † Risks of invasive disease relative to risk of in situ disease, adjusted for age and presence or absence of oncogenic human papillomavirus (HPV) type or other and unknown HPV type or no HPV (three categories). ‡ In situ cases included in comparisons of cases versus controls/in comparisons of invasive versus in situ cases. § OR, odds ratio; CI, confidence interval; DMPA, depot-medroxyprogesterone acetate. ¶ Excluding women who had never been pregnant. View Large TABLE 6. Age-adjusted odds ratios of invasive and in situ squamous cell cervical carcinoma in relation to nonsexual, nonhormonal factors, Bangkok, Thailand, 1991–1993 Factor  Invasive   In situ   Invasive   In situ   Invasive/in situ*   Cases (no.)  Controls (no.)  Cases† (no.)  Controls (no.)  OR‡  95% CI‡  OR  95% CI  OR  95% CI  Ever used an IUD‡                       No  168  231  59/67  101  1.0  Reference  1.0  Reference  1.0  Reference   Yes  22  60  6/8  23  0.5  0.3, 0.9  0.4  0.2, 1.1  1.2  0.5, 3.1  Tubal ligation                       No  138  187  38/46  70  1.0  Reference  1.0  Reference  1.0  Reference   Yes  52  104  27/29  54  0.7  0.4, 1.0  0.9  0.5, 1.7  0.4  0.2, 0.8  Months since last cervical smear                       No cervical smear  165  192  50/56  75  1.0  Reference  1.0  Reference  1.0  Reference   1–12  7  1  11/11  30  0.2  0.1, 0.4  0.6  0.3, 1.3  0.3  0.1, 0.8   >12  5  2  4/7  14  0.3  0.1, 0.6  0.4  0.1, 1.3  0.3  0.1, 1.1   Unknown months  13  2  0  5  0.5  0.3, 1.0          Ever smoked ≥100 cigarettes§                       No  175  276  61/69  121  1.0  Reference  1.0  Reference  1.0  Reference   Yes  14  15  4/6  3  1.4  0.7, 2.0  2.2  0.5, 10  1.2  0.4, 3.7  Ever drank alcoholic beverages                       No  94  148  39/44  64  1.0  Reference  1.0  Reference  1.0  Reference   Yes  96  143  26/31  60  1.0  0.7, 1.5  0.7  0.4, 1.3  1.1  0.6, 2.0  Ever had a chest radiographic examination                       No  158  204  42/46  84  1.0  Reference  1.0  Reference  1.0  Reference   Yes  32  87  23/29  40  0.5  0.3, 0.7  1.2  0.6, 2.3  0.3  0.2, 0.6  Attended school                       No  31  27  3/4  10  1.0  Reference  1.0  Reference  1.0  Reference   Yes  159  264  62/73  114  0.6  0.3, 1.0  1.7  0.5, 6.6  0.5  0.2, 1.6  Factor  Invasive   In situ   Invasive   In situ   Invasive/in situ*   Cases (no.)  Controls (no.)  Cases† (no.)  Controls (no.)  OR‡  95% CI‡  OR  95% CI  OR  95% CI  Ever used an IUD‡                       No  168  231  59/67  101  1.0  Reference  1.0  Reference  1.0  Reference   Yes  22  60  6/8  23  0.5  0.3, 0.9  0.4  0.2, 1.1  1.2  0.5, 3.1  Tubal ligation                       No  138  187  38/46  70  1.0  Reference  1.0  Reference  1.0  Reference   Yes  52  104  27/29  54  0.7  0.4, 1.0  0.9  0.5, 1.7  0.4  0.2, 0.8  Months since last cervical smear                       No cervical smear  165  192  50/56  75  1.0  Reference  1.0  Reference  1.0  Reference   1–12  7  1  11/11  30  0.2  0.1, 0.4  0.6  0.3, 1.3  0.3  0.1, 0.8   >12  5  2  4/7  14  0.3  0.1, 0.6  0.4  0.1, 1.3  0.3  0.1, 1.1   Unknown months  13  2  0  5  0.5  0.3, 1.0          Ever smoked ≥100 cigarettes§                       No  175  276  61/69  121  1.0  Reference  1.0  Reference  1.0  Reference   Yes  14  15  4/6  3  1.4  0.7, 2.0  2.2  0.5, 10  1.2  0.4, 3.7  Ever drank alcoholic beverages                       No  94  148  39/44  64  1.0  Reference  1.0  Reference  1.0  Reference   Yes  96  143  26/31  60  1.0  0.7, 1.5  0.7  0.4, 1.3  1.1  0.6, 2.0  Ever had a chest radiographic examination                       No  158  204  42/46  84  1.0  Reference  1.0  Reference  1.0  Reference   Yes  32  87  23/29  40  0.5  0.3, 0.7  1.2  0.6, 2.3  0.3  0.2, 0.6  Attended school                       No  31  27  3/4  10  1.0  Reference  1.0  Reference  1.0  Reference   Yes  159  264  62/73  114  0.6  0.3, 1.0  1.7  0.5, 6.6  0.5  0.2, 1.6  * Risks of invasive disease relative to risk of in situ disease, adjusted for age and presence or absence of oncogenic human papillomavirus (HPV) type or other and unknown HPV type or no HPV (three categories). † In situ cases included in comparisons of cases versus controls/in comparisons of invasive versus in situ cases. ‡ OR, odds ratio; CI, confidence interval; IUD, intrauterine device. § One case with unknown smoking history was excluded. View Large As shown in table 4, the risk of both invasive and in situ disease tended to decline with increasing age at first intercourse. The risk of both tumor types was only weakly associated with positive serologic tests for herpes simplex virus type 2 antibodies, syphilis or hepatitis B antibodies, or surface antigen, and no associations with herpes simplex virus type 1 were found (not shown). Few women in this study admitted to having had more than one sexual partner, and no increase in risk of either tumor type was observed in relation to this variable. None of the sexual factors considered was significantly more strongly related to invasive than in situ disease. Only one woman with invasive disease and three with CIN-III were positive for the human immunodeficiency virus test (not shown). The risk of both in situ and invasive disease increased with increasing number of pregnancies (table 5). The risk of neither tumor type was significantly associated with the other hormonal factors considered, except for an increase in risk of in situ carcinoma in users of depot-medroxyprogesterone acetate for contraception. Use of oral contraceptives was weakly associated with both tumor types. None of the hormonal factors shown in table 5 was significantly more strongly associated with invasive than in situ disease. Ages at menarche and first livebirth were also not associated with either tumor type (not shown). Women who had used an intrauterine device and who had a history of prior cervical smears were at reduced risk of both invasive and in situ disease, and the risk of both types was slightly but not significantly increased in smokers (table 6). The risk of invasive but not of in situ carcinoma was lower in women with than without a history of a screening chest radiographic examination and formal schooling. The risk of invasive diseases was low relative to the risk of in situ disease in women with a tubal ligation, prior cervical smears, and a prior chest radiographic examination, and possibly in women who had attended school. When the analyses for the last columns in tables 4–6 were restricted to data on study subjects who had never had a cervical smear, to eliminate differences between women with invasive and in situ disease that may be due to prior screening behavior, the results were not appreciably different from those presented. Moreover, the same analyses were performed separately for invasive and in situ cases with an oncogenic HPV type (150 invasive and 42 in situ cases) and without an oncogenic HPV type (26 invasive and 18 in situ cases), and no significant differences (p > 0.05) were observed in the comparable odds ratio estimates from the two sets of analyses. DISCUSSION Oncogenic HPV DNA was found in a high proportion of invasive squamous cell cervical carcinomas, with type 16 being the predominant type. This is in accordance with prior observations (1, 6–8, 19) in which sensitive polymerase chain reaction-based methods were used to detect HPV DNA in cervical cancer cells. The oncogenic types that were assayed were demonstrated in a lower proportion of the in situ than invasive tumors, and untyped HPV DNA was found more frequently in the intraepithelial lesions. These observations are also in accordance with prior studies (1, 10, 20, 21) and are compatible with observations by others that the prevalence of oncogenic HPV types in tumors is directly related to the severity of the lesion (1, 22–26). There is little reason to suggest that the odds ratio estimates in this study in relation to HPV type are biased: cervical scrapings were obtained from a high proportion of both the invasive and in situ cases; both groups of women resided in the same areas of Thailand; and the prevalence of the HPV types that were assayed is similar to that observed by others (1). The odds ratio estimates shown in table 2 thus provide evidence that HPV types 16 and 18 are associated with a fourfold increase in the risk of progression to invasion, that types 31/33/35/39 are less strongly predictive of progression, and that the other (untyped) HPV types are not indicative of subsequent invasion. This interpretation is consistent with the observation (27) that squamous intraepithelial lesions with oncogenic HPV DNA tend to be monoclonal, and those with other HPV DNA types tend to be polyclonal. Although assays for types of HPV other than those that were investigated in this study are now available, and although some of these more recently identified types may be oncogenic, it is unlikely that their omission from this study influenced the results that pertain to types 16, 18, and 31/33/35/39, because only 14 women with invasive disease and 12 with in situ tumors had untyped HPV DNA in their cervical scrapings. The observation that untyped HPV types are more strongly associated with in situ than invasive disease would probably be strengthened if we had tested for the more recently recognized oncogenic types and removed them from the group with untyped DNA; our conclusion that there are HPV types associated with in situ disease that do not progress to invasion would not change. After controlling for age and HPV type, the epidemiologic features of in situ and invasive cervical carcinomas were found to be quite similar. The age at first intercourse was a strong risk factor for both tumor types in this largely monogamous population, suggesting either that early exposure to oncogenic types of HPV may enhance the risk of persistent infection and carcinogenesis, or that the sexual behavior of the husbands of women with early and late onset of sexual activity may differ (17). Serologic indices of exposure to sexually transmitted agents other than HPV were generally not strongly related to either invasive or in situ disease, and none was significantly more strongly related to one tumor type or the other. This is as expected, because these factors are more likely indicators of risk of initial infection with HPV rather than cofactors operative after neoplastic changes. Although quite possibly due to chance, the risk in relation to serologic evidence of hepatitis B surface antigen (table 4) was slightly greater for invasive than in situ disease. Because persistent hepatitis B virus antigenemia may suggest a deficient immune response, this observation is compatible with a role for immunodeficiency in the persistence of HPV infection that can lead to invasive carcinoma. An increase in the risk of both in situ and invasive disease with parity and use of oral contraceptives was observed, but the associations were not significantly stronger for invasive disease, suggesting that these hormonal factors operate prior to the development of in situ lesions. The risk of carcinoma in situ, but not of invasive disease, was elevated in users of depot-medroxyprogesterone acetate. This same observation was previously observed in a larger study conducted in Thailand and elsewhere (28, 29) and is compatible with the interpretation that any effect of depot-medroxyprogesterone acetate on the risk of cervical carcinoma in situ is reversible or that the lesions induced by this product have a low invasive potential. Smoking was weakly associated with the risk of both invasive and in situ lesions, but the prevalence of smoking among the study subjects was low and the odds ratio estimate had wide confidence limits. If smoking is a cofactor, it operates prior to the development of in situ disease and is not an important determinant of risk in Thailand. Having had a screening chest radiographic examination and having ever attended school were both associated with a reduced risk of invasive disease but not of in situ disease. These associations persisted after controlling for the frequency of prior cervical smear screening, suggesting that these associations are not due to more cervical cancer screening in the more educated women and in women who are screened for tuberculosis. Another possible explanation is that women of higher socioeconomic status may tend to seek care earlier than women of lower status and, hence, are more likely to have their cervical neoplasia diagnosed when still in a preinvasive stage. However, the observation that the presenting symptoms of women with in situ and invasive disease did not differ is not supportive of this interpretation. Other studies have also shown odds ratios in relation to school attendance to be less than unity for invasive but not in situ disease (10, 14, 15, 30), and the consistency of these results and ours suggests that factors associated with education or other measures of socioeconomic status, such as nutritional factors, may inhibit invasion. Prior cervical smears were shown to be related to a reduced risk of cervical carcinoma and, as expected, to be more strongly protective against invasive than in situ disease. As has also been observed elsewhere (31), reduced risks in relation to use of an intrauterine device and a tubal ligation most likely reflect screening for cervical cancer at the time of intrauterine device insertion or removal and at the time of tubal surgery. Although there has been only one prior report of studies in which direct comparisons of risk factors for in situ and invasive carcinomas were made after controlling for HPV type in the tumor (10), there have been other investigations in which women with both in situ and invasive disease have been studied using comparable methodology but without inclusion of HPV testing (13–15, 30). The results of all these studies are generally consistent with our findings in showing no sexual or hormonal factors, or smoking, to be more strongly associated with invasive than in situ carcinomas. Additional studies of the role of these factors in the genesis of cervical carcinoma should focus on the more proximal preinvasive stages of the carcinogenic process. Several possible sources of bias could have influenced the odds ratio estimates in this study. If women with carcinoma in situ were more likely than women with invasive carcinoma to have been screened for cervical cancer, then observed differences between women with these two tumor types could reflect factors associated with screening behavior rather than with development of invasive disease. This is an unlikely explanation for our results. There is no mass screening of asymptomatic women in Thailand, and cases of in situ disease in this study did not differ from women with invasive disease with respect to the presenting symptoms that lead to diagnosis. In addition, controlling odds ratio estimates for history of prior cervical smears had no appreciable effect on the results. Confounding by HPV status could have influenced the odds ratio estimates based on comparisons of cases and controls, although the consistency of our results with those of others suggests that this is not a likely explanation for the findings. More importantly, the direct comparisons of women with invasive and in situ carcinomas were controlled for HPV type, thus eliminating confounding by these viruses as an explanation for the few associations observed in those analyses. The comparison of invasive with in situ disease is an investigative approach that other investigators should consider in an attempt to identify factors that may enhance or prevent progression of intraepithelial lesions to invasive disease. Because HPV type was the only predictor of progression identified, other studies should focus on additional factors not adequately considered in this and other investigations. Correspondence to Dr. David B. Thomas, Program in Epidemlogy, Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue North, Seattle, WA 98109 (e-mail: [email protected]). This study was supported by grant CA49044 from the US National Institutes of Health. The assistance of Drs. Anna Marie Beckmann and Larry Corey is gratefully acknowledged. REFERENCES 1. International Agency for Research on Cancer. Human papillomaviruses. IARC Monogr Eval Carcinog Risks Hum  1995; 64: 1–409. Google Scholar 2. Cavalcanti SM, Deus FC, Zardo LG, et al. Human papillomavirus infection and cervical cancer in Brazil: a retrospective study. Mem Inst Oswaldo Cruz  1996; 91: 433–40. Google Scholar 3. Shah KV, Viscidi RP, Alberg AJ, et al. Antibodies to human papillomavirus 16 and subsequent in situ or invasive cancer of the cervix. Cancer Epidemiol Biomarkers Prev  1997; 6: 233–7. Google Scholar 4. Lehtinen M, Dillner J, Knekt P, et al. Serologically diagnosed infection with human papillomavirus type 16 and risk for subsequent development of cervical carcinoma: nested case-control study. BMJ  1996; 312: 537–9. Google Scholar 5. Bosch FX, Muñoz N, de Sanjosé S, et al. Risk factors for cervical cancer in Colombia and Spain. Int J Cancer  1992; 52: 750–8. Google Scholar 6. Chaouk N, Bosch FX, Muñoz N, et al. The viral origin of cervical cancer in Rabat, Morocco. Int J Cancer  1998; 75: 546–54. Google Scholar 7. Chichareon S, Herrero R, Muñoz N, et al. Risk factors for cervical cancer in Thailand: a case-control study. J Natl Cancer Inst  1998; 90: 50–7. Google Scholar 8. Ngelangel C, Muñoz N, Bosch FX, et al. Causes of cervical cancer in the Philippines: a case-control study. J Natl Cancer Inst  1998; 90: 43–9. Google Scholar 9. Daling JR, Madeleine MM, McKnight B, et al. The relationship of human papillomavirus-related cervical tumors to cigarette smoking, oral contraceptive use, and prior herpes simplex virus type 2 infection. Cancer Epidemiol Biomarkers Prev  1996; 5: 541–8. Google Scholar 10. Moreno V, Muñoz N, Bosch FX, et al. Risk factors for progression of cervical intraepithelial neoplasms grade III to invasive cervical cancer. Cancer Epidemiol Biomarkers Prev  1995; 4: 459–67. Google Scholar 11. Muñoz N, Bosch FX, de Sanjosé S, et al. Risk factors for cervical intraepithelial neoplasia grade III/carcinoma in situ in Spain and Colombia. Cancer Epidemiol Biomarkers Prev  1993; 2: 423–31. Google Scholar 12. Ferrera A, Baay MF, Herbrink P, et al. A sero-epidemiological study of the relationship between sexually transmitted agents and cervical cancer in Honduras. Int J Cancer  1997; 73: 781–5. Google Scholar 13. Parazzini F, La Vecchia C, Negri E, et al. Reproductive factors and the risk of invasive and intraepithelial cervical neoplasia. Br J Cancer  1989; 59: 805–9. Google Scholar 14. Jones CJ, Brinton LA, Hamman RF, et al. Risk factors for in situ cervical cancer: results from a case-control study. Cancer Res  1990; 50: 3657–62. Google Scholar 15. Brinton LA, Tashima KT, Lehman HF, et al. Epidemiology of cervical cancer by cell type. Cancer Res  1987; 47: 1706–11. Google Scholar 16. Begg CB, Zhang ZF. Statistical analysis of molecular epidemiology studies employing case-series. Cancer Epidemiol Biomarkers Prev  1994; 3: 173–5. Google Scholar 17. Thomas DB, Ray RM, Koetsawang A, et al. Human papillomaviruses and cervical cancer in Bangkok. I. Risk factors for invasive cervical carcinomas with human papillomavirus types 16 and 18 DNA. Am J Epidemiol  2001; 153: 723–31. Google Scholar 18. Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol I. The analysis of case-control studies. Lyon, France: International Agency for Research on Cancer, 1980:325–38. (IARC scientific publication no. 32). Google Scholar 19. Bosch FX, Muñoz N, de Sanjosé S, et al. Human papillomavirus and cervical intraepithelial neoplasia grade III/carcinoma in situ: a case-control study in Spain and Colombia. Cancer Epidemiol Biomarkers Prev  1993; 2: 415–22. Google Scholar 20. Ho GY, Kadish AS, Burk RD, et al. HPV 16 and cigarette smoking as risk factors for high-grade cervical intraepithelial neoplasia. Int J Cancer  1998; 78: 281–5. Google Scholar 21. Kjaer SK, van den Brule AJ, Bock JE, et al. Human papillomavirus—the most significant risk determinant of cervical intraepithelial neoplasia. Int J Cancer  1996; 65: 601–6. Google Scholar 22. van den Brule AJ, Walboomers JM, du Maine M, et al. Difference in prevalence of human papillomavirus genotypes in cytomorphologically normal cervical smears is associated with a history of cervical intraepithelial neoplasia. Int J Cancer  1991; 48: 404–8. Google Scholar 23. Becker TM, Wheeler CM, McGough NS, et al. Cervical papillomavirus infection and cervical dysplasia in Hispanic, Native American, and non-Hispanic white women in New Mexico. Am J Public Health  1991; 81: 582–6. Google Scholar 24. Cuzick J, Terry G, Ho L, et al. Human papillomavirus type 16 DNA in cervical smears as predictor of high-grade cervical cancer. Lancet  1992; 339: 959–60. Google Scholar 25. Lungu O, Sun XW, Felix J, et al. Relationship of human papillomavirus type to grade of cervical intraepithelial neoplasia. JAMA  1992; 267: 2493–6. Google Scholar 26. Coker AL, Jenkins GR, Busnardo MS, et al. Human papillomaviruses and cervical neoplasia in South Carolina. Cancer Epidemiol Biomarkers Prev  1993; 2: 207–12. Google Scholar 27. Park TJ, Richart RM, Sun XW, et al. Association between human papillomavirus type and clonal status of cervical squamous intraepithelial lesions. J Natl Cancer Inst  1996; 88: 355–8. Google Scholar 28. The WHO Collaborative Study of Neoplasia and Steroid Contraceptives. Depot-medroxyprogesterone acetate (DMPA) and risk of invasive squamous cell cervical cancer. Contraception  1992; 45: 299–312. Google Scholar 29. Thomas DB, Ye Z, Ray RM, et al. Cervical carcinoma in situ and use of depot-medroxyprogesterone acetate (DMPA). Contraception  1995; 51: 25–31. Google Scholar 30. Kjaer SK, Dahl C, Engholm G, et al. Case-control study of risk factors for cervical neoplasia in Denmark. II. Role of sexual activity, reproductive factors, and venereal infections. Cancer Causes Control  1992; 3: 339–48. Google Scholar 31. Li HQ, Thomas DB, Jin SK, et al. Tubal sterilization and use of an IUD and risk of cervical cancer. J Womens Health Gend Based Med  2000; 9: 303–10. Google Scholar

Thomas, David B.;Ray, Roberta M.;Kuypers, Jane;Kiviat, Nancy;Koetsawang, Amorn;Ashley, Rhoda L.;Qin, Qin;Koetsawang, Suporn

doi: 10.1093/aje/153.8.740pmid: 11296145

Abstract Between September 1991 and September 1993, husbands of women with and without cervical neoplasia and commercial sex workers in one brothel and one massage parlor in Bangkok, Thailand, were interviewed; serologic tests for sexually transmitted infections were performed; and cervical and penile scrapings were tested for human papillomavirus (HPV) DNA. The risks of cervical carcinoma in monogamous women and of oncogenic HPV in their husbands were associated with the men's having unprotected intercourse with prostitutes. The prevalence of oncogenic HPV was higher in commercial sex workers than in women attending gynecologic and family planning clinics. Oncogenic HPV prevalence declined with age in human immunodeficiency virus (HIV)-negative, but not in healthy HIV-positive, commercial sex workers and was weakly associated with hepatitis B antigenemia, suggesting that persistence of HPV infection is due to subtle changes in immunity. Associations of HPV with recent pregnancy and oral contraceptive use suggest that hormonal factors may increase the risk of cervical neoplasia by enhancing persistence of HPV infection. The prevalence of high-grade squamous intraepithelial lesions was strongly related to oncogenic HPV types and weakly to HIV infection only in their presence. Commercial sex workers in Bangkok are reservoirs of oncogenic HPV, and cervical cancer in monogamous Thai women develops in part as a result of transmission of these viruses to them by their husbands from prostitutes. cervix neoplasms, papillomavirus, human, prostitution, sexual partners DMPA, depot-medroxyprogesterone acetate, HBsAg, hepatitis B surface antigen, HGSIL, high-grade squamous intraepithelial lesions, HIV, human immunodeficiency virus, HPV, human papillomavirus. It is now generally accepted that certain types of human papillomavirus (HPV) play an important role in the genesis of cervical carcinoma (1). These viruses are sexually transmitted (1, 2) and, in cultures in which most women are virgins at marriage and monogamous afterward, the sexual behavior of men is more important than the sexual behavior of women in the transmission of oncogenic types of HPV and the genesis of cervical cancer in their wives (3–16). The prevalence of oncogenic types of HPV in men has been associated with contacts with prostitutes (17, 18), number of sexual partners (17–19), and evidence of prior sexually transmitted disease (18, 19), providing further evidence of the importance of male sexual behavior in the transmission of these agents. In a prior study of cervical cancer in Thailand (10), based on interviews of the husbands of monogamous cases and controls, the risk of invasive squamous cell cervical carcinoma was shown to be associated with the husbands' having visited prostitutes without using a condom when the husbands were less than 30 years old. The prostitutes presumably serve as reservoirs of oncogenic types of HPV, and the husbands serve as vectors of transmission to their wives. This hypothesis is supported by observations that prostitutes are at increased risk of cervical cancer (20) and have an elevated prevalence of cervical intraepithelial neoplasia (21) and HPV infection (17). This is a report of results from interviews of husbands of women with and without cervical cancer and from a survey of prostitutes in Bangkok that were conducted to further clarify the role of prostitutes as reservoirs of oncogenic HPV types and the determinants of HPV infections and their early neoplastic consequences in these heavily exposed women. MATERIALS AND METHODS Commercial sex workers in Bangkok work in a variety of establishments that cater to different types of men. Massage parlors tend to serve local men of middle and high socio-economic status and foreigners. Brothels provide sexual services largely to local men of low socioeconomic status. This study was conducted in one large massage parlor and one brothel in Bangkok, Thailand. We included in this study 170 women who were working as masseuses and prostitutes in the massage parlor and 84 prostitutes who were working in the brothel at the time of the study workers' visits. Specially trained and experienced female study workers administered a standardized questionnaire to obtain information on date and place of birth, marital status, educational level, sexual history, history of sexually transmitted diseases, reproductive history, use of steroid and barrier contraceptives, and use of tobacco and alcohol. Information on history of sexually transmitted diseases was supplemented by reviewing medical records in clinics where the study subjects were treated. In two case-control studies conducted in conjunction with this investigation (22, 23), women with histologically confirmed in situ or invasive cervical cancer, diagnosed between September 1991 and September 1993, who were born after 1929, who had resided in Thailand for at least the previous year, and who were treated in Siriraj Hospital, Bangkok, were recruited as cases. Two controls were selected for each woman with invasive disease from otolaryngology and general surgery wards of Siriraj Hospital, and two controls for each case of carcinoma in situ were selected from among women with normal cervical smears who attended the same clinics from which the corresponding cases came. An attempt was made to interview the current husbands of all women. The same questionnaire that was used in a prior study in Thailand (10) was used to ascertain detailed information on the husbands' sexual history during each decade of adult life, including numbers of partners, visits to prostitutes, and use of condoms. Information was also ascertained on history of sexually transmitted diseases, genital hygiene, circumcision status, reproductive history, and use of tobacco and alcohol. The prevalence of HPV DNA in the commercial sex workers was compared with the prevalence in women attending gynecologic clinics and in women attending family planning clinics at Siriraj Hospital. Controls for the case-control study of cervical intraepithelial neoplasia (23) were selected from these clinics, and the two comparison groups included those controls, plus additional women selected in the same manner from each of the two clinics. Thai women who had resided in Thailand for at least the past year, who were born after 1929, who were returning to the clinics to learn the results of a recent cervical smear, and whose cervical smear revealed no suspicion of neoplastic change were randomly sampled and interviewed using the same questionnaire that was used in the case-control study. A 15-ml blood specimen was obtained from the interviewed men and women. Four aliquots of serum were stored at −70°C, two of which were periodically shipped to Seattle on dry ice. As described previously (22), serum samples were assayed for antibodies to human immunodeficiency virus (HIV), hepatitis B, Treponema pallidum, and herpes simples virus types 1 and 2, as well as for hepatitis B surface antigen (HBsAg). Cervical smears were obtained from all women on the same day they were interviewed, as were scrapings for HPV DNA assays by scraping the cervix, including the cervical os, with a Teflon (E. I. du Pont de Nemours and Company, Wilmington, Delaware)-coated swab. An attempt was also made to obtain penile specimens for HPV assays from the interviewed husbands. The coronal sulcus was cleaned of visible smegma with dry gauze, and the area was then scraped with a Teflon-coated swab, as was the urethral opening. The Teflon-coated swabs were broken off into tubes of specimen transport medium (Digene Diagnostics, Inc., Beltsville, Maryland), and the specimens were stored at −70°C and periodically shipped to Seattle, Washington, on dry ice. All cervical smears from the commercial sex workers were read by a single pathologist in Bangkok. Slides with suspicious lesions were reviewed by a pathologist in Seattle, and those designated high-grade squamous intraepithelial lesions (HGSIL) by the latter were considered as such for this report. As detailed previously (22), cervical and penile scrapings were tested for evidence of any HPV DNA and for type-specific DNA of types 6 and 11, type 16, type 18, types 31, 33, 35, and 39, and type 45 using polymerase chain reaction-based assays. The cervical scrapings from four commercial sex workers who were HIV positive and who had HGSIL and from 30 HIV-positive sex workers with normal cervical cytology were tested for HIV type 1 DNA by nested polymerase chain reaction of the gag gene (24). The sexual behavior and HPV status of the husbands of the cases and controls were compared, as were the results of HPV assays in women and their husbands. When appropriate, odds ratios, as estimates of relative risks, were calculated using unconditional logistic regression methods (25). Data from the commercial sex workers were analyzed to estimate the prevalence odds ratios and their 95 percent confidence intervals for specific types of HPV and for HGSIL using unconditional logistic regression (25), and 95 percent confidence intervals were estimated using exact logistic regression when analysis included cells with zero observations (26). RESULTS Study of husbands Interviews were conducted with 50 (28.6 percent) of the current husbands of 175 women with invasive squamous cell cervical carcinoma who gave a history of a single sexual partner and with 98 (36.0 percent) of the husbands of 272 allegedly monogamous hospital controls. Consistent with previous observations from a larger study in Thailand with higher response rates (10), a trend of increasing risk of invasive squamous cell cervical carcinoma with increasing estimated number of lifetime visits to prostitutes by the women's husbands was observed (table 1), and risk was associated with the husband's number of visits to prostitutes per year when he was in his teens and twenties, but not at a later age, and with his using condoms less than 10 percent of the time at those ages (not shown). After controlling for lifetime visits to prostitutes, risk was not associated with husbands' history of sexually transmitted diseases, number of sexual partners other than prostitutes, herpes simples virus type 2 antibodies in the husband's serum, or his smoking history (not shown). TABLE 1. Odds ratios of invasive squamous cell cervical cancer in monogamous women in relation to their husband's experience with prostitutes, Bangkok, Thailand, 1991–1993 Estimated lifetime visits to prostitutes  No. of women   OR*,†,‡  95% CI‡  Cases  Controls  None  10  32  1.0  Reference  1–80  15  29  1.8  0.6, 5.2  81–280  8  15  1.7  0.5, 5.8  >280  15  15  3.2  1.0, 10.9  Unknown number  2  7      Estimated lifetime visits to prostitutes  No. of women   OR*,†,‡  95% CI‡  Cases  Controls  None  10  32  1.0  Reference  1–80  15  29  1.8  0.6, 5.2  81–280  8  15  1.7  0.5, 5.8  >280  15  15  3.2  1.0, 10.9  Unknown number  2  7      * Adjusted for woman's age and husband's history of sexually transmitted diseases. † p value for test of trend = 0.08. ‡ OR, odds ratio; CI, confidence interval. View Large Penile scrapings were obtained from 57 husbands of women with in situ or invasive disease (39 of whom were monogamous) and from 68 husbands of control women (56 being monogamous). Only eight men (6.4 percent) tested positive for any oncogenic HPV (including types 16, 18, 31, 33, 35, and 39), and the prevalence was not significantly higher in the husbands of the cases (7.0 percent) than in the husbands of the controls (5.9 percent). Moreover, the 44 husbands of cases with any oncogenic HPV did not have a higher prevalence of an oncogenic HPV type (6.8 percent) than did the 13 husbands of cases with no oncogenic type (7.7 percent). Compared with 79 interviewed husbands with no oncogenic HPV, a higher proportion of the seven interviewed husbands with such an infection reported a history of gonorrhea, multiple sexual partners, visits to prostitutes, and no use of condoms during such visits, although none of the differences reached statistical significance (not shown). The 45 husbands married to women with invasive disease had a lower prevalence of oncogenic HPV (4.4 percent) than did the 12 husbands of women with in situ disease (16.7 percent), although the difference could have occurred by chance (p > 0.05). Study of commercial sex workers Women who worked in the brothel were younger than the women who worked in the massage parlor (table 2), both the mean and median ages being 19 and 30 years, respectively. The prevalence of all types of HPV considered was significantly (p < 0.05) higher in the brothel workers than in the massage parlor workers. With the exception of HPV types 6/11, the prevalence of HPV DNA in cervical scrapings decreased with age in the brothel workers, but not appreciably so in the women in the massage parlor. Even the relatively low prevalence of HPV DNA in the massage parlor workers was higher than that in women of comparable age attending gynecologic and family planning clinics at Siriraj Hospital (table 3). Although women from those two clinics with smears suspicious of neoplasia are not represented, when the seven women from the massage parlor with HGSIL were omitted from the analyses, the results were similar to those presented. TABLE 2. Age-specific prevalence (%) of specific types of human papillomavirus DNA in cervical scrapings of commercial sex workers from a brothel and a massage parlor, Bangkok, Thailand, 1991–1993 Place of employment and age (years)  Total no. tested  Subjects tested positive by HPV* type   6/11   16   18   31/33/35/39   45   No.  %  No.  %  No.  %  No.  %  No.  %  Brothel†                         15–19  48  13  27.1  15  31.3  6  12.5  14  29.2  3  6.3   20–24  34  10  29.4  6  17.6  1  2.9  8  23.5  1  2.9  Massage parlor‡                         20–24  40  1  2.5  4  10.0  1  2.5  3  7.5  0  0   25–29  39  1  2.6  3  7.7  3  7.7  4  10.3  1  2.6   30–34  41  1  2.4  3  7.3  1  2.4  2  4.9  1  2.4  ≥35  49  1  2.0  4  8.1  3  6.1  4  8.2  0  0  Place of employment and age (years)  Total no. tested  Subjects tested positive by HPV* type   6/11   16   18   31/33/35/39   45   No.  %  No.  %  No.  %  No.  %  No.  %  Brothel†                         15–19  48  13  27.1  15  31.3  6  12.5  14  29.2  3  6.3   20–24  34  10  29.4  6  17.6  1  2.9  8  23.5  1  2.9  Massage parlor‡                         20–24  40  1  2.5  4  10.0  1  2.5  3  7.5  0  0   25–29  39  1  2.6  3  7.7  3  7.7  4  10.3  1  2.6   30–34  41  1  2.4  3  7.3  1  2.4  2  4.9  1  2.4  ≥35  49  1  2.0  4  8.1  3  6.1  4  8.2  0  0  * HPV, human papillomavirus. † Excluding two women aged 25–29 years. ‡ Excluding one woman aged 15–19 years. View Large TABLE 3. Age-specific prevalence (%) of specific types of oncogenic human papillomavirus DNA in cervical scrapings from commercial sex workers in a massage parlor and from women attending gynecologic and family planning clinics, Bangkok, Thailand, 1991–1993 Age (years)  No. of women tested   Prevalence (%)   HPV* 16/18   HPV 31/33/35/39   Massage parlor  Gynecologic clinic  Family planning clinic  Massage parlor  Gynecologic clinic  Family planning clinic  Massage parlor  Gynecologic clinic  Family planning clinic  25–29  39  21  5  15.4  9.5  0.0  10.3  9.5  20.0  30–34  41  40  31  7.3  5.0  3.2  4.9  0.0  3.2  ≥35  49  67  48  14.3  6.0  0.0  8.2  0.0  2.1  Age (years)  No. of women tested   Prevalence (%)   HPV* 16/18   HPV 31/33/35/39   Massage parlor  Gynecologic clinic  Family planning clinic  Massage parlor  Gynecologic clinic  Family planning clinic  Massage parlor  Gynecologic clinic  Family planning clinic  25–29  39  21  5  15.4  9.5  0.0  10.3  9.5  20.0  30–34  41  40  31  7.3  5.0  3.2  4.9  0.0  3.2  ≥35  49  67  48  14.3  6.0  0.0  8.2  0.0  2.1  * HPV, human papillomavirus. View Large Possible risk factors for HPV types 6/11, 16, and 31/33/35/39 are shown in tables 4, 5, and 7. Too few women were infected with other types for meaningful analysis. All prevalence odds ratios are adjusted for age and place of employment using the strata shown in table 2. The prevalence of none of the HPV types was significantly related to the number of customers per day, the days worked per month, the age at which the woman began working in the sex industry, or the frequency of condom use. The frequency of condom use was ascertained for each decade of sexual activity, and women whose average frequency of condom use was 50 percent or greater are in the “more frequently” category in the table; the remaining women are in the “less frequently” group. The prevalence of type 16 and of types 31/33/35/39 was higher in women with than without history of a sexually transmitted disease in the past year, although the 95 percent confidence limits of the prevalence odds ratios include unity. The risk of prevalent HPV types 6/11 and 31/33/35/39 was lower in women who had worked as prostitutes for 2 or more years than for newer employees, but those results could also readily be due to chance. TABLE 4. Prevalence odds ratios* of human papillomavirus types 6/11, 16, and 31/33/35/39 DNA in cervical scrapings from commercial sex workers in relation to sexual variables ascertained from interviews, Bangkok, Thailand, 1991–1993 Factor  Total no. tested  HPV† type   6/11   16   31/33/35/39   Tested positive (no.)  POR†  95% CI†  Tested positive (no.)  POR  95% CI  Tested positive (no.)  POR  95% CI  Customers/day                       1–2  107  13  1.0  Reference  12  1.0  Reference  9  1.0  Reference   3  94  9  0.4  0.1, 1.0  17  1.4  0.6, 3.4  18  2.2  0.9, 5.5  ≥4  50  5  0.3  0.1, 1.1  6  0.7  0.2, 2.0  8  1.4  0.5, 4.3  Days work/month                       <20  122  6  1.0  Reference  12  1.0  Reference  12  1.0  Reference   21–25  70  12  0.9  0.2, 3.3  11  0.7  0.2, 2.1  13  0.8  0.3, 2.4  ≥26  59  9  0.6  0.2, 2.5  12  1.0  0.3, 2.8  10  0.7  0.2, 2.0  Age began working as CSW†                      ≥23  116  5  1.0  Reference  8  1.0  Reference  8  1.0  Reference   19–22  68  6  1.1  0.01, 1.1  12  2.2  0.5, 8.7  12  2.2  0.5, 8.5   15–18  67  16  0.4  0.02, 3.0  15  1.1  0.2, 6.3  15  1.5  0.2, 8.2  Years since began work as CSW                      ≥2  147  7  1.0  Reference  12  1.0  Reference  11  1.0  Reference   1  50  9  1.2  0.3, 4.3  13  2.1  0.7, 5.9  10  2.0  0.6, 6.1   <1  54  11  1.4  0.5, 4.1  10  1.0  0.3, 3.5  14  2.3  0.7, 7.8  Condom use                       More frequently  208  18  1.0  Reference  28  1.0  Reference  28  1.0  Reference   Less frequently  43  9  1.7  0.6, 4.3  7  0.9  0.3, 2.2  7  0.9  0.3, 2.1  History of STD† in past year                       No  172  15  1.0  Reference  17  1.0  Reference  17  1.0  Reference   Yes  79  12  0.9  0.4, 2.2  18  1.8  0.8, 4.1  18  1.9  0.8, 4.5  Factor  Total no. tested  HPV† type   6/11   16   31/33/35/39   Tested positive (no.)  POR†  95% CI†  Tested positive (no.)  POR  95% CI  Tested positive (no.)  POR  95% CI  Customers/day                       1–2  107  13  1.0  Reference  12  1.0  Reference  9  1.0  Reference   3  94  9  0.4  0.1, 1.0  17  1.4  0.6, 3.4  18  2.2  0.9, 5.5  ≥4  50  5  0.3  0.1, 1.1  6  0.7  0.2, 2.0  8  1.4  0.5, 4.3  Days work/month                       <20  122  6  1.0  Reference  12  1.0  Reference  12  1.0  Reference   21–25  70  12  0.9  0.2, 3.3  11  0.7  0.2, 2.1  13  0.8  0.3, 2.4  ≥26  59  9  0.6  0.2, 2.5  12  1.0  0.3, 2.8  10  0.7  0.2, 2.0  Age began working as CSW†                      ≥23  116  5  1.0  Reference  8  1.0  Reference  8  1.0  Reference   19–22  68  6  1.1  0.01, 1.1  12  2.2  0.5, 8.7  12  2.2  0.5, 8.5   15–18  67  16  0.4  0.02, 3.0  15  1.1  0.2, 6.3  15  1.5  0.2, 8.2  Years since began work as CSW                      ≥2  147  7  1.0  Reference  12  1.0  Reference  11  1.0  Reference   1  50  9  1.2  0.3, 4.3  13  2.1  0.7, 5.9  10  2.0  0.6, 6.1   <1  54  11  1.4  0.5, 4.1  10  1.0  0.3, 3.5  14  2.3  0.7, 7.8  Condom use                       More frequently  208  18  1.0  Reference  28  1.0  Reference  28  1.0  Reference   Less frequently  43  9  1.7  0.6, 4.3  7  0.9  0.3, 2.2  7  0.9  0.3, 2.1  History of STD† in past year                       No  172  15  1.0  Reference  17  1.0  Reference  17  1.0  Reference   Yes  79  12  0.9  0.4, 2.2  18  1.8  0.8, 4.1  18  1.9  0.8, 4.5  * Prevalence odds ratio adjusted for age and place of employment. † HPV, human papillomavirus; POR, prevalence odds ratio; CI, confidence interval; CSW, commercial sex worker; STD, sexually transmitted disease. View Large As shown in table 5, the prevalence of all three groups of HPV types was only slightly higher in HIV-positive than HIV-negative women; somewhat more strongly associated with HBsAg; and slightly lower for women with than without a positive test for syphilis, antibodies to herpes simplex virus type 2, and antibodies to hepatitis B surface and core antigens. However, all of the prevalence odds ratio estimates have 95 percent confidence intervals that include one. Although the overall prevalence of the three HPV groups was not higher in the brothel workers with than without HIV infection (table 6), the prevalence of HPV types 16 and 31/33/35/39 declined with age in women without HIV infection but not in women who were HIV positive. A similar analysis could not be performed for the massage parlor workers because only three of them were HIV positive. TABLE 5. Prevalence odds ratios* of human papillomavirus types 6/11, 16, and 31/33/35/39 DNA in cervical scrapings from commercial sex workers in relation to serologic factors, Bangkok, Thailand, 1991–1993 Factor  Total no. tested  HPV† type   6/11   16   31/33/35/39   Tested positive (no.)  POR†  95% CI†  Tested positive (no.)  POR  95% CI  Tested positive (no.)  POR  95% CI  HIV† serology                       Negative  214  17  1.0  Reference  26  1.0  Reference  25  1.0  Reference   Positive  37  10  1.1  0.4, 2.9  9  1.2  0.4, 3.2  10  1.2  0.5, 3.3  HSV-2† antibody‡                       Negative  20  3  1.0  Reference  1  1.0  Reference  6  1.0  Reference   Positive  224  22  0.9  0.3, 4.6  30  –§    28  0.4  0.1, 1.4  Syphilis serology¶                       Negative  211  25  1.0  Reference  32  1.0  Reference  31  1.0  Reference   Positive  39  2  0.4  0.1, 1.8  3  0.6  0.1, 1.4  4  0.7  0.2, 2.2  HBsAg†                       Negative  230  23  1.0  Reference  30  1.0  Reference  30  1.0  Reference   Positive  21  4  1.4  0.3, 4.6  5  1.6  0.5, 4.7  5  1.6  0.5, 4.6  Anti-HBs antibody†                       Negative  89  17  1.0  Reference  20  1.0  Reference  19  1.0  Reference   Positive  162  10  0.5  0.2, 1.1  15  0.5  0.2, 1.0  16  0.6  0.3, 1.4  Anti-HBc antibody†                       Negative  42  7  1.0  Reference  10  1.0  Reference  10  1.0  Reference   Positive  209  20  0.9  0.3, 2.5  25  0.6  0.3, 1.4  25  0.6  0.3, 1.4  Factor  Total no. tested  HPV† type   6/11   16   31/33/35/39   Tested positive (no.)  POR†  95% CI†  Tested positive (no.)  POR  95% CI  Tested positive (no.)  POR  95% CI  HIV† serology                       Negative  214  17  1.0  Reference  26  1.0  Reference  25  1.0  Reference   Positive  37  10  1.1  0.4, 2.9  9  1.2  0.4, 3.2  10  1.2  0.5, 3.3  HSV-2† antibody‡                       Negative  20  3  1.0  Reference  1  1.0  Reference  6  1.0  Reference   Positive  224  22  0.9  0.3, 4.6  30  –§    28  0.4  0.1, 1.4  Syphilis serology¶                       Negative  211  25  1.0  Reference  32  1.0  Reference  31  1.0  Reference   Positive  39  2  0.4  0.1, 1.8  3  0.6  0.1, 1.4  4  0.7  0.2, 2.2  HBsAg†                       Negative  230  23  1.0  Reference  30  1.0  Reference  30  1.0  Reference   Positive  21  4  1.4  0.3, 4.6  5  1.6  0.5, 4.7  5  1.6  0.5, 4.6  Anti-HBs antibody†                       Negative  89  17  1.0  Reference  20  1.0  Reference  19  1.0  Reference   Positive  162  10  0.5  0.2, 1.1  15  0.5  0.2, 1.0  16  0.6  0.3, 1.4  Anti-HBc antibody†                       Negative  42  7  1.0  Reference  10  1.0  Reference  10  1.0  Reference   Positive  209  20  0.9  0.3, 2.5  25  0.6  0.3, 1.4  25  0.6  0.3, 1.4  * Prevalence odds ratio adjusted for age and place of employment. † HPV, human papillomavirus; POR, prevalence odds ratio; CI, confidence interval; HIV, human immunodeficiency virus; HSV-2, herpes simplex virus type 2; HBsAg, hepatitis B surface antigen; anti-HBs antibody, anti-hepatitis B surface antibody; anti-HBc antibody, anti-hepatitis B core antibody. ‡ Seven women with unknown type of herpes simplex virus antibody were excluded. § –, indeterminate. ¶ One woman with no syphilis serology available was excluded. View Large TABLE 6. Age-specific prevalence (%) of specific types of human papillomavirus DNA in cervical scrapings of brothel workers with and without positive tests for human immunodeficiency virus, Bangkok, Thailand, 1991–1993 HIV* serology  Age (years)  Total no. tested  HPV† type tested positive   6/11   16   31/33/35/39   No.  %  No.  %  No.  %  Positive  15–19  16  5  31.2  4  25.0  4  25.0    20–24  18  5  27.8  5  27.8  6  33.3    Total  34  10  29.4  9  24.5  10  29.4  Negative  15–19  32  8  25.0  11  34.4  10  31.2    20–24  16  5  31.2  1  6.2  2  12.5    Total  48  13  27.1  12  25.0  12  25.0  HIV* serology  Age (years)  Total no. tested  HPV† type tested positive   6/11   16   31/33/35/39   No.  %  No.  %  No.  %  Positive  15–19  16  5  31.2  4  25.0  4  25.0    20–24  18  5  27.8  5  27.8  6  33.3    Total  34  10  29.4  9  24.5  10  29.4  Negative  15–19  32  8  25.0  11  34.4  10  31.2    20–24  16  5  31.2  1  6.2  2  12.5    Total  48  13  27.1  12  25.0  12  25.0  * HIV, human immunodeficiency virus; HPV, human papillomavirus. View Large As shown in table 7, the risk of prevalent HPV type16 is associated with having ever been pregnant, especially if within the past 6 years. No trend with the number of livebirths is evident. A similar pattern is seen for types 31/33/35/39 but not for types 6/11. A possible increase in risk of all types is seen in relation to the current use of oral contraceptives, but these observations could be due to chance. No relation of risk to use of the injectable progestational contraceptive, depot-medroxyprogesterone acetate (DMPA), is evident. There were no associations of risk with alcohol or tobacco use (not shown). TABLE 7. Prevalence odds ratios* of human papillomavirus types 6/11, 16, and 31/33/35/39 DNA in cervical scrapings from commercial sex workers in relation to hormonal factors, Bangkok, Thailand, 1991–1993 Factor  Total no. tested  HPV† type   6/11   16   31/33/35/39   Tested positive (no.)  POR†  95% CI†  Tested positive (no.)  POR  95% CI  Tested positive (no.)  POR  95% CI  Pregnancy history                       Never pregnant  136  24  1.0  Reference  22  1.0  Reference  24  1.0  Reference   Ever pregnant  115  3  0.8  0.1, 5.1  13  3.5  1.0, 16  11  1.9  0.6, 7.5  No. of livebirths                       0  41  0  –‡    4  2.7  0.5, 15  5  2.4  0.6, 10   1  39  2  1.7  0.2, 15  6  4.9  1.1, 27  5  2.4  0.6, 11  ≥2  35  1  1.2  0.1, 14  3  3.0  0.5, 21  1  0.5  0.03, 4.0  Years since last pregnancy                       <6  51  1  0.5  0.03, 4.5  9  4.7  1.3, 23  5  1.8  0.5, 7.7  ≥6  64  2  1.3  0.1, 13  4  1.7  0.3, 10  6  2.1  0.5, 10  Use of oral contraceptives                       Never  35  1  1.0  Reference  1  1.0  Reference  1  1.0  Reference   Past  30  1  1.1  0.04, 31  2  2.4  0.2, 54  4  5.3  0.7, 107   Current  186  25  2.1  0.3, 40  32  4.4  0.8, 80  30  3.7  0.7, 70  DMPA† use                       Never  171  24  1.0  Reference  33  1.0  Reference  29  1.0  Reference   Past  52  2  0.8  0.1, 3.5  2  0.2  0.04, 0.9  2  0.3  0.1, 1.3   Current  28  1  0.5  0.02, 2.9  0  –    4  1.4  0.4, 4.3  Factor  Total no. tested  HPV† type   6/11   16   31/33/35/39   Tested positive (no.)  POR†  95% CI†  Tested positive (no.)  POR  95% CI  Tested positive (no.)  POR  95% CI  Pregnancy history                       Never pregnant  136  24  1.0  Reference  22  1.0  Reference  24  1.0  Reference   Ever pregnant  115  3  0.8  0.1, 5.1  13  3.5  1.0, 16  11  1.9  0.6, 7.5  No. of livebirths                       0  41  0  –‡    4  2.7  0.5, 15  5  2.4  0.6, 10   1  39  2  1.7  0.2, 15  6  4.9  1.1, 27  5  2.4  0.6, 11  ≥2  35  1  1.2  0.1, 14  3  3.0  0.5, 21  1  0.5  0.03, 4.0  Years since last pregnancy                       <6  51  1  0.5  0.03, 4.5  9  4.7  1.3, 23  5  1.8  0.5, 7.7  ≥6  64  2  1.3  0.1, 13  4  1.7  0.3, 10  6  2.1  0.5, 10  Use of oral contraceptives                       Never  35  1  1.0  Reference  1  1.0  Reference  1  1.0  Reference   Past  30  1  1.1  0.04, 31  2  2.4  0.2, 54  4  5.3  0.7, 107   Current  186  25  2.1  0.3, 40  32  4.4  0.8, 80  30  3.7  0.7, 70  DMPA† use                       Never  171  24  1.0  Reference  33  1.0  Reference  29  1.0  Reference   Past  52  2  0.8  0.1, 3.5  2  0.2  0.04, 0.9  2  0.3  0.1, 1.3   Current  28  1  0.5  0.02, 2.9  0  –    4  1.4  0.4, 4.3  * Prevalence odds ratio adjusted for age and place of employment. † HPV, human papillomavirus; POR, prevalence odds ratio; CI, confidence interval; DMPA, depot-medroxyprogesterone acetate. ‡ –, indeterminate. View Large Commensurate with the higher prevalence of oncogenic HPV types in the brothel workers than in the massage parlor workers, HGSIL were found in six (8.3 percent) of 72 women in the brothel with an adequate cervical smear but in only seven (4.3 percent) of 161 successfully screened women in the massage parlor. Unlike the prevalence of oncogenic HPV, the prevalence of HGSIL was higher in the older than in the younger women represented in both groups of commercial sex workers (not shown). Eight (61.8 percent) of the 13 women with HGSIL tested positive for HPV type 16 compared with 13.7 percent of all 233 tested women (prevalence odds ratio = 19.2; 95 percent confidence interval: 5.3, 77). Four of these eight women also had other oncogenic HPV types. Three of the remaining women with HGSIL had only an untyped HPV (prevalence odds ratio = 1.8; 95 percent confidence interval: 0.4, 6.2), and two tested negative for any HPV type. To provide larger numbers for analysis to evaluate the possible effect of HIV infection on the risk of HGSIL in the presence and absence of oncogenic HPV types, we combined the groups of women with HPV types 16 and 18. The risk of prevalent HGSIL was not associated with HIV infection in the absence of HPV type 16 or type 18 (table 8). In their presence, the risk of HGSIL is about doubled in women who also are infected with HIV, although this observation could have occurred by chance. TABLE 8. Prevalence (%) and prevalence odds ratios of high-grade squamous intraepithelial lesions of the cervix in commercial sex workers in relation to human papillomavirus type 16 or 18 DNA in cervical scrapings and human immuno-deficiency virus serologic test results, Bangkok, Thailand, 1991–1993 Test result   Total no. of women*  Women with HGSIL†   POR†  95% CI†,‡  HPV† 16/18  HIV†  No.  %  −  −  169  4  2.3  1.0  Reference  −  +  23  1  4.3  1.2  0.1, 13  +  −  31  5  16.1  10.2  2.4, 46  +  +  10  3  30.0  20.1  2.4, 204  Test result   Total no. of women*  Women with HGSIL†   POR†  95% CI†,‡  HPV† 16/18  HIV†  No.  %  −  −  169  4  2.3  1.0  Reference  −  +  23  1  4.3  1.2  0.1, 13  +  −  31  5  16.1  10.2  2.4, 46  +  +  10  3  30.0  20.1  2.4, 204  * A total of 18 women with inadequate cervical smears were excluded. † HGSIL, high-grade squamous intraepithelial lesions; POR, prevalence odds ratio adjusted for age and place of employment; CI, confidence interval; HPV, human papillomavirus; HIV, human immunodeficiency virus. ‡ pvalue of test for interaction = 0.7. View Large The cervical scrapings from the four women with HGSIL who were serologically positive for HIV and the scrapings from 30 of the women without HGSIL who were similarly HIV positive were tested for the presence of HIV DNA. Scrapings from all four women with HGSIL, but from just nine (30 percent) of the 30 women without HGSIL, were positive for HIV DNA, giving an odds ratio of infinity with a lower 95 percent confidence interval of 1.2. DISCUSSION Although the response rate of the husbands in this study was low, the results based on interview data are consistent with those from a larger study in Thailand that utilized similar methodology (10) and achieved a higher level of participation. Together, they provide strong evidence that the risk of cervical cancer in monogamous Thai women is enhanced if their husbands had unprotected sexual intercourse with prostitutes when the husbands were in their teens and twenties. The possible relation of HPV infection in the husbands to their having unprotected intercourse with prostitutes provides direct evidence for the role of husbands as vectors of HPV transmission from prostitutes, and observations by us and others (17) of a higher prevalence of these viruses in commercial sex workers than in other sexually active women confirm that prostitutes serve as reservoirs of infection. Since most women with cervical cancer in this study were monogamous and harbored an oncogenic type of HPV, the husbands likely transmitted the offending virus to their wives early in their marriage. The few cases with multiple partners could have contracted their HPV infection from another man and transmitted HPV to their husbands, so it is reasonable to conclude that virtually all of the husbands of cases had previously been infected by an oncogenic HPV. The low prevalence of oncogenic HPV DNA in the penile scrapings from the husbands of cases is consistent with the observation that men tend to clear acute HPV infections at higher rates than do women (27). Positive tests for HPV DNA in men, even if based on highly sensitive polymerase chain reaction-based technology, are poor indicators of prior infections and, as also noted by others (16), not reflective of the infection status of the man at the time relevant to his role in the genesis of cervical cancer in his wife. A low level of concordance between HPV in women's cervical samples and their husbands' penile samples was observed in this study and by others (18, 28–30). In Thailand, women with cervical cancer are clearly not important sources of demonstrable HPV infection in their husbands, probably because symptoms of cervical carcinoma tend to preclude sexual intercourse. The lower prevalence of HPV in men married to women with invasive carcinoma than in men married to women with in situ disease is supportive of this explanation. Consistent with other studies (27, 31–36), the prevalence of oncogenic HPV infection in brothel workers decreased with age, although the trend was less apparent in the women in the massage parlor. HPV infection was also associated with recent initial employment in the sex industry and a history of treatment for a sexually transmitted disease only within the past year, which are findings consistent with observations by others that risk in nonprostitutes is inversely related to years since initiating sexual activity (31–33) and directly to the number of partners in the past year (31, 32, 37) but not to the total number of partners. Women thus tend to develop an immune response to HPV infections with the passage of time since initial exposure. The absence of a decline in the prevalence of oncogenic HPV types with age in women with an HIV infection and the higher prevalence of HPV infections in women with than without hepatitis B antigenemia, but not in women with than without a demonstrable antibody response to hepatitis B and several other sexually transmitted agents, suggest that women who have a compromised immunologic system are at increased risk of persistent oncogenic HPV. A weak association of HPV type16-associated invasive cervical carcinomas with HBsAg positivity (22) is supportive of this hypothesis. Although the prevalence (38, 39) and persistence over time (40, 41) of HPV infection have been related to low levels of CD4 T lymphocytes, women in this study were generally healthy and actively working, suggesting that only subtle changes in immune status are necessary for persistence of oncogenic HPV types. Additional studies to characterize the nature of the immune status permissive to HPV persistence, especially in the absence of HIV infection, are warranted. Our observations that the prevalence of oncogenic HPV types was weakly associated with a relatively recent pregnancy and current use of oral contraceptives are consistent with previous reports that the risk of cervical neoplasia is associated with use of oral contraceptives (22, 23, 42), and in some studies with parity (43, 44), although not in others (45, 46), and suggest that the mechanism for these associations, if causal, is by enhancing the persistence of HPV infection. Prior studies that have not shown an association of HPV infection with parity (31–33, 47) or oral contraceptive use (27, 33, 47) did not report risk in relation to recency of pregnancy and oral contraceptive exposure, and the influence of these hormonal factors on risk could be of short duration. Two studies in which cases of in situ and invasive cervical carcinoma were directly compared (23, 34) provide evidence that neither of these hormonal factors operates to enhance progression of intraepithelial lesions to invasion. The long recognized, but incompletely understood, interaction between steroidal sex hormones and both cell-mediated and hormonal immunity (48) may provide the basis for a unifying explanation for these hormonal risk factors for HPV persistence and the relation of HPV infection to HIV and HbsAg, and it further suggests that studies of immunity in apparently healthy women may provide clues to the causes of HPV persistence and hence cervical carcinoma. The absence of an association between DMPA use and HPV infection is consistent with studies showing no association of DMPA use with invasive cervical carcinoma (49) and is not supportive of an increase in risk of carcinoma in situ in DMPA users previously observed (23, 50). The absence of an association of oncogenic HPV prevalence with smoking is consistent with most recent observations (27, 3132–33, 47), suggesting that if smoking enhances the risk of cervical carcinoma it does so by mechanisms other than immunosuppression and enhancement of HPV persistence. The absence of evidence for a protective effect of condoms against HPV infections in prostitutes is also consistent with previous reports (4, 32, 37, 47). Inconsistent use is unlikely to provide protection against cervical cancer in women who are heavily exposed to oncogenic HPV. This observation is not inconsistent with our findings that men who seldom or never use condoms when visiting prostitutes are more likely to harbor oncogenic HPV DNA in exfoliated penile cells than men who do use condoms and with the observation that the risk of cervical cancer in monogamous Thai women is enhanced if their husbands have unprotected sexual intercourse with prostitutes (10). Condoms regularly used by men may well provide them some protection against HPV infection. As expected, the risk of prevalent HGSIL was strongly associated with oncogenic types of HPV, especially type 16. The prevalence of HGSIL did not decline with age in our study or in Costa Rica (35), presumably because HGSIL develop in women who fail to clear their HPV infection. HPV-infected women were at greater risk of HGSIL if they were HIV positive than if they were HIV negative, but the risk of HGSIL was not associated with HIV status in women without oncogenic HPV DNA. These observations support conclusions from prospective studies (51, 52) that HIV enhances the risk of intraepithelial cervical neoplasia by suppressing the immune system and thus permitting oncogenic HPV infection to persist. A likely explanation for the association of cervical cancer with HIV in the absence of HPV in a recent meta-analysis (53) is the presence of undetected HPV in cases classified as HPV negative. An increase in the risk of HGSIL was observed in women with a positive serologic HIV test in relation to the presence of HIV DNA in exfoliated cervical cells. This is consistent with the observation that the risk of cervical intraepithelial neoplasia in HPV-infected women with HIV is associated with the degree of immunosuppression (41, 54), although more HIV-infected CD4 T cells could be shed from women with than without HGSIL because of greater bleeding or inflammation unrelated to immunosuppression. In order to minimize the possibility of selection bias in this study, visits were made to the brothel and massage parlor on multiple days of the week. Although no sampling scheme was utilized, few women who were asked to participate refused. The investigators in Thailand have conducted many previous studies of the commercial sex workers in Bangkok and have established a level of rapport and trust to give us confidence in the validity of the answers to questions asked. Although we cannot ensure that the women employed in the two establishments in which this study was conducted are representative of all commercial sex workers in Bangkok, the consistency of our results with those of others and the biologic reasonableness of the findings give us confidence in the reliability and generalizability of the results. Prostitutes in Bangkok are reservoirs of oncogenic HPV infection, both because of their high level of exposure to these sexually transmitted agents and because of their high level of HIV infection that facilitates persistence of HPV in the cervical mucosa. They consequently are at high risk of HGSIL. Dr. David B. Thomas, Program in Epidemiology, Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue North, Seattle, WA 98109 (e-mail: [email protected]). This study was supported by grant CA49044 from the US National Institutes of Health. The assistance of Drs. Anna Marie Beckmann and Larry Corey is gratefully acknowledged. REFERENCES 1. International Agency for Research on Cancer. Human papillomaviruses. IARC Monogr Eval Carcinog Risks Hum  1995; 64: 1–409. Google Scholar 2. Schneider A, Koutsky LA. Natural history and epidemiological features of genital HPV infection. In: Muñoz N, Bosch FX, eds. The epidemiology of human papillomavirus and cervical cancer. Lyon, France: International Agency for Research on Cancer, 1992. (IARC scientific publication no. 119). Google Scholar 3. Skegg DC, Corwin PA, Paul C, et al. Importance of the male factor in cancer of the cervix. Lancet  1982; 2: 581–3. Google Scholar 4. Wang PD, Lin RS. Risk factors for cervical intraepithelial neoplasia in Taiwan. Gynecol Oncol  1996; 62: 10–18. Google Scholar 5. Bosch FX, CastellsaguéX, Muñoz N, et al. Male sexual behavior and human papillomavirus DNA: key risk factors for cervical cancer in Spain. J Natl Cancer Inst  1996; 88: 1060–7. Google Scholar 6. Zunzunegui MV, King MC, Coria CF, et al. Male influences on cervical cancer risk. Am J Epidemiol  1986; 123: 302–7. Google Scholar 7. Donnan SP, Wong FW, Ho SC, et al. Reproductive and sexual risk factors and human papilloma virus infection in cervical cancer among Hong Kong Chinese. Int J Epidemiol  1989; 18: 32–6. Google Scholar 8. Zhang ZF, Parkin DW, Yu SZ, et al. Risk factors for cancer of the cervix in a rural Chinese population. Int J Cancer  1989; 43: 762–7. Google Scholar 9. Pridan H, Lilienfeld AM. Carcinoma of the cervix in Jewish women in Israel, 1960–67. Isr J Med Sci  1971; 7: 1465–70. Google Scholar 10. Thomas DB, Ray RM, Pardthaisong T, et al. Prostitution, condom use, and invasive squamous cell cervical cancer in Thailand. Am J Epidemiol  1996; 143: 779–86. Google Scholar 11. Buckley JD, Harris RW, Doll R, et al. Case-control study of the husbands of women with dysplasia or carcinoma of the cervix uteri. Lancet  1981; 2: 1010–15. Google Scholar 12. Kjaer SK, deVilliers EM, Dahl C, et al. Case-control study of risk factors for cervical neoplasia in Denmark. I. Role of the “male factor” in women with one lifetime sexual partner. Int J Cancer  1991; 48: 39–44. Google Scholar 13. Agarwal SS, Sehgal A, Sardana S, et al. Role of male behavior in cervical carcinogenesis among women with one lifetime sexual partner. Cancer  1993; 72: 1666–9. Google Scholar 14. Muñoz N, Bosch FX, de SanjoséS, et al. Risk factors for cervical intraepithelial neoplasia grade III/carcinoma in situ in Spain and Colombia. Cancer Epidemiol Biomarkers Prev  1993; 2: 423–31. Google Scholar 15. Brinton LA, Reeves WC, Brenes MM, et al. The male factor in the etiology of cervical Cancer among sexually monogamous women. Int J Cancer  1989; 44: 199–203. Google Scholar 16. Muñoz N, CastellsaguéX, Bosch FX, et al. Difficulty in elucidating the male role in cervical Cancer in Colombia, a high-risk area for the disease. J Natl Cancer Ins  1996; 88: 1068–75. Google Scholar 17. Bosch FX, Muñoz N, de SanjoséS, et al. Importance of human papillomavirus endemicity in the incidence of cervical Cancer: an extension of the hypothesis on sexual behavior. Cancer Epidemiol Biomarkers Prev  1994; 3: 375–9. Google Scholar 18. CastellsaguéX, Ghaffari A, Daniel RW, et al. Prevalence of penile human papillomavirus DNA in husbands of women with and without cervical neoplasia: a study in Spain and Colombia. J Infect Dis  1997; 176: 353–61. Google Scholar 19. Hippeläinen M, Syrjänen S, Hippeläinen M, et al. Prevalence and risk factors of genital human papillomavirus (HPV) infections in healthy males: a study on Finnish conscripts. Sex Transm Dis  1993; 20: 321–8. Google Scholar 20. Hulka BS. Risk factors for cervical Cancer. J Chronic Di  1982; 35: 3–11. Google Scholar 21. de SanjoséS, Palacio V, Tafur L, et al. Prostitution, HIV, and cervical neoplasia: a survey in Spain and Colombia. Cancer Epidemiol Biomarkers Prev  1993; 2: 531–5. Google Scholar 22. Thomas DB, Ray RM, Koetsawang A, et al. Human papillomaviruses and cervical Cancer in Bangkok. I. Risk factors for invasive cervical carcinomas with human papillomavirus types 16 and 18 DNA. Am J Epidemiol  2001; 153: 723–31. Google Scholar 23. Thomas DB, Qin Q, Kuypers J, et al. Human papillomaviruses and cervical Cancer in Bangkok. II. Risk factors for in situ and invasive squamous cell cervical carcinomas. Am J Epidemiol  2001; 153: 732–9. Google Scholar 24. Moss GB, Overbaugh J, Welch M, et al. Human immunodeficiency virus DNA in urethral secretions in men: association with gonococcal urethritis and CD4 cell depletion. J Infect Dis  1995; 172: 1469–74. Google Scholar 25. Breslow NE, Day NE, eds. Statistical methods in Cancer research. Vol I. The analysis of case-control studies. Lyon, France: International Agency for Research on Cancer, 1980:325–38. (IARC scientific publication no. 32). Google Scholar 26. Tritchler D. An algorithm for exact logistic regression. J Am Stat Assoc  1984; 79: 709–74. Google Scholar 27. Van Doornum GJ, Prins M, Juffermans LH, et al. Regional distribution and incidence of human papillomavirus infections among heterosexual men and women with multiple sexual partners: a prospective study. Genitourinary Med  1994; 70: 240–6. Google Scholar 28. Baken LA, Koutsky LA, Kuypers J, et al. Genital human papillomavirus infection among male and female sex partners: prevalence and type-specific concordance. J Infect Dis  1995; 171: 429–32. Google Scholar 29. Strand A, Rylander E, Wilander E, et al. HPV infection in male partners of women with squamous intraepithelial neoplasia and/or high-risk HPV. Acta Derm Venereol  1995; 75: 312–16. Google Scholar 30. Kyo S, Inoue M, Koyama M, et al. Detection of high-risk human papillomavirus in the cervix and semen of sex partners. J Infect Dis  1994; 170: 682–5. Google Scholar 31. Svare EI, Kjaer SK, Worm AM, et al. Risk factors for HPV infection in women from sexually transmitted disease clinics: comparison between two areas with different cervical Cancer incidence. Int J Cancer  1998; 75: 1–8. Google Scholar 32. Kjaer SK, van den Brule AJ, Bock JE, et al. Determinants for genital human papillomavirus (HPV) infection in 1000 randomly chosen young Danish women with normal cervical smear: are there different risk profiles for oncogenic and nononcogenic HPV types? Cancer Epidemiol Biomarkers Prev  1997; 6: 799–805. Google Scholar 33. Svare EI, Kjaer SK, Smits HL, et al. Risk factors for HPV detection in archival Pap smears. A population-based study from Greenland and Denmark. Eur J Cancer  1998; 34: 1230–4. Google Scholar 34. Moreno V, Muñoz N, Bosch FX, et al. Risk factors for progression of cervical intraepithelial neoplasms grade III to invasive cervical Cancer. Cancer Epidemiol Biomarkers Prev  1995; 4: 459–67. Google Scholar 35. Herrero R, Hildesheim A, Bratti C, et al. Population-based study of human papillomavirus infection and cervical neoplasia in rural Costa Rica. J Natl Cancer Ins  2000; 92: 464–74. Google Scholar 36. Ho GY, Bierman R, Beardsley L, et al. Natural history of cervicovaginal papillomavirus infection in young women. N Engl J Med  1998; 338: 423–8. Google Scholar 37. Young TK, McNicol P, Beauvais J. Factors associated with human papillomavirus infection detected by polymerase chain reaction among urban Canadian aboriginal and non-aboriginal women. Sex Transm Dis  1997; 24: 293–8. Google Scholar 38. Miotti PG, Dallabetta GA, Daniel RW, et al. Cervical abnormalities, human papillomavirus, and human immunodeficiency virus infections in women in Malawi. J Infect Dis  1996; 173: 714–17. Google Scholar 39. Ho GY, Burk RD, Fleming I, et al. Risk of genital human papillomavirus infection in women with human immunodeficiency virus-induced immunosuppression. Int J Cancer  1994; 56: 788–92. Google Scholar 40. Sun XW, Kuhn L, Ellerbrock TV, et al. Human papillomavirus infection in women infected with the human immunodeficiency virus. N Engl J Med  1997; 337: 1343–9. Google Scholar 41. Ahdieh L, Muñoz A, Vlahov D, et al. Cervical neoplasia and repeated positivity of human papillomavirus infection in human immunodeficiency virus-seropositive and -seronegative women. Am J Epidemiol  2000; 151: 1148–57. Google Scholar 42. International Agency for Research on Cancer. Hormonal contraception and post-menopausal hormonal therapy. IARC Monogr Eval Carcinog Risks Hum  1999; 72: 1–660. Google Scholar 43. Chaouk N, Bosch FX, Muñoz N, et al. The viral origin of cervical Cancer in Rabat, Morocco. Int J Cancer  1998; 75: 546–54. Google Scholar 44. Ngelangel C, Muñoz N, Bosch FX, et al. Causes of cervical Cancer in the Philippines: a case-control study. J Natl Cancer Ins  1998; 90: 43–9. Google Scholar 45. Chichareon S, Herrero R, Muñoz N, et al. Risk factors for cervical Cancer in Thailand: a case-control study. J Natl Cancer Ins  1998; 90: 50–7. Google Scholar 46. Bosch FX, Muñoz N, de SanjoséS, et al. Risk factors for cervical Cancer in Colombia and Spain. Int J Cancer  1992; 52: 750–8. Google Scholar 47. Giuliano AR, Papenfuss M, Schneider A, et al. Risk factors for high-risk type human papillomavirus infection among Mexican-American women. Cancer Epidemiol Biomarkers Prev  1999; 8: 615–20. Google Scholar 48. Grossman CJ. Interactions between the gonadal steroids and the immune system. Science  1985; 227: 257–61. Google Scholar 49. The WHO Collaborative Study of Neoplasia and Steroid Contraceptives. Depot-medroxyprogesterone acetate (DMPA) and risk of invasive squamous cell cervical Cancer. Contraception  1992; 45: 299–312. Google Scholar 50. Thomas DB, Ye Z, Ray RM, et al. Cervical carcinoma in situ and use of depot-medroxyprogesterone acetate (DMPA). Contraception  1995; 51: 25–31. Google Scholar 51. Ylitalo N, Sørensen P, Josefsson AM, et al. Consistent high viral load of human papillomavirus 16 and risk of cervical carcinoma in situ: a nested case-control study. Lancet  2000; 355: 2194–8. Google Scholar 52. Frisch M, Biggar RJ, Goedert JJ, et al. Human papillomavirus-associated Cancers in patients with human immunodeficiency virus infection and acquired immunodeficiency syndrome. J Natl Cancer Ins  2000; 92: 1500–10. Google Scholar 53. Mandelblatt JS, Kanetsky P, Eggert L, et al. Is HIV infection a cofactor for cervical squamous cell neoplasia? Cancer Epidemiol Biomarkers Prev  1999; 8: 97–106. Google Scholar 54. Rezza G, Giuliani M, Branca M, et al. Determinants of squamous intraepithelial lesions (SIL) on Pap smear: the role of HPV infection and of HIV-I-induced immunosuppression. Eur J Epidemiol  1997; 13: 937–43. Google Scholar

Thomas, David B.; Ray, Roberta M.; Kuypers, Jane; Kiviat, Nancy; Koetsawang, Amorn; Ashley, Rhoda L.; Qin, Qin; Koetsawang, Suporn

doi: N/Apmid: N/A

Between September 1991 and September 1993, husbands of women with and without cervical neoplasia and commercial sex workers in one brothel and one massage parlor in Bangkok, Thailand, were interviewed; serologic tests for sexually transmitted infections were performed; and cervical and penile scrapings were tested for human papillomavirus (HPV) DNA. The risks of cervical carcinoma in monogamous women and of oncogenic HPV in their husbands were associated with the men's having unprotected intercourse with prostitutes. The prevalence of oncogenic HPV was higher in commercial sex workers than in women attending gynecologic and family planning clinics. Oncogenic HPV prevalence declined with age in human immunodeficiency virus (HIV)-negative, but not in healthy HIV-positive, commercial sex workers and was weakly associated with hepatitis B antigenemia, suggesting that persistence of HPV infection is due to subtle changes in immunity. Associations of HPV with recent pregnancy and oral contraceptive use suggest that hormonal factors may increase the risk of cervical neoplasia by enhancing persistence of HPV infection. The prevalence of high-grade squamous intraepithelial lesions was strongly related to oncogenic HPV types and weakly to HIV infection only in their presence. Commercial sex workers in Bangkok are reservoirs of oncogenic HPV, and cervical cancer in monogamous Thai women develops in part as a result of transmission of these viruses to them by their husbands from prostitutes.

Yeh, Hsin-chieh;Matanoski, Genevieve M.;Wang, Nae-yuh;Sandler, Dale P.;Comstock, George W.

doi: 10.1093/aje/153.8.749pmid: 11296146

Abstract A population from a hearing clinic in Washington County, Maryland, in 1943–1960 was followed to assess the risk of developing neoplasms from radium treatment of the nasopharynx for adenoid hypertrophy. Of the 2,925 subjects who attended the clinic, 904 received radium treatment. A nonconcurrent prospective study compared the cancer incidence among the irradiated persons with that among persons with other treatments. Seven brain tumor cases (three malignant and four benign) were identified in the irradiated group versus none in the nonirradiated group (relative risk = 14.8, 95% confidence interval: 0.76, 286.3). A nonsignificant excess risk of thyroid cancer was detected in the irradiated group based on two cases in the exposed group and one case in the nonexposed group (relative risk = 4.2, 95% confidence interval: 0.38, 46.6). Decreased risks of breast cancer, female genital cancers, and prostate cancer were observed among the irradiated individuals, although these deficits were not statistically significant individually. The decreased risk of sex hormone-related cancers in the irradiated group suggests possible radiation damage to the pituitary, with consequent reduction in pituitary hormone output and alterations in sexual and other hormonal development in early life. This hypothesis needs further evaluation. nasopharynx, neoplasms, radiation, radium CI, confidence interval, RR, relative risk Acute upper respiratory infection often causes hypertrophy of nasopharyngeal lymphoid tissue. The hypertrophied tissue can result in obstruction of the eustachian tube and may induce hearing loss. Before the use of antibiotics, tonsillectomy and adenoidectomy were common treatments. However, incomplete surgical removal often allowed subsequent hypertrophy of the residual tissue. In 1924, Crowe (1) pioneered the use of radiation to treat hyperplasia of lymphoid elements of the nasopharynx. Radium treatment of the nasopharynx was widely used in Europe and the United States from the 1940s to the mid-1960s. The Centers for Disease Control and Prevention estimates that approximately 500,000 to two million individuals, mostly children, received radium treatment in the United States between 1946 and 1961 (2). Based on short-term observations, several studies have reported that radiation treatment was effective in reducing adenoid size and possibly preventing deafness and other ear, nose, and throat symptoms (3–5). Three cohort studies assessed the long-term risk of developing cancer after radiation. Hazen et al. (6) conducted a questionnaire follow-up study of exposed subjects and their siblings as a comparison. A study in the Netherlands assessed the mortality of exposed patients compared with matched, nonexposed controls from the same clinic (7). These two studies did not observe any significantly increased cancer risk among the irradiated subjects. A third study conducted in 1978 in Washington County, Maryland, found a slight excess risk of tumors of the head and neck among irradiated individuals compared with nonirradiated children treated in the same clinic (8, 9). The radiation doses for populations treated in the study by Hazen et al. (6) and the study in the Netherlands (7) were approximately 14–28 percent of that in the Washington County population. All three studies were conducted when irradiated subjects had not reached the high-risk age for cancer. Our study continues the evaluation of the individuals treated in Washington County to provide an additional 17-year follow-up period with observations of study participants aged 39–92 years (median, 46 years). MATERIALS AND METHODS Study population Study subjects were identified from the records of the Clinic for Prevention of Deafness in Children in Washington County, Maryland, a semirural area in western Maryland where population mobility is low. The clinic was established in 1943 by the health department in collaboration with state, federal, and voluntary agencies to serve as both a primary health care resource and a referral service for the diagnosis and treatment of hearing defects (10). The clinic also operated in conjunction with a screening program in both public and private county schools to identify potential patients. Irradiation therapy was given at the clinic if elimination of nasopharyngeal lymphoid tissues was desirable and tonsillectomy and adenoidectomy were not indicated. If there was a large mass of adenoid tissue, surgical removal combined with irradiation was recommended (11). All persons first seen at the clinic from January 1, 1940 to January 1, 1960, were included as the initial study population (n = 2,925). The exposed group included patients who had nasopharyngeal radium treatment (n = 904); the nonexposed group included individuals who had been given other treatments (n = 2,021), mainly tonsillectomy and adenoidectomy, to relieve their symptoms. Table 1 lists the number of subjects included in each step of the current and previous studies by exposure status. Analysis of cancer incidence in this study was restricted to persons who were successfully located in the initial 1978 study (n = 2,627). Among these individuals, 83 had died by the time of the earlier follow-up, and 2,135 responded to a questionnaire survey at that time. These previous respondents were traced and contacted again in 1994–1995 to obtain updated health information. Those who refused or did not respond in the earlier study (n = 409) were traced only by searching death certificates and the cancer registry files. No direct personal contact with this group of subjects was attempted. Persons who were not traced in the earlier follow-up (n = 298) were excluded because there was no new information to allow further contact (12). TABLE 1. Number of subjects included in each step of the study by exposure status, Washington County, Maryland, 1940–1995 Major steps  Exposed  Non-exposed  Total  Original population  904  2,021  2,925  Lost to follow-up in 1978  96  202  298  Located in 1978 study  808  1,819  2,627   Dead  35  48  83   Alive  773  1,771  2,544    No reply to 1978 questionnaire  106  303  409    Responded to 1978 questionnaire  667  1,468  2,135     Located in 1994–1995  618  1,383  2,001      Dead  35  87  122      Alive  583  1,296  1,879       Responded to 1994–1995 questionnaire  545  1,158  1,703  Major steps  Exposed  Non-exposed  Total  Original population  904  2,021  2,925  Lost to follow-up in 1978  96  202  298  Located in 1978 study  808  1,819  2,627   Dead  35  48  83   Alive  773  1,771  2,544    No reply to 1978 questionnaire  106  303  409    Responded to 1978 questionnaire  667  1,468  2,135     Located in 1994–1995  618  1,383  2,001      Dead  35  87  122      Alive  583  1,296  1,879       Responded to 1994–1995 questionnaire  545  1,158  1,703  View Large Follow-up in 1994–1995 Incident cancer cases were identified from a questionnaire survey, the Washington County cancer registry, and death certificates. During 1994–1995, a questionnaire was mailed to subjects who were not known to have died and whose addresses could be located. Resources utilized for tracing subjects included telephone directories, city directories, and Washington County private censuses. Located subjects received up to two mailings followed by a telephone reminder, if needed. The questionnaire included queries about diseases such as malignant and benign tumors, reproductive history for women, and other factors that could be potential confounding variables for cancer. Each subject was asked to sign a permission form allowing the retrieval of hospital or doctor's records for verification of disease history. The Washington County cancer registry was used for both case identification and verification. The registry includes resident and locally diagnosed nonresident cancer cases from 1948 to the present. It identifies cases through review of records at the only hospital in the county as well as review of death certificates for county residents. Vital status was obtained by searching Washington County mortality files as well as matching with Social Security death records. If subjects were not found in either source, they were assumed to be alive as of December 31, 1994. Death certificates were obtained for decedents to determine the causes of death. Specific causes of death were coded according to the International Classification of Diseases, Ninth Revision. If the date of the cancer diagnosis was not available from other sources, the date of death from cancer was used as the incidence date. All cancer cases reported on the questionnaires were checked without knowledge of exposure status. Diagnoses were verified by matching with the cancer registry or by requesting medical records through hospitals and physicians. More than 92 percent of the cancers reported on the questionnaire were verified. Radium treatment The treatment provided by the clinic used a radium applicator that consisted of an 8-inch (203.2-mm) flexible rod with 50 mg of radium in its tip and a 0.3-mm monel metal (a nickel alloy) filter (8). This allowed emissions consisting of approximately 30 percent beta particles and 70 percent gamma rays (13). Treatment consisted of two applicators, one in each nostril, which were inserted along the floor of the nasal cavity beneath the inferior turbinate bone to eventually lie just medial to the pharyngeal ostium of the eustachian tube. The applicators were kept in place for approximately 12 minutes; three such doses, administered at 14-day intervals, constituted the standard course of therapy. The total exposure time was 60 mg-hour. Additional radium treatments were occasionally given if symptoms persisted or reappeared. Among the irradiated patients, 11 percent received dosages of less than 60 mg-hours, 74 percent received the standard course of treatment, and the remaining 15 percent received dosages of more than 60 mg-hours. Approximate distances from the tips of the radium applicators to different sites varied by age and individual. The distances from various tissue sites were estimated in the 1978 study by using skull films of children at different ages (8). Distances measured in this way were used to estimate the exposures for thyroid, pituitary, and salivary glands, the sites of major interest for the risk of cancer. The doses to these sites were then estimated by taking into account the varying energies and absorption fractions of the different products of radium decay and known properties of radium application (14). Table 2 shows the absorbed doses at different ages and distances. On the basis of these estimations, the pituitary gland and lower portion of the brain received approximately 0.78 Gy from the standard three treatment series. In most instances, the thyroid gland was 9 cm from the applicators and received exposures of approximately 0.09 Gy. TABLE 2. Estimated radiation doses to different anatomic sites from a standard course of therapy (three treatments) calculated by absorbed fraction method, Washington County, Maryland, 1940–1995 Sites and age at exposure* (years)  Distances (cm)  Estimated dose (grays)  Thyroid       ≤3  4–5  0.26, 0.44   4–7  5–8  0.09, 0.26   8–10  8–10  0.07, 0.09   11–12  10–12  0.04, 0.07   ≥13  ≥12  ≤0.04  Pituitary       ≤9  2–3  0.78, 1.70   ≥10  3–4  0.44, 0.78  Salivary       ≤6  5–6     7–9  6–7  0.09, 0.26   ≥10  7–8    Sites and age at exposure* (years)  Distances (cm)  Estimated dose (grays)  Thyroid       ≤3  4–5  0.26, 0.44   4–7  5–8  0.09, 0.26   8–10  8–10  0.07, 0.09   11–12  10–12  0.04, 0.07   ≥13  ≥12  ≤0.04  Pituitary       ≤9  2–3  0.78, 1.70   ≥10  3–4  0.44, 0.78  Salivary       ≤6  5–6     7–9  6–7  0.09, 0.26   ≥10  7–8    * Differences in age grouping were due to different distances from radium applicator to specific sites by age. View Large Data analysis After examination of the distribution of all variables, the chi-square test for differences in proportions was used to compare the frequency distributions of selected variables in study subgroups. The beginning of follow-up was defined as the first date of radium treatment for the exposed group and the first date of a clinical visit for the nonexposed group. To calculate cancer incidence, we determined the end of follow-up date by the date of diagnosis, the date of death obtained from the death certificate (if there was no prior record of a cancer diagnosis), or the last date known to be free of cancer. In analyzing the incidence of endometrial and ovarian cancers, women with a known history of hysterectomy and/or oophorectomy were removed from person-year contributions at the time of surgery, based on the assumption that they were no longer at risk. Skin cancer was excluded from the analysis because of difficulty in confirming cases from the registry. Internal comparisons were performed by Poisson regression to calculate the relative risks of tumor incidence, adjusting for possible confounding factors. Survival analysis was further carried out to gain insight on cancer incidence over the course of follow-up. Associations between exposure status and the cumulative incidence of tumors during the follow-up period were first evaluated by the Kaplan-Meier method. The significance levels of the difference between the two groups were assessed with the log-rank test. The Cox proportional hazards model was used for the multivariate analysis when the data met the proportionality assumption implicit in the model. All tests of significance were two-tailed, with an alpha level of 0.05. Since breast cancer, ovarian cancer, and endometrial cancer may be considered as hormone-related cancers and share similar characteristics, these were also combined as the outcome in an additional set of regression analyses. Prostate cancer was added to this combined outcome category of genital cancers in both sexes. RESULTS In the current follow-up, 2,001 of the total 2,135 individuals included in active tracing were successfully located. Tracing was similarly successful for exposed and nonexposed study subjects, with 92.7 and 94.2 percent of each group, respectively, being traced (p = 0.16). There was no difference for demographic variables between the two groups. Questionnaires were mailed to 1,882 persons who were found to be still alive, and 1,703 replied to the survey. This represents a 93.5 percent response among the exposed and 89.1 percent among the nonexposed persons. Seventy-six subjects did not return the questionnaires even after several contacts, and 103 individuals refused to participate in the study. A comparison of questionnaire respondents and nonrespondents indicated no differences in the percent exposed in the two groups (32 percent of respondents vs. 31 percent of nonrespondents received radium treatment). The respondents were, however, older than nonrespondents by an average of 2 years. Baseline characteristics Table 3 lists the frequency distribution of baseline characteristics in the exposed and nonexposed subjects who were eligible for cancer incidence analysis regardless of questionnaire response status. The two groups did not differ significantly by sex, race, or age at the beginning of follow-up, but they differed by year of birth and year at start of follow-up. The length of follow-up was longer in the exposed group by an average of 2 years (p < 0.01). Exposed and nonexposed individuals did not differ significantly by family history of deafness, yet more irradiated subjects had a history of adenoidectomy with or without tonsillectomy before the start of treatment at the clinic. Mild and severe hearing loss was also more common in the irradiated group at the first clinic visit. TABLE 3. Percent distribution of baseline characteristics of exposed and nonexposed subjects, Washington County, Maryland, 1940–1995 Characteristics  Exposed (%) (n = 808)  Non-exposed (%) (n = 1,819)  p value  Year of birth         <1935  6.6  5.9  0.00   1935–1939  27.2  17.6     1940–1944  39.0  36.1     ≥1945  25.0  40.4    Year started follow-up         1940–1944  8.2  1.5  0.00   1945–1949  29.7  14.9     1950–1954  52.7  54.4     1955–1960  9.4  29.3    Age at start of follow-up (years)         ≤6  26.2  26.6  0.52   7–9  37.9  36.1     10–12  19.7  21.8     ≥13  14.3  15.5    Sex         Male  56.1  52.8  0.12   Female  43.9  47.2    Race         White  98.1  97.3  0.20   Other  1.9  2.7    Family history of deafness         No  78.8  81.1  0.14   Yes  12.6  12.5     Unknown  8.5  6.4    History of adenoidectomy         No  92.3  98.8  0.00   Yes  7.7  1.2    History of T&A*         No  19.7  50.5  0.00   Yes, before start of follow-up  76.4  14.0     Yes, after start of follow-up  3.9  35.5    Hearing loss at first visit         None  50.2  72.4  0.00   Mild  34.8  14.0     Severe  5.6  2.9     Not recorded  9.4  10.8    Characteristics  Exposed (%) (n = 808)  Non-exposed (%) (n = 1,819)  p value  Year of birth         <1935  6.6  5.9  0.00   1935–1939  27.2  17.6     1940–1944  39.0  36.1     ≥1945  25.0  40.4    Year started follow-up         1940–1944  8.2  1.5  0.00   1945–1949  29.7  14.9     1950–1954  52.7  54.4     1955–1960  9.4  29.3    Age at start of follow-up (years)         ≤6  26.2  26.6  0.52   7–9  37.9  36.1     10–12  19.7  21.8     ≥13  14.3  15.5    Sex         Male  56.1  52.8  0.12   Female  43.9  47.2    Race         White  98.1  97.3  0.20   Other  1.9  2.7    Family history of deafness         No  78.8  81.1  0.14   Yes  12.6  12.5     Unknown  8.5  6.4    History of adenoidectomy         No  92.3  98.8  0.00   Yes  7.7  1.2    History of T&A*         No  19.7  50.5  0.00   Yes, before start of follow-up  76.4  14.0     Yes, after start of follow-up  3.9  35.5    Hearing loss at first visit         None  50.2  72.4  0.00   Mild  34.8  14.0     Severe  5.6  2.9     Not recorded  9.4  10.8    * T&A, tonsillectomy and adenoidectomy. View Large Cancer incidence Relative risks of cancer incidence associated with exposure are presented in table 4 for the total period of observation. A total of 41 cancer cases were detected among the 808 exposed patients. In the nonexposed group, 83 cancer cases were identified in a population of 1,819 persons. Two cases of thyroid cancer were detected in the exposed group, and only one case of thyroid cancer was detected in the nonexposed group. No malignant salivary gland tumors developed in this population. Three malignant and four benign brain tumors were identified in the exposed group versus none in the nonexposed group. The relative risk of malignant brain tumor is 14.8 (95 percent CI: 0.76, 286.3), and for the combined category of benign and malignant tumors, it is 30.9 (95 percent CI: 1.87, 541.7). Two of the malignant cases were astrocytomas, but sites in the brain were unknown. Among the benign cases, one was identified as a hemangioblastoma in the right cerebellar lobe, and one was a neurilemmoma in the posterior fossa of the cranial cavity. The medical records were not available to verify the other two tumors. The three malignant cases occurred within 12, 18, and 25 years after radiation. The four benign cases were diagnosed 35–44 years after radiation. While the exposed group had a higher risk of cancers of the head and neck, the rates for cancers of breast, endometrium, ovary, and prostate in those who were exposed were lower than for the nonexposed (table 4). The risk of developing this combined group of sex hormone-related cancers was 56 percent lower in the irradiated group (relative risk = 0.44, 95 percent CI: 0.20, 0.96). TABLE 4. Relative risks of cancer incidence and 95 percent confidence intervals of selected cancers among exposed compared with nonexposed subjects using Poisson regression, Washington County, Maryland, 1940–1995 Cancer  No. of cases in exposed group (person-years = 31,005)  No. of cases in nonexposed group (person-years = 65,502)  Adjusted RR*,†  95% CI*  All neoplasms except skin  41  83  1.02  0.70, 1.50   Oral cavity and pharynx  2  1  4.22  0.38, 46.59   Digestive system  9  10  1.65  0.65, 4.17   Respiratory system  9  15  1.30  0.56, 3.01    Lung  9  12  1.60  0.66, 3.86   Breast (female)‡  6  19  0.62  0.24, 1.59   Endometrium and ovary‡,§  1  8  0.28  0.04, 2.28   Breast, endometrium, and ovary‡,§  7  27  0.51  0.21, 1.19   Prostate¶  0 (0.5)#  4 (4.5)#  (0.22)#  (0.01, 3.13)#   Breast, endometrium, ovary, and prostate§  7  3  0.39  0.17, 0.91   Brain  3 (3.5)#  0 (0.5)#  (14.79)#  (0.76, 286.3)#   Thyroid  2  1  4.22  0.38, 46.59   All lymphopoietic  4  12  0.65  0.21, 2.09  Cancer  No. of cases in exposed group (person-years = 31,005)  No. of cases in nonexposed group (person-years = 65,502)  Adjusted RR*,†  95% CI*  All neoplasms except skin  41  83  1.02  0.70, 1.50   Oral cavity and pharynx  2  1  4.22  0.38, 46.59   Digestive system  9  10  1.65  0.65, 4.17   Respiratory system  9  15  1.30  0.56, 3.01    Lung  9  12  1.60  0.66, 3.86   Breast (female)‡  6  19  0.62  0.24, 1.59   Endometrium and ovary‡,§  1  8  0.28  0.04, 2.28   Breast, endometrium, and ovary‡,§  7  27  0.51  0.21, 1.19   Prostate¶  0 (0.5)#  4 (4.5)#  (0.22)#  (0.01, 3.13)#   Breast, endometrium, ovary, and prostate§  7  3  0.39  0.17, 0.91   Brain  3 (3.5)#  0 (0.5)#  (14.79)#  (0.76, 286.3)#   Thyroid  2  1  4.22  0.38, 46.59   All lymphopoietic  4  12  0.65  0.21, 2.09  * RR, relative risk; CI, confidence interval. † Relative risks were adjusted for year of birth and year follow-up was begun. The following categories were not adjusted because of small numbers of cases: oral cavity and pharynx, all female genital, prostate, brain, and thyroid. ‡ Female only. Person-years = 13,791.2 in the exposed women and 31,377.8 in the nonexposed women. § Subjects were censored at the time of hysterectomy and/or oophorectomy. Person-years = 12,338.8 in the exposed females and 27,402.8 in the nonexposed females. ¶ Male only. Person-years = 17,213.8 in the exposed men and 34,124.3 in the nonexposed men. # To obtain an estimated relative risk, a correction factor of 0.5 cases was added to each group (Haldane (28)). View Large To examine further the difference in cancer incidence over the course of follow-up, we performed survival analysis for selected cancers with five or more cases in both groups. The results of this additional analysis were similar to those estimated by Poisson regression. For the combined category of breast, endometrium, ovary, and prostate cancers, the estimated Cox relative risk was 0.29 (95 percent CI: 0.13, 0.65). Figure 1 shows the cumulative incidence of this combined category for both exposed and nonexposed groups estimated with the Kaplan-Meier product limit method. FIGURE 1. View largeDownload slide Cumulative incidence of breast, endometrium, ovary, or prostate cancer among exposed and nonexposed patients, Washington County, Maryland, 1940–1995. FIGURE 1. View largeDownload slide Cumulative incidence of breast, endometrium, ovary, or prostate cancer among exposed and nonexposed patients, Washington County, Maryland, 1940–1995. Reproductive characteristics The exposed and nonexposed groups were compared with respect to other characteristics identified from the questionnaire to determine whether any of these factors might explain the observed differences in cancers of reproductive organs. As shown in table 5, the exposed women were older, had completed more years in school, and were less often current cigarette smokers than were nonexposed women. The differences in age at menopause and at first full-term pregnancy were suggestive but nonsignificant, with the exposed women being older. The two groups were not significantly different in family history of breast cancer, percentage of subjects who had ever been pregnant, age at menarche, history of taking oral contraceptives, history of receiving hormone replacement therapy, infertility, and number of children. In men, those who were exposed were older, but there was no difference in number of children, infertility, and family history of prostate cancer between the two groups (data not shown.) The similarity in the reproductive characteristics of the two groups makes it unlikely that those factors played a role in the difference in cancer risk. TABLE 5. Percent distribution of characteristics of female respondents by exposure status, Washington County, Maryland, 1940–1995 Characteristics  Exposed % (n = 235)  Non-exposed % (n = 579)  p value  Year of birth         <1935  8.4  6.3  0.00   1935–1939  30.3  18.0     1940–1944  35.9  37.5     ≥1945  25.5  38.2    Education (grades)         ≤8  4.0  3.4  0.00   9–12  66.2  81.5     >12  27.9  13.0     Unknown  1.8  1.2    Smoking         Never  40.8  38.1  0.014   Ex-smokers  37.7  33.8     Current smokers  21.6  28.2    Had blood relatives with breast cancer         No  92.5  92.8  0.89   Yes  7.5  7.2    Ever pregnant         Ever  89.6  88.9  0.73   Never  10.4  11.1    Ever seen doctors for difficulty in having a child         Ever  4.1  4.7  0.71   Never  95.9  95.3    No. of children*         None  11.4  11.8  0.65   ≤3  63.4  65.9     >3  25.2  22.3    Age at 1st pregnancy (years)*         <20  44.0  51.7  0.11   20–29  52.8  44.4     ≥30  2.2  3.9    Age at menarche (years)         <12  21.0  18.9  0.63   12–14  66.1  66.1     >14  12.9  15.0    Still menstruate         Yes  23.9  29.9  0.09   No  74.9  68.0     Unknown  1.2  2.1    Age at menopause (years)†         Natural          <45  12.8  16.8  0.07    45–50  56.8  58.0      >50  31.4  25.2    Surgical          <35  27.0  39.6  0.09    35–40  27.0  25.2      >40  46.1  35.2    Ever taken oral contraceptives         Yes  54.4  54.0  0.99   No  43.2  43.7     Unknown  2.4  2.3    Ever received hormone replacement therapy         Yes  37.5  33.3  0.47   No  59.8  64.3     Unknown  2.8  2.5    Characteristics  Exposed % (n = 235)  Non-exposed % (n = 579)  p value  Year of birth         <1935  8.4  6.3  0.00   1935–1939  30.3  18.0     1940–1944  35.9  37.5     ≥1945  25.5  38.2    Education (grades)         ≤8  4.0  3.4  0.00   9–12  66.2  81.5     >12  27.9  13.0     Unknown  1.8  1.2    Smoking         Never  40.8  38.1  0.014   Ex-smokers  37.7  33.8     Current smokers  21.6  28.2    Had blood relatives with breast cancer         No  92.5  92.8  0.89   Yes  7.5  7.2    Ever pregnant         Ever  89.6  88.9  0.73   Never  10.4  11.1    Ever seen doctors for difficulty in having a child         Ever  4.1  4.7  0.71   Never  95.9  95.3    No. of children*         None  11.4  11.8  0.65   ≤3  63.4  65.9     >3  25.2  22.3    Age at 1st pregnancy (years)*         <20  44.0  51.7  0.11   20–29  52.8  44.4     ≥30  2.2  3.9    Age at menarche (years)         <12  21.0  18.9  0.63   12–14  66.1  66.1     >14  12.9  15.0    Still menstruate         Yes  23.9  29.9  0.09   No  74.9  68.0     Unknown  1.2  2.1    Age at menopause (years)†         Natural          <45  12.8  16.8  0.07    45–50  56.8  58.0      >50  31.4  25.2    Surgical          <35  27.0  39.6  0.09    35–40  27.0  25.2      >40  46.1  35.2    Ever taken oral contraceptives         Yes  54.4  54.0  0.99   No  43.2  43.7     Unknown  2.4  2.3    Ever received hormone replacement therapy         Yes  37.5  33.3  0.47   No  59.8  64.3     Unknown  2.8  2.5    * Includes ever-pregnant women only. † Includes postmenopausal women only. View Large DISCUSSION Although there was an excess risk of brain tumors associated with radiation, a dose-response effect could not be demonstrated because of the small number of cases. However, previous studies have shown a possible etiologic role of low-dose radiation in the development of brain tumors. Children who received x-ray treatment for tinea capitis (15, 16) and atomic bomb survivors were found to have elevated risks of malignant and benign brain tumors from radiation doses as low as 0.27 Gy (17). The finding of this study regarding risk of brain tumors is consistent with evidence from previous studies. The average dose to pituitary fossa was 0.78 Gy in the study population. The brain doses, as estimated by Land (18) were between 0.15 to 0.4 Gy, a value in the same range as the lower end of exposures in other reports. Radiogenic brain tumor types have often been identified as meningiomas but glioma, astrocytoma, hemangioblastoma, and acoustic neuroma have also been reported (15, 16, 19). The histologic types observed in this study were not different from those in other studies. One interesting observation was that all of the malignant cases occurred within the first 25 years after radiation (age range, 25–39 years), and all of the benign cases were diagnosed 35 years or more after the beginning of follow-up (age range, 44–75 years). In the general population, the mean age of presentation of astrocytoma is 45 years (20). Physiologically, sensitivity to radiation may vary among different brain tissues. The induction period after irradiation may be shorter for malignant tumors than for benign tumors. On the other hand, benign tumors might be diagnosed late in life instead of having a late onset. They may have existed for a very long period of time without clinical symptoms but were incidentally detected during other ear examinations. Two thyroid cancer cases occurred in the exposed group, and one case developed in the nonexposed group. These cases represented a higher risk in the exposed population, but little weight can be placed on the finding because of the small number of cases. The thyroid cases received radiation doses similar to the group's median thyroid dose of 0.09 Gy. Significantly increased risks of thyroid cancer have been reported in many radiation studies, including A-bomb survivors (17), children exposed to x-rays for tinea capitis in Israel (21), and patients treated for thymus enlargement (22). Among these populations, the thyroid doses ranged from 0.09 Gy in the Israeli tinea capitis cohort (21) to 1.36 Gy in the Rochester Thymus Study (22). Another tinea capitis cohort in New York did not find any thyroid cancer cases (15) but did report six thyroid adenomas in irradiated patients versus none in controls. The current cohort received estimated thyroid doses similar to those of the children with tinea capitis. Ron et al. (23) proposed a linear dose-response model for risk of developing thyroid cancer. On the basis of the model derived from Israeli tinea capitis patients, 1.86 cases were expected for the group exposed to radium, which was identical to the number observed in the study (two cases). The irradiated women had lower rates of breast cancer than did nonirradiated women despite the fact that the exposed group had a slightly higher frequency of risk factors commonly associated with the risk of breast cancer. The exposed group was older, better educated, older at the first pregnancy, younger at menarche, and older at menopause, and yet they had a lower risk of breast cancer (table 5). A decreased risk of breast cancer in irradiated individuals was also observed in the nasopharyngeal radium irradiation study in the Netherlands (7) as well as in the study by Hazen et al. (6). The Netherlands study found no breast cancer deaths in exposed women but two deaths in the nonexposed, for rates of zero and eight per 100,000, respectively. Hazen et al. reported no cases of breast cancer in the exposed and three cases in the sibling controls. Although these differences are not statistically significant in any of the studies, the consistency of findings across the studies is important. Furthermore, our study found that the risks of endometrial, ovarian, and prostate cancers were lower in the irradiated group. The findings lead to speculation that alterations of hormone level from pituitary irradiation might have been responsible for the decreased cancer rates among the exposed subjects. Because of the proximity of the radium applicator to the pituitary, the radiation from the treatment might have resulted in cell damage in the pituitary and might have caused subsequent mild hypopituitarism and lower levels of luteinizing hormone and follicle-stimulating hormone. This might result in lower levels of estrogen secretion via the ovary in the irradiated women due to insufficient stimulation from the pituitary. Since high levels of estrogen have been indicated as a risk factor for female cancers, the lower levels in irradiated females might have resulted in a lower risk of those hormone-related cancers. Testosterone levels might have been similarly affected in men, resulting in changes in prostate cancer. These speculations need further support and additional testing through hormone studies of the population. The relation between low-dose pituitary irradiation and hormone-related cancers has not been documented in other studies. However, an inverse relation between risk of breast cancer and pelvic radiotherapy (24–26) has been found in several populations. Serologic studies further revealed that estrogen concentrations were reduced among irradiated women (27), which might explain the reduced risk of breast cancer. Similar hormonal changes might occur after pituitary irradiation. Although this cohort is small, one of the strengths of this Washington County study is the availability of childhood clinic records and a relatively stable population that can be followed for many years. The exposure status was clearly recorded for all subjects in the fixed cohort of hearing clinic patients, and it was, therefore, possible to construct an internal comparison group. It has been difficult to establish a large cohort of children exposed to nasopharyngeal radiation, since treatment records are available in only a very few places. However, consistency of findings in different study populations would be a strong argument for a causal association. Most studies of radiation have focused on the increased cancer effects of such exposures. However, the results of follow-up of this population with nasopharyngeal exposure suggest the hypothesis that exposure of hormone-regulating organs to radiation in a period prior to puberty may result in a low risk of breast, uterus, ovary, and prostate cancers in later years. Further studies to evaluate the hormonal profiles in irradiated and nonirradiated persons would advance our understanding of the mechanism of hormone-induced cancers. Reprint requests to Dr. Hsin-chieh Yeh, Department of Epidemiology, Johns Hopkins School of Hygiene and Public Health, 2024 E. Monument Street, Suite 2-600, Baltimore, MD 21205. Supported in part by grants from the Texaco Fund, the Johns Hopkins University Summer Epidemiology Program Scholarship, and the Matanoski Gift Fund. Dr. Comstock is supported by Research Career Award HL 21670 from the National Heart, Lung, and Blood Institute. The authors thank the staff of the Johns Hopkins Training Center for Public Health Research for their extensive help with data collection and the Washington County Department of Health for the long-term support of the research. REFERENCES 1. Crowe SJ. Irradiation of the nasopharynx. Trans Am Acad Ophthalmol Laryngol  1946; 51: 29–35. Google Scholar 2. Mellinger-Birdsong AK. Estimates of numbers of civilians treated with nasopharyngeal radium irradiation in the Unites States. Otolaryngol Head Neck Surg  1996; 115: 429–32. Google Scholar 3. Canfield N, Sudarsky D. Radium therapy in partial hearing loss. Ann Otol Rhinol Laryngol  1949; 58: 957–73. Google Scholar 4. Dow CH. The use of radium in therapy of hypertrophic lymphoid tissue in the nasopharynx. Arch Otolaryngol  1949; 50: 417–28. Google Scholar 5. Bilchick EB, Kolar AR. Radium therapy for lymphoid tissue in the nasopharynx. Ann Otol Rhinol Laryngol  1950; 59: 78–89. Google Scholar 6. Hazen RW, Pifer JW, Toyooka ET, et al. Neoplasms following irradiation of the head. Cancer Res  1966; 26: 305–11. Google Scholar 7. Verduijn PG, Hayes RB, Looman C, et al. Mortality after nasopharyngeal radium irradiation for eustachian tube dysfunction. Ann Otol Rhinol Laryngol  1989; 98: 839–44. Google Scholar 8. Sandler DP, Comstock GW, Matanoski GM. Neoplasms following childhood radium irradiation of the nasopharynx. J Natl Cancer Inst  1982; 68: 3–8. Google Scholar 9. Sandler DP. Nasopharyngeal radium irradiation: The Washington County, Maryland, Study. Otolaryngol Head Neck Surg  1996; 115: 409–14. Google Scholar 10. Willard WR, Proctor DF. Results and problems after four years of a conservation of hearing program. Am J Public Health  1948; 38: 1424–33. Google Scholar 11. Proctor DF, Willard WR. Washington County (Maryland) program for the prevention of deafness in children. Arch Otolaryngol  1946; 43: 462–72. Google Scholar 12. Yeh HC, Matanoski GM, Comstock GW. A follow-up study of childhood nasopharyngeal radium irradiation in Washington County, Maryland. Otolaryngol Head Neck Surg  1996; 115: 415–16. Google Scholar 13. Bordley JE, Hardy WG. The use of radium in the aerotitis control program of the Army Air Force. Ann Otol Rhinol Laryngol  1945; 54: 650–60. Google Scholar 14. Brownell GL, Ellett WH, Reddy AR. Absorbed fractions for photon dosimetry. J Nucl Med  1968; (suppl 1): 29–39. Google Scholar 15. Shore RE, Albert RE, Pasternack BS. Follow-up study of patients treated by x-ray epilation for tinea capitis resurvey of post-treatment illness and mortality experience. Arch Environ Health  1976; 31: 21–8. Google Scholar 16. Ron E, Modan B, Boice JD Jr, et al. Tumors of the brain and nervous system after radiotherapy in childhood. N Engl J Med  1988; 319: 1033–9. Google Scholar 17. Thompson DE, Mabuchi K, Ron E, et al. Cancer incidence in atomic bomb survivors. Part II: Solid tumors, 1958–1987. (Review). Radiat Res  1994; 137: S17–67. (Published erratum appears in Radiat Res  1994; 139: 129). Google Scholar 18. Land CE. Carcinogenic effects of radiation on the human digestive tract and other organs. In: Upton AC, Albert RE, Burns FJ, et al., eds. Radiation carcinogenesis. New York, NY: Elsevier, 1986:347–78. Google Scholar 19. Pressig SH, Bohmfalk GL, Reichel GW. Anaplastic astrocytoma following radiation for a glomus jugulare tumor. Cancer  1979; 43: 2243–7. Google Scholar 20. American Cancer Society. Clinical oncology. 2nd ed. Atlanta, GA: American Cancer Society, 1995. Google Scholar 21. Ron E, Modan B, Preston D, et al. Thyroid neoplasia following low-dose radiation in childhood. Radiat Res  1989; 120: 516–31. Google Scholar 22. Shore RE, Hildreth N, Dvoretsky P, et al. Thyroid cancer among persons given x-ray treatment in infancy for an enlarged thymus gland. Am J Epidemiol  1993; 137: 1068–80. Google Scholar 23. Ron E, Lubin JH, Shore RE, et al. Thyroid cancer after exposure to external radiation: a pooled analysis of seven studies. Radiat Res  1995; 141: 259–77. Google Scholar 24. Ryberg M, Lundell M, Nilsson B, et al. Malignant disease after radiation treatment of benign gynecological disorders. Acta Oncol  1990; 29: 563–7. Google Scholar 25. Wangoner JK. Leukemia and other malignancies following radiation therapy for gynecological disorders. In: Boice JD Jr, Fraumeni JF Jr, eds. Radiation carcinogenesis: epidemiology and biological significance. New York, NY: Raven Press, 1984:153–9. Google Scholar 26. Darby SC, Reeves G, Key T, et al. Mortality in a cohort of women given x-ray therapy for metropathia haemorrhagica. Int J Cancer  1994; 56: 793–801. Google Scholar 27. Boice JD Jr, Blettner M, Kleinerman RA, et al. Radiation dose and breast cancer risk in patients treated for cancer of the cervix. Int J Cancer  1989; 44: 7–16. Google Scholar 28. Haldane JBS. The estimation and significance of the logarithm of a ratio of frequencies. Ann Hum Genet  1956; 2: 309–11. Google Scholar

Yeh, Hsin-chieh; Matanoski, Genevieve M.; Wang, Nae-yuh; Sandler, Dale P.; Comstock, George W.

doi: N/Apmid: N/A

A population from a hearing clinic in Washington County, Maryland, in 1943–1960 was followed to assess the risk of developing neoplasms from radium treatment of the nasopharynx for adenoid hypertrophy. Of the 2,925 subjects who attended the clinic, 904 received radium treatment. A nonconcurrent prospective study compared the cancer incidence among the irradiated persons with that among persons with other treatments. Seven brain tumor cases (three malignant and four benign) were identified in the irradiated group versus none in the nonirradiated group (relative risk = 14.8, 95% confidence interval: 0.76, 286.3). A nonsignificant excess risk of thyroid cancer was detected in the irradiated group based on two cases in the exposed group and one case in the nonexposed group (relative risk = 4.2, 95% confidence interval: 0.38, 46.6). Decreased risks of breast cancer, female genital cancers, and prostate cancer were observed among the irradiated individuals, although these deficits were not statistically significant individually. The decreased risk of sex hormone-related cancers in the irradiated group suggests possible radiation damage to the pituitary, with consequent reduction in pituitary hormone output and alterations in sexual and other hormonal development in early life. This hypothesis needs further evaluation.