JAMA Internal Medicine

Larson, Eric B.

doi: 10.1001/archinte.158.9.941pmid: 9588426

Hanzlick, Randy; Hutchins, Grover M.

doi: 10.1001/archinte.158.9.947pmid: 9588427

Gotsman, Mervyn S.; Admon, Dan

doi: 10.1001/archinte.158.9.949pmid: 9588428

Whitley, Richard J.; Jacobson, Mark A.; Friedberg, Dorothy N.; Holland, Gary N.; Jabs, Douglas A.; Dieterich, Douglas T.; Hardy, W. David; Polis, Michael A.; Deutsch, Thomas A.; Feinberg, Judith; Spector, Stephen A.; Walmsley, Sharon; Drew, W. Lawrence; Powderly, William G.;

Sulmasy, Daniel P.; Linas, Benjamin P.; Gold, Karen F.; Schulman, Kevin A.

doi: 10.1001/archinte.158.9.974pmid: 9588430

ObjectiveTo explore the relationship between general internists' tendency to conserve medical resources and their willingness to participate in physician-assisted suicide (PAS).Design and ParticipantsSurvey of a random sample of general internists in 6 urban areas of the United States.MeasurementsWe assessed the physicians' use of medical resources by constructing a scale based on 6 hypothetical clinical scenarios in which respondents were given a choice between resource-intensive and resource-conserving options. We then presented a scenario of a competent terminally ill patient with breast cancer making stable and persistent requests for PAS.ResultsSixty-seven (33%) of the 206 respondents indicated that they would participate in the suicide of the depicted patient. In a multivariate model, physicians who were more conservative with resources were 6.4 times more likely than their resource-intensive counterparts to prescribe the requested drugs (P=.02); minority physicians were less willing than whites to participate in PAS (odds ratio, 0.34; P=.03). Physicians' number of years in practice, location, sex, reported percentage of fee-for-service patients, and self-reported strength and direction of financial incentives in the respondents' practices were not associated with willingness to prescribe drugs for PAS.ConclusionsMost general internists, especially minority physicians, are personally reluctant to participate in PAS. While the characteristics of their practices do not affect PAS, physicians who tend to practice resource-conserving medicine are significantly more likely than their resource-intensive counterparts to provide a lethal prescription at the request of a terminally ill patient.THE MORAL, legal, and social issues surrounding the question of whether physicians should assist terminally ill patients with suicide have proved divisive for the profession and the American public. The US Supreme Court recently ruled that it is not unconstitutional for states to prohibit physician-assisted suicide (PAS).However, since the Supreme Court did not rule that states are required to prohibit PAS, the debate has effectively been shifted to the state governments. The state of Oregon has passed legislation legalizing the practice,and a recent referendum to repeal this law was defeated in November 1997.Many Americans support the legalization of PAS.The American Medical Association remains officially opposed to legalization,but several surveys have shown that a significant minority of physicians have actually engaged in this practiceand many physicians would support legalization.It has been suggested that pressures for cost-control have the potential for a deleterious effect on end-of-life decision making,and that legalization of PAS might be unwise because the presently intense pressure placed on health care professionals to control costs could result in undue influence on terminally ill patients to opt for PAS.Several studieshave surveyed physician attitudes and practices regarding PAS in the United States. To the best of our knowledge, only 1 studyat a single institution asked physicians about the "importance of cost-benefit analysis" in decisions about PAS, but this study did not examine any linkage between resource-conserving attitudes and physician willingness to provide PAS.To assess this relationship, we surveyed physicians from 6 major metropolitan areas in the United States.METHODSSURVEYSData were collected via a self-administered, mailed questionnaire. Surveys presented respondents with 7 medical scenarios as well as items collecting information about personal and practice characteristics. The first 6 scenarios depicted hypothetical patients who presented with symptoms commonly encountered by primary care physicians, including asthma, dysuria, headache, persistent cough, chest pain, and dementia. The seventh scenario presented a 72-year-old retired university professor with metastatic breast cancer refractory to further chemotherapy who requests PAS, and meets proposed guidelines for morally acceptable PAS.SCENARIO REGARDING PASA 72-year-old retired university professor with metastatic breast cancer, now refractory to further chemotherapy, is suffering from severe pain due to spinal metastases. Vigorous efforts at pain relief using oral narcotics, nonsteroidal anti-inflammatory drugs, tricyclic antidepressants, and an epidural pump have been only partially successful. The patient is a long-standing member of the Hemlock Society and has a living will. She has never married and has no family. She deplores the idea of hospice as an invasion of her privacy. A month ago, she asked for your help in arranging "death with dignity" through an overdose. You asked her to give the pain control methods a chance and that she see a psychiatrist. The psychiatrist found her competent. She now says that the very chronicity of the residual pain has become unbearable, and she is repeating her request for assistance in finding death with dignity, "before it's too late." The choices are as follows. (1) Explain that you understand how she feels, that you are grateful that she has been patient with your request that she tries other means first, and that you will now support her decision by giving her instructions and the prescription that she has requested. (2) Tell her that although you understand how she feels, you are morally opposed to assisted suicide, and offer her the opportunity to find another physician if she cannot accept this limitation on what you are prepared to do for her. (3) Other (please briefly explain).The 6 primary care scenarios were designed to be toss-up situations in which 2 answer choices (one resource conserving and the other resource intensive) would both be considered medically appropriate. They were presented in sufficient clinical detail to allow a reasonable medical judgment. For example, in the dysuria scenario, an otherwise healthy, nonpregnant, 32-year-old woman with no allergies telephones the physician on call during the weekend reporting the new onset of a temperature of 38°C and flank pain following 3 days of dysuria. Choices were to prescribe oral antibiotics via the telephone or to refer the patient to the local emergency department for diagnosis and treatment. If respondents did not feel comfortable with the provided treatment options, they could indicate an "other" treatment option with a brief explanation. These other responses were then independently coded by 2 blinded researchers and discrepancies were reconciled after discussion. In coding other responses to the PAS scenario, any explicit indication that the physician would be willing to prescribe the requested medication (eg, "I would write the prescription but not encourage the patient to take it") was coded as indicating willingness to prescribe the requested drugs. (The complete instrument is available on request from us).To help offset framing effects, respondents were randomly assigned to 1 of 2 groups. Half were asked to choose the treatment they would recommend for the patient depicted in each scenario. The other half were asked to consider a caseload of 100 patients similar to the patient depicted by each scenario and to indicate the percentage of such patients for whom they would recommend each treatment option. The order of the first 6 scenarios was also randomly varied.In addition to hypothetical scenarios, surveys collected information about respondents' practice characteristics, such as self-estimated percentage of fee-for-service (FFS) patients and their perceptions of the strength and direction of the financial incentives in their personal medical practices.The survey was pilot tested and revised to ensure clarity and variability in responses. Based on the pilot data, we calculated that a respondent pool of 200 physicians would be an adequate sample size for our regression analysis.We used the official American Board of Medical Specialists directory to select randomly 550 general internist physicians, stratified by 6 large metropolitan areas in the United States: New York, NY, Philadelphia, Pa, Chicago, Ill, Los Angeles, Calif, Atlanta, Ga, and Washington.STATISTICAL ANALYSISTo assess the physicians' orientation with respect to resource conservation, we first constructed a Resource Conservation Index (RCI). The answers of those respondents who were given a dichotomous choice between resource-intensive and resource-conserving options were scored as 0 or 1.0, respectively. The answers of those respondents given the caseload response option were dichotomized and scored as 0 if the respondent indicated that the resource-intensive option would be chosen in 50% or more cases and scored as 1.0 if the respondent indicated that the resource-intensive option would be chosen for less than 50% of cases. The RCI is the unweighted mean response, on a scale of 0 to 1.0, with higher scores indicating a more resource-conserving response pattern. Using χ2analyses and multivariate logistic regression, we assessed the relationship between RCI and responses to the assisted suicide scenario.In addition, we assessed the relationship between reported financial incentives and responses to the assisted suicide scenario using multivariate logistic regression and analysis of variance.RESULTSOf the original sample of 550 physicians, we failed to contact 60 because they had either moved or retired, leaving an effective sample size of 490. Following multiple mailings and telephone calls to nonresponders, we received 207 responses, all but 1 of which were complete, yielding a 42% rate of return of usable questionnaires. Twenty-four percent of scenario responses were other and required recoding as described in the "Methods" section. The agreement rate between the 2 recoders was 93%.DEMOGRAPHICSResponding physicians did not differ significantly in age or number of years in practice from nonresponders. Responders were more likely to practice medicine in a group practice environment (P<.01). Philadelphia had the highest response rate (55%) and Los Angeles the lowest (31%).The mean age of the respondent pool was 47 years (range, 31-83 years). The respondent pool was 166 (81%) men and 171 (83%) white. Thirty-one (15%) respondents were from New York, 32 (16%) from Philadelphia, 41 (20%) from Chicago, 17 (8%) from Atlanta, 33 (16%) from Los Angeles, and 50 (25%) from Washington. Sixty (29%) respondents reported practicing in a solo private practice, 86 (42%) in a single-specialty group practice, 41 (20%) in a multispecialty group practice, and 19 (9%) in another practice.RESOURCE UTILIZATIONThe response scores for the individual items on the RCI are shown in Table 1. As illustrated, there was substantial variability in the respondents' choices, indicating a range of responses that might be considered appropriate care.Table 1. Resource Conservation Index*See table graphicPRACTICE FINANCIAL CHARACTERISTICSRespondents reported that, on average, 62% of their patients offer payment on an FFS basis. The percentage of patients paying on an FFS basis was significantly and positively correlated with the number of years that a physician had been in medical practice (r=0.23; P=.001).The respondents' perceptions of the strength and direction of the financial incentives under which they practice indicated a slight tendency toward incentives to limit tests, treatments, and referrals, with 39% of respondents reporting strong or moderate incentives to limit tests, treatments, and referrals, 48% reporting neutral financial incentives, and 13% reporting that they practice under strong or moderate incentives to increase therapies. As expected, the percentage of patients paying on an FFS basis was significantly correlated with perceptions of incentives to increase tests, treatments, and referrals (r=0.30; P=.001).ASSISTED SUICIDESixty-seven (33%) respondents indicated that they would participate in the suicide of the patient depicted in the seventh scenario. There was a tendency for a resource-conserving response on 4 of the 6 individual items to be associated with willingness to participate in PAS. The odds ratios for willingness to prescribe were 2.33, 1.33, 1.32, 1.45, 0.94, and 1.03 for the pyelonephritis, asthma, chest pain, dementia, headache, and cough scenarios, respectively. Only the pyelonephritis scenario was significantly associated with PAS as an individual item (P=.005). The mean RCI for physicians who would assist with suicide was 0.42, while it was 0.35 for those who would not (P=.03).As shown in Figure 1, there was a strong, linear relationship between the number of items on which the individual chose the resource-conserving options and the likelihood that he or she would be willing to participate in PAS. The χ2value for linear trend was 4.90 (P=.03).Percentage of physicians who would assist in suicide according to resource conservation score (see the "Methods" section of the text).In a multivariate logistic regression model (Table 2) controlling for respondent race, sex, response format (case load vs individual), and number of years in medical practice, physicians who tended to choose resource-conserving treatment strategies in the first 6 medical scenarios were 6.4 times more likely than their resource-intensive counterparts to prescribe the requested drugs (P=.02). In addition, minority physicians were less likely to prescribe the requested lethal dose than were whites (odds ratio, 0.34; P=.03). The physician's sex (P=.94), number of years in medical practice (P=.93), and response format (P=.49) were not associated with willingness to prescribe. Adding self-reported percentage of FFS and perceived financial incentives to the model increased the odds ratio for the RCI variable to 7.4 but did not increase the model's overall predictive power and so were excluded from the final model. Likewise, city and practice type were not significant in bivariate analyses and did not increase the predictive power of the model and were thus also excluded from the final model.Table 2. Multivariate Logistic Regression Model of Characteristics Associated With Willingness to Provide Assistance With Suicide to a Terminally Ill Woman*See table graphicIn a separate multiple logistic regression model, also controlling for physician sex, race, and number of years in medical practice, neither the reported percentage of patients paying on an FFS basis (P=.82) nor the physicians' perceptions of the strength and direction of financial incentives in medical practices (P=.22) were associated with willingness to prescribe. Compared with physicians in Philadelphia, physicians in New York (P=.61), Chicago (P=.62), Los Angeles (P=.73), Washington (P=.73), and Atlanta (P=.96) were neither more nor less likely to prescribe the medication.COMMENTSurveys have reported a marked division among American physicians regarding their support for PAS.In our survey of general internists, 33% indicated a willingness to write a lethal prescription for a competent, terminally ill patient suffering from pain not relieved by substantial analgesic interventions.This study has identified a significant, strong, linear association between the tendency of general internists to choose resource-conserving treatment options and their willingness to assist a terminally ill patient with suicide, as assessed by their responses to hypothetical cases. This association is independent of physician sex, race, years in practice, or geographic location.Our study is limited because it involved only a small number of physicians in a single specialty. However, it has the advantage of constituting a national sample of general internists—a group likely to be involved in the practice of PAS if it is legalized. Rural physicians were excluded, but both the cost-containment and PAS debates seem centered in the urban and suburban settings. While response bias cannot be completely excluded, respondents did not differ from nonrespondents in age or number of years in practice. Furthermore, even if the relationship we have shown were to hold only among those who responded, the fact that such a subpopulation can be identified is itself of interest.The validity of the findings from this study might be questioned. However, we minimized bias in the recoding of other responses by blinding and by using 2 recoders. We used carefully pretested scenarios and questions, and relied on a scale of resource conservation rather than a single response. We found an association or a trend toward an association between PAS and resource-conserving responses on 4 of the 6 individual items on the scale. Moreover, the relationship between the number of resource-conserving options chosen and willingness to participate in PAS was linear. These facts strongly suggest that a definite underlying attitude was identified.Neither the physicians' reported percentage of FFS patients nor their perceptions of the strength and direction of the financial incentives in their practices correlated with the physicians' willingness to prescribe lethal medication to the terminally ill patient depicted in the scenario. This might suggest that physicians' attitudes regarding resource conservation and PAS are intrinsic characteristics of the physicians themselves and are not related to the financial environment of the practice. However, our data on the financial characteristics of the practice are based on unvalidated physician reports. In addition, the reported percentage of FFS (62%) was high in these practices, and only 39% reported fiscal incentives to limit costs.We found that minority physicians showed less willingness than their white colleagues to prescribe lethal medication to a terminally ill patient. This corroborates the minority opposition to euthanasia that Blendon et alhave reported in their surveys of the general public.It would be a mistake to conclude automatically from this study that physicians view PAS as a cost-containment mechanism. Demonstrating an association between a resource-conserving practice style and willingness to participate in PAS does not establish a causal relationship. While it cannot be ruled out that PAS may be viewed by these physicians, explicitly or implicitly, as a means of conserving resources, other explanations are also plausible. It is possible, for instance, that some physicians are more resource intensive because they are averse to risk, and might therefore be less likely than their resource-conserving colleagues to participate in PAS. It is also possible that physicians who are more resource conserving might simply be more likely to view it as irrational to insist that patients who choose to end their lives should be prohibited from doing so when such acts will, in fact, conserve scarce medical resources. This view of rationality and justice, however, does not meet with universal approval.The association we report does not constitute proof that abuse of PAS will result from the legalization of this practice in a cost-constrained environment. Nonetheless, it suggests a sobering degree of caution in legalizing PAS in a medical care environment that is characterized by increasing pressure on physicians to control the cost of care. This information should be carefully considered and studied further as policymakers, lawyers, and judges debate whether PAS should be legalized.Not AvailableVacco v Quill.WL 348037 (US) (1997).Not AvailableWashington v Glucksberg.WL 348094 (US) (1997).Not AvailableNot AvailableOregon Death With Dignity Act (1994).TEganIn Oregon, opening a new front in the world of medicine.New York Times.November 6, 1997:A26.RJBlendonUSSzalayRAKnoxShould physicians aid their patients in dying? the public perspective.JAMA1992;267:2658-2662.KMSandrickAMA rigidly opposed to assisted suicide.Med Tribune.1996;37:1-5.ALBackJIWallaceHEStarksRAPearlmanPhysician assisted suicide and euthanasia in Washington State: patient requests and physician responses.JAMA.1996;275:919-925.DJDoukasDWaterhouseDWGorenfloJSeidAttitudes and behaviors on physician assisted death: a study of Michigan oncologists.J Clin Oncol.1995;13:1055-1061.JSCohenSDFihnEJBoykoARJonsenRWWoodAttitudes towards assisted suicide and euthanasia among physicians in Washington State.N Engl J Med.1994;331:89-94.JGBachmanKHAlcserDJDoukasRLLichtensteinADCorningHBrodyAttitudes of Michigan physicians and the public towards legalizing physician-assisted suicide and euthanasia.N Engl J Med.1996;334:303-309.RSShapiroARDerseMGottliebDSchiedermayerMOlsonWillingness to perform euthanasia: a survey of physician attitudes.Arch Intern Med.1994;154:575-584.TRFriedMDSteinPSO'SullivanDWBrockDHNovackLimits of patient autonomy: physician attitudes and practices regarding life-sustaining treatments and euthanasia.Arch Intern Med.1993;153:722-728.MALeeHDNelsonVPTildenLGanziTASchmidtSWTolleLegalizing assisted suicide: views of physicians in Oregon.N Engl J Med.1996;334:310-315.SHMilesEPWeberRKoeppEnd-of-life treatment in managed care: the potential and the peril.West J Med.1995;163:302-305.HJBursztajnABrodskyClear, convincing, and authentic advance directives in the context of managed care?J Clin Ethics.1994;5:364-366.DPSulmasyManaged care and managed death.Arch Intern Med.1995;155:133-136.MCKavenyJPLanganThe doctor's call.New York Times.July 15, 1996:A13.TSmithCheap, managed death.BMJ.1995;310:744.SMWolfPhysician-assisted suicide in the context of managed care.Duquesne Law Rev.1996;35:455-479.EJEmanuelDLFaircloughERDanielsBRClarridgeEuthanasia and physician-assisted suicide: attitudes and experiences of oncology patients, oncologists, and the public.Lancet.1996;347:1805-1810.PVCaralisJSHammondAttitudes of medical students, housestaff, and faculty physicians towards euthanasia and termination of life-sustaining treatment.Crit Care Med.1992;20:683-690.TEQuillCKCasselDEMeierCare of the hopelessly ill: proposed clinical criteria for physician-assisted suicide.N Engl J Med.1992;327:1380-1384.ATrollopeThe Fixed Period.New York, NY: Oxford University Press Inc; 1993.AMacIntyreWhose Justice? Which Rationality?Notre Dame, Ind: University of Notre Dame Press; 1988.Accepted for publication August 29, 1997.This research was supported by a starter grant from the Georgetown University Department of Medicine Research. Dr Sulmasy receives support as a Soros Faculty Scholar of the Project on Death in America.Reprints: Daniel P. Sulmasy, OFM, MD, PhD, Center for Clinical Bioethics, Room 236, Bldg D, Georgetown University Medical Center, Washington, DC 20007 (e-mail: [email protected]).

Chandra, Nisha C.; Ziegelstein, Roy C.; Rogers, William J.; Tiefenbrunn, Alan J.; Gore, Joel M.; French, William J.; Rubison, Michael

doi: 10.1001/archinte.158.9.981pmid: 9588431

BackgroundTo determine whether there are sex differences in the demographics, treatment, and outcome of patients with acute myocardial infarction in the United States, data from the National Registry of Myocardial Infarction-I from September 1990 to September 1994 were examined.MethodsThe National Registry of Myocardial Infarction-I is a national observational database consisting of 1234 US hospitals in which each hospital submits data from each patient with acute myocardial infarction to a central data collection center. For these analyses, the following variables were examined in 354435 patients with acute myocardial infarction: demographics; use of medical therapy including thrombolytic agents; use of procedures including cardiac catheterization, percutaneous transluminal coronary angioplasty, and coronary artery bypass surgery; length of hospital stay; adverse events (stroke, major bleeding, or recurrent myocardial infarction); and causes of death.ResultsIn comparison with men, women experiencing acute myocardial infarction in the United States are older, with 55.7% older than 70 years. Women have a higher mortality rate than men even when controlled for age and die less often from arrhythmia but more often from cardiac rupture whether or not thrombolytic therapy is used. Treatment with aspirin, heparin, or β-blockers is less frequent in women. When thrombolytic therapy is used, women are treated an average of almost 14 minutes later than men and experience a greater incidence of major bleeding. Cardiac catheterization, percutaneous transluminal coronary angioplasty, and coronary artery bypass surgery are used less often in women.ConclusionsObservations from the National Registry of Myocardial Infarction-I document important sex differences in demographics, treatment, and outcome of patients with acute myocardial infarction in the United States.CORONARY artery disease is the leading cause of death and is responsible for more than one third of all deaths in women in the United States.While the clinical manifestations of coronary artery disease in women are delayed by approximately a decade compared with men, this sex difference nearly disappears in older age.Studies of acute myocardial infarction (AMI) from the prethrombolytic era demonstrate that women have a worse prognosis than men.This observation has been supported by more recent large clinical trials of thrombolytic agents in both placebo and thrombolysis groupsand was emphasized recently by the GUSTO-I investigators.This difference in outcome may be explained in part by older age, more advanced disease, higher Killip class, and a greater prevalence of underlying medical illness in women than in men.Recent attention has also focused on the role of sex differences in diagnostic and therapeutic strategies in patients with known or suspected coronary artery disease. Manybut not allstudies have reported that physicians are less likely to pursue an aggressive approach in women than in men.No recent large study has focused on the demographics and routine clinical use of drug therapy or procedures in women experiencing AMIs in the United States, including both those treated and not treated with thrombolytic therapy. In this article, we describe observations relating to the treatment and outcome of men and women with AMI in the United States based on data from more than 350000 patients in the National Registry of Myocardial Infarction-I (NRMI-I) from September 1990 to September 1994.METHODSThe detailed methods for this study have been reported previously.In brief, the NRMI is a voluntary observational database sponsored by Genentech Inc, South San Francisco, Calif. The purpose of the NRMI is to collect uniform prospective data on the treatment of patients with AMI that can be used to analyze practice patterns and treatment outcomes. Participating sites receive quarterly information on local care to facilitate quality improvement. Phase I of the registry (NRMI-I), from which this study is drawn, entered 354435 patients from September 1990 to September 1994. A second registry, NRMI-II, is ongoing.Demographic, procedural, and outcome data on patients with AMI are recorded by participating hospitals throughout the United States. To facilitate uniform data collection, each registry data coordinator received in-service training on data collection methods and terminology as defined in the 38-page manual for form completion for NRMI. Thus, standardized terminology and criteria for item entry are defined for each hospital. Data from each hospital are tabulated by a central data collection center (ClinTrials Research Inc, Lexington, Ky). Participating hospitals receive quarterly summaries of the cumulative study-wide NRMI-I data together with confidential, individualized, parallel tabulations of the hospital data.Approval of the NRMI-I data collection process at participating hospitals included review by individual institutional committees on human research as determined by local policies and procedures.DEFINITIONSTo be included in NRMI-I, patients had to have an AMI documented as the discharge diagnosis according to local hospital criteria, encompassing elevated levels of cardiac enzymes (total or creatine kinase–MB), electrocardiographic (ECG) changes, or findings supported by coronary angiography. Time of chest pain onset was defined as the time when chest pain intensified or became prolonged or intolerable such that the patient decided to seek treatment. Time of initial presentation was defined as the time that the patient arrived at the NRMI-I hospital or the referral hospital if that was earlier. Patient initials and birth dates were recorded to minimize double counting of patients who might have been transferred from one registry hospital to another. For purposes of this analysis, MI location was classified as anterior, nonanterior, or other, including nonspecific ST- or T-wave abnormalities or no ECG evidence of MI. The location of MI, as defined by the attending physician, was determined by chart review. The use of concomitant medications (intravenous [IV] heparin, inotropic agents, IV β-blockers, aspirin, calcium channel blockers, or IV nitroglycerin), invasive procedures, and adverse events were recorded if they were implemented or occurred at any time during the hospitalization. Peri-MI arrhythmias included any arrhythmia occurring within 24 hours of the index infarction. Recurrent MIs required the patient to evolve ECG changes and elevated levels of creatine kinase indicative of a second infarct. Drug-induced hypotension was described as hypotension requiring at least fluid replacement therapy. Cause of death was determined by review of the patient's chart and/or death certificate (as noted by the attending physician) and categorized as being due to sudden cardiac arrest (collapse in an unmonitored situation); cardiogenic shock (low blood pressure [usually <90 mm Hg] with signs of hypoperfusion, cool, clammy skin, oliguria or altered sensorium, and nonresponsiveness to fluid resuscitation or pressor agents); recurrent MI (as defined earlier); arrhythmia (to include supraventricular tachycardia, ventricular tachycardia or fibrillation, or any arrhythmia resulting in hemodynamic compromise); cardiac rupture and/or electromechanical dissociation (EMD) (based either on autopsy findings, clinical course, or the sudden development of pulseless electrical activity); other cardiac causes (pericarditis or valvular insults); or noncardiac causes of death (eg, pneumonia, pulmonary embolism, or sepsis).STATISTICAL METHODSGroup differences based on categorical variables were assessed using the χ2test. Differences based on continuous variables (such as age and weight) were assessed using the 2-sample ttest, and differences based on time intervals were assessed using the nonparametric Wilcoxon signed rank test. When appropriate, data were expressed as mean (± SD). Multivariate logistic regression analyses were performed to assess sex differences after accounting for the presence of those potential confounding variables that were recorded in NRMI-I. All statistical analyses were performed using SAS 6.06 statistical package programs (SAS Institute Inc, Cary, NC). To adjust for the large number of comparisons, a Bonferroni adjustment was applied.This adjustment identified a Pvalue of .003 as corresponding to an α level of .05. Thus, significance was defined as P≤.003.RESULTSDEMOGRAPHICS: GENERAL CHARACTERISTICSFrom September 1990 to September 1994, 354435 patients were enrolled in the NRMI-I from 1234 contributing hospitals. Table 1lists the main demographic features of the patients enrolled and divides them by sex and use of thrombolytic therapy. The majority of the NRMI-I population was male (63.7%). Only 34.8% of all patients received thrombolytic therapy. Those who received thrombolytic agents were younger than those who did not. The average time to presentation for all patients was more than 5 hours (mean ± SD, 307.9 ± 444 minutes) and was approximately 3 hours longer for those not treated with thrombolytic therapy than for those who received thrombolytic agents (mean ± SD, 397.9 ± 513.3 minutes vs 176.8 ± 267.3 minutes) (P<.001). The average hospital stay was a mean ± SD of 9.2 ± 15.3 days and was longer for those not treated with thrombolytic therapy (mean ± SD, 9.5 ± 16.1 days vs 8.7 ± 15.6 days) (P<.001). Patients who received thrombolytic therapy were more likely to have an MI in which location could be specified compared with those who did not receive thrombolytic therapy. In those who received thrombolytic therapy, the initial ECG was commonly used to make the diagnosis of MI, whereas in those who did not receive thrombolytic agents, the subsequent ECG and determination of serial cardiac enzymes were more likely to be used to establish the diagnosis.Table 1. Demographics*See table graphicDEMOGRAPHICS: SEX DIFFERENCESAs noted in Table 1, women were substantially older than men for both thrombolytic and nonthrombolytic groups. The majority of women (55.7%) were 70 years or older compared with only 32.6% of the men. Women presented later than men in both thrombolytic (mean ± SD, 194.9 ± 281 minutes vs 169.5 ± 262 minutes) and nonthrombolytic (mean ± SD, 408.2 ± 517 minutes vs 390.9 ± 511 minutes) groups. The time from presentation to thrombolytic therapy was almost 14 minutes longer for women than for men. Women treated with thrombolytic therapy weighed more than those who were not. Of those who received thrombolytic agents, 12979 (37.5%) of 34569 women were older than 70 years compared with only 16876 (19.5%) of the 86508 men (P<.001). Nevertheless, among the NRMI-I patients older than 70 years, men received thrombolytic therapy more often than women (16876 [23.3] of 72312 men vs 12979 [18.5] of 70221 women) (P<.001).USE OF ADDITIONAL PROCEDURESFor both men and women, additional procedures (cardiac catheterization, percutaneous transluminal coronary angioplasty, or bypass surgery) were more likely to be used in those who received thrombolytic therapy compared with those who did not (Table 2). For example, 69.1% of those who received thrombolytic agents underwent cardiac catheterization compared with only 48.5% of those who were not treated with thrombolytic therapy (P<.001). Almost twice as many patients who received thrombolytic therapy underwent percutaneous transluminal coronary angioplasty (29.9%) as those who were not treated with thrombolytic agents (16.9%) (P<.001). However, in both treatment groups women were less likely than men to undergo additional procedures. Of those not treated with thrombolytic therapy, women were less likely to undergo cardiac catheterization (40% vs 54.4% of men), percutaneous transluminal coronary angioplasty (13.8% vs 19% of men), and coronary artery bypass grafting (8% vs 12.6% of men) (P<.001 for all). Similarly, in those who received thrombolytic therapy, these procedures were used less often in women than in men (64.3% vs 71% of men for catheterization; 28.3% vs 30.6% of men for percutaneous transluminal coronary angioplasty; and 11% vs 13.5% of men for coronary artery bypass grafting) (P<.001).Table 2. Procedures and Adverse Events*See table graphicADVERSE EVENTSAll adverse events recorded for stroke, major bleeding, and recurrent MIs were more common in patients receiving thrombolytic therapy. Women had a higher occurrence of stroke, major bleeding, and recurrent MIs than men, whether or not they received thrombolytic agents. Allergic and anaphylactic reactions were reported only in patients receiving thrombolytic therapy (Table 2).MORTALITY AND CAUSE OF DEATH AFTER MYOCARDIAL INFARCTIONThe mortality rate was substantially lower in men than in women for similar treatment strategies and age groups. However, the mortality in both treatment groups increased as a function of age (Table 3and Figure 1). The most common causes of death in all patients with AMI were cardiogenic shock and sudden cardiac arrest. A lower mortality rate was observed for men and women who received thrombolytic therapy compared with those who did not (Figure 1). Cardiogenic shock and rupture and/or EMD were more common causes of death in those who received thrombolytic therapy, while sudden cardiac arrest and death due to noncardiac causes were less common in these patients. In women, death from arrhythmias occurred less frequently than in men, but a higher frequency of rupture and/or EMD was noted (Table 3). Comparing cause of death in men treated with thrombolytic therapy with men who did not receive such therapy, similar relationships were found, with a higher incidence of cardiogenic shock and myocardial rupture, although a lower frequency of sudden cardiac arrest and death due to noncardiac causes was noted in patients treated with thrombolytic agents. For those who did not receive thrombolytic therapy, death from recurrent MIs and rupture and/or EMD was more common and death from arrhythmia less common in women than in men.Table 3. Mortality*See table graphicFigure 1.Mortality rates in females (F) and males (M) with no thrombolytic therapy (No TT) and in those treated with thrombolytic agents (TT) in patients aged 50 years and younger, 50 to 60 years, 60 to 70 years, 70 to 80 years, and older than 80 years. Absolute mortality in each category is also displayed. Mortality figures were not adjusted for other confounding variables (hypertension, diabetes, tobacco use, or hyperlipidemia).CONCOMITANT MEDICATIONSThe NRMI database tracked the use of IV heparin, inotropic agents, IV or oral β-blockers, aspirin, calcium channel blockers, and IV nitroglycerin. In both men and women treated with thrombolytic agents, the use of heparin, β-blockers, aspirin, and IV nitroglycerin was substantially higher than in those not treated with thrombolytic therapy (Table 4). In all patients not treated with thrombolytic therapy, the use of calcium channel blockers was more common. In both those treated with thrombolytic therapy and those who did not receive thrombolytic agents, survivors were more likely to have received heparin, β-blockers, aspirin, and IV nitroglycerin compared with those who died. However, irrespective of treatment with thrombolytic agents, the use of these concomitant medications was always lower in women with the least frequent use being noted in women who died. Mean time to death in all patients was longer than 5 days (Table 3), suggesting that the reason for not using concomitant therapy was unlikely to be very early death. When the median time to death was examined (3.75 days for all patients) similar observations were noted (median time: 4.1 days for men not treated with thrombolytic therapy, 4.0 days for women not treated with thrombolytic therapy, 2.6 days for men treated with thrombolytic therapy, and 2.3 days for women treated with thrombolytic therapy).Table 4. Concomitant Medication Use*See table graphicMULTIVARIATE LOGISTIC REGRESSION ANALYSISA multivariate logistic regression analysis was performed to address the relative importance of the following variables in predicting mortality: age, sex, anterior MI, and use of IV heparin, oral or IV β-blockers, aspirin and IV nitroglycerin, calcium channel blockers, or thrombolytic therapy (Figure 2). Although mortality was most dependent on age and use of thrombolytic therapy, female sex still emerged as a predictor of poor outcome (Figure 2) (P<.001; odds ratio, 0.78; 95% confidence interval [CI] 0.77-0.80). Further analyses were performed to test the interaction of age and sex, since most women were older than 70 years. A significant interaction effect was detected between ages 65 to 75 years and sex (P<.001; odds ratio, 1.35; 95% CI, 1.26-1.44) and also between ages older than 75 years and sex (P<.001; odds ratio, 1.59; 95% CI, 1.49-1.69).Figure 2.Odds ratios and 95% confidence intervals for mortality effects for the variables listed. IV indicates intravenous; MI, myocardial infarction.TEMPORAL TRENDSAs depicted in Figure 3, the number of patients recruited per year into this observational database has increased. Between 1990 and 1993, 39%, 40%, 39%, and 39% of men were treated with thrombolytic agents compared with 26%, 27%, 27%, and 28% of women for comparable years. The mortality rate remained comparable for men and women during this time.Figure 3.Total patient recruitment in thousands for the years 1990 to 1993 for all patients (first bar); all males (M) (second bar); and all females (F) (third bar). The dark portion of each bar reflects the number of patients receiving thrombolytic therapy (TT). The percentage of patients treated with thrombolytic therapy each year for all patients and for all men and women is noted. The mortality rate in those treated with and those not treated (no TT) with thrombolytic agents is also noted.COMMENTCoronary artery disease is the leading cause of death in women in the United States.This observational study of more than 350000 patients reflects clinical practice in more than 1200 hospitals nationwide and demonstrates that, compared with men, women with AMI have a higher mortality rate whether or not they receive thrombolytic therapy; die more often from cardiac rupture; and are less likely to be treated with aspirin, heparin, or β-blockers. Several studies have demonstrated that women who sustain an AMI have a worse prognosis than men.While several studies examining practice trends at individual hospitals or in specific states suggest that a sex difference exists in the treatment of patients with AMI, this is the first study to examine and report on national practice patterns. The initial observations in this article from 354435 patients with AMIs in the United States suggest that important sex differences exist with regard to demographics, mortality rate, cause of death, and the use of anti-ischemic and thrombolytic therapy, cardiac catheterization, percutaneous transluminal coronary angioplasty, and coronary artery bypass surgery.When coronary artery disease is suspected based on abnormal findings from a radionuclide exercise test, women are much less likely to be referred for coronary angiography than men.Even when women are hospitalized for coronary heart disease they appear to undergo fewer diagnostic and therapeutic procedures than men.Fewer women in the Survival and Ventricular Enlargement Trialunderwent cardiac catheterization or coronary artery bypass surgery than men. This was evident despite women having had more symptoms consistent with greater functional disability from angina.Some authorshave suggested that the higher operative mortality rate of women undergoing coronary artery bypass surgery may be explained in part by a difference in referral patterns, with women being considered for revascularization at a later stage in their disease. Other investigators have focused on the use of medical therapy in men and women to explain the differences in outcome. Despite the documented benefit of thrombolytic therapy for AMI, it may often be underused in women. Only 14.4% of women in the MITI trial received thrombolytic therapy compared with 26% of men, although use was comparable in those with ST wave elevation.In the Western Washington trial, only 16% of women were considered eligible to receive thrombolytic therapy compared with 25% of men.Even when considered eligible, a significantly lower percentage of women (55%) in the Western Washington trial received thrombolytic therapy compared with eligible men (78%).In the International Tissue Plasminogen Activator/Streptokinase Mortality Study,even when women received thrombolytic therapy they were given treatment considerably later (18 minutes) than men, a finding similar to our observations. While there was no sex difference in the use of nitrates, β-blockers, and calcium channel blockers in the Survival and Ventricular Enlargement Trialmore men were receiving aspirin before the index infarction.Advanced age has been previously identified as a predictor of poor outcome following MIand our data confirm this observation (Figure 1). Our data demonstrate that in comparison with men, women with AMI in NRMI-I were substantially older, with 55.7% older than 70 years. However, for similar age and treatment groups, the mortality rate was higher in women (Figure 1). The overall in-hospital mortality rate was higher for women than for men, whether treated with thrombolytic therapy (9.3% vs 4.5% of men) or not (16.0% vs 10.9% of men). This finding agrees with those of other studies that predated the use of thrombolytic agentsand from other data obtained from both the placebo and treatment arms of most large randomized clinical trials.The present study also documents important differences in the cause of death, with women being more likely to die of cardiac rupture. In the group of patients in the NRMI-I not treated with thrombolytic therapy, women (compared with men) were more likely to die of recurrent MI and rupture. While a cause and effect relationship cannot be established in an observational study like this one, women were also treated less often with aspirin, heparin, or β-blockers, especially those who died. The low administration rates of these drugs cannot be ascribed to early death since the mean time to death was 5 days or longer, certainly more than enough time to initiate therapy with heparin or aspirin. The use of heparin (38%), β-blockers (14%), and aspirin (36%) was remarkably low in women who died and did not receive thrombolytic therapy. These data also illustrate that the use of such beneficial concomitant therapy was much lower than that recommended for patients not receiving thrombolytic therapy. The low use of aspirin is particularly noteworthy. The multivariate analysis demonstrates the importance of concomitant therapy but also shows that despite controlling for thrombolytic therapy, age, anterior MI, and use of concomitant therapy, female sex is an independent predictor of poor outcome.These data demonstrate that even when the constraints and guidelines for therapy of rigorous clinical trials are removed, cardiac catheterization and revascularization procedures are used more often in men than in women, with less use of these procedures in those not treated with thrombolytic agents.The NRMI-I database, by virtue of its ongoing data collection and tabulation, allows temporal trends to be examined. Despite the increasing awareness of the value of thrombolytic therapy and the publication of various thrombolytic megatrials, Figure 3demonstrates that there has been no substantial change in the percentage of patients being treated with thrombolytic therapy from 1990 to 1993, and also no substantial change in the mortality rate either in those receiving thrombolytic agents or not. For all treatment modalities, women have a higher mortality rate, although it appears that the number of women being treated with thrombolytic therapy may be increasing. However, these mortality figures are unadjusted for variables such as diabetes and hypertension, which were not recorded on the NRMI-I data collection form. Other studies have reported a higher incidence of hypertension and diabetes in women, which would favor increasing mortality.This study has several limitations, similar to those reported by Rogers et al.Although large, the NRMI-I is an observational database rather than a randomized trial, and it is therefore more valuable for documenting practice patterns and temporal trends than for comparing effectiveness of various treatment interventions. Like many of the recent megatrials in thrombolysis, the data obtained on each NRMI-I patient had no independent validation of data forms, and there exists the potential for underreporting adverse events or for the enrollment of nonconsecutive patients. Also, although encompassing more than 1000 hospitals nationwide, NRMI-I hospitals are not necessarily representative of all US hospitals and likely reflect practice in larger, more procedure-related centers. The modest collection of demographic information in NRMI-I limits data interpretation. Although the percentage of use of aspirin, β-blockers, and other adjunctive therapy is likely to be correct, the presence of contraindications to these agents (that may, in part, explain their low use) was not recorded. These data are also limited in that potentially important confounding variables, such as the prevalence of diabetes, hypertension, tobacco use, and hyperlipidemia, were not recorded and hence could not be studied. Despite these limitations, the NRMI-I data reflect recent practice trends in more than 350000 patients and make compelling observations regarding the limited use of recommended adjunctive therapy. In the GUSTO trial,the use of β-blockers, heparin, and aspirin was standard protocol and the use of β-blockers in appropriate candidates was as high as 46% IV and 71% orally. Intravenous nitroglycerin was used in 77% of all patients. Recently, the GUSTO investigators reported outcomes in women,confirming their higher mortality rate compared with men. However, the authors did not comment on the percentage of use of such adjunctive therapy, which becomes particularly relevant (given the proven benefits of β-blocker use) if morbidity, in addition to mortality, is considered. In this article using data from the NRMI-I database, only in-hospital mortality is recorded and reported. When studying sex differences the GUSTO investigatorsreported 30-day mortality rates, limiting our ability to compare the NRMI-I data with those from the GUSTO trial. Furthermore, 30-day mortality rates may be additionally affected by adjunctive medical therapy. Our data suggest that in clinical practice in the United States, there may be an inappropriately conservative approach to the care of patients with AMI, especially in terms of using adjunctive therapy. This becomes particularly relevant in the case of women with MI, in which the actual use of β-blockers has been less than 50% of that reported in clinical trials. As supported by the multivariate analysis, this conservatism is associated with increased mortality and should be the focus of future studies.CONCLUSIONSThese observations from more than 350000 patients hospitalized in the United States over a 4-year period suggest that important sex differences exist in demographics, treatment, and outcome of patients with AMI. The less frequent use in women of thrombolytic therapy, cardiac catheterization, coronary artery bypass surgery, aspirin, heparin, and β-blockers may, in part, explain their higher mortality rate compared with men.American Heart Association1997 Heart and Stroke Statistical Update.Dallas, Tex: American Heart Association National Center; 1996.DJLernerWBKannelPatterns of coronary heart disease morbidity and mortality in the sexes: a 26-year follow-up of the Framingham population.Am Heart J.1986;111:383-390.MPulettiLSunseriMCurioneSMErbaCBorgiaAcute myocardial infarction: sex-related differences in prognosis.Am Heart J.1984;108:63-66.GHToflerPHStoneJEMullerEffect of sex and race on prognosis after myocardial infarction: adverse prognosis for women, particularly black women.J Am Coll Cardiol.1987;9:473-482.PGreenlandHReicher-ReissUGoldbourtSBeharand the Israeli SPRINT InvestigatorsIn-hospital and 1-year mortality in 1524 women after myocardial infarction.Circulation.1991;83:484-491.Gruppo Italiana per lo Studio della Streptochinasi nell'Infarcto Miocardico (GISSI)Effectiveness of intravenous thrombolytic treatment in acute myocardial infarction.Lancet.1986;1:349-360.ISIS-2 (Second International Study of Infarct Survival) Collaborative GroupRandomised trial of intravenous streptokinase, oral aspirin, both or neither among 17187 cases of suspected acute myocardial infarction: ISIS-2.Lancet.1988;2:349-360.RGWilcoxCGOlssonAMSkeneGvon der LippeGJensenJRHamptonfor the ASSET Study GroupTrial of tissue plasminogen activator for mortality reduction in acute myocardial infarction.Lancet.1988;1:525-530.JWKennedyGVMartinKBDavisThe Western Washington Intravenous Streptokinase in Acute Myocardial Infarction Randomized Trial.Circulation.1988;77:345-352.JWKennedyJLRitchieKBDavidMLStadiusCMaynardJFritzThe Western Washington Randomized Trial of Intracoronary Streptokinase in Acute Myocardial Infarction: a 12-month follow-up report.N Engl J Med.1985;312:1073-1078.WDWeaverHDWhiteRGWilcoxComparisons of characteristics and outcomes among women and men with acute myocardial infarction treated with thrombolytic therapy.JAMA.1996; 275:777-782.HDWhiteGIBarbashMModeanAfter correcting for worse baseline characteristics, women treated with thrombolytic therapy for acute myocardial infarction have the same mortality and morbidity as men except for a higher incidence of hemorrhagic stroke.Circulation.1993;88(pt 1):2097-2103.SBEysmannPSDouglasReperfusion and revascularization strategies for coronary artery disease in women.JAMA.1992;268:1903-1907.RMSteingartMPackerPHammSex differences in the management of coronary artery disease.N Engl J Med.1991;325:226-230.JZAyanianAMEpsteinDifferences in the use of procedures between women and men hospitalized for coronary heart disease.N Engl J Med.1991;325:221-225.CMaynardPELitwinJSMartinWDWeaverGender differences in the treatment and outcome of AMI: results from the MITI registry.Arch Intern Med.1992;152:972-976.CMaynardRAlthouseMCerqueiraMOlsufkaJWKennedyUnderutilization of thrombolytic therapy in eligible women with acute myocardial infarction.Am J Cardiol.1990;68:529-530.JNTobinSWassertheil-SmollerJPWexlerSex bias in considering coronary bypass surgery.Ann Intern Med.1987;107:19-25.SSKhanSNessimRGrayLCzerAChauxJMatloffIncreased mortality of women in coronary artery bypass surgery: evidence for referral bias.Ann Intern Med.1990;112:561-567.DBMarkLKShawERDeLongRMCaliffDBPryorAbsence of sex bias in the referral of patients for cardiac catheterization.N Engl J Med.1994;330:1101-1106.WJRogersLJBowlbyNCChandraTreatment of myocardial infarction in the United States (1990 to 1993).Circulation.1994;90:2103-2114.DEMatthewsVTFarewellA Formal Discussion of Multiple Comparisons in Using and Understanding Medical Statistics.2nd rev ed. Farmington, Conn: S Karger Ag; 1988.ISUdvarhelyiCGatsonisAMEpsteinCLPashosJPNewhouseBJMcNeilAcute myocardial infarction in the Medicare population.JAMA.1992;268:2530-2536.The GUSTO InvestigatorsAn international randomized trial comparing four thrombolytic strategies for acute myocardial infarction.N Engl J Med.1993;329:673-682.Accepted for publication August 20, 1997.Reprints: Nisha C. Chandra, MD, Division of Cardiology, Johns Hopkins Bayview Medical Center, 4940 Eastern Ave, Baltimore, MD 21224 (e-mail: [email protected]).

Langlois, Jean Ann; Visser, Marjolein; Davidovic, Lara S.; Maggi, Stefania; Li, Guohua; Harris, Tamara B.

doi: 10.1001/archinte.158.9.990pmid: 9588432

BackgroundChange in body weight is a potentially modifiable risk factor for hip fracture in older women but, to our knowledge, its relationship to risk in older men has not been reported previously.ObjectiveTo investigate the effects of weight loss and weight gain from age 50 years to old age on the risk of hip fracture among elderly men.MethodsThe association between weight change and risk of hip fracture was studied in a cohort of 2413 community-dwelling white men aged 67 years or older from 3 sites of the Established Populations for Epidemiologic Study of the Elderly.ResultsThe older men in this study, observed for a total of 13620 person-years during the 8 years of follow-up, experienced 72 hip fractures, yielding an overall incidence rate of 5.3 per 1000 person-years. Extreme weight loss (≥10%) beginning at age 50 years was associated in a proportional hazards model with increased risk of hip fracture (relative risk, 1.8; 95% confidence interval, 1.04-3.3). Weight loss of 10% or more was associated with several indicators of poor health, including physical disability, low mental status score, and low physical activity (P<.05). Weight gain of 10% or more beginning at age 50 years provided borderline protection against the risk of hip fracture (relative risk, 0.4; 95% confidence interval, 0.1-1.00).ConclusionsDespite differences between older men and women in the incidence of and risk factors for hip fracture, weight history is also an important determinant of the risk of hip fracture among older men. Weight loss of 10% or more beginning at age 50 years increases the risk of hip fracture in older white men; weight gain of 10% or more decreases the risk of hip fracture. The relationship between extreme weight loss and poor health suggests that weight loss is a marker of frailty that may increase the risk of hip fracture in older men. Physicians should include weight history in their assessment of the risk of hip fracture among older men.HIP FRACTURES in men are projected to account for one third of all hip fracturesworldwide and are associated with higher mortality than in women.Relatively few studies,however, have investigated risk factors for hip fracture in men, especially those aged 65 years and older who are at highest risk of a hip fracture.Because the risk of hip fracture in men is not linked to accelerated bone loss attributable to a distinct menopause equivalent, understanding the reasons for the increase in the risk of hip fracture with aging in men and identifying prevention strategies presents a greater challenge.Changes in body weight occur commonly with aging.Low body weight is a risk factor for hip fracture among men.Weight loss may explain at least part of the risk of hip fracture associated with low weight because weight loss is related both to decreased bone mineral densityand to declines in health and physical functionthat may increase the risk of falls. Among women aged 65 years and older, weight loss is associated with an increased risk of hip fractureand other osteoporotic fractures.Weight loss also is associated with an increased risk of hip fracture in middle-aged men.Because substantially greater weight loss among men occurs between middle age and old age than during middle age,it is important to investigate directly the association between weight loss and the risk of hip fracture among older men. To our knowledge, the relationship between change in body weight between middle and old age and the risk of hip fracture in older men has not been studied.We used data from the Established Populations for Epidemiologic Studies of the Elderly (EPESE) to study the risk of hip fracture associated with weight change between middle (50 years) and old age in white men aged 67 years and older.SUBJECTS AND METHODSData collected about older white men from 3 sites of the EPESE were studied. The EPESE, sponsored by the National Institute on Aging, is a longitudinal study initially designed to identify predictors of mortality, hospitalization, chronic diseases, and disability in community-dwelling elderly individuals. Community surveys were conducted on all people aged 65 years or older living in East Boston, Mass, and Iowa and Washington counties, Iowa. The third population, from New Haven, Conn, was drawn from a stratified random sample defined according to sex, to ensure an adequate number of men, and housing status (public housing for the elderly, private housing for the elderly, or community residence). Details of the study have been described elsewhere.During 1981 through 1983, household interviews were conducted with more than 80% of the more than 10000 eligible study participants. From 1983 through 1989, annual follow-up interviews were conducted. The third and sixth in-person interviews updated the detailed information about health characteristics that were obtained in the initial interview. The remaining brief telephone surveys ascertained health status and health events from the previous year. More than 90% of the people initially interviewed participated in each of the follow-up interviews. The occurrence of hip fracture was ascertained from Medicare hospitalization data for 8 years (1985-1992) beginning with the third follow-up interview. Thus, the third follow-up serves as the baseline for our analyses.STUDY POPULATIONThe study population consisted of 2878 white men aged 67 years and older who participated in the baseline interview. Eliminated from the analyses were 21 men who could not be matched to the Medicare hospitalization files, 82 men who reported previous hip fracture, 347 men with missing body weight data (269 at age 50 years and 78 at the third follow-up), and 15 men with missing height data (used to calculate body mass index [BMI], which is a measure of weight in kilograms divided by the square of the height in meters, at age 50 years). The men excluded because of missing data were significantly older, reported a greater number of medical conditions and greater physical disability, had lower mental status scores, were less physically active, and were less likely to be current smokers or to have consumed alcohol in the past year than those included in the analyses (P<.05). The final sample included 2413 men aged 67 through 104 years (943 from East Boston, 876 from Iowa, and 594 from New Haven). Nonwhite men were excluded from the analyses because there was insufficient statistical power to examine them separately.HIP FRACTURE AND VITAL STATUSMen hospitalized with a hip fracture (principal diagnosis code 820 in the International Classification of Diseases, Ninth Edition, Clinical Modification)occurring between January 1, 1985, and December 31, 1992, were included in the study. Hospitalization data obtained from the Medicare Provider Analyses and Review Files maintained by the Health Care Financing Administration were matched to EPESE data by using identification codes or Social Security numbers. Details of the matching procedure have been described elsewhere.Data from these files are estimated to include 100%, 95.5%, and 86.7% of the hip fracture hospitalizations among individuals aged 65 years or older in Iowa, Connecticut, and Massachusetts, respectively.In the final study sample, the total number of hip fracture cases was 72 (29 in East Boston, 21 in Iowa, and 22 in New Haven).Vital status was ascertained for almost 100% of the population at each study site. Vital status was determined for men from East Boston through the sixth follow-up period, New Haven through the seventh follow-up, and Iowa through 1992 using information collected at annual interviews and by local surveillance procedures, including monitoring obituary notices and review of state vital statistics records. The remaining data on vital status through 1992 were obtained from the National Death Index.WEIGHT CHANGEWeight change was estimated by using the reported weight at baseline and the reported weight at age 50 years. The percentage of weight change was calculated as ([weight at age 50 years (weight at baseline)/weight at 50 years] × 100). Weight change was categorized as a loss of 10% or more, a loss of 5% to less than 10%, a loss or gain of less than 5%, a gain of 5% to less than 10%, and a gain of 10% or more. The percentage of weight change was used because it provides a more meaningful comparison between individuals with widely differing weights and is a more clinically relevant measurement than absolute weight change.COVARIATESBody mass index at age 50 years was calculated based on reported weight at age 50 years and height reported at the initial EPESE interview. Body mass index at age 50 years was used as a continuous variable.Other risk factors assessed at baseline that were used in the analyses were age; number of medical conditions from a self-reported history of having been told by a physician or other health care provider that they had cancer, diabetes, heart attack, or stroke; low mental status score (≥3 errors on the Short Portable Mental Status Questionnaire); physical disability, determined using measures of mobility and activities of daily living (ADL): no disability, mobility disability (report of the need for help with walking 12mile (0.8 km), climbing a flight of stairs, or doing heavy housework, but no limitations in ADL), and ADL disability (report of the need for help with ≥1 ADL that included bathing, dressing, eating, walking across a room, and transferring from bed to chair). Physical activity level was measured using 3 items concerning the frequency of walking, gardening, and doing vigorous exercise.For each activity, subjects were classified into 3 categories: frequently (≥3 times per week), sometimes (twice weekly, weekly, or several times per month), and rarely or never. For the multivariable analysis, a value of 3 was assigned when the activity was performed frequently, 2 when the activity was performed sometimes, and 1 when the activity was performed rarely or never. The scores for each activity were summed to create a total score for physical activity and categorized into 3 groups based on the distribution of scores: low physical activity level (score, 3-5), moderate (score, 6), and high (score, 7-9). Other behavioral variables included cigarette smoking (current, former, or never); use of alcohol (none or any consumed in the past year); and use of thiazide diuretics (current or none). Data about disability, physical activity, smoking, and alcohol were missing for fewer than 2%, mental status score data were missing for 8.0%, and use of thiazide diuretics data were missing for 15.8% of the men in our study.STATISTICAL METHODSThe incidence rates of hip fracture per 1000 person-years were calculated by age. Person-years were calculated by summing the time from baseline until the occurrence of a hip fracture, death, or the end of follow-up, whichever came first. Age-adjusted means and percentages of men with risk factors for hip fracture were compared by hip fracture status and by level of weight change using the SAS general linear models procedure.Adjusted relative risks (RRs) for hip fracture by category of weight change from age 50 years were calculated from a Cox proportional hazards model stratified by community using the STRATA option of the SAS proportional hazards regression procedure to obtain a summary estimate across the communities.Men reporting little change in weight (loss or gain <5%) were used as the reference group. The model controlled for BMI at age 50 years (ie, BMI at the start of the weight change interval) and all other covariates. Statistical interaction between weight change and both age at baseline and BMI at age 50 years was tested by entering the interaction terms separately into the final model.RESULTSCHARACTERISTICS OF THE STUDY POPULATIONThe 2413 men were observed for a total of 13620 person-years during the 8-year follow-up. During the follow-up period, 72 men had a hip fracture, yielding an overall incidence rate of 5.3 per 1000 person-years. The incidence rate of hip fracture was substantially higher among men aged 75 years and older (9.0 per 1000 person-years) than among men aged 65 to 74 years (2.6 per 1000 person-years).Men who had hip fractures were significantly older than those who did not (Table 1). Although the men with and without hip fracture were not significantly different with respect to mean weight at baseline or at age 50 years, those with hip fractures were nearly twice as likely to have been obese (BMI >30) at age 50 years. Men with hip fracture were also nearly twice as likely to have a low mental status score and ADL disability at baseline, and were less likely to have consumed alcohol in the past year. A larger proportion of the men with hip fracture had mobility disability and a low level of physical activity; however, the differences were not statistically significant. The 2 groups also were not significantly different (P>.05) in their reported number of medical conditions, cigarette smoking, and use of thiazide diuretics.Table 1. Age-Adjusted Percentages of White Men Aged 67 to 104 Years With Risk Factors for Hip Fracture According to Hip Fracture Status, Established Populations for Epidemiologic Studies of the Elderly, 1985-1986See table graphicOverall, one third of the older men in our study lost weight and one third gained weight between age 50 years and old age. Extreme weight loss was more common among very old men. A total of 26.5% of men aged 75 years and older reported loss of 10% or more of their body weight since age 50 years, compared with 16.0% of men aged 67 to 74 years (P=.001).Men who lost 10% or more body weight between age 50 years and the study baseline (extreme weight loss) were older and weighed less at baseline than those with little change in weight (Table 2). Men who lost weight reported heavier weights and were nearly 4 times as likely to have been obese (BMI >30) at age 50 years. They also were more likely to report a low level of physical activity. Those who had extreme weight loss were more likely to have 2 or more medical conditions, a low mental status score, and to have mobility and ADL disability; they were less likely to have never smoked or to have consumed alcohol in the past year. Men who gained weight reported heavier weights at baseline and lower weights at age 50 years than those with little change in weight. Those who gained 5% to less than 10% of their weight since age 50 years were more likely to have consumed alcohol and to have reported use of thiazide diuretics at baseline. Men with extreme weight gain (≥10% weight from age 50 years) were more likely to have been underweight at age 50 years (BMI <19), to have a low physical activity level at baseline, and to be former smokers.Table 2. Age-Adjusted Percentages of White Men Aged 67 to 104 Years With Risk Factors for Hip Fracture According to Weight Change After Age 50 Years, Established Populations for Epidemiologic Studies of the Elderly, 1985-1986See table graphicWEIGHT CHANGE AND RISK OF HIP FRACTUREThe older men who lost 10% or more body weight between age 50 years and old age (extreme weight loss) had the highest incidence rate of hip fracture (Table 3). The unadjusted RR of hip fracture among older men with extreme weight loss was 3 times that of both men with lesser weight loss and those with little change in weight. Men who gained 10% or more body weight had the lowest risk of hip fracture.Table 3. Incidence Rates of Hip Fracture by Category of Weight Change Between Age 50 Years and Old Age Among 2413 White Men Aged 67 to 104 Years, Established Populations for Epidemiologic Studies of the Elderly, 1985-1992See table graphicAfter adjustment for other risk factors for hip fracture, including BMI at age 50 years, the older men with a weight loss of 10% or more between age 50 years and the study baseline had a significantly increased risk of hip fracture (RR, 1.85; 95% confidence interval [CI], 1.04-3.31) (Figure 1). Conversely, men with extreme weight gain had a borderline significant decreased risk of hip fracture (RR, 0.38; 95% CI, 0.14-1.00). Men with lesser decreases in weight (5%-<10%) did not have a significantly increased risk of hip fracture (RR, 0.78; 95% CI, 0.33-1.83). Likewise, men with lesser increases in weight did not have a significantly decreased risk of hip fracture (RR, 0.60; 95% CI, 0.24-1.48).Association between change in body weight after the age of 50 years and the risk of hip fracture. The relative risks are adjusted for age at baseline, body mass index (a measure of weight in kilograms divided by the square of the height in meters) at age 50 years, cigarette smoking, alcohol consumption in the past year, number of medical conditions, impaired mobility and disability in activities of daily living, use of thiazide diuretics, physical activity level, and mental status score. Men with a weight loss or gain of less than 5% served as the reference group. The T bars indicate the upper and lower 95% confidence intervals.The interactions of weight change and age at baseline and weight change and BMI at age 50 years were not statistically significant (P>.05). Stratified analysis of the association between weight change and risk of hip fracture by tertiles of BMI at age 50 years revealed an elevated RR only among the men who were thinnest at age 50 years (RR, 2.13; 95% CI, 0.88-5.14; P=.09) although the RR was not statistically significant.COMMENTThe results of this study show that weight loss of 10% or more from age 50 years to old age is associated with a nearly 2-fold increased risk of hip fracture among white men aged 67 years or older. The association of weight loss with several indicators of poor health suggests that weight loss is a marker of frailty that increases the risk of hip fracture among older men. Weight gain of 10% or more is associated with a borderline significant reduction in the risk of hip fracture. The incidence rate of hip fracture in our study, 5.3 per 1000 person-years, was similar to the rates of 4 to 5 per 1000 person-years reported in previous studies.A major strength of our study is the relatively large population of nearly 2500 community-dwelling older men that allowed us to investigate specifically the risk of hip fracture as opposed to all osteoporotic fractures.Our findings of an increased risk of hip fracture with weight loss are consistent with those from a study that included middle-aged Norwegian men.In this study, men who lost more than 3 kg of body weight during an average 11 years of follow-up had a 2.5-fold increased risk of hip fracture compared with a reference group of men who gained between 1.3 and 5.5 kg.Of note, the comparison of men who lost weight with a reference group of men who gained weight resulted in a substantially higher estimate of the RR than would have been obtained if, as in our study, the reference group were composed of men with little change in weight. Although older men have a lower overall risk of hip fracture than older women,we found that the pattern among older men of greater risk of hip fracture associated with greater weight loss, and lesser risk associated with greater weight gain, from age 50 years to old age was comparable with the findings for older women from the same (EPESE) population.This suggests that the overall influence of weight change on risk of hip fracture is similar in both sexes.In both cross-sectionaland longitudinalstudies, weight loss has been shown to be associated with decreased bone mineral density. In a cross-sectional study,weight loss accounted for one third of the decline in bone mineral density with aging among older men.Weight loss may contribute directly to a decline in bone mineral density by decreasing the frequency or intensity of stimuli to bone formation, probably as a result of reduced mechanical loading.The specific changes in soft tissue body composition that accompany weight loss, ie, decreased lean (muscle) and fat mass, may also be important in determining bone mineral density. Heavier individuals carry more weight and have both greater lean body mass and greater fat mass.Weight loss that results in the loss of lean mass may result in decreased strength, decreased physical activity, and reductions in mechanical stress on bone.This may be a relatively more important mechanism among menbecause of their greater muscle massand thus the greater likelihood of muscle loss with weight losscompared with women. Fat mass appears to be of greater importance to the maintenance of bone mineral density among postmenopausal womenthan among men.Among overweight women, weight loss that results in loss of fat may result in decreased bone mineral density through decreased adipose-based production of estrogenor through decreases in the efficiency of absorption and utilization of calcium.Because men do not have a menopause equivalent with related acceleration in bone loss, testosterone or one of its conversion products, estradiol,may be more important to maintenance of bone mineral density in menthan estrogen from adipose tissue; however, the relationship between fat mass and hormonal levelsand the influence of both on maintenance of bone in menis not fully understood. Weight loss that includes loss of fat from around the hip may result in less natural protective padding to reduce the risk of hip fracture in the event of a fall.Weight loss may be associated with reduced dietary intake of nutrientssuch as protein and calcium that are necessary for maintenance of bone mineral density. Protein-energy undernutrition may cause both reduced weight and accelerated bone losssince adequate protein intake is necessary for production of skeletal growth factors. Higher levels of calcium intake and combined calcium and vitamin D supplementationhave been shown to be associated with decreased risk of hip fracture among older men.Perhaps most important in this population of older men, involuntary weight loss may be a marker of increased risk of hip fracture associated with impaired health because underlying illness may be both the cause of weight loss and a risk factor for decreased bone mineral density, falls, and hip fracture. Poor et alfound that medical conditions accounted for nearly three fourths of the hip fractures among the men in their study. Although we did not specifically determine whether weight loss was voluntary or involuntary, the men who lost the most weight in our study had more indicators of poor health than those with little change in weight. These indicators included a greater number of medical conditions and a higher proportion with physical disability, a low mental status score, a low level of physical activity, or current smoking. Even after adjustment for health status, however, weight loss of 10% or more between age 50 years and old age was associated with an increased risk of hip fracture.Diet-induced weight loss also has been shown to result in decreased bone mineral density in a study of healthy older women.Although the need for information on the efficacy of weight loss regimens, especially pharmacotherapeutic interventions, has received increasing attention,little is known about their health effects, including their effects on bone, in older people. Future studies must address this issue. In the interim, weight loss programs for older men should encourage appropriate levels of physical activity and nutrient intake, including calcium, to maintain bone.The finding of a higher percentage of older men who were obese (BMI >30) at age 50 years among those with hip fracture than among those with no hip fracture is intriguing. Greater BMI has been shown to be associated with lower growth hormone secretion and clearance in men,and lower growth hormone secretion may be associated with decreased bone mineral density.Heavier weight in middle age also may be associated with the development of a greater burden of disease in old age, resulting in involuntary weight loss and disability that increase the risk of hip fracture. These hypotheses are speculative and should be investigated in other studies of men with follow-up of body composition and health status from middle through old age.We found that a weight gain of 10% or more from age 50 years to old age was associated with a borderline significant reduction in the risk of hip fracture among the older men in our study. Contrary to weight loss, weight gain may contribute to decreased risk of hip fracture by increasing bone mineral density through increased gravitational force on bone, increased adipose-based production of estrogens, or reduced likelihood of injury in the event of a fall because of increased protective padding around the hip.Men who gained weight also may have had a reduced exposure to the risk of falling associated with a lower physical activity level. Use of thiazide diuretics by some men who gained weight may have contributed to a reduced risk of hip fracture.Of particular interest, the older men who gained 10% or more weight from age 50 years were less likely to be current smokers and more likely to be former smokers than those with little change in weight. In addition to contributing to lower body weight,smoking may increase hip fracture riskthrough its association with decreased bone mineral density.In older men as in older women, the positive effects of smoking cessation, including the avoidance of further weight loss and the slowing or halting of bone loss even among those who quit smoking cigarettes in old agemay decrease the risk of hip fracture. Although heavier weight may contribute to a reduced risk of hip fracture, the negative health effects, including increased risk of a wide range of chronic health conditionsand physical disability,outweigh the potential benefits. To promote overall health and maintenance of bone, physicians and other health care providers should encourage older men to achieve and to maintain an appropriate weight.Some limitations of our study should be considered. First, the older men (16%) excluded from the analysis because of missing weight and/or height data were older and were less healthy than the men included in the analysis. The exclusion of men with poorer health likely resulted in an underestimate of the association between weight change and hip fracture in our study. Second, we were limited to the use of self-reported data from only 2 points in time in selecting a reference group with little change in weight (ie, men whose weight at baseline was only 5% more or less than their weight at age 50 years). The inclusion in our reference group of some men whose weight may have increased after age 50 years and then returned in old age to within 5% of their weight at age 50 years may have resulted in a bias toward the null in the association of weight loss with risk of hip fracture. Third, the body weight data in our study were self-reported. Data from the older men participating in the sixth follow-up of the Iowa EPESE revealed a high correlation between self-reported and measured weights (Pearson r= 0.96). Weight recalled even from many years earlier by elderly men has been found to be relatively accuratealthough men may be more likely to overestimate their past weight.Such a recall bias may have resulted in an overestimate of the weight loss and an underestimate of the weight gain in our study. Future studies should obtain measured weights at frequent intervals to identify a truly weight-stable reference group, and should determine the circumstances related to changes in weight among older men.Weight loss is common among older men and contributes to a substantial increase in the risk of hip fracture. Physicians and other health care providers should obtain a weight history as part of an overall assessment of the risk of hip fracture in older men and should encourage them to achieve and maintain a healthy weight.CCooperGCampionHip fractures in the elderly: a worldwide projection.Osteoporos Int.1992;2:285-289.LJMelton IIIBLRiggsEpidemiology of age-related fractures.In: Avioli LV, ed. The Osteoporotic Syndrome.New York, NY: Grune & Stratton Inc; 1983:45.LForsenABjorndalKBjartveitInteraction between current smoking, leanness, and physical inactivity in the prediction of hip fracture.J Bone Miner Res.1994;9:1671-1678.JAGrissoGYChiuWCSteinmannJPortaleRisk factors for hip fractures in men: a preliminary study.J Bone Miner Res.1991;6:865-868.DHemenwayDRAzraelEBRimmDFeskanichWCWillettRisk factors for hip fracture in US men aged 40 through 75 years.Am J Public Health.1994;84:1843-1845.HEMeyerATverdalJAFalchRisk factors for hip fracture in middle-aged Norwegian women and men.Am J Epidemiol.1993;137:1203-1211.HEMeyerATverdalJAFalchChanges in body weight and incidence of hip fracture among middle aged Norwegians.BMJ.1995;311:91-92.GPoorEJAtkinsonWMO'FallonLJMelton IIIPredictors of hip fractures in elderly men.J Bone Miner Res.1995;10:1900-1907.DFWilliamsonHSKahnPLRemingtonRFAndaThe 10-year incidence of weight gain in US adults.Arch Intern Med.1990;150:665-672.DTFelsonYZhangMTHannanJJAndersonEffects of weight and body mass index on bone mineral density in men and women: the Framingham Study.J Bone Miner Res.1993;8:567-573.JECompstonMALaskeyPICroucherACoxonSKreitzmanEffect of diet-induced weight loss on total body bone mass.Clin Sci.1992;82:429-432.TLHolbrookEBarrett-ConnorThe association of lifetime weight and weight control patterns with bone mineral density in an adult community.Bone Miner.1993;20:141-149.NWGlynnJMZmudaJACauleyWeight change and bone mineral density in older men: a longitudinal study [abstract].J Bone Miner Res.1996;11:S155.JFischerMAJohnsonLow body weight and weight loss in the aged.J Am Diet Assoc.1990;90:1697-1706.JALangloisTBHarrisACLookerJMadansWeight change between age 50 years and old age is associated with hip fracture in white women aged 67 years and older.Arch Intern Med.1996;156:989-994.RGCummingRJKlinebergCase-control study of risk factors for hip fracture in the elderly.Am J Epidemiol.1994;139:493-503.SRCummingsMCNevittWSBrownerRisk factors for hip fracture in white women.N Engl J Med.1995;332:767-773.KEEnsrudJCauleyRLipschutzSRCummingsWeight change and fractures in older women.Arch Intern Med.1997;157:857-863.JCCornoni-HuntleyAMOstfeldJOTaylorRBWallaceEstablished Populations for Epidemiologic Studies of the Elderly: Resource Data Book.Bethesda, Md: National Institute on Aging, National Institutes of Health; 1986:1-10. NIH publication 86-2433.JCCornoni-HuntleyAMOstfeldJOTaylorRBWallaceEstablished populations for epidemiologic studies of the elderly: study design and methodology.Aging (Milano).1993;5:27-37.Not AvailableInternational Classification of Diseases, Ninth Revision, Clinical Modification.Ann Arbor, Mich: Commission on Professional and Hospital Activities; 1986.MPahorJGuralnikMESaliveEAChrischillesAMantoRBWallaceDisability and severe gastrointestinal hemorrhage.J Am Geriatr Soc.1994;42:816-825.ESFisherJABaronDJMalenkaJBarrettTABubolzOvercoming pitfalls in the use of Medicare data for epidemiologic research.Am J Public Health.1990;80:1487-1490.RRoubenoffJJKehayiasThe meaning and measurement of lean body mass.Nutr Rev.1991;49:163-175.EPfeifferA Short Portable Mental Status Questionnaire for the assessment of organic brain deficit in elderly patients.J Am Geriatr Soc.1975;23:433-441.AZLaCroixJMGuralnikLFBerkmanRBWallaceSSatterfieldMaintaining mobility in late life, II: smoking, alcohol consumption, physical activity, and body mass index.Am J Epidemiol.1993;137:858-869.SAS Institute IncSAS/STAT User's Guide, Version 6.4th ed. Cary, NC: SAS Institute Inc; 1989;2:891-1026.SAS Institute IncSAS Technical Report P-217, SAS/STAT Software: The PHREG Procedure, Version 6.Cary, NC: SAS Institute Inc; 1991:1-20.WEBaconGSSmithSPBakerGeographic variation in the occurrence of hip fractures in the elderly white US population.Am J Public Health.1989;79:1556-1558.SJJacobsenJGoldbergTPMilesJABrodyWStiersAARimmHip fracture incidence among the old and very old: a population-based study of 745,435 cases.Am J Public Health.1990;80:871-873.TVNguyenJAEismanPJKellyPNSambrookRisk factors for osteoporotic fractures in elderly men.Am J Epidemiol.1996;144:255-263.HMaySMurphyKKhawAge-associated bone loss in men and women and its relationship to weight.Age Ageing.1994;23:235-240.SLEdelsteinEBarrett-ConnorRelation between body size and bone mineral density in elderly men and women.Am J Epidemiol.1993;138:160-169.GDFosterTAWaddenJLMullenResting energy expenditure, body composition, and excess weight in the obese.Metabolism.1988;37:467-472.NPocockJEismanTGwinnMuscle strength, physical fitness, and weight but not age predict femoral neck bone mass.J Bone Miner Res.1989;4:441-448.RNBaumgartnerPMStauberKMKoehlerLRomeroPJGarryAssociations of fat and muscle masses with bone mineral in elderly men and women.Am J Clin Nutr.1996;63:365-372.WRFronteraVAHughesKJLutzWJEvansA cross-sectional study of muscle strength and mass in 45- to 78-year-old men and women.J Appl Physiol.1991;71:644-650.GBForbesLean body mass-body fat interrelationships in humans.Nutr Rev.1987;45:225-231.IRReidRAmesMCEvansDeterminants of total and body regional bone mineral density in normal postmenopausal women: a key role for fat mass.J Clin Endocrinol Metab.1992;75:45-51.IRReidLDPlankMCEvansFat mass is an important determinant of whole body bone density in premenopausal women but not in men.J Clin Endocrinol Metab.1992;75:779-782.MVisserDPKielJLangloisMuscle mass and fat mass in relation to bone mineral density in very old men and women: the Framingham Heart Study.Appl Radiat Isot.In press.JMGrodinPKSiiteriPCMacDonaldSource of estrogen production in post-menopausal women.J Clin Endocrinol Metab.1973;36:207-214.CHassagerCChristiansenInfluence of soft tissue body composition on bone mass and metabolism.Bone.1989;10:415-419.NHBellSEpsteinAGreeneJSharyMJOexmannSShawEvidence for alteration of the vitamin-D endocrine system in obese subjects.J Clin Invest.1985;76:370-373.CLongcopeJHPrattSEFriebergAromatization of androgens by muscle and adipose tissue in vivo.J Clin Endocrinol Metab.1978;46:146-152.CLongcopeJHPrattSHSchneiderSEFinebergThe in vivo metabolism of androgens by muscle and adipose tissue of normal men.Steroids.1976;28:521-533.JAJacksonMKleerekoperOsteoporosis in men: diagnosis, pathophysiology, and prevention.Medicine.1990;69:137-152.AVermeulenJMKaufmanVAGiagulliInfluence of some biological indexes on sex hormone binding globulin and androgen levels in aging or obese males.J Clin Endocrinol Metab.1996;81:1821-1826.HKKleyTDesekaersHPeerenboomHLKruskemperEnhanced conversion of androstenedione to estrogens in obese males.J Clin Endocrinol Metab.1980;51:1128-1132.WCHayesSJPiazzaPKZyssetBiomechanics of fracture risk prediction of the hip and spine by quantitative computed tomography.Radiol Clin North Am.1991;29:1-18.SLGreenspanERMyersLAMaitlandNMResnickWCHayesFall severity and bone mineral density as risk factors for hip fracture in ambulatory elderly.JAMA.1994;271:128-133.ESOrwollThe effects of dietary protein insufficiency and excess on skeletal health.Bone.1992;13:343-350.BDawson-HughesSSHarrisEKrallGDallalEffect of calcium and vitamin D supplementation on bone density in men and women 65 years of age or older.N Engl J Med.1997;337:670-676.TLHolbrookEBarret-ConnorDLWingardDietary calcium and risk of hip fracture: 14-year prospective population study.Lancet.1988;2:1046-1049.ELauSDonnanDJPBarkerCCooperPhysical activity and calcium intake in fracture of the proximal femur in Hong Kong.BMJ.1988;297:1441-1443.National Task Force on the Prevention and Treatment of ObesityLong-term pharmacotherapy in the management of obesity.JAMA.1996;276:1907-1915.AIranmeshGLizarraldeJDVeldhuisAge and relative adiposity are specific negative determinants of the frequency and amplitude of growth hormone (GH) secretory bursts and the half-life of endogenous GH in healthy men.J Clin Endocrinol Metab.1991;73:1081-1088.DRudmanAGFellerGAGergansEffects of human growth hormone in men over 60 years old.N Engl J Med.1990;323:1-6.DFWilliamsonJMadansRFAndaJCKleinmanGAGiovinoTByersSmoking cessation and severity of weight gain in a national cohort.N Engl J Med.1991;324:739-745.KAHollenbachEBarrett-ConnorSLEdelsteinTHolbrookCigarette smoking and bone mineral density in older men and women.Am J Public Health.1993;83:1265-1270.TVNguyenPJKellyPNSambrookCGilbertNAPocockJAEismanLifestyle factors in the elderly: implications for osteoporosis prevention.J Bone Miner Res.1994;9:1339-1346.National Institutes of Health Consensus Development Panel on the Health Implications of ObesityHealth implications of obesity: National Institutes of Health Consensus Development Conference Statement.Ann Intern Med.1985;103:1073-1977.FXPi-SunyerHealth implications of obesity.Am J Clin Nutr.1991;53:1595S-1603S.JStevensJEKeilLRWaidPCGazesAccuracy of current, 4-year, and 28-year self-reported body weight in an elderly population.Am J Epidemiol.1990;132:1156-1163.AMustWCWillettWHDietzRemote recall of childhood height, weight and body build by elderly subjects.Am J Epidemiol.1993;138:56-64.GSPerryTEByersAHMokdadMKSerdulaDFWilliamsonThe validity of self-reports of past body weights by US adults.Epidemiology.1995;6:61-66.Accepted for publication September 25, 1997.We thank Eric Orwoll, MD, and Cliff Rosen, MD, for their valuable comments on the manuscript.Reprints: Jean A. Langlois, ScD, MPH, Division of Acute Care, Rehabilitation Research and Disability Prevention, National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, 4770 Buford Hwy NE (MS-F-41), Atlanta, GA 30341-3724.

Ravid, Mordchai; Brosh, David; Ravid-Safran, Dorit; Levy, Zohar; Rachmani, Rita

doi: 10.1001/archinte.158.9.998pmid: 9588433

BackgroundThe control of hyperglycemia is of major importance in the treatment of patients with type 1 diabetes mellitus. However, there is no consensus about the required degree of metabolic control in patients with type 2 diabetes mellitus and about the role of hyperglycemia in diabetic nephropathy and in the development of atherosclerosis in relation to other risk factors.Patients and MethodsA prospective, long-term follow-up study was conducted on 574 patients, aged 40 to 60 years, with recent onset of type 2 diabetes mellitus. Patients were initially normotensive and had normal renal function and a normal urinary albumin excretion rate (<30 mg/24 h). The patients were followed up for 2 to 9 years (mean ± SD, 7.8 ± 0.9 years). Levels of hemoglobin A1cand plasma lipids, mean blood pressure, and body mass index (calculated as the weight in kilograms divided by the square of the height in meters) were determined periodically. Cigarette smoking and socioeconomic status were recorded. Renal status was evaluated by the logarithm of the final urinary albumin excretion rate and by the decline in reciprocal creatinine values. Definite clinical events including death, nonfatal myocardial infarction, angina pectoris, congestive heart failure, and peripheral vascular disease were recorded.ResultsAt the end of the study the urinary albumin excretion rate remained normal (<30 mg/24 h) in 373 patients (65%), 111 (19%) had microalbuminuria (30-300 mg/24 h), and 90 (16%) had overt albuminuria (>300 mg/24 h). Logistic regression models demonstrated that the correlation between hemoglobin A1clevels and the risk of albuminuria is exponential. Multiple logistic regression analysis indicated that levels of total cholesterol, mean blood pressure, and hemoglobin A1cwere the main factors associated with the decrease in renal function and with the increase in albuminuria. The combination of values higher than the 50th percentile of all 3 factors defined a high-risk patient population. These high-risk patients had an odds ratio of 43 (95% confidence interval, 25-106) for microalbuminuria and 15 (95% confidence interval, 9-25) for clinical events related to arteriosclerosis compared with the rest of the group. Low levels of high-density lipoprotein, body mass index, cigarette smoking, low socioeconomic status, and male sex were all significantly associated with diabetic nephropathy, as well as with the manifestations of arteriosclerosis.ConclusionsThe combination of blood pressure values in the high-normal range with moderately elevated levels of total cholesterol and hemoglobin A1cdefines a high-risk group for the progression to diabetic nephropathy and for clinical events related to arteriosclerotic cardiovascular disease.DIABETIC nephropathy is a major cause of morbidity and is associated with increased cardiovascular mortality in type 2 diabetes mellitus.The specific pathological changes in the kidney, the clinical course, and the overall risk to develop nephropathy are quite similar in both types of diabetes.Conclusive evidence exists that strict control of hyperglycemia lowers the risk of nephropathy and of other diabetic complications in type 1 diabetes mellitus.The association between the risk of nephropathy and hyperglycemia was found to be nonlinear and continuous, withor withouta threshold value of hemoglobin A1c. However, data regarding this association in type 2 diabetes mellitus are inconclusive and conflicting.The decline in renal function over time has been associated with the initial glomerular filtration rate, initial urinary albumin excretion rate (UAE), hyperglycemia, and age.In a cross-sectional study of patients with type 2 diabetes mellitus, UAE was positively associated with hyperglycemia.The Oklahoma Indian Diabetes Study of patients with type 2 diabetes mellitusalso found that hyperglycemia and elevated blood pressure were associated with an increased risk for renal failure. The results of a prospective study from Japanshowed a reduction in the risk of nephropathy in patients with type 2 diabetes mellitus who were treated with an intensive insulin regimen. However, these patients were lean and sensitive to insulin. Therefore, it is uncertain whether these results could be extrapolated to the typical overweight, insulin-resistant patient with type 2 diabetes mellitus. In a long-term follow-up study of a mixed type 1 and type 2 diabetes mellitus cohort, Hellman and associatesshowed that a comprehensive diabetes treatment program resulted in reduced all-cause mortality. The large United Kingdom Prospective Diabetes Studyhas not yet provided data on this issue. The St Vincent Declaration of 1994states that there is no convincing evidence that strict glucose control helps to slow the progression of nephropathy.The present report summarizes data of a 2- to 9-year follow-up study of a large group of patients with type 2 diabetes mellitus in the Tel Aviv, Israel, area. All patients initially had normal UAE and were normotensive. We examined the influence of glucose control, mean blood pressure, plasma lipid concentration, and other potential risk factors on the development and progression of diabetic nephropathy. We also sought correlations between these parameters and clinical presentations of arteriosclerotic cardiovascular disease.PATIENTS AND METHODSPATIENTSThe UAE was determined during 1986 in more than 750 patients with type 2 diabetes mellitus, diagnosed according to World Health Organization criteria,whose urine was negative for protein using a dipstick test (Labstick, Ames, Iowa).One hundred eight patients who had microalbuminuria were enrolled in a prospective study of the influence of angiotensin-converting enzyme inhibition on diabetic nephropathy.The patients with normal UAE (<30 mg/24 h) comprised the present series.Additional inclusion criteria were as follows: age, 40 to 60 years; diagnosis of diabetes after age 40 years; known duration of diabetes of less than 5 years with a history of initial glucose control by diet or oral agents during the first 6 months; no evidence of systemic, renal, cardiac, or hepatic disease; normal blood pressure values on 2 consecutive examinations (systolic, ≤140 mm Hg and diastolic, ≤90 mm Hg; mean blood pressure, ≤107 mm Hg); serum creatinine level less than 124 µmol/L (1.4 mg/dL); and body mass index (BMI; calculated as the weight in kilograms divided by the square of the height in meters) less than 35 kg/m2. Initially there were 621 eligible patients (men/women, 319/302) with a mean (± SD) age of 47.7 ± 4.5 years. The duration of diabetes was 0 to 5 years (mean ± SD duration, 1.92 ± 1.2 years). The mean (± SD) BMI was 24.9 ± 3.4 kg/m2. Sixty-two patients received insulin, 235 were taking oral hypoglycemic agents, and 324 were using diet to control their diabetes.PROTOCOLInformed consent of the patients to use their data for research was obtained. The patients were followed up by their primary care physicians, who agreed to administer the patients' records in accordance with the protocol.The physicians were advised to try to maintain patients' blood pressures within the normal range. However, there were no interventions in practical therapeutic decisions.The minimum requirements included semiannual determinations of hemoglobin A1c, serum creatinine, UAE, and blood pressure and annual determinations of plasma lipid levels and BMI.Blood pressure was measured with mercury sphygmomanometers with the patients sitting after a 5-minute rest; the average of 2 determinations was recorded. The diastolic pressure was determined at Korotkoff phase V.Antihypertensive therapy with calcium channel blockers and low-dose thiazides was initiated when mean blood pressure values of 105 mm Hg or higher were recorded on at least 2 consecutive visits.Relevant clinical events were recorded, including death (all cause), nonfatal myocardial infarction, angina pectoris, congestive heart failure, and objectively verified peripheral vascular disease. The majority (95%) of the hospitalizations occurred in 2 regional hospitals. The records were reviewed by one of us (M.R.) and the diagnosis verified.The follow-up was concluded in 1995 or earlier if an angiotensin-converting enzyme inhibiting agent was prescribed. The patients who died during the study were included until the time of death. The length of the follow-up ranged from 2 to 9 years (mean ± SD, 7.8 ± 0.9 years).MEASUREMENTSAll laboratory examinations were done centrally and the assays were unchanged during the study period. Creatinine levels were determined using a routine kinetic automated method, as described by Bartels et al.Hemoglobin A1clevels were measured using affinity chromatography (Isolab, Akron, Ohio). The normal range of this assay is 0.035 to 0.056. Based on data from 8 different lot numbers, the coefficient of variation (intra-assay and interassay) was calculated to be less than 3%. Levels of total cholesterol and triglycerides were determined by the colorimetric enzymatic method of Allain et alwith the modifications of Badham and Trinder,using an autoanalyzer (Hitachi 747, Hitachi, Japan). Levels of high-density lipoprotein (HDL) cholesterol were determined using a phosphotungstic acid precipitation step. The UAE was measured using 24-hour urine samples with an automated immunoturbidimetric assay.The intraassay and interassay coefficients of variation of this method are 2.9% and 7.6%, respectively. All together, 8 to 15 collections of urine, 4 to 20 determinations of hemoglobin A1c, 14 to 20 determinations of serum creatinine, and 5 to 10 determinations of plasma lipids were available for each patient. The mean blood pressure values were calculated for each visit (mean pressure was defined as diastolic pressure plus one third of the pulse pressure). The reciprocal creatinine value (100/creatinine value) was calculated for each visitand the decline in renal function was expressed as a percentage of the initial value of the same patient (ΔCr). The socioeconomic status was defined as high or low according to the area of residence. The 50th percentile of the 3 main modifiable baseline characteristics, namely, levels of hemoglobin A1c, total cholesterol, and mean blood pressure, were calculated. Patients in whom all 3 values were at the 50th percentile or higher were termed high risk, while all other patients were considered low risk.STATISTICAL ANALYSISAll data are expressed as mean (± SD). Significance was defined as P<.05. The rate of decrease of reciprocal creatinine as a percentage of the initial value and the increase in the logarithm of albuminuria were calculated using linear regression analysis. The comparison between the groups of patients with normoalbuminuria, microalbuminuria, and macroalbuminuria, as well as those with different mean hemoglobin A1cvalues, was performed using pooled variance Student ttests for independent variables. Stepwise logistic regression analysis was used to determine the correlations between the independent and the dependent variables. The relation between the mean values of hemoglobin A1cand the risk for microalbuminuria was examined using 3 alternative models: simple exponential,a threshold,and a change point model.Survival curves were plotted to compare high- and low-risk patients. Data were stored and processed using computer software (SPSS-PC for Windows, SPSS Inc, Chicago, Ill).RESULTSTen patients discontinued the follow-up and could not be located. The data for 15 patients were incomplete. Twenty-two patients died during the follow-up period. The cause of death was related to coronary heart disease in 13 patients, cerebrovascular accident in 2, malignancy in 2, a motor vehicle crash in 1, and unknown in 4. Thus, the analysis of renal outcome was determined for 574 patients.NEPHROPATHYIn 373 patients (65%), the normoalbuminuria persisted throughout the study (UAE, <30 mg/24 h). Two hundred one patients (35%) developed microalbuminuria, and 90 (16%) of these patients progressed to macroalbuminuria (UAE, >300 mg/24 h). The initial main characteristics of the patients in the 3 groups are outlined in Table 1. The annual mean values of total cholesterol, mean blood pressure, and hemoglobin A1cin these 3 groups during the follow-up period are shown in Figure 1. The patients who progressed to nephropathy had significantly higher initial plasma values of hemoglobin A1c, total cholesterol, low-density lipoprotein cholesterol, and triglycerides and lower values of HDL than those who maintained normal UAE. Also, the initial mean blood pressure was significantly higher in the patients who developed nephropathy. The impact of the individual baseline parameters on the risk for microalbuminuria is outlined in Table 2. A comparison of high- and low-risk patients showed that the odds ratio for microalbuminuria was almost 43 (95% confidence interval [CI], 25-106; P<.001). Thus, these 3 parameters (hemoglobin A1c≥0.09, total cholesterol ≥5.24 mmol/L [203 mg/dL], and mean blood pressure ≥95 mm Hg) when present together define patients at high risk for diabetic nephropathy. Furthermore, the risk (odds ratio) for progression to macroalbuminuria during follow-up was 18 (95% CI, 11-33; P=.001) among high- vs low-risk patients. Univariate analyses showed significant correlations between the decline in renal function (ΔCr as a percentage of the initial value in the same patient) and the mean study values of total cholesterol concentration (r=0.60; P<.001), mean blood pressure (r=0.57; P<.001), and hemoglobin A1c(r=0.46; P<.001). Also, levels of low-density lipoprotein cholesterol, HDL, triglyceride concentration, BMI, cigarette smoking, low socioeconomic class, and male sex were significantly associated with ΔCr. Similar correlations also were found with albuminuria (expressed as the mean of the last 2 values).Table 1. Baseline Characteristics in 574 Patients With Type 2 Diabetes Mellitus and Initially Normal UAE Rate*See table graphicFigure 1.Annual mean values of total cholesterol, mean blood pressure, and hemoglobin A1cduring follow-up in the 3 groups of patients formed by their final urinary albumin excretion rate (I, <30 mg/24 h; II, 30-300 mg/24 h; and III, >300 mg/24 h). BP indicates blood pressure.Table 2. Association of Baseline Parameters With the Risk to Develop Microalbuminuria Among Patients With Type 2 Diabetes Mellitus*See table graphicThe relative impact on the kidney of these variables was examined using multiple logistic regression analysis. Total cholesterol, mean blood pressure, and hemoglobin A1cwere the main determinants of the subsequent decline in renal function. The degree of albuminuria was influenced also by the BMI, HDL values, and age. The risk for microalbuminuria was predicted by the initial values of total cholesterol, mean blood pressure, hemoglobin A1c, HDL, and BMI. The duration to the development of microalbuminuria was mainly determined by the mean blood pressure, BMI, and HDL values.Patients were analyzed according to treatment group at baseline (diet, oral agents, or insulin). When included in the regression models, the mode of treatment itself had no predictive power for subsequent renal outcome. The introduction of insulin treatment during the study was strongly associated with the patient being at high risk at baseline (r=0.53; P=.001) and with the subsequent development of microalbuminuria (odds ratio, 9.55:1; 95% CI, 4.12-23.56; P=.001 for patients receiving insulin vs those not receiving insulin, respectively). However, when the treatment mode was included in the multivariate model this correlation was no longer apparent. The introduction of antihypertensive treatment had no further association with renal outcome when regulated for the average mean blood pressure values.The glycemic control of the patients remained fairly stable throughout the study (Figure 1). Overall there was a gradual increase of less than 0.01 in hemoglobin A1cvalues. There was an inverse correlation between the initial hemoglobin A1cvalues and the increase of those values during follow-up (r=−0.44; P=.001). Table 3shows the mean study hemoglobin A1cvalues by quintiles. There was no consistent pattern of progression with age or duration of diabetes mellitus from the low to high values of hemoglobin A1c. The patients with low mean values of hemoglobin A1calso had lower mean total cholesterol and higher HDL cholesterol values than those with poor glycemic control. They also had a lower BMI and lower mean blood pressure values. The correlation between any pair of these parameters was highly significant: mean blood pressure vs total cholesterol values, r=0.54 (P<.001); mean blood pressure vs hemoglobin A1c, r=0.34 (P<.001); mean blood pressure vs BMI, r=0.50 (P<.001); mean blood pressure vs HDL, r=0.45 (P<.001); total cholesterol vs hemoglobin A1c, r=0.62 (P<.001); BMI vs hemoglobin A1c, r=0.55 (P<.001); and BMI vs total cholesterol, r=0.65 (P<.001).Table 3. Decline in Kidney Function and Final Albuminuria in 5 Subgroups of Patients According to Mean Values of Hemoglobin A1c*See table graphicTable 3demonstrates that more patients developed albuminuria with each 0.01 increase in the hemoglobin A1clevel, from as low as 0.06 in patients with hemoglobin A1cbelow 0.08 to as high as 0.75 in the patients with hemoglobin A1clevels above 0.11. The possibility for a nonlinear relation between the risk of microalbuminuria and hemoglobin A1cvalues was examined by grouping the hemoglobin A1cvalues in small intervals of 0.0045 and 0.009 in the tails. The hemoglobin A1cvalues were modeled with indicator variables in a logistic regression model of the prevalence of microalbuminuria with covariates to adjust for age at onset of diabetes, the duration of diabetes, mean blood pressure, total cholesterol values, and sex. The reference group for the adjusted relative odds was the group of patients with hemoglobin A1cvalues of 0.060 to 0.078. Three logistic regression models were tested,and the results were similar. The exponential model confirmed the impression of nonlinearity (r2=0.73; P=.001) and seemed to represent the data most closely.ARTERIOSCLEROSISSixty-two patients encountered definite clinical events related to arteriosclerosis: 36 had a nonfatal myocardial infarction, 22 developed unequivocal angina pectoris, 4 had congestive heart failure, and 10 developed peripheral vascular disease. These events, grouped with the 22 patients who died, were correlated with the baseline characteristics. Event-free survival was significantly higher in the low-risk patients (P<.001), as shown in Figure 2. The odds ratio for any of the cardiovascular end points (including all-cause mortality) was 14.75 (95% CI, 8.72-24.95; P<.001) among the high-risk vs low-risk patients. The odds ratios for cardiovascular end points according to individual risk factors are detailed in Table 4.Figure 2.Survival curves for the low- and high-risk patients. An event was defined as death, nonfatal myocardial infarction, new-onset angina pectoris, congestive heart failure, or peripheral vascular disease. The numbers below each curve at the end of follow-up indicate the estimates of the cumulative incidence of event-free survival for each group (P<.001 by the log-rank test). High-risk patients were those with a cholesterol concentration of 5.25 mmol/L or higher (≥203 mg/dL), a mean blood pressure of 95 mm Hg or higher, and a hemoglobin A1clevel of 0.09 or higher.Table 4. Association of Baseline Parameters With the Risk to Reach Cardiovascular End Points in Patients With Type 2 Diabetes Mellitus*See table graphicCOMMENTIn this study the risk factors for diabetic nephropathy in type 2 diabetes mellitus are borne out from long-term, noninterventional follow-up of a large and initially uniform group of patients. Logistic regression analysis highlighted the role of glucose control along with the levels of total cholesterol and mean blood pressure as joint major risk factors for the subsequent renal outcome. The importance of blood pressure control in this context is widely accepted and needs no further underscoring.The predictive role of plasma cholesterol concentration in diabetic nephropathy has also been documented in type 1and type 2 diabetes mellitus.However, the importance of lowering blood glucose levels was previously uncertain.Among our patients, elevated hemoglobin A1clevels were associated both with crossing the threshold to nephropathy and the degree of progression of renal impairment. The pattern of the correlation between values of hemoglobin A1cand the risk for albuminuria was exponential but without a definite threshold value, as is most probably the case in type 1 diabetes mellitus.One hundred fifty-five patients had baseline values higher than the 50th percentile of all 3 major risk factors, namely, total cholesterol (>5.25 mmol/L [203 mg/dL]), mean blood pressure (>95 mm Hg), and hemoglobin A1c(>0.09). These patients comprised 27% of the total patient population and their odds ratio for diabetic nephropathy was 43, compared with those who had 2 or fewer risk factors in the 50th percentile or higher. These higher-risk patients also had a 15-to-1 risk for clinical end points of arteriosclerotic cardiovascular disease compared with the rest of the study population. The definition of a high-risk group with only 3 parameters is easy and enables the development of preventive strategies concentrating on these patients. The association between clinical phenomena of arteriosclerosis and these risk factors is time honored. The new aspect highlighted by this study is that these characteristics are all within the normal or previously accepted range for this population. Their combination in a single patient, however, is associated with a very high risk for microvascular and macrovascular complications of diabetes.The inclusion criteria of our study required that diabetes be diagnosed after age 40 years and was initially regulated by diet or oral hypoglycemic agents. Thus, the probability of finding cases of type 1 diabetes mellitus among these patients was low. Other studies that found a correlation between hemoglobin A1cvalues and albuminuria were mostly cross-sectional observations.A recent 6-year follow-up study from Finlandfound that the subsequent development of albuminuria was best predicted by the initial values of serum insulin. The Japanese study by Ohkudo and colleaguesexamined a relatively small group but was well designed and long-term. Their results seem to support the introduction of intensive metabolic control in patients with type 2 diabetes mellitus. However, their patients were relatively young, lean (BMI, 19-21 kg/m2), treated with insulin, and must have been exceptionally cooperative to maintain near normal levels of glycosylated hemoglobin (hemoglobin A1c, 0.06-0.07). Furthermore, the insulin requirements were modest and there was no mention of weight gain during the study period. The applicability of the results of their study to patients with type 2 diabetes mellitus in the western hemisphere is therefore uncertain.The recommendations of the American Diabetes Associationadvocating hemoglobin A1clevels of 0.07 as a therapeutic goal in patients with type 2 diabetes mellitus are based mainly on the extrapolation of the Diabetes Control and Complications Trial.In type 2 diabetes mellitus, however, this goal will require substantially larger doses of insulin with as yet uncertain effects. The recently published feasibility results of the Veterans Affairs Cooperative Study on Glycemic Control and Complications in NIDDMindicate that excellent glycemic control in men with type 2 diabetes mellitus is possible for a limited period. However, the benefit-risk ratio of such intensive therapy has not yet been established. The applicability of these conclusions to women is also uncertain.Our data have several drawbacks. First, this was an uncontrolled observational study; second, the follow-up periods were variable; third, the clinical treatment of the patients was not uniform; and fourth, the method of assessment of renal function is inaccurate. However, the long period of observation, the uniformity of the initial data, the relative stability of mean blood pressure, hemoglobin A1c, and cholesterol values throughout the study, and the use of each patient as his/her own control enable these data to be used in the evaluation of their association with diabetic nephropathy. Furthermore, this study is probably one of the last of its kind. The increasing use of angiotensin-converting enzyme inhibitors in patients with diabetes will obscure the natural course of nephropathy. Also, despite the lack of direct evidence derived from controlled studies, withholding available modalities of intervention for the management of risk factors becomes increasingly difficult from the point of view of medical ethics.Finally, our data indicate that the progression of diabetic nephropathy is truly multifactorial. The list of risk factors includes (in a declining order of significance) elevated levels of plasma total cholesterol, small increments in mean blood pressure, hyperglycemia, high BMI, low levels of HDL, high levels of low-density lipoprotein, cigarette smoking, a low socioeconomic class, and male sex. The similarity of this list to the risk factors for atherosclerosis is striking. Indeed, the patients who were at high risk for microalbuminuria also had a 15-times higher risk for clinical end points of arteriosclerotic cardiovascular disease.CEMogensenMicroalbuminuria predicts clinical proteinuria and early mortality in maturity onset diabetes.N Engl J Med.1984;310:356-360.RKleinBEKKleinSEMossThe incidence of gross proteinuria in people with insulin-dependent diabetes mellitus.Arch Intern Med.1991;151:1344-1348.SFDinneenHCGersheinThe association of microalbuminuria and mortality in non-insulin-dependent diabetes mellitus.Arch Intern Med.1997;157:1413-1418.SMMauerBMChaversA comparison of kidney disease in type I and type II diabetes.In: Vranic M, Hollenberg CH, Steiner G, eds. Comparison of Type I and Type II Diabetes: Similarities in Etiology, Pathogenesis, and Complications.New York, NY: Plenum Press; 1985:299-303. Advances in Experimental Medicine and Biology; Vol 186.CHasslacherERitzPWahlCMichaelSimilar risks of nephropathy in patients of type I or type II diabetes mellitus.Nephrol Dial Transplant.1989;4:859-863.JAPughRMedinaMRamirezComparison of the course to end-stage renal disease of type I (insulin-dependent) and type II (non-insulin-dependent) diabetic nephropathy.Diabetologia.1993;36:1094-1098.The Diabetes Control and Complications Trial Research GroupThe effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus.N Engl J Med.1993;329:977-986.BFeldt-RasmussenERMathiesenTJensenTLauritzenTDeckertEffect of improved metabolic control on loss of kidney function in type I (insulin-dependent) diabetic patients: an update of the Steno Studies.Diabetologia.1991;34:164-170.OBrinchmann-HansenKDahl-JorgensenLSadvikKFHanssenBlood glucose concentrations and progression of diabetic retinopathy: the seven year results of the Oslo Study.BMJ.1992;304:19-22.PReichardBYNilssonMRosenqvistThe effect of long-term intensified insulin treatment on the development of microvascular complications of diabetes mellitus.N Engl J Med.1993;329:304-309.ASKrolewskiLMBLaffelMKrolewskiMQuinnJHWarramGlycosylated hemoglobin and the risk of microalbuminuria in patients with insulin-dependent diabetes mellitus.N Engl J Med.1995;332:1251-1255.The Diabetes Control and Complications Trial Research GroupThe absence of glycemic threshold for the development of long term complications: the perspective of the Diabetes Control and Complications Trial.Diabetes.1966;215:1289-1298.GCVibertiA glycemic threshold for diabetic complications?N Engl J Med.1995;332:1293-1295.SPinho-SilveiroRFriedmanJLGrossGlomerular hyperfiltration in NIDDM patients without overt proteinuria.Diabetes Care.1993;16:115-119.SPinho-SilveiroRFriedmanMJobin de AzeredoLHCanatiJLGrossFive-year prospective study of glomerular filtration rate in normofiltering and hyperfiltering normoalbuminuric NIDDM patients.Diabetes Care.1996;19:171-174.SSuranitiFBledAGiraultPFressinaudMMarreSerum lipids and urinary albumin excretion in non-insulin-dependent diabetics.Mol Cell Biochem.1992;109:197-200.ETLeeVSLeeMLuJSLeeDRusselJYehIncidence of renal failure in NIDDM: the Oklahoma Indian Diabetes Study.Diabetes.1994;4:572-579.YOhkudoHKishikawaEArakiIntensive insulin therapy prevents the progression of diabetic microvascular complications in Japanese patients with non-insulin-dependent diabetes mellitus: a randomized prospective 6-year study.Diabetes Res Clin Pract.1995;28:103-117.RHellmanJReganHRosenEffect of intensive treatment of diabetes on the risk of death or renal failure in NIDDM and IDDM.Diabetes Care.1997;20:258-264.RTurnerCCullRHolmanUnited Kingdom Prospective Diabetes Study 17: a 9-year update of a randomized, controlled trial on the effect of improved metabolic control on complications in non-insulin-dependent diabetes mellitus.Ann Intern Med.1996;124(1 pt 2):136-145.GCVibertiCEMogensenPPassaRBilousRMangiliSt Vincent Declaration 1994: guidelines for the prevention of diabetic renal failure.In: Mogensen CE, ed. The Kidney and Hypertension in Diabetes Mellitus.Boston, Mass: Kluwer Academic Publishers; 1994:522.WHO Expert Committee on Diabetes MellitusSecond Report of the WHO Expert Committee on Diabetes Mellitus.Geneva, Switzerland: World Health Organization; 1980. Technical Reports Series No. 646.MRavidHSavinIJutrinTBentalBKatzMLishnerLong-term stabilizing effect of angiotensin-converting enzyme inhibition on plasma creatinine and on proteinuria in normotensive type II diabetic patients.Ann Intern Med.1993;118:577-581.MRavidRLangRRachmaniMLishnerLong term renoprotective effect of angiotensin-converting enzyme inhibition in non-insulin-dependent diabetes mellitus: a 7-year follow-up study.Arch Intern Med.1996;156:286-289.HBartelsMBohmerCHeierlSerum creatinine determination without protein precipitation.Clin Chim Acta.1972;37:193-197.CCAllainLSPoonCSGChanWRichmondPCFuEnzymatic determination of total serum cholesterol.Clin Chem.1974;10:470-475.DBadhamPTrinderAn improved color reagent for the determination of blood glucose by the oxidose system.Analyst.1972;97:142-145.PSathianathanVPRegeJLBarronImmunoturbidimetric determination of urinary albumin with a centrifugal analyzer.Clin Chem.1986;32:202.NEMitchMNalserGABuffingtonJLemman JrA simple method of estimating progression of chronic renal failure.Lancet.1976;2:1326-1328.Not AvailableSAS/STAT Software: Calis and Logistic Procedures, Release 6.04.Cary, NC: SAS Institute Inc; 1990.KUlmA statistical method for assessing a threshold in epidemiological studies.Stat Med.1991;10:341-349.RHJonesBAMolitorisA statistical method for determining the breakpoint of two lines.Anal Biochem.1984;141:287-290.National High Blood Pressure Education Program Working Group report on hypertension in diabetesHypertension.1994;23:145-158.MRavidHSavinRLangIJutrinSLudvinovskyMLishnerProteinuria, renal impairment, metabolic control and blood pressure in type 2 diabetes mellitus: a 14-year follow-up report on 195 patients.Arch Intern Med.1992;152:1225-1229.HMulecSAJohnsenOWiklundSBjorckCholesterol: a renal risk factor in diabetic nephropathy?Am J Kidney Dis.1993;22:196-201.MRavidLNeumanMLishnerPlasma lipids and the progression of nephropathy in diabetes mellitus type II: effect of ACE inhibitors.Kidney Int.1995;47:907-910.ORWirtaAIPasternackJTMustorenTAKoivulaAHarmoinenUrinary albumin excretion rate and its determinants after six years in non-insulin-dependent diabetic patients.Nephrol Dial Transplant.1996;11:449-456.JAColwellDCCT findings: applicability and implications for NIDDM.Diabetes Rev.1994;2:277-291.JAColwellThe feasibility of intensive insulin management in non-insulin-dependent diabetes mellitus: implications of the Veterans Affairs Cooperative Study on Glycemic Control and Complications in NIDDM.Ann Intern Med.1996;124(1 pt 2):131-135.Accepted for publication September 25, 1997.Supported in part by the Tyomkin Research Grant, provided by Yehudit and Avi Tyomkin, Kfar Shmariahu, Israel (Dr Ravid).Reprints: Mordchai Ravid, MD, Department of Medicine, Meir Hospital, Kfar-Sava 44281, Israel.

White, Richard H.; Zhou, Hong; Romano, Patrick S.

doi: 10.1001/archinte.158.9.1005pmid: 9588434

BackgroundCurrent guidelines suggest that all patients with acute deep venous thrombosis should be treated with intravenous heparin for at least 5 days, overlapping with warfarin sodium for 4 to 5 days.MethodsUsing linked state of California hospital discharge records from 1991 to 1994 we identified patients with acute deep venous thrombosis without pulmonary embolism, and determined the 6-month cumulative incidence of rehospitalization for recurrent thromboembolism. Coding was validated by reviewing the charts of 218 patients matched with the statewide data from 4 local hospitals.ResultsA total of 36924 linked records met study criteria. In the validation group, objectively confirmed thrombosis that was treated with intravenous heparin followed by warfarin was noted in 20%, 65%, 94%, and 95% of the patients who were hospitalized for 2 or fewer days or 3, 4, or 5 or more days, respectively. Statewide, among patients hospitalized for 3, 4, 5, and 6 days, the 6-month cumulative incidence of hospitalization for recurrent thromboembolism was 5.4%, 5.1%, 5.4%, and 6.0%, respectively. Multivariate modeling of patients hospitalized for 3 to 10 days revealed that recurrent thromboembolism was associated with the length of hospitalization (odds ratio [OR], 1.06 each additional day; 95% confidence interval [CI], 1.04-1.08), presence of malignancy (OR, 1.58; 95% CI, 1.46-1.68), age (OR, 0.85 each 10 years; 95% CI, 0.84-0.86), dementia (OR, 0.38; 95% CI, 0.26-0.49), hospitalization for multiple injuries within 3 months (OR, 0.46; 95% CI, 0.32-0.60), and surgery within 3 months (OR, 0.84; 95% CI, 0.78-0.90).ConclusionsWe found no evidence that a stay of 4 days for treatment of deep venous thrombosis was associated with a higher rate of recurrent thromboembolism compared with hospitalization for 5 or more days. Although the evidence was not as strong, the incidence of recurrent thromboembolism after a stay of 3 days appeared comparable with that after a stay of 5 days. These findings suggest that fewer than 5 days of intravenous heparin overlapping with warfarin may provide effective initial treatment for deep venous thrombosis among patients deemed ready for hospital discharge.EVIDENCE-BASED guidelines for the treatment of deep venous thrombosis currently recommend at least 5 days of intravenous heparin overlapping with at least 4 or 5 days of warfarin sodium or until the international normalized ratio (INR) is 2.0 or more.However, in the 1990s health care systems are moving rapidly to reduce health care expenditures, particularly the cost of hospital care. For patients with symptomatic deep venous thrombosis this means shortening the length of hospital stay by instituting oral anticoagulant therapy on the day of admission and discontinuing heparin as soon as a therapeutic INR is achieved. However, there are no published data regarding the efficacy of intravenous heparin therapy for fewer than 5 days.Using a large linked hospital discharge data set, we sought to determine the relationship between the duration of hospitalization for treatment of acute deep venous thrombosis and the incidence of rehospitalization for recurrent venous thromboembolism during the ensuing 6 months. We hypothesized that: (1) the median length of hospitalization for treatment of deep venous thrombosis decreased between 1991 and 1994, and (2) the incidence of recurrent deep venous thrombosis was higher among patients with a hospital stay shorter than 5 days than among those with longer stays (who presumably had a longer duration of intravenous heparin therapy).SUBJECTS AND METHODSDISCHARGE COHORTThis study was approved by the California Health and Welfare Agency Committee for the Protection of Human Subjects and the University of California, Davis, Human Subjects Committee in Sacramento. We used a linked hospital discharge data set compiled by the California Office of Statewide Planning and Development (OSHPD) to determine the incidence and date of recurrent thromboembolic events.All nonfederal licensed hospitals, both acute care and rehabilitation hospitals, are required by the state of California to submit information about each inpatient after hospital discharge that include the admission and discharge date, birth date, race, ZIP code of residence, Social Security number, and sex. In addition, OSHPD requires that hospitals report the principal diagnosis (reason prompting the admission) and up to 24 additional diagnoses, the principal procedure and up to 20 additional procedures, the date of each procedure, and the discharge disposition (eg, home or death). All diagnoses and procedures are coded using the International Classification of Diseases, Ninth Revision, Clinical Modification(ICD-9-CM).Since July 1, 1990, it has been possible to link serial hospital records for nearly all California residents using encrypted Social Security numbers, except for patients using out-of-state or federal hospitals.The OSHPD defines the length of hospitalization as the difference between the admission date and the discharge date. Thus, 1 day of hospitalization means that the patient was discharged the day after admission, which represents, for example, a stay of 1 to 48 hours and 2 days means 25 to 72 hours. Because transfer to another hospital or health care facility would artificially shorten the length of stay at the first hospital, we defined length of hospitalization as the total number of days of contiguous inpatient care. Deaths were determined using the discharge disposition code and a linked death registry.PATIENTSWe included in our cohort only patients admitted to the hospital for acute deep venous thrombosis who were not coded as having pulmonary embolism. We included only patients with a diagnosis related group of 128, 130, or 131 (consistent with a principal diagnosis of venous thrombosis without major surgery), and a principal diagnosis of venous thrombosis defined by ICD-9-CMcodes 451.1x, 451.2, 451.81, 451.9, 453.1, 453.2, 453.8, 453.9, or 997.2 (postoperative complication, if accompanied by a secondary diagnosis of venous thrombosis).We excluded patients who had a secondary diagnosis of pulmonary embolism (ICD-9-CMcode 415.1). The diagnosis related group requirement was designed to exclude patients who underwent major surgery during the index hospitalization, which might have prevented standard anticoagulation treatment. We analyzed the length of hospital stay for all patients hospitalized in 1991 through 1994, but restricted our outcome analysis to patients hospitalized between January 1, 1991, and June 30, 1994, allowing a minimum of 6 months' follow-up. We analyzed only the first hospitalization for venous thrombosis in each linked record and, to minimize the likelihood that patients were being treated for venous thrombosis with warfarin at the cohort's inception, we excluded patients diagnosed (principal or secondary) as having pulmonary embolism or venous thrombosis in the 6 months before the index hospitalization. Finally, we excluded patients treated during the index hospitalization with an inferior vena cava filter (ICD-9-CMcode 38.7).We defined several conditions associated with deep venous thrombosis, including cancer (ICD-9-CMcodes 141 through 208, except code 173, nonmelanoma skin cancers); multiple trauma (2 ICD-9-CMcodes between 800 and 959); lower extremity fracture (codes 820.x-828.0); fracture of the pelvis (code 808.x); and surgery within 3 months (all surgical diagnosis related group codes). Specific codes were used to identify comorbid conditions using the modified Charlson index,a widely used measure of comorbidity.OUTCOMESThe principal outcome measure was the incidence of rehospitalization with a principal diagnosis of deep venous thrombosis or pulmonary embolism during the first 6 months following the index hospitalization. Death within 6 months was also determined.VALIDATION OF CODING AND LENGTH OF STAYWe reviewed charts from 4 local hospitals (1 university, 1 staff model health maintenance organization, and 2 private) to determine the number of patients meeting our entry criteria who had typical deep venous thrombosis. Typical was defined as venous thrombosis that was (1) objectively diagnosed in the deep veins of the upper or lower extremity using compression ultrasonography or venography, (2) treated with intravenous heparin, and (3) treated and discharged on a regimen of warfarin. Patients who were taking warfarin at the time of admission were classified as typical cases if a new venous thrombosis was documented and intravenous heparin was administered. All other cases were deemed atypical. These included patients who did not have objectively defined deep venous thrombosis or were not treated with both heparin and warfarin.The charts of patients discharged between January 1, 1991, and December 31, 1994, with a principal diagnosis of venous thrombosis who were hospitalized for 0 to 5 days and matched with the OSHPD cohort were identified by the medical records departments at 4 hospitals. Specially trained research assistants reviewed the charts of all patients hospitalized for 0 to 5 days, as well as a random sample of patients hospitalized for 6 to 15 days. In addition to documenting the presence of venous thrombosis and the use of heparin and warfarin, we specifically noted (1) a history of venous thrombosis, (2) the use of warfarin at the time of admission, (3) the number of days of warfarin therapy, (4) the INR at the time of hospital discharge, and (5) the number of hours of intravenous heparin therapy.DATA ANALYSIS AND STATISTICSWe restricted our statistical analysis to patients admitted for 3 to 10 days because the majority of patients admitted for 0 to 2 days had atypical venous thrombosis, and fewer than 10% of patients were hospitalized for 11 or more days. We used bivariate analyses and stepwise logistic regression modeling to examine the relationship between recurrent venous thrombosis within 6 months of the index hospitalization and several risk factors. Variables significant at the P<.10 level in bivariate analyses were tested in the stepwise model. Candidate risk factors included the length of hospitalization, age (10-year increments), sex, presence of malignancy, surgery within 3 months, hospitalization for multiple trauma within 3 months, pelvic or lower extremity fracture within 3 months, and a dummy variable for each year to account for any secular trends. In addition, we used several risk factors that are included in a widely used comorbidity measure called the modified Charlson index.This index is based on the presence of 17 different conditions, as defined by ICD-9-CMcodes from the index hospitalization, including disorders such as renal disease, liver disease, malignancy, recent myocardial infarction, heart failure, and hemiplegia. Patients were removed from analysis at the time of death or hospitalization for major surgery or recurrent thromboembolism. Differences in means during 4 years were analyzed using linear regression analysis. Differences in categorical data were analyzed using the χ2test. Data were analyzed using SAS statistical software (SAS Institute Inc, Cary, NC) on a mainframe computer.RESULTSDEEP VENOUS THROMBOSIS COHORTFrom January 1, 1991, through December 31, 1994, there were 36824 linked records that met entry criteria. Clinical characteristics of the study cohort are shown in Table 1. The mean (± SD) length of hospitalization for deep venous thrombosis decreased progressively between 1991 and 1994, from 6.6 ± 2.5 days in 1991 to 5.7 ± 2.3 days in 1994 (P≤.001). The percentage of patients hospitalized for 0 to 2 days was small, increasing from 4.2% in 1991 to 5.6% in 1994. Figure 1shows the increases in the percentages of patients hospitalized for 3 or 4 days and 5 or 6 days during this period and the corresponding decreases in the percentages of patients hospitalized for 7 to 10 days and more than 11 days. Only 1.3% to 2.0% of all patients were transferred directly to another hospital or facility each year.Table 1. Clinical Characteristics of the Study Cohort Stratified by YearSee table graphicFigure 1.Distribution, by year, of the percentage of patients with acute deep venous thrombosis who were hospitalized in California from 1991 to 1994 for 3 or 4 days, 5 or 6 days, 7 to 10 days, and more than 11 days.VALIDATION OF CODING AND LENGTH OF STAYThe charts of 218 patients who were matched with the OSHPD cohort and who were hospitalized at the University of California, Davis Medical Center (n=76), the Kaiser Permanente Medical Center (n=68), Sutter General Hospital (n=35), or Sutter Memorial Hospital, (n=39), all located in Sacramento, were reviewed. Table 2shows the percentage of patients with typical deep venous thrombosis in this cohort, stratified by length of hospitalization. Overall, of 218 patients, 191 (88%) had objectively defined deep venous thrombosis and 181 (83%) met criteria for typical venous thrombosis.Table 2. Validation Cohort of 218 Patients With a Diagnosis of DVT From 4 Area Hospitals in Sacramento, Calif*See table graphicOutcomes were determined for the 32059 patients hospitalized before June 30, 1994. Among the 30560 patients hospitalized for 3 or more days, the cumulative incidence of rehospitalization for recurrent deep venous thrombosis or pulmonary embolism during the first 6 months of follow-up was 1800 (5.9%); 87% of the patients with recurrences had venous thrombosis and 13% had pulmonary embolism. Figure 2shows the incidence of recurrent thromboembolism among patients hospitalized for 3 to 10 days. The 6-month incidence of recurrent thromboembolism was 110 (5.4%) of 2055 patients hospitalized for 3 days, 203 (5.1%) of 3993 patients hospitalized for 4 days, 283 (5.4%) of 5288 patients hospitalized for 5 days, and 335 (6.0%) of 5554 patients hospitalized for 6 days. Comparing hospital stays of 4 and 5 days, the difference in the 6-month incidence of recurrent thromboembolism was 0.27% (95%, −0.64% to 1.18%); there was 80% power to detect a rate of recurrent thromboembolism as high as 6.5% in the 4-day group with α=.05. In addition, the difference in the incidence of recurrent thromboembolism between patients hospitalized for 3 or 4 days (5.2%) and patients hospitalized for 5 or 6 days (5.7%) was not significant (difference=0.53%; 95%, −0.18% to 1.23%).Figure 2.Effect of the length of hospitalization for 3 to 10 days on the 6-month incidence of recurrent deep venous thrombosis or pulmonary embolism.In bivariate analyses, a higher incidence of recurrent thromboembolism was significantly associated with longer length of hospitalization and presence of malignancy, whereas age, dementia, hospitalization within 3 months for multiple injuries, trauma of the lower extremity within 3 months, and any surgery within the previous 3 months were associated with a lower incidence of recurrent thromboembolism. The following measures of comorbidity were not significantly (P>.05) associated with recurrent thromboembolism: renal disease, prior myocardial infarction, peripheral vascular disease, congestive heart failure, chronic obstructive pulmonary disease, diabetes, peptic ulcer disease, liver disease, rheumatic disease, paralysis, varicose veins, and a combined measure called the modified Charlson index. The results of stepwise multivariate logistic regression analysis among patients hospitalized for 3 to 10 days is shown in Table 3. With the exception of trauma of the lower extremity within 3 months, all the measures associated with recurrent thromboembolism in bivariate analyses entered the multivariate model, including longer length of hospitalization.Table 3. Logistic Regression Model of Factors Associated With Recurrent Thromboembolism Using 26165 Patients Hospitalized for 3 to 10 Days in California From 1991 to 1994See table graphicTo explore the possibility that longer length of stay was serving as a proxy for illness severity, we performed logistic regression analysis using death within 6 months of initial hospitalization as the dependent variable, forcing the length of hospitalization in the model, and allowing all measures of comorbidity to enter in a stepwise fashion. In this model, length of stay was significantly associated with death within 6 months (odds ratio, 1.1 for each hospital day; 95% confidence interval, 1.09-1.13), together with 8 measures of comorbidity, such as presence of malignancy or liver disease. This finding suggests that in our analysis of recurrent thromboembolism, length of hospitalization may have served as a measure of comorbidity that we were unable to explain by ICD-9-CMcodes.COMMENTSeveral important issues regarding the treatment of acute deep venous thrombosis remain poorly defined and therefore controversial. One such issue is the optimal duration of intravenous heparin therapy and the number of days that treatment with warfarin should overlap with heparin. There are a number of reasons why fewer than 5 days of heparin together with 4 to 5 days of warfarin might result in a higher incidence of recurrent thromboembolism in the first 6 months following discharge. First, although Hull and coworkersfound no significant difference in the incidence of recurrent thromboembolism after 5 days of heparin therapy compared with 10 days, this small clinical trial of 200 patients had only 15% power to detect a 3% difference in the 6-month incidence of recurrent venous thromboembolism (eg, 7% vs 10%).Thus, a large trial might have shown that 5 days of heparin therapy is less effective compared with 10 days of heparin. Second, 2 studieshave shown that the efficacy of warfarin depends on depletion of clotting factors X and II, both of which have long half-lives compared with factor VII, which is the principal factor that affects the INR.These findings provide the rationale for the recommendation that warfarin therapy be overlapped with heparin for at least 4 to 5 days, as well as the widely held view that an INR of 2.0 or higher during the first 2 or 3 days of warfarin therapy may not provide sufficient anticoagulation to prevent progression of thrombosis.Third, a clinical trialin the Netherlands found that oral anticoagulation therapy alone, without any heparin, was associated with a 3-fold higher 6-month incidence of recurrent venous thrombosis (20%) compared with combined therapy (7%).There is some evidence suggesting that fewer than 5 days of heparin treatment may be as effective as treatment for longer periods, particularly if warfarin treatment is instituted early and an INR of 2.0 or higher is quickly achieved. In a studythat detected microemboli flowing cephalad from acute deep venous thrombi, the emboli could not be detected after 72 hours of heparin therapy. Reduction in factor VII levels, as evidenced by an INR of 2.0 or higher, may be more important in inhibiting coagulation than a reduction in factor II or factor X levels. Studies have shown that the level of factor VII correlates with risk for ischemic heart disease,and factor VII activity correlates with activation of thrombosis, as measured using the activation product of prothrombin, F1+2.In this study we used a large linked data set to determine the relationship between the duration of initial hospitalization for treatment of acute venous thrombosis and the incidence of rehospitalization for recurrent thromboembolism. Use of this linked data set allowed us to identify all subsequent hospitalizations within California, except when hospitalization occurred in a military or veterans hospital. Prior validation studieshave shown high sensitivity and specificity for the coding of principal diagnoses.However, we were concerned about the accuracy of the coding for venous thrombosis among patients admitted for inexplicably short periods, particularly 0 to 3 days.The concern that patients admitted for short periods would have atypical venous thrombosis proved to be warranted. In our validation study, only 5 of 20 patients hospitalized for 1 or 2 days had documented venous thrombosis that was treated with heparin and warfarin. Among 37 patients who were hospitalized for 3 days, 24 (65%) met our criteria for typical venous thrombosis. This percentage was lower than that seen among patients hospitalized for 4 days (94%) or 5 or more days (95%). However, 29 (78%) of 37 of the validation patients hospitalized for 3 days were treated with warfarin and were presumably at risk for recurrent thromboembolism. The lower prevalence of typical venous thrombosis in patients hospitalized for 3 days could have biased the observed 6-month incidence of recurrent thromboembolism if atypical patients had a significantly lower risk of recurrent thromboembolism. The percentage of patients discharged with a therapeutic INR was similar among the patients with typical venous thrombosis who were treated for 3 or 4 days compared with 5 or more days, suggesting that physicians did not generally discharge patients early unless the INR was close to or higher than 2.0.In our analysis of the state cohort we found that the incidence of rehospitalization for recurrent thromboembolism among patients hospitalized for 3 days (5.4%) or 4 days (5.1%) was approximately the same as the incidence among patients hospitalized for 5 days (5.4%). Because the patients in our validation cohort who were hospitalized for 4 days were similar to those who were hospitalized for 5 days (Table 2), our findings suggest that the 6-month incidence of recurrent thromboembolism is no higher for patients hospitalized for 4 days compared with hospitalization for 5 or more days. Assuming that one third of all patients statewide who were hospitalized for 3 days had atypical cases, and that none of these atypical cases developed recurrent thromboembolism, the true cumulative incidence among patients hospitalized for 3 days could have been as high as 8.3%. For the reasons cited above, the rate was likely to be much closer to the observed rate of 5.4%.Overall, we found that there was a statistically significant, but probably clinically unimportant, increase in the 6-month incidence of rehospitalization for recurrent thromboembolism associated with longer duration of hospitalization. This paradoxical finding may be explained by the presence of confounding related to unmeasured illness severity or resistance to oral anticoagulation therapy. The significant association between length of stay and death within 6 months supports a relationship between length of hospitalization and greater comorbidity.The 6-month incidence of recurrent thromboembolism that we observed, 5.9%, is consistent with the findings of other studies, which have reported 8.6%,6.7%,and 5.1%although the exact incidence depends on the prevalence of underlying risk factors for thrombosis as well as the adequacy of anticoagulant treatment.Our lower rate probably reflects some migration out of the state of California, loss of follow-up because of admission to a veterans or military hospital, and elimination of patients who underwent intervening surgery.The predictors of recurrent thromboembolism that we noted in our multivariate model are similar to those found by Prandoni and coworkers,who followed up an inception cohort of 355 patients with venous thrombosis. This group noted that cancer, recent surgery, and recent trauma were associated with recurrent thromboembolism, with relative hazards of 1.72, 0.36, and 0.51, respectively. These values are similar to the odds ratios of 1.58, 0.84, and 0.46 that we noted for cancer, recent surgery, and recent trauma, respectively. It is not surprising that we found an odds ratio closer to 1.0 for recent surgery because we defined surgery based on diagnosis related group codes that included minor operations. Our finding that younger age was associated with greater odds of developing recurrent thrombosis has not been previously reported. This may reflect the higher rate of recurrent thromboembolism that has been noted among patients with idiopathic venous thrombosis, a condition seen more frequently in younger individuals.The close similarity between our findings and those of Prandoni et alindirectly supports the validity of our administrative data set.In conclusion, using a large, linked administrative data set combined with a validation sample, we found that the percentage of patients with deep venous thrombosis who were hospitalized for fewer than 5 days increased significantly between 1991 and 1994. The 6-month cumulative incidence of rehospitalization for recurrent thromboembolism among patients hospitalized for 4 days to treat acute venous thrombosis was no higher than the rate noted among patients hospitalized for 5 days. Although the evidence was not as strong, it appeared that the incidence of recurrent thromboembolism among patients treated for 3 days was not substantially higher than that among patients treated for 5 days. These findings suggest that fewer than 5 days of intravenous heparin with overlapping warfarin therapy may provide effective initial treatment for patients with deep venous thrombosis who are judged to be ready for discharge from the hospital.A randomized controlled trial would be necessary to provide a definitive answer to the question of whether fewer than 5 days of heparin therapy for treatment of acute deep venous thrombosis is safe and effective.TMHyersJGWegAntithrombotic therapy for venous thromboembolism.Chest.1995;108:(suppl 4):335S-351S.EMeuxEncrypting personal identifiers.Health Serv Res.1994;29:247-256.Not AvailableInternational Classification of Diseases, Ninth Revision, Clinical Modification.Washington, DC: Public Health Service, US Dept of Health and Human Services; 1988.FAAnderson JrHBWheelerRJGoldbergA population-based perspective of the hospital incidence and case-fatality rates of deep vein thrombosis and pulmonary embolism: the Worcester DVT Study.Arch Intern Med.1991;151:933-938.PSRomanoLLRoosJGJollisAdapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives.J Clin Epidemiol.1993;46:1075-1079.RDHullGERaskobDRosenbloomHeparin for 5 days as compared with 10 days in the initial treatment of proximal venous thrombosis.N Engl J Med.1990;322:1260-1264.SWesslerSNGitelWarfarin: from bedside to bench.N Engl J Med.1984;311:645-652.AZivelinLVRaoSIRapaportMechanism of the anticoagulant effect of warfarin as evaluated in rabbits by selective depression of individual procoagulant vitamin K-dependent clotting factors.J Clin Invest.1993;92:2131-2140.JHirshVFusterGuide to anticoagulant therapy, II: oral anticoagulants—American Heart Association.Circulation.1994;89:1469-1480.LHarrisonMJohnstonMPMassicotteMCrowtherKMoffatJHirshComparison of 5-mg and 10-mg loading doses in initiation of warfarin therapy.Ann Intern Med.1997;126:133-136.DPBrandjesHHeijboerHRBullerMde RijkHJagtJWten CateAcenocoumarol and heparin compared with acenocoumarol alone in the initial treatment of proximal-vein thrombosis.N Engl J Med.1992;327:1485-1489.SCNichollsJKO'BrianMGSuttonVenous thromboembolism: detection by duplex scanning.J Vasc Surg.1996;23:511-516.GJMillerHCWilkesTWMeadeKABauerSBarzegarRDRosenbergHaemostatic changes that constitute the hypercoagulable state.Lancet.1991;338:1079.YNemersonTissue factor and the initiation of blood coagulation.Adv Exp Med Biol.1987;214:83-94.EFMeuxSAStithAZachReport of Results From the OSHPD Reabstracting Project: An Evaluation of the Reliability of Selected Patient Discharge Data, July Through December 1988.Sacramento, Calif: Office of Statewide Health Planning and Development; 1990.PPrandoniAWLensingACogoThe long-term clinical course of acute deep venous thrombosis.Ann Intern Med.1996;125:1-7.The Columbus InvestigatorsLow molecular-weight heparin in the treatment of patients with venous thrombosis.N Engl J Med.1997;337:657-662.MNLevineJHirshMGentOptimal duration of oral anticoagulant therapy: a randomized trial comparing four weeks with three months of warfarin in patients with proximal deep vein thrombosis.Thromb Haemost.1995;74:606-611.SSchulmanASRhedinPLindmarkerA comparison of six weeks with six months of oral anticoagulant therapy after a first episode of venous thromboembolism: duration of Anticoagulation Trial Study Group.N Engl J Med.1995;332:1661-1665.Accepted for publication September 18, 1997.Reprints: Richard H. White, MD, Division of General Medicine, University of California, Davis, Room 3107, PCC, 2221 Stockton Blvd, Sacramento, CA 95817 (e-mail: [email protected]).

Cady, Roger C.; Ryan, Robert; Jhingran, Priti; O'Quinn, Stephen; Pait, D. Gayla

doi: 10.1001/archinte.158.9.1013pmid: 9588435

ObjectiveTo evaluate the impact of sumatriptan succinate injection compared with placebo on productivity loss during a migraine attack in the workplace.DesignRandomized, double-blind, placebo-controlled, parallel-group clinical trial.SettingFifteen clinical centers in the United States.PatientsOne hundred thirty-five patients 18 years and older diagnosed as having migraine according to International Headache Society criteria.InterventionsPatients self-administered sumatriptan injection (6 mg) or matching placebo to treat a moderate or severe migraine occurring within the first 4 hours of a minimum 8-hour work shift.Main Outcome MeasuresMean productivity loss 2 hours after dosing and across the work shift; percentages of patients returning to normal work performance within 2 hours after dosing and across the work shift; percentages of patients experiencing headache relief (reduction of moderate or severe predose pain to mild or no pain) 1 and 2 hours after dosing.ResultsMean productivity loss was significantly (P≤.002) lower in the sumatriptan group compared with the placebo group both during the 2-hour postdose period (sumatriptan, 39 minutes; placebo, 54 minutes) and across the work shift (sumatriptan, 86 minutes; placebo, 168 minutes). Significantly (P<.001) greater percentages of patients in the sumatriptan group compared with the placebo group returned to normal work performance by 2 hours after dosing (sumatriptan, 52%; placebo, 9%) and across the work shift (sumatriptan, 66%; placebo, 18%). Significantly (P≤.001) greater percentages of patients in the sumatriptan group compared with the placebo group experienced headache relief 1 hour after dosing (sumatriptan, 69%; placebo, 18%) and 2 hours after dosing (sumatriptan, 79%; placebo, 32%).ConclusionSumatriptan reduced migraine-associated productivity loss during a minimum 8-hour work shift by approximately 50% compared with placebo and alleviated headache in more than three fourths of patients.IT IS estimated that more than 10 million individuals in the United States suffer from moderately to severely debilitating migraine,in which patients may be bedridden for days with headache accompanied by symptoms such as nausea, vomiting, and sensitivity to light and sound. Many migraine sufferers are working-age individuals,who incur substantial costs in lost workplace productivity due to absenteeism or to reduced effectiveness during migraine attacks. By one estimate,10 million US migraineurs were bedridden for more than 3 million days per month and experienced 74.2 million restricted activity-days per year in 1989 due to migraine. A 1992 Canadian population surveyshowed that employers annually lost 7 million working days because of patients experiencing migraine attacks. This loss of productive time in the workplace exacts a large economic burden, with employers' annual outlays due to migraine-associated lost workplace productivity estimated at $5.6 billion to $17 billion in the United States.By relieving symptoms and reducing disability, an effective migraine therapy should counteract migraine-associated productivity loss. In fact, the 5-hydroxytryptamine–agonist sumatriptan succinate used for 6 months in each of 2 open-label studieswas associated with 30% to 38% reductions in lost workplace productivity compared with 2 to 4 months of patients' usual (nonsumatriptan) therapy. Although these data suggest that effective treatment of migraine with sumatriptan reduces productivity losses attributed to migraine, the open-label design of these studies does not allow unequivocal attribution of the reduction in lost workplace productivity to sumatriptan.The study described herein is, to our knowledge, the first double-blind, placebo-controlled evaluation of the effects of a migraine medication on workplace productivity loss. This randomized, double-blind, parallel-group study evaluated productivity loss among patients treating 1 migraine attack in the workplace with either sumatriptan injection (6 mg) or matching placebo.PATIENTS AND METHODSPATIENTSMen or women 18 years or older with at least a 1-year history of moderate to severe migraine with or without aura diagnosed according to International Headache Society criteriawere eligible for the study. Patients had to have experienced between 1 and 6 migraine attacks per month and to have treated at least 1 disabling migraine in the workplace in the past 60 days. In addition, patients had to be working 8-hour (minimum) shifts at their jobs. Patients fulfilling any of the following criteria were excluded from the study: confirmed or suspected ischemic heart disease or Prinzmetal angina, uncontrolled hypertension (systolic pressure of ≥140 mm Hg, diastolic pressure of ≥90 mm Hg), Raynaud syndrome, basilar or hemiplegic migraine, pregnancy, lactation, previous use of sumatriptan (any formulation), or treatment with monoamine oxidase inhibitors within 2 weeks before screening. All patients provided written informed consent before participating in the study.PROCEDURESThe protocol for this double-blind, placebo-controlled, single-attack, parallel-group study was approved by an institutional review board for the 15 US study sites. During a screening visit, patients underwent physical examinations and reviews of medical and migraine histories. Patients were given instructions about the use of diary cards for recording productivity loss and efficacy assessments and the use of an autoinjector for self-administering study medication.Patients treated the first migraine occurring after the screening visit with their usual (nonsumatriptan) medication and used a practice diary to record efficacy and productivity assessments. The practice diaries were reviewed with patients when they returned to the clinic for the randomization visit. Patients were randomized 1:1 during this visit to use sumatriptan injection (6 mg) or matching placebo in the workplace to treat a moderate or severe migraine occurring within the first 4 hours of a minimum 8-hour work shift.Patients were instructed not to take ergotamine-containing medications or sumatriptan within 24 hours before or after they administered study medication, and analgesics, antiemetics, or other acute migraine medications within 6 hours before they administered study medication. Patients could take rescue medication (with the exception of ergotamine-containing medications or sumatriptan) for intolerable pain beginning 2 hours after dosing with study medication. Patients not using rescue medication and experiencing headache recurrence (defined as return of moderate or severe pain, where moderate or severe predose pain had been reduced to mild or none at 2 hours after initial dosing) in the workplace could use a second, identical dose of study medication.Within 14 days of using study medication to treat a migraine in the workplace, patients returned to the clinic for the exit visit, during which they returned their completed diaries and were queried about the occurrence of adverse events (defined as any untoward medical occurrence regardless of its suspected relationship to administration of study medication). Patient diaries were reviewed to ensure that they were complete with no discrepancies.PHARMACOECONOMIC ASSESSMENTSPatients used diary cards to record (1) time missed from work because of migraine symptoms (defined as any time patients temporarily or permanently left the work area for any migraine-related reason beginning with the first administration of study medication through the end of the work shift); (2) time worked with migraine symptoms (recorded on an hourly basis beginning with the first dose of study medication); (3) percent effectiveness while working with migraine symptoms (recorded immediately before administration of the first dose of study medication and hourly thereafter for the remainder of the work shift); and (4) the date and time at which patients returned to patient-defined normal work performance.The primary pharmacoeconomic end point was mean productivity loss 2 hours after dosing with study medication; secondary pharmacoeconomic end points included mean productivity loss across the work shift, the percentage of patients returning to normal work performance by 2 hours after dosing and across the work shift, and the median time to return to normal work performance.Last observations were carried forward to account for rescue medication use and any missing observations for the hourly assessments, but not for the normal work performance assessment. If a patient used rescue medication before returning to normal work performance, the patient was considered not to have returned to normal work performance. Productivity loss was computed using the following formula:P=Σi[Ti× (100−Peff/100)] + H,where i is either 1 or 2, denoting the hourly assessments at 1 and 2 hours after the first injection (for calculating productivity loss 2 hours after the first dose of study medication), or the hourly assessments from 1 hour after the first injection until the end of the work shift (for calculating productivity loss across the work shift); Tiindicates the time worked with migraine symptoms at the ith hour after treatment; H, either the total time missed from work because of migraine symptoms in the 2 hours after the first injection (for calculating productivity loss 2 hours after the first dose of study medication) or until the end of the work shift (for calculating productivity loss across the work shift); and Peff, the average of the percent effectiveness assessments recorded at the ith hour and (i−1)st hour after treatment. Differences between the sumatriptan and placebo groups in productivity loss at 2 hours after treatment and across the work shift were tested using an analysis of variance (controlling for treatment, center, and treatment-by-center interaction). Differences between the sumatriptan and placebo groups in percentages of patients returning to normal work performance within 2 hours after dosing and across the work shift were tested using the Cochran-Mantel-Haenszel test controlling for center.EFFICACY ASSESSMENTSPatients used a 4-point scale (0, no pain; 1, mild pain; 2, moderate pain; 3, severe pain) to record migraine pain severity on diary cards immediately before dosing and hourly hereafter for the remainder of the work shift. Patients also recorded any use of rescue medication as well as the time to patient-defined meaningful relief, measured with a stopwatch started at dosing and stopped when relief became meaningful to the patient, after the first dose and second dose (if taken for headache recurrence) of study medication.Efficacy end points included the percentages of patients experiencing headache relief (defined as reduction of moderate or severe pain predose to mild or no pain), complete relief (defined as reduction of moderate or severe pain predose to no pain), and meaningful relief (patient defined) after the first dose of study medication as well as the percentages of patients using rescue medication. Differences between the sumatriptan and placebo groups in the percentages of patients experiencing headache relief and complete relief 1 and 2 hours after dosing, in the percentages of patients experiencing meaningful relief by 2 hours after dosing, and in the percentages of patients using rescue medication from initial dosing through the end of the work shift were tested using the Cochran-Mantel-Haenszel test, controlling for center. The difference between the sumatriptan and placebo groups in the median time to meaningful relief was tested using the log-rank test. For analyses of headache relief and complete relief, the last nonmissing assessment was carried forward to account for any missing assessments. Patients were considered not to have experienced relief, complete relief, or meaningful relief if they used rescue medication before the assessment. Descriptive statistics only were computed for the percentages of patients experiencing headache recurrence and meaningful relief of headache recurrence within 24 hours of the second dose of study medication.SAFETY ASSESSMENTSThe primary safety end point was the percentage of patients reporting adverse events. The percentages of patients reporting adverse events were tabulated for each treatment group. For each adverse event occurring in more than 5% of sumatriptan-treated patients, differences between the sumatriptan group and the placebo group were tested using the Fisher exact test.RESULTSPATIENTSOne hundred thirty-five patients enrolled in the study and were included in the safety analyses; 132 of the 135 completed the study and were included in the pharmacoeconomic and efficacy analyses. Three patients, each randomized to placebo, withdrew prematurely from the study. One patient failed to return to the clinic and 2 did not use treatment in accordance with the study protocol.Demographics, clinical characteristics, and categories of primary occupation were similar between the sumatriptan group and the placebo group (Table 1). The 2 most common occupations were administrative support and professional specialty (Table 1). Concomitant medications were used by all 135 patients; categories of concomitant medications used by at least 20% of patients in either treatment group are listed in Table 1. The mean time between initial dosing with study medication and the scheduled end of the work shift was similar (P<.05) between the sumatriptan group (6.3 hours; SD, 1.9) and the placebo group (6.8 hours; SD, 1.9).Table 1. Demographics, Clinical Characteristics, and Categories of Primary Occupation Among Patients Using Sumatriptan Succinate Injection or Placebo to Treat a Migraine in the Workplace*See table graphicEFFECTS OF SUMATRIPTAN VS PLACEBO ON PRODUCTIVITY IN THE WORKPLACEMean Productivity LossMean productivity loss was significantly (P≤.002) lower in sumatriptan-treated patients compared with placebo-treated patients both 2 hours after dosing and across the work shift (Figure 1). Considered separately, each of the components contributing to productivity loss (reduced effectiveness while working with symptoms and missing work because of migraine symptoms) was lower in sumatriptan-treated patients compared with placebo-treated patients. Mean (SD) time lost because of reduced effectiveness while working with symptoms was 55.2 (57.2) minutes in the sumatriptan group compared with 108.8 (82.1) minutes in the placebo group. Mean (SD) time lost due to missing work because of migraine symptoms was 31.3 (71.2) minutes in the sumatriptan group compared with 69.3 (119.3) minutes in the placebo group.Figure 1.Mean productivity loss 2 hours after dosing and across an 8-hour work shift in patients treating a migraine in the workplace with either sumatriptan succinate injection (6 mg) or matching placebo. Asterisk indicates P≤.002 vs placebo.Return to Normal Work PerformanceSignificantly (P<.001) greater percentages of sumatriptan-treated patients compared with placebo-treated patients returned to normal work performance both within 2 hours of dosing and across the work shift (Figure 2). Similar results were obtained (sumatriptan vs placebo, P<.001) when the cumulative percentage of patients returning to normal work performance over the work shift was determined post hoc as a supplemental means of examining these data (Figure 3). (The end-of-work shift values for the cumulative distribution graph [Figure 3] are slightly different than those for the bar graph [Figure 2] because patients' time to return to normal work performance was set in the cumulative distribution graph to the time to the end of the scheduled work shift for patients not returning to normal work performance.)Figure 2.Percentage of patients returning to normal work performance 2 hours after dosing and across an 8-hour work shift after treating a migraine in the workplace with either sumatriptan succinate injection (6 mg) or matching placebo. Asterisk indicates P<.001 vs placebo.Figure 3.Cumulative proportion of patients returning to normal work performance across an 8-hour work shift after treating a migraine in the workplace with either sumatriptan succinate injection (6 mg) or matching placebo. (Patients' time to return to normal work performance has been set to the time to the end of the scheduled work shift for patients not returning to normal work performance.)The median time to return to normal work performance was 120 minutes in the sumatriptan group. Because fewer than 50% of placebo-treated patients returned to normal work performance, the median time to return to normal work performance could not be defined in the placebo group.CLINICAL EFFICACY OF SUMATRIPTAN VS PLACEBOHeadache ReliefPredosing headache pain was scored as moderate or severe for all patients in both treatment groups. Significantly (P≤.002) greater percentages of sumatriptan-treated patients compared with placebo-treated patients experienced headache relief and complete relief both 1 hour and 2 hours after dosing. Similarly, significantly (P<.001) greater percentages of sumatriptan-treated patients compared with placebo-treated patients experienced meaningful relief within 2 hours (sumatriptan, 51 [76] of 67; placebo, 21 [32] of 65) and across the work shift (sumatriptan, 57 [85] of 67; placebo, 26 [40] of 65). The median time to meaningful relief was 40 minutes in the sumatriptan group (P<.001). Because fewer than 50% of placebo-treated patients experienced meaningful relief, the median time to meaningful relief could not be defined in the placebo group.Rescue Medication UseA significantly (P<.001) smaller percentage of sumatriptan-treated patients (5 [7] of 67) compared with placebo-treated patients (20 [31] of 65) used rescue medication.Headache RecurrenceA smaller percentage of sumatriptan-treated patients (15%) compared with placebo-treated patients (33%) experienced headache recurrence during the work shift. Among patients using a second dose of study medication to treat recurrence (n=8, sumatriptan; n=7, placebo), 6 (75%) of 8 sumatriptan-treated patients compared with 3 (43%) of 7 placebo-treated patients reported meaningful relief of recurrence.SAFETY AND TOLERABILITY OF SUMATRIPTAN VS PLACEBONo patient withdrew from the study because of an adverse event. Thirty-five (52%) of 67 patients in the sumatriptan group and 14 (21%) of 68 patients in the placebo group experienced an adverse event. Adverse events experienced by more than 5% of patients in the sumatriptan group are depicted in Table 2. The most frequently reported adverse events among sumatriptan-treated patients were warm or hot sensations and nausea and vomiting. Among the adverse events experienced by more than 5% of sumatriptan-treated patients (Table 2), the only adverse events that were significantly (P<.05) more common in the sumatriptan group compared with placebo were warm or hot sensation and pressure sensation. (The adverse event "pressure sensation" could include pressure reported in any portion of the body except the chest. Pressure in the chest was reported as a chest symptom.)Table 2. Patients Experiencing Adverse Events*See table graphicCOMMENTThe results of this study provide the first randomized, double-blind, placebo-controlled demonstration that sumatriptan injection reduces lost workplace productivity among migraineurs. The results indicate that sumatriptan injection offered a clear advantage compared with placebo in reducing workplace productivity loss attributable to migraine. On average, lost workplace productivity 2 hours after dosing for sumatriptan-treated patients was only 39 minutes compared with 54 minutes for placebo-treated patients. Across the entire work shift, sumatriptan-treated patients lost less than 1.5 hours due to migraine, whereas placebo-treated patients lost nearly 3 hours. By the end of the work shift, two thirds of sumatriptan-treated patients compared with less than one fifth of placebo-treated patients had returned to normal work performance.Sumatriptan injection was more effective than placebo at reducing both components of overall productivity loss—time lost due to reduced effectiveness while working with migraine symptoms and time lost due to missing work because of migraine symptoms. Of the 2 components, mean time lost due to reduced effectiveness at work was higher than mean time lost because of missing work due to migraine symptoms for both the sumatriptan group and the placebo group. This finding highlights the importance of measuring reduced effectiveness in the workplace as well as time missed from work in studies evaluating the effects of disease and therapy on workplace productivity.After sumatriptan injection, patients quickly returned to normal work performance. More than 50% of patients treated with sumatriptan injection compared with 9% of placebo-treated patients indicated that they had returned to normal work performance by 2 hours after dosing—the earliest sampled time point in this study. This rapid effect is consistent with the time between dosing and onset of relief: approximately 70% of patients in the sumatriptan group experienced reduction of moderate or severe pain predose to mild or no pain by 1 hour after dosing, and the median time to meaningful relief among sumatriptan-treated patients was 40 minutes. (The median time to meaningful relief was "undefined" for the placebo-treated patients, because fewer than 50% of them reported that they achieved meaningful relief.) For a migraine medication taken in the workplace, speed of action is important in realizing effects on lost workplace productivity. The shorter the time between dosing and onset of relief, the greater the likelihood that the medication will impact workplace productivity loss during the shift on which the medication is taken. Similarly, the more promptly after onset of symptoms that a migraine medication is used in the workplace, the greater the likelihood that the medication will affect workplace productivity during the shift on which the medication is taken. Although any beneficial effects of a migraine medication that does not become effective until after the end of the work shift will be realized by the patient, the impact of the medication in terms of reduction of lost workplace productivity may not accrue to the employer.The reduction in lost workplace productivity among sumatriptan-treated patients in this study is attributable to the clinical efficacy of sumatriptan. Consistent with data from previous clinical trials,approximately 70% and 80% of sumatriptan-treated patients (compared with 18% and 32% of placebo-treated patients) experienced headache relief 1 hour and 2 hours after dosing, respectively. In a post hoc analysis, the strength of relationship between productivity measures and clinical efficacy measures in this study was assessed using Spearman ρ. Spearman ρ for productivity loss 2 hours after dosing and the mean of pain scores at baseline and 1 and 2 hours after dosing was 0.613 (P<.001). Thus, positive clinical response was associated with favorable workplace performance in both sumatriptan- and placebo-treated patients.Complementing these efficacy data, sumatriptan's side-effect profile in this study did not interfere with patients' return to activity. The most commonly reported adverse event during treatment with sumatriptan injection was warm or hot sensation. The adverse event data are consistent with previous studies of sumatriptan injection.The data from this placebo-controlled study are consistent with the results of surveys in which patients were asked to recall workplace-related effects of sumatriptan administered across a number of migraine attacks.For example,83% of 160 patients in a health maintenance organization reported that they missed fewer days from work during the 6 months after sumatriptan was added to the formulary. Similarly, patients using open-label sumatriptan injection for up to 24 months reported missing fewer work days during sumatriptan therapy (mean, 1.4 days in the past 4 weeks) compared with baseline (presumatriptan) (mean, 2.5 days in the past 4 weeks).The data from this placebo-controlled study also corroborate the results of 2 open-label studiesthat used lost workplace productivity measures similar to those in the present trial. Mushet and colleaguesfound that lost workplace productivity was 38% lower over a 6-month period in which 43 patients used sumatriptan injection (6 mg) to treat migraines compared with the preceding 12- to 18-week period during which patients used their usual (nonsumatriptan) therapy. The decrease in lost workplace productivity with sumatriptan therapy was attributed to reductions in time missed from work due to symptoms and time worked with symptoms. Similar results were obtained in a studyof 220 nurses using sumatriptan tablets (100 mg) to treat migraines for 6 months after their usual (nonsumatriptan) therapy had been used for 2 months. Lost workplace productivity was 30% lower during sumatriptan therapy compared with usual therapy. The total annual cost of migraine to employers for this sample of nurses was estimated in a post hoc analysis to be $244634 with usual therapy and $116625 with sumatriptan.Considered together, the data from the open-label studies and the present placebo-controlled trial consistently demonstrate that sumatriptan reduces productivity loss due to migraine. These data warrant careful consideration by the clinician, the health care administrator, and the employer who seek to reduce the humanistic and economic costs of migraine.WFStewartRBLiptonDDCelentanoPrevalence of migraine headache in the United States: relation to age, income, race, and other sociodemographic factors.JAMA.1992;267:64-69.PEStangJTOsterhausImpact of migraine in the United States: data from the National Health Interview Survey.Headache.1993;33:29-35.WPryse-PhillipsHFindlayPTugwellJEdmeadsTJMurrayRFNelsonA Canadian population survey on the clinical, epidemiologic, and societal impact of migraine and tension-type headache.Can J Neurol Sci.1992;19:333-339.JTOsterhausDLGuttermanJRPlachetkaHealthcare resource use and lost labor costs of migraine headache in the United States.Pharmacoeconomics.1992;2:67-76.GRMushetDWMillerBClementsGPaitDLGuttermanImpact of sumatriptan on workplace productivity, nonwork activities, and health-related quality of life among hospital employees with migraine.Headache.1996;36:137-143.JRAdelmanMSharfmanRJohnsonImpact of oral sumatriptan on workplace productivity, health-related quality of life, healthcare use, and patient satisfaction with medication in nurses with migraine.Am J Manage Care.1996;2:1245-1254.Headache Classification Committee of the International Headache SocietyClassification and diagnostic criteria for headache disorders, cranial neuralgias and facial pain.Cephalalgia.1988;8(suppl 7):1-98.MDrummondCost-of-illness studies: a major headache!Pharmacoeconomics.1992;2:5-7.The Subcutaneous Sumatriptan International Study GroupTreatment of migraine attacks with sumatriptan.N Engl J Med.1991;325:316-321.RKCadyJKWendtJFKirchnerJDSargentJFRothrockHSkaggsTreatment of acute migraine with subcutaneous sumatriptan.JAMA.1991;265:2831-2835.JSchoenenJBulckeJCaekebekeSelf-treatment of acute migraine with subcutaneous sumatriptan using an auto-injector device: comparison with customary treatment in an open, longitudinal study.Cephalalgia.1994;14:55-63.The Sumatriptan Auto-Injector Study GroupSelf-treatment of acute migraine with subcutaneous sumatriptan using an auto-injector device.Eur Neurol.1991;31:323-331.DLGreinerSNAddySumatriptan use in a large group-model health maintenance organization.Am J Health Syst Pharm.1996;53:633-638.PJhingranRKCadyJRubinoDWMillerRGriceDLGuttermanImprovements in health-related quality of life with sumatriptan treatment for migraine.J Fam Pract.1996;42:36-42.Accepted for publication September 24, 1997.This work was sponsored by Glaxo Wellcome Research Institute, Research Triangle Park, NC.Presented at a meeting of the American Academy of Neurology, Boston, Mass, April 15, 1997, and at meetings of the International Headache Congress, Amsterdam, the Netherlands, June 14, 1997, and American Association for the Study of Headache, New York, NY, June 22, 1997.The following investigators are gratefully acknowledged for their participation in the clinical trial described in this article: James Adelman, MD, Greensboro, NC; John Cochran, MD, Alexandria, Va; Gregory Collins, MD, Charlotte, NC; W. Travis Ellison, MD, Greer, SC; Michael Finkel, MD, Eau Claire, Wis; Thomas Klein, MD, Wichita, Kan; Mark Linden, MD, Boca Raton, Fla; Frank Maggiacomo, DO, Providence, RI; Ninan Mathew, MD, Houston, Tex; Naynesh Patel, MD, Centerville, Ohio; John A. H. Porter, Milwaukee, Wis; John Rubino, MD, Raleigh, NC; and Joel Saper, MD, Ann Arbor, Mich.Reprints: Roger C. Cady, MD, Headache Care Center, 1230 E Kingsley, Springfield, MO 65804 (e-mail: [email protected]).