Research JAMA Pediatrics | Original Investigation Association of Oral Corticosteroid Bursts With Severe Adverse Events in Children Tsung-Chieh Yao, MD, PhD; Jiu-Yao Wang, MD, DPhil; Sheng-Mao Chang, PhD; Yen-Chen Chang, MS; Yi-Fen Tsai, MS; Ann Chen Wu, MD, MPH; Jing-Long Huang, MD; Hui-Ju Tsai, MPH, PhD Supplemental content IMPORTANCE The adverse effects from the long-term use of oral corticosteroids are known, CME Quiz at but, to our knowledge, few studies have reported the risk of corticosteroid bursts, particularly jamacmelookup.com and CME among children. Questions page 755 OBJECTIVE To quantify the associations of corticosteroid bursts with severe adverse events, including gastrointestinal (GI) bleeding, sepsis, pneumonia, and glaucoma, in children. DESIGN, SETTING, AND PARTICIPANTS This study used data derived from the National Health Insurance Research Database in Taiwan from January 1, 2013, to December 31, 2017, on children younger than 18 years of age and used a self-controlled case series design. Data were analyzed from January 1 to July 30, 2020. EXPOSURE Oral corticosteroid bursts (defined as oral corticosteroid use for14 days). MAIN OUTCOMES AND MEASURES Incidence rates were calculated of 4 severe adverse events (GI bleeding, sepsis, pneumonia, and glaucoma) in children who did or did not receive corticosteroid bursts. Conditional fixed-effect Poisson regression was used to estimate incidence rate ratios (IRRs) of severe adverse events within 5 to 30 days and 31 to 90 days after initiation of corticosteroid bursts. RESULTS Among 4 542 623 children, 23% (1 064 587; 544 268 boys [51.1%]; mean [SD] age, 9.7 [5.8] years) were prescribed a single corticosteroid burst. The most common indications were acute respiratory tract infections and allergic diseases. The incidence rate differences per 1000 person-years between children administered a single corticosteroid burst and those not prescribed corticosteroids were 0.60 (95% CI, 0.55-0.64) for GI bleeding, 0.03 (95% CI, 0.02-0.05) for sepsis, 9.35 (95% CI, 9.19-9.51) for pneumonia, and 0.01 (95% CI, 0.01-0.03) for glaucoma. The IRRs within 5 to 30 days after initiating corticosteroid bursts were 1.41 (95% CI, 1.27-1.57) for GI bleeding, 2.02 (95% CI, 1.55-2.64) for sepsis, 2.19 (95% CI, 2.13-2.25) for pneumonia, and 0.98 (95% CI, 0.85-1.13) for glaucoma; the IRRs within the subsequent 31 to 90 days were 1.10 (95% CI, 1.02-1.19) for GI bleeding, 1.08 (95% CI, 0.88-1.32) for sepsis, 1.09 (95% CI, 1.07-1.11) for pneumonia, and 0.95 (95% CI, 0.85-1.06) for glaucoma. CONCLUSIONS AND RELEVANCE This study suggests that corticosteroid bursts, which are commonly prescribed for children with respiratory and allergic conditions, are associated with a 1.4- to 2.2-fold increased risk of GI bleeding, sepsis, and pneumonia within the first month after initiation of corticosteroid therapy that is attenuated during the subsequent 31 to 90 days. Author Affiliations: Author affiliations are listed at the end of this article. Corresponding Authors: Tsung-Chieh Yao, MD, PhD, Division of Allergy, Asthma, and Rheumatology, Department of Pediatrics, Chang Gung Memorial Hospital, 5 Fu-Hsin Street, Kweishan, Taoyuan 33305, Taiwan (yao@adm. cgmh.org.tw); Hui-Ju Tsai, MPH, PhD, Institute of Population Health Sciences, National Health Research Institutes, 35 Keyan Rd, Zhunan, JAMA Pediatr. 2021;175(7):723-729. doi:10.1001/jamapediatrics.2021.0433 Miaoli County 35053, Taiwan (tsaihj@ Published online April 19, 2021. Corrected on July 6, 2021. nhri.edu.tw). (Reprinted) 723 Research Original Investigation Association of Oral Corticosteroid Bursts With Severe Adverse Events in Children ral corticosteroids are the bedrock of treatment for sev- eral inflammatory diseases, such as rheumatoid ar- Key Points O thritis, inflammatory bowel disease, and asthma, as Question Are there potential harms associated with oral 1,2 recommended by international guidelines. It has been well corticosteroid bursts (defined as the use of oral corticosteroids for recognized for more than a half century that long-term use of 14 or fewer days) in children? oral corticosteroids is associated with subsequent adverse Findings In this nationwide population-based study of 1 064 587 events, including Cushingoid features, gastrointestinal (GI) children who received a single corticosteroid burst, a burst was bleeding, infections, glaucoma, hyperglycemia, cardiovascu- associated with 1.4- to 2.2-fold increased risk of gastrointestinal 3-8 lar diseases, and osteoporosis. Clinicians therefore caution bleeding, sepsis, and pneumonia within the first month after against long-term use of oral corticosteroids unless the poten- corticosteroid initiation. tial benefits outweigh the potential risks. Meaning This study suggests that clinicians should be aware of To our knowledge, scant data are available about the po- potentially severe adverse events associated with corticosteroid tential harms of corticosteroid bursts, which are defined as bursts in children. 9-19 courses of oral corticosteroids for 14 or fewer days. Now- adays, use of corticosteroid bursts are considered harmless, an approximately 23 million individuals covered in the National assumption supported by years of clinical data linking expo- sure duration with toxic effects. Clinicians currently pre- Health Insurance Program (NHIP) in Taiwan. Approximately 99% of the Taiwanese population has been registered and cov- scribe short courses of oral corticosteroids to 21% of the gen- eral adult population in the US and up to 17% of the general ered by the NHIP. In this study, we used the deidentified medi- cal claims records and prescription data from the entire NHIRD adult population in France. Corticosteroid bursts are typi- cally prescribed for treating non–life-threatening conditions, from January 1, 2013, to December 31, 2017. The institutional review board of the National Health Research Institutes, Tai- such as upper respiratory tract infections, bronchitis, rashes, 10,22 and low-back pain. A population-based study by Waljee wan, approved this study protocol, and informed consent was waived because all data were encrypted. et al showed increased rates of adverse events, including sep- sis, venous thromboembolism, and fracture, among adults in the US who were treated with oral corticosteroids for fewer than Study Design and Populations 30 days. A recent longitudinal analysis of 15 million adults in In this study, we undertook a self-controlled case series to quantify the risks of 4 severe adverse events, GI bleeding, Taiwan by Yao and colleagues is the first, to our knowledge, to report potential harms of corticosteroid bursts by using a sepsis, pneumonia, and glaucoma, after initiation of a corti- costeroid burst. In a self-controlled case series, each partici- self-controlled case series design. Yao et al demonstrated in- creased risks of GI bleeding, sepsis, and heart failure in a gen- pant serves as his or her own control, given unmeasured time-invariant variables automatically controlled for in the eral adult population receiving corticosteroid bursts. How- succeeding analysis. The risks of each severe adverse event ever, to our knowledge, data regarding the potential harms of short-term oral corticosteroids in children remain limited. within the pretreatment period (the reference period defined as 5-90 days prior to initiation of a corticosteroid To address this knowledge gap, we used a self-controlled case series design and conducted a nationwide population- burst) were compared with the risks within each of 2 post- treatment periods (5-30 days and 31-90 days after initiation based study in Taiwan to evaluate the association of cortico- steroid bursts in children with 4 adverse events available in of a corticosteroid burst) among participants who received a single corticosteroid burst (Figure 1). We excluded partici- our database, GI bleeding, sepsis, pneumonia, and glaucoma. pants who received more than 1 corticosteroid burst during the observation period. We used a conservative approach by including a 4-day washout period. As such, the severe Methods adverse events that occurred during a 4-day window both before and after corticosteroid use were dismissed because Data Source the severe adverse events observed among those partici- The National Health Insurance Research Database (NHIRD) comprises medical claims records and prescription data from pants might be due to other factors. Figure 1. Graphic Presentation of Self-controlled Case Series Design Baseline period: –5 to –90 d Risk period 2: 31 to 90 d Risk period 1: 5 to 30 d Washout period: ±4 d Start of observation Start of a corticosteroid burst End of observation –90 90 Time, d The observation periods are the baseline period and the 2 risk periods. 724 JAMA Pediatrics July 2021 Volume 175, Number 7 (Reprinted) jamapediatrics.com Association of Oral Corticosteroid Bursts With Severe Adverse Events in Children Original Investigation Research Participants who were enrolled in the NHIP 1 year prior to tions, we performed sensitivity analyses to examine (1) the in- the study period and during the entire study period were in- clusion of participants with prescriptions of topical cortico- cluded. Exclusion criteria were (1) 18 years of age or older in steroids prior to the study period and (2) the different durations 2013;(2)prescriptionofsystemicortopicalcorticosteroidsprior of observation periods, with 180 days as the maximum post- to2013;(3)diagnosisofGIbleeding,sepsis,pneumonia,orglau- exposure time (reference period defined as 5-180 days prior coma prior to 2013; (4) more than 1 corticosteroid burst ad- to initiation of a corticosteroid burst and 2 posttreatment pe- ministered during the observation period; (5) continuous oral riods defined as 5-60 days and 61-180 days after initiation of corticosteroid prescription for more than 14 days; and (6) con- a corticosteroid burst). We further used E-values to evaluate genital anomalies or catastrophic illnesses. the association of potential unmeasured confounding. Sub- group analyses were performed to investigate the number of Exposure and Study Outcomes days of corticosteroid bursts by classifying participants into 2 Data on the exposure to corticosteroid bursts were obtained groups: those who received corticosteroid bursts for less than from the NHIRD. We summed all successive corticosteroid pre- 7 days vs those who received corticosteroid bursts for 7 days scriptiondayssincethefirstcorticosteroidprescriptionthrough or more. All analyses were performed using SAS, version 9.2 all prescription records in the posttreatment period as “cumu- (SAS Institute Inc) and R package, version 3.6.3 (R Group for lative use days” and identified corticosteroid bursts as con- Statistical Computing). tinuous use of oral corticosteroids for 14 days or less. To ascertain standardized doses, we converted the investigated corticosteroids into a daily dose based on prednisone equiva- Results lent doses (eTable 1 in the Supplement). 3-8 Previous studies have reported the adverse effects of Baseline Characteristics of the Study Participants long-term corticosteroid use on the GI, immune, and ophthal- The total number of study participants younger than 18 years mologic systems; however, to our knowledge, it remains was 4 542 623. Among those, 1 897 858 (42%) received at least unknown whether corticosteroid bursts are associated with 1 corticosteroid burst during the 5-year study period. In this adverse effects on these systems, especially in children. Thus, study, 1 064 587 participants (23%; 544 268 boys [51.1%] and we chose 4 severe adverse events (GI bleeding, sepsis, pneu- 520 319 girls [48.9%]; mean [SD] age, 9.7 [5.8] years) who re- monia, and glaucoma) as the outcomes of interest in this study. ceived a single corticosteroid burst were included; and 91% had Episodes of syncope were treated as a negative control out- a Charlson Comorbidity Index score of 0. Table 1 shows the come. Gastrointestinal bleeding, sepsis, pneumonia, glau- baseline characteristics of the study participants who re- coma, and syncope were defined based on International ceived a single corticosteroid burst or 1 or more corticoste- Classification of Diseases, Ninth Revision, Clinical Modifica- roid bursts during the observational period. Table 1 suggests tion codes for encounters between 2013 and 2015 and Inter- comparable baseline characteristics between these 2 cohorts national Statistical Classification of Diseases and Related Health and shows the most common indications for use of cortico- Problems, Tenth Revision, Clinical Modification codes for 2016 steroid bursts: acute respiratory tract infections (acute upper and 2017 (eTable 2 in the Supplement). respiratory infections [10.2% vs 10.5%], acute bronchitis and bronchiolitis [9.1% vs 10.1%], acute sinusitis [5.6% vs 6.0%], acute tonsillitis [3.4% vs 3.3%], acute laryngitis and tracheitis Covariates The complete list of time-varying covariates included the top [3.1% vs 3.0%], and acute nasopharyngitis [2.9% vs 3.2%]) and 10 diagnosed acute conditions and concomitant medication allergic diseases (urticaria [11.9% vs 11.0%], contact dermati- use for the severe adverse events (eg, nonsteroidal anti- tis and eczema [10.3% vs 9.1%], asthma [5.2% vs 7.1%], and al- inflammatory drugs [NSAIDs] and proton pump inhibitors for lergic rhinitis [3.4% vs 3.2%]). These indications accounted for GI bleeding, NSAIDs and systemic immunosuppressive agents 65% of all reasons that corticosteroid bursts were prescribed for sepsis and pneumonia, and NSAIDs for glaucoma). for participants who received a single corticosteroid burst. The top 5 physician specialties associated with the prescriptions Statistical Analysis of corticosteroid bursts were pediatrics, dermatology, otolar- Data were analyzed from January 1 to July 30, 2020. We com- yngology, family practice, and internal medicine, accounting puted incidence rates per 1000 person-years of the 4 severe for 93% of corticosteroid bursts prescribed to participants who adverse events for participants prescribed corticosteroid bursts received a single corticosteroid burst. and participants not prescribed corticosteroids. We calcu- lated incidence rate ratios (IRRs) by comparing the incidence Incidence Rates of 4 Adverse Events rates of the severe adverse events within each posttreatment The incidence rates per 1000 person-years of the 4 severe period with the incidence rates of the severe adverse events adverse events (GI bleeding, sepsis, pneumonia, and glau- within the reference period. We performed the analyses using coma) for participants prescribed a single corticosteroid burst conditional fixed-effect Poisson regression. For each severe ad- and for participants not prescribed corticosteroids are pre- verse event and negative control event, stepwise selection was sented in Table 2. The incidence rates per 1000 person-years applied to determine the corresponding list of time-varying co- of the 4 severe adverse events among participants adminis- variates adjusted in the analytical models (eMethods in the tered a single corticosteroid burst were greater than those Supplement). To assess the robustness of observed associa- among participants not prescribed corticosteroids. The jamapediatrics.com (Reprinted) JAMA Pediatrics July 2021 Volume 175, Number 7 725 Research Original Investigation Association of Oral Corticosteroid Bursts With Severe Adverse Events in Children Table 1. Characteristics of Children With Corticosteroid Bursts Children, No. (%) 1 Corticosteroid burst All corticosteroid burst(s) Characteristic (n = 1 064 587) (n = 1 897 858) Age, mean (SD), y 9.7 (5.8) 9.5 (5.7) Female 520 319 (48.9) 897 193 (47.3) Corticosteroid use Daily dose, median (IQR), mg/d 6.00 (1.50-15.00) 8.17 (1.25-15.00) Duration, median (IQR), d 3.00 (3.00-3.00) 3.00 (3.00-3.00) Incidence rate per 1000 person-years (95% CI) GI bleeding 2.48 (2.44-2.52) 2.54 (2.51-2.57) Sepsis 0.37 (0.35-0.39) 0.41 (3.96-4.22) Pneumonia 25.74 (25.59-25.88) 27.86 (27.75-27.98) Glaucoma 0.62 (0.60-0.64) 0.65 (0.64-0.65) Diagnosis of the top 10 acute conditions (ICD-9-CM and ICD-10-CM codes) Urticaria (708.xx and L50.xxxx) 126 290 (11.9) 20 8957 (11.0) Contact dermatitis and other eczema (692.xx, L23.xxxx, 109 113 (10.3) 172 276 (9.1) L24.xxxx, and L25.xxxx) Acute Upper respiratory tract infections (465.xx and J06.xxxx) 108 082 (10.2) 198 685 (10.5) Bronchitis and bronchiolitis (466.xx, J20.xxxx, 96 514 (9.1) 191 826 (10.1) and J21.xxxx) Sinusitis (461.xx and J01.xxxx) 59 329 (5.6) 113 428 (6.0) Asthma (493.xx and J45.xxxx) 55 064 (5.2) 13 4658 (7.1) Allergic rhinitis (477.xx and J30.xxxx) 35 879 (3.4) 60 884 (3.2) Acute Tonsillitis (463.xx and J03.xxxx) 35 683 (3.4) 6640 (3.3) Abbreviations: GI, gastrointestinal; Laryngitis and tracheitis (464.xx, J04.xxxx, 32 541 (3.1) 56 086 (3.0) ICD-9-CM, International Classification and J05.xxxx) of Diseases, Ninth Revision, Clinical Nasopharyngitis (460.xx and J00.xxxx) 30 847 (2.9) 60 384 (3.2) Modification; ICD-10-CM, International Statistical Classification Physician specialty of Diseases, Tenth Revision, Clinical Pediatrics 283 996 (26.7) 558 859 (29.5) Modification; IQR, interquartile range. Dermatology 276 286 (26.0) 434 936 (22.9) ICD-9-CM codes were used to define Otolaryngology 189 174 (17.8) 336 791 (17.8) the conditions in 2013 to 2015, and ICD-10-CM codes were used to Family practice 159 178 (15.0) 277 690 (14.6) define the conditions in 2016 and Internal medicine 78 516 (7.4) 135 956 (7.2) 2017 (eTable 2 in the Supplement). Table 2. Incidence Rates of Gastrointestinal Bleeding, Sepsis, Pneumonia, and Glaucoma in Children With or Without Corticosteroids Corticosteroid bursts No corticosteroids Incidence rate per Incidence rate per Rate difference No. of No. of 1000 person-years No. of No. of 1000 person-years per 1000 person-years Adverse event cases person-years (95% CI) cases person-years (95% CI) (95% CI) Gastrointestinal 13 078 5 273 004 2.48 (2.44-2.52) 31 466 16 706 990 1.88 (1.86-1.90) 0.60 (0.55-0.64) bleeding Sepsis 1966 5 306 732 0.37 (0.35-0.39) 5628 16 785 555 0.34 (0.33-0.34) 0.03 (0.02-0.05) Pneumonia 121 143 4 706 896 25.74 (25.59-25.88) 250 122 15 261 762 16.39 (16.32-16.45) 9.35 (9.19-9.51) Glaucoma 3279 5 303 115 0.62 (0.60-0.64) 10 200 16 771 943 0.61 (0.60-0.62) 0.01 (0.01-0.03) incidence rate differences per 1000 person-years between the than the reference period. The IRR for sepsis in the first post- 2 groups were 0.60 (95% CI, 0.55-0.64) for GI bleeding, 0.03 treatment period was significantly greater than the reference (95% CI, 0.02-0.05) for sepsis, 9.35 (95% CI, 9.19-9.51) for pneu- period, but not in the second posttreatment period. During the monia, and 0.01 (95% CI, 0.01-0.03) for glaucoma (Table 2). first posttreatment period, the IRR was 1.41 (95% CI, 1.27- 1.57) for GI bleeding, 2.02 (95% CI, 1.55-2.64) for sepsis, 2.19 (95% CI, 2.13-2.25) for pneumonia, and 0.98 (95% CI, 0.85- IRRs From Self-controlled Case Series Analysis Figure 2 shows that the IRRs for GI bleeding and pneumonia 1.13) for glaucoma. During the second posttreatment period, across 2 posttreatment periods (5-30 days and 31-90 days the IRR was 1.10 (95% CI, 1.02-1.19) for GI bleeding, 1.08 (95% after initiating corticosteroid bursts) among participants who CI, 0.88-1.32) for sepsis, 1.09 (95% CI, 1.07-1.11) for pneumo- received a single corticosteroid burst were significantly higher nia, and 0.95 (95% CI, 0.85-1.06) for glaucoma. The results in 726 JAMA Pediatrics July 2021 Volume 175, Number 7 (Reprinted) jamapediatrics.com Association of Oral Corticosteroid Bursts With Severe Adverse Events in Children Original Investigation Research Figure 2. Association Between Exposure to Corticosteroid Bursts and Figure 3. Association Between Exposure to Corticosteroid Bursts and Gastrointestinal Bleeding, Sepsis, Pneumonia, and Glaucoma in Children Gastrointestinal Bleeding, Sepsis, Pneumonia, and Glaucoma in Children Based on Alternative Inclusion and Exclusion Criteria and Different No Durations of Observation Periods increased Increased Period, d IRR (95% CI) risk risk No Gastrointestinal increased Increased bleeding Period, d IRR (95% CI) risk risk 5-30 1.41 (1.27-1.57) Alternative inclusion and exclusion criteria 31-90 1.10 (1.02-1.19) Gastrointestinal Sepsis bleeding 5-30 2.02 (1.55-2.64) 5-30 1.45 (1.32-1.60) 31-90 1.08 (0.88-1.32) 31-90 1.12 (1.04-1.21) Pneumonia Sepsis 5-30 2.19 (2.13-2.25) 5-30 1.82 (1.46-2.27) 31-90 1.09 (1.07-1.11) 31-90 1.23 (1.04-1.44) Glaucoma Pneumonia 5-30 0.98 (0.85-1.13) 5-30 2.16 (2.09-2.23) 31-90 0.95 (0.85-1.06) 31-90 1.23 (1.20-1.27) 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.4 2.8 Glaucoma IRR (95% CI) 5-30 1.05 (0.93-1.19) 31-90 0.94 (0.85-1.04) Incidence rate ratios (IRRs) and corresponding 95% CIs for 4 severe adverse Different durations of observation events in 2 posttreatment periods (5-30 days and 31-90 days after initiation of a periods corticosteroid burst). Gastrointestinal bleeding eTable 3 in the Supplement reveal no association of cortico- 5-60 1.36 (1.26-1.48) 61-180 1.27 (1.20-1.34) steroid bursts with the risk of syncope, the negative control Sepsis outcome. 5-60 1.78 (1.46-2.18) 61-180 1.04 (0.89-1.22) Sensitivity Analyses Pneumonia Sensitivity analyses were performed to investigate different 5-60 2.19 (2.15-2.24) inclusion and exclusion criteria and different durations of ob- 61-180 1.20 (1.18-1.22) servational periods. The results in Figure 3 were comparable Glaucoma to those in Figure 2, indicating the robustness of the ob- 5-60 0.94 (0.85-1.06) served associations. We further calculated E-values to evalu- 61-180 1.01 (0.93-1.09) ate unmeasured confounding for the IRRs reported for the 4 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.4 2.8 severe adverse events. The E-values ranging from 2.17 to 3.80 IRR (95% CI) for the point estimate of GI bleeding, sepsis, and pneumonia IRR indicates incidence risk ratio. within the first month after corticosteroid initiation sug- With inclusion of participants with prescriptions of topical corticosteroids prior gested no substantial unmeasured confounding (eTable 4 in to the study period. the Supplement). With 180 days as the maximum postexposure time (reference period defined as 5-180 days prior to initiation of a corticosteroid burst and 2 posttreatment Subgroup Analysis periods defined as 5-60 days and 61-180 days after initiation of a corticosteroid burst). The results of subgroup analysis in 2 groups are comparable to those reported in the whole study, although some results were not statistically significant in the group of children who used corticosteroids for 7 days or more, probably owing to the This study demonstrates the potential harms of prescribing cor- decreased sample size (eTable 5 in the Supplement). ticosteroid bursts to children and calls for the prudent use of corticosteroid bursts. To our knowledge, this is the first and only nationwide, longitudinal, population-based study quantifying the asso- Discussion ciation of corticosteroid bursts with risks of severe adverse In this nationwide population-based study of more than 4 mil- events in children. Using a US health care database, Waljee lionchildren,42%wereexposedtoatleast1corticosteroidburst et al reported increased risks of sepsis, venous thromboem- during the 5-year study period. Corticosteroid bursts were typi- bolism, and fracture among adults receiving short-term oral cally prescribed for children with acute respiratory tract in- corticosteroids for fewer than 30 days. Using Taiwan’s NHIRD, fections (34%) and allergic diseases (31%). Corticosteroid bursts Yaoetal indicated that corticosteroid bursts in adults are as- were significantly associated with a 1.4- to 2.2-fold increase sociated with increased risks of GI bleeding, sepsis, and heart of GI bleeding, sepsis, and pneumonia, but not glaucoma, failure. Our study extends the risks of severe adverse events within the first month after initiation of corticosteroid therapy. associated with corticosteroid bursts from adults to children jamapediatrics.com (Reprinted) JAMA Pediatrics July 2021 Volume 175, Number 7 727 Research Original Investigation Association of Oral Corticosteroid Bursts With Severe Adverse Events in Children and provides supportive evidence that treatment with corti- a self-controlled case series design, which is robust to control costeroid bursts is associated with increased risk of GI bleed- for time-invariant risk factors. The E-values for GI bleeding, ing, sepsis, and pneumonia within the first month after ini- sepsis, and pneumonia suggest that it is very unlikely that tiation of corticosteroid therapy for children. unmeasured confounding can explain the observed associa- The findings have several clinical implications. First, sep- tion of corticosteroid bursts with these severe adverse events. sis is a rare but potentially life-threatening event. Despite the Second, previous studies report that the adverse effects of cor- small observed incidence rate difference in sepsis between chil- ticosteroids include the GI, immune, and ophthalmologic sys- dren with and children without prescriptions of corticoste- tems. Our study assessed the association of corticosteroid roid bursts, corticosteroid bursts were associated with a 2-fold bursts with 4 severe adverse events, GI bleeding, sepsis, pneu- increased risk of sepsis during the first month after starting monia, and glaucoma, among children in Taiwan. Further stud- treatment. Particular caution is therefore needed when ad- ies are needed to assess the validity of these findings and other ministering corticosteroid bursts to children. Second, this study corticosteroid-associated adverse events in other pediatric populations. Third, medication noncompliance is a potential provides evidence that corticosteroid bursts are not innocu- ous but may pose potentially serious health risks, such as GI concern for studies based on registry data. However, noncom- bleeding, sepsis, and pneumonia, to children. Clinicians pre- pliance is independent of subsequent severe adverse events scribing corticosteroid bursts to children need to weigh the ben- and may attenuate the observed risk estimates toward the null. efits against the risks of severe adverse events. Third, the Fourth, we did not explore whether the prescriptions of anti- present findings call for a careful reevaluation regarding the biotics, a broader marker of infection, increased after initia- prudent use of corticosteroid bursts in children because of the tion of the corticosteroid bursts. The prescriptions of antibi- substantial proportion of children administered corticoste- otics after the corticosteroid bursts will be worth further roid bursts in the world. investigation. In this study, we were able to control for time- Among children receiving corticosteroid bursts in our invariant risk factors in individual-level variability but not study, 91% had no baseline comorbid condition. Most of population-level variability owing to the features of self- the corticosteroid bursts were prescribed for non–life- controlled case series design. threatening conditions, including acute respiratory tract in- fections and allergic diseases. A clinical practice guideline for the management of sore throat indicates a weak recommen- Conclusions dation for the use of oral corticosteroids in children aged 5 years 25 22 or older and in adults. Dvorin et al estimated that 11% of This nationwide population-based study demonstrates that adult outpatients with acute respiratory tract infections across oral corticosteroid bursts are commonly prescribed to chil- the US are treated with systemic corticosteroids. Although dren for non–life-threatening conditions, including acute some studies showed that corticosteroid bursts mitigated ear- respiratory tract infections and allergic diseases. Treatment lier symptoms of acute pharyngitis, clinical trials showed no with corticosteroid bursts is associated with a 1.4- to 2.2-fold efficacy of corticosteroid bursts for acute lower respiratory tract increased risk of GI bleeding, sepsis, and pneumonia within 27 28 infection and sinusitis. Further research is necessary to con- the first month after initiation of corticosteroid therapy firm the high frequency of use of corticosteroid bursts in among children. Clinicians should be aware of these rare but children with acute respiratory tract infections or other potentially serious adverse events associated with use of non–life-threatening diseases. corticosteroid bursts for children, particularly during the first month after corticosteroid initiation. These findings provide real-world evidence for clinicians and guideline Limitations Several limitations deserve mention in our analysis. First, life- developers to implement strategies with optimal benefit to style factors, including exposure to tobacco smoke and body risk ratios for preventing avoidable harms from the use of mass index, are not available in the NHIRD. We therefore used corticosteroid bursts for children. ARTICLE INFORMATION Center for Allergy and Clinical Immunology Cheng Hospital, Chang Gung Memorial Hospital, Research, College of Medicine, National Cheng New Taipei, Taiwan (Huang). Accepted for Publication: January 27, 2021. Kung University, Tainan, Taiwan (Wang); Author Contributions: Drs Yao and H.-J. Tsai had Published Online: April 19, 2021. Department of Pediatrics, National Cheng Kung full access to all of the data in the study and take doi:10.1001/jamapediatrics.2021.0433 University Hospital, Tainan, Taiwan (Wang); responsibility for the integrity of the data and the Correction: This article was corrected on July 6, Department of Statistics, National Cheng Kung accuracy of the data analysis. 2021, to fix errors in the Abstract, Key Points, and University, Tainan, Taiwan (S.-M. Chang); Institute Concept and design: Yao, Wang, Wu, H.-J. Tsai. text. of Population Health Sciences, National Health Acquisition, analysis, or interpretation of data: Yao, Research Institutes, Zhunan, Taiwan (Y.-C. Chang, Open Access: This is an open access article Wang, S.-M. Chang, Y.-C. Chang, Y.-F. Tsai, Wu, Y.-F. Tsai, H.-J. Tsai); Precision Medicine and distributed under the terms of the CC-BY License. Huang. Translational Research Center, Department of © 2021 Yao TC et al. JAMA Pediatrics. Drafting of the manuscript: Yao, Wang, H.-J. Tsai. Population Medicine, Harvard Pilgrim Health Care Critical revision of the manuscript for important Author Affiliations: Division of Allergy, Asthma, Institute and Harvard Medical School, Boston, intellectual content: All authors. and Rheumatology, Department of Pediatrics, Massachusetts (Wu); Department of Pediatrics, Statistical analysis: S.-M. Chang, Y.-C. Chang, Chang Gung Memorial Hospital, Taoyuan, Taiwan Children’s Hospital, Boston, Massachusetts (Wu); Y.-F. Tsai. (Yao); School of Medicine, Chang Gung University Department of Pediatrics, New Taipei Municipal Tu Obtained funding: Yao, H.-J. Tsai. College of Medicine, Taoyuan, Taiwan (Yao, Huang); 728 JAMA Pediatrics July 2021 Volume 175, Number 7 (Reprinted) jamapediatrics.com Association of Oral Corticosteroid Bursts With Severe Adverse Events in Children Original Investigation Research Supervision: Yao, Wang, Wu, H.-J. Tsai. literature. Eur Respir J. 2018;52(4):1800703. adverse events and economic impact associated doi:10.1183/13993003.00703-2018 with oral corticosteroids in asthma. J Asthma. 2019; Conflict of Interest Disclosures: Dr Wu reported 56(12):1334-1346. doi:10.1080/02770903.2018. receiving grants from GlaxoSmithKline outside the 5. Fardet L, Kassar A, Cabane J, Flahault A. submitted work. No other disclosures were Corticosteroid-induced adverse events in adults: reported. frequency, screening and prevention. Drug Saf. 17. Waljee AK, Wiitala WL, Govani S, et al. 2007;30(10):861-881. doi:10.2165/00002018- Corticosteroid use and complications in a US Funding/Support: This work was supported by 200730100-00005 inflammatory bowel disease cohort. PLoS One. grants from National Health Research Institutes, 2016;11(6):e0158017. doi:10.1371/journal.pone. Taiwan (PH-109-PP-08, Dr Tsai), Ministry of Science 6. Nashel DJ. Is atherosclerosis a complication of and Technology of Taiwan (MOST long-term corticosteroid treatment? Am J Med. 107-2314-B-400-031-MY3, Dr Tsai; and MOST 1986;80(5):925-929. doi:10.1016/0002-9343(86) 18. Ernst P, Coulombe J, Brassard P, Suissa S. The 106-2314-B-182-051-MY3 and MOST 90639-X risk of sepsis with inhaled and oral corticosteroids in 109-2314-B-182-042-MY3, Dr Yao); research grants patients with COPD. COPD. 2017;14(2):137-142. 7. Walsh LJ, Wong CA, Pringle M, Tattersfield AE. from the Headquarters of University Advancement, doi:10.1080/15412555.2016.1238450 Use of oral corticosteroids in the community and National Cheng Kung University, Tainan, Taiwan the prevention of secondary osteoporosis: a cross 19. Bloechliger M, Reinau D, Spoendlin J, et al. (Dr Wang), Chang Gung Medical Foundation sectional study. BMJ. 1996;313(7053):344-346. Adverse events profile of oral corticosteroids (CMRPG3F1711-3, CMRPG3F0361, CMRPG3J0121, doi:10.1136/bmj.313.7053.344 among asthma patients in the UK: cohort study and CMRPG3K1371, Dr Yao), and the National with a nested case-control analysis. Respir Res. 8. Van Staa TP, Leufkens HG, Abenhaim L, Zhang B, Institutes of Health/Eunice Kennedy Shriver 2018;19(1):75. doi:10.1186/s12931-018-0742-y Cooper C. Use of oral corticosteroids and risk of National Institute of Child Health and Human fractures. J Bone Miner Res. 2000;15(6):993-1000. 20. Wallace BI, Waljee AK. Burst case scenario: why Development (1R01HD085993-01, Dr Wu). doi:10.1359/jbmr.2000.15.6.993 shorter may not be any better when it comes to Role of the Funder/Sponsor: The funding source corticosteroids. Ann Intern Med. 2020;173(5):390- 9. Yao TC, Huang YW, Chang SM, Tsai SY, Wu AC, had no role in the design and conduct of the study; 391. doi:10.7326/M20-4234 Tsai HJ. Association between oral corticosteroid collection, management, analysis, and bursts and severe adverse events: a nationwide 21. Bénard-Laribière A, Pariente A, Pambrun E, interpretation of the data; preparation, review, or population-based cohort study. Ann Intern Med. Bégaud B, Fardet L, Noize P. Prevalence and approval of the manuscript; and decision to submit 2020;173(5):325-330. doi:10.7326/M20-0432 prescription patterns of oral glucocorticoids in the manuscript for publication. adults: a retrospective cross-sectional and cohort 10. Waljee AK, Rogers MA, Lin P, et al. Short term Disclaimer: This study is based in part on data from analysis in France. BMJ Open. 2017;7(7):e015905. use of oral corticosteroids and related harms the National Health Insurance Research Database doi:10.1136/bmjopen-2017-015905 among adults in the United States: population provided by the Bureau of National Health based cohort study. BMJ. 2017;357:j1415. 22. Dvorin EL, Lamb MC, Monlezun DJ, Boese AC, Insurance of the Ministry of Health and Welfare, doi:10.1136/bmj.j1415 Bazzano LA, Price-Haywood EG. High frequency of Taiwan. The interpretation and conclusions systemic corticosteroid use for acute respiratory contained in this article do not represent those of 11. Narum S, Westergren T, Klemp M. tract illnesses in ambulatory settings. JAMA Intern the Bureau of National Health Insurance or the Corticosteroids and risk of gastrointestinal Med. 2018;178(6):852-854. doi:10.1001/ Ministry of Health and Welfare. bleeding: a systematic review and meta-analysis. jamainternmed.2018.0103 BMJ Open. 2014;4(5):e004587. doi:10.1136/ Additional Contributions: We thank staff members bmjopen-2013-004587 23. Petersen I, Douglas I, Whitaker H. Self in the Data Science Center of the Ministry of Health controlled case series methods: an alternative to and Welfare, Taiwan, for their data management 12. Rogers MAM, Lin P, Nallamothu BK, Kim C, standard epidemiological study designs. BMJ. 2016; and maintenance. Waljee AK. Longitudinal study of short-term 354:i4515. doi:10.1136/bmj.i4515 corticosteroid use by working-age adults with REFERENCES diabetes mellitus: risks and mitigating factors. 24. VanderWeele TJ, Ding P. Sensitivity analysis in J Diabetes. 2018;10(7):546-555. doi:10.1111/1753- observational research: introducing the E-Value. 1. McDonough AK, Curtis JR, Saag KG. The 0407.12631 Ann Intern Med. 2017;167(4):268-274. doi:10.7326/ epidemiology of glucocorticoid-associated adverse M16-2607 events. Curr Opin Rheumatol. 2008;20(2):131-137. 13. Christiansen CF, Christensen S, Mehnert F, doi:10.1097/BOR.0b013e3282f51031 Cummings SR, Chapurlat RD, Sørensen HT. 25. Aertgeerts B, Agoritsas T, Siemieniuk RAC, et al. Glucocorticoid use and risk of atrial fibrillation or Corticosteroids for sore throat: a clinical practice 2. Fuhlbrigge AL, Lemanske RF Jr, Rasouliyan L, flutter: a population-based, case-control study. guideline. BMJ. 2017;358:j4090. doi:10.1136/bmj. Sorkness CA, Fish JE. Practice patterns for oral Arch Intern Med. 2009;169(18):1677-1683. j4090 corticosteroid burst therapy in the outpatient doi:10.1001/archinternmed.2009.297 management of acute asthma exacerbations. 26. Sadeghirad B, Siemieniuk RAC, Allergy Asthma Proc. 2012;33(1):82-89. 14. Sullivan PW, Ghushchyan VH, Globe G, Schatz Brignardello-Petersen R, et al. Corticosteroids for doi:10.2500/aap.2012.33.3499 M. Oral corticosteroid exposure and adverse effects treatment of sore throat: systematic review and in asthmatic patients. J Allergy Clin Immunol. 2018; meta-analysis of randomised trials. BMJ. 2017;358: 3. Freyberg RH, Traeger CH, Patterson M, Squires 141(1):110-116.e7. doi:10.1016/j.jaci.2017.04.009 j3887. doi:10.1136/bmj.j3887 W, Adams CH. Problems of prolonged cortisone treatment for rheumatoid arthritis; further 15. Dix D, Cellot S, Price V, et al. Association 27. Hay AD, Little P, Harnden A, et al. Effect of oral investigations. J Am Med Assoc. 1951;147(16):1538- between corticosteroids and infection, sepsis, and prednisolone on symptom duration and severity in 1543. doi:10.1001/jama.1951.03670330030008 infectious death in pediatric acute myeloid nonasthmatic adults with acute lower respiratory leukemia (AML): results from the Canadian tract infection: a randomized clinical trial. JAMA. 4. Volmer T, Effenberger T, Trautner C, Buhl R. Infections in AML research group. Clin Infect Dis. 2017;318(8):721-730. doi:10.1001/jama.2017.10572 Consequences of long-term oral corticosteroid 2012;55(12):1608-1614. doi:10.1093/cid/cis774 therapy and its side-effects in severe asthma in 28. Venekamp RP, Thompson MJ, Hayward G, et al. adults: a focused review of the impact data in the 16. Al Efraij K, Johnson KM, Wiebe D, Sadatsafavi Systemic corticosteroids for acute sinusitis. M, FitzGerald JM. A systematic review of the Cochrane Database Syst Rev. 2014;(3):CD008115. jamapediatrics.com (Reprinted) JAMA Pediatrics July 2021 Volume 175, Number 7 729 eMethods. eTable 1. Equivalent Doses of Corticosteroids Investigated in This Study eTable 2. ICD-9-CM and ICD-10-CM Codes of Four Adverse Events and Negative Control Outcome eTable 3. Incidence Rate Ratios for Syncope (Negative Control Outcome) Associated With Corticosteroid Bursts in Children eTable 4. Incidence Rate Ratios and E-Values for Four Adverse Events Associated With Corticosteroid Bursts in Children eTable 5. Subgroup Analyses of Children With Various Days of Corticosteroid Use (Less Than 7 Days Versus 7 Days or More) eMethods. Time-varying covariates concomitant medication use adjusted time-varying covariates included in the final analytical models are: nonsteroidal anti-inflammatory drugs (NSAIDs), proton-pump inhibitors, for ; NSAIDs, systemic immunosuppressive agents, for sepsis; NSAIDs, systemic immunosuppressive agents, for pneumonia; and NSAIDs, for glaucoma. eTable 1. Equivalent Doses of Corticosteroids Investigated in This Study Corticosteroid Equivalent Dose Betamethasone 0.6 mg Dexamethasone 0.75 mg Methylprednisolone 4 mg Triamcinolone 4 mg Prednisone 5 mg Prednisolone 5 mg Hydrocortisone 20 mg Cortisone 25 mg eTable 2. ICD-9-CM and ICD-10-CM Codes of Four Adverse Events and Negative Control Outcome J12 -J18 Pneumonia Glaucoma 780.2x, 992.1x R55 Syncope eTable 3. Incidence Rate Ratios for Syncope (Negative Control Outcome) Associated With Corticosteroid Bursts in Children * * 5-30 Days 31-90 Days Duration of Using No. of Daily Dose (mg/day), Adverse Event Corticosteroids Incidence Rate Ratio Incidence Rate Ratio Events Median (IQR) (days), Median (IQR) (95% CI) (95% CI) Syncope 204 10.00 (1.50-15.00 ) 3.00 (3.00 -3.00 ) 1.03 (0.77-1.38) 0.92 (0.74-1.15) IQR = interquartile range; CI = confidence interval. Syncope model was adjusted for NSAIDs, acute bronchitis and bronchiolitis, acute laryngitis and tracheitis, asthma, and allergic rhinitis. eTable 4. Incidence Rate Ratios and E-Values for Four Adverse Events Associated With Corticosteroid Bursts in Children Duration of Using Daily Dose 5-30 Days 31-90 Days No. of Corticosteroids Adverse Event (mg/day), Median Events (days), Median Incidence Rate Ratio Incidence Rate Ratio (IQR) E-value (CI ) E-value (CI ) (IQR) (95% CI) (95% CI) 1,176 10.00 (1.50-15.00) 3.00 (3.00-3.00) 1.41 (1.27-1.57) 1.10 (1.02-1.19) 1.43 (1.16) GI Bleeding 230 2.38 (0.60-8.33) 3.00 (3.00-3.00) 2.02 (1.55-2.64) 1.08 (0.88-1.32) 1.37 (1.00) Sepsis Pneumonia 19,579 2.03 (0.50-9.33) 3.00 (3.00-3.00) 2.19 (2.13-2.25) 3.80 (3.68) 1.09 (1.07-1.11) 3.21 (1.34) Glaucoma 657 8.00 (1.50-15.00) 3.00 (3.00-3.00) 0.98 (0.85-1.13) 0.95 (0.85-1.06) 1.29 (1.00) E-values for lower limit of the CI when relative risk>1; E-values for upper limit of the CI when relative risk<1. GI bleeding model was adjusted for nonsteroidal anti-inflammatory drugs (NSAIDs), proton-pump inhibitors, Sepsis model was adjusted for NSAIDs, systemic immunosuppressive agents, . Pneumonia model was adjusted for NSAIDs, systemic immunosuppressive agents, Glaucoma model was adjusted for NSAIDs, . eTable 5. Subgroup Analyses of Children With Various Days of Corticosteroid Use (Less Than 7 Days Versus 7 Days or More) 5-30 Days 31-90 Days 5-30 Days 31-90 Days Adverse Event Incidence Rate Ratio Incidence Rate Ratio Incidence Rate Ratio Incidence Rate Ratio (95% CI) (95% CI) (95% CI) (95% CI) GI Bleeding 1.40 (1.26-1.56) 1.12 (1.03-1.21) 1.49 (1.08-2.07) 0.89 (0.67-1.18) Sepsis 2.01 (1.53-2.63) 1.07 (0.87-1.32) 1.78 (0.50-6.35) 0.88 (0.34-2.27) Pneumonia 2.22 (2.16-2.28) 1.10 (1.08-1.13) 1.72 (1.52-1.93) 0.85 (0.77-0.94) Glaucoma 0..99 (0.85-1.15) 0.96 (0.85-1.07) 0.94 (0.55-1.63) 0.88 (0.58-1.34) GI bleeding model was adjusted for nonsteroidal anti-inflammatory drugs (NSAIDs), proton-pump inhibitors, Sepsis model was adjusted for NSAIDs, systemic immunosuppressive agents, . Pneumonia model was adjusted for NSAIDs, systemic immunosuppressive agents, Glaucoma model was adjusted for NSAIDs, .
JAMA Pediatrics – American Medical Association
Published: Jul 19, 2021