Oral Glucocorticoids and Incident Treatment of Diabetes Mellitus, Hypertension, and Venous Thromboembolism in Children

Abstract

Rates of incident treatment were quantified in this study for diabetes mellitus, hypertension, and venous thromboembolism (VTE) associated with oral glucocorticoid exposure in children aged 1–18 years. The retrospective cohort included more than 930,000 children diagnosed with autoimmune diseases (namely, inflammatory bowel disease, juvenile idiopathic arthritis, or psoriasis) or a nonimmune comparator condition (attention-deficit/hyperactivity disorder) identified using US Medicaid claims (2000–2010). Associations of glucocorticoid dose per age- and sex-imputed weight with incident treated diabetes, hypertension, and VTE were estimated using Cox regression models. Crude rates were lowest for VTE (unexposed: 0.5/million person-days (95% confidence interval (CI): 0.4, 0.6); currently exposed: 15.6/million person-days (95% CI: 11.8, 20.1)) and highest for hypertension (unexposed: 6.7/million person-days (95% CI: 6.5, 7.0); currently exposed: 74.4/million person-days (95% CI: 65.7, 83.9)). Absolute rates for all outcomes were higher in unexposed and exposed children with autoimmune diseases compared with those with attention-deficit/hyperactivity disorder. Strong dose-dependent relationships were found between current glucocorticoid exposure and all outcomes (adjusted hazard ratios for high-dose glucocorticoids: for diabetes mellitus, 5.93 (95% CI: 3.94, 8.91); for hypertension, 19.13 (95% CI: 15.43, 23.73); for VTE, 16.16 (95% CI: 8.94, 29.22)). These results suggest strong relative risks, but low absolute risks, of newly treated VTE, diabetes, and especially hypertension in children taking high-dose oral glucocorticoids.

Abbreviations

ADHD

attention-deficit/hyperactivity disorder

aHR

adjusted hazard ratio

CI

confidence interval

DM

diabetes mellitus

HTN

hypertension

IBD

inflammatory bowel disease

ICD

International Classification of Diseases, Ninth Revision

JIA

juvenile idiopathic arthritis

PSO

psoriasis

VTE

venous thromboembolism

Oral glucocorticoids are commonly used drugs prescribed for children, taken yearly by approximately 1% of children and youth younger than 20 years each month and, in 1 US study, by up to 40% of children with asthma (1, 2). Despite the use of locally acting steroid preparations (e.g., inhaled, topical) and other steroid-sparing therapies (e.g., immunosuppressants) for many conditions, oral glucocorticoid use has remained high over time (1–4). Valued for fast-acting anti-inflammatory and immunosuppressive properties, glucocorticoids also have well-characterized, incompletely understood toxicities, including effects on mood, sleep, skin, bones, metabolism, eyes, and the cardiovascular system (5). Many toxicities are dose dependent and reversible, whereas others (e.g., cataracts, osteonecrosis) are not (5, 6).

More serious glucocorticoid-related toxicities include cardiometabolic complications such as diabetes mellitus (DM), hypertension (HTN), and venous thromboembolism (VTE). The absolute rates of these cardiometabolic conditions in children are low: incidence of DM (type 1 and type 2 combined) in children is approximately 3/10,000 person-years (7); prevalence of HTN in children is 0.3%–3% (8–10); and incidence of VTE in children is approximately 0.3/10,000 person-years (11). However, the rates of these complications with glucocorticoids in large populations have beencharacterized predominantly in adults (12–18). Unlike adults, who often take oral glucocorticoids chronically (19), children are more likely to take glucocorticoids in short, high-dose courses (e.g., for asthma) (20). Furthermore,findings from adult populations are less applicable to children because of underlying differences in physiology, absolute risk, and comorbidities (21). Glucocorticoid-associated cardiometabolic complications have been examined only in select pediatric populations—mostly those with malignancies or transplantation, or hospitalized children (22–28). These findings, too, are unlikely to generalize to other pediatric populations because of dramatically different glucocorticoid exposure patterns, potentially toxic cotreatments, disease-specific risks, and selection and referral biases.

Very little is known about glucocorticoid-associated VTE in children, and no population-based study, to our knowledge, has examined any cardiometabolic complications in children. We quantified rates of new-onset DM, HTN, and VTE associated with different oral glucocorticoid exposure patterns in a broad pediatric population affected by autoimmune diseases or a nonimmune reference condition, hypothesizing dose- and time-dependent toxicities that were more likely with autoimmune diseases.

METHODS

Design and setting

We performed a retrospective cohort study using 50-state Medicaid Analytic eXtract files (2000–2010). Medicaid Analytic eXtract files contain administrative claims data from publicly insured, predominantly low-income or disabled individuals in the United States, including demographic, clinical, drug-dispensing, and health use information. This study was approved or deemed nonhuman subjects research by the authors’ affiliations’ respective institutional review boards.

Study population

The study cohort comprised children 1–18 years old with at least 9 months of observable time with continuous enrollment, pharmacy claims at least every 6 months, and 1 of 4 chronic disease diagnoses: inflammatory bowel disease (IBD), juvenile idiopathic arthritis (JIA), psoriasis (PSO), or attention-deficit/hyperactivity disorder (ADHD; a common, nonimmune reference condition for which glucocorticoids are not indicated). Diagnosis of IBD (International Classification of Diseases, Ninth Revision (ICD-9) codes 555, 556) or JIA (ICD-9 codes 714, 720, 696.0) required at least 2 diagnostic codes 7–183 days apart or at least 1 diagnosis code followed by an immunosuppressant medication (e.g., methotrexate, 6-mercaptopurine, or a biologic). Diagnosis of ADHD (ICD-9 code 314.0) or PSO (ICD-9 code 696.1) required at least 2 ICD-9 codes 7–183 days apart. Most data on children with ADHD came from 2006–2010. Patients diagnosed at any time with 1 or more qualifying disorder of interest were classified hierarchically as having IBD > JIA > PSO > ADHD, on the basis of clinical practices for children with more than 1 disease.

We excluded children with prior malignancy, chemotherapy, organ or hematopoietic transplantation, human immunodeficiency virus, or a non-JIA rheumatic disease (e.g., systemic lupus) (Web Table 1, available at https://academic.oup.com/aje). We also excluded children diagnosed before age 1 year, because of questions about diagnostic validity, and applied additional outcome-specific exclusions (Web Table 1).

The index date occurred on or after 274 days of enrollment with a qualifying diagnosis of inclusion (i.e., ADHD, IBD, JIA, or PSO) or with an oral glucocorticoid claim (i.e., an exposure) (Figure 1). To achieve a new-user design (29), children whose follow-up began with a glucocorticoid claim needed to have at least 183 prior days without a prescribed glucocorticoid. The 9-month baseline period and 6-month glucocorticoid-free requirement were designed to account for a theoretical, left-censored, 3-month glucocorticoid course starting immediately before enrollment. Because chronic diseases may be diagnosed months or years after symptoms develop and are treated, we did not require glucocorticoid-exposed individuals to have qualifying diagnoses before the index date as long as a qualifying diagnosis was made during the study period. This approach ensured capture of biologically relevant exposure periods.

Figure 1.

Study design. Cohort entry required at least 9 months of baseline time after enrollment and began A) with or after a qualifying inclusion diagnosis (Dx; i.e., a Dx followed by a second diagnosis or prescription drug (Rx) claim within 7–183 days), B) the first glucocorticoid (GC) claim, or C) 6 months after most recent glucocorticoid exposure, if the initial claim(s) occurred within 9 months after enrollment. Children entering the study at the time of qualifying diagnosis without concurrent or prior glucocorticoid claim were considered unexposed (A, gray dashed line). Children entering the study with a glucocorticoid claim were required to have a qualifying diagnosis within either the C) baseline or B) follow-up period. Glucocorticoid dose was updated over time and characterized by current dose (low, medium, or high, as represented by thickness of solid lines), recently exposed within the last 1–180 days (short-dashed line), or remotely exposed over 180 days ago (dotted line). In models of ever-exposure versus never-exposure, any time after the first glucocorticoid claim during the study period was considered exposed, whereas any time preceding the first glucocorticoid claim during the study period was considered unexposed. Children were followed until occurrence of an outcome or censoring event (i.e., loss of continuous insurance eligibility, no pharmacy claim after 6 months, interval diagnosis of exclusion, or death).

Exposure

We characterized oral glucocorticoid exposure by date, dose, and duration on the basis of available information from pharmacy claims on strength, quantity, daily dose, and days’ supply. For overlapping prescriptions, refills were conservatively interpreted to connote completion of previously dispensed glucocorticoids. In models of ever-exposure versus never-exposure, any time after the first glucocorticoid claim during the study period (i.e., on or after the index date) was considered exposed, whereas any time after the index date and preceding the first glucocorticoid claim during the study period was considered unexposed (Figure 1). We classified current glucocorticoid dose as low (<0.25 mg/kg/day), medium (0.25–0.99 mg/kg/day), or high (≥1 mg/kg/day) on the basis of prednisone-equivalent dosage (30) and age- and sex-imputed weights (31). Glucocorticoid dosage was time varying and updated on a person-day basis. We established dose category cutoffs a priori on the basis of clinical standards. Timing of prior glucocorticoid use was categorized as recent (most recent dose, 1–180 days ago) or remote (most recent dose, > 180 days ago). In models of glucocorticoid dose, an unexposed cohort included children who entered the study with a qualifying disorder of interest and no recorded prior glucocorticoid exposure, including during the baseline period (Figure 1).

Outcomes

We defined primary outcomes based on incident treatment of DM (type 1 or 2), HTN, or VTE (Web Table 1). We alsorequired HTN diagnoses for new users of less-specific antihypertensives (Web Table 1). We also considered more specific secondary definitions of each outcome (Web Table 1). The study cohort was censored for loss of continuous insurance eligibility, no pharmacy claim after 6 months, interval diagnosis of exclusion, or death.

Covariates

For all outcomes, we considered a wide range of confounders, including demographics, chronic disease diagnosis (i.e., ADHD, IBD, JIA, or PSO), and health-care use within 6 months before the index date (Web Table 1). We evaluated additional, outcome-specific comorbidity and treatment confounders (Web Table 1).

Statistical analysis

We compared cohorts using standard descriptive statistics. Crude rates of primary and secondary outcomes were estimated for each condition of interest and for different levels of exposure (i.e., unexposed, remote, recent, or current exposure). We estimated associations between oral glucocorticoid exposure and each outcome using multivariable Cox proportional hazards regression adjusted for confounders (Web Table 1) and expressed as adjusted hazard ratios with 95% confidence intervals. Of note, because of relatively fewevents meeting the secondary definition for VTE, we used a reduced model with fewer degrees of freedom (Web Table 1). We considered 2 primary models of time-varying glucocorticoid exposure: ever- versus never-exposed during the study period; and current dose (low, medium, or high) or timing of prior use (recent or remote) compared with no prior exposure. Additional secondary analyses were restricted to children aged 11 years or older (who experienced all outcomes at higher rates) and stratified by disease.

Sensitivity analyses examined the strength of unmeasured confounding that could nullify observed results and lower bound of confidence intervals, using the E-value (32). Analyses were performed using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina) and R, version 3.4.0 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

We identified 932,517 eligible individuals (n = 198,855 (21.3%) glucocorticoid exposed) followed for 1.6 million person-years before outcome-specific exclusions (0.7%–1.5%of overall cohort) (Web Figure 1). Compared with unexposed individuals, glucocorticoid-exposed children tended to be younger and have longer follow-up time, higher rates of prior oral glucocorticoid and nonsteroidal anti-inflammatory exposure, less stimulant use, and higher levels of health-care use (Table 1). ADHD was the most common condition across exposed and unexposed cohorts, but children with IBD and JIA were more likely to receive glucocorticoids than children with ADHD or PSO. Across the study population, the median glucocorticoid course was 5 (interquartile range, 5–7) days at 0.66 (interquartile range, 0.31–1.12) mg/day prednisone-equivalents (Web Table 2). However, patterns of glucocorticoid exposure varied considerably by disease (Table 2, Web Table 2). The median length of glucocorticoid courses and total follow-up time spent receiving glucocorticoids was longer among children with IBD (27 days; 6.7%) and JIA (9 days; 4.8%) than among children with ADHD (5 days; 0.4%) or PSO (6 days; 0.8%). Of note, more than half of glucocorticoid-exposed children (63.8%), including those with ADHD (65.5%) or PSO (57.8%), were also diagnosed with asthma or prescribed albuterol either in the baseline period or during follow-up.

Table 1.

Baseline Characteristics of Children in This Study From US Medicaid Data, 2000–2010^a

Characteristic	No GCs b  (n = 733,662)		Any GCs (n = 198,855)		Standardized Difference c
	No.	%	No.	%
Age, years
1–5	66,163	9.0	41,538	20.9	0.34
6–10	378,304	51.6	91,782	46.2	−0.11
11–15	230,821	31.5	51,262	25.8	−0.13
16–18	58,374	8.0	14,273	7.2	−0.03
Female sex	217,842	29.7	61,202	30.8	0.02
Race
White	437,043	59.6	122,551	61.6	0.04
Black	147,710	20.1	33,893	17.0	−0.08
Hispanic	60,115	8.2	16,480	8.3	0.00
Other/unknown	88,794	12.1	25,931	13.0	0.03
Years of follow-upd	1.5 (1.3)		2.6 (1.5)		0.77
MAX eligibility
Poverty	333,899	45.5	93,145	46.8	0.03
Disability	112,779	15.4	35,834	18.0	0.07
Foster care status	68,175	9.3	13,787	6.9	−0.09
Other	218,809	29.8	56,089	28.2	−0.04
Year of entry
2000–2005	23,933	3.3	14,212	7.1	0.18
2006–2010	709,729	96.7	184,643	92.9
Chronic diseasee
ADHD	699,801	95.4	178,947	90.0	−0.21
IBD	5,877	0.8	7,186	3.6	0.19
JIA	10,790	1.5	7,500	3.8	0.14
PSO	17,194	2.3	5,222	2.6	0.02
Other baseline conditions
Asthma	90,015	12.3	66,150	33.3	0.52
Central catheter	2,240	0.3	1,494	0.8	0.06
Diabetes mellitus	5,010	0.7	1,603	0.8	0.01
Fracture	42,523	5.8	11,988	6.0	0.01
Hypertensiona	6,742	0.9	2,822	1.4	0.05
Mood disorder	129,509	17.7	33,660	16.9	−0.02
Obesity	16,427	2.2	4,757	2.4	0.01
VTE	235	0.0	125	0.1	0.01
Baseline medications
Albuterol	128,520	17.5	82,070	41.3	0.54
Antipsychotics	114,401	15.6	30,784	15.5	0.00
Calcineurin inhibitors	151	0.0	274	0.1	0.04
Oral contraceptives	11,095	1.5	4,330	2.2	0.05
GCsb	86,637	11.8	60,713	30.5	0.47
NSAIDs	127,995	17.4	52,079	26.2	0.21
Stimulants	663,906	90.5	137,747	69.3	−0.55
Health-care use
Hospitalization, last 6 months	28,744	3.9	11,815	5.9	0.09
No. of medicationsd	1.8 (1.2)		2.0 (1.4)		0.11
No. of office visitsd	3.4 (2.6)		4.2 (3.4)		0.25

Abbreviations: ADHD, attention deficit/hyperactivity disorder; GC, glucocorticoid; IBD, inflammatory bowel disease; IQR, interquartile range; JIA, juvenile idiopathic arthritis; MAX, Medicaid Analytic eXtract; NSAID, nonsteroidal anti-inflammatory drug; PSO, psoriasis; SD, standard deviation; VTE, venous thromboembolism.

^a Characteristics apply to the population eligible for the study of at least 1 outcome; individuals with outcome-specific exclusions (e.g., pre-existing diabetes for the diabetes outcome) are counted in this table.

^b Some children in the unexposed cohort received GCs before the index date.

^c Standardized differences > 0.1 or less than −0.1 suggest imbalance.

^d Values are expressed as mean (standard deviation).

^e Children diagnosed with > 1 condition were classified hierarchically: IBD > JIA > PSO > ADHD.

Table 2.

Duration and Prevalence of Glucocorticoid Exposure During the Entire Study Period for the Diabetes Outcome in US Medicaid Data, 2000–2010^a

Glucocorticoid Use Pattern	Overall		ADHD		IBD		JIA		PSO
	Person-Years	%	Person-Years	%	Person-Years	%	Person-Years	%	Person-Years	%
None	1,230,659	76.01	1,180,498	77.56	8,801	35.02	17,902	48.64	23,457	66.80
Remote (>180 days ago)	232,225	14.34	205,755	13.52	8,616	34.29	10,529	28.60	7,326	20.86
Recent (within prior 180 days)	145,901	9.01	129,182	8.49	6,029	23.99	6,627	18.01	4,063	11.57
Current, low doseb	2,029	0.13	939	0.06	344	1.37	697	1.89	48	0.14
Current, medium doseb	5,174	0.32	3,412	0.22	981	3.90	649	1.76	133	0.38
Current, high doseb	3,093	0.19	2,243	0.15	358	1.42	403	1.10	89	0.25

Abbreviations: ADHD, attention deficit/hyperactivity disorder; IBD, inflammatory bowel disease; JIA, juvenile idiopathic arthritis; PSO, psoriasis.

^a Usage patterns varied trivially for the outcomes of hypertension and venous thromboembolism.

^b Dose was defined by daily estimated dose based on pharmacy claims and age- and sex-imputed weights: low, < 0.25 mg/kg/day; medium, 0.25–0.99 mg/kg/day; high (≥1 mg/kg/day). See text for more detail.

Crude rates of the outcomes varied considerably by disease and glucocorticoid exposure (Web Tables 3–5). Crude rates were lowest for VTE treatment, occurring in 0.5 (95% confidence interval (CI): 0.4, 0.6) per million person-days of unexposed follow-up time and 15.6 (95% CI: 11.8, 20.1) per million person-days of currently exposed follow-up time. Crude rates were highest for anti-HTN drug use, whose incidence was 6.7 (95% CI: 6.5, 7.0) per million person-days of unexposed follow-up time and 74.4 (95% CI: 65.7, 83.9) per million person-days of currently exposed follow-up time. Crude rates of antidiabetic drug use were 5.3 (95% CI: 5.1, 5.5) per million person-days of unexposed follow-up time and 24.5 (95% CI: 19.7, 30.0) per million person-days of currently exposed follow-up time. Crude rates of secondary outcomes requiring diagnoses were approximately 20%–50% of the primary outcomes, depending on the outcome. Compared with those with ADHD, absolute rates for all outcomes were generally higher in children with autoimmune diseases, more markedly so among those with current glucocorticoid exposure (Web Tables 3–5).

After adjusting for demographics, comorbidities, medication use, and health-care use, we found significant associations between any oral glucocorticoid exposure and the development of DM, HTN, and VTE (Figure 2). Examination of the associations of dose and timing revealed a notable dose response with current exposure: high-doseglucocorticoid exposure was strongly associated with DM(adjusted hazard ratio (aHR) = 5.93, 95% CI: 3.94, 8.91), HTN (aHR = 19.13, 95% CI: 15.43, 23.73), and VTE (aHR = 16.16, 95% CI: 8.94, 29.22). Hazard ratios were smaller but still substantial for low-dose exposure, ranging from approximately 3 (DM, HTN) to approximately 6 (VTE). Prior glucocorticoid exposure of any dose or duration was associated with smaller associations than current exposure at any level; glucocorticoid use within the prior 180 days corresponded to approximately doubling of hazard ratios, whereas use more than 180 days ago was associated with approximately 50% elevations in rates of DM and HTN. Models evaluating secondary outcome definitions (Figure 2) and children aged 11 years or older (Web Table 6) showed associations of similar magnitude. Analyses stratified by chronic disease suggested potentially stronger associations of high-dose-exposure on rates of DM and VTE among children with IBD and higher rates of HTN in high-dose–exposed children with JIA (Web Table 7). However, confidence intervals of disease-specific point estimates overlapped considerably.

Figure 2.

Dose- and time-dependent associations between glucocorticoid exposure and incident cardiometabolic outcomes. Forest plots showing adjusted hazard ratios (HRs; black diamonds) with 95% confidence intervals (CIs; bars) for primary (models 1 and 3) and secondary (models 2 and 4) outcomes based on any glucocorticoid exposure (models 1 and 2) and timing of most recent exposure or dose of current exposure (models 3 and 4). Results are shown by outcome: A) diabetes mellitus; B) hypertension; C) venous thromboembolism. Daily dose was estimated on the basis of pharmacy claims and age- and sex-imputed weights: low, <0.25 mg/kg/day; medium, 0.25–0.99 mg/kg/day; high, ≥1 mg/kg/day. Estimates came from Cox regression models adjusted for demographics, comorbidities, medication use, and health-care use. See text and Web Table 1 for more details.

In sensitivity analyses, the primary findings could be nullified only by very strong, unmeasured sources of confounding: aHR  ≥3 for low-dose exposure (corresponding to the E-value for the lower bound of the confidence interval for new antidiabetic drug use) and aHR  ≥7.3 for high-dose exposure (corresponding to the E-value for the lower bound of the confidence interval for new antidiabetic drug use).

DISCUSSION

In a large population-based study of publicly insured children, we found strong dose- and time-dependent associations between oral glucocorticoid exposure and 3 potentially serious cardiometabolic toxicities: DM, HTN, and VTE. Results were consistent for more specific secondary outcome definitions and analyses limited to children aged 11 years or older. In absolute terms, DM and HTN were more common complications, whereas the magnitude of glucocorticoid-associated HTN and VTE was stronger than for glucocorticoid-associated DM. Absolute rates of all complications at all glucocorticoid dose and duration levels appeared greater in children with autoimmune diseases (i.e., IBD, JIA, or PSO). The associations observed for prior glucocorticoid exposure were attenuated, but remote exposures longer than 6 months prior were associated with small increases in risk.

Prior population-based research on glucocorticoid-associated cardiometabolic toxicities focused on adult populations (12–16). One large study of glucocorticoid-associated VTE included children, but no pediatric-specific rates were reported (17). Compared with these prior studies, we found higher relative rates of all 3 cardiometabolic complications in glucocorticoid-exposed children, reflecting, in part, the lower baseline incidence in pediatric populations. Studies of glucocorticoid-associated complications in children have focused on high-risk pediatric populations, predominantly children with malignancies or transplant recipients (hematopoietic or solid organ). Depending on study population, design, and duration of follow-up, reported rates have ranged from 3% to 30% for DM (24, 26, 33–36), from 45% to 73% for HTN (25, 26, 37, 38), and from 2% to 12% for VTE (22, 23, 39–41)—rates with similar relative frequency (HTN > DM > VTE) but markedly higher than in our study, due to unique disease- and treatment-specific risks that do not apply to children without these diseases. The conditions included in our study have been associated with cardiometabolic complications in studies generally agnostic to glucocorticoid exposure: DM, HTN, and VTE in children and/or adults with psoriasis (42–45); VTE in children and adults with IBD (46, 47); and hypercoagulable states in children with JIA (48). In 1 population-based study of adults, VTE was associated with IBD (aHR = 1.84) and, to a lesser extent, rheumatoid arthritis (aHR = 1.21) and asthma (aHR = 1.22). However, these results were not adjusted for current or recent glucocorticoid exposure, which, in separate models, was associated with greater risk (aHR = 3.05–4.68) (16). Although stimulants can increase blood pressure slightly, their contribution to clinically relevant HTN appears minimal (49, 50), and in our study, glucocorticoid-exposed children with ADHD were less likely to use stimulants.

As with the outcomes studied here, many glucocorticoid-related toxicities are dose-dependent and resolve with drug tapers and discontinuation. We did not examine to what extent glucocorticoid-associated cardiometabolic toxicities persist in children after glucocorticoid withdrawal. In a single-center study of pediatric survivors of allogeneic hematopoietic cell transplantation, approximately one-third of those with treatment-associated DM or HTN required ongoing treatment after 2 years (26). Even if DM or HTNresolved after glucocorticoid discontinuation, it is unclear whether these children are predisposed to future cardiometabolic disease, akin to long-term risks among women with gestational DM and HTN (51, 52). In some children (e.g., those with type 1 DM), glucocorticoids may have hastened disease development or diagnosis. Risks of these complications were approximately double for children exposed to glucocorticoids within the past 6 months and up to 50% higher in children with more remote glucocorticoid exposures. These observations may have several potential explanations, including delayed or residual effects of glucocorticoids, delays in diagnosis, misclassification of glucocorticoid exposure within claims data, or bias from unmeasured confounding.

Our study had multiple strengths. The use of national Medicaid data enabled us to produce population-based estimates of serious though uncommon glucocorticoid-associated cardiometabolic toxicities that occur too rarely to detect in typical pediatric clinical trials (53). This setting minimized the impact of selection bias commonly seen in referral center–based studies and enabled adjustment for various sources of confounding. These estimates will be useful for clinicians to consider when counseling patients and families about the relative benefits and risks of glucocorticoids versus glucocorticoid-sparing treatments. In practice, clinicians prescribe glucocorticoids variably, even when treating the same conditions. Better understanding of the time course of glucocorticoid-associated cardiometabolic toxicities can improve monitoring practices for these complications.

Our study also had several limitations. Our estimations of glucocorticoid dose based on filled prescriptions may have differed from doses actually taken, for various reasons (e.g., nonadherence, assumptions about refills and weight), leading to overly precise estimates for specific doses. Nonetheless, given that our approach likely overestimated doses consumed, our results likely underestimated the true population-wide associations. Our use of broad dose categories probably reduced the impact of dose misclassification as well. The small numbers of outcomes observed within disease- and dose-specific strata may have limited our ability to distinguish effect modification by underlying disease on the multiplicative scale. Our models also did not enable us to examine associations of glucocorticoid course duration with the outcomes or whether nonlinear relationships may have existed. We were also limited in studying age-specific associations for DM, HTN, and VTE—cardiometabolic complications that increase with age—although associations observed among children aged 11 years or older were similar to those for all children. As an observational study, bias may have resulted from unmeasured confounding, including undiagnosed or uncoded obesity, disease severity, and other time-varying confounders. For example, children with high levels of systemic inflammation may have higher inherent risks of DM, HTN, or VTE; higher complication rates with high-dose glucocorticoid exposure may reflect these underlying risks. In our study, the predominant indication for glucocorticoid exposure was likely asthma. To our knowledge, severe asthma does not substantially increase the risk of DM, HTN, or VTE directly. Given the strong associations seen among children in all disease cohorts and our adjustment for proxies of unmeasured confounders such as disease severity (e.g., hospitalization, number of distinct drug classes taken) (54), unmeasured confounding is unlikely to have explained the strong observed associations. Our primary outcomes were not validated, and the medications examined (antidiabetic, antihypertensive, and antithrombotic) may have been used for other indications. However, we addressed this in primary analyses (e.g., exclusion of children with polycystic ovarian syndrome to study DM, requiring HTN diagnosis for users of β-blockers and clonidine), and more specific outcome definitions in secondary analyses supported the primary findings, including a previously used definition of type 2 DM more restrictive than 1 validated in children (55, 56). Our data set may not have included children treated with newer medicines, but we do not expect recent treatments to have fundamentally changed the relationship between glucocorticoid patterns and the outcomes, particularly given the similar associations seen across diseases. Finally, given that in this study we focused on lower-income and disabled children in the United States with chronic conditions, our results may not generalize to all pediatric populations.

In a large, population-based study of glucocorticoid exposure in children with various chronic conditions, we report novel, population-based estimates of dose- and time-dependent risks of cardiometabolic complications—namely, DM, HTN, and VTE. Associations appear to be highest during periods of ongoing glucocorticoid exposure, but small risks for these complications may persist after glucocorticoid exposure. Absolute rates of glucocorticoid-associated DM, HTN, and VTE vary widely by underlying disease, with higher rates in children with autoimmune diseases. In absolute terms, HTN is a more common glucocorticoid-related complication than DM or VTE. Fortunately, all these treatment complications are uncommon in children.