To assess whether increased genetic risk of type 2 diabetes (T2D) is associated with the development of hyperglycemia after glucocorticoid treatment.
We performed a retrospective analysis of individuals with no diagnosis of diabetes who received a glucocorticoid dose of ≥10 mg prednisone. We analyzed the association between hyperglycemia and a T2D global extended polygenic score, which was constructed through a meta-analysis of two published genome-wide association studies.
Of 546 individuals who received glucocorticoids, 210 developed hyperglycemia and 336 did not. T2D polygenic score was significantly associated with glucocorticoid-induced hyperglycemia (odds ratio 1.4 per SD of polygenic score; P = 0.038).
Individuals with increased genetic risk of T2D have a higher risk of glucocorticoid-induced hyperglycemia. This finding offers a mechanism for risk stratification as part of a precision approach to medical treatment.
Introduction
Glucocorticoids are commonly prescribed medications with a known adverse effect of hyperglycemia resulting from their effect on glucose metabolism (1–3); however, it is unclear why only 10–50% of individuals receiving glucocorticoids develop overt hyperglycemia (4). Because the pathophysiology of glucocorticoid-induced hyperglycemia (GIH) overlaps with that of type 2 diabetes (T2D) (1), we hypothesized that shared genetic factors may modulate the risk of both conditions. In this study, we assessed whether individuals with increased genetic risk of T2D are predisposed to develop hyperglycemia after glucocorticoid treatment.
Research Design and Methods
We analyzed medical records from participants in the Mass General Brigham Biobank, which is linked to electronic health records dating from 1988. We examined individuals with no diagnosis code for diabetes who received a glucocorticoid dose of ≥10 mg prednisone and had glucose levels checked within 7 days of glucocorticoid administration. Hyperglycemia was defined as fasting blood glucose ≥126 mg/dL or random blood glucose ≥200 mg/dL. Because fasting status is not typically recorded in the medical record, we defined fasting blood glucose as any glucose value drawn between 4:00 a.m. and 6:59 a.m.
We performed manual record reviews to verify the dose and timing of glucocorticoid administration, the timing of glucose measurement, and the lack of diabetes diagnosis. Glucocorticoid dose was converted to the equivalent dose of prednisone using established formulas (1). Only oral or intravenous glucocorticoid regimens were included.
We obtained clinical information from record review, including age at the time of glucocorticoid administration, sex, self-reported race and ethnicity, BMI, and baseline creatinine. BMI was calculated using the weight listed in the medical record within 1 year of glucocorticoid administration; if not available, we used the median BMI from the most recent 5 years of data (2016–2021). Individuals with widely fluctuating creatinine levels were excluded on the basis of the following criteria extracted from the medical record: new kidney transplant in the past 7 days, temporary or permanent hemodialysis, or acute kidney injury stage ≥2 (serum creatinine >2.0 times baseline) (5). Estimated glomerular filtration rate (eGFR) was derived from serum creatinine values using the Chronic Kidney Disease Epidemiology Collaboration equation without race correction (6).
Detailed information on genotyping and polygenic score construction is provided in the Supplementary Methods. Briefly, we constructed two T2D polygenic scores, which represent a weighted average of the effects of numerous variants across the genome: a global extended score (comprising ∼1 million variants, created using the Polygenic Risk Score–Continuous Shrinkage method [7]) and a restricted-to-significant score (comprising ∼400 variants reaching genome-wide significance) (8). We determined the weights for the polygenic scores by conducting a meta-analysis of genome-wide association study summary statistics from two prior publications (9,10). We performed logistic regression to analyze the association between polygenic score and GIH, controlling for age, sex, BMI, eGFR, glucocorticoid dose and duration, and the first 10 principal components to account for differences in genetic ancestry.
Results
We screened 1,755 individuals in the Mass General Brigham Biobank who had never been diagnosed with diabetes and who had ever received a glucocorticoid (Fig. 1). After manual record review, we confirmed that 758 individuals met the inclusion criteria for this study. We excluded individuals who had missing genomic data or widely fluctuating creatinine levels (see Research Design and Methods), leading to a final list of 546 participants.
Flowchart for selection of study participants. Participants were selected from the Mass General Brigham Biobank according to specified inclusion and exclusion criteria.
Flowchart for selection of study participants. Participants were selected from the Mass General Brigham Biobank according to specified inclusion and exclusion criteria.
Within this cohort, 60% were female, 92% self-identified as non-Hispanic, and 88% self-identified as White (Table 1). At the time of glucocorticoid administration, the mean age was 53.6 years, and the mean BMI was 27.6 kg/m2. After receiving glucocorticoids, 210 individuals developed hyperglycemia, whereas 336 individuals did not. The baseline characteristics of the two groups are displayed in Table 1.
Baseline characteristics of study participants
. | All . | Without hyperglycemia . | With hyperglycemia . | P . |
---|---|---|---|---|
N of participants | 546 | 336 | 210 | — |
Sex, n (%) | ||||
Female | 326 (59.7) | 206 (61.3) | 120 (57.1) | 0.33 |
Male | 220 (40.3) | 130 (38.7) | 90 (42.9) | 0.33 |
Self-reported ethnicity, n (%) | ||||
Hispanic | 7 (1.3) | 6 (1.8) | 1 (0.5) | 0.19 |
Non-Hispanic | 505 (92.5) | 307 (91.4) | 198 (94.3) | 0.21 |
Unknown/declined | 34 (6.2) | 23 (6.8) | 11 (5.2) | 0.45 |
Self-reported race, n (%) | ||||
Asian | 6 (1.1) | 4 (1.2) | 2 (1.0) | 0.80 |
Black | 29 (5.3) | 16 (4.8) | 13 (6.2) | 0.47 |
White | 478 (87.5) | 296 (88.1) | 182 (86.7) | 0.62 |
Other | 24 (4.4) | 15 (4.5) | 9 (4.3) | 0.92 |
Unknown/declined | 9 (1.6) | 5 (1.5) | 4 (1.9) | 0.71 |
Mean age, years (SD) | 53.6 (15.9) | 52.0 (15.8) | 56.3 (15.8) | 2.1 × 10−3 |
Mean BMI, kg/m2 (SD) | 27.6 (6.6) | 27.6 (6.7) | 27.6 (6.6) | 0.89 |
Mean eGFR, mL/min/1.73 m2 (SD) | 90.0 (28.2) | 94.7 (24.0) | 82.5 (32.6) | 4.2 × 10−6 |
Median glucocorticoid dose, mg of prednisone (IQR) | 50.0 (20.0–100.0) | 40.0 (17.5–60.0) | 72.5 (40.0–179.7) | 3.2 × 10−6 |
Median glucocorticoid duration, days (IQR) | 3 (1–11) | 4 (2–21) | 2 (1–4) | 0.10 |
. | All . | Without hyperglycemia . | With hyperglycemia . | P . |
---|---|---|---|---|
N of participants | 546 | 336 | 210 | — |
Sex, n (%) | ||||
Female | 326 (59.7) | 206 (61.3) | 120 (57.1) | 0.33 |
Male | 220 (40.3) | 130 (38.7) | 90 (42.9) | 0.33 |
Self-reported ethnicity, n (%) | ||||
Hispanic | 7 (1.3) | 6 (1.8) | 1 (0.5) | 0.19 |
Non-Hispanic | 505 (92.5) | 307 (91.4) | 198 (94.3) | 0.21 |
Unknown/declined | 34 (6.2) | 23 (6.8) | 11 (5.2) | 0.45 |
Self-reported race, n (%) | ||||
Asian | 6 (1.1) | 4 (1.2) | 2 (1.0) | 0.80 |
Black | 29 (5.3) | 16 (4.8) | 13 (6.2) | 0.47 |
White | 478 (87.5) | 296 (88.1) | 182 (86.7) | 0.62 |
Other | 24 (4.4) | 15 (4.5) | 9 (4.3) | 0.92 |
Unknown/declined | 9 (1.6) | 5 (1.5) | 4 (1.9) | 0.71 |
Mean age, years (SD) | 53.6 (15.9) | 52.0 (15.8) | 56.3 (15.8) | 2.1 × 10−3 |
Mean BMI, kg/m2 (SD) | 27.6 (6.6) | 27.6 (6.7) | 27.6 (6.6) | 0.89 |
Mean eGFR, mL/min/1.73 m2 (SD) | 90.0 (28.2) | 94.7 (24.0) | 82.5 (32.6) | 4.2 × 10−6 |
Median glucocorticoid dose, mg of prednisone (IQR) | 50.0 (20.0–100.0) | 40.0 (17.5–60.0) | 72.5 (40.0–179.7) | 3.2 × 10−6 |
Median glucocorticoid duration, days (IQR) | 3 (1–11) | 4 (2–21) | 2 (1–4) | 0.10 |
Continuous variables are presented as mean with SD in parentheses. Glucocorticoid dose and duration had skewed distributions, so these variables are presented as median (interquartile range [IQR]). P values represent the statistical significance of the deviation between those with and without hyperglycemia, as assessed by two-sample test of proportions (categorical variables) or t test (continuous variables).
To assess the relationship between genetic variants and GIH, we implemented a T2D global extended polygenic score, which captures the cumulative risk of T2D conferred by numerous variants (∼1 million) across the entire genome. In a logistic regression model, T2D polygenic score was significantly associated with GIH (P = 0.038), with an odds ratio (OR) of 1.44 per SD (95% CI 1.02–2.04) (Fig. 2A). Other significant covariates included eGFR (OR 0.87 per 10 mL/min/1.73 m2) and glucocorticoid dose (OR 1.08 per 50 mg prednisone). Age, sex, BMI, glucocorticoid duration, and principal components of genetic ancestry were not significantly associated with GIH. We tested for interactions in the logistic regression model between T2D polygenic score and each of the clinical covariates, but none of these interactions were statistically significant.
Factors that modulate the risk of glucocorticoid-induced hyperglycemia. A: We used a logistic regression model to test the association between a T2D global extended polygenic score and GIH, while simultaneously controlling for the listed covariates. The 95% CI is displayed for each data point. B: We constructed receiver operating curves to distinguish between individuals with and without GIH. The initial regression model (black) included only those covariates that were significantly associated with GIH (eGFR, glucocorticoid dose). The revised regression model (red) additionally included the T2D global extended polygenic score. The area under the curve (AUC) is displayed for each model. P value was obtained using the DeLong test.
Factors that modulate the risk of glucocorticoid-induced hyperglycemia. A: We used a logistic regression model to test the association between a T2D global extended polygenic score and GIH, while simultaneously controlling for the listed covariates. The 95% CI is displayed for each data point. B: We constructed receiver operating curves to distinguish between individuals with and without GIH. The initial regression model (black) included only those covariates that were significantly associated with GIH (eGFR, glucocorticoid dose). The revised regression model (red) additionally included the T2D global extended polygenic score. The area under the curve (AUC) is displayed for each model. P value was obtained using the DeLong test.
The T2D polygenic score improved the ability of a clinical model to predict GIH. In a baseline model including only significant covariates (eGFR and glucocorticoid dose), the area under the receiver operating curve was 0.65. After incorporating T2D polygenic score, the area under the receiver operating curve improved significantly to 0.68 (DeLong test P = 0.029) (Fig. 2B). Furthermore, in a linear regression model, T2D polygenic score was significantly associated with glucose level (increase of 8.9 mg/dL glucose per SD of polygenic score; P = 0.034).
As a sensitivity analysis, we repeated the logistic regression model after removing each covariate individually. In each instance, the relationship between T2D polygenic score and phenotype of GIH remained statistically significant (P < 0.05). In addition, we noted that designating fasting status based on time of day may have been inaccurate. Therefore, we repeated our analysis after removing 151 individuals with glucose level ≥126 mg/dL and <200 mg/dL, because the presence of hyperglycemia in these individuals depends on fasting status. The association between T2D polygenic score and GIH remained statistically significant (P = 0.038).
Finally, we implemented a restricted-to-significant polygenic score (comprising ∼400 variants that reached genome-wide significance in a prior T2D genome-wide association study [11]), but this polygenic score was not significantly associated with GIH (P = 0.46).
Conclusions
In this study, we demonstrated that individuals who do not have a diagnosis of diabetes but who carry an increased genetic risk of T2D have a higher risk of GIH, supporting the hypothesis that the two conditions share a common pathophysiology. Prior studies have suggested that individuals with GIH have an increased risk of developing T2D (12); indeed, the concept that glucocorticoids may be used as a stress test to predict T2D onset was first proposed many decades ago (13). Additionally, we previously demonstrated that variants in glucocorticoid-related genes are associated with T2D (14). Here, we used a complementary approach to demonstrate that variation in T2D-related genes is associated with GIH.
Prior studies have implicated various clinical factors that increase the risk of GIH, including dose and duration of glucocorticoids, age, and BMI (2). Here, we found that glucocorticoid dose was significantly associated with risk of hyperglycemia, whereas age and BMI were not significantly associated. Notably, glucocorticoid duration did not demonstrate a significant association with hyperglycemia; this could be because glucocorticoid dose and duration were inversely related or because our power was not large enough to detect an association with glucocorticoid duration. In addition, the median duration of exposure was relatively short (3 days), which may have limited our ability to detect a significant association. We also confirmed a previously reported association between decreased renal function and GIH (15), which may be related to decreased renal clearance of glucocorticoid metabolites (16).
The careful phenotyping of the cohort in the form of clinical record review is a strength of our study, allowing us to have confidence in our phenotype. In addition, we demonstrated the strength of a global extended polygenic score to fully capture T2D genetic risk. Notably, a more limited polygenic score was not significantly associated with GIH, consistent with previous evidence that restricted-to-significant polygenic scores are less powerful than global extended polygenic scores (7).
Nevertheless, caution must be taken when generalizing our findings. In particular, this study excluded participants with kidney disease, and a majority of participants were non-Hispanic White; however, polygenic scores derived mainly from cohorts with European ancestry may have decreased performance in other populations (17). In addition, a majority of early morning laboratory draws occurred in the inpatient setting, but findings in hospitalized participants may not translate to nonhospitalized individuals. Finally, we were unable to assess whether elapsed time between glucocorticoid administration and glucose measurement was associated with GIH.
Furthermore, it is well established that prediabetes is a strong predictor of T2D risk (18); unfortunately, very few participants had documented HbA1c levels before the administration of glucocorticoids, so we did not include prediabetes status in our prediction models. However, because our goal was to create a clinical tool to identify individuals at high risk of GIH, the polygenic score is useful even for individuals who are subsequently found to have prediabetes (or undiagnosed T2D).
Overall, we demonstrated that a T2D polygenic score can help predict the onset of GIH. As genetic information becomes more widely available, this tool may be used to help identify individuals at high risk of this complication, potentially prompting clinicians to reduce the dose of glucocorticoids, identify suitable therapeutic alternatives, or monitor more closely for hyperglycemia.
This article contains supplementary material online at https://doi.org/10.2337/figshare.23264648.
S.K. is currently affiliated with the Department of Surgery, Virginia Mason Medical Center, Seattle, WA.
Article Information
Funding. A.J.D. was supported by National Institutes of Health (NIH)/National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant T32DK007028. L.N.B was supported by NIH/NIDDK grant 1F32DK115086-01A1. J.M.M. was supported by American Diabetes Association Innovative and Clinical Translational Award 1-19-ICTS-068, American Diabetes Association grant 11-22-ICTSPM-16, and National Human Genome Research Institute grant U01HG011723.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. A.J.D. performed the statistical analyses and wrote the initial draft of the manuscript. A.J.D., S.K., F.E., and L.N.B. performed record reviews and extracted data for the study. A.J.D., J.C.F., and L.N.B. designed the study. P.H.S., R.M., and J.M.M. performed imputation and quality control for genomic data in the Mass General Brigham Biobank and created the T2D polygenic score. M.S.U., J.C.F., and L.N.B. supervised the study and edited the manuscript. All authors approved the final version of the manuscript. L.N.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this study were presented at the 83rd Scientific Sessions of the American Diabetes Association, San Diego, CA, 23–26 June 2023.