Association Between Cannabis Use and Risk for Diabetic Ketoacidosis in Adults With Type 1 Diabetes

Cannabis use is increasing with the shifts in legality and public perceptions in the United States.¹ Studies have reported improvement in insulin sensitivity and pancreatic beta cell function with cannabis use,^2,3 generating widespread media attention suggesting cannabis as a potential therapeutic agent for treatment of type 2 diabetes. By contrast, we published a case series⁴ reporting recurrent diabetic ketoacidosis (DKA) with cannabis use in patients with type 1 diabetes (T1D). Because little is known about cannabis use and its contribution to DKA in T1D, we investigated the characteristics of cannabis use among adults with T1D and the association of cannabis use with DKA.

Between June 2017 and January 2018, adults aged 18 years or older with T1D attending the Barbara Davis Center for Diabetes, the largest T1D treatment center in Colorado, where cannabis is legal for medical and recreational use, were invited to complete an in-person questionnaire on their cannabis use. Patients with diabetes other than T1D, pregnancy, and repeat follow-up visits within the study duration were excluded. A questionnaire was used to collect demographic characteristics, diabetes history and complications, severe hypoglycemia requiring assistance, and cannabis use information. Point-of-care hemoglobin A_1c level (HbA_1c; DCA Vantage Analyzer) was measured during the clinic visit. Scores on the Cannabis Use Disorder Identification Test-Revised⁵ were used to define hazardous cannabis use (score ≥8 and <12) and possible cannabis use disorder (score ≥12). The Colorado Multiple Institutional Review Board (Aurora, Colorado) approved this study, and all participants provided written informed consent.

The primary outcome was DKA hospitalization during the preceding 12 months. All self-reported DKA hospitalizations were confirmed by medical record review. Comparison of categorical variables was conducted with 2-tailed χ² tests, and 2-sample t tests were used to test normally distributed continuous variables. A logistic regression model was built to calculate the odds of DKA hospitalization by cannabis use. Clinical evidence-based risk factors for DKA, such as age, sex, diabetes duration, income, educational level, HbA_1c level, and insurance (derived from income), were modeled, and a stepwise selection method was used to confirm the final model. Model fit was assessed by Akaike information criterion. Sensitivity analyses were performed using propensity score matching of cannabis users and nonusers, adjusting for age, sex, ethnicity, tobacco and alcohol use, educational level, income, employment, diabetes duration, HbA_1c level, and use of diabetes technology. Residual confounding was controlled by including significant variables in the final logistic regression model. All analyses were conducted using SAS, version 9.4 (SAS Institute Inc). A 2-sided P < .05 was considered statistically significant.

Of the 631 eligible patients, 450 (71.3%) responded to the survey. The mean (SD) age (36.8 [13.8] vs 34.6 [11.9] years, P = .07), sex (female, 201 [44.6%] vs 86 [47.5%]; P = .10), mean (SD) diabetes duration (19.5 [12.9] vs 18.3 [12.1] years, P = .22) and mean (SD) HbA_1c level as percent of total hemoglobin (7.8% [1.8%] vs 7.8% [1.2%], P = .72; to convert HbA_1c to proportion of total hemoglobin, multiply by 0.01) were similar between responders and nonresponders. Among 450 T1D participants, 134 (29.8%) reported using cannabis. Cannabis use duration, frequency, and form are shown in Table 1. Cannabis users were younger (entire cohort mean [SD] age, 31.3 [11.1] vs 39.1 [14.2] years), had lower income and educational levels, and shorter duration of diabetes (entire cohort mean [SD] duration, 16.3 [11.2] vs 20.9 [13.6] years) than nonusers (Table 1).

In adults with T1D, cannabis use within the previous 12 months was associated with an increased risk of DKA compared with no cannabis use (entire cohort OR, 1.98; 95% CI, 1.01-3.91) (Table 2). Compared with nonusers with T1D, the mean (SD) HbA_1c level (8.4% [2.0%] vs 7.6% [1.6%], P < .01) was higher but there was no difference in severe hypoglycemia (21 of 134 [15.6%] vs 64 of 316 [20.3%], P = .17) among cannabis users with T1D. Cannabis users had a mean 0.41% higher HbA_1c level than nonusers when adjusted for insulin delivery method, income, and age (β, 0.41; 95% CI, 0.38-0.43).

In the present survey study, the prevalence of cannabis use was 30% among adults with T1D. Cannabis use was associated with a higher risk for DKA in this population. Cannabinoids alter gut motility and cause hyperemesis, which may play a role in increased risk for DKA in T1D.^4,6 A small sample size, single-center, and self-reported diabetes outcomes are limitations of the present study. The possibility of unmeasured confounders, such as access to health care, cannot be excluded. Further research is needed to confirm these findings and understand the effects and adverse consequences of cannabis use in patients with T1D.

Accepted for Publication: August 3, 2018.

Corresponding Author: Viral N. Shah, MD, Barbara Davis Center for Diabetes, University of Colorado Anschutz Medical Campus, 1775 Aurora Ct, Room M20-1318, Aurora, CO 80045 (viral.shah@ucdenver.edu).

Published Online: November 5, 2018. doi:10.1001/jamainternmed.2018.5142

Author Contributions: Dr Shah had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Akturk, Camsari, Shah.

Acquisition, analysis, or interpretation of data: Akturk, Taylor, Rewers, Kinney, Shah.

Drafting of the manuscript: Akturk, Taylor, Camsari, Rewers, Shah.

Critical revision of the manuscript for important intellectual content: Akturk, Rewers, Kinney, Shah.

Statistical analysis: Taylor, Kinney, Shah.

Administrative, technical, or material support: Rewers.

Supervision: Kinney, Shah.

Conflict of Interest Disclosures: Dr Shah reported receiving research support through the University of Colorado from Sanfo US, Dexcom Inc, Eyenuk Inc, and T1D Exchange clinic registry and consulting fees through the University of Colorado from Sanofi US and Dexcom Inc, all outside the submitted work. No other disclosures were reported.