Second Generation Antipsychotics and Diabetes Outcomes in Patients with Depression and Diabetes

This dissertation evaluated the use of and comparative effects of second generation antipsychotics (SGAs) and non-SGA depression pharmacotherapies [bupropion, lithium, mirtazapine, tricyclic antidepressants (TCAs) and thyroid hormone] on diabetes outcomes in patients with pre-existing diabetes and depression who previously used a selective serotonin reuptake inhibitor or selective norepinephrine reuptake inhibitor. Three clinical studies compared 1) non-adherence and non-persistence to SGA versus non-SGA depression therapies, 2) non-adherence and non-persistence to oral antidiabetic drugs (OAD) between SGA and non-SGA users, and 3) SGA and non-SGA users on the risk of diabetes-related hospitalization or diabetes drug intensification. Use of SGAs was associated with a 1.8 times higher odds of non-adherence and a 1.4 times higher risk of non-persistence to therapy compared to non-SGA use. Also, SGA users had a 30-40% higher odds of a 10% or greater decline in OAD adherence compared to non-SGA users, while persistence to OAD therapy was similar between groups. Risk of diabetes of diabetes-related hospitalization or diabetes drug intensification was no different comparing SGA versus non-SGA users; however, there were differences when comparing specific treatment subgroups: bupropion was associated with a 15% reduced risk of diabetes-related hospitalization or treatment intensification compared to TCAs, quetiapine was associated with a 18% reduced risk of events compared to mirtazapine, and quetiapine was associated with a 16% reduced risk of events compared to TCAs. Differences for other subgroup comparisons between aripiprazole, quetiapine, bupropion, mirtazapine and TCAs were small and non-significant. Future studies are needed to access the impact of SGA and non-SGA therapies on other diabetes outcomes, including hemoglobin A1c, diabetic complications, and mortality. A fourth methodological study assessed the performance of full-cohort high dimensional propensity score (HDPS) matching approaches versus subgroup-specific HDPS approaches. The full-cohort HDPS matching approaches sometimes resulted in non-overlapping propensity score distributions, more imbalance in potential confounders, and greater than 10% change in effect estimates compared to the subgroup-specific approach. Therefore, examining covariate balance after matching to ensure that patient subgroups are balanced with respect to potential confounders is recommended if one of the full-cohort HDPS approaches are used.