Table 3.
Probability of antidiabetic drug dispensation after index date in naïve patients
| Dementia versus Dementia-free (Whole-cohort analysis) | ||
| Cox regression Adjusted HR (95% CI) | Competing risk regression Adjusted sHR (95% CI) | |
| Insulin (13,755 new dispensations) | 1.28 (1.21–1.35) † | 1.22 (1.15–1.29) † |
| Metformin (4,734) | 1.03 (0.93–1.15) | 1.02 (0.91–1.14) |
| Sulfonylureas (3,832) | 0.93 (0.82–1.05) | 0.90 (0.80–1.03) |
| TZD (335) | 0.64 (0.40–1.03) | 0.62 (0.39–1.00) |
| DPP-4i (9,802) | 0.88 (0.81–0.95)* | 0.84 (0.78–0.90) † |
| GLP-1a (2,096) | 0.59 (0.49–0.71) † | 0.56 (0.47–0.67) † |
| SGLT-2i (2,794) | 0.54 (0.45–0.64) † | 0.50 (0.43–0.60) † |
| Dementia versus Dementia-free (PS-matched analysis) | ||
| Cox regression Crude HR (95% CI) | Competing risk regression Crude sHR (95% CI) | |
| Insulin (2,543 new dispensations) | 1.29 (1.19–1.40)† | 1.21 (1.11–1.31)† |
| Metformin (829) | 0.90 (0.79–1.03) | 0.88 (0.77–1.02) |
| Sulfonylureas (560) | 0.95 (0.80–1.12) | 0.90 (0.76–1.07) |
| TZD (45) | 0.72 (0.40–1.30) | 0.68 (0.38–1.23) |
| DPP-4i (1,789) | 0.75 (0.69–0.83)† | 0.72 (0.66–0.79)† |
| GLP-1a (370) | 0.54 (0.44–0.67)† | 0.51 (0.41–0.63)† |
| SGLT-2i (462) | 0.48 (0.39–0.58)† | 0.44 (0.36–0.54)† |
HR, hazard ratio; sHR, subdistribution hazard ratio; CI, confidence interval; TZD, thiazolidinediones; DPP-4i, dipeptidyl-peptidase-4 inhibitors; GLP-1a, glucagon-like peptide-1 agonists; SGLT-2i, sodium-glucose cotransporter-2 inhibitors; PS, propensity score; Table represents time-to-first-dispensation of specific antidiabetic drugs with dementia status as primary exposure; Only subjects without history of specific antidiabetic drug use prior to index date were considered in the analyses (naïve subjects); Attained-age was used as time-scale; Death was considered competing event; Whole-cohort analyses were adjusted for dementia status, index year, sex, diabetes duration and type, Charlson comorbidity index, renal failure, cardiovascular, antithrombotic, antipsychotic, antidepressant, hypnotic/sedative and anxiolytic drugs, education, income group, usage of other antidiabetic drugs and index year; Other antidiabetic medication was considered at the time or prior to index date; Non-diabetic medication use was determined in the interval index date and three-years prior; PS-matched models included one predictor - dementia status; Matching variables included variables used in the whole-cohort analyses, with addition of age; *p < 0.05; †p < 0.001.