Descriptive statistics of patients on T2DM second-line drug classes for the HbA1c outcome
| Drug class | Patients (n) | Treatment change* | Missing outcome* | Average age† | % Female* |
| Sulfonylurea | 26 684 | 4336 (16%, 3e−152) | 12 269 (46%, –) | 61.2 (2e−98) | 47.7% (2e−15) |
| Thiazolidinedione | 4794 | 1145 (24%, 2e−12) | 2235 (47%, –) | 59.6 (0.001) | 48.2% (–) |
| Glucagon-like peptide-1 receptor agonists | 1532 | 398 (26%, 4e−9) | 735 (48%, –) | 52.8 (5e−113) | 66.6% (3e−44) |
| Dipeptidyl peptidase 4 | 7861 | 2314 (29%, 3e−118) | 3405 (43%, 5e−6) | 58.9 (2e−32) | 51.1% (8e−5) |
Per-confounder statistics appear in online supplementary table 2.
*Proportion test. Missing entries (–) are not significant with FDR <0.05.
†Wilcoxon rank-sum test. Missing entries (–) are not significant with FDR <0.05.
FDR, false discovery rate; HbA1c, glycated hemoglobin; T2DM, type 2 diabetes mellitus.