Table 3 External validation in CPRD: Observed treatment effects across strata defined by clinical cut-offs of predicted treatment benefit. Estimates are adjusted for clinical features in the treatment selection model (to improve precision and control for potential differences in covariate balance within subgroups)
| Â | Observed treatment difference (mmol/mol; negative favors SGLT2-i) | ||||
|---|---|---|---|---|---|
Predicted HbA1c difference | N patients | Treatment difference (mean) | Lower 95% CI | Upper 95% CI | p-value |
Penalized regression model external validation | |||||
 Overall | 18,741 | -4.5 | -5.0 | -4.1 | < 0.001 |
 Strata | |||||
  SGLT2-i benefit by any mmol/mol | 15321 | -5.2 | -5.7 | -4.7 | < 0.001 |
  SGLT2-i benefit by ≥ 10 mmol/mol | 1173 | -10.7 | -13.0 | -8.4 | < 0.001 |
  SGLT2-i benefit by 5–10 mmol/mol | 4503 | -7.6 | -8.6 | -6.6 | < 0.001 |
  SGLT2-i benefit by 3–5 mmol/mol | 3517 | -5.0 | -5.9 | -4.0 | < 0.001 |
  SGLT2-i benefit by 0–3 mmol/mol | 6128 | -2.3 | -3.0 | -1.6 | < 0.001 |
  DPP4-i benefit by any mmol/mol | 3420 | 0.3 | -0.7 | 1.3 | 0.506 |
  DPP4-i benefit by 0–3 mmol/mol | 2972 | 0.0 | -1.0 | 1.1 | 0.937 |
  DPP4-i benefit by ≥ 3 mmol/mol | 448 | 2.9 | -0.6 | 6.5 | 0.107 |
Causal forest external validation | |||||
 Overall | 18741 | -4.5 | -4.9 | -4.0 | 0.003 |
 Strata | |||||
  SGLT2-i benefit by any mmol/mol | 18689 | -4.5 | -5.0 | -4.1 | < 0.001 |
  SGLT2-i benefit by ≥ 10 mmol/mol | 0 | NA | NA | NA | NA |
  SGLT2-i benefit by 5–10 mmol/mol | 2175 | -8.7 | -10.0 | -7.3 | < 0.001 |
  SGLT2-i benefit by 3–5 mmol/mol | 8676 | -6.0 | -6.6 | -5.3 | < 0.001 |
  SGLT2-i benefit by 0–3 mmol/mol | 7838 | -1.1 | -1.7 | -0.4 | 0.001 |
  DPP4-i benefit by any mmol/mol | 52 | 10.3 | -8.6 | 29.3 | 0.269 |
  DPP4-i benefit by 0–3 mmol/mol | 52 | 10.3 | -8.6 | 29.3 | 0.269 |
  DPP4-i benefit by ≥ 3 mmol/mol | 0 | NA | NA | NA | NA |