Table 2.
Comparison of treatment effects for binary outcome in case study
| Method of estimation | Estimated treatment effect (95% confidence interval) |
|---|---|
| Odds ratio | |
| Logistic regression adjustment | 0.73 (0.64, 0.83) |
| Relative risk | |
| Zhang-Yu substitution method | 0.79 (0.72, 0.87) |
| Poisson regression | 0.81 (0.73, 0.90) |
| Poisson regression – sandwich variance estimation | 0.81 (0.74, 0.89) |
| Conditional standardization by centering covariates | 0.77 (0.68, 0.86) |
| Marginal probabilities | 0.81 (0.75, 0.88) |
| Propensity score matching | 0.79 (0.71, 0.89) |
| Propensity score stratification | 0.80 (0.73, 0.89) |
| Inverse probability of treatment weighting | 0.81 (0.73, 0.91) |
| Risk difference | |
| Bender and Blettner | −0.063 (−0.087, −0.040) |
| Marginal probabilities | −0.054 (−0.076, −0.034) |
| Imbens | −0.053 (−0.076, −0.029) |
| Propensity score matching | −0.055 (−0.082, −0.029) |
| Propensity score stratification | −0.057 (−0.081, −0.033) |
| Inverse probability of treatment weighting | −0.054 (−0.077, −0.031) |
| Number needed to treat | |
| Bender and Blettner | 16.0 (11.5, 26.4) |
| Marginal probabilities | 18.5 (13.2, 29.4) |
| Imbens | 18.9 (13.2, 34.5) |
| Propensity score matching | 18.2 (12.2, 34.5) |
| Propensity score stratification | 17.5 (12.3, 30.3) |
| Inverse probability of treatment weighting | 18.5 (13.0, 32.3) |