A large retrospective cohort study evaluates the association between severe nocturnal continuous positive airway pressure (CPAP) non-adherence and the 5-year incidence of major adverse cardiovascular events (MACE).
Initial analysis reveals a crude relative risk (RR) of 2.8 (95% CI: 2.1 - 3.7) for MACE in CPAP non-adherent patients compared to adherent patients. The investigators perform stratified analyses for Body Mass Index (BMI) and concurrent beta-blocker therapy:
| Stratification Variable | Stratum | Relative Risk (95% CI) |
|---|---|---|
| BMI | 35 kg/m | 1.1 (0.8 - 1.4) |
| 35 kg/m | 1.2 (0.9 - 1.6) | |
| Beta-blocker therapy | Active use | 1.0 (0.7 - 1.3) |
| No use | 4.5 (3.2 - 6.1) |
Based on these stratified analyses, which of the following biostatistical approaches are mathematically appropriate for handling these variables in the final analysis? (Select all that apply)
- For BMI, report a single adjusted relative risk using multivariable regression or Mantel-Haenszel pooling.Answer
- BFor concurrent beta-blocker use, calculate and report an adjusted relative risk to control for its confounding effect.
- For concurrent beta-blocker use, report the stratum-specific relative risks separately to demonstrate the differing association.Answer
- DFor BMI, report the stratum-specific relative risks separately because the variation between the crude and stratified risks indicates effect modification.
- EReport the crude relative risk of 2.8 as the primary finding without adjustment, as stratification introduces selection bias.
Answer
The researchers should report a single adjusted relative risk for BMI (a confounder) and report stratum-specific relative risks separately for beta-blocker use (an effect modifier).
BMI is a confounder because the crude relative risk (2.8) differs significantly from the stratified relative risks (1.1 and 1.2), which are similar to each other. In the analysis phase, confounding is managed by calculating an adjusted (pooled) estimator to remove the distortion. Conversely, beta-blocker use is an effect modifier because its stratified relative risks (1.0 and 4.5) differ significantly from each other. Effect modifiers represent true interactions and should never be pooled; instead, their stratum-specific risks must be reported separately to show how the exposure's effect changes based on the modifier.
Step-by-Step Solution
Key Concept
Distinguishing Confounding from Effect Modification and applying appropriate statistical reporting methods for each.