Bias Control

The main analysis used a 3-day lag after NSAID dispensing, with miscarriages during the lag counted as unexposed. The main Cox model treated NSAID exposure as time-dependent to account for exposure changes and avoid immortal time bias. The study addressed immortal time bias and protopathic bias as methodological problems in earlier research. Longer lag periods reduced estimated hazard ratios, but the association remained statistically significant for every lag tested. The lag-period pattern supports the importance of controlling protopathic bias, though lagging exposure could underestimate immediate NSAID effects. Surgeons and operating room staff cannot be blinded because the steerable sheath has a visible deflection lever.