In Reply We agree with Dr Bhupathiraju and colleagues that debates over analytic approach are often surrogates for debating which scientific questions are of greatest interest. That said, they take issue with a sentence in our Editorial that characterized the difference between the original NHS analyses1,2 and the 2008 reanalysis3 as due to a design bias rather than a difference in causal question. They question whether the differences observed might be due to effects of age and estrogen type rather than the distinctions we drew between analyses of new users vs current users.
Goodman SN, Schneeweiss S. Posing Causal Questions When Analyzing Observational Data—Reply. JAMA. 2017;318(2):201–202. doi:10.1001/jama.2017.6235