Observational Studies Analyzed Like Randomized Experiments

We present comparisons between these estimates and previously reported Nurses’ Health Study estimates

Miguel A. Hernán; Alvaro Alonso; Roger Logan; Francine Grodstein; Karin B. Michels; Walter C. Willett; JoAnn E. Manson; James M. Robins


Scholarcy highlights

  • The Women’s Health Initiative randomized trial found greater coronary heart disease risk in women assigned to estrogen/progestin therapy than in those assigned to placebo
  • If the true adherence-adjusted hazard ratio is constant over time, this method produces valid estimates provided that discontinuing the baseline treatment is unrelated to unmeasured risk factors for CHD incidence within levels of the covariates, and that the logistic model used to estimate the inverse probability weights is correctly specified
  • We excluded 14,764 women who received a form of hormone therapy other than oral estrogen plus progestin in all of the Nurses’ Health Study “trials,” or did not provide information on the type of hormone therapy in any of the “trials.” Of the remaining 63,030 women, we excluded 17,146 who received hormone therapy in the two years before the baseline of all the “trials.” Of the remaining 45,884 women, we excluded 11,309 who did not have an intact uterus in 1984
  • The estimated average hazard ratio of CHD for initiators versus noninitiators was 0.96 when the entire follow-up time was included in the analysis
  • When the start of follow-up is defined in this way, the selected group of initiators differs from the initiator group in Table 3 because it does not include women who, during the 2 year-interval before the baseline questionnaire, either initiated and stopped hormone therapy or survived a CHD event occurring after initiation
  • The ITT hazard ratios of CHD for therapy initiation were 1.42 in the NHS versus 1.68 in the WHI9 during the first two years, and 1.00 in the NHS versus approximately 1.24 in the WHI8 during the first 8 years
  • As a consequence, observationalrandomized discrepancies cannot be automatically attributed to randomization itself

Need more features? Save interactive summary cards to your Scholarcy Library.