Generalizing Findings in Nested Cohorts

Area of Science:

• Reproductive Health
• Epidemiology
• Biostatistics

Background:

• Research findings often derive from nested cohorts, limiting generalizability to the original population.
• Generalizing results from subsets back to the parent cohort is infrequently performed.

Purpose of the Study:

• To demonstrate a "proof of concept" for generalizing findings from a nested cohort.
• To apply simple methods to generalize the association between prepregnancy obesity and preeclampsia.

Main Methods:

• Utilized the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers to Be (nuMoM2b) as the target sample.
• Employed the Heart Health Study (nuMoM2b-HHS) as a source sample, a nested subset with extended follow-up.
• Constructed inverse probability of selection weights to generalize findings from the source to the target sample.

Main Results:

• Estimated risk ratios for obesity and preeclampsia in the target sample (2.0 [1.7, 2.3]).
• Estimated risk ratios in the source sample (2.2 [1.8, 2.6]).
• Estimated risk ratios in the weighted source sample generalized to the target sample (2.0 [1.6, 2.4]).

Conclusions:

• Simple methods, such as inverse probability weighting, can successfully generalize findings from nested cohorts.
• Researchers should consider and implement methods for generalizing nested cohort findings when applicable.
• Accurate generalization enhances the external validity of epidemiological research.