Recommendations to improve the state of the literature

- Investigate the distributional tails of AFB, especially very early and very late age at first birth.  - Justify AFB exposure categorization. How were categories selected (theory, biological evidence, statistical considerations, etc.)?  - Reference category selection should reflect underlying hypotheses about exposed and unexposed groups.  - Assure sufficient sample size and follow up time, taking into account cohort-age (fewer events should be expected for younger cohorts). For smaller studies, a priori power calculations are advisable.  - Improved use of theory to direct covariate choice and modelling approaches. Where appropriate, future research would benefit from more sophisticated use of tools for investigating pathways (e.g. directed acyclic graphics, mediation modelling, structural equation modelling, etc.)  - Ideally, prospective studies that start well before reproductive age could help answer questions about selection effects (for example, are observed associations between early AFB and CVD linked to some common earlier life event such as obesity in childhood or a “risk-taking” personality)