New models for large prospective studies: is there a risk of throwing out the baby with the bathwater?

Manolio et al. (Am J Epidemiol. 2012;175:859-866) proposed that large cohort studies adopt novel models using "temporary assessment centers" to enroll up to a million participants to answer research questions about rare diseases and "harmonize" clinical endpoints collected from administrative records. Extreme selection bias, we are told, will not harm internal validity, and "process expertise to maximize efficiency of high-throughput operations is as important as scientific rigor" (p. 861). In this article, we describe serious deficiencies in this model as applied to the United States. Key points include: 1) the need for more, not less, specification of disease endpoints; 2) the limited utility of data collected from existing administrative and clinical databases; and 3) the value of university-based centers in providing scientific expertise and achieving high recruitment and retention rates through community and healthcare provider engagement. Careful definition of sampling frames and high response rates are crucial to avoid bias and ensure inclusion of important subpopulations, especially the medically underserved. Prospective hypotheses are essential to refine study design, determine sample size, develop pertinent data collection protocols, and achieve alliances with participants and communities. It is premature to reject the strengths of large national cohort studies in favor of a new model for which evidence of efficiency is insufficient.