Design Strategies for Generating Highly Influential Studies

Abstract

BACKGROUND AND AIM: When epidemiologists plan a study, they aspire to make that study as influential as possible when the totality of the research on that topic is assessed. Expert assessments constitute an essential step in the pathway by which research informs policy. Where prior studies are nonexistent or limited, the expectations are lower than for extensively studied topics since the magnitude of contribution is measured in relation to what has come before. METHODS: In environmental epidemiology, the primary limiting factors and opportunities for marked advancement of knowledge tend to follow a predictable pattern, suggesting where advances are needed and would have substantial influence on the integrated assessment of the evidence if implemented effectively: RESULTS:1) Identification of populations with well-defined, large exposure contrasts when most research has considered a narrow range of exposure, illustrated by arsenic and bladder cancer. 2) Selecting a study setting with an atypical confounding structure when previous studies have been unable to fully adjust for strong, difficult-to-measure confounders, illustrated with research on air pollution and pregnancy outcomes in New York City. 3) Conducting studies that examine a clinically significant endpoint when previous studies have largely focused on subclinical outcomes, illustrated with studies of temperature and stillbirth. 4) Demonstrating biological responses associated with low-level exposures that are suspected of causing adverse effects, illustrated with early work on air pollution and cardiovascular disease endpoints. CONCLUSIONS:It is also useful to note standard strategies that rarely make major contributions such as pure replication, simply making the study larger, refining standard instruments for data collection such as questionnaires, or expanding the array of covariates. Greater attention should be given to the likely influence of a given study, assessing whether what is feasible is what is needed. Flawed studies may be highly contributory and high quality studies may add little depending on the context. KEYWORDS: Methods environmental epidemiology, Methods methodological study design