To the editor—In your May issue, John Bell highlights the demise of traditional clinical research in recent decades and the move from bedside, clinical research to benchside molecular biological research1. He describes the extent to which the limitations of traditional clinical research—"the measurement of complex variables in crudely defined disease states"—has driven the move to molecular biology that seeks insights into disease causation. Surprisingly, he seems not to have noticed an equally dramatic shift from bedside, clinical research to population-based epidemiological research.

The main limitation of traditional clinical research in addressing the causes of disease was not the lack of molecular biology but the focus on cases (numerators), without reference to the population (denominators) from which cases are derived. To elucidate the causes of disease we focus more and more on epidemiological methods. We must measure and compare disease incidence rates in defined populations among persons exposed and not exposed to putative causal factors (including gene markers). Perusal of the major peer-reviewed clinical research journals over the last two decades confirms the extent to which traditional bedside research has been displaced by epidemiological research on causation. Molecular biology, in contrast, is crucial in the elucidation of disease mechanisms (but contributes to our understanding of causation by refining case definitions in heterogeneous syndromes, facilitating work on gene-environment interaction and enhancing the precision of exposure measurements).

Bell reminds us that all disease depends on an interaction between genetic and environmental causal factors. We need to move, therefore, from simplistic, hierarchical concepts of causation that ascribe a more fundamental role to genes than environment in the development of disease2. The mysteries of human disease clearly will not yield to laboratory based investigation of cells and molecules alone. The clinical researcher of the future (or the researcher concerned with clinical questions) will need to combine biological, clinical and epidemiological approaches to study the full spectrum of disease in defined populations. It would also be helpful to discard naïve concepts of what constitutes 'basic' and 'applied' research. We should instead think in terms of 'laboratory', 'clinical' and 'population-based' research, each of which may be basic or applied, depending on the nature of the problem addressed. Clinical research might then benefit from the best of three worlds.