To what extent can trends in risk factor levels in populations explain trends in disease rates? Some methodological difficulties in answering this question include weaknesses of ecological studies, small sample size, time lags and multiple sources of variability. Two possible approaches are to apply predictive equations derived from longitudinal studies of individuals to average data for populations, or to use results from overviews of randomized, controlled trials and cohort studies to estimate the magnitude of effects achievable through various risk factor changes. These methods are illustrated with Australian data on trends in mortality from ischaemic heart disease and cigarette smoking, blood pressure and cholesterol during the 1980s. If it is assumed that time lags are short between reductions in risk factor levels and reductions in risk, then both methods suggest that about half the mortality decline in men, and less in women, may be related to changes in these three well-established risk factors. Improvements in population levels of blood pressure and, for men, reductions in smoking appear to be the main contributors. It is argued that such ecological analyses can provide valid results as long as care is taken to avoid various epidemiological pitfalls.