Ecological studies are a poor means of testing aetiological hypothesesBMJ 1998; 317 doi: https://doi.org/10.1136/bmj.317.7159.678 (Published 05 September 1998) Cite this as: BMJ 1998;317:678
- Shah Ebrahim, Professor of clinical epidemiology,
- George Davey Smith, Professor of clinical epidemiology
- Royal Free Hospital School of Medicine, University College London Medical School, London NW3 2PF
- Department of Social Medicine, University of Bristol, Bristol BS 2PR
EDITOR—The declines in cardiovascular diseases in Poland, “apparently without precedent in peacetime,”1 are similar to the declines seen in Britain over the same period.2 The table shows secular declines in mortality from cardiovascular diseases in Britain: between 1991 and 1994, death rates from both ischaemic heart disease and stroke fell dramatically, particularly among men at all ages and women aged over 55 years. In Britain these death rates are continuing to fall, in contrast to Poland. Furthermore, as mortality from ischaemic heart disease in Britain plateaued between 1980 and 1985 and then fell, the declines observed between 1985 and 1989 were as striking as those in Poland: 18.5% and 15% for men and women aged 55-64 years respectively, compared with Polish declines between 1991 and 1994 of 20% and 16% for men and women aged 45-64 years respectively. In contrast, mortality from stroke in Poland did not fall nearly so much as in Britain: 4% for men and no change in women. Over such a short time, these trends are unlikely to be due to artefacts of coding or practice of certification in either country.
It is probable that a lag will occur between secular trends in disease rates and changes in any explanatory risk factors. In Poland consumption of animal fats fell and consumption of vegetable fats and fruit rose at the same time as mortality declined, making these dietary intakes less plausible explanatory factors. In Britain mortality from ischaemic heart disease has been falling since the 1980s, suggesting that explanatory changes in risk factors must have been occurring before this time. Between 1950 and 1980, personal consumption of total fats fell by only 3% and consumption of fresh fruit increased by 45%; more striking was the 19-fold increase in consumption of poultry. Social and economic changes were also marked during this period: the rate of car ownership increased almost fourfold, and unemployment rose 4.5-fold.3 From this evidence, it is unlikely that anyone would conclude that increase in poultry consumption or rising unemployment explained falls in rates of heart disease.
Ecological analyses are a weak means of testing aetiological hypotheses, and only limited inferences can be made about concurrent (or lagged) changes in possible risk factors. Using individual data on risk factors and clinical events, it is probable that about half of the declining secular trend may be explained by concomitant falls in major risk factors for cardiovascular disease (that is, blood pressure, smoking, and blood cholesterol).4 It is also likely that trends in both risk factors and mortality are due to social and economic factors that operate at a population, rather than individual, level and also due to the lagged effects of such socioeconomic change. The contributions of specific risk factors to secular trends in cardiovascular disease are likely to vary between countries and time periods, and be affected by the proportion of the population exposed, making causal inferences of general applicability difficult.5