Editorial: There are no shortcuts to finding out what works – population laboratories are essential tools

There is no doubt that cardiovascular diseases (CVD) are fast becoming a major cause of mortality and morbidity in many transitional and low-income countries (Beaglehole 1992). In poorer countries, the loud political voice of relatively affluent urban dwellers is leading to the diversion of scarce health care resources into secondary care for noncommunicable chronic diseases. These pressures have led to the familiar suggestion that we should focus our efforts on primary prevention of cardiovascular disease, on the assumption that this will be less expensive than dealing with the consequences of established disease (Pearson et al. 1993). Community-based CVD prevention programmes, along the lines of the North Karelia study (Puska et al. 1976), have been put forward as the best means of achieving disease control. The proponents of setting up similar preventive programmes in poorer countries consider it unnecessary to measure the impact of such programmes in terms of reduction in mortality and morbidity (as was done in the North Karelia and other studies). Simply evaluating the programme effects on risk factor changes would, they suggest, be a sufficient guide to their impact and thus guide policy. However, despite their assertions, CVD shares many characteristics of the interventions and diseases that have been successfully evaluated using population laboratories: preventive interventions are given at defined times, CVD outcomes occur within a few years of intervention and are relatively frequent, and the relationship between risk factor change and outcome is generally assumed to be linear. There is a consensus over the importance of reducing cardiovascular risk factors (e.g. smoking cessation, weight control, dietary change and increased physical activity) in controlling cardiovascular diseases, and this flows from the substantial body of epidemiological data demonstrating associations between changes in risk factors and reductions in disease events. There is no consensus on community-based CVD prevention programmes as the best means of CVD control. We do not have a body of knowledge that amounts to ‘known preventive measures’ which can be applied and will achieve sufficient risk factor control through lifestyle modification at either an individual or population level. The community programmes that adopted quasi-experimental designs demonstrated concurrent reductions in risk factors and mortality in both intervention and control regions (and this has occurred in both the original (Valkonen 1992) and later, better designed studies(Farquhar et al. 1990; Ebrahim & Davey Smith 1996). Furthermore, the substantial number of randomized controlled trials of lifestyle intervention (often augmented by antihypertensive and lipid-lowering drugs) failed to demonstrate a clear impact on either risk factors or disease events ( Ebrahim & Davey Smith 1997). There are three major agendas for the control of CVD in poorer countries. The first concerns the implementation of fiscal and legal measures – termed health protection– for the control of tobacco products, a battle that requires constant vigilance as tobacco profits decline in more affluent countries and tobacco companies turn their attention to expanding markets of the youth of poorer countries. The second concerns relatively small-scale research on understanding the determinants of behaviour change in individuals and groups in an effort to find more effective means of persuasion which would then need testing in clinical trials. The third area concerns establishing the reasons for the massive secular trends that are occurring in cardiovascular disease risk. Understanding secular trends demands collaborative research between affluent and poorer countries and requires population laboratories. Far from being too inappropriate and expensive to contemplate, population laboratories are the essential building blocks of rational approaches to making health and social policy. Issues of costs of maintenance and the size of population laboratories have to be considered in the light of their multiple uses to make and monitor policy in a wide range of diseases and health problems. The major costs are those of maintaining demographic surveillance, counting fatal events and assigning causes to deaths. The sample sizes of population laboratories required to give reasonable estimates of disease risk will depend on the uses to which the data will be put. If complete populations are surveyed, it is apparent from experience in Matlab in Bangladesh (Fauveau 1990) that a population of 200 000 is both feasible to maintain over many years and can provide invaluable information on topics as diverse as maternal and infant mortality to evaluation of fertility control methods, to mortality among adults living alone. Such a population would also yield a very substantial number of CVD deaths (as many as 500 per year) which could be used both for monitoring and evaluation of interventions. It is worth remembering that the Framingham study (Dawber & Kannel 1966), which provided the early estimates of CVD risk, comprised less than 5000 people and not hundreds of thousands! Future population laboratories might well be more modest in their size and if they were focused purely on noncommunicable disease in adults, would require sample sizes of the order of 10 000 people and follow-up of around five years to generate useful amounts of information. Substituting ‘action’ research, risk factor surveys, demonstration projects, and process evaluations would not necessarily be any cheaper and might fail to deliver interpretable data. Indeed the costs of laboratory assays and standardization of measurement of risk factors such as blood pressure are expensive and technically difficult. By comparison, counting deaths is a relatively easy task with much less room for imprecision and inaccuracy. In Mauritius, evaluation of the effects of a noncommunicable disease intervention programme using sequential cross-sectional surveys 5 years apart showed large changes in blood cholesterol levels; but making a causal inference was difficult, as changes in the composition of a widely used cooking oil occurred concurrently (Dowse et al. 1995). That study was neither cheap to conduct nor capable of judging whether the intervention was of benefit or not (a case of poor value for money?) In the affluent countries of the north, we are now on a sharp downward slope where CVD mortality risk is falling by 4–5% per year (Charlton et al. 1997). Medical care cannot take the credit and falls in risk factors probably explain, at most, only half the decline (Davey Smith 1997; Hunink 1997). In poorer countries, and interestingly in the socio-economically fragmented countries of central and eastern Europe, CVD mortality rates are rising. Unravelling the causes of these secular trends is much more feasible when trends in different populations are in opposite directions. This work could ultimately bring us closer to realizing the goal of better means of controlling the burdens of CVD at a population level.