Prevention of cardiovascular diseases
– a scientific dilemma
Av Dag Thelle, Göteborg

---

Prevention is better than cure, Galen said almost 2000 years ago. We still quote him, even if the statement should be assessed in light of his time and ability to cure. Given the quality and capability of curative methods at Galen’s time, preventive efforts would be the preferable option when choosing between prevention and cure for most diseases.

    The situation with regard to cardiovascular disease (CVD) today is more difficult to assess. As the present methods for treating coronary heart disease (CHD) have radically changed the outlook for the patients, what is the current status of preventive cardiology?
    The question may be more complicated than just an assess-ment of the rationality of preven-tion versus cure as a number of reports have emerged the last the decade suggesting that neither screening and high-risk approaches nor community interventions have measurable effects on the mortality rates.

SBU report
    One of these reports stems from the Swedish agency for medical technical evaluation (SBU). In 1997 the agency published a report of more than three hundred pages on the prevention of cardiovascular diseases through community actions and population strategies (1). Three of the main conclusions will be highlighted here:
1. There is no doubt that dietary habits, smoking, hypertension, lack of physical activity and social factors are of utmost importance as risk factors for cardiovascular diseases.
2. There is no scientific basis for starting new population and community programmes similar to those which have been undertaken throughout the world during the 1970s and 80s.
3. The assessment of eight community programmes showed no effects on cardiovascular risk factors and disease incidence beyond the secular changes already taking place.

    Such conclusions however stand in seemingly contrast to the rapid changes in the mortality rates which have been observed in a number of industrialised countries, and which partly coincide with changes in the prevalence of the major risk factors (2-3).
    At the same time we are continuously being confronted with glossy advertisements in high ranking medical journals on the benefits of blood pressure treatment, and hypolipemic drugs. The advertisements are referring to results from large randomised controlled trials which have been properly published and undergone critical scrutiny from the scientific community. Based upon these trials we are even in a position to indicate how many who have to be identified as high risk subjects and to be treated in order to postpone a cardiovascular event by say ten years.

Understanding
   Is it possible to reconcile the optimistic tunes from the glossy papers with the cooler attitudes reflected in the Swedish conclusions concerning the population approach? This question becomes even more relevant when taking into account the huge efforts laid down to establish international guidelines for preventive cardiology and the realisation that a second wave of the cardiovascular epidemic is on its way in most developing countries and eastern Europe (4).    There is an urgent need for understanding whether cardiovascular diseases can be stemmed at a population level and if the decline in cardiovascular mortality observed in many industrialised countries is at all associated with health policy making or merely reflecting secular trends caused by hitherto unrecognised factors.

    In this paper we will discuss whether the lack of scientific evidence is sufficient to dismiss the community programmes as feasible and effective public health actions in preventive cardiology. Some consideration will be given to concepts such as etiology, causal factors, effect-modification, risk and life expectancy.

Etiology - causal factors
   By the mid 1980s more than 200 single variables had been identified as risk factors for cardiovascular diseases. Some of these are obviously etiologically important factors whereas others are acting as indirect markers for hitherto unknown variables. The final identification of a risk factor as an etiological cause depends upon our ability to assess whether changes in exposure are followed by changes in cardiovascular disease risk, preferably in a randomised controlled trial.

    For preventive purposes the definition of an etiological factor is one which is directly associated with the risk of disease in the sense that a change in exposure to the risk factor will invariably be followed by a change in the incidence of hard endpoints. According to this definition only a minority of the recognised risk factors can be called etiological factors. Still for a number of risk factors trials to assess their etiological role is unfeasible, and for some we even conclude that there is a causal relationship in spite of negative or inconclusive randomised trials.

    An example of the latter is the role of cigarette smoking which is well established as a cardiovascular risk factor, but where the only randomised trial using hard endpoints came up with inconclusive results (5).

Other factors
    That other factors, in addition to total cholesterol, blood pressure and cigarette smoking are important, has been demonstrated by the seemingly inconsistent results from a number of cross-cultural and international studies. Most impressive is the follow-up of the Seven Countries Study which showed large variations in risk for coronary heart disease at the same levels of total cholesterol between Japan, the Mediterranean countries and northern Europe (6).

    Such discrepancies only emphasise the existence of modifying factors that by themselves may provide new means for preventing or postponing heart attacks. The multitude of risk factors and potential causal factors demonstrates the multifactorial causal web, which constitutes the pathological process underlying the clinical disorder. Most of the community intervention however, has focused on the three classical risk factors, blood lipids, blood pressure and smoking.

The population approach and the individual strategy
    Preventive medicine can be divided into a population and an individually aimed strategy. The population strategy does not identify individuals but has as its goal to shift the whole distribution of disease risk to the left on the distribution curve and is aiming at the whole population. The major prerequisites for preventive programmes are that they have an impact on the problem, and that the measures taken are acceptable to the population who will be exposed to the programme and usually paying the bill. The population approach implies that high and low risk subjects are treated alike, resulting in a large proportion of the individuals being «unnecessarily» exposed to the preventive measures.

Even then the impact from such programmes should theoretically have larger effects on public health parameters than only approaching a small proportion of high risk person due to the larger number of people exposed to the preventive measures and the slope of the risk curve. The larger number of disease events occur in the majority of the population who are not at so-called high risk. These observations and the theoretical considerations by the late professor G. Rose lead to the concept of the prevention paradox which says that a measure aiming at whole populations will have small impact on the average individual risk, but a large impact on a public health parameter.

    The opposite is also true, that measures taken against high risk subjects may have a large impact upon their health (provided the measures are effective), but very little on the public health parameter because the high risk subjects only constitute a small proportion of the population (7).

Community Programmes
   This thinking forms the rationale for community program-mes and health policy actions such as nutrition and food policy, tobacco policy and public health promotion campaigns. Theoretically these programmes would be more effective in stemming the cardiovascular epidemic than the high risk approach. The word Community programme implies that these systematic actions were not experiments or trials. Only a few trials have been undertaken, and even fewer were organised as randomised controlled trials. The trial design has been so called quasi experimental in the sense that the intervention area has been compared to a control area selected by convenience and availability.

The Swedish report assessed a number of community programmes and came up with eight which fulfilled their inclusion criteria:
1. The intervention is a mass strategy
2. The intervention is aiming at multiple risk factors
3. The evaluation design is controlled which implies that control populations are defined
4. The trial shall have an a priori research hypothesis
5. The results shall be available in international journals with referees.

The overall assessment of the effects of these eight programmes was, as mentioned above, that virtually no effect was observed which could be ascribed the programmes as such.

Among the reasons for this lack of effect was the possible “contamination” of the control area with the health education message, and that declining trends in cardiovascular mortality had already started before the programmes were established. The report also underlines the statistical methodological problems associated with examining effects on individual health when whole communities are observation units. One of the recommendations is that future studies should be more focused and less comprehensive.

Comments on reasons for lack of effect
Some of the principal reasons for not observing interven-tion effects in a trial are listed below:

1. The intervention was introduced too late
2. The intervention was not accepted by large segments of the participants (inefficient methods)
3. The control population changed its life-style (“contamination”)
4. Data on end-points were missing in both control and intervention group (reducing power)
5. Lack of power due to declining disease rates
6. Lack of power in general
7. The intervention measures did not attack the biologically relevant mechanisms (ineffective methods)
8. Bad luck

All these arguments can be assessed with regard to the community trials even if the last one is the least likely. But there are aspects on the etiology of cardiovascular diseases which may be of relevance for this discussion. Cardiovascular diseases are often categorised as multifactorial diseases. The theoretical concepts concerning multifactorial diseases, were developed by Rothman in the early 1980s (8). He suggested the use of the terms necessary and sufficient causes to describe effectmodification and interaction between different potentially causal factors.

    With regard to atherosclerotic diseases, blood lipids, and especially LDL-cholesterol may be considered a necessary factor in the sense that the likelihood for disease at a lower level of LDL-cholesterol is extremely small. Cigarettes on the other hand, may be included in a complex of factors constituting a sufficient set of causes, but its absence does not guarantee against atherosclerotic disease. Similar arguments can be put forward for most of the suggested risk factors, even if the mechanisms still remain obscure for a number of them. The incidence of cardiovascular disease will therefore depend both upon the prevalence of the different risk factors as well as the relative distribution of the factors within the population.

    Let us assume that one of the risk factors e.g. LDL-cholesterol is undergoing a secular change due to unknown cohort effects. A community programme aims at reducing smoking in half of the population, but the effects of this effort may not be measurable against the more powerful LDL-cholesterol decline determining the cardiovascular risk. So called “spontaneous” or secular trends have been used as explanations as to why it is so difficult to assess the effects of community programmes.

    This is really a special case of lack of statistical power. What we are trying to detect is the difference in the incidence rates between two populations which are both declining, and such differences are likely to be small. The number of subjects that are needed in order to detect the rate-differences that we may be observing is in the order of tens of thousands. If we sum up the eight arguments listed above and assess the different community program-mes accordingly, we are likely to conclude that almost all of them are relevant, and may have played a role in determining the outcome.

Conclusions
   The rapid changes in cardiovascular mortality rates suggest together with the identification of potentially amenable risk factors that these diseases are preventable. This has been demonstrated in intervention trials on high risk individuals. The effects of community interventions as well as politically oriented measures are not possible to demonstrate in scientific settings. The changing trends in incidence rates, mortality, case- fatality and secondary attack rate should be examined on a individual level using cohorts with already known risk factor levels as well as disease endpoints. The lack of effect in the community programmes emphasises the need for more focused and larger trials where individual data are included in the analysis.

Dag Thelle. professor
Hjärtlunginstitutionen, SU/Sahlgrenska,
Göteborg
E-post: Dag.Thelle@hjl.gu.se

References
1. Att förebygga sjukdom. SBU, Stockholm 1997
2. Wilhelmsen L, Rosengren A, Johansson S, Lappas G. Coronary heart disease attack rate, incidence and mortality 1975-1994 in Göteborg, Sweden. Eur Heart J 1997;18:572-581
3. Rosengren A, Eriksson H, Larsson B, Svärdsudd K, Tibblin G, Welin S, Wilhelmsen L. Secular changes in cardiovascular risk factors over 30 years in Swedish men aged 50: The Study of Men Born in 1913, 1923, 1933 and 1943. J Internal Med 1999;246:0-0
4. Wood D. Guidelines—a missed opportunity.Atherosclerosis 1999 ;143 Suppl 1: S7-12
5. Rose G, Colwell L. Randomised controlled trial of anti-smoking advice: final (20 year) results. J Epidem Comm Health 1992;46:75-77
6. Kromhout D. On the waves of the Seven Countries Study: a public health perspective on cholesterol.Eur Heart J 1999 ;11: 796-802
7. Rose G. Sick individuals and sick populations. In J Epid 1985;141:32-38
8. Rothman KJ. Modern epidemiology.Little, Brown and company. Boston/Toronto 1986