Epidemiology is an essential component of all public health activities that implement the organized efforts of society to promote, protect, and restore health. These comprise: clinical medicine, which deals with individual patients; prevention and early diagnosis at the population level; and empowering people to exercise responsibility for their health through adopting health-promoting habits. ‘From epidemiology to medicine, prevention, and public health’ argues that systematic reviews complemented by meta-analyses of randomized controlled trials are particularly valuable for clinical medicine, contributing to the continuously evolving body of evidence-based medicine which guides doctors' everyday practice. More generally, the quantitative and probabilistic traits of epidemiology pervade clinical medicine.
Working for the health of all
Epidemiology is at heart a field of applied research with the improvement of the health of all as the key aim. As such, epidemiology is an essential component of all public health activities that implement the organized efforts of society to promote, protect, and restore health.
This concept of public health has no relation to how societal efforts to improve health are or should be organized; it does, however, imply that some kind of explicit organization should exist, rather than just dispersed and uncoordinated initiatives, for society to successfully tackle health problems.
As shown in Figure 14, three broad activities contribute to people’s health. In clinical medicine, doctors and other health personnel deal individually with each patient. They provide preventive measures such as drugs to control high cholesterol or elevated blood pressure, or deliver advice and psychological support to stop smoking. They intervene to diagnose, treat, and when possible cure, diseases with procedures ranging from the simple prescription of an antibiotic to a complex liver or heart transplant. Finally, they offer individual rehabilitation to people with disabling diseases.
p. 98p. 99↵Prevention and early diagnosis at the population level form the second field of activity. Prevention addresses the root causes of disease, environmental or genetic. It embraces a vast array of regulations spanning control of pollutants in air, water, and the workplace, to traffic speed limits and safety requirements in home appliances. It includes compulsory and optional vaccination programmes as well as campaigns to foster healthy diet and behaviour. When it targets genetic causes of diseases, for example the screening of all newborns for genetic defects, primary prevention uses medical diagnostic tools, as do organized programmes of early diagnosis and treatment of diseases. These have proved effective and are operational in many countries for a limited number of high-impact diseases such as cancers of the uterine cervix and of the breast.
The third activity consists in the empowerment of people to exercise responsibility for their health through adoption of health-promoting habits and participation in the decision processes that shape health policies. The latter in turn may reinforce or inhibit people’s empowerment, the development of which depends on formal and informal education and on updated and accurate information.
Public health also coordinates these activities in relation to other societal actions, external to the health system, which strongly influence health, for example income and housing policies. In the coordination process, public health administrators and policy makers usually demand that the benefits and adverse effects of proposed policies be subject to economic analysis, in which epidemiologists play a specific role jointly with other specialists.
Channelling the research results into practice, whether in clinical medicine, in population prevention, or for people’s empowerment, requires as a first step the aggregation of the results of multiple studies to consolidate the total evidence available on a specific question, for example whether vitamin C protects against p. 100↵cancer in humans. This is done by critically reviewing the studies’ reports, comparing methods and results, and drawing a general ‘best’ answer to the question at hand. In the last two decades, the approach and methods used in a review, previously entirely left to the reviewer’s discretion, have been refined and made more objective and rigorous under the heading of systematic reviews.
Systematic reviews, with and without meta-analysis
A systematic review is a review carried out using a systematic approach to minimize bias and random errors, a process which is explicitly documented in the methods section of the review itself. It usually offers a more objective appraisal of the available evidence than traditional reviews, conducted as narrative commentaries on the studies. In a systematic review, each study is scrutinized to assess its quality in respect of a number of criteria fixed in advance, e.g. how well the population is defined, whether the study responses were assessed blindly or not, and so on. This makes it possible to consider separately studies judged of higher and lower quality, rather than all of them together, and see whether the results of the lower-quality studies point in the same direction (e.g. towards a reduction or an increase in risk) as the higher-quality ones. Broadly consistent results can be combined in a statistical analysis, a meta-analysis, to provide a single summary estimate of risk. This analysis, in which each study is given a ‘weight’ proportional to the number of disease cases it contributes, may cause a clear-cut result to emerge, while the individual studies, particularly if small in size, may each present a result statistically non-significant that is difficult to interpret.
Combining studies often permits the evaluation of rare events, too few of which occur in a single study. A typical case is that p. 101↵of side effects of new drugs, which occur infrequently, say once in a thousand treated patients. However, if a side effect is serious, for instance a major heart problem, it will have considerable impact when the drug is put on the market and used by hundreds of thousands or millions of people. Yet such an effect will be hard to detect in a randomized experiment of a size of, say, a few hundred subjects, which would be more than adequate to measure a much more frequent therapeutic effect. It is only by combining all available data from different randomized experiments that a sufficiently large number of patients is reached to allow the adverse event to become detectable. A telling example is Rofecoxib, a drug commercialized in 1999 as an anti-inflammatory remedy for rheumatic and muscular disorders. It was withdrawn from the market by the manufacturer in September 2004 on account of an increased risk of heart attacks, when an estimated 80 million people had already used it. However, if the manufacturer or the drug licensing authorities had conducted a timely meta-analysis, they would have detected the increased risk more than three years earlier, in 2000, as Figure 15 clearly shows (the meta-analysis of this figure was retrospectively performed in 2004 by independent academic epidemiologists).
Systematic reviews complemented by meta-analyses of randomized controlled trials are most valuable for clinical medicine. They have helped to develop the continuously evolving body of evidence-based medicine which guides doctors’ everyday practice. They have also helped to put the evidence from randomized preventive trials carried out in populations on a firm basis, for example the prevention of myocardial infarction with cholesterol-lowering drugs.
Meta-analyses have also been extended to observational epidemiology studies directly relevant to public health. Combining results from observational studies in which confounding factors and biases have usually been dealt with in a different way in each p. 102↵study in a statistical analysis is, however, problematic. As we know from Chapter 5, in randomized controlled trials bias and confounding are prevented by randomization and do not impinge on a meta-analysis, a condition that does not apply to observational studies. For these studies, systematic reviews are iany case necessary while the worth of meta-analyses has to be assessed case by case.
Systematic reviews form an important part of clinical epidemiology, but more generally the quantitative and probabilistic traits of epidemiology pervade clinical medicine. It is common to find today in standard textbooks of medicine references to ‘NNTs’ and schemes of ‘diagnostic decision trees’. Comparing treatment options is helped by computing the NNT, or number needed to treat. In severe hypertensive subjects, the risk of a major adverse outcome (such as death or stroke) in the coming three years may be as high as 20%. A treatment may, however, reduce it to 15%. The risk reduction obtained with the treatment is 20 ‐ 15 = 5%, which means that out of 100 subjects treated, 5 avoid the major adverse outcome they would have otherwise suffered. This is the same as saying that for one subject to avoid a major adverse event, the number needing treatment is 100/5 = 20. Should a new treatment reduce the risk to 4%, it would be necessary to treat only 6 ˜= 100 / (20 ‐ 4) patients to avoid one adverse event. Comparing the number of people who need to be treated for the two treatments, 20 against 6, conveys tangible information on the merits of the two treatments, the second being clearly superior (provided all other aspects are the same, for instance the frequency of side effects, but these can be dealt with in terms similar to NNT).
A diagnostic decision tree is designed to assist the physician in formulating a diagnosis. If a young man presents with a sudden vague but aching and recurrent pain in the left chest, one diagnostic possibility is coronary artery disease, the narrowing of the coronary arteries that supply blood to the heart. Given the young age of the patient and the absence of any other sign, this condition appears a priori unlikely, but being very serious it could be disastrous to miss it. The patient can thus undergo an exercise stress test whereby his electrocardiogram is monitored during controlled physical effort. A negative test would be reassuring; unfortunately the test is not perfect and sometimes it turns out falsely p. 104↵negative even in presence of the disease, in the same way that it can be falsely positive in its absence. If narrative terms like ‘a priori unlikely’, ‘sometimes falsely positive’, ‘sometimes falsely negative’ are replaced with figures of probabilities (derived from specific studies), a map, or decision tree, can be built of all possible courses of diagnostic actions. One course may be to dismiss straight away the diagnosis of coronary artery disease because the type of pain found in an otherwise healthy and young man makes the diagnosis less than 5% probable. The alternative course is to proceed to the stress test knowing, however, that it has a 30% probability of false negative results (i.e. it has a sensitivity of 100 – 30 = 70%) and a 10% probability of false positive results (i.e. it has a specificity of 100 – 10 = 90%). Combining these figures makes it possible to calculate the probability, or predictive power, that each alternative will correctly identify the disease if present or dismiss it if absent. A comparison of these probabilities, and of the penalties involved in a wrong diagnosis, helps the physician to analyse the diagnostic process, which often involves not just one but many possible tests, and to choose an optimal diagnostic strategy (these calculations are based on Bayes’ theorem, a fundamental tool for drawing inferences of probabilistic nature from empirical observations, established as early as the mid-18th century by the Reverend Thomas Bayes).
Prevention and early diagnosis
In a strict technical sense, ‘prevention’ denotes the activities aimed at directly modifying the root determinants of disease, which fall only into two broad categories: genes and environment, or in more archaic wording ‘nature and nurture’. Early diagnosis, on the other hand, aims at detecting and treating diseases before they become manifest through symptoms. These two neatly separated activities, both organized at the level of the whole population, have, however, a major bridge in the diagnosis of host risk factors, like high blood cholesterol or high blood pressure, that are not yet ‘diseases’ but increase the chance of disease occurrence; on the one side, the host risk factors share this property with a p. 105↵person’s genes predisposing to disease, while on the other they are themselves the result, like early disease, of a complex interplay of genes and environment.
Some early disease diagnosis tests are carried out as ‘opportunistic screening tests’ by individual doctors when they examine a patient: for instance, the PSA test for prostate cancer discussed in Chapter 5 has become, rightly or wrongly, popular in several developed countries even in the absence of firm evidence of net benefit. Only screenings for which this evidence exists do, however, qualify for systematic adoption in the population in the form of ‘organized screening programmes’, such as those for colon cancer or for cervical and breast cancer in women, now implemented on a substantial scale in many countries. Screening programmes aimed at early diagnosis in apparently healthy populations are evaluated in the same ways as the diagnostic procedures in symptomatic patients previously discussed. Programmes for different diseases can be compared or different alternatives of a programme, for instance screening for cervical cancer using either the cytological ‘Pap test’ or the assay detecting the human papilloma virus. For this purpose, indexes such as the predictive power and the number needed to screen (NNS) are calculated. The latter is closely similar to the number needed to treat (NNT) and tells how many subjects one needs to test in order to avoid one death or other major adverse event within a period of time. It depends not only, as NNT does, from the probability that a treatment successfully avoids death but also from the probability that an apparently healthy subject turns out to have the disease without symptoms. NNS are usually in the range from several hundreds to, more often, several thousands.
Screening for host factors, genetic or acquired, that may predispose to a disease stands on the basic assumption that subjects who will develop the disease can be distinguished from subjects who will not, so that any preventive intervention, for example a change in diet, can be concentrated on the former (should the distinction p. 106↵prove impossible, there would be no point in screening and any intervention would simply need to be applied to everybody). Looking closely at one of these risk factors, blood cholesterol, throws light on how far the basic assumption is justified and illustrates at the same time some general principles of prevention, taken in the wide and generic sense of any measure able to prevent at any point the progression from health to disease and death.
Today, few will be surprised if a heavy smoker comes down with lung cancer. Many may be surprised, however, if told that avoiding heavy smoking will not wipe out the burden of lung cancer in the population because a substantial number of cases occur in fact in people who regularly smoke only moderately. What is true for smoking holds even more for blood cholesterol, as this set of figures shows: p. 107↵People with frankly anomalous cholesterol levels, say above 6.5 millimoles per litre, represent 6 + 3 + 2 = 11% of the population in which it has been found that 13 + 9 + 8 = 30% of the deaths from heart attacks occur (in case you feel more comfortable with milligrams per 100 millilitres, 6.5 millimoles is about 250 milligrams). Intervening on this ‘high-risk’ fraction of the population, about one-tenth of the total, would prevent – assuming an intervention that is 100% effective – just one-third of the deaths, leaving untouched the other two-thirds. Why these disappointing results? Because the risk is not concentrated solely in people ‘at high risk’, with cholesterol levels above 6.5 millimoles, but involves everybody to some degree. As cholesterol levels increase over the very lowest levels (category 0–3.9), the risk of disease increases by small increments, with no abrupt jumps.
Categories of cholesterol levels in millimoles per litre
Percentage of the population
Percentage of deaths from heart attacks in each category
0.0 – 3.9
4.0 – 4.4
4.5 – 4.9
5.0 – 5.4
5.5 – 5.9
6.0 – 6.4
6.5 – 6.9
7.0 – 7.4
7.5 – 8.0
As a consequence, the many people with only modest elevations in cholesterol who are also at a modestly increased risk produce more cases of heart attacks than the minority of people at high risk. This ‘paradox of prevention’ implies that the bulk of cases could be prevented by moderately reducing the cholesterol level, hence the risk, of everybody. Abating cholesterol only in people with high levels is certainly beneficial to them but cannot do the public health job of preventing the mass of cases in the population. Many disease determinants have been found to increase the risk of some diseases in a smooth, continuous way like cholesterol, for example blood pressure for heart attacks, hydraulic pressure in the eye for glaucoma, or alcohol consumption for cancer of the oesophagus or liver cirrhosis.
The graded distribution over the whole population of risk generated by these determinants, rather than its exclusive concentration in some groups, stresses their role as population disease determinants, discussed in Chapter 4 in contrast to individual determinants. The susceptibility of each person, rooted in their genetic make-up, plays – as does chance – a role in determining who becomes diseased, but the number affected p. 108↵will depend to a major extent on the population determinants. For example, there are no known populations with a high frequency of heart attacks without also an average (over the whole population) high level of cholesterol. The next question then becomes: why do population determinants differ from one population to another? Cholesterol level is diet dependent and, like alcohol consumption, is conditioned by available foods (or alcoholic drinks), traditional tastes, and behaviour influenced by marketing and by economic constraints. For infectious diseases, the proportion of people vaccinated is a typical population determinant of how often a disease will occur, because vaccinated people do not fall ill and at the same time they interrupt the chain of transmission of the contagion.
For most diseases, multiple, rather than single, determinants are recognized. For example, blood cholesterol level, blood pressure, tobacco smoking, diabetes, and obesity are main population determinants of heart attacks. Interventions acting in turn on these determinants aim at promoting healthy habits, behaviours, foods, and to limit the availability of harmful products. This population strategy of prevention, based on a variable mix of incentives, education, and regulation, is beneficial to everybody, whatever one’s known or unknown susceptibility or level of risk. It can be complemented by specific preventive actions, often involving the use of drugs (e.g. to lower cholesterol or blood pressure) for people known to be at definitely high risk. Recently the idea has dawned that a combination in a single pill (‘polypill’) of low doses of several drugs controlling cholesterol level, blood pressure, and blood clotting propensity could be used in a population prevention strategy by offering it to most or all middle-aged and older people. Whether this is an effective, safe, and realistic possibility remains to be explored. The general principle is that before being launched on a grand scale, a preventive measure must have been clearly shown to work. This involves research covering a large number (Figure 16) of disease determinants, from proximate biological and genetic factors, to personal behavior p. 109↵traits, and to the ‘determinants of the determinants’ operating at the level of the social or of the global environment.
Attention to the global environment has markedly increased in recent years. Localized ‘heat waves’ have caused clearly documented excesses of mortality and fluctuations in urban air pollutants, especially fine particulates, which have been shown to increase hospital admissions for respiratory and cardiovascular ailments and to precipitate deaths from a variety of causes. Protocols to prevent these adverse effects affecting in particular vulnerable, already sick people have been put in place in a number of countries.
In contrast to these meteorological episodes, the health consequences of the foreseen global climatic change are a completely new chapter for epidemiological investigation. A likely temperature increase of anything between 2°C and p. 110↵5°C by the end of this century may be reflected in a sea-level elevation of 20 centimetres to 60 centimetres, involving a change in coastlines with consequent exposure of populations to flooding, already regularly experienced in a country like Bangladesh. Tropical cyclones, to which more than 300 million people are currently exposed, are expected to become more intense. The biological cycles of parasites are sensitive to climate changes, so that hundreds of millions of additional people will be infected by diseases like malaria. A further likely consequence is increased under-nutrition caused by droughts and rural poverty that, like the other sequels of climatic change, will induce mass migrations, themselves a source of severe health problems (as just one example, keeping well controlled a serious case of diabetes, a delicate but everyday routine task in developed countries, may become hopeless in a moving refugee population). Today, these effects can be identified but their probable impact on health (currently quite modest) remains to be quantified through research that combines available epidemiological data, for example on malaria in different regions, with models simulating how the disease may evolve under various hypotheses of temperature and other environmental changes.
In an equal rights society, every citizen ought to be empowered to take part in decisions affecting her or his health and, through democratic processes (on which more in Chapter 10), in deliberations concerning the health of the population. This can come about through information, conditioning, or education. Therare innumerable sources of information: newspapers, magazines, books, television, and, most prominently, the Internet. There are close to 100,000 sites on the Web dealing with health matters and major issue is the accuracy of the information. Studies are being done to measure the risk of encountering inaccurate sources, and private and public accreditation systems are being developed.
p. 111Almost everybody with access to the Web searches it occasionally or regularly on health, usually in relation to actual or possible health problems. Texts found for this reason or for curiosity or cultural interest need to be interpreted in the light of two considerations. First, most descriptions are inherently probabilistic, based as they must be on risks and rates of success of a preventive measure or of a treatment, or rates and risks of harm from side effects of a drug, from an unhealthy food, or from smoking marijuana, say. Second, the presentation is usually influenced by who is providing the information and for what declared or implicit purpose. It may be impartial and strictly to the point or framed in a wider educational context (as the guidance for the wider public of the National Institute for Health and Clinical Excellence in the UK), or it may instead lean towards propaganda to condition people to buy and use some product often by heightening worries about health.
There are, however, some rules of thumb that may be usefully applied to screen health information from Web searches or in the media:
Trust new findings only if replicated Frequently repeated claims such as, for instance, that a newly identified protein in tomatoes reduces the risk of colon cancer by half, should be treated with great caution, not because the result arises from a flawed study (it could) but, as stressed many times in this book, because a causal link can only be established through replicated, separate, and concordant investigations. Replicated investigations means several different studies, not the result of the same study echoed with various delays by several different media.
Trust only findings qualified by their uncertainty We all prefer a black-and-white image of reality, if nothing else because decisions to be taken are perforce yes-or-no. Yet most often there is some margin of uncertainty in the results and black-and-white p. 112descriptions hide an essential part of the relevant information. Dogmatic statements should be treated with caution.
Trust findings only if placed in context Enabling people, in professional as in ordinary life, to transform information into empowering knowledge implies that information is not isolated nor randomly connected to other elements, but placed in context. For example, findings of studies on the possible cancer risks from mobile phone use should be discussed in the context of other health effects, including the risk of car accidents from use while driving. This can be done within the text itself or by links to external references. Such contextualization is essential in developing the reader’s personal appropriation and interpretation of information; it does not substitute for it nor should it try to do so.
Trust findings only if not framed as advertisements Advertising is a signal necessary to draw attention to the substance, i.e. new or important findings. In commercial, sales-promoting reporting, however, the roles are completely reversed, the advertisement being the substance. Selling genetic profiles on the Web pretending, based on questionable or no evidence, to predict which diseases you will suffer in the future, has become a profitable enterprise. A safe rule would be to ignore it altogether: by ignoring such ventures, you lose nothing and when sound evidence about genetic factors predisposing to a disease becomes available, you will learn of them anyway from other, non-commercial sources.
Trust findings and recommendations only if concordant This is perhaps the most crucial guiding rule. With information, as with most other circumstances in life, there is no free meal; there are low-cost fast meals but they are of unknown quality. It is only by taking the time and effort of cross-checking the information from different sources, carefully looking at details, that one can be reasonably confident about the quality and validity of the information.
The health system is the common name for the complex of all activities directly dealing with health, although in most, if not all, countries this ensemble is more a complicated aggregate of many component systems than a unique organization. Public health coordination, itself one of the components operated chiefly by central, regional, and local health authorities, frames and interrelates the systems of hospitals (public, private for profit, private non-profit), general practices, clinical specialists, prevention units, and all other health-related activities.
Administrators at all levels of the health system, as well as political decision makers, constantly face the issue of comparing benefits and costs of interventions and services. Economic analyses may focus on exploring different ways of performing the same intervention, for instance the same number of renal dialyses, in order to identify the least costly procedure (cost-minimization analysis). Or they may compare the cost and the result, in terms of a common outcome like prolongation of life, of different interventions such as renal dialysis versus kidney transplant (cost-effectiveness analysis). Finally, they may compare costs and benefits of different interventions for the same or different conditions (hypertension treatment or influenza immunization?) in monetary terms or in some measure of ‘value’ as perceived by individuals (cost-utility). Epidemiologists intervene in these analyses by providing evidence on health benefits and ill effects as evaluated by systematic reviews of biomedical and epidemiological studies. The same standard of rigorous scrutiny applied to this evidence also needs to be used for assessing the economic evidence. Short of this there is no guarantee that the health of all, including the most vulnerable, will stay ahead of other societal interests, industrial, financial, or ideological.