The paper reviews the strengths and weaknesses of the epidemiologic methodology using examples from observational studies and experimental studies (natural and true or man-made) The epidemiological methodology uses the scientific method in the generation and testing of hypotheses. Epidemiologic reasoning is empirical, inductive, and refutative. Epidemiological studies are much cheaper in terms of material or human resources and provide generalizable results in a shorter time than comparable laboratory-based studies. The biggest strength of the epidemiologic methodology is ability to offer viable practical interventions that solve problems even before detailed causal mechanisms are known. Unlike laboratory experiments, epidemiological studies can not control extraneous or confounding factors perfectly. Epidemiological findings are rarely conclusive.

NATURE OF THE EPIDEMIOLOGIC METHODOLOGY

Epidemiology uses the procedures of the scientific method: stating a hypothesis, collecting and analyzing data to test the hypothesis, and reaching conclusions about the hypothesis. The epidemiologic methodology ensures continuous improvement of knowledge by generating and testing new hypotheses. An epidemiological hypothesis is formulated to relate two phenomena: the disease and the putative cause of the disease (the exposure or risk factor). Hypotheses must be specific and testable. Empirical data is from experimentation and observation. The conclusions from testing a hypothesis can be rejection/non-rejection and never acceptance. A new hypothesis is generated from the conclusion and the process is repeated. Use of the scientific method implies among other things that epidemiological knowledge is never stable. It keeps on changing and getting nearer the truth as new information is discovered.

Epidemiological methodology is empirical, inductive, and refutative. Empiricism refers to reliance on physical proof. Epidemiology relies on and respects only empirical findings. There is no room for preventive action being based on pure reasoning, rationalism, not subjected to empirical verification by collection and analysis of data. Both inductive and deductive logic are used in epidemiological reasoning. The inductive is used more because it is more in line with empirical experimental verification. Induction is a type of reasoning that starts from one observation and generalizes. Refutation is a type of scientific reasoning introduced by the philosopher Karl Popper (1) that emphasizes rejection of suppositions rather than accepting them. Rejecting an idea opens the way to test other ideas. Epidemiology can refute a finding but can never offer conclusive proof. This is in line with the spirit of empirical inquiry that knowledge and understanding grow continuously and facts accepted and interpreted in one way today may be rejected and interpreted in another way tomorrow.

NATURAL EXPERIMENTS

Natural experiments are not deliberately designed by humans. They can be analysed to provide insight at little cost to humans because they just involve observation of events. They however are rarely conclusive. Natural experiments can be divided into 2 types. Some involve no human agency at all like earth-quakes, floods, and cosmic radiation. Others involve human action which is not deliberate or planned in any systematic way.

The following are examples of natural experiments:

• Smog episode in Danora, PA. In 1948 (2)
• The London fog of 1952 that killed 4000 persons in 10 days (3)
• Snow's study of the relation between polluted water and cholera in London (4)
• Occurrence of polio after tonsillectomy (5)
• Cancer in accidental chemical exposure ( 6)
• Cancer in survivors of atomic explosions in Hiroshima and Nagasaki (7, 8)
• Heart attacks and asthma following the Athens earthquake of 1981 (9).
• Thalidomide disaster (10)
• En epidemic of deafness in Australia (11)

TRUE EXPERIMENTS

True experiments involve deliberate human action or intervention whose outcome is then observed. Their objective is to establish the definitive causal relation. True experiments are more involved and expensive that observational studies or natural experiments but are still cheaper and easier than laboratory-based research. The main strength of experimental studies is good control of extraneous or confounding factors that might make interpretation of study results difficult. The main weaknesses is that well controlled experiments on humans are difficult. It is difficult to put humans under full experimental conditions where they can be observed for 24 hours. Ethical controversies and violation of human rights always arise in such studies.

The following are examples of true experiments

• Lind's trial of fruit juice for scurvy in 1747 (12)
• Jenner's cowpox vaccination in 1796 (13)
• Induction of pellagra by Goldberger and Wheeler in 1915 (14)
• Discovery of the relation between rice and beri-beri by Fletcher in Kuala Lumpur in 1905 (15)

COMMUNITY INTERVENTION STUDIES

A community intervention study is designed to test whether a certain public health intervention such as health education or water fluoridation has an effect on a given outcome measure. Two or more similar communities are randomly allocated to receive different interventions and the outcome is then measured. Random allocation ensures comparability. The strength of the community intervention study is that it can evaluate a public health intervention is natural field circumstances. The weakness is that people in the control community may receive the intervention under study on their own because tight control as occurs in laboratory experimental or animal studies is not possible with humans.

Examples of community intervention studies

• Tests of the prevention of dental caries by fluoridation carried out by the United States Public Health Service (USPHS) in 1970 (16).
• Trials of vaccination efficacy

OBSERVATIONAL STUDIES

Observational studies, descriptive or analytic, allow nature to take its course with no human interference. They usually precede and prepare for definitive experimental studies. Cross-sectional (prevalence), case-control, and follow-up studies are the main types of observational studies in epidemiology. The advantages of observational studies are their low cost and lack of ethical controversies. A cheap study is made of a wide variety of human experiences by just observing and recording information. This is much cheaper than experimental studies in which people must be subjected to various treatments and exposures at the experimenter's cost. Ethical problems in observational studies are much less than those in experimental studies because the human subject is not exposed to any major physical risk. Observational studies have three main disadvantages. It is not possible to study etiology directly because the investigator does not manipulate the exposures. Etiology is studied only indirectly by comparing disease experience in the group exposed to a putative risk factor with the group that was not exposed. Information on the variable of interest may not be available recourse being made to surrogate variables. Several unplanned co-factors (giving rise to confounding, interaction, or effect modification) are involved making interpretation difficult. Experimental studies, unlike observational studies, collect systematic information on these co-factors rendering study interpretation easier.

Most of epidemiological study is observational. The following are a few examples

• Hippocrates and observations on the relation between disease and environment (17)
• Observations by Graunt on the London bills of mortality (18)
• Observations by William Farr based on vital statistics of England and Wales (19)
• Semmelweiss's observation of the relation between washing hands and child-bed fever (20)
• Relation between working in the dye industry and bladder tumors (21)
• Cancer mortality after irradiation for ankylosing spondylitis (22, 23)
• Smoking and lung cancer (24, 25, 26, 27, 28, 29, 30)
• Cancer of the cervix and circumcision (31)
• Social class and mental illness (32)
• Mortality of American radiologists (33)
• Neoplasia in children treated with x-rays (34)
• Cancer in uranium miners (35)

PREVENTIVE ACTION BEFORE FULL UNDERSTANDING

It is a strength of the epidemiological methodology that effective preventive action against disease can be undertaken even before the complete mechanism of etiology is known as shown in the following examples

• Smoking and lung cancer
• High lipid diet and cardio-vascular disease
• Low fiber diet and colon cancer
• HBV vaccination and hepatocellular carcinoma

THE PROBLEM OF CONFOUNDING BIAS

Miettinen has dealt with the theoretical basis of confounding (36). Confounding bias arises when the disease-exposure relationship is disturbed by an extraneous factor called the confounding variable. The confounding variable is not actually involved in the exposure-disease relationship. It is however predictive of disease but is unequally distributed between exposure groups. Being related both to the disease and the risk factor, the confounding variable could lead to a spurious apparent relation between disease and exposure.

Confounding can be understood by the following example. Alcohol consumption confounds the relation between smoking and lung cancer. There is an indirect relation between alcohol consumption and cancer of the lung. We observe that those who have lung cancer also consume alcohol. This is because of the non-causal relation between alcohol consumption and cigarette smoking. The two are part of the same lifestyle and tend to occur together. The direct causal relationship between cigarette smoking and lung cancer could be distorted in a study in which alcohol consumption is not balanced between the smoking and non-smoking exposure groups. A negative relationship between cigarette smoking and lung cancer will be seen if study subjects are selected predominantly from the non-smoking population.

Confounding can be prevented or minimized. Prevention of confounding at the design stage by eliminating the effect of the confounding factor can be achieved using 3 strategies are used: pair-matching, stratification, and randomisation. Multivariate techniques can be used to adjust for the effects of confounding at the analysis stage of the study.

BALANCE OF STRENGTHS AND WEAKNESSES:

In a 1979 seminal discussion of the strengths and weaknesses of epidemiological methodology, Brian McMahon, Professor of epidemiology at Harvard, summarized the strengths as obtaining observational data cheaply and with a wide wide range of human exposures. The humans both choose and pay for the exposures and what the epidemiologist does is to collect the data. He identified the major weakness as inability to offer conclusive proof of the disease-exposure relationship. In this sense epidemiological findings open the way for further work and verification by laboratory science. As more epidemiologic studies indicate the same etiology, increasing convincing evidence is obtained. However the final proof will have to come from the laboratory. In some cases the final proof is never obtained. Using the limited epidemiological knowledge on causation, public health interventions may be undertaken and they result in elimination of the disease before the laboratory workers have a chance to say their word.

TABLE #1: MORTALITY FROM CHOLERA IN THE DISTRICTS OF LONDON SUPPLIED BY SOUTHWALK & VAUXALL COMPANY and the LAMBETH COMPANY, July  8 to August 26, 1854

Source: MacMahon, B. et al. Epidemiology: Principles and Methods. 2^nd Edition. Little, Brown, and Company, Boston  1996.

TABLE #2: MORTALITY FROM CHOLERA IN LONDON, July 8 to August 26, 1854 Related to the Water Supply of Individual Houses in Districts Served by Both the SOUTHWALK & VAUXALL COMPANY and the LAMBETH COMPANY, July  8 to August 26, 1854

Source: MacMahon, B. et al. Epidemiology: Principles and Methods. 2^nd Edition. Little, Brown, and Company, Boston  1996.

TABLE #3: RELATION BETWEEN CURED RICE AND BERIBERI

Source: MacMahon, B. et al. Epidemiology: Principles and Methods. 2^nd Edition. Little, Brown, and Company, Boston  1996.

* Cured rice is parboiled in its husks before milling which allows thiamine to diffuse and stay in the rice and is not removed by subsequent milling.

TABLE #4: RELATIVE and ATTRIBUTABLE MORTALITY FROM SELECTED CAUSES ASSOCIATED WITH HEAVY CIGARETTE SMOKING BY BRITISH MALE PHYSICIANS 1951-1961

TABLE #5: DEATH RATES FROM LUNG CANCER ACCORDING TO SMOKING HABITS OF BRITISH MALE PHYSICIANS AGE 35 and OVER, 1951-1956