Commentary: Causality—the Achilles' heel of observational studies

Salonen et al use a nested case-control design to correlate risk of myocardial infarction for different paraoxonase genotypes. The method identified incident cases and matched them to controls in a ratio of 1 to 2. This approach uses a subset of subjects from a larger cohort to address a particular hypothesis.

Case-control studies are one of the weakest methods to infer a causal association between a risk factor and a disease. On the other hand, they are simple to perform and useful for generating hypotheses Problems of bias arise because of difficulty in matching cases to controls and the presence of multiple known and unknown confounding variables. Apparently significant results can arise by chance because of multiple testing, and some statistical approaches can be tipped in favour of significance in borderline cases.

In this study associations between genotype and risk of myocardial infarction were comparatively weak (lower limit of 95% confidence interval 1.17 (P=0.025) in the unadjusted analysis and 1.03 (P=0.044) in the adjusted model), which does not give great confidence in the causality of the findings. The main problem is that the number of subjects in the analysis is small. The hypothesis is intriguing and deserves a fair appraisal.

The correct next step would be to replicate these findings in a similar cohort study that had obtained DNA samples. Collaboration among groups with well documented cohorts would provide a major advance in the assessment of promising new markers of risk in both epidemiological and interventional studies after the publication of the main results of individual studies. This approach would help to reinforce or refute observations of borderline significance.

The importance of an association between a particular genotype and risk of disease allows people at high risk to be counselled appropriately and modifiable risk factors to be treated more aggressively. In certain circumstances, correcting a protein deficiency with supplements may be helpful. Genetic manipulation, although intuitively attractive, is associated with practical, technical, and ethical problems but may be a treatment of the future. Most cases of myocardial infarction are unlikely to be due to a single gene defect. Nevertheless, the report by Salonen et al helps to increase understanding of genotype and disease and to identify promising areas of research.

Commentary: How high density lipoprotein protects against heart disease

High density lipoprotein protects against cardiovascular disease. Protection is conferred in at least two ways The first way is by transporting cholesterol back from organs such as arteries to the liver, thus protecting the arteries from further atheromatous plaque formation. The second way is by acting as an antioxidant.1 Lipid peroxidation encourages atheroma formation, so that anything which limits lipid peroxidation should reduce the risk of cardiovascular events. The component of high density lipoprotein that is most likely to provide antioxidant activity is paraoxonase.

In this study Salonen et al examined the paraoxonase gene of 165 men One genetic mutation appeared quite commonly. In this mutation methionine is substituted by leucine at one particular position in the enzyme's protein structure, increasing the enzyme's antioxidant activity. Men who did not carry this mutation (that is, those who were homozygous for the M allele) had three times the risk of myocardial infarction of those who did.

These observations suggest that paraoxonase is more active when the mutation is present and less active when it is not and that paroxonase might be protective against cardiovascular disease.