ViewpointWhen are observational studies as credible as randomised trials?
Section snippets
Benefits of randomisation
Two benefits are expected from randomisation: unbiased allocation of treatment, because of easier concealment of the allocation scheme,1 and application of statistical theory on the basis of random sampling.2
The first benefit was the reason that randomisation was introduced in medicine.3 In the first half of the 20th century, clinical investigators who tried to achieve a fair comparison between treatments often used alternation: patients received treatment in the order in which they presented.
Adverse drug reactions: breaking the link between prescription and prognosis
The average randomised drug trial is too small, and does not have sufficient follow-up to detect adverse effects that are fewer than about one per 200 per year, or that take longer than 1 year to develop. To investigate adverse drug reactions, either case-control studies or large-scale observational follow-up studies are needed.
The key to understand how observational studies about adverse drug reactions might approach a similar level of unbiasedness of allocation as if patients were randomised,
Genetic research: mendelian randomisation
When we read that people with BRCA1 mutations develop breast cancer more often than those without the mutation, or that people with factor V Leiden have more venous thrombosis than those without, we almost never ask the question that we would always ask about a randomised trial: is there a table of baseline characteristics of people with and without this gene to make certain that these people are comparable for known risk factors for the disease? Were they of similar age, sex, pregnancy status,
A proposal for epidemiology and clinical research: a three-pronged restriction
Observational researchers often seem to assume that it suffices to enter potential confounders into a statistical model, and that everything that is then left by way of effect for an exposure of interest will be an estimate of a causal association.12 From the debacles on hormone replacement therapy and β carotene, we have learned otherwise—many confounders are insufficiently known or are unquantifiable. A better strategy might be to use restrictions that directly guarantee the unbiasedness of
Conclusion
Although not all restrictions will always be feasible, and in each instance uncertainties will remain—as with any research undertaking—restriction in research topics, design, and analysis helps observational research to attain the desired benefits of randomisation, and gives observational research the chance to be as credible as randomised evidence.
References (20)
- et al.
Allocation concealment in randomised trials: defending against deciphering
Lancet
(2002) - et al.
Risk of idiopathic cardiovascular death and nonfatal venous thromboembolism in women using oral contraceptives with differing progestagen components
Lancet
(1995) - et al.
Mendelian randomisation: a new spin or real progress?
Lancet
(2003) - et al.
Evidence from randomised trials on the long-term effects of hormone replacement therapy
Lancet
(2002) - et al.
Effect of simultaneous pancreas-kidney transplantation on mortality of patients with type-1 diabetes mellitus and end-stage renal failure
Lancet
(1999) The design of experiments. 8th edn, 1971
Why transition from alternation to randomisation in clinical trials was made
BMJ
(1999)- et al.
Case-control studies of drug induced illness
Am J Epidemiol
(1978) The need for randomization in the study of intended effects
Stat Med
(1983)Bayesian inference for causal effects: the role of randomization
Ann Statistics
(1978)
Cited by (449)
The bidirectional association between frailty index and cardiovascular disease: A Mendelian randomization study
2024, Nutrition, Metabolism and Cardiovascular DiseasesMise à jour technique n<sup>o</sup> 439: Corticothérapie prénatale en période de prématurité tardive
2023, Journal of Obstetrics and Gynaecology CanadaTechnical Update No. 439: Antenatal Corticosteroids at Late Preterm Gestation
2023, Journal of Obstetrics and Gynaecology CanadaCardiorenal Outcomes Among Patients With Atrial Fibrillation Treated With Oral Anticoagulants
2023, American Journal of Kidney Diseases