Using twin studies to label disease as genetic or environmental is inappropriate
BMJ 2002; 324 doi: https://doi.org/10.1136/bmj.324.7345.1100/b (Published 04 May 2002) Cite this as: BMJ 2002;324:1100
- Alex J MacGregor, Arthritis Research Campaign senior fellow (alex.macgregor{at}kcl.ac.uk),
- Jerry Lanchbury, reader in molecular immunogenetics,
- Alan S Rigby, senior lecturer in statistics and epidemiology,
- Jaakko Kaprio, professor of genetic epidemiology,
- Harold Snieder, genetic epidemiologist
- Twin Research and Genetic Epidemiology Unit, St Thomas's Hospital, London SE1 7EH
- Guy's, King's and St Thomas's School of Medicine, Guy's Hospital, London SE1 9RT
- University of Sheffield School of Medicine, Sheffield Children's Hospital, Sheffield S10 2TH
- Department of Public Health, PO Box 41, FIN-00014 University of Helsinki, Finland
- Georgia Prevention Institute, Medical College of Georgia, Augusta, GA 30912-3715, USA
EDITOR—The cover of the BMJ on 2 February poses the question “Rheumatoid arthritis: is it genetic?” and answers “Probably not.” Neither the question nor the response is valid.
The Danish twin study on which this conclusion is based identified no concordant monozygotic twin pairs and only two concordant dizygotic pairs from a sample of 37 338.1 By modelling the reported numbers of confirmed cases of rheumatoid arthritis, we calculate that the study had 80% power to detect a heritability of 65% and 90% power to detect a heritability of 75%. Thus there was insufficient power to detect a genetic influence on rheumatoid arthritis equivalent to the 60% heritability estimated in the two most recent twin studies of the disease.2
The point estimates of concordance are also cause for concern. The prevalence of rheumatoid arthritis at 0.15% seems low, indicating possible deficiencies in screening. The absence of the middle aged birth cohort severely limits the sample's representativeness. The failure to identify a single monozygotic concordant pair (we know that they exist: cases identified in recent studies have been characterised in detail3) indicates bias in the zero concordance estimate.
The difficulties inherent in designing twin studies of rheumatoid arthritis inspired the two largest and most recent studies in Finland and the United Kingdom, which yielded remarkably similar results despite having contrasting sampling strategies.2 Potential limitations of these studies are incorrectly highlighted in the Danish report. Cases of ankylosing spondylitis were specifically excluded in the Finnish study. The classification of rheumatoid arthritis entailed applying specified criteria. In the United Kingdom study the similarity in concordance among pairs recruited in the media and by general practitioners gave no reason to suspect disproportionate recruitment of concordant monozygotic pairs.
Using the results of twin studies to attach the label “genetic” or “environmental” to disease is inappropriate. This is readily appreciated by considering the classical examples of phenylketonuria (where an inherited disease is expressed only in the context of specific environmental exposure) and tuberculosis (which shows a greater concordance in monozygotic than dizygotic twins).4 The contemporary value of estimating heritability from twins is in establishing the likely success of specific strategies to detect the action of individual genes. In rheumatoid arthritis the established importance of genetic variation in HLA and the identification of new genetic regions linked to the disease is ample support for an important genetic contribution.5