Authors respond to criticism of rheumatoid arthritis twin study

BMJ 2002; 325 doi: https://doi.org/10.1136/bmj.325.7360.391/a (Published 17 August 2002) Cite this as: BMJ 2002;325:391
  1. Anders Svendsen, consultant,
  2. Niels V Holm, associate professor,
  3. Kirsten Kyvik, associate professor,
  4. Per Hyltoft Petersen, master of science,
  5. Peter Junker, professor
  1. Department of Rheumatology, Odense University Hospital, 5000 Odense C, Denmark
  2. Department of Radiotherapy and Clinical Oncology, Odense University Hospital
  3. Danish Twin Registry, Department of Public Health, Epidemiology, University of Southern Denmark, Odense
  4. Department of Clinical Chemistry, Odense University Hospital
  5. Department of Rheumatology, Odense University Hospital

    EDITOR—The letter by MacGregor et al in response to our twin study focuses on three points: heritability, bias, and the contribution of specific genes to the aetiology of rheumatoid arthritis.1

    MacGregor et al state that two previous twin studies on rheumatoid arthritis have reported comparable heritability estimates at approximately 60%. But a heritability estimate is a theoretical concept relying on several assumptions which cannot at present be met in rheumatoid arthritis. Besides, heritability estimates are population specific since the variation of environmental and genetic effects may not be identical in different geographical areas and ethnic populations.

    Embedded Image


    In our study, twins younger than 42 were 92% of the total population, but an age adjusted estimate accords well with the latest population based study from Norway in which the prevalence of rheumatoid arthritis between 20 and 79 years was 0.437.2 Prevalences exceeding 1% were found only in people older than 60. Macgregor et al state that the concordance estimate of zero is by definition biased since monozygotic concordance pairs do exist. But bias is a systematic error originating from inherent problems including selection, information, and confounding. In our study all possible measures were taken to minimise bias—for example, by contacting all available people in the entire Danish twin population irrespective of disease status, by using multiple sources of information and by doing specialist examination of all candidate probands.

    Furthermore, whether concordant pairs exist in excess of coincidence is questionable. In the study by Silman et al 19 monozygotic and 39 dizygotic twin pairs would be anticipated to be concordant by chance.3 This study, however, identified only 14 monozygotic and four dizygotic concordant pairs—that is, there were no concordant monozygotic pairs in excess of what would be expected by chance. Therefore, we cannot accept the inference that our study is biased by definition because we failed to identify concordant monozygotic pairs, given that the literature on twin methodology points to the risk of bias towards overascertainment of monozygotic pairs, and concordant monozygotic pairs in particular.4

    MacGregor et al argue that the genetic variation in HLA in rheumatoid arthritis and the recent identification of genetic regions linked to the disease is evidence of an aetiological contribution from specific genes. The finding of several possible genetic risk factors is interesting but their definite significance still remains to be settled, just as for the bulk of putative environmental effectors proposed so far.5 Experience from diabetes, which has a much larger genetic contribution in twin studies, suggests that this will be a tedious process.


    View Abstract

    Sign in

    Log in through your institution

    Free trial

    Register for a free trial to thebmj.com to receive unlimited access to all content on thebmj.com for 14 days.
    Sign up for a free trial