Letters Type 2 diabetes and cancer

Authors’ reply to editorial linked to their umbrella review of meta-analyses of observational studies on type 2 diabetes and cancer

BMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h711 (Published 11 February 2015) Cite this as: BMJ 2015;350:h711
  1. Konstantinos K Tsilidis, assistant professor of epidemiology1,
  2. John PA Ioannidis, professor2
  1. 1Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, 45110, Greece
  2. 2Stanford Prevention Research Center, Department of Medicine, and Department of Health Research and Policy, Stanford University School of Medicine, and Department of Statistics, Stanford University School of Humanities and Sciences, Stanford, CA, USA
  1. ktsilidi{at}cc.uoi.gr

We thank Satija and colleagues for their thoughtful editorial on our umbrella systematic review.1 2 They interpret the evidence linking type 2 diabetes to multiple cancers as being strong overall,1 but we think that more caution is needed. Prediction intervals are increasingly thought to be the best way to recognise uncertainty in meta-analyses.3 Broad prediction intervals and the large between study heterogeneity seen for some of the associations between type 2 diabetes and cancer should therefore temper statements about the presence of strong and consistent effects. Although biological heterogeneity is certainly plausible, subgroup analyses in epidemiology have notoriously poor credibility. As for evidence on mortality, it should simply be acknowledged that it is more limited.

Our major concern, however, is their conclusion that the association between type 2 diabetes and cancer should be seen as causative and should lead to screening interventions and lifestyle changes in patients with diabetes. This is a bold statement that would require randomised evidence to justify it. Many proposed screening tests lack evidence of effectiveness or have been shown to be useless even in the general population,4 let alone in subgroups. The reputation of epidemiology has been damaged by interpreting observational associations as causative and basing interventions and public policy recommendations on them.5 We believe that this recurrent misconception should be avoided.


Cite this as: BMJ 2015;350:h711


  • Competing interests: None declared.


View Abstract