Self-selection bias, a study that illustrates it

March 21, 2021

Cécile Bour, MD

One of our faithful and wise readers, whom we thank, asked our opinion on a study by Tabar and Duffy, recently published but we had not mentioned. 

https://pubs.rsna.org/doi/10.1148/radiol.2021203935

According to this publication, there would be a significant reduction in mortality from breast cancer in patients who were monitored.

Here are our comments. We did not mention this study because there is a huge and well-known selection bias, namely that women who do not participate in screening are very different from those who do; and this bias can explain the results as well as the screening itself in this population.

On Medscape we can read:

https://www.medscape.org/viewarticle/933105

One of the experts who was approached by Medscape Medical News to comment on the new study, Philippe Autier, MPH, MD, PhD, from the University of Strathclyde Institute of Global Public Health at the International Prevention Research Institute, Dardilly, France, questioned the methodology of the study. "This method is incorrect simply because women attending screening are different from women not attending screening," he said. "The former are more health aware and have healthier behaviors than the latter, and this is a well-known fact and supported by the literature."

Dr Autier emphasized that it is practically impossible to control for that bias, which is known as confounding by indication.

"The statistical methods used for attenuating the so-called self-selection are very approximate and based on unverified assumptions," he said. "For this reason, the Handbook on Breast Cancer Screening produced by the International Agency for Research on Cancer clearly stated that 'observational studies based on individual screening history, no matter how well designed and conducted, should not be regarded as providing evidence for an effect of screening,' and the methodology in this paper has never been recommended by the [agency]."

A better way of conducting this type of study would have been to show the incidence trends of advanced-stage breast cancer in Sweden for the entire female population aged 40 years and older, he asserts. Dr Autier used that methodology in his own study in the Netherlands, as previously reported by Medscape Medical News.[4]That study foundt hat in the Netherlands, screening mammography over a period of 24 years among women aged 50 to 74 years had little effect on reducing rates of advanced breast cancer or mortality from the disease.

Cancer Rose est un collectif de professionnels de la santé, rassemblés en association. Cancer Rose fonctionne sans publicité, sans conflit d’intérêt, sans subvention. Merci de soutenir notre action sur HelloAsso.


Cancer Rose is a French non-profit organization of health care professionals. Cancer Rose performs its activity without advertising, conflict of interest, subsidies. Thank you to support our activity on HelloAsso.