Affordable Access

Publisher Website

Study design and statistics in the epidemiology of breast cancer

Authors
Publication Date
Identifiers
DOI: 10.1038/sj.bjc.6603370
Keywords
  • Letters To The Editor
Disciplines
  • Biology
  • Design
  • Education
  • Medicine

Abstract

Letter to the Editor Study design and statistics in the epidemiology of breast cancer L Maraqa*,1 and M Lansdown1 1Breast Unit, Lincoln Wing, St James’s University Hospital, Beckett Street, Leeds LS9 7TF, UK British Journal of Cancer (2006) 95, 1301. doi:10.1038/sj.bjc.6603370 www.bjcancer.com Published online 3 October 2006 & 2006 Cancer Research UK � � � � � � � � � � � � � � � � � Sir, Your recent article by Garcı´a-Closas et al (2006) raises interesting discussion points regarding tumour behaviour in Poland. If correct, these results may have implications for health care in Poland, but we disagree that the findings can be generalised to other populations, particularly when the reported ER negativity was 41%. There were several flaws in the study design and analysis that make such a conclusion unsafe. Firstly, we understand mammographic screening is not routine in Poland (reflected in the low incidence (6%) of DCIS). The authors should have made clear whether mammography was in fact offered to equal numbers in the study and control groups. It is possible that the results were biased by differential access to and uptake of mammography between the groups; perhaps, better educated women were both more likely to take HRT and seek mammograms. Indeed, the absolute differences for better education between the two arms (10%, unadjusted OR¼ 1.89) and history of a previous mammogram (8%, un- adjusted OR¼ 1.39) support this suggestion. Another potential error lies in the histology. The rate of agreement between the original Polish and the subsequent American analysis varied from 18 to 80%, depending on tissue type. However, nearly one in five (n¼ 428) did not benefit from an American review, yet were still included in the study. With such high rates of discordance, it was hazardous to assume the original diagnosis was correct in the nonreviewed group. If the ratios for error are the same as the reviewed group, then at least 4.9% of all samples would have been misclassified, thus chall

There are no comments yet on this publication. Be the first to share your thoughts.