Case selection bias affected results

BMJ 1996;312:1156 (4 May)

Letters

Case selection bias affected results

EDITOR,--Charles R Gillis and David J Hole provide clear evidencethat patients of specialist surgeons have a higher survivalrate from breast cancer than those of non-specialist surgeons.¹However, they go on to conclude that this association is oneof cause and effect, rather than due to differences in the casesreferred to the two groups of surgeons (case selection bias).Their reason is that the difference in outcome persisted afteradjustment for some prognostic factors; but statistical techniquescan correct for case selection bias under only three conditions,none of which held in the study reported.

Firstly, the correction must take account of all prognosticfactors, but Gillis and Hole did not include whether metastaticdisease was present, and information on histological grade wasavailable in only a minority of patients. Furthermore, the differencein prognosis between breast cancer patients is only partiallyexplained by the known prognostic factors, . . . [Full text of this article]

Add to CiteULike CiteULike Add to Complore Complore Add to Connotea Connotea Add to Del.icio.us Del.icio.us Add to Digg Digg Add to Reddit Reddit Add to StumbleUpon StumbleUpon Add to Technorati Technorati What's this?

Relevant Article

Survival outcome of care by specialist surgeons in breast cancer: a study of 3786 patients in the west of Scotland: Charles R Gillis and David J Hole
BMJ 1996 312: 145-148. [Abstract] [Full Text]