Effects of observer variation in population screening for cervical carcinoma

Abstract

A stratified random sample of 100 cervical smears was taken from routine population screening material and graded by three cytological laboratories. The sensitivity of the smear test for detecting cervical malignancies varied from 72 to 88%—a difference of 16%; for all serve epithelial abnormalities the sensitivity varied from 52 to 71% (a difference of 19%). The histological material pertaining to these smears was re-examined `blindly' and then sent for diagnosis to a second pathologist. Using each set of histodiagnoses as reference, the sensitivity of the test for detecting cervical malignancies varied from 77 to 85%—a difference of 8%; the sensitivity for all epithelial abnormalities varied from 62 to 76% (a difference of 14%). The significance of these observer variations on the efficacy of screening for cervical carcinoma is discussed.