The objective of this study was to quantify the incidence of underlying cervical intraepithelial neoplasia (CIN) among women referred for colposcopy with three consecutive inadequate smears. The design was a retrospective cohort study analysing data from a regional colposcopy database at Cervical Screening Wales. Women who were referred to all the colposcopy clinics in Wales with three consecutive inadequate smears, the third inadequate smear being taken between 1 April 2001 and 31 March 2002 constituted the study population. The results of the fourth smear taken at the colposcopy clinic after three consecutive inadequate smears, the worst biopsy results from any of the subsequent colposcopies and the relationship between the result of the fourth smear taken at colposcopy clinic and any histology result were the main outcome measures. The accuracy of the colposcopic opinion was also examined. Of the 433 women identified as having been referred because of three consecutive inadequate smears, 302 were linked to either a subsequent smear and/or a biopsy result. An adequate smear result was available for 85% of these women when the smear was taken in the colposcopy clinic; 77% were reported as negative and 8% were abnormal. Of the 347 women seen in the colposcopy clinic, high-grade CIN was seen in 3% and low-grade lesion in 8%. The sensitivity and specificity of the fourth inadequate smear test in predicting underlying CIN were 15% and 84% respectively, with a positive predictive value of 8%. The sensitivity and specificity of colposcopy in predicting histological CIN among patients with three inadequate smears was 70% and 49%, respectively, and the positive predictive value was 44%. This study raises the question as to whether three consecutive inadequate smears should be considered as an indication for colposcopy, or merely for a further smear to be taken in circumstances where there is a greater likelihood getting an adequate result.