Over 60,000 women are treated for cervical intraepithelial neoplasia (CIN) each year in England, most by excision. Management of women who have incomplete excision is controversial and the subject of much debate. Consequently, the completeness of excision is often ignored in the planning of subsequent treatment. We aimed to assess the effect of completeness of excision on the risk of post-treatment disease. We undertook a meta-analysis of studies published between Jan 1, 1960, and Jan 31, 2007, that studied the risk of post-treatment disease (ie, CIN of any grade or invasive cancer) in relation to completeness of excision. Studies were included if they described treatment of CIN by excision; numbers of women with involved margins; prevalence of and numbers of women with post-treatment disease in relation to margin status. Criteria for post-treatment disease had to be stated as a defined abnormal cytology or histology. Studies were excluded if they described treatment of cervical glandular intraepithelial disease (CGIN); if all or nearly all women had reflex hysterectomy done soon after initial treatment; if women were immunosuppressed (eg, if they were HIV-positive); or if no control group with disease-free margins was used. The endpoint of our analysis was the relative risk (RR) of post-treatment disease in those whose treatment histology suggested that excision was complete compared with those in whom excision was incomplete or uncertain. RR meta-analysis was done by use of a random effects model. The initial Medline search identified 1756 publications, from which 125 publications were short-listed. Of these, 65 and one unpublished study met our inclusion criteria; therefore, 66 studies were included in this meta-analysis. These studies described findings in 35,109 women of whom 8091 (23%) had at least one margin of the excision biopsy involved with disease. After incomplete excision, RR of post-treatment disease of any grade was 5.47 (95% CI 4.37-6.83) and RR of high-grade disease (ie, CIN 2 or 3, or high-grade squamous intraepithelial lesion) was 6.09 (3.87-9.60) compared with the reference group who had complete excision. High-grade post-treatment disease occurred in 597 of 3335 (18%) women who had incomplete excision versus 318 of 12 493 (3%) women who had complete excision. Incomplete excision of CIN exposes women to a substantial risk of high-grade post-treatment disease. Some of these women would be safer with a second treatment, especially if deep margins are involved, but most will need close follow-up for at least 10 years. Every effort should be made to avoid incomplete excision. Adding extensive ablation in the treatment crater to compensate for inadequate excision should be avoided because this might delay detection of inadequately treated invasive disease and because the effectiveness of additional ablation to destroy any residual CIN cannot be assessed. Furthermore, extensive ablation does not decrease any risk of preterm delivery in subsequent pregnancies.