AIMS--To determine the histological features in CIN3 associated with or predictive of subsequent microinvasion. METHODS--The histological appearances of CIN3 accompanying 120 cases of microinvasive carcinoma of the uterine cervix were retrospectively studied. Major features were defined as those present in greater than 80% of cases of microinvasive carcinoma (MICA) and less than 10% of control cases of CIN3. One hundred cases of CIN3, 36 showing all, and 64 lacking all, of the major features associated with microinvasion, as defined in the retrospective study, were prospectively studied. Deeper levels were cut to exclude the presence of microinvasion in the original biopsy specimen and negative cases were followed up for a period of up to 18 months in order to assess rates of recurrence or progression. RESULTS--The major features identified in CIN3 associated with microinvasive carcinoma were extensive involvement of surface epithelium and deep endocervical crypts by expansile CIN3, luminal necrosis, and intraepithelial squamous maturation. Other features more commonly present in MICA associated CIN3 than in controls included frequent mitosis and apoptosis, pericryptal concentric fibroplasia, pericryptal inflammatory infiltrate, pronounced cellular pleomorphism, nuclear changes (distinct nucleoli and chromatin clearing), and the emergence of streams of darkly stained spindle cells oriented at right angles to the basement membrane. In the prospective study 83% of cases illustrating the major MICA associated features revealed evidence of MICA or frank invasion either on serial sections of the original biopsy or on subsequent biopsy. None of the 64 cases of CIN3 that lacked these features showed evidence of invasion on serial sections or on further follow up over 18 months. CONCLUSIONS--The data strongly suggest that cases of CIN3 which have a higher probability of association with or rapid progression to invasive disease can be identified. When these features are present in a biopsy specimen of CIN3, serial sections should be performed to exclude the presence of microinvasion. Closer clinical follow up of these patients may be needed.