Grading of malignancy in squamous cell carcinomas of the uterine cervix is based on qualitative, morphologic examination and suffers from poor reproducibility. Using modern stereology, unbiased estimates of the three-dimensional, volume-weighted mean nuclear volume (nuclear vv), were obtained in pretreatment biopsies from 51 patients treated for cervical cancer in clinical Stages I through III (mean age of 56 years, follow-up period greater than 5 years). In addition, conventional, two-dimensional morphometric estimates of nuclear and mitotic features were obtained. DNA indices (DI) were estimated by flow cytometry. Finally, the semiquantitative malignancy grade score value (MGS) was determined according to previously published methods. Estimates of nuclear vv were on average increased in euploid lesions (2P = 0.01), but the overall relationship between nuclear vv and DI was poor. Different clinical stages of disease did not differ with regard to nuclear vv (2P = 0.99) and DI (2P = 0.56). No relationship was disclosed between MGS and nuclear vv (2P = 0.85). Single-factor analysis showed prognostic impact of clinical stage of disease (2P = 0.0001) and DI (2P = 0.04), whereas estimates of nuclear vv were only of marginal prognostic significance (2P = 0.07). However, Cox multivariate regression analysis showed independent prognostic value of patient age and nuclear vv along with clinical stage and DI. All other investigated variables were rejected from the model. A prognostic index with highly distinguishing capacity between prognostically poor and favorable cases was constructed (2P = 1.9 x 10(-7)). It is concluded that realistic estimates of nuclear volume are independent of nuclear DNA content and are of prognostic value for objective malignancy grading in patients with squamous cell carcinoma of the uterine cervix.