Epidemiologic studies have demonstrated strong and consistent associations between the detection of human papillomavirus (HPV) type 16 DNA and the risk of cervical intraepithelial neoplasia (CIN) and cervical cancer. However, HPV16 is also the most common type of HPV in the normal population, and only a minority of women with HPV16 infection develop cervical cancer. Studies of genomic heterogeneity in HPV16 have demonstrated the presence of multiple variant forms in all human populations examined to date. It is conceivable that the natural variants of HPV16 in a given population may not have the same biologic behavior. This study was designed to determine the association between natural variants of HPV16 and the risk of biopsy-confirmed CIN 2 or 3, the most important precancerous lesions of the uterine cervix. Prospective studies were conducted among 1) women attending a university and 2) women presenting to a sexually transmitted disease clinic. Subjects were eligible for inclusion in this investigation if the initial cytologic findings did not reveal CIN 2-3 and HPV16 DNA was detected by means of a polymerase chain reaction (PCR)-based method in one or more cervical or vulvovaginal samples. Eligible subjects were followed every 4 months with cervical Pap smears and colposcopic examinations. Women were referred for biopsy if cytology or colposcopy suggested CIN 2-3. Two groups of HPV16 variants, prototype-like and nonprototype-like, were determined by means of single-strand conformation polymorphism (SSCP) analysis of PCR products from the noncoding region of the viral genome. Representative SSCP patterns from HPV16 variants were further characterized by direct DNA sequencing of the PCR products. Relative risks (RRs) and 95% confidence intervals (CIs) were calculated by Cox regression analysis. Prototype-like variants accounted for 79% of the HPV16 detected in university students and 86% of the virus detected in patients presenting to the sexually transmitted disease clinic. CIN 2-3 was confirmed by biopsy in nine of 57 HPV16-positive women attending the university and in 10 of 66 HPV16-positive women presenting to the sexually transmitted disease clinic. Among university students, those with HPV16 nonprototype-like variants were 6.5 (95% CI = 1.6-27.2) times more likely to develop CIN 2-3 than those with prototype-like variants. A similar association was observed among women presenting to the sexually transmitted disease clinic (RR = 4.5; 95% CI = 0.9-23.8). This study suggests that the risk of developing CIN 2-3 is not the same with all variants of HPV16 and that nonprototype-like variants confer a greater risk compared with prototype-like variants. The important genomic differences underlying this increased risk of CIN 2-3 remain to be determined.