According to the worldwide cancer incidence database maintained by the World Health Organization, cervical cancer is the third most common malignancy in women worldwide, exceeded in incidence only by breast and colorectal cancers (1). Cervical cancer is an important cause of mortality in women worldwide, and the cervix is a well-established clinical and histopathologic model of carcinogenesis. Human papillomavirus (HPV) is the major etiologic agent in cervical cancer. The cervix is easily

more »

... ible for examination, and colposcopy provides a visual model of angiogenesis and tumor development, making the cervix a good model for preventive interventions (2). The precursor lesions to cervical cancer are squamous intraepithelial lesions (SILs) or cervical intraepithelial neoplasia (CIN). SILs are used in the cytologic classification, and CINs are used in the histopathologic classification. Several chemoprevention trials have been conducted in women with SIL/CIN by use of retinoids, micronutrients, ␣-difluoromethylornithine, and indole-3-carbinol. Although the results of phase I/IIa studies have appeared to be promising, the phase IIb/III studies have usually been negative (3-13) ( Table 1) . 1 The failure of the phase IIb/III studies to demonstrate an effect may be due to several factors not being adequately considered in the design of these trials. Advances in the design and implementation of clinical trials have been made over the last decade. Factors that need to be considered in the design of cervical chemoprevention trials are as follows: the natural history of the disease in the absence of any intervention, the optimal range of anticipated clinical response to the randomly assigned intervention, and the validity and predictiveness of the primary (histologic regression) and secondary (surrogate endpoint biomarker modulation) outcome measures (14-16). The likelihood that phase II cervical chemoprevention trials will be uninformative can be minimized with careful attention to critical features of study design: enrollment of a sufficient number of patients to permit the study to reveal differences in response rates; careful selection of the type and dose of chemopreventive agent, based on results from preclinical studies and phase I and IIa clinical trials; accurate classification of patients' disease status, both at enrollment and at study end; and selection of appropriate primary and secondary outcome measures. To ensure that cervical cancer chemoprevention trials have appropriate statistical power, the natural history of the disease must be taken into account. Patton wrote an extensive review of the natural history of SIL/CIN in the 1950s [reviewed in (17,18)], and two recent reviews of the natural history of SIL/CIN have also been reported (17, 18) . Estimated rates of regression of CIN vary widely, both within each CIN grade and among studies. The regression rate decreases as the CIN grade increases, but estimates of spontaneous regression based on randomized trials vary from 27% to 66%, depending on the study and the entry criteria (Table 1) . Apparent spontaneous regression of CIN has been observed from the natural history studies, from 57% regression of CIN1 and 43% regression of CIN2 to 32% regression of CIN3 (18). Thus, with the exception of the studies by Meyskens et al. (3), de Vet et al. (8), and Childers et al. (13), many of the published phase IIb trials did not have sufficient power to detect a clinically significant difference in response rate. The dose and duration of medications used in chemoprevention trials should be selected on the basis of results from phase