We enrolled 85 patients with invasive cervical cancer and collected cervicovaginal lavage samples at each clinical visit for diagnosis, staging, treatment, and follow-up. Lavage samples were tested by L1 consensus polymerase chain reaction for human papillomavirus (HPV). Results were compared with HPV demonstrated in tumor tissue and the clinical status at time of sample collection. Sensitivity and specificity of the lavage for detection of tumor HPV, determined on the basis of results of tests on lavage samples collected prior to therapy, were found to be 56% and 76.9%, respectively. The proportion of lavage samples detecting tumor HPV decreased significantly with treatment, from 0.54 at diagnosis to 0.03 at complete response (P < .001). Local treatment failure was associated with increased detection of tumor HPV; however, no samples were positive prior to clinically detected treatment failure. These results suggest that cervicovaginal lavage is not an effective sampling method for epidemiological analysis of HPV in cervical tumors.