Re: Comparison of three management strategies for patients with atypical squamous cells of undetermined significance: baseline results from a randomized trial.

In their study of alternative management strategies for women with equivocal Pap smears, Solomon et al. (1) concluded that “HC 2 [Hybrid Capture 2TM] testing for cancer-associated HPV [human papillomavirus] DNA . . . has greater sensitivity to detect CIN3 [cervical intraepithelial neoplasia grade 3] or above and specificity comparable to a single additional cytologic test indicating ASCUS [atypical squamous cells of undetermined significance] or above.” Solomon et al. (1) reported that the HPV test had a 96.3% sensitivity to detect CIN3 or higher (CIN3+) and a 10.0% positive predictive value in a population with a 5.1% prevalence of CIN3+. However, these data imply that the specificity of the HPV test was only 53.4%. Solomon et al. further reported that a repeat Pap smear based on the LSIL+ (i.e., low-grade squamous intraepithelial lesion or higher) had a 64.0% sensitivity to detect CIN3+ and a 14.3% positive predictive value in the same population. From the latter data, I calculate that the specificity of the follow-up Pap smear based on the LSIL+ criterion was actually 79.4%. My calculations are given in Table 1, which shows the distributions of HPV test results (panel A) and follow-up Pap smear results based on the LSIL+ criterion (panel B) in a standard population of 10 000 women with atypical squamous cells on initial Pap smear. The trade-off between the higher sensitivity of the HPV test and the higher specificity of the repeat cytology based on LSIL+ becomes especially important when the prevalence of disease is as low as 5.1%. For every 10 000 women with equivocal Pap smears, comparison of panels A and B shows that the HPV test will correctly identify 165 additional cases of high-grade cervical neoplasia (i.e., 491 − 326). However, at the same time, the test will mistakenly send to colposcopy an additional 2465 women who do not have precancerous or cancerous cervical lesions (i.e., 4419 − 1954). The results reported by Solomon et al. (1) were widely misinterpreted by the media (2,3), which highlighted the finding that the HPV DNA test had a negative predictive value of 99.5% in the detection of high-grade cervical neoplasia. However, the negative predictive value is the probability that a patient with a negative HPV test did not have highgrade cervical neoplasia. By contrast, the specificity is the probability that a woman without high-grade cervical neoplasia had a negative HPV test (4). Unless the HC 2 test can be improved to identify the specific oncogenic subtypes of HPV DNA, it may not be superior to repeat cytology. Barring such improvements, we will need to balance the benefits of early detection of cervical neoplasia in 165 additional cases per 10 000 women against the higher costs of the HC 2 test and the costs and trauma of unnecessary colposcopy in 2465 additional cases per 10 000 women.