Recent reports from large cytopathology laboratories in USA and China, linking prior testing of cervical cell specimen with the subsequent occurrence of invasive cervical cancer revealed significantly higher false-negativity rates for HPV testing than for the combination of HPV testing and cytology. The authors concluded that co-testing should be the preferred screening policy.
Prior meta-analyses on the accuracy of HPV and cervical cytology to detect underlying CIN2+ or CIN3+ were updated and extended for the outcome of invasive cervical cancer. The current report assesses the sensitivity of high-risk HPV testing and compares it with that of co-testing. Standard random effects models for pooling of binomial test accuracy data were used.
The sensitivity pooled from 16 studies was 85% (95% CI: 80-90%, I2=26%, p for heterogeneity<0.001) and 95% (95% CI: 91-99%, I2=84%, p<0.001) for hrHPV testing and co-testing, respectively. Co-testing was 11% more sensitive than hrHPV (relative sensitivity=1.11 [95% CI: 1.08-1.15]). The absolute sensitivity varied by test and interval between test and cancer diagnosis (p<0.001), however relative sensitivity did not show significant heterogeneity (p=0.26).
Although the meta-analysis confirms the aforementioned reports, flaws on the recent linkage studies should be noted on design and conclusions. Linkage studies should only include screening histories and exclude recent tests immediately preceding a diagnosis of cancer. Moreover, sensitivity estimates for cancer are not the most relevant criterion for decision making. Using the KPNC screening, we estimated a NPV for cervical cancer of 99.995% vs 99.998% of hrHPV vs cotesing, respectively. This would correspond with ~three additional cases of cancer detected but also ~4,500 more false-positive cases per 100,000 screened with co-testing compared to a general hrHPV test only.
In a HPV-based cervical cancer screening, an extensive virological audit of archived specimen from each HPV-negative case of cancer case should be set up.