The KPNC co-testing program provided us with an unparalleled opportunity to evaluate the real-world clinical effectiveness of concurrent HPV testing with cytology in a large and diverse U.S. screening population followed for several years. Women testing HPV-negative had extremely low risk of CIN3+ or cancer over five years, regardless of whether they had normal cytology or minor abnormalities. Although HPV-positive women with cytologic abnormalities had the highest risks, HPV-positive women with normal cytology accrued substantial risk of CIN3+ and cancer over 5 years. Co-testing was less able to identify women at high risk at their second visit in approximately 3 years after an HPV-negative/Pap-negative enrollment test. Their co-test results were down-staged to safer co-tests and the CIN3+ risk for each possible second co-test result was generally diminished from the risk associated with that co-test result at enrollment.
Enrollment HPV-negative/Pap-negative women had an extremely low cervical cancer risk of 3.2/100,000 women/year over five years. This risk is comparable to the risk of vulvar cancer in northern California in women aged 30 and older (3.1/100,000 women/year; SEER 2003-2007), another potentially preventable cancer that is too rare to justify organized efforts at prevention. This finding supports lengthening the screening interval for HPV-negative/Pap-negative women to five years, as has already been done in many European countries40
. The 80% lower risks of CIN3+ among HPV-negative/Pap-negative women aged 50 or older (vs. aged 30-34), or for HPV-negative/Pap-negative women with no history of previous abnormal cytology, raise the possibility that some subgroups of HPV-negative/Pap-negative women could be safely screened at an even longer interval.
A single HPV-negative test sufficed to reassure a woman of extremely low risk of CIN3+ or cancer for five years. We found that negative cytology provided no extra reassurance against cancer beyond that conferred by an HPV-negative test result. The practically equal 5-year risks of CIN3+ or cancer for HPV-negative/Pap-negative women and HPV-negative/ASC-US women provides powerful evidence that women with HPV-negative/ASC-US could be safely screened in routine clinical practice at the same extended interval as HPV-negative/Pap-negative women.41-43
These findings are consonant with the biological fact that carcinogenic HPV is necessary to cause almost every cervical cancer. Finally, the low yields of LSIL or worse Pap smear (0.5%) and ASC-H/HSIL/SCC Pap smears (0.05%) in HPV-negative women may not be sufficient to justify Pap smear tests for HPV-negative women. Our findings strongly suggest that primary HPV testing, with HPV-positive tests triaged by cytology (or other tests with high specificity), a strategy which might preserve nearly all the safety of co-testing while reducing the number of Pap tests by 95% in our population, could be more efficient than co-testing (as has been suggested by others13, 40
Although cytologic abnormalities indicated prevalent disease, HPV testing predicted future disease much better than cytology. For example, HPV-positive/Pap-negative women, the majority of HPV-positive women (73%), had low prevalent disease risk at enrollment (in part because these women were rarely sent for immediate colposcopy), yet accrued considerable incident disease risks over five years that were comparable to women with ASC-H/HSIL/SCC cytology. Importantly, 17 of the 27 adenocarcinomas were found in HPV-positive/Pap-negative women. Follow-up of HPV-positive/Pap-negative women must strike the difficult balance of being stringent yet avoiding excessive early intervention45-46
. Thus, it is imperative to develop new biomarkers (such as HPV genotyping43, 47
} or p16INK4a
) to better triage HPV-positive/Pap-negative women into finer risk categories49
, especially risk of adenocarcinoma, whose incidence is on the rise in the U.S.50
We expected that the ability of co-testing to separate high-risk women from low-risk women would diminish at the second co-test after an enrollment HPV-negative/Pap-negative co-test because there are always fewer high-risk women with prevalent CIN3+ at the second and subsequent screens. Therefore, women HPV-positive at the second co-test approximately three years after an enrollment HPV-negative/Pap-negative had lower CIN3+ and cancer risks than women testing HPV-positive at enrollment. In contrast, a second consecutive HPV-negative/Pap-negative test (at a median of 2.9 years after the first) appeared to offer no detectable additional reassurance against CIN3 or cervical cancer than the first HPV-negative/Pap-negative co-test. These findings suggest the futility of the common practice of yearly HPV testing for HPV-negative women “just to be sure”18
and support guidelines15-16
that HPV-negative/Pap-negative women should not be re-screened before at least 3 years.
Although KPNC practices are not typical in many ways, because KPNC serves a large and diverse population using providers without special qualifications or special training to participate, we believe that the KPNC experience serves as a large-scale “demonstration project” of what could realistically be achieved in real-life clinical practice. Consequently, non-compliance by clinicians or patients to protocols, imperfect diagnostic testing, and other imperfections are real-life complexities that should be appropriately reflected in our risk estimates. The value of our study is to reassure women, clinicians, and screening guidelines committees that the benefits of HPV testing seen in clinical trials and research cohorts, which might not represent routine clinical practice in a general population, are indeed observed in general practice with all its attendant complexities.
However, the KPNC experience may be difficult to apply to substantially different screening programs. For example, most women at KPNC with abnormal cytology were referred for colposcopy at enrollment, but 73% of HPV-positive women had normal cytology and few of them were referred for colposcopy at enrollment. Thus, an alternate screening program that sends more HPV-positive/Pap-negative women to colposcopy would likely observe greater CIN3+ risk at enrollment than we did in KPNC. For another example, although we observed low 5-year cancer rates in HPV-negative/Pap-negative women, a program that actually implements a 5-year screening interval for these women would observe a slightly higher cancer rate than that observed in KPNC because the second co-test at 3 years, while having little effect, still excised the few CIN2 or CIN3 lesions that could have progressed to cancer in two years. For a final example, the cancer risk for all HPV-negative women in a primary HPV testing program where all HPV-negative women have extended screening intervals should be somewhat higher than the cancer risk in all HPV-negative women in KPNC, because KPNC referred the 0.5% of HPV-negative women with LSIL or worse abnormalities for colposcopy where excision of CIN2 or CIN3 lesions may have prevented a few future cancers.
In summary, our findings demonstrate that adding HPV testing to cytology screening promoted earlier identification of the women at high risk of cervical cancer (especially adenocarcinoma) and allowed safe 3-year screening intervals for HPV-negative/Pap-negative women that reduced the burden of screening on patients and clinicians. Furthermore, our findings suggest that 5-year screening intervals for HPV-negative/Pap-negative women may be safe and that HPV testing without adjunctive cytology may be sufficiently sensitive for primary cervical cancer screening. The results of co-testing in 330,000 women over five years at KPNC definitively demonstrates that concurrent HPV testing and cytology can be feasibly implemented in routine clinical practice to provide powerful prevention of cervical cancer.