Cain and Howett state that even low-risk subtypes of HPV can be found in cancers. In studies of more than 2000 invasive cervical cancers, however, only three were identified as having a low-risk type of HPV in the absence of a high-risk type (2-5). These studies would seem to suggest that detecting low-risk types of HPV may be unimportant, and the focus of screening should be to detect only high-risk types of HPV. Another point that needs clarification is the statement that pubescent and adolescent girls are "most likely to need access to screening": The annual incidence of invasive cervical cancer is 0.3 to 2.6 per 100,000 in women 15 to 24 years old living in the United States, compared with 14.9 to 16.8 per 100,000 in women older than 40 years (6). Indeed, a universal criticism of current screening in this country is the relative overscreening of low-risk young women compared with high-risk older women (7).

Cain and Howett also state that the "frequency of HPV testing will depend on the demographics of new infections." Yet screening frequency is primarily determined by the probability of a test being negative in a woman with disease--that is, on the false-negative fraction--and the average time required for a precursor to progress to an invasive cancer. Finally, the statement that "a positive HPV test does not confirm actual disease" implies that a positive result from cytology (a positive Pap smear) does confirm the presence of disease. Neither test, however, is diagnostic; instead, both cytology and HPV testing are screening tests, the purpose of which is to identify women with the highest probability of having a cervical lesion--not to confirm the presence of the lesion itself.

HPV testing and cytology need to be evaluated in terms of the test being positive in women with a high risk of disease and negative in women at low risk. To assess the real-world performance of these tests requires combining information on both test performance and disease prevalence. Several recently published studies have compared the relative performance of cytology and HPV testing using the Hybrid Capture II (HCII) HPV DNA assay (Digene Corporation, Gaithersburg, MD) for detecting high-grade cervical disease in women 35 years of age or older (8-11). In these studies, HPV testing showed a higher sensitivity for detecting high-grade SIL/cancer, ranging from 88 to 95%, than did cytology, the sensitivity of which ranged from 78 to 79%. The specificity of HPV testing, which ranged from 82 to 94%, was somewhat lower than that of cytology, which ranged from 97 to 99%. Although these results are promising, determining the real-world clinical effectiveness of using HPV DNA testing as part of a screening program will require analytical modeling methods to synthesize data on the costs and benefits of alternative screening strategies. Such models can combine data from multiple sources, extrapolate health benefits beyond the time horizon of a single clinical study, and evaluate different combinations of screening tests, screening intervals, and management options.

In the absence of such analytical modeling, the assessment by Cain and Howett (1) of the barriers to implementing HPV testing seems prematurely pessimistic. One of the principal barriers mentioned was the greater cost of a HPV test, at an average of $60 per test, compared with cytology, at $20 to $40. The cost differential, however, is less relevant than the cost-effectiveness of the screening program incorporating the test. A test is not cost-effective or cost-ineffective in and of itself; the cost-effectiveness of screening strategies will depend on the additional cost of a specific screening strategy relative to its additional clinical benefit, compared with the next least expensive strategy. Such a cost-effectiveness analysis conducted from a societal perspective, as recommended by the Panel on Cost-Effectiveness in Health and Medicine when evaluating health care policy, should include all costs and all consequences regardless of to whom they accrue (12). For example, it would need to incorporate the costs of litigation associated with cytology.

We are particularly concerned about statements by Cain and Howett (1) regarding possible psychological consequences of a positive HPV test, and the potential for overtreatment because of "false positives." According to Cain and Howett, "sensitizing even informed women to every subsequent variation in normal menstrual cycles as though they augured the onset of cancer is a concern." These ambiguous statements overlook the fact that women who test positive for high-risk types of HPV are not "false positives"; they are infected with a DNA tumor virus associated with more than 98% of cervical cancers (3). Although the majority of HPV-infected women will spontaneously clear their infections over time, these women are at significantly increased risk for having cervical disease or of developing a precursor over the next 2 to 5 years (13-15). It is difficult to believe that in this day and age most clinicians would adopt the paternalistic position of withholding information on a factor that greatly enhances a woman's risk of developing cervical disease out of concern that knowledge of her HPV status would convey insupportable anxiety. We have confidence that both the women undergoing testing and the clinicians providing it can be educated on the significance of a positive HPV DNA test and on the appropriate clinical response.

Tom C. Wright, Jr.
Department of Pathology
College of Physicians and Surgeons
Columbia University
New York, NY 10032, USA
Sue J. Goldie
Department of Health
Policy and Management
Harvard School of Public Health
Boston, MA 02115, USA

REFERENCES AND NOTES

1. J. M. Cain and M. K. Howett, Science 288, 1753 (2000) [Free Full Text] .
2. F. X. Bosch, et al., J. Natl. Cancer Inst. 87, 779 (1995) [Free Full Text] .
3. J. M. Walboomers, et al., J. Pathol. 189, 12 (1999) [CrossRef] [Web of Science] [Medline] .
4. C. Ngelangel, et al., J. Natl. Cancer Inst. 90, 43 (1998) [Abstract/Free Full Text] .
5. S. Chichareon, et al., J. Natl. Cancer Inst. 90, 50 (1998) [Abstract/Free Full Text] .
6. NCI Surveillance, Epidemiology, and End Results Program, Cancer Statistics Review: 1973-1996, www-seer.ims.nci.nih.gov, accessed 30 July 2000.
7. A. B. Miller, Am. J. Public Health 85, 795 (1995) [Abstract/Free Full Text] .
8. D. D. Davey, S. Woodhouse, P. Styer, J. Stastny, D. Mody, Arch. Pathol. Lab. Med. 124, 203 (2000) [Web of Science] [Medline] .
9. J. Cuzick, et al., Br. J. Cancer 81, 554 (1999) [CrossRef] [Web of Science] [Medline] .
10. M. Schiffman, et al., JAMA 283, 87 (2000) [Abstract/Free Full Text] .
11. L. Kuhn, et al., J. Natl. Cancer Inst. 92, 818 (2000) [Abstract/Free Full Text] .
12. T. C. Wright, L. Denny, L. Kuhn, A. Pollack, A. Lorincz, JAMA 283, 81 (2000) [Abstract/Free Full Text] .
13. M. R. Gold, J. E. Siegel, L. B. Russell, M. C. Weinstein, Cost-Effectiveness in Health and Medicine (Oxford Univ. Press, New York, 1996).
14. T. V. Ellerbrock, et al., JAMA 283, 1031 (2000) [Abstract/Free Full Text] .
15. G. Y. Ho, R. Bierman, L. Beardsley, C. J. Chang, R. D. Burk, N. Engl. J. Med. 338, 423 (1998) [Abstract/Free Full Text] .
16. T. C. Wright is currently a principal investigator on a clinical trial being conducted at Columbia University that is funded by Digene, the maker of an HPV DNA diagnostic test.

Response: We acknowledge that low-risk HPV types are rarely found in invasive cervical cancers, but disagree that detection is clinically unimportant. Low-risk HPVs are often asymptomatic, yet can cause significant problems, particularly for maternal/fetal transmission and immunocompromised women (1-3). The pattern of acquisition of low- versus high-risk subtypes may hold valuable information for developing cervical cancer screening strategies (4, 5).

Wright and Goldie state that young women are overscreened by the Pap test relative to high-risk older women, implying that HPV testing will be limited to older women. Although this might be a sound policy, it is not currently recommended. The HPV test is being recommended as a substitute for Pap screening; hence, the same demographic group presumably would be screened--and, indeed, women in all age groups have been screened in clinical trials of the HPV test. The United Kingdom has had great success reducing the incidence of cervical cancer by recommending a first cytology screening at age 20 followed by screening at least every 5 years until age 64 (in England and Wales) or age 50 (in Scotland). The American Cancer Society recommends first cytology at age 18 (or when first sexually active); following three negative annual screenings, women can be screened less often at the decision of their physician (6). Women are not screened exclusively to determine if they have cervical cancer but also to determine if they are harboring HPV pre-malignant lesions or other infections and if they are potentially infectious to others.

According to Wright and Goldie, screening frequency should be based on the magnitude of the "false-negative fraction" and the "average time required for a precursor to progress to an invasive cancer." The demographics of infection, including relative prevalence of low- and high-risk types and number and frequency of sexual partners for a given population, determine HPV infection risk and the risk of progression to cancer. Thus, young, sexually active women are at higher risk for acquisition of HPV infections than are older, monogamous, or sexually abstinent women. The younger women will, for the most part, clear their infections. Risk of acquiring HPV and risk of developing disease are thus distinct, and ideal screening frequency for either test is not determined. The use of HPV testing per se will not solve the problem of overscreening the young population.

We agree that Pap smears are not diagnostic; however, Pap smears are better prognosticators of disease than HPV testing. The specificity of cytology is higher (not equivalent as suggested by Wright and Goldie), and HPV testing is more likely to be positive in the absence of histologic abnormality. A 1999 study found that classical cytology had a sensitivity of 85.3% and a specificity of 94.9%, whereas HCII had a sensitivity of 97.1% and a specificity of 85.2% (7). A recent study of Costa Rican women yielded a sensitivity of 88.4%, a specificity of 89%, and a colposcopy referral rate of 12.3% for HCII, compared with a sensitivity of 77.4%, a specificity of 94.2%, and a colposcopy referral rate of 6.9% for conventional Pap smearing (8). In a study in Cape Town, South Africa, the false-positive rate for HCII was 17.1% (95% confidence interval, 15.1 to 19.3%), compared with 12.3% (95% confidence interval, 10.5 to 14.2%) for Pap testing (9). As to cost-effectiveness, no data have been put forward to suggest that the HCII assay could be performed less frequently and, therefore, at less cost than the Pap smear--and, although there may indeed be litigation associated with cytology, there will also be litigation associated with HPV testing.

It remains unclear how an HPV test will alter follow-up even when used concurrently with Pap smears, except potentially in women with ASCUS (atypical squamous cells of undetermined significance). These women incur both costs, because the HPV test is consequent to the cytology, but could avoid the cost of colposcopy if the HPV test is negative. One study found that HPV testing did not add to management decisions when low-grade squamous intraepithelial lesions (LSIL) were already found on Pap smears because the rate of positives was so high (8). Routine HPV testing is more expensive, even in a technologically advanced society; it seems extremely unlikely that the health care system can deliver effective HPV testing in developing countries, where effective delivery even of Pap screening has not been accomplished. In addition, many developing countries lack facilities for treatment of pre-cancers or cancers, even if a positive cytology or HPV test were obtained. In the face of these facts, it is unrealistic to suggest any one plan for screening for developing countries (10-12).

Current data comparing the specificity of cytology and HPV testing are based on a single sampling, rather than on repeated sampling, which would seem important in an infection that has both a high rate of spontaneous resolution and a long latent period for cancer development. And, although Wright and Goldie stress that "women who test positive for high-risk types of HPV are not `false positives,' " but "are at a significantly increased risk for having cervical disease or of developing a precursor over the next two to five years"--statements that would seem counter to their argument that testing of young women is unnecessary--we are not sure that a false positive always means that the subject is infected. There are other ways in which a test can be positive, including technical errors. Whatever the source of false positives, the data suggest that, even in controlled clinical studies, women are more likely, in the absence of tissue pathology, to have a positive HPV test than to have an abnormal Pap smear. Such problems are likely to be amplified if the test is generalized to the entire health care delivery system and to countries lacking sophisticated clinical laboratory infrastructures. We certainly do not advocate withholding information on risk factors, but we do suggest that conveying the results of a false positive test can unnecessarily raise patient anxiety.

Many of the objections by Wright and Goldie stem from studying groups from among the U.S. population. Unfortunately, the world is not demographically similar with respect to the prevalence of the virus or the age at first intercourse, which changes the risk profile for the disease by age. Indeed, there are clearly subpopulations even within the U.S. population that require different strategies (13-15). The pattern of acquisition of HPV subtypes in women age 30 and above, too, can change with changing sexual behaviors. The best combination of screening options is not yet evident, and the proper role for HPV testing is clearly still in flux.

Joanna M. Cain
Department of Obstetrics and Gynecology
College of Medicine
Penn State University
Hershey, PA 17033, USA
E-mail: jcain{at}psu.edu
Mary K. Howett
Department of Microbiology and Immunology
College of Medicine
Penn State University

REFERENCES AND NOTES

1. W. Eppel, et al., Obstet. Gynecol. 96, 337 (2000) [CrossRef] [Web of Science] [Medline] .
2. D. Bishai, H. Kashima, K. Shah, Arch. Otolaryngol. Head Neck Surg. 126, 935 (2000) .
3. P. Tenti, et al., Obstet. Gynecol. 93, 475 (1999) [CrossRef] [Web of Science] [Medline] .
4. K. K. Thomas, et al., J. Infect. Dis. 182, 1097 (2000) [CrossRef] [Web of Science] [Medline] .
5. H. Richardson, et al., Sex. Transm. Dis. 27, 79 (2000) [Web of Science] [Medline] .
6. R. A. Smith, C. J. Mettlin, K. J. Davis, H. Eyre, Cancer J. Clin. 50, 34 (2000) [Abstract].
7. C. Clavel, et al., Br. J. Cancer 89, 1306 (1999) .
8. M. Schiffman, et al., JAMA 283, 87 (2000) .
9. T. C. Wright Jr., L. Denny, L. Kuhn, A. Pollack, A. Lorincz, JAMA 283, 81 (2000) .
10. R. Sankaranarayanan, P. Pisani, in New Developments in Cervical Cancer Screening and Prevention, E. Franco, J. Monsonego, Eds. (Blackwell, Oxford, 1997), pp. 70-83.
11. E. McGoogan, in New Developments in Cervical Cancer Screening and Prevention, E. Franco, J. Monsonego, Eds. (Blackwell, Oxford, 1997), pp. 317-322.
12. R. Bailie, P. Barron, G. Learmonth, S. Afr. Med. J. 85, 30 (1995) [Web of Science] [Medline] .
13. K. Shy, J. Chu, M. Mandelson, B. Greer, D. Figge, Obstet. Gynecol. 74, 838 (1989) [Web of Science] [Medline] .
14. L. Kuhn, et al., J. Natl. Cancer. Inst. 92, 818 (2000) .
15. Kuhn et al. (14) noted that the specificity of Paps is statistically significant (6.8% versus 87.8%).