Making the Most of Pap Testing

Following dramatic newspaper and television accounts that clinical laboratories often failed to detect abnormal Papanicolaou (Pap) smears, Congress amended the Clinical Laboratory Improvement Act (CLIA) in 1988. CLIA '88 included provisions designed to reduce the false-negative rate for Pap smears by detecting and correcting deficiencies in laboratory procedures.¹ Most laboratories have complied with the CLIA Pap smear requirements by taking a second look at ("rescreening") 10% of all Pap smears initially classified as "within normal limits."

CLIA-mandated rescreening may have stimulated laboratories to improve the quality and accuracy of testing. But when 90% of the slides read as normal are not rescreened, random rescreening can raise the overall sensitivity of testing by little. Newer approaches to reduce the false-negative rate, such as computer-based technologies to automate or assist in the interpretation of Pap smears, have the potential to do better. Automation might be expected to lower the costs and increase the accuracy of both initial screening and rescreening. Unlike human beings who examine smear after smear in rapid succession, computerized devices are not subject to fatigue or distraction. In this issue of THE JOURNAL, O'Leary and colleagues assess whether these potential advantages of computer-assisted interpretation of Pap smears are real.² They examine PAPNET, a computerized system that is applied to all smears classified on initial manual screening as normal, and that selects the subset with suspicious areas for manual review.

O'Leary and colleagues report that Papnet-assisted rescreening detected more abnormalities than manual reexamination of all normal slides. But the additional abnormalities are few, and nearly all are atypical squamous cells of undetermined significance or atypical glandular cells of undetermined significance, both of which have prognostic importance andmanagement implications that are unclear.³ Although PAPNET reduces the number of slides that must be reexamined manually, the cost is $6 per slide more than manual rescreening, assuming PAPNET processing alone costs $7.50 per slide. According to the authors, at the $40 cost for processing quoted in PAPNET advertisements, PAPNET-assisted rescreening costs $37 more per slide than manual rescreening. Although the authors did not perform a conventional cost-effectiveness analysis that expresses outcomes in life-years or quality-adjusted life-years, the findings of O'Leary and colleagues imply that 100% manual rescreening is a more cost-effective alternative to 10% rescreening than PAPNET-assisted rescreening.

In other settings, PAPNET-assisted rescreening might be more cost-effective. Laboratories with higher false-negative rates or with cytotechnologists who earn more or are less efficient would have more to gain from the technology. PAPNET-assisted rescreening also might have a greater yield in a population with a higher prevalence of cytological abnormalities.

Because this study focuses on a comparison between PAPNET-assisted and 100% manual rescreening, it does not directly address related questions about the best approach to preventing cervical cancer. Do other computer-assisted screening and rescreening aids perform better? How valuable are other methods for improving sensitivity? Inadequate sampling, for example, is an important cause of false-negative results. Better training of physicians and nurses and the use of such devices as endocervical brushes can mitigate sampling problems, and can even be cost-saving.⁴ Monolayer and thin-layer slide preparations also may improve sensitivity by making it easier to view and interpret slides.

Although the typical sensitivity of 80% implies that the Pap smear has a high false-negative rate,⁵ screening women as infrequently as every 3 years reduces the incidence of invasive cervical cancer by more than 90%.⁶ How can such an imperfect test be so effective? Unlike many other cancers, most cervical cancers become invasive only after a very long "preclinical interval" when they are detectable but still curable.⁷ For such a slowly progressive condition, periodic screening presents multiple opportunities to detect a missed case of disease before it becomes incurable.

However, those opportunities are only available to women who undergo screening. Unfortunately, many American women do not have Pap smears regularly, and women at high risk of developing cervical cancer are among those who are least likely to be screened. As O'Leary and colleagues note, failure to be tested is the greatest obstacle to prevention of cervical cancer in the United States. Expense is also an important barrier to compliance with screening programs. Thus, costly efforts to improve sensitivity must be balanced against their effects on participation. With further development, computer-assisted evaluation of Pap smears may become more effective and less costly, thereby improving both test performance and accessibility. But according to O'Leary and colleagues, PAPNET-assisted rescreening today is more expensive than manual rescreening and identifies few unequivocally significant abnormalities. We should ensure that our quest to decrease the false-negative rate does not increase costs enough to jeopardize efforts to persuade more women to take advantage of this simple and highly effective screening test.