OUP user menu

Toward Objective Cervical Cancer Screening
Maybe the Eyes Do Have It

Mark H. Stoler MD
DOI: http://dx.doi.org/10.1309/AJCP6OORTVEZRK7Y 5-6 First published online: 1 July 2010

Several years ago, I wrote an editorial for the Journal entitled, “Toward Objective Quality Assurance: The Eyes Don’t Have It.”1 The point of that commentary was to clarify some of the confusion over the relative utility of visual rescreening methods for quality control in the cytopathology laboratory that are often highly subjective and limited by issues of selection bias. In contrast, what was advocated in that editorial was the use of human papillomavirus (HPV) testing as an independent and objective quality assurance metric for assessing diagnostic interpretations. In the last decade, HPV testing has become the standard care for clinical triage of equivocally abnormal Papanicolaou (Pap) smears. Implicit in its use is the fact that in any population, HPV testing has a significantly higher sensitivity than a single Pap smear for the detection of prevalent cervical intraepithelial neoplasia (CIN) grade 3. Multiple retrospective or cross-sectional studies, authoritative meta-analyses, and, now, randomized control trials have confirmed this fact.2 However, one of the main criticisms of HPV testing is that despite its CIN 3 sensitivity, which drives the entire clinical decision-making process, HPV tests are often criticized for somewhat lower specificity when compared with cytology. Of course it is almost impossible for a test to have terrific sensitivity and spectacular specificity at the same time. Thus, while a single high-risk HPV test will identify approximately 95% of prevalent CIN 3, it still refers at least twice as many people to colposcopy as may truly have clinically significant disease detectable within any reasonable period of follow-up.

The reasons for the superior sensitivity of HPV testing have been widely speculated. The arguments usually include the fact that a higher proportion of cells are potentially sampled by the molecular methods compared with the cells on the Pap slide. The morphologic correlate of such a statement is that the cells responsible for the positive molecular result are simply not on the slide. If the cells are truly not on the slide, then a good fraction of the gynecologic cytology samples that are falsely negative may simply be due to problems with sampling. But if the total explanation for false-negatives rested simply with the cells just not being on the slide, then cytotechnologists and cytopathologists would naturally be off the hook for the consequences of false-negative Pap smears. Unfortunately, a growing backlog of medicolegal experience suggests that this is not the case. Logic then dictates that for a proportion of missed cases, the cells are on the slide and are simply misinterpreted or not identified in the screening process. It is with this thought in mind that I highly recommend to the readers of the Journal the article by Denton and colleagues3 in this issue. This is a large, well-designed, retrospective study of the utility of p16INK4a immunocytochemical staining of cytologic samples compared with HPV testing for the detection of high-grade cervical disease in the triage of atypical squamous cells of undetermined significance (ASC-US) and low-grade squamous intraepithelial lesion (LSIL) Pap cytology results. The study was designed and commissioned in collaboration with the European CINtec Cytology Study Group, which is, of course, supported by the manufacturer of CINtec p16 immunocytochemical staining products. Yes, this is a manufacturer-supported study presented in an academic and respected journal. Furthermore, I openly declare that I have periodically been a consultant for the manufacturer, mtm laboratories. Yet, despite any perception of conflict of interest, I believe the science speaks for itself, which is, of course, why it is being presented, after thorough peer review, in the Journal.

The starting point for the analysis was a statistical power calculation on the number of high-grade CIN lesions necessary to come to a meaningful conclusion, and the investigators used as their end point independently adjudicated, biopsy-proven disease, a standard now used by virtually all active HPV clinical trials. These elements make the analysis somewhat of a model for future studies attempting to demonstrate in a preliminary manner the performance of any potential biomarker, be it in the HPV space or elsewhere. The authors also went to great pains to balance their arguments and point out the advantages and limitations of the study. Briefly, some of the limitations are that this was a retrospective analysis; the analysis was limited to specimens that had adequate follow-up and adequate materials in the liquid-based cytology vials and in the pathology blocks to do the required analyses (a selection bias); and, finally, a further bias that should have worked against the study, that many of the vials were more than 1 year and some even up to 3 years old, yet still the residual material was adequate for both immunomorphologic and molecular analysis.

Despite these limitations and biases, the punch line is clear: p16 immunocytochemical analysis as described in this study is just as sensitive as high-risk HPV testing for the identification of biopsy-proven high-grade CIN with approximately twice the specificity. This is a remarkable result because it implies that potentially, if demonstrated by a proper prospectively designed clinical trial that addresses some of the limitations, p16 triage of ASC-US and, maybe even more importantly, LSIL cytologic results would find just as many CIN 3 cases as the current clinical practice under American Society for Colposcopy and Cervical Pathology guidelines while referring approximately 25% to 50% fewer patients to colposcopy.4

This is a remarkable result, and the data presented in this article reinforce some of the concepts alluded to in my original editorial and clarify some of the hypotheses that have been long debated about which factors impact the performance of cervical cytology. First and foremost, the way the analysis was done clearly shows that pathologists need their cytotechnologists. Pathologists are not really very good at screening, especially at finding rare events on slides, apparently even when they are colorimetrically highlighted. Not only were the cytotechnologists significantly more sensitive in identifying p16+ cells and, therefore, properly classifying cases as positive or negative, but pathologists essentially always agreed with the cytotechnologists when these cases were brought to adjudication. Even more important, this analysis clearly demonstrates the cells of interest are, almost always, on the slide. They may be rare or they may be buried under inflammation or blood or in thick groups, but a properly designed immunocytochemical stain can sufficiently highlight and bring out these cells such that a well-trained cytotechnologist can find them with a much higher degree of accuracy in terms of predicting outcome than a routinely stained and screened Pap smear from the same vial, even if the vial is years old. If these data are substantiated in further trial work, then perhaps, now 8 years later, the pendulum may be swinging back toward morphologic studies, at least perhaps toward molecularly enhanced cytology.

Yet, as noted in the article by Denton and colleagues,3 a cost-effectiveness analysis has not been done based on these data. Such analyses will, by necessity, weigh the cost impact of having a skilled professional morphologically interpret a slide vs the potential medical benefit those skills and costs contribute to the process, all of which will factor heavily into the ultimate form of screening and triage that is used for women at risk for cervical cancer. Historically, the interpretation of sophisticated immunohistochemical stains or in situ hybridizations is associated with significant clinical laboratory charges, and rightly so when they add justifiable medical benefit. However, the current tension that exists between proponents of traditional morphologic cervical screening as opposed to proponents of molecular cervical screening is rooted exactly in this argument between the benefits of morphologic studies vs the cost and risks of keeping or adding humans into the equation compared with a potentially automatable system that takes the subjectivity and perhaps some of the cost out of that process, especially when the process is high volume. Indeed, it is the volume aspect of cervical cancer screening with the potential to capture a slice of a large reimbursement pie that has no doubt driven the ongoing technical innovation in the field of cervical screening. Ultimately, whether “the eyes have it or not” still remains to be seen, but clearly based on the article by Denton et al3 and the potential of upcoming clinical trials, the eyes are at least back in the game.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
View Abstract