This issue of Cancer Cytopathology contains a collaborative study by authors Chute, Lim, and Kong analyzing Stanford University data documenting the performance of the FocalPoint Slide Profiler in Papanicolaou (Pap) test screening for atypical glandular cells.1 As noted by the authors and others, Pap testing in general has shown significantly less effectiveness in screening for and preventing cervical or endometrial glandular neoplasia as compared with screening for and preventing cervical squamous neoplasia.2-4 In fact, it was recently reported that as the rate of decline for cervical squamous carcinoma has accelerated in the United States since 1997, coincident with the more widespread use of liquid-based cytology, the incidence rates of cervical adenocarcinoma have continued to increase, with the greatest rate of increase in younger white women.5
Limited data have been published on the efficacy of increasingly used computer-assisted screening devices in possibly improving Pap test screening for atypical glandular cells, endocervical adenocarcinoma in situ (AIS), cervical adenocarcinoma, or endometrial adenocarcinoma. The Stanford data, based on a test set of 66 SurePath Pap tests interpreted as atypical glandular cells with available follow-up histopathology, underscores both the value and challenges of computer-assisted imaging in screening for atypical glandular cells of cervical or endometrial origin. None of 27 review-qualified atypical glandular cells cases with abnormal (cervical intraepithelial neoplasia [CIN] 2 of 3/AIS/cancer) histopathologic follow-up was scored by the FocalPoint Slide Profiler as requiring no further review; however, these cases were not preferentially ranked in the highest quintile. Accordingly, the authors conclude that FocalPoint Slide Profiler-screened slides need to be carefully and fully manually rescreened for glandular abnormalities, regardless of quintile ranking.
The authors note in their discussion that the FocalPoint Slide Profiler was primarily optimized for the detection of squamous cell abnormalities, and that the initial US Food and Drug Administration (FDA) clinical validation study included only 8 atypical glandular cells cases with no clinical follow-up available on the single atypical glandular cells case ranked as No Further Review.6 The authors also reference a 2002 report from Cedars-Sinai Medical Center in Los Angeles that studied performance of the AutoPap system (now the FocalPoint Slide Profiler) on 14 atypical glandular cells–endometrial conventional Pap smear cases.7 One atypical glandular cells–endometrial smear without available histopathology follow-up was classified as No Further Review. A Massachusetts General Hospital FocalPoint imaging study that included 32 seeded atypical glandular cells SurePath Pap tests and 5 seeded AIS SurePath Pap tests is also discussed.8 Of the 37 seeded examples of glandular abnormalities, only 1 atypical glandular cells SurePath Pap test case was ranked by the FocalPoint Slide Profiler as No Further Review, but no histopathological follow-up was available. In another study at the University of Utah, investigators concluded that manual screening of slides designated by FocalPoint Slide Profiler as No Further Review and lacking endocervical cells appears to be necessary.9 Clearly, additional investigation and documentation of FocalPoint field performance in detection of histopathologically confirmed cases of glandular neoplasia will be useful. Also, because most of the atypical glandular cells cases in the current study and in the Cedars-Sinai study reflected endometrial rather than endocervical glandular abnormalities, additional studies on FocalPoint Slide Profiler performance in screening for cervical glandular neoplasia would be especially useful. Data on detection of glandular abnormalities using the location-guided BD FocalPoint GS Imaging System, relatively recently approved by the FDA on December 3, 2008, are limited at this time to data in the product insert and in recently published results from the product clinical trials.10 With the FocalPoint GS Imaging System, there are no No Further Review low scoring slides that are not manually reviewed. FocalPoint imaging is reportedly now being used in 47% to 50% of the almost 30% of the liquid-based cytology Pap test market estimated by the manufacturer for the SurePath Pap test and in 7% of the estimated 5% to 8% conventional Pap smear market.
Data on screening for glandular abnormalities have also been reported, reflecting use of the other FDA-approved computer-assisted Pap screening system, the ThinPrep Imaging System. ThinPrep Imaging System imaging is now reportedly being used in 65% to 70% of the roughly 70% of the US Pap market share estimated by the manufacturer for the ThinPrep Pap test. The authors of the current Cancer Cytopathology article discuss some of the available data on ThinPrep Imaging System-assisted detection of glandular abnormalities. As with the FocalPoint Slide Profiler, limited data on computer-assisted detection of glandular abnormalities were published in FDA premarket approval data.11 Two of 4 trial cases determined to be atypical glandular cells of undetermined significance by adjudication were classified as negative in the unadjudicated imager review arm of the trial; however, all 10 cases of glandular cancer and the 1 case of AIS in the restricted analytic cancer study portion of the premarket approval report were identified as abnormal by both cytotechnologists and pathologists examining the 22 ThinPrep Imaging System-selected fields of view. A larger independent study conducted by the Screening and Test Evaluation Program of the School of Public Health of the University of Sydney in New South Wales, Australia included 3 cases of histopathologically confirmed AIS and 3 cases of histopathologically confirmed adenocarcinoma, all of which were identified as abnormal in ThinPrep Imaging System-assisted reviews.12 The Cancer Cytopathology authors also reference a New York Memorial Sloan-Kettering Cancer Center study of ThinPrep Imaging System performance on 70 cases of histologically confirmed adenocarcinoma with abnormal glandular cells on a preceding ThinPrep Pap test.13 ThinPrep Imaging System-assisted reviews detected abnormal glandular cells in 68 of 70 cases, including all 11 cases of endocervical adenocarcinoma, but the ThinPrep Imaging System-selected 22 fields of view did not select fields with abnormal glandular cells in 2 cases of endometrial adenocarcinoma. One other study on ThinPrep Imaging System-assisted performance in detection of atypical glandular cells was published after submission of the current Cancer Cytopathology article. A University of Kentucky study of 39 cases reported as having glandular abnormalities found the most abnormal cells in the 22 ThinPrep Imaging System fields of view in 32 of the 39 cases.14 Apparently, however, other cells presented in the 22 fields of review were sufficient to trigger full cytotechnologist manual review and identification of the glandular abnormalities.
A limitation of the above studies and of most Pap test screening sensitivity studies in general lies in the issue of verification bias.15 As the selective process of performing diagnostic histopathologic testing primarily on the subset of patient cases with abnormal screening test results introduces verification bias by excluding or under-representing patients with negative screening test results, it is important to try to estimate the prevalence of underlying significant histopathologically verifiable disease in patients with negative screening test results. In prospective clinical trials, this has generally been accomplished by obtaining biopsies of a subset of patients with negative screening test results and extrapolating the histopathologic findings to estimate true disease prevalence in the screened population. By using this method, European and North American screening trials have estimated verification bias-adjusted conventional Pap smear screening sensitivity for detection of histopathological CIN2+ to vary widely between as low as 19% and as high as 77%, with a mean of around 53%.16 The same trials showed human papillomavirus (HPV) DNA screening sensitivity for CIN2+ to vary from a low of 85% to a high of 100%, with a mean of 96%. Because selection bias will not take place unless case selection is based on a variable (Pap screening results) that has a strong causal effect of the test variable of interest (CIN2+ histopathological findings), another reported approach takes advantage of the finding that hysterectomy findings provide a subset of patients with tissue findings largely independent of Pap test-driven verification bias.17 By using this approach, Zhao et al at the University of Pittsburgh recently estimated verification bias-adjusted sensitivity at 93% for a cohort of >400,000 women primarily (98%) screened with the ThinPrep Imaging System. In the same study, the authors reported histopathological follow-up findings in 662 patients with ThinPrep Imaging System-imaged Pap test results, the largest atypical glandular cells follow-up study published to date. This publication and related studies18 support conjecture that computer-assisted screening has the potential to significantly enhance detection of significant precancerous (CIN2 of 3/AIS) and malignant cervical disease, including detection of glandular neoplasms. Such favorable cytology screening results are also consistent with recent dramatic findings in the prospective multicenter UK ARTISTIC Trial, where conservative manual workload-based Pap screening (32 per day maximum in a 4-hour allotted daily screening time) over 2 screening rounds with liquid-based cytology (ThinPrep Pap test) achieved sensitivity in detection of histopathologically documented CIN2 of 3+ disease equivalent to that achieved with HPV testing.19 This is the first prospective European clinical trial to compare HPV screening results with liquid-based cytology. No ongoing or previous trials have yet assessed the impact of computer-assisted cytology screening in comparison to HPV screening.
Whether or not these promising developments in cytology will translate in the future into long hoped for population data showing decreased incidence and down-staging of cervical glandular neoplasia remains to be documented in future population studies. It has been noted that current recommendations for cervical screening intervals have so far been based almost exclusively on data reflecting the effectiveness of Pap screening in preventing cervical squamous carcinoma incidence and mortality.20 As data emerge on the ability of enhanced cytologic methodologies to favorably impact detection of atypical glandular cells and glandular neoplasia, the increase in and natural history of cervical glandular neoplasia and the impact of Pap screening on incidence, morbidity, and mortality because of glandular neoplasia need to be reflected in proposed and evolving screening algorithms.