Cervical cytology screening has led to a reduction in cancer mortality

Transcription

1 CANCER CYTOPATHOLOGY 105 ThinPrep Pap Test Performance and Biopsy Follow-Up in a University Hospital A. Betts Carpenter, M.D., Ph.D. Diane D. Davey, M.D. Department of Pathology and Laboratory Medicine, University of Kentucky Chandler Medical Center, Lexington, Kentucky. Cytyc Corporation provided free training and a portion of supplies were provided at decreased cost, an offer made available to all laboratories in late The authors thank Jacqueline Emery, M.D., and Helen Banks, CT (ASCP), for assistance with implementing the ThinPrep technique. Dr. Carpenter s current affiliation: Department of Pathology, King s Daughters Medical Center, Ashland, Kentucky. Address for reprints: Diane D. Davey, M.D., Department of Pathology and Laboratory Medicine, MS 117, University of Kentucky Medical Center, Lexington, KY Received November 3, 1998; revision received February 12, 1999; accepted February 24, BACKGROUND. The ThinPrep Pap Test (TP), a liquid-based cervical cytology preparation, was approved for use in the U.S. in The purpose of this study was to compare TP performance and biopsy follow-up studies with a similar population of high risk patients sampled by conventional Papanicolaou (Pap) smear (CS). METHODS. Diagnostic and specimen adequacy interpretations for 2727 TP directto-vial Pap tests from a high risk university hospital practice were compared with 5000 CS preparations from the same physicians taken 1 year previously. Biopsy follow-up studies for the categories of squamous intraepithelial lesion (SIL), carcinoma, and atypical squamous cells of undetermined significance (ASCUS) for each time period and technique were contrasted. RESULTS. The SIL/carcinoma detection rate increased from 7.7% to 10.5% (P 0.01) and the ASCUS rate decreased from 12.5% to 6.9% (P 0.01); the percentage of satisfactory but limited specimens decreased from 19.4% to 10.5% (P 0.01). Low grade SIL cases increased by 57% (P 0.01) whereas the 26% increase in high grade SIL cases was not statistically significant. Greater than 90% of ungraded SIL, high grade SIL, and carcinoma cases had abnormal biopsies by both the TP and CS methods. The number of biopsy-confirmed high grade dysplasias and carcinomas was similar in the two groups. A low grade SIL detected by TP was less likely to have an abnormal biopsy (70% vs. 85% for CS). Nevertheless, the 57% increase in low grade SIL diagnoses by TP resulted in more TP patients with dysplastic biopsy diagnoses. Follow-up studies for ASCUS cases diagnosed by either TP or CS were similar, and 21 24% of patients eventually were found to have dysplasia. CONCLUSIONS. The TP technique appears to lead to the increased detection of low grade SIL lesions, decreased satisfactory but limited samples, and fewer equivocal specimens. No increase in biopsy-confirmed high grade dysplasias and carcinomas was found. Follow-up studies for the ASCUS category were nearly identical to those for CS. Cancer (Cancer Cytopathol) 1999;87: American Cancer Society. KEYWORDS: ThinPrep, cervical cytology, smears, cervix, carcinoma, Papanicolaou, screening, dysplasia. Cervical cytology screening has led to a reduction in cancer mortality in the U.S. and several other countries, but traditional screening methods are associated with occasional false-negative results. Several innovative technologies recently have been introduced in an attempt to decrease false-negative results due to sampling, screening, and interpretive problems. The ThinPrep Pap Test (TP) (Cytyc Corporation, Boxborough, MA) is a liquid-based cell collection method that was approved by the U. S. Food and Drug Administration (FDA) as a replacement for the conventional cervical smear in The clinical trial leading to FDA approval involved samples split for 1999 American Cancer Society

2 106 CANCER (CANCER CYTOPATHOLOGY) June 25, 1999 / Volume 87 / Number 3 conventional Papanicolaou (Pap) smear (CS) and TP in six laboratories, and an expert pathologist served as the final arbitrator of discordant results. 1 A statistically significant increase in the detection rate of squamous intraepithelial lesions (SIL) was noted in the TP arm, with the greatest increase in the screening centers. The three hospital laboratories with a high abnormality rate exhibited only a slightly increased detection rate (6%). Biopsy follow-up studies were not reported for this clinical trial. Specimen adequacy was improved, with an 11% increase in satisfactory samples. Since FDA approval, TP has been used as a directto-vial Pap test replacement in a variety of laboratory settings. To our knowledge only a few published articles have detailed biopsy follow-up results, 2-4 and very little has been reported regarding follow-up of atypical squamous cells of undetermined significance (ASCUS) diagnoses using the TP technique. The purpose of this study was to compare TP results and follow-up data with those of CS specimens from the same physician group obtained 1 year previously in a university hospital setting with a high rate of abnormalities. METHODS The University of Kentucky Medical Center cytology laboratory introduced the TP technique in late After a training program at Cytyc Corporation, the method was introduced to a few gynecologists with high risk patient populations. Split samples initially were performed, and physicians switched to the direct-to-vial TP procedure when both the laboratory and clinicians were comfortable with procedures and assessments. By February 1997, the majority of the gynecology division faculty was performing TP Pap tests on the majority of their patients. Other selected clinicians from the student health service and family practice department began to use TP a few months later. Patients were not charged out-of-pocket for TP during this study so this is a relatively unbiased population sample from these clinics. The study population included all direct-to-vial TP specimens (total of 2727) performed by these clinicians between February 1 and July 31, Some physicians performed nearly all TP on their patients over this entire 6-month block, whereas others used TP for fewer months. All TPs diagnosed as a SIL or carcinoma during this 6-month period were identified by the pathology computer system. Follow-up biopsy specimens obtained 9 12 months after the index TP were identified by computer records. If patients had multiple biopsies, the most abnormal biopsy diagnosis was used for analysis. For comparison, CS performed by the same physicians 1 year previously were analyzed. Diagnostic and specimen adequacy results were compared for the first 5000 CS performed during this time period. Any differences in proportions were analyzed statistically using a chi-square distribution. Biopsy follow-up studies regarding CS diagnosed as SIL or carcinoma were followed similarly for 12 months. Because this group of physicians performed more CS over the same period in 1996, only 4 months of CS were investigated to obtain an equivalent number. The corresponding total number of CS for these 4 months was The number of patients sampled from each physician or clinic was similar proportionately for CS in 1996 and TP in The University of Kentucky uses the Bethesda system nomenclature and criteria, with some modifications. 5 Cases termed as mild-moderate dysplasia on cytology exhibited predominately low grade SIL (LSIL), with occasional cells suggestive of moderate dysplasia. Ungraded SIL cases had features intermediate between LSIL and high grade SIL (HSIL), or had relatively few or poorly preserved SIL cells. The suspicious category was used most often for cases suspicious for adenocarcinoma. The surgical pathology laboratory generally reports cervical biopsies using dysplasia terminology: moderate dysplasia, severe dysplasia, and carcinoma in situ (CIS). For data analysis purposes, these terms all were combined and referred to as high grade dysplasia. For the assessment of adequacy, the Bethesda system criteria were used for CS. Criteria for judging adequacy for TP were identical for transformation zone component and obscuring factors. The criteria for a scantly cellular TP were modified; if 40% of fields showed significant squamous material, the specimen was judged unsatisfactory. It should be noted that this was a visual estimate and cell counts were not performed. For analysis of ASCUS cases, 100 patients with ASCUS on TP during the 1997 time period were compared with 100 CS patients from the same physician group 1 year previously. Computer files were searched for follow-up cytology and/or biopsies. The follow-up time was 1 year, and the most abnormal biopsy or cytology follow-up was assumed to represent the correct diagnosis. The diagnosis of atypical glandular cells of undetermined significance (AGUS) is used relatively infrequently at the University of Kentucky. Its use has decreased with improved cytologic criteria for benign glandular entities and also with the recognition that many of these cases represent squamous lesions. More frequently, these processes are described as atypical metaplastic/endocervical cells (or epithelial groups). The small number of AGUS cases are in-

3 ThinPrep Performance and Follow-Up/Carpenter and Davey 107 TABLE 1 Cytologic Diagnoses for ThinPrep Pap Test Cohort versus Conventional Papanicolaou Smear Cohort ThinPrep Conventional Change Year Satisfactory 89.2% 80% 11.5% Unsatisfactory 0.3% 0.6% P 0.05 Limited (total) 10.5% 19.4% 46% (P 0.01) No transformation zone 9.8% 5.5% Obscuring, other 0.7% 13.9% ASCUS a 6.9% 12.5% 45% (P 0.01) Low grade SIL 6.9% 4.4% 57% (P 0.01) Ungraded SIL 1.0% 1.1% High grade SIL 2.4% 1.9% 26% (P 0.10) Carcinoma/suspicious 0.2% 0.3% Total SIL 10.5% 7.7% 36% (P 0.01) ASCUS/SIL Total number FIGURE 1. ThinPrep Papanicolaou test showing a low grade squamous intraepithelial lesion (LSIL) from a 16-year-old female. A previous cytology result also was found to be LSIL, but the follow-up biopsy was negative (Papanicolaou stain, 132). ASCUS: atypical squamous cells of undetermined significance; SIL: Squamous intraepithelial lesion; SIL : all SIL, carcinoma, and suspicious cases. a Figure also includes a small number of atypical glandular cells of undetermined significance cases. cluded with the overall ASCUS statistics in Table 1. Cases designated purely as AGUS or suspicious for adenocarcinoma from the same time periods then were separated for follow-up analysis. Finally, surgical pathology records were searched for endocervical adenocarcinoma specimens. RESULTS Cytologic diagnoses and adequacy determinations for 2727 TP performed between February and July 1997 were compared with results reported for 5000 CS obtained by the same physician group in 1996 (Table 1). The rate of completely satisfactory specimens increased from 80% to 89.2% with the implementation of TP, an 11.5% improvement. The 46% decrease in limited specimens was statistically significant, and entirely due to a marked decrease in partially obscured specimens. Indeed, the limited rate due to no transformation zone component was higher for TP than CS (5.5% vs. 9.8%). The unsatisfactory rate did not change significantly with TP use, but the number of unsatisfactory cases was 1% for both methods. The detection rate for SIL and more serious lesions (SIL ) increased from 7.7% to 10.5% with TP implementation. The majority of this change was due to a 57% increase in LSIL detection (P 0.01); this category also included cases described as mild-moderate dysplasia in the cytology report (Fig. 1). The HSIL detection rate increased 26% from 1.9% to 2.4%, but this change was not statistically significant (Fig. 2). FIGURE 2. ThinPrep Papanicolaou test showing a high grade squamous intraepithelial lesion from an 18-year-old female. The follow-up biopsy revealed severe dysplasia (Papanicolaou stain, 132). There was no significant change in ungraded SIL, carcinoma, or suspicious cases. If the latter cases were combined with the HSIL category, the improved detection rate attributable to TP was 10%. The percentage of ASCUS cases decreased by 45% from an initial rate of 12.5% to 6.9% (P 0.01), and the ASCUS to SIL ratio decreased from 1.62 to Of the 286 SIL TP specimens reported, follow-up information was available for 206 specimens (72%). Biopsies were obtained on 156 specimens (55%) (Table 2). An additional 50 patients underwent only repeat cytology during the time period. The majority were patients with initial LSIL on TP; 31 had negative repeat cytology and 19 had abnormal cytology. Of patients with LSIL on TP followed by biopsy, 21 of 71 (30%) had negative biopsies, 43 of 71 (61%) had mild

4 108 CANCER (CANCER CYTOPATHOLOGY) June 25, 1999 / Volume 87 / Number 3 TABLE 2 ThinPrep Pap Test Biopsy Follow-Up Studies Cytologic diagnosis Biopsy follow-up Low grade SIL Mild-moderate dysplasia High grade SIL Ungraded SIL Carcinoma Negative Mild dysplasia a High grade dysplasia b Dysplasia, not graded Carcinoma Totals SIL: squamous intraepithelial lesion. a Includes human papillomavirus/condylomatous change. b Includes moderate/severe dysplasia and carcinoma in situ. TABLE 3 Conventional Smear Biopsy Follow-Up Studies Cytologic diagnosis Biopsy follow-up Low grade SIL Mild-moderate Dysplasia High grade SIL Ungraded SIL Carcinoma Negative Mild dysplasia a High grade dysplasia b Dysplasia, not graded Carcinoma Totals SIL: squamous intraepithelial lesion. a Includes human papillomavirus/condylomatous change. b Includes moderate/severe dysplasia and carcinoma in situ. dysplasia, and the remainder had high grade or ungraded dysplasia. Those patients with mild-moderate dysplasia on TP most often had mild dysplasia (11 of 21; 52%), but 9 of 21 (43%) either had ungraded or high grade dysplasia on biopsy. Patients with HSIL most often had high grade dysplasia on biopsy (35 of 51; 69%), with only 1 patient having negative biopsy follow-up. Follow-up data were available for 209 patients with CS obtained during the 4-month period in 1996 (total CS of 4660). Repeat cytology was the only study performed on 27 patients; 16 were negative and 11 were abnormal. The biopsy data from 182 women are illustrated in Table 3. Of the 59 patients with LSIL on CS who had biopsies, 9 (15%) had negative biopsies, 38 (64%) had mild dysplasia, and the remainder had high grade or unspecified dysplasia. Patients with mild-moderate dysplasia were nearly equally likely to have mild versus high grade or ungraded dysplasia on biopsy. Patients with HSIL on CS predominately had high grade dysplasia on biopsy (63 of 78; 81%), with only 1 patient showing negative histology. Relatively few cases of carcinoma were observed on either TP or CS, and the majority revealed squamous or adenocarcinoma on biopsy. One CS case revealed marked radiation change on biopsy and most likely represented a false-positive cytology result. The abnormal biopsy rate after various cytologic diagnoses are compared for TP and CS in Table 4. Abnormal biopsy rates were 90% after either TP or CS diagnoses of HSIL or ungraded SIL. The positive predictive value after a CS diagnosis of LSIL was higher than for a TP LSIL diagnosis (85% vs. 70%). The total number of biopsy-proven dysplasias and carcinomas after a SIL cytology diagnosis was 130 (4.8%) in the TP cohort and 170 (3.6%) in the CS cohort. The TP cohort had a significantly higher yield of abnormal biopsies (P 0.02), despite the lower predictive value of LSIL cytology diagnoses on TP. This most likely was due to the 36% increase in SIL cy-

5 ThinPrep Performance and Follow-Up/Carpenter and Davey 109 TABLE 4 Predictive Value: Percent with Abnormal Biopsy for ThinPrep Pap Test versus Conventional Papanicolaou Smear Cytologic diagnosis ThinPrep Conventional Low grade SIL 70% 85% Ungraded SIL a 94% 97% High grade SIL/carcinoma 98% 98% All SIL/carcinoma 85% 93% SIL: squamous intraepithelial lesion. a Includes mild-moderate dysplasia. TABLE 5 ASCUS Follow-Up Studies ThinPrep Negative biopsy 9% 10% Mild dysplasia, biopsy 10% 17% High grade dysplasia, biopsy 4% 3% Dysplasia, not graded, biopsy 0% 1% SIL on cytology 7% 3% ASCUS on cytology 9% 10% Negative cytology 39% 31% No follow-up 22% 25% Total negative follow-up 48% 41% Total SIL/dysplasia 21% 24% Conventional ASCUS: atypical squamous cells of undetermined significance; ThinPrep: ThinPrep Pap Test ; Conventional: conventional Papanicolaou smear; SIL: squamous intraepithelial lesion. tology diagnoses, with more patients biopsied for a SIL indication. The increase in abnormal biopsies was accounted for entirely by diagnoses of mild dysplasia in the TP cohort. The total yield of high grade dysplasia/carcinoma biopsies after a SIL diagnosis appeared slightly higher for the CS group compared with the TP group (96 [2.1%] vs. 50 [1.8%]); the difference was not statistically significant. This analysis provides another way of viewing the data, but has several limitations. Not all patients with SIL cytology underwent biopsies at the University of Kentucky during the study, and dysplastic biopsies after an ASCUS diagnosis were not considered in the calculation. Biopsy and cytology follow-up studies for ASCUS cases diagnosed by TP and CS are compared in Table 5. Greater than 70% of ASCUS cases diagnosed by either method underwent follow-up studies, and slightly more patients with ASCUS on CS were biopsied. When both cytology and biopsy diagnoses were considered, slightly more CS ASCUS patients had SIL/ dysplasia on follow-up (24%) than TP patients (21%). The majority of dysplasias on follow-up were low grade, with 5% high grade processes detected. Slightly 50% of the patients with an ASCUS diagnosis had a history of a prior abnormality, but the follow-up results were similar for both groups. Five cases (0.2%) of pure AGUS or suspicious glandular lesions were noted on TP analysis. Two were from the same patient diagnosed with adenocarcinoma in situ and squamous CIS on cone biopsy. Two patients had endometrial carcinoma and one had negative follow-up. Seventeen cases (0.3%) of pure AGUS or suspicious glandular lesions were noted on conventional cytology during the study period. Four patients had endometrial carcinoma, four had mild dysplasia on biopsy, eight had negative follow-up, and one patient was lost to follow-up. Except for the first case described, all cases of endocervical adenocarcinoma seen in surgical pathology during this period were from patients without previous or concurrent cervical cytology specimens in our laboratory. DISCUSSION Evaluation of new cytology technologies is plagued by difficulties in defining the gold standard or truth for a cytologic specimen. Patients with negative cytology results by both conventional and new methods generally are not followed long enough to determine whether they have a false-negative result. Cytologic interpretation is subject to interobserver variation, especially for low grade and atypical processes. Biopsy follow-up studies may be limited by colposcopic sampling problems and regression or progression of lesions prior to biopsy. 6 The Intersociety Working Group for Cytology Technologies has recommended panel adjudication of discordant cytology results for automated screening instruments. 7 The TP clinical trials utilized an expert pathologist to define the truth for discordant split sample pairs. 1,8 Human papillomavirus studies also have been correlated with cytology results. 8 Once a laboratory has converted to direct-to-vial liquid cervical cytology, methods other than split sample comparison must be used to compare their performance with that of CS. Our approach was to compare both TP cytologic results and biopsy follow-up with a cohort of patients with CS from the same physician and clinic group. Comparison of cytologic results alone could be subject to differences in the cytologic interpretation of TP and CS preparations. Some users have noted slight differences in cytologic presentation on TP, 9 and there also may be a subconscious tendency to diagnose preparations from a new technology more definitively. Biopsy follow-up likely is most useful for the subset of patients with HSIL and carcinoma. Such lesions are less likely to regress, and previous cytohistologic correlation studies have confirmed a high positive predictive value for CS

6 110 CANCER (CANCER CYTOPATHOLOGY) June 25, 1999 / Volume 87 / Number 3 Although biopsy is not necessarily the gold standard for the subset of patients with LSIL, we assumed that both the TP and CS methods would be affected similarly by biopsy sampling problems and clinical regression. Several previous studies utilizing TP have found an increased detection rate for SIL compared with CS The clinical trial leading to FDA approval was a split sample study conducted in three screening centers and three hospitals. The most impressive increase in detected precursor lesions was noted in the screening centers characterized by low abnormal rates (LSIL rate of 4%). 1 The hospital centers, characterized by high risk populations (LSIL rate of 9%), showed only a modest 6% gain in detection rates. Direct-to-vial studies from laboratories with average risk populations have confirmed this increased rate of detection. 2,3 Although the majority of improvement is related to increased LSIL detection, some split-sample and direct-to-vial studies also have found higher detection rates for HSIL. 3,15,16 Our study differs from previously reported directto-vial studies in that it analyzes a high risk university hospital laboratory population. The SIL rate in the University of Kentucky Cytology Laboratory was 7.7% prior to the introduction of TP for the group of physician users, and the ASCUS rate was 12.5%. These abnormal rates are 90% of laboratories participating in recent questionnaire surveys conducted by the College of American Pathologists. 17,18 Although the total rate of abnormalities is high at the University of Kentucky, the ASCUS/SIL ratio is below the median national reporting rate in the same survey. 18 Other direct-to-vial studies have been reported from laboratories with higher ASCUS/SIL ratios. 2,3 Finally, to our knowledge, very little information has been reported regarding ASCUS follow-up for the TP technique, with some studies presented in abstract form. 19,20 The 36% increase in SIL detection for the TP cohort in the high risk population served by the University of Kentucky laboratory confirms the significantly improved detection of precursor lesions noted previously in average risk population laboratories. The direct-to-vial method may permit the increased detection of SIL in all types of patient populations. Low grade SIL accounted for the majority of the increase; there was no significant difference in HSIL detection. A larger series of cases or a multiinstitutional trial would be useful to demonstrate the improved detection of both HSIL and carcinoma cases. Reporting rates for ASCUS decreased by 45% in the TP cohort. A decrease in ASCUS rates also was suggested in split sample FDA clinical trials. 1 Other direct-to vial studies have shown both increased and decreased ASCUS rates. 2,3,14 Papillo et al. reported decreased ASCUS rates with TP, but the ASCUS/SIL ratio in that patient population was higher than the national average. 3 Biopsy follow-up studies comparing TP with CS were similar for the HSIL and ungraded SIL categories. The number of carcinoma cases was too small to draw definite conclusions, but appeared equivalent. The positive biopsy rate after LSIL was less than for HSIL, and appeared to be lower for TP than for CS. Large CS series have shown fewer positive biopsies after LSIL than after HSIL. 10,11 Positive biopsy rates after HSIL and LSIL cytology results were 93.5% and 86.5%, respectively, in a large cytohistologic correlation Q-Probe study. 11 Many of the LSIL cases with negative biopsies may represent very small colposcopic lesions or transient regressing lesions. 12 The observation that the TP cases in our series more frequently had negative biopsies could be related to the natural history of LSIL in our patient population. The TP technique may allow cases with few abnormal cells to be presented more clearly and diagnosed definitively. On CS, such cases could be classified as ASCUS or missed altogether. Other investigators have found that LSIL cases with relatively few dysplastic cells have fewer abnormal biopsies. 21 Another possibility is that such TP LSIL cases were overcalled by the cytopathologists. However, these diagnoses were confirmed by other pathologists during quality assurance reviews. The percentage of patients with abnormal biopsies in the TP cohort was higher than for the CS cohort (4.8% vs. 3.6%), despite the lower predictive value of LSIL cytology diagnoses on TP. This finding relates to the even greater number of patients diagnosed with LSIL by TP. The clinical significance of this is uncertain because high grade lesions are most important for treatment purposes and carcinoma prevention. Our study could not demonstrate an increased yield of biopsy-proven high grade or malignant biopsies in the TP group. Other studies incorporating biopsy follow-up information are limited and have been reported mainly in populations with relatively low rates of abnormalities. Papillo et al. reported a higher rate of abnormal biopsies for TP compared with the cohort of CS (81% vs. 72%). Of note, an increase in the HSIL detection rate was noted and 93% of HSIL cytology results on TP with biopsy follow-up were confirmed as high grade dysplasia. 3 Bolick and Hellman had biopsy follow-up information available for fewer specimens, but also reported a higher positive biopsy rate for the TP cohort. 2 In a large split sample study, Roberts et al. reported an improved detection of high grade lesions on TP that were confirmed by biopsy follow-up. 16

7 ThinPrep Performance and Follow-Up/Carpenter and Davey 111 Although ASCUS diagnoses decreased with TP, follow-up cytology and biopsy studies showed findings nearly identical to CS. This suggests that the implications of the ASCUS diagnosis on TP, as well as clinical follow-up protocols, should be similar to those reported for CS. 17 One abstract has described improved subclassification of ASCUS diagnoses by use of the TP technique. 20 The University of Kentucky laboratory subclassifies relatively few ASCUS cases, and mainly does so to direct specific patient management. Atypical cases in which HSIL is favored are subclassified, as are cases accompanied by a reactive/infectious process or atrophy. Although the SIL detection rate increased, the decreased ASCUS detection rates resulted in an overall decrease in all types of reported abnormalities for the TP cohort (17.4% ASCUS vs. 20.2% ASCUS for CS). This could be judged in either a positive or negative way. If the goal is to maximize sensitivity for all types of processes, the TP technique could lead to lower detection rates for low grade abnormalities in some laboratory settings. However, follow-up studies for both the SIL and ASCUS categories suggest similar rates of biopsy-proven abnormalities. Certainly, the rate of abnormal biopsies was no higher for an abnormal TP diagnosis. We postulate that the decrease in ASCUS rates is due to shifts both to the SIL and negative reporting categories, presumably because of improved cytologic preservation and presentation. This hypothesis would explain the similar ASCUS follow-up findings despite fewer overall ASCUS cases. The overall decline in atypical rates in a high risk population with coexistent inflammatory processes should not be an unexpected result. If this change is due to fewer atypical processes that actually are inflammatory and not preneoplastic, the decrease in costs and patient anxiety would be beneficial to the cervical carcinoma screening program. Future research using costeffectiveness methods could investigate whether the decrease in ASCUS diagnoses and attendant follow-up costs is sufficient to pay for the increased costs of this methodology. Some cytopathologists have raised concern regarding the detection and diagnosis of glandular lesions on liquid cytology. 22 The AGUS category includes a wide spectrum of benign processes, glandular neoplasia, and squamous lesions. Very few AGUS cases were noted on TP in this study, and the number was slightly lower than noted on CS. However, the majority of CS AGUS cases either were benign or showed SIL on follow-up. In contrast, the majority of TP AGUS cases demonstrated either endometrial or endocervical glandular neoplasia. Comparing overall numbers of cases by the use of a historic control most likely is not valid for the AGUS category because fewer AGUS diagnoses currently are made on any cytology preparations in our laboratory. Many such processes currently are included in the ASCUS category or descriptively signed out as possible HSIL. In the current study, we could find no evidence that significant glandular lesions had been missed on either TP or CS, although a longer time period would be necessary to detect false-negative cytology specimens. Glandular neoplasia detection via liquid cytology should be investigated by larger or multiinstitutional studies. The rate of completely satisfactory specimens increased from 80% to 89.2% with the use of TP. The satisfactory but limited rate was 46% lower, and was caused entirely by a decrease in obscuring factors. To our knowledge the decrease in satisfactory but limited specimens has not led to any demonstrable changes in screening patterns thus far, because the majority of clinicians promote annual screening for their patients. Of note, the satisfactory but limited rate due to no transformation zone component actually increased. The split sample clinical trial also revealed more specimens with no transformation zone component, whereas one direct-to-vial study revealed equivalent detection of a transformation zone component. 23 Our laboratory noted a tendency for cytotechnologists to miss small numbers of endocervical and metaplastic cells on the TP early in implementation, and the majority of the limited samples were rescreened by another cytotechnologist during the period of this study. It is more likely that the actual TP procedure or new sampling method was responsible for the lower transformation zone result. This study was conducted using mostly broom-style sampling devices for TP, whereas clinicians at the University of Kentucky traditionally have used a spatula/endocervical brush for CS. Whatever the reason for the lower detection rate of transformation zone elements, this shift in the type of limited specimen could be a topic for future research. Given the similar yield of HSIL in both groups, it is not evident whether the absence of obscuring factors or the presence of transformation zone elements plays a more important role in assessing adequacy. As with other direct-to-vial studies, this analysis has certain limitations. Patient populations for the 2 years of this study may not have been matched exactly. Although patients from the same clinician group exactly 1 year earlier were selected, it is possible that changes in clinical and laboratory practices occurred. The same group of cytotechnologists screened both cohorts of slides, but one staff pathologist differed for the 2 years of the study. Sampling devices also varied during the course of this retrospective study. To our knowledge the current study is one of the

8 112 CANCER (CANCER CYTOPATHOLOGY) June 25, 1999 / Volume 87 / Number 3 largest studies providing biopsy follow-up of TP in a high risk university hospital setting and the first published study that we are aware to provide follow-up data regarding ASCUS cases diagnosed by TP. An increase in LSIL detection was noted on TP. The improvement in HSIL detection was not statistically significant, and the total yield of biopsy-proven grade dysplasias and carcinomas was unchanged. The positive biopsy rate for a LSIL diagnosis on TP was slightly lower than for CS, but the significant increase in LSIL cytology diagnoses still resulted in a higher percentage of dysplasias in the TP cohort. These observations may be due to a shift in cases from the ASCUS to the LSIL category in the TP cohort, with some representing transient human papillomavirus infections. ASCUS rates decreased with TP implementation, but ASCUS follow-up studies were similar, suggesting that the clinical management of such cases should remain the same. REFERENCES 1. Lee KR, Ashfaq R, Birdsong GG, Corkill ME, McIntosh KM, Inhorn SL. Comparison of conventional Papanicolaou smears and fluid-based, thin-layer system for cervical cancer screening. Obstet Gynecol 1997;90: Bolick DR, Hellman DJ. Laboratory implementation and efficacy assessment of the ThinPrep cervical cancer screening system. Acta Cytol 1998;42: Papillo JL, Zarka MA, St. John PL. Evaluation of the ThinPrep Pap test in clinical practice: a seven month, 16,314-case experience in northern Vermont. Acta Cytol 1998;42: Sheets EE, Constantine NM, Dinisco S, Dean B, Cibas ES. Colposcopically directed biopsies provide a basis for comparing the accuracy of ThinPrep and Papanicolaou smears. J Gynecol Tech 1995;1: National Cancer Institute Workshop. The Bethesda System for reporting cervical/vaginal cytologic diagnosis. Acta Cytol 1993;37: Tritz DM, Weeks JA, Spires SE, Sattich M, Banks H, Cibull ML, et al. Etiologies for non-correlating cervical cytologies and biopsies. Am J Clin Pathol 1995;103: Intersociety Working Group for Cytology Technologies. Proposed guidelines for primary screening instruments for gynecologic cytology. Am J Clin Pathol 1997;109: Sherman ME, Mendoza M, Lee KR, Ashfaq R, Birdsong GG, Corkill ME, et al. Performance of liquid-based, thin-layer cervical cytology: correlation with reference diagnoses and human papillomavirus testing. Mod Pathol 1998;11: Wilbur DC. New technology in gynecologic cytology: automation in cytology screening and thin-layer preparation techniques. In: Bonfiglio TA, Erozan YS, editors. Gynecologic cytopathology. Philadelphia: Lippincott-Raven Publishers, 1997: DiBonito L, Falconieri G, Tomasic G, Colautti I, Bonifacio D, Dudine S. Cervical cytopathology. An evaluation of its accuracy based on cytohistologic comparison. Cancer 1993;72: Jones BA, Novis DA. Cervical biopsy-cytology correlation. A College of American Pathologists Q-probes study of correlations in 348 laboratories. Arch Pathol Lab Med 1996; 120: Richart RM, Wright TC. Controversies in the management of low-grade cervical intraepithelial neoplasia. Cancer 1993;71: Austin RM, Ramzy I. Increased detection of epithelial cell abnormalities by liquid-based gynecologic cytology preparations. A review of accumulated data. Acta Cytol 1998;42: Dupree WB, Suprun HZ, Beckwith DG, Shane JJ, Lucente V. The promise and risk of a new technology. The Lehigh Valley Hospital s experience with liquid-based cervical cytology. Cancer (Cancer) Cytopathol 1998;84: Linder J, Zahniser D. ThinPrep Papanicolaou testing to reduce false negative cervical cytology. Arch Pathol Lab Med 1998;122: Roberts JM, Gurley AM, Thurloe JK, Bowditch R, Laverty CR. Evaluation of the ThinPrep Pap test as an adjunct to the conventional Pap smear. Med J Aust 1997;167: Davey DD, Nielsen ML, Naryshkin S, Robb JA, Cohen T, Kline TS. Atypical squamous cells of undetermined significance. Current laboratory practices of participants in the College of American Pathologists Interlaboratory Comparison Program in Cervicovaginal Cytology. Arch Pathol Lab Med 1996;120: Interlaboratory Comparison Program in Cervicovaginal Cytology PAP supplemental questionnaire, laboratory practice: ASCUS/AGUS. Northfield, IL: College of American Pathologists, Johnson JE, Jones HW, Conrad KA, Huff BC. Increased rate of SIL detection with excellent biopsy correlation after implementation of direct-to-vial ThinPrep liquid-based preparation of cervicovaginal specimens at a University Medical Center. Acta Cytol 1998;5: Papillo JL, St. John TL, Zarka MA. Atypical squamous cells of undetermined significance (ASCUS): widening the differential between reactive and neoplastic processes utilizing the ThinPrep technique. Acta Cytol 1998;5: Hall S, Wu TC, Soudi N, Sherman ME. Low-grade squamous intraepithelial lesions: cytologic predictors of biopsy confirmation. Diagn Cytopathol 1994;10: Austin RM. Implementing liquid-based gynecologic cytology: balancing marketing, financial, and scientific issues. Cancer (Cancer) Cytopathol 1998;84: Corkill M, Knapp D, Martin J, Hutchinson ML. Specimen adequacy of ThinPrep sample preparations in a direct-tovial study. Acta Cytol 1997;41:39 44.