Fax: (612) 273-1142
Liquid-based Papanicolaou test (SurePath) interpretations before histologic diagnosis of endometrial hyperplasias and carcinomas
Study of 272 cases classified by the 2001 Bethesda system†
Version of Record online: 28 JUN 2007
Copyright © 2007 American Cancer Society
Volume 111, Issue 4, pages 217–223, 25 August 2007
How to Cite
Thrall, M., Kjeldahl, K., Gulbahce, H. E. and Pambuccian, S. E. (2007), Liquid-based Papanicolaou test (SurePath) interpretations before histologic diagnosis of endometrial hyperplasias and carcinomas. Cancer, 111: 217–223. doi: 10.1002/cncr.22770
This study was presented in part at the 54th Annual Scientific Meeting of the American Society of Cytopathology, Toronto, Ontario, Canada, November 3–8, 2006.
- Issue online: 13 AUG 2007
- Version of Record online: 28 JUN 2007
- Manuscript Accepted: 21 FEB 2007
- Manuscript Revised: 20 FEB 2007
- Manuscript Received: 27 DEC 2006
- abnormal glandular cells;
- benign-appearing endometrial cells;
- cervicovaginal cytology;
- complex hyperplasia
In addition to the reporting of atypical glandular cells (AGC) and adenocarcinoma (ADCA), the 2001 Bethesda System requires the reporting of benign-appearing endometrial cells in women aged >40 years (BAEMC). In this study, the authors evaluated the contribution of each of these reporting categories to the sensitivity and specificity of a liquid-based Papanicolaou test for endometrial carcinoma or hyperplasia.
Over the 3-year study period, in the setting of a large, multihospital health care system, the authors analyzed the results from liquid-based Papanicolaou tests that were performed within the 6 months that preceded a histologic diagnosis of endometrial carcinoma or hyperplasia and that were reported according to the 2001 Bethesda System.
Two hundred seventy-two women had a histologic diagnosis of endometrial hyperplasia (n = 199) or malignancy (n = 73) within 6 months after a Papanicolaou test. In total, 188,594 Papanicolaou tests (91,385 from women aged >40 years) were interpreted during the study period and resulted in 3810 diagnoses of BAEMC, 326 diagnoses of AGC, and 30 diagnoses of ADCA. Only 28 of 73 women (38.4%) with endometrial carcinoma had cytologically AGC or ADCA reported on a previous Papanicolaou test. The reporting of BAEMC increased this sensitivity by only 5.5% (4 additional tests) but decreased the specificity of the Papanicolaou test for endometrial malignancy from 99.8% to 96%. For endometrial hyperplasias, the sensitivity of the Papanicolaou test was even lower (39 of 198 tests; 19.7%), but BAEMC represented the majority of endometrial-type cells reported (36 of 39 tests).
The reporting of BAEMC led to an only marginal increase in sensitivity that had to be weighed against the significant loss in specificity of the Papanicolaou test for endometrial neoplasia. Cancer (Cancer Cytopathol) 2007. © 2007 American Cancer Society.
The Papanicolaou cervicovaginal test (Pap test) was designed to screen for squamous pathology of the cervix. In that regard, it has been a resounding success for decades. Endometrial cells are present incidentally on many Pap tests, providing cytopathologists an opportunity to examine these cells in specimens that may have been obtained for other reasons. Sometimes, such cells raise suspicion for significant abnormalities of the endometrium that otherwise may go undetected. However, Papanicolaou himself noted long ago that the vaginal smear method is not as accurate for diagnosing carcinoma of the fundus as it is for diagnosing carcinoma of the cervix.1 A number of older studies, in which various cervical, vaginal, and cervicovaginal sampling techniques were used, examined the potential for cervicovaginal cytology to serve as a screening test for endometrial carcinoma.2–4 Those studies demonstrated that the sensitivity of the Pap test for endometrial lesions was too low to justify the procedure for this indication. Direct cytologic sampling of the endometrium also has been studied. Those methods showed better results than cervicovaginal sampling but were not sufficiently promising to gain widespread acceptance.5 Although it is acknowledged widely that the Pap test has limited utility as a screening tool for endometrial lesions, there are occasional instances in which cells are present in these preparations that correctly suggest endometrial pathology. For this reason, cytopathologists are compelled to report endometrial cells that are observed in Pap tests under some circumstances if there is a reasonable likelihood that those cells correspond to underlying endometrial pathology. With this in mind, the 2001 Bethesda System has mandated multiple reporting categories meant to apply to endometrial cells. The well-accepted categories of adenocarcinoma and atypical glandular cells (AGC) underwent only minor revisions, whereas the reportable category of benign-appearing endometrial cells was broadened to include all such cells in women aged ≥40 years.6
A number of studies have examined the findings on surgical pathology follow-up of benign, atypical, and malignant endometrial cells in Pap tests. Recently, this literature was reviewed thoroughly for benign7, 8 and atypical9 endometrial cells. To our knowledge, however, no studies have examined the impact of the 2001 Bethesda System on these relations using an analysis of Pap tests preceding a histologic diagnosis of endometrial hyperplasia or malignancy. For the current study, we identified a cohort of women who, over 3 years, had a diagnosis of endometrial hyperplasia or adenocarcinoma made in biopsy, curettage, or hysterectomy specimens. Among those women, we identified the subset with Pap tests performed within 6 months prior to their initial surgical pathology diagnosis. Our objective was to determine, under current conditions, the relative contribution of the 2001 Bethesda System reporting categories, including adenocarcinoma, AGC, and especially benign-appearing endometrial cells, in women aged ≥40 years, to the detection of endometrial adenocarcinoma and hyperplasia.
MATERIALS AND METHODS
This study was performed on specimens that were submitted to the centralized Fairview cytology laboratory, mostly from within the Fairview Health System. Fairview is a community-based, not-for-profit health care organization with 7 hospitals, 31 primary care clinics, 25 specialty clinics, 5 urgent care clinics, and a wide range of specialty service centers across Minnesota. The patient population is predominantly suburban, but the system also includes a tertiary care center that is affiliated with the University of Minnesota. The demographics of the patients served by Fairview resemble those of the state as a whole. The Institutional Review Board of the University of Minnesota approved this study.
The Fairview cytopathology laboratory implemented the 2001 Bethesda System on December 3, 2002. Endometrial biopsies, curettings, and hysterectomies that were diagnosed as endometrial hyperplasia or adenocarcinoma were identified by a search of the surgical pathology database over a 3-year and 4-month time frame that extended to March 31, 2006. Pap tests that were performed within 6 months prior to a histologic diagnosis of endometrial hyperplasia or malignancy were identified by cross-referencing the cytology database with the results from the previous histology search for the 3-year period that extended to November 30, 2005. During the study period, >97% of all Pap test specimens that were processed by Fairview were SurePath liquid-based preparations. The remaining specimens were a mix of conventional smears and ThinPrep liquid-based preparations.
The type of hyperplasia, as classified according to the World Health Organization system10 (simple or complex hyperplasia with or without atypia), was noted along with the histologic type and the International Federation of Gynecology and Obstetrics (FIGO) grade for adenocarcinomas. The FIGO stage was recorded for all endometrial adenocarcinomas in women who underwent definitive surgery within our health care system. Among the women who had multiple histology specimens, the specimen that had the most thorough sampling (usually a hysterectomy) was selected for this study.
The Pap tests that were included in this study were reported according to the 2001 Bethesda System. Benign-appearing endometrial cells were reported in all women aged ≥40 years. Glandular cells with cytologic atypia that fell short of outright carcinoma were reported as AGC, with an accompanying favored site of origin or with the qualifier not otherwise specified. Cytologically malignant glandular cells were designated as adenocarcinoma. This interpretation also was accompanied by a favored site of origin or by the qualifier not otherwise specified. There were no Pap tests with an interpretation of adenocarcinoma in situ in the study group.
The Pap test interpretations were made by 14 pathologists, including 12 community practice pathologists and 2 academic pathologists, on samples that were screened by 12 cytotechnologists. The histopathology diagnoses were made by 17 pathologists, including 12 community pathologists and 5 academic pathologists. Neither cytologic nor histologic diagnoses were reviewed. Two-by-two contingency tables were constructed to calculate the sensitivity and specificity of the Pap test for endometrial carcinoma and hyperplasia, and 95% confidence intervals for proportions were calculated according to the efficient-score method (corrected for continuity) described by Newcombe, based on the procedure outlined by Wilson.11
From December 3, 2002 to November 30, 2005, the Fairview central cytopathology laboratory processed 188,594 Pap tests, including 91,385 tests from women aged ≥40 years. Over the 3-year study period (starting December 3, 2002 and extending to the November 30, 2005), 26 Pap tests were interpreted as adenocarcinoma, and 219 Pap tests were interpreted as AGC. Over the same period, in total, there were 3458 Pap tests from women aged ≥40 years that mentioned benign-appearing endometrial cells. Cross-referencing the results from the histopathology search with Pap tests performed within the preceding 6 months identified 272 women with endometrial hyperplasia and malignancy, including 8 women who had prior Pap tests interpreted as adenocarcinoma, 23 women who had prior Pap tests interpreted as AGC, 40 women who had prior Pap tests that mentioned benign endometrial cells, and 201 women who had prior Pap tests that did not have any endometrial cells present (Table 1).
|No.||Age range (mean), y||No. of patients||Age range (mean), y||No. of patients||Age range (mean), y|
|Adenocarcinoma||8||52–82 (69)||8||52–82 (69)||0||NA|
|AGC||23||41–79 (60)||20||43–79 (62)||3||41–52 (48)|
|BEC||40||42–74 (52)||4||52–74 (62)||36||42–66 (51)|
|None||201||40–90 (55)||41||41–90 (63)||160||40–81 (54)|
Thirty-two of the 73 women (43.8%) who had endometrial malignancies diagnosed over this period with Pap tests performed during the prior 6 months showed endometrial-type cells (4 benign-appearing endometrial cells, 20 AGC, and 8 adenocarcinomas) in the Pap test. Among those 73 endometrial malignancies, there were 38 FIGO grade 1 tumors, 19 FIGO grade 2 tumors, and 16 FIGO grade 3 tumors (Table 2). During the same period, there were 198 endometrial hyperplasias identified in women who had a previous Pap test in the preceding 6 months (Table 2), including 126 women with simple hyperplasia and 72 women with complex hyperplasia. Among the Pap tests that showed hyperplasia, 39 tests (18.1%) showed endometrial-type cells. Of these, 36 were interpreted as benign-appearing, 3 were interpreted as AGC, and none were interpreted as adenocarcinoma.
|Pap interpretation||No. of women (%)|
|Hyperplasia||Malignancy, FIGO grading|
|Simple||Complex||Atypical*||Grade 1||Grade 2||Grade 3||Subtypes†|
|Adenocarcinoma||0 (0)||0 (0)||0 (0)||0 (0)||2 (10.5)||6 (37.5)||6 (35.3)|
|AGC||2 (1.6)||1 (1.4)||0 (0)||11 (28.9)||6 (31.6)||3 (18.8)||2 (11.8)|
|BEC||20 (15.9)||16 (22.2)||4 (22.2)||3 (7.9)||1 (5.3)||0 (0)||0 (0)|
|None||104 (82.5)||55 (76.4)||14 (77.8)||24 (63.2)||10 (52.6)||7 (43.7)||9 (52.9)|
Considering the presence of any endometrial-type cell as abnormal, the overall sensitivity of the Pap test for endometrial carcinoma was 26.1% (95% confidence interval [95% CI], 21–32%). The frequency of reporting any endometrial cells, AGC, or adenocarcinoma cells on Pap tests from women who had subsequent diagnoses of endometrial hyperplasia or adenocarcinoma increased with the aggressiveness of the lesion. For simple hyperplasia, the sensitivity was only 17.6%, and neither of the 2 women who had simple hyperplasias with atypia had preceding endometrial cells. Endometrial cells preceded complex hyperplasia in 23.6% of lesions, and only 4 of 16 women who had atypical results showed prior endometrial cells, all of which were reported as cytologically benign.
The sensitivity of the Pap test increased to 43.8% (95% CI, 32.4–55.9%) for adenocarcinomas, with the percentage that showed previous endometrial cells rising from 36.8% for grade 1, to 47.4% for grade 2, and to 56.3% for grade 3 adenocarcinomas. Among the subtypes with high-risk histology, the rate of detection also was relatively high at 47.1%.
Among the women who had a biopsy diagnosis of endometrial carcinoma, 56 of 73 women (76.7%) underwent hysterectomies performed within the Fairview system and, thus, had surgical pathology reports available for review (Table 3). In this group of 56 women, 39 women (69.6%) had stage I disease, 4 women (7.1%) had stage II disease, 12 women (21.4%) had stage III disease, and 1 woman (1.8%) had stage IV disease. Only 30.8% of women with stage 1 disease had endometrial cells observed on a preceding Pap test, whereas that figure rose to 70.6% among women with higher stage disease.
|Pap interpretation||No. of women (%)|
|FIGO stage||Other parameters|
|Stage 1||Stage 2||Stage 3||Stage 4||>50% Inv*||LV Inv†|
|Adenocarcinoma||3 (7.7)||0 (0)||5 (41.7)||0 (0%)||2 (12.5)||4 (50)|
|AGC||6 (15.4)||2 (50)||4 (33.3)||0 (0)||4 (25.0)||1 (12.5)|
|BEC||3 (7.7)||1 (25)||0 (0)||0 (0)||2 (12.5)||1 (12.5)|
|None||27 (69.2)||1 (25)||3 (25)||1 (100)||8 (50.0)||2 (25)|
Of the 4 women who had endometrial malignancy and only benign-appearing endometrial cells on a previous Pap test, only 1 woman was aged <55 years. Among the 36 women who had endometrial hyperplasia and a prior Pap test that showed only benign-appearing endometrial cells, 5 women (13.9%) were aged <45 years, 10 women (27.8%) were between ages 45 years and 49 years, and 12 women (33.3%) were between ages 50 years and 54 years. All 4 women who had atypical hyperplasias and benign-appearing endometrial cells on a previous Pap test were aged <55 years, and 3 of those women were aged <50 years. Thus, unlike adenocarcinomas, the majority of hyperplasias that were diagnosed after the presence of benign-appearing endometrial cells on a Pap test occurred in relatively young women (between ages 40 years and 55 years).
The current results indicate that the sensitivity of the Pap test for endometrial malignancy is quite low; and, for endometrial hyperplasia, it is even lower. The low sensitivity for endometrial malignancy was improved only marginally (from 38.3% to 43.8%) by the addition of the 2001 Bethesda System reporting category of benign endometrial cells in women aged >40 years. The increase in sensitivity was much more substantial for endometrial hyperplasias than for malignancies after the inclusion of benign endometrial cells in women aged >40 years, because only 1.5% women with hyperplasia had atypical cells observed on their prior Pap test. However, the sensitivity only increased to 19.7%.
These modest improvements in sensitivity were realized at the cost of a decrease in specificity of the Pap test for the diagnosis of endometrial hyperplasia or malignancy from 99.8% to 96%. The specificity remains 96% for malignancies alone when benign-appearing endometrial cells are considered to be a positive test result. A test with a specificity of only 96%, when applied to a large population with a low pretest probability of disease, will produce an enormous number of false-positive results. In our population of 91,385 women aged >40 years, there were 3458 Pap tests that showed benign endometrial cells, of which only 40 tests (1.2%) corresponded to endometrial abnormalities, including only 4 malignancies (0.1%).
Using publicly available cancer demographics data, it is possible to estimate the positive predictive value of an isolated finding of benign-appearing endometrial cells in women aged >40 years. In 2002, the incidence of cancer of the uterine fundus in Minnesota was 28.5 per 100,000 women.12 The 2000 US Census data show that women aged >40 years constituted 43.8% of the population of the state.13 Combining these figures yields an estimated incidence of endometrial carcinoma in our study population of 65 per 100,000 women. In the absence of clinical information, this is the pretest probability of disease for women aged >40 years who undergo routine Pap test screening in Minnesota. Taking the sensitivity of the finding of benign-appearing endometrial cells alone for endometrial malignancy from our data (9.1%), the calculated positive predictive value is only 0.15%.
Our study included both endometrial carcinomas and hyperplasias, because endometrial hyperplasia generally is considered the precursor of type I endometrial carcinomas and because the prevalence of underlying endometrial carcinoma in women with a biopsy diagnosis of atypical endometrial hyperplasia may be very high.14 The pathologists within our system did not use the recently proposed alternative classification system of endometrial intraepithelial neoplasia, which may correspond more closely to endometrial carcinoma precursors.15–17 The natural history of endometrial hyperplasia is not understood fully; and a high proportion of simple and complex endometrial hyperplasias, especially those without atypia, regress without treatment. The length of time over which such regression may occur is unclear. Similarly, the time scale over which benign endometrium progresses to hyperplasia or over which endometrial hyperplasias progress to carcinoma also is uncertain. Although previous studies have used longer follow-up,18–20 we chose a 6-month time frame to minimize the impact of regression or progression of endometrial hyperplasias and the development of new endometrial hyperplasias or malignancies, thus avoiding the potentially larger bias induced by longer follow-up.21
The only previous study that matched our methodology, in which conventional Pap tests were classified according to the Bethesda II System, reported a similarly low sensitivity of 47% for endometrial malignancy.18 Many other studies have examined the cervical cytology findings in women with endometrial adenocarcinoma. Those published prior to 1970 have been well summarized elsewhere,3 with sensitivities ranging from 32% to 92%. Large changes in the diagnosis and treatment of endometrial carcinoma, as well as in Pap test techniques and classification, make these results difficult to compare with more recent findings. Studies since 1970 have focused predominantly on the potential role of cervical cytology as a predictor of grade, stage, or other clinical parameters (Table 4).3, 18, 22–34 The sensitivities of those studies range from 20% to 90%. Many report higher sensitivity than we found in the current study; however, in most instances, this well may have resulted from a review of Pap tests by cytologists who were aware of the underlying malignancy, with a concomitant change in the perception of abnormal.3, 24, 26, 27, 29, 32 Comparisons between those studies also are hampered by significant differences in the classification of Pap tests, including a lack of consensus about whether benign-appearing endometrial cells constitute normal or abnormal findings. In addition, differentiating between benign, atypical, and malignant endometrial cells in Pap tests may be difficult.35 Therefore, findings that were classified as benign endometrial cells in 1 study may have been classified as atypical or even malignant in another. Only 2 studies have examined the prior Pap test findings in women who were diagnosed later with endometrial hyperplasia, and those investigators reported very low sensitivities of 22%3 and 10%.24
|Study||Year||No. of patients||BEC*||Sensitivity, %|
|Total no.||Normal results||Abnormal results|
|Burk et al.23||1974||154||123||31||Unknown||20|
|Bibbo et al.22||1974||197||105||92||Normal||47|
|Bibbo et al.24||1979||33||10||23||Normal||70|
|Lozowski et al.25||1986||58||23||35||Abnormal||60|
|Schneider et al.26||1986||217||88||169||Normal||78|
|Kuebler et al.27||1989||34||12||22||Unknown||65|
|Larson et al.28||1994||164||94||70||Normal||43|
|Zuna and Erroll29||1996||61||25||36||Unknown||59|
|Eddy et al.18||1997||112||59||53||Abnormal||47|
|Fukuda et al.30||1999||99||68||31||Normal||31|
|Gu et al.31||2001||76||34||42||Unknown||55|
|Morimura et al.32||2002||53||15||38||Normal||72|
|DuBeshter et al.33||2003||300||109||191||Abnormal||64|
|Brown et al.34||2005||412||139||273||Abnormal||66|
Despite the increased attention to endometrial cells as a result of the application of the 2001 Bethesda reporting system, in the current study, we did not observe any notable improvement in sensitivity compared with the results from older studies. We believe that the persistent finding low sensitivity of Pap tests for endometrial lesions results primarily from a lack of potentially diagnostic cells in these preparations. Pap tests sample the endometrium only incidentally and inconsistently. Furthermore, many endometrial hyperplasias and malignancies have small volumes at the time of detection, decreasing the likelihood of substantial cell shedding. Our study suffers from the weakness of being a retrospective correlation of pathology results from within a single health care organization. We have no way of knowing whether women who had Pap tests prior to a histologic diagnosis of an endometrial lesion systematically differed from women who did not, although this seems unlikely. Furthermore, we do not have Pap tests preceding every diagnosis of endometrial hyperplasia or malignancy, nor do we have histologic follow-up of every woman with endometrial cells reported on a Pap test. To determine the true sensitivity of Pap testing for endometrial lesions, a prospective study with standardized follow-up of Pap tests reports would be necessary. The sensitivity of Pap tests for endometrial hyperplasia and malignancy remains quite low despite the introduction of liquid-based Pap tests. The implementation of a requirement to report benign-appearing endometrial cells in women aged >40 years has led to a slight increased in sensitivity while dramatically reducing specificity of the Pap test for endometrial lesions. Increased reporting of benign-appearing endometrial cells may lead to the detection of a greater number of otherwise occult endometrial malignancies and premalignant lesions. However, in the absence of clear consensus and guidelines regarding clinical management, it also could produce a substantial increase in the number of investigations of healthy, middle-aged women with a concomitant rise in office visits, diagnostic procedures, and patient anxiety. The expected benefits and harms of reporting benign-appearing endometrial cells must be weighed against one another. A recent report indicated that routine endometrial sampling of asymptomatic, premenopausal women is not cost effective,36 indicating the need for a more targeted approach. Our current results suggest that clinical actions spurred by a finding of benign-appearing endometrial cells will produce a reasonable yield only in women who have a substantially increased pretest probability of finding disease. The systematic application of the knowledge of patient menstrual status and bleeding symptoms held by the physicians performing Pap tests is the most effective means of selecting women for biopsy. With this in mind, we support the recent recommendation to perform follow-up studies for benign-appearing endometrial cells only in postmenopausal women.8
- 1Diagnosis of Uterine Cancer by the Vaginal Smear. New York, NY: The Commonwealth Fund; 1943., .
- 12U.S. Cancer Statistics Working Group. United States Cancer Statistics: 2002 Incidence and Mortality. Atlanta, Ga: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention and National Cancer Institute; 2005.
- 13U.S. Census Bureau. Census 2000 Data for the State of Minnesota. Sex by Single Years of Age (PCT12). Washington, DC: Government Printing Office; 2002.