CANCER 254 CYTOPATHOLOGY Adenocarcinoma In Situ with a Small Cell (Endometrioid) Pattern in Cervical Smears A Test of the Distinction from Benign Mimics Using Specific Criteria Kenneth R. Lee, M.D. Department of Pathology, Brigham and Women’s Hospital, Boston, Massachusetts. BACKGROUND. Papanicolaou smears have been less effective in preventing cervical adenocarcinoma than in preventing squamous carcinoma. One reason may be a lack of awareness of certain smear patterns of adenocarcinoma in situ (AIS) such as those with crowded small cells (endometrioid pattern). METHODS. A test set of 29 smears (17 AIS with an endometrioid pattern, 12 benign mimics) was reviewed by 11 cytologists (4 experienced cytotechnologists, 3 cytopathology fellows, and 4 cytopathologists with varying levels of experience). Participants were blinded as to the actual diagnosis and the number of cases in each category and were instructed to diagnose either AIS or a benign lesion. Results of this review were not disclosed before a second review conducted after instruction in specific criteria for “endometrioid” AIS. Results were compiled using kappa statistics. RESULTS. In the first round, the ability to distinguish these lesions was poor for 8 of the 11 reviewers, and no reviewer was in excellent agreement with the actual diagnosis. In the second round, only 1 reviewer had a poor rating, and 4 of 11 were in the excellent category. Misdiagnoses in both rounds were more commonly the result of underdiagnosis of AIS than overdiagnosis of benign cases. CONCLUSIONS. The presentation of AIS in smears as groups of crowded small cells is prone to underdiagnosis. Awareness of this problem and use of criteria improves sensitivity. [See editorial on pages 243– 4, this issue.] Cancer (Cancer Cytopathol) 1999;87:254 – 8. © 1999 American Cancer Society. KEYWORDS: cervix, adenocarcinoma, adenocarcinoma in situ, cytology, Papanicolaou smear, cervicovaginal smears. n a prior study1 it was found that in cases of adenocarcinoma in situ (AIS) that were underdiagnosed in cervicovaginal smears, there were two cytologic patterns. In one, cells resemble reactive endocervical cells; in the other, the cells are small, mimicking endometrial cells or endocervical cells from high in the canal and/or with tubal metaplasia. To study the latter problem, a test set of 29 smears composed of AIS cases containing small (endometrioid) cells and cases containing their benign mimics was reviewed by three pathologists with the goal of developing criteria to separate benign from malignant cells in these diagnostically difficult cases.2 In that study statistically significant criteria selected as favoring AIS were as follows: I The author thanks Elizabeth Allred, Ph.D., for statistical assistance. Address for reprints: Kenneth R. Lee, M.D., Department of Pathology, Brigham and Women’s Hospital, 75 Francis St., Boston, MA 02115. Received October 7, 1998; revisions received March 5, 1999; accepted March 18, 1999. © 1999 American Cancer Society 1. Absence of endometrial stromal cells and endometrial-like tubules 2. Coarse chromatin Testing Criteria for “Endometrioid” AIS/Lee TABLE 1 Criteria Favoring “Endometrioid” AIS vs. Benign Mimics Absent endometrial stroma Coarse nuclear chromatin Extreme nuclear crowding Mitotic figures “Feathering” Absent endometrial tubules Rosettes Strips of columnar cells Modified from Lee KR, Genest DR, Minter LJ, Granter SR, Cibas ES. Adenocarcinoma in situ in cervical smears with a small cell (endometrioid) pattern: distinction from cells directly sampled from the upper endocervical canal or lower segment of the endometrium. Am J Clin Pathol 1998;109:738–42. 3. Extreme nuclear crowding (lack of honeycombing) 4. Mitotic figures 5. Feathering. Rosettes of cells and strips of columnar cells also favored AIS but were not statistically significant (Table 1).2 The current study uses the same 29 cases to test the ability of cytotechnologists and pathologists to discriminate among them before and after having been instructed in the use of the above criteria. 255 TABLE 2 Number of Incorrect Diagnoses of 29 Unknown Cervicovaginal Smears of Either AIS or Benign Glandular Cells Reviewed by 11 Study Participants Mean Median Range First round Second round 9.4 11 5–14 6.2 6 2–15 TABLE 3 Kappa Statistic for Each Individual Review of 29 Cases Cytotechnologists Cytology fellows Cytopathologists First round Second round 0.35 0.19 0.03 0.30 0.66 0.19 0.35 0.30 0.38 0.68 0.65 0.10 0.46 0.58 0.47 0.73 0.56 0.78 0.60 0.53 0.72 0.86 Kappa , 0.4 5 poor, 0.4 to 0.7 5 fair to good, . 0.7 5 excellent. MATERIALS AND METHODS Thirty-five smears (20 AIS,; 15 benign) were selected from a teaching collection accrued from the files of the Brigham and Women’s Hospital, Boston, Massachusetts, and the Medical Center Hospital of Vermont, Burlington ,Vermont, during the interval May 1982 to December 1996. Cases were selected by the author from a larger group that originally had caused difficulty in the differential diagnosis between AIS and cells from the upper endocervix, sometimes with tubal metaplasia, or from the endometrium. These were reviewed, and cases in which the diagnosis was judged straightforward and those in which diagnostic cells were scant or poorly preserved were excluded. The 20 AIS smears were selected from 12 patients. The diagnosis of AIS without a concomitant high-grade squamous lesion was confirmed by cone biopsy, hysterectomy, or both. The 15 smears with benign lesions were from 15 different patients, 7 of whom had undergone a previous cone biopsy (high-grade squamous lesion, 5; AIS, 1; tubal metaplasia, 1.). Follow-up in these 15 patients included one to three negative smears in 7 patients (4 with negative results of endocervical curretage as well) and a cone biopsy or hysterectomy with negative results in 6 (4 with tubal metaplasia of the cervix). Two smears were recent examples of direct endometrial sampling that had caused difficulty in the distinction from AIS. Three smears from each of the two categories were selected as teaching slides to be used after all participants had reviewed the remaining 29 cases for the first time. The 29 smears were reviewed independently by 11 individuals (4 experienced cytotechnologists, 3 cytopathology fellows, and 4 cytopathologists of varying levels of experience). Participants were not informed of the number of cases in each category and were instructed to record their diagnosis as either benign or AIS. After the first review, participants were instructed in the differential diagnosis, first by a lecture format in which the criteria found to be most useful in the prior study (Table 1.) were illustrated and discussed and then by small group instruction using a multiheaded microscope with the teaching set of six cases. Several months later participants re-examined the 29 smears without knowledge of how they had performed in the original review. Incorrect diagnoses for each individual for the first and second review were recorded and analyzed by kappa statistics, which account for the skewed probability for an incorrect diagnosis resulting from the imbalance of cases with AIS and benign changes (kappa , 0.4 5 poor, 0.4 to 0.7 5 fair to good, . 0.7 5 excellent). After results were tallied, smears from individual cases that were particularly problematic (6 or more of the 11 participants with an incorrect diagnosis 256 CANCER (CANCER CYTOPATHOLOGY) October 25, 1999 / Volume 87 / Number 5 FIGURE 1. Representative groups from four difficult AIS cases. (A) Small cells with extreme crowding, nuclear coarseness. Wispy cytoplasmic “tails” are seen at the bottom of the group. (B) Two groups with extreme nuclear crowding. The group on the left is a strip of columnar cells. Nuclear coarseness is less evident in these groups. (C) AIS group with feathering, extreme nuclear crowding, and chromatin coarseness. (D) A small rosette is present at bottom right of a group containing crowded, hyperchromatic, small nuclei. in either round) were reviewed by the author to assess possible reasons for the difficulties. RESULTS The number of incorrect diagnoses for the entire group in each of the two rounds is given in Table 2. In Table 3 the diagnoses for each individual relative to the expected diagnosis is given in terms of kappa values and stratified by cytotechnologists, fellows and cytopathologists. These data may be summarized as follows: the ability to separate AIS with small (endometrioid) cells from benign lesions in the first round was poor for 8 of the 11 reviewers, whereas in the second round, it remained poor for only 1 of the 11. Conversely, no reviewer had an excellent rating in the first round, whereas, in the second round 4 of 11 were in the excellent category. There was a tendency towards underdiagnosis of AIS rather than overdiagnosis of benign lesions. In the first round, considering all diagnoses from the 11 observers, 40% of AIS cases were misdiagnosed versus 21% of the benign cases. In the second round, these values were 26% and 14%, respectively. In the first round all six cases in which there were six or more incorrect diagnoses were AIS. In the second round there was only one case, also AIS, with six incorrect diagnoses. Upon review of the six AIS cases that were particularly prone to underdiagnosis, the common theme appeared to be the mistaken identification of small groups of crowded cells as either benign endometrial or endocervical cells. In some of these cases, the abnormal cells were confined to small areas of the slide. However, in no case were they extremely scant or poorly preserved. Representative cell groups from four of these difficult AIS cases are illustrated in Figure 1. From this figure, the difficulties in suspecting AIS may be appreciated. Some of the cell groups are quite small, and there is a lack of significant nuclear pleomorphism and obviously malignant nuclear features. However, the cells are generally well preserved, and most groups demonstrate some of the findings that Testing Criteria for “Endometrioid” AIS/Lee 257 FIGURE 2. Two benign cases demonstrate findings that discriminate them from AIS. (A) An endometrial tubule is seen at the top, and a sheet of evenly spaced endocervical cells is present at the bottom. The nuclear chromatin is fine and evenly dispersed. (B) A folded sheet of cells containing evenly spaced small nuclei with fine, evenly dispersed chromatin. help distinguish AIS from benign endometrial and endocervical cells (illustrated in Fig. 2), such as extreme nuclear crowding, nuclear hyperchromasia, and slight coarsening of the chromatin in conjunction with subtle endocervical cell differentiation (feathering, columnar cells, rosettes). DISCUSSION Invasive endocervical adenocarcinoma is increasing in proportion to squamous carcinoma.3,4Evidence exists that Papanicolaou smears do not afford the same degree of protection from the development of adenocarcinoma as they do for squamous carcinoma.5–7 To increase the effectiveness of the Papanicolaou smear in the prevention of adenocarcinoma, it is critical that cytotechnologists and cytopathologists are able to recognize the patterns of presentation of the precursor lesion, adenocarcinoma in situ. It has been shown that AIS smears with groups of small (endometrioid) cells are particularly prone to underdiagnosis.1 The results in the first round of the current study reinforce that observation, while at the same time the round two results demonstrate that instruction in the recognition of this pattern can improve diagnostic accuracy. The principal differential diagnosis is with small glandular cells, which may be derived either from the upper endocervix, sometimes with tubal or tuboendometrial metaplasia,8,9 or from the endometrium.10,11 Smears with such benign cells are more commonly encountered than those with AIS, and it is crucial that cytologists recognize this distinction in attempting to increase sensitivity for AIS without a large increase in falsely positive smears. The value of the criteria previously developed2 and education in the use of these criteria has been demonstrated in this study. However, the differential diagnosis remains a difficult one. It is hoped that through the educational function of exercises such as this one and the dissemination of information relative to the less well known features of adenocarcinoma in situ, awareness will improve and the Papanicolaou smear will become a more powerful tool in the prevention of cervical adenocarcinoma. REFERENCES 1. 2. 3. 4. 5. 6. 7. Lee KR, Minter LJ, Granter SR. Papanicolaou smear sensitivity for adenocarcinoma in situ of the cervix: a study of 34 cases. Am J Clin Pathol 1997;107:30 –5. Lee KR, Genest DR, Minter LJ, Granter SR, Cibas ES. Adenocarcinoma in situ in cervical smears with a small cell (endometrioid) pattern: distinction from cells directly sampled from the upper endocervical canal or lower segment of the endometrium. Am J Clin Pathol 1998;109: 738 – 42. Hopkins MP, Morley GW. A comparison of adenocarcinoma and squamous carcinoma of the cervix. Obstet Gynecol 1997; 77:912– 7. Peters RK, Chow A, Mack TM, Bernstein TD, Henderson BE. Increased frequency of adenocarcinoma of the uterine cervix in young women in Los Angeles County. J Natl Cancer Inst 1986;76:423– 8. Nieminen F, Kallio M, Hajama M. The effect of mass screening on incidence and mortality of squamous and adenocarcinoma of the cervix uteri. Obstet Gynecol 1995; 85:1017–25. Boon ME, Guilloud JCD, Kok LT, Olthof AM, van Erp EJM. Efficacy of screening for cervical squamous and adenocarcinoma: the Dutch experience. Cancer 1987;59:862– 6. Mitchell H, Medley G, Gordon I, Giles G. Cervical cytology reported as negative and risk of adenocarcinoma of the cervix: no strong evidence of benefit. Br J Cancer 1995;71: 894 –7. 258 8. CANCER (CANCER CYTOPATHOLOGY) October 25, 1999 / Volume 87 / Number 5 Novotny DB, Maygarden SJ, Johnson DE, Frable WJ. Tubal metaplasia: a frequent potential pitfall in the cytologic diagnosis of endocervical glandular dysplasia on cervical smears. Acta Cytol 1992;36:1–10. 9. Ducatman BS, Wang HH, Jonasson JG, Hogan CL, Antonioli DA. Tubal metaplasia: a cytologic study with comparison to other neoplastic and non-neoplastic conditions of the endocervix. Diagn Cytopathol 1993;9:98 –105. 10. Babkowski RC, Wilbur DC, Rutkowski MA, Facik MS, Bon- figlio TA. The effects of endocervical canal topography, tubal metaplasia, and high canal sampling on the cytologic presentation of non-neoplastic endocervical cells. Am J Clin Pathol 1996;105:403–10. 11. dePeralta-Venturino MN, Purslow MJ, Kini SR. Endometrial cells of the “lower uterine segment” (LUS) in cervical smears obtained by endocervical brushings: a source of potential diagnostic pitfall. Diagn Cytopathol 1995; 12:263–71.