Accuracy of endocervical cytological tests in diagnosing preinvasive lesions of the cervical canal in patients with type 3 transformation zone: a retrospective observational study

ABSTRACT BACKGROUND: Cervical cancer screening in Brazil is done using Pap smears. Women who are most likely to have a preinvasive lesion or cervical cancer are immediately referred for colposcopy. OBJECTIVE: The aim of this study was to evaluate the diagnostic performance of endocervical cytological tests in diagnosing preinvasive cervical lesions in women with initial high-grade squamous intraepithelial lesions (HSIL), or atypical squamous cells in which high-grade lesions could not be ruled out (ASC-H), or atypical glandular cells (AGC), and whose colposcopy did not show any abnormalities, with no fully visible transformation zone (types 2 and 3). DESIGN AND SETTING: Retrospective observational study conducted in Rio de Janeiro, Brazil. METHODS: Data from women who came to the cervical pathology outpatient clinic between January 2012 and April 2017 were analyzed. The results from endocervical cytological tests were compared with the final diagnosis, which was obtained through examination of a surgical specimen or, among women who did not undergo an excisional procedure, after cytological and colposcopic follow-up for two years. RESULTS: We included 78 women. The sensitivity of endocervical cytological tests was 72.7%; specificity 98.5%; positive and negative predictive values 88.9% and 95.6%, respectively; and positive and negative likelihood ratios 48.7 and 0.28. CONCLUSION: Endocervical cytological tests are simple, inexpensive and noninvasive, and form a reliable method for determining management among patients with HSIL, ASC-H and AGC cytological findings and negative colposcopic findings without visualization of the squamocolumnar junction.


INTRODUCTION
Cervical cancer is the fourth most common type of cancer among women worldwide, accounting for 570,000 new cases and 305,000 deaths in 2018. 1 In Brazil, 16,370 new cases were expected for that same year. In 2016, it was the third most frequent tumor, and the fourth largest cause of cancer-related death, excluding non-melanoma skin cancer. 2 While mortality has declined over recent years, 3 cervical cancer remains a public health problem, and reducing its incidence and mortality is still an ongoing challenge.
Periodic conventional Pap smears remain the most widely adopted strategy for cervical cancer screening in Brazil. When the results from this test show alterations that are most likely to represent the presence of cancer or its preinvasive lesions, immediate referral for colposcopy is necessary. 4 These lesions can include presence of atypical squamous cells in which high-grade lesions cannot be ruled out (ASC-H); atypical glandular cells (AGC); high-grade squamous intraepithelial lesions (HSIL); high-grade squamous intraepithelial lesions in which microinvasion cannot be ruled out; invasive squamous carcinoma; adenocarcinoma in situ (AIS); and invasive cancer.
When the squamocolumnar junction (SCJ) is not completely visible (type III transformation zone, TZ), colposcopy cannot rule out the presence of endocervical disease. In these cases, the most accurate method for doing so, or for treating it if present, is type III excision (cervical conization). However, this procedure increases the risk of obstetric complications 5 or other disorders, such as endocervical canal stenosis, thereby impairing these patients' follow-up, 6

METHODS
This was an observational study, in which information was obtained from our institution's cervical pathology outpatient database, complemented by data extracted from medical records.
Women who first came to our clinic between January 2012 Women with an indication for colposcopy are referred for diagnosis and treatment at units that can provide care for cases of higher complexity, such as the institution where the study was carried out.
This institution has one of the referral units for diagnosing and treating preinvasive cervical lesions in the state of Rio de Janeiro.
Endocervical cytological specimens were obtained by means of endocervical brushing, which was done after the colposcopic examination. These specimens were processed using the conventional method (Papanicolaou).
Type III excision procedures (cervical conizations) were performed by means of electrosurgery, using a 2 cm loop electrode or a straight wire electrode to reach a depth of at least 2 cm in the endocervical canal. 10 The length of follow-up for women who did not undergo the excisional procedure was two years, with semiannual cytological tests and colposcopy. After this period, if there was no evidence of preinvasive disease or cancer, they were dis-

Study population
All the women who were referred to our institution from primary healthcare facilities and were received presenting either of the following were included in this study: (a) initial HSIL or ASC-H cytological and colposcopic findings without significant abnormalities, and with non-fully visible or partially visible SCJ (type III TZ); or (b) with initial AGC cytological findings, regardless of the type of TZ or abnormal colposcopic findings. In all these cases, the women underwent endocervical cytological tests to define the case management.
Patients whose endocervical tests done during the initial approach at our institution showed the same positive results as in the referral test, or more relevant results than those of the referral test, were considered positive. It was recommended that these patients should undergo an excisional procedure (type III excision, i.e. conization), although in reality not all of these patients underwent the procedure, while some others did so even though the new test had been negative. For patients who were considered negative (i.e. those with other results), it was recommended that they should be followed up for at least two years after the initial assessment.
Pregnant women were not included in this study because of the difficulty in following them up and the impossibility of performing an immediate excisional procedure. Women who were lost to follow-up, despite repeated attempts to contact them via letters or by telephone, were excluded.

Diagnostic criteria
Endocervical cytological tests that showed diagnoses of HSIL, ASC-H or AGC were considered positive, since these results, according to the Brazilian Guidelines for Cervical Cancer Screening, indicate that an excisional procedure is needed in order to obtain the diagnosis. In cases in which squamous preinvasive disease is present, no further treatment is needed. 4 There were no cases with AIS in the endocervical cytological findings, or with cancer. All others, in which satisfactory material was obtained (including at least squamous and glandular or metaplastic epithelium) and did not show any of these cytological diagnoses, were considered negative.

Gold standard
The results from the endocervical cytological tests were compared with the definitive diagnosis that had been obtained through one of the following methods: (a) among women undergoing type III excision (conization), histopathological examination of the specimen obtained as part of the initial approach (through a positive endocervical cytological test); or (b) among women undergoing cytological and colposcopic follow-up (not undergoing excisional procedure as the initial approach), negative results from cytological and colposcopic examinations that were performed semiannually for two years or from histopathological examination of the specimen obtained through an excisional procedure performed after a new cytological or colposcopic examination showed a suspected preinvasive lesion during follow-up.
The final diagnosis took into consideration whether a preinvasive cervical lesion was present: either squamous (cervical intraepithelial neoplasia, CIN 2 or 3) or glandular (AIS). No cases of cancer were found. All other diagnoses were taken to be absence of preinvasive disease or cancer.

Data analysis
The data on the women included in this study were transferred to a spreadsheet. After checking for nonconformities, they were corrected and supplemented using information from the physical medical charts. Performance measurements were calculated from contingency tables that were constructed using the Statistical Package for the Social Sciences (SPSS), version 21.

Ethical matters
This was a retrospective study. The stored data were identified only by the numbers of the medical records, thus preserving the women's identity. This study was approved by the local research ethics committee under the number CAE 03847218.0.0000.5269, on December 20, 2018.

RESULTS
Ninety-one women seen between January 2012 and April 2017 were found to be eligible. Of these, 13 were excluded because they had not completed a two-year follow-up. Hence, 78 patients remained in the study, among whom nine women were deemed to be positive, since they had endocervical cytological findings that maintained the cytological diagnosis of their referral or another result suggestive of preinvasive disease. These women underwent type III excision.
The remaining 69 women whose endocervical cytological findings suggested diagnoses of lesser importance than the initial one were deemed to be negative (Figure 1). Twelve women, out of the 78 women who remained in the study, underwent an excisional procedure after the early colposcopy (three despite negative endocervical cytological results) and the other 66 (66/78) This woman underwent an excisional procedure that confirmed the presence of preinvasive disease (CIN 3) and her case was considered to be a false negative from the endocervical cytological test that had been performed during the initial assessment.
A total of 13 women underwent an excisional procedure (12 after the initial approach and one during follow-up). Among these, 11 had findings compatible with a preinvasive lesion (CIN 2-3 or CIN 2-3 with concomitant AIS) and two were negative (no abnormalities or CIN 1).
The characteristics of the women included, the initial colposcopic findings and the final diagnoses are shown in Table 1. Baseline cytological tests showing AGC were more frequent; the patients' mean age was close to 50 years; and HSIL was the preinvasive disease most often found among the women with baseline cytological tests.   Table 2). It can also be considered that the excisional procedure was still useful when it did not show the presence of preinvasive lesions or cancer, since it served as a diagnostic method, thereby ensuring the absence of endocervical preinvasive disease in women who had had at least two positive cytological examinations (baseline and the one done during early colposcopy).
The large confidence intervals that were observed in the diagnostic performance measurements ( Table 2) resulted from the small sample size. Despite the long study period of five years, few women met the requirements for using endocervical cytological findings as a predictor for endocervical disease.
The exclusion of women who had not completed two years of follow-up after negative endocervical cytology results contributed to this limitation. However, this also had the aim of preventing the bias that would result from their inclusion, since they would not have had enough time to show one of the outcomes.
We calculated the diagnostic performance measurements separately between women whose baseline cytological tests showed squamous atypia (ASC-H or HSIL) and those whose   Negative predictors are essential because they reduce the risk of overtreatment, which could lead to reproductive problems in the future for these women (such as miscarriage and premature birth), 5 along with cervical canal stenosis. Such situations would impair the follow-up among these women or lead to secondary dysmenorrhea or even hematometra, thus requiring new surgical procedures. 6 The mean age that was observed in our sample resulted from the inclusion criteria through which the women were selected.
This characteristic did not represent bias, since these were the women who would benefit most from endocervical cytological testing, to whom these results can be applied, following the rec- Our study showed results that were close to those obtained by the only paper identified in the systematic review cited above.
In the study thus identified, Goksedef et al. 11 compared the performance of endocervical brushing with that of cervical curettage. In their study, endocervical cytological tests showed slightly higher sensitivity than what we observed (83.3%), and similar specificity (96.5%). The difference in sensitivity may be explained in terms of random error due to the small sample size of both studies, in association with other variations, such as their inclusion criteria, since they only included women with LSIL as referral cytological results, among whom preinvasive lesions are known to present low prevalence. Moreover, those authors used a different brushing technique for sample collection: the material obtained through brushing was fixed in formaldehyde and processed for embedding in paraffin blocks, followed by histological analysis (rather than using conventional cytological analysis). This may produce performance differing from that of the usual technique. Hence, the results reported by Goksedef et al. 11 can only be extrapolated to sample collections using the same method. Apart from these factors, they did not define the method and depth of treatment excision that would be considered the gold standard in the study (which may have missed some positive results in the case of endocervical lesions) and did not find the diagnoses that were eventually obtained during the follow-up, when results that had been lost in the early evaluation could be obtained, as we did in our study, which decreased the sensitivity of the test in our research.

CONCLUSION
We can conclude that, in our practice, conventional endocervical brushing is a reliable test for determining the management that should be used among patients with HSIL, ASC-H and AGC screening cytological and colposcopic tests without abnormal findings and without a full view of SCJ (Type III TZ) as recommended in the Brazilian Guidelines for Cervical Cancer Screening. The likelihood ratios found ensured that endocervical cytological testing presented good diagnostic performance, regardless of the prevalence of the disease among these women.
Further studies with larger sample sizes and in other scenarios are required to increase the accuracy and consistency of these results and to strengthen the recommendations for using this test to investigate women in the situation studied here.
Endocervical cytological testing is a simple low-cost noninvasive method that should be encouraged in referral centers that receive these patients. Its use can avoid unnecessary hospitalizations and invasive procedures, thereby reducing the potential complications of type III excision (conization).