Diagnostic accuracy of intraoperative frozen section at radical abdominal trachelectomy for early-stage cervical cancer.

Cervical cancer is one of the most common gynecological malignancies among young women worldwide [ 1 ]. Although the incidence of invasive cervical cancer has decreased in most Western countries, the number of patients diagnosed with early-stage cervical cancer who are of reproductive age is increasing [ 2 ]. Additionally, as the age at which women have their first child increases, significant efforts have been directed towards fertility-sparing approaches for the treatment of early-stage cervical cancer [ 3 ]. The National Comprehensive Cancer Network guidelines recommend radical trachelectomy as a treatment option for patients with early-stage cervical cancer who wish to preserve their fertility [ 4 ]. The use of fertility-sparing trachelectomy has increased in patients with early-stage cervical cancer and is widely recognized as a safe and effective treatment [ 5 6 7 8 ]. Patients with tumors of <2 cm in size and without lymph-node metastasis were considered eligible for radical trachelectomy [ 3 ]. Based on the Surveillance, Epidemiology, and End Results database, there was no difference in 5-year disease-specific survival between the fertility-sparing surgery group and the non-fertility sparing surgery (any type of hysterectomy) group [ 8 ]. The ConCerv trial [ 9 ] and the SHAPE trial [ 10 ] have reported oncological outcomes of less radical procedures for cervical cancer, and less radical surgery appears oncologically safe for selected low-risk patients with early-stage cervical cancer. In the Concerv trial [ 9 ], the “low-risk” criteria were tumor size of <2 cm, squamous cell or adenocarcinoma (grade 1 or 2 only) histology, no lymphovascular space invasion, invasion depth of <10 mm, and negative conization margins. Radical trachelectomy is an indispensable therapeutic option for those who desire to preserve fertility and do not satisfy the “low-risk” criteria. We adopted radical trachelectomy in 2002 and have since reported on its oncological and obstetric outcomes [ 11 12 ]. Intraoperative frozen sections for surgical margins and lymph nodes are essential to confirm the surgical criteria and to ensure optimal oncological outcomes for this procedure [ 13 ]. A clear endocervical surgical margin is critical to prevent local recurrence in radical trachelectomy. Compared with hysterectomy, while nodal recurrence rates were comparable, 27% of recurrences following trachelectomy occurred in the cervix [ 14 ]. Conversely, residual uterine length has been reported as a risk factor for cervical stenosis following trachelectomy [ 15 ]. Therefore, frozen section is crucial for determining an adequate resection margin. Previous reports have described techniques for frozen section during this procedure [ 16 17 18 19 20 ]; however, no reliable consensus guidelines or large case-series data exist on diagnostic performance. Thus, this study was conducted to evaluate the diagnostic accuracy of frozen sections, with a detailed investigation of false-negative cases among patients undergoing radical abdominal trachelectomy (RAT) at our institution.

In total, 326 patients with stage IA1–IB1 cervical cancer who met the inclusion criteria were planned to undergo RAT to preserve fertility, of whom 30 (9.2%) required conversion to RAH based on intraoperative findings. The demographics of the patients planned to undergo RAT are listed in Table 1 . Among these patients, 28 (8.6%) had FIGO 2008 stage IA1, 27 (8.3%) had stage IA2, and 271 (83.1%) had stage IB1. The median follow-up was 71 months (range, 1–232 months). Tumor histology revealed SCC, adenocarcinoma, and adenosquamous carcinoma in 251 (77.0%), 67 (20.6%), and 8 (2.5%) patients, respectively. Tumor size was determined either by preoperative assessment or final pathological specimen analysis, with a median tumor diameter of 10 mm (range, 1–38 mm). Cervical conization was performed prior to the procedure in 242 patients (74.2%), primarily to confirm histology. Among these patients, residual tumors were found in 45 (13.8%). Of the patients initially planned to undergo RAT, 28 (8.6%) were converted to RAH; 14 (50.0%) due to positive surgical margins, 10 (35.7%) due to pelvic lymph-node metastases, 3 (10.7%) due to gross tumor size >20 mm, and one (3.5%) each due to vaginal wall invasion, pelvic dissemination, and prominent LVSI. All these results were confirmed by frozen section examination ( Table S1 ). Values are presented as median (range) or number (%). FIGO, International Federation of Gynecology and Obstetrics; LN, lymph node; LVSI, lymphovascular space involvement; SCC, squamous cell carcinoma. In total, 361 frozen sections of surgical margins were performed on 315 patients, excluding 11 patients who were converted to hysterectomy due to lymph-node metastases or peritoneal dissemination. Among the 326 patients planned to undergo RAT, 214 (65.6%) underwent intraoperative lymph-node biopsies, with a total of 446 lymph nodes biopsied. We conducted a comprehensive chart review of frozen section examinations for patients with positive surgical margins and lymph-node metastases. We compared the frozen section diagnoses with the permanent section diagnoses and identified 4 false-negative cases of positive surgical margins. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were 93.5%, 100%, 100%, and 98.7%, respectively. No false-positive results were observed. Fig. S1 shows the sensitivities of the frozen sections of the surgical margins in each period; there were no significant differences over time. In cases where lymph-node biopsy was performed intraoperatively on apparently enlarged lymph nodes, there was one false-negative result. No false-positive results were observed. The sensitivity, specificity, PPV, and NPV were 94.1%, 100%, 100%, and 99.8%, respectively. Table 2 summarizes the diagnostic performance of frozen sections of the surgical margins and lymph nodes. Values are presented as percentage (95% confidence interval). LN, lymph node; NPV, negative predictive value; PPV, positive predictive value. Frozen section analysis was positive in 10 out of 214 (4.7%) patients, while lymph-node metastases were present in 21 patients (6.4%) according to the final pathology of systematic PLND. We identified lymph-node metastases on final pathology of PLND specimens in 2.0% (4/204) of patients with negative intraoperative lymph-node biopsy results and 6.3% (7/112) of patients who did not undergo lymph-node biopsy ( Fig. 2 ). Among the 21 cases with lymph-node metastases, 10 were detected by biopsy of enlarged lymph nodes. The sensitivity was 47.6% (95% confidence interval [CI]=28.3%–67.6%), which indicated that evaluation based solely on lymph-node appearance during surgery could not detect approximately half of the lymph-node metastases. LN, lymph-node; PLND, pelvic lymph-node dissection; RAT, radical abdominal trachelectomy. Table S2 details the false-negative cases identified on frozen section analysis of the surgical margins. These cases were diagnosed as margin-negative on frozen section; however, tumor cells were identified on paraffin sections from the same specimen (2 cases of SCC, one of adenocarcinoma in situ [AIS], and one of cervical intraepithelial neoplasia grade 3 [CIN3]). Fig. 3 illustrates a false-negative case where SCC was not recognized on the frozen section. The frozen section revealed findings suggestive of inflammatory cell infiltration into the stroma, with large reactive fibroblasts and vascular endothelial cells, making the morphological identification of tumor cells challenging on hematoxylin-eosin (H&E) slides ( Fig. 3A and B ). In the paraffin sections, cellular details were more clearly visualized due to inferior staining quality on frozen section, making tumor cells with enlarged nuclei and distinct nucleoli in the same area more apparent ( Fig. 3C and D ). Additionally, p16 and cytokeratin staining revealed the presence of tumor cell infiltration ( Fig. 3E and F ). H&E, hematoxylin-eosin. Fig. 4 presents another false-negative case where AIS was identified at the surgical margin. No atypical glands were present on the frozen section, but AIS appeared on the paraffin section. This discrepancy may have been due to a slight difference in the cross-section or the presence of a discontinuous AIS lesion (“skip lesion”). H&E, hematoxylin-eosin. Fig. S2 shows another false-negative case where CIN3 was located at the surgical margin. Although focal CIN3 was present in the frozen sections ( Fig. S2A and B ), the lesion was unclear compared to the paraffin section due to tissue folding that occurred during the frozen sectioning process ( Fig. S2C and D ). Aside from the case shown in Fig. 4 , artifacts induced by freezing or inferior staining quality on frozen section could have led to misdiagnosis. However, these limitations are inherent to frozen section analysis and are not unique to the radical trachelectomy procedure. In the 4 cases with false-negative surgical margins, 2 patients received adjuvant treatment, while the other 2 did not, because additional slices were resected at the surgeon’s discretion following the frozen section slice, and the surgical margins of the additional slices were negative on the final pathology. All patients were closely followed up without any evidence of recurrence. Fig. S3 shows a false-negative lymph-node biopsy case. The frozen section displayed cracking artifacts, but no obvious metastases ( Fig. S3A ). A micrometastasis (160 µm) was identified on the paraffin section ( Fig. S3B and C ) and confirmed with cytokeratin staining ( Fig. S3D ). The patient received adjuvant chemotherapy and remains without any signs of recurrence.

In this study, we retrospectively reviewed the frozen sections of 326 cases where RAT was planned; this represents one of the largest and longest observational studies of patients with planned RAT. We examined the pathological findings in false-negative cases and identified a case where it was difficult to distinguish between inflammatory and neoplastic changes due to differences in staining quality and artifacts in the frozen sections. Defects in the epithelium of the cervical canal and thermal distortion can complicate the assessment of surgical margins. Artifacts in frozen sections can alter the original morphology and impact diagnostic accuracy. Marginal artifacts, such as folding, tissue gaps, and irregularities in frozen sections, may hinder the microscopic evaluation of specimen margins. Benign mimics of endocervical adenocarcinoma, including tuboendometrioid metaplasia, endometriosis, and endometrium in the lower uterine segment, can be challenging to differentiate from malignancy due to their mucin-depleted glands and varying degrees of nuclear pseudo-stratification and proliferation [ 21 22 ]. Several sectioning protocols for surgical margin assessment in radical trachelectomy have been reported ( Table S3 ). To date, this study presents the largest reported case series. The sectioning protocol described by Dargent et al. [ 23 ] using sectioning perpendicular to the cervical canal, is the most commonly used method. Zhang et al. [ 20 ] reported a similar protocol, which closely aligns with our approach. Another protocol involves dividing the proximal 1-cm segment into 10–12 longitudinal sections for frozen section examination [ 17 ]. While longitudinal sectioning allows measurement of the marginal distance from the tumor, it requires identifying the direction in which the tumor extends most proximally, which is both time-consuming and resource-intensive intraoperatively. Park et al. [ 18 ] and Tanguay et al. [ 19 ] proposed protocols that combine longitudinal and transverse sections to assess marginal distance, helping to identify the tumor's location and spread. We encountered a suspected skip lesion of AIS at the surgical margin, where the transverse-sectioning protocol may have missed discontinuous lesions; longitudinal sectioning could address this issue. While longitudinal sectioning may be considered to reduce the risk of missing skip lesions, particularly in adenocarcinoma, the rarity of skip lesions and the additional time and cost required for frozen section do not strongly support its routine use. In our protocol, surgeons resect a slice for frozen section examination at the surgical margin, followed by the resection of an additional 5 mm or more to ensure marginal clearance. If the frozen section is positive, a more proximal margin of the remaining corpus is submitted for frozen section, and another resection is performed to create a safety margin. In other reported protocols, the entire resected cervix is submitted, and pathologists section a transverse slice 5–10 mm away from the proximal end [ 18 19 20 ]. Tanguay et al. [ 19 ] reported a case where the marginal distance was 12 mm on the frozen section, but the tumor was located less than 5 mm from the endocervical margin on paraffin-embedded sections, likely due to tissue retraction after fixation. This discrepancy is specific to evaluation methods using longitudinal sections. In reports by Tanguay et al. [ 19 ] and Chênevert et al. [ 16 ], frozen sections were omitted in cases without gross lesions, and they cautioned that microscopic lesions might be found in the final pathology in such cases. Relying solely on gross inspection is too subjective and increases the risk of false negatives. It remains challenging to determine which sectioning protocol is more accurate due to the small number of cases in each report. The diagnoses from frozen and permanent sections of lymph nodes were well-aligned in our study. Compared to results of PLND, frozen section of grossly enlarged lymph nodes detected only 47.6% (95% CI=28.3%–67.6%) of cases with lymph-node metastases. This finding suggests that the biopsy of grossly enlarged lymph nodes may carry a risk of sampling errors. Tu et al. [ 24 ], in a meta-analysis of 31 studies (1,887 patients with early-stage cervical cancer), reported that the pooled sensitivity of frozen section confirmed by permanent section diagnosis for SLN biopsy was 77% (66%–85%). When compared to the pooled sensitivity of studies on SLN biopsy, biopsies of grossly enlarged lymph nodes showed lower sensitivity, although without statistical significance (odds ratio=0.28, p=0.246). The results varied by sectioning protocol; the pooled sensitivity for frozen section using a protocol that bisected SLNs along the longitudinal axis (13 studies, 650 patients) was 59% (46%–72%), compared to 86% (79%–91%) with a protocol that sectioned along the short axis at short intervals (2–5 mm) (13 studies, 1,047 patients). When only macrometastases (>2 mm) were considered, pooled sensitivities were 97% and 86% for the longitudinal and perpendicular sections, respectively. Although bisecting SLNs along the longitudinal axis reduced the detection rate of micrometastases, the detection rate for macrometastases was comparable between the 2 protocols. Several studies suggest that missing micrometastases in SLNs on frozen sections does not impact oncological outcomes [ 25 26 27 28 29 ]. In Japan, since insurance coverage for technetium phytate as a radioisotope tracer for SLN identification in cervical cancer began in March 2023, SLN biopsy is expected to become more widely adopted. In frozen sections, original tissue morphology can change during fixation, and staining quality may deteriorate compared to that in paraffin sections. Artifacts, such as tissue folding, stretching, and wrinkles, can occur in frozen sections. Additionally, the quality of H&E staining in frozen sections is generally inferior to that in paraffin sections. Differences in staining quality can cause variations in nuclear appearance between frozen and paraffin-embedded samples [ 30 31 ]. Frozen sections are typically thicker than paraffin sections, which can affect staining quality and make it difficult to clearly visualize nuclear details. Pathologists should consider these differences when evaluating frozen sections, as they can make diagnosis more challenging. Limitations of this retrospective study include that the diagnostic accuracy presented is based on the sectioning protocol used at our institution. It is important to note that diagnostic accuracy may vary depending on the protocol, and further case accumulation is needed to compare the diagnostic accuracy of different protocols and identify the most appropriate method. Additionally, diagnostic accuracy may be influenced by the proportion of cases with lesions in the removed cervix. In this study, 60% of patients had no residual lesions after conization, a percentage similar to or higher than that in other studies. In conclusion, this study reviewed the performance of intraoperative frozen sections in one of the largest series of patients planned to undergo RAT. The accuracy of frozen sections in RAT is acceptable, making it a reliable tool for ensuring the oncological outcomes of RAT. However, freezing artifacts and differences in staining quality can lead to incongruent diagnoses on frozen sections compared to paraffin sections. Recognizing the pitfalls of frozen section examination can enhance diagnostic accuracy.

We performed RAT following preoperative criteria at our institution, a tertiary referral center. We reviewed pathology reports and obtained clinical information from the medical records of patients who underwent RAT or radical abdominal hysterectomy (RAH) between December 1, 2002, and December 31, 2021. Furthermore, we retrospectively assessed the diagnostic performance of frozen section of surgical margin and lymph-node biopsy. The inclusion criteria for RAT were as follows: (1) a strong desire to preserve future fertility; (2) International Federation of Gynecology and Obstetrics (FIGO) 2008 stage IA1 with lymphovascular space involvement (LVSI) to stage IB1 with tumor size of ≤2 cm; (3) histological confirmation of squamous cell carcinoma (SCC), adeno- or adenosquamous carcinoma; (4) tumor confined to the cervix without evidence of metastasis, as confirmed by radiologic studies; (5) sufficient estimated remaining uterine cavity length following surgery; and (6) age <45 years [ 13 ]. Patients were evenly divided into 4 groups (groups 1–4) based on the date of surgery to compare the sensitivity over periods. This study was approved by Institutional Review Board of Keio University School of Medicine (registration numbers 20030107 and 20110275), and the requirement for informed consent was waived owing to the retrospective nature of the study. After partial resection of the cervix, an additional thin slice was excised from the cut surface of the uterus and sent for frozen section analysis to evaluate whether the resection margins were tumor-free. The proximal surgical margin was assessed by confirming the absence of residual tumor on the frozen section, where the entire circumferential cervical canal was visible. To maintain the original orientation, the vaginal side of the specimen was marked with sutures immediately after resection. After confirming that the frozen section was negative, additional sections of ≥5 mm were obtained to ensure a safe margin. The length of the uterine cavity was routinely measured using a uterine sound to confirm that the remaining length was at least 50 mm ( Fig. 1 ). During systematic pelvic lymph-node dissection (PLND), lymph-node biopsies were performed if grossly enlarged lymph nodes were detected, and these were submitted for intraoperative diagnosis. For frozen section analysis, lymph-node specimens were bisected along the longitudinal axis. Immunohistochemistry was not routinely performed. We reviewed all surgical cases and identified false-negative cases. Two experienced gynecologic pathologists (U.A. and O.H.) retrospectively reviewed both frozen and paraffin-embedded specimens. We also compared the frozen section results from lymph-node biopsies with the final pathology results of systematic PLND. Statistical analyses were performed using SPSS (version 29.0; SPSS Inc., Chicago, IL, USA) and R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was set at p<0.05. Continuous variables are expressed as medians, and statistical significance was determined using the Mann–Whitney U test. Qualitative variables were analyzed using the χ 2 test and Fisher’s exact test. To compare the sensitivity of biopsy for grossly enlarged lymph nodes with that of sentinel lymph-node (SLN) biopsy reported in the literature, a generalized linear mixed model with a binomial distribution was used, incorporating random effects for each study; statistical differences were assessed using likelihood ratio testing.