Accuracy of Frozen Section Biopsy in the Diagnosis of Endometrial Cancer: A Systematic Review and Meta-Analysis

Simple Summary Endometrial cancer continues to be among the most prevalent cancers affecting the female reproductive system. The timely and precise diagnosis of endometrial cancer is crucial for the survival of individuals affected by it. The aim of the present systematic review and meta-analysis was to consolidate and assess the findings concerning the diagnostic precision of frozen section analysis of endometrial tissues for diagnosing endometrial cancer and atypical hyperplasia. The method has demonstrated high reliability in diagnosing endometrial cancer and even greater accuracy for atypical hyperplasia. This could significantly influence clinical practice, as a less invasive procedure such as frozen section biopsy could benefit a substantial number of women. Abstract The early and accurate diagnosis of endometrial cancer is of paramount importance for the survival of these patients. The aim of this study was to systematically appraise the available data regarding the accuracy of frozen section biopsy in diagnosing endometrial cancer. A thorough literature search was performed in PubMed/Medline, Scopus and the Cochrane Central Register of Controlled Trials databases from inception up to January 2023, with the use of specific, relevant key terms. A quality evaluation for each study was performed with the QUADAS-2 tool, whereas a bivariate random-effect model was performed to generate a summary receiver-operated curve. Heterogeneity was evaluated with Cochrane Q and Higgins’ I2 statistics. Subgroup analyses were performed for studies focused on atypical hyperplasia and those focused on endometrial cancer. The search yielded 47 studies, involving 7790 patients with endometrial cancer. Among them, only 11 could be included in the quantitative analysis. QUADAS-2 evaluation resulted in rather high quality among the included studies. Quantitative synthesis resulted in a pooled sensitivity of 0.863 and pooled specificity of 0.916. The AUC was 0.948, the Q statistic was 10.488 (10 df, p = 0.399) and Higgins’ I2 (4.655%) reported no significant heterogeneity. Subgroup analyses based on the diagnosis revealed a pooled sensitivity 0.886, specificity 0.862 and AUC 0.934 for endometrial cancer versus a sensitivity of 0.816, specificity of 0.962 and AUC 0.939 for atypical hyperplasia. Frozen section appears as a valid and reliable diagnostic tool for endometrial cancer. Its reliability seems to be even higher for the diagnosis of atypical hyperplasia. Therefore, this method may be considered in clinical practice and in settings with appropriate resources.


Introduction
Endometrial cancer is one of the most common cancer types affecting women in both high-and low/middle-income countries [1].The disease burden of endometrial cancer appears to be increasing as annual rates have increased from 0.58 to 0.89% between 1990 and 2017 worldwide [1].The widely accepted gold standard treatment for endometrial cancer is a comprehensive procedure known as total hysterectomy combined with bilateral salpingo-oophorectomy.This surgery aims to remove the uterus along with the fallopian tubes and ovaries [2].Various surgical approaches can be employed, including abdominal, laparoscopic, or robotic methods.It is noteworthy that minimally invasive techniques such as laparoscopic and robotic surgeries have been found to be equally effective as abdominal hysterectomy, offering patients potentially quicker recovery times and reduced postoperative discomfort [3].In addition, lymphadenectomy, with pelvic/paraaortic lymph nodes removal, may be performed, according to the stage of the disease [2].However, the extent and its therapeutic benefits on survival rates are still under investigation, especially in early cancer stages; some clinicians favor the removal of lymph nodes from all patients, whereas others choose to proceed with this procedure only in selected cases [4,5].Sentinel lymph biopsy may be a safe alternative to systematic lymphadenectomy [6].
In most patients, tumor grade and histotype are determined preoperatively via endometrial curettage or biopsy.However, pathological examination during surgery demonstrates high sensitivity and specificity and may play a crucial role in surgical decisions [7,8].In particular, relevant research has shown that frozen sections of endometrial tissues are highly predictive of the final diagnosis and the degree of surgical invasion required for each patient [9].
The aim of the present study was to synthesize and quantify findings regarding the diagnostic accuracy of endometrial tissues' frozen section for the diagnosis of endometrial cancer.

Materials and Methods
This systematic review and meta-analysis was carried out in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines [9].The study protocol was registered with PROSPERO international prospective register of systematic reviews (protocol number: CRD42023389536).

Search Strategy
Two independent researchers (S.K. and A.A.) separately reviewed online databases, including PubMed/Medline, Scopus and the Cochrane Central Register of Controlled Trials (CENTRAL), from inception to 26 January 2023, searching for studies examining the diagnostic accuracy of frozen section for endometrial cancer.All references were inserted in a reference manager tool (Zotero) to identify and remove any duplicate studies.The initial search was performed by screening the title and abstract of each study, followed by a full text review by two independent researchers.Any disagreements were resolved by a third investigator (I.T.).
The literature search was performed using the following keywords: "frozen section, biopsy, endometrial cancer, endometrial malignancy, endometrial tumor".The reference lists of each study included in this review were also carefully examined to identify potentially relevant papers that were not traced during the initial search.
With respect to the inclusion criteria, a study had to be published in English and to evaluate frozen section for the diagnosis of endometrial cancer compared to other diagnostic methods.A study was excluded when a different type of cancer was investigated, if the study aimed to evaluate the depth of invasion, or in cases where the provided data were insufficient.Moreover, a study that included both endometrial cancer and atypical hyperplasia, could be included in the review only in the case where it presented separate results regarding the two pathologies.Exclusion criteria related to the year of publication were not applied.Furthermore, the electronic registry of systematic reviews was examined to identify any previous meta-analyses on the topic.

Data Extraction
Data extraction was performed by using a standard predefined data form created in a datasheet file.Data collected included the first author, year of publication, journal, origin, as well as the basic characteristics of each study sample, data related to the frozen section and final biopsy procedure, the key findings of each study and any additional information necessary to assess the quality of the studies.In addition, the number of true positives, true negatives, false positives and false negatives was retrieved, and 2 × 2 tables were created for each study.

Quality Evaluation
A quality evaluation of each study was performed with the QUADAS-2 tool.QUADAS-2 is a well-known tool used for systematic reviews in order to assess risk of bias and applicability in primary research of diagnostic accuracy.QUADAS-2 consists of four main areas: sample selection, diagnostic criteria, reporting method and flow and timing.Each was assessed for risk of bias, with the first three also used to evaluate risk related to study implementation [10].

Quantitative Synthesis and Meta-Analysis
A bivariate random-effect meta-analysis was conducted according to the method described by Reitsma et al. [11].This generated a summary receiver operating characteristic (SROC) with a calculated area under the curve (AUC) and a summary estimate of sensitivity and specificity, with confidence intervals (CI) creating a 95% confidence region ellipse on the SROC.A diagnostic odds ratio (DOR) was also calculated.The heterogeneity was evaluated with Cochrane Q and Higgins' I 2 statistics.Spearman's correlation analysis between sensitivity and false positive rate was also calculated to consider threshold effect (r ≥ 0.6 generally indicates considerable threshold effect).A subgroup analysis based on the type of diagnosis (endometrial cancer or atypical hyperplasia) was also performed.All analyses were conducted in R using the mada package [12].

Study Selection
The initial search yielded 885 articles.After excluding duplicates, 574 articles remained.Subsequently, each title and abstract were screened, and 478 articles were excluded as they were either irrelevant to the topic or were published in other languages other than English.The 96 articles were retrieved as full-text and the eligibility criteria were applied by the researchers.The final step of the literature search yielded 47 studies conducted between 1993 and 2022 (the coefficient of agreement between the two reviewers was Cohen's k: 0.911).The study selection process is presented in Figure 1.
Most of the studies were retrospective cross-sectional studies that used data from medical records.In total, the 47 studies included 8353 patients; 41 of these studies included only patients with endometrial cancer (7790 cases), 7 of them encompassed 563 patients and included only cases of atypical hyperplasia, and 1 study reported separate results for the two diagnoses.Overall, only 11 studies were eligible for quantitative synthesis.Frozen section pathology was compared with the final pathology report.Regarding the surgical techniques, laparoscopic or abdominal total hysterectomy was performed in all cases.Frozen section was used to intra-operatively estimate the depth of myometrial invasion and decide subsequently whether to proceed to lymphadenectomy.The main characteristics of study groups are presented in detail in Tables 1 and 2. Most of the studies were retrospective cross-sectional studies that used data from medical records.In total, the 47 studies included 8353 patients; 41 of these studies included only patients with endometrial cancer (7790 cases), 7 of them encompassed 563 patients and included only cases of atypical hyperplasia, and 1 study reported separate results for the two diagnoses.Overall, only 11 studies were eligible for quantitative synthesis.Frozen section pathology was compared with the final pathology report.Regarding the surgical techniques, laparoscopic or abdominal total hysterectomy was performed in all cases.Frozen section was used to intra-operatively estimate the depth of myometrial invasion and decide subsequently whether to proceed to lymphadenectomy.The main characteristics of study groups are presented in detail in Tables 1 and 2.

Quality Evaluation
As already mentioned, the evaluation of the studies was carried out by applying the QUADAS-2 tool.The results of the evaluation revealed that the overall quality of the included studies was rather high, as the only issues raised included the study design and the sampling procedure, which was not randomized.The results of the evaluation are presented in Supplementary Table S1.

Quantitative Analysis
Among the included studies, 11 provided adequate data to be included in the statistical analysis.The pooled sensitivity was 0.863 (range 0.768-0.923),the pooled specificity was 0.916 (range 0.822-0.963),the overall DOR was 77.2 and the AUC was 0.948.Figure 2 shows the paired forest plot for the data.Figure 3 shows the confidence interval regions for the estimates of the primary studies, while Figure 4 presents the forest plot of the meta-analysis, using the DOR and the SROC curve.The diagnostic accuracy ratios for each of the included studies are presented in Figure 5.            Regarding heterogeneity, the results (Cochran's Q: 10.488, 10 df, p = 0.399 and Higgins' I 2 : 4.655%) revealed no significant heterogeneity, whereas the correlation analysis between sensitivity and false positive rate (rho: 0.306) indicated a low threshold effect possibility.
The subgroup analysis focused on endometrial cancer showed a pooled sensitivity of 0.886 (range 0.739 to 0.955) and a specificity of 0.862 with a range 0.055-0.695.The overall DOR was 54.9.The AUC was calculated to be 0.934.Cochran's Q test was 6.212 (6 df, p = 0.4) and Higgin's I 2 was 3.42%.Moreover, the analysis for the diagnosis of atypical Regarding heterogeneity, the results (Cochran's Q: 10.488, 10 df, p = 0.399 and Higgins' I 2 : 4.655%) revealed no significant heterogeneity, whereas the correlation analysis between sensitivity and false positive rate (rho: 0.306) indicated a low threshold effect possibility.
The subgroup analysis focused on endometrial cancer showed a pooled sensitivity of 0.886 (range 0.739 to 0.955) and a specificity of 0.862 with a range 0.055-0.695.The overall DOR was 54.9.The AUC was calculated to be 0.934.Cochran's Q test was 6.212 (6 df, p = 0.4) and Higgin's I 2 was 3.42%.Moreover, the analysis for the diagnosis of atypical hyperplasia resulted in a pooled sensitivity of 0.816 (range 0.744-0.872)and a pooled specificity of 0.962 (range 0.987-0.895).The overall DOR was 109, the AUC was 0.939, Cochran's Q test was 2.942 (3 df, p = 0.401) and Higgin's I 2 was 0%.

Discussion
According to the findings of the present study, frozen section shows high accuracy in the diagnosis of endometrial cancer.Additionally, with regard to atypical hyperplasia, our results revealed similar reliability for the examined diagnostic method.The finding of significantly high sensitivity and specificity in frozen section biopsy for the identification of both endometrial cancer and atypical hyperplasia not only underscores the validity of this diagnostic method but also highlights its clinical relevance.This enhanced sensitivity and specificity suggest that frozen section biopsy is a promising, reliable tool for accurately diagnosing hyperplasia or cancer within endometrial tissues.
Furthermore, the close approximation of the calculated area under the curve (AUC) values to 1 further reinforces the robustness and precision of frozen section biopsy.The AUC values, which serve as a quantitative measure of the diagnostic test's discriminatory power, approach 1, which indicates an exceptionally high level of accuracy in distinguishing between affected and unaffected individuals.
These findings collectively emphasize the diagnostic superiority and clinical utility of frozen section biopsy in the context of endometrial cancer and atypical hyperplasia detection.As such, this method holds significant potential for informing clinical decisionmaking processes and facilitating timely and accurate interventions for patients with these conditions.Moreover, intraoperative evaluation has been shown to be less time-consuming and is of low cost [59].The necessity of intraoperative pathologist consultation (IC) is vital to assist in directing immediate surgical decisions.Such consultations offer surgeons vital insights that can potentially influence the course of a surgical procedure, including the possibility of modification or cessation.Frozen sections conducted during ICs serve various purposes such as defining the characteristics and scope of a lesion, assessing surgical margins, and ensuring that tissue sampling for subsequent investigations is adequate.Of note, other diagnostic methods should not be neglected, as they possess specific advantages.As an example, serum biomarkers (i.e., CA 125) have the capacity to identify endometrial cancer in patients experiencing abnormal uterine bleeding [60].Nevertheless, the fact that frozen section is more accurate with a lower probability of yielding false-negative cases, an element that was also confirmed by the correlation test performed in our study, should be mentioned as one of its strongest points.
The early stages of endometrial cancer refer to those circumstances where cancer is confined within the uterus, not affecting adjacent tissues and organs.The standard clinical practice for treating stage I endometrial cancer is surgery, including the removal of the uterus, fallopian tubes and ovaries, as well as any nearby lymph nodes.According to the relevant literature, the five-year survival rate for stage IA is about 90%, with the prognosis being excellent and the cancer being highly curable with the surgical procedure alone [61].Likewise, survival rates for IB stage demonstrate similar percentages [62].Even though the prognosis of this stage remains high, the risk of a relapse is present, and in these cases, women may have to undergo additional treatment, such as radiation or chemotherapy.
An interesting point that strengthens the results of this meta-analysis is the design of the majority of the studies.In particular, all the included studies were observational; well-designed observational studies, whether retrospective or prospective, can provide valuable insights and evidence, especially in situations where randomized controlled trials are not practical or ethical.Furthermore, a strong feature of the present work is the identification of low heterogeneity between studies, a fact that facilitates and strengthens the generalizability of the results [63].
A strong feature of this meta-analysis is that, to our knowledge, no similar study has been carried out so far.On the other hand, this study bears certain limitations that need to be considered and addressed.The studies of this meta-analysis have included samples without calculating the required number of participants to obtain statistically significant results that can be generalized.In addition, the fact that different diagnostic techniques have been compared with endometrial frozen sections (e.g., transvaginal ultrasound, MRI) should not be neglected when synthesizing the findings.

Conclusions
The results of the present systematic review and meta-analysis showed that frozen section biopsy is a valid and reliable diagnostic tool for the diagnosis of endometrial cancer and atypical hyperplasia.In addition, it could provide information about tumor grade, which may have an important impact on clinical decision-making.Although other methods should not be overlooked, since they can be useful in clinical environments that do not have the possibility of applying other diagnostic methods, the revealing reliability of frozen section biopsy is unquestionable.In this regard, the clinical implementation of this method could enhance the quality of life and the level of care provided to women undergoing surgery.

Figure 1 .
Figure 1.Flow chart of the included studies.

Figure 1 .
Figure 1.Flow chart of the included studies.

Figure 2 .
Figure 2. Sensitivity and specificity results for each study are presented.

Figure 3 .
Figure 3.The ROC (receiver operating characteristic) curve is a graphical representation used to assess the performance of a classification model or a diagnostic test.It illustrates the trade-off between sensitivity (true positive rate) and specificity (true negative rate) across various thresholds.A perfect classifier would have an ROC curve that passes through the upper-left corner of the plot, indicating 100% sensitivity and 100% specificity.

Figure 3 .
Figure 3.The ROC (receiver operating characteristic) curve is a graphical representation used to assess the performance of a classification model or a diagnostic test.It illustrates the trade-off between sensitivity (true positive rate) and specificity (true negative rate) across various thresholds.A perfect classifier would have an ROC curve that passes through the upper-left corner of the plot, indicating 100% sensitivity and 100% specificity.

Figure 4 .
Figure 4. Plots with confidence regions for primary study estimates.

Figure 4 .
Figure 4. Plots with confidence regions for primary study estimates.

Figure 4 .Figure 5 .
Figure 4. Plots with confidence regions for primary study estimates.

Table 1 .
Characteristics of the studies focusing on endometrial cancer that were included in this systematic review.

Table 1 .
Characteristics of the studies focusing on endometrial cancer that were included in this systematic review.

Table 2 .
Characteristics of the studies focusing on atypical hyperplasia that were included in this systematic review.