Complete surgical removal of a malignant tumour with histologically negative margins is an important determinant of local disease control, recurrence and survival. A positive surgical margin indicates the presence of tumour cells at or close to the resection boundary and may necessitate additional excision, wider surgery, adjuvant treatment or a second operative procedure. Intraoperative assessment of tumour margins therefore provides an opportunity to identify residual malignancy and modify the extent of surgery during the same operative session. Frozen-section examination is among the most widely used techniques for this purpose because it provides a rapid histopathological diagnosis while the patient remains under anaesthesia. During frozen-section examination, fresh tissue is rapidly frozen, cut in a cryostat, stained and examined microscopically. The technique can provide information about margin involvement within approximately 15-30 minutes and may reduce the frequency of repeat operations. Frozen sections are employed in the surgical management of tumours involving the breast, oral cavity, head and neck, thyroid, ovary, pancreas and other organs. A systematic review of intraoperative margin-assessment techniques in breast cancer demonstrated that frozen section was among the established methods with favourable diagnostic performance, although its use was limited by turnaround time and resource requirements [1]. Similarly, a meta-analysis of intraoperative margin assessment in head-and-neck squamous cell carcinoma reported pooled sensitivity and specificity of approximately 79.8% and 99.1%, respectively, for frozen-section examination [2]. Despite its clinical utility, frozen-section diagnosis is affected by several technical and interpretative limitations. Freezing artefacts, tissue folding, thick sections, cautery-related changes, fatty or calcified tissue, small tumour foci and inadequate sampling may interfere with microscopic interpretation. Sampling error is particularly important because only selected portions of a potentially extensive surgical margin can be examined intraoperatively. Pancreatic frozen-section studies have demonstrated low overall discrepancy rates, but diagnostic deferrals and occasional errors remain unavoidable [3]. The performance of frozen-section assessment may also differ according to tumour site, histological type, margin characteristics, specimen-orientation practices and the experience of the reporting pathologist. Formalin-fixed, paraffin-embedded permanent-section histopathology provides superior morphology and permits more extensive sampling; hence, it is considered the reference standard for final margin evaluation. However, the result becomes available only after completion of surgery. Although frozen sections can show high specificity, their sensitivity may be lower in some anatomical sites, resulting primarily in false-negative diagnoses [4]. Discrepancies may be attributable to sampling, technical or interpretative errors. Moreover, an apparently negative revised frozen margin does not invariably eliminate the risk associated with an initially involved margin [5]. Regular comparison of frozen-section findings with permanent-section diagnoses is therefore an essential component of pathology quality assurance. AIM To evaluate the diagnostic accuracy of intraoperative frozen-section examination for tumour-margin assessment using permanent-section histopathology as the reference standard. OBJECTIVES 1. To determine the sensitivity, specificity, positive predictive value, negative predictive value and overall accuracy of frozen-section examination in detecting tumour-positive surgical margins. 2. To assess the concordance between frozen-section and permanent-section diagnoses and determine the frequencies of false-positive, false-negative and deferred diagnoses. 3. To identify the tumour-related, specimen-related, technical and interpretative factors associated with discordance between frozen-section and permanent-section findings.

Source of data The study data were obtained from patients who underwent surgical excision of a suspected or histologically confirmed tumour and for whom intraoperative frozen-section examination of one or more surgical margins was requested. The corresponding permanent-section histopathology reports, operation records, requisition forms, frozen-section records and pathology registers were reviewed. Each eligible case had paired observations consisting of the intraoperative frozen-section diagnosis and the final permanent-section diagnosis of the corresponding sampled margin. Study design This was a hospital-based observational diagnostic-accuracy study in which the findings of intraoperative frozen-section examination the index test were compared with formalin-fixed, paraffin-embedded permanent-section histopathology, which was considered the reference standard. The study followed a paired diagnostic design because both frozen-section and permanent-section examinations were performed on corresponding tissue from the same surgical margin. Study location The study was conducted in the Department of Pathology in collaboration with the surgical departments of Vedanta Institute of Medical Sciences, Dahanu, Maharashtra, India. Study duration The study was conducted over two years, from January 2024 to December 2025, including case identification, specimen processing, retrieval of permanent-section results, data collection and statistical analysis. Study population The study population consisted of patients undergoing surgery for malignant or potentially malignant tumours in whom intraoperative frozen-section assessment of surgical margins was performed. Sample size A total of 200 eligible cases with paired frozen-section and permanent-section findings were included. If multiple margins were received from the same patient, the patient constituted the primary study unit, while individual margins were additionally recorded for margin-level analysis. The final sample comprised consecutive eligible cases available during the defined study period. All cases satisfying the eligibility criteria were included until the required sample size of 200 was reached. Sampling technique A consecutive sampling method was used. Every eligible case received during the study period was included sequentially until the predetermined sample size was achieved. This approach minimized arbitrary case selection and reflected routine institutional frozen-section practice. Inclusion criteria Patients of any adult age and either sex who underwent surgical excision of a suspected or confirmed tumour. Cases in which intraoperative frozen-section examination was specifically requested for tumour-margin assessment. Cases with an unequivocal frozen-section diagnosis of positive or negative margin, or a documented deferred diagnosis. Cases in which the corresponding tissue was subsequently processed for permanent-section examination. Cases for which the frozen-section report, permanent-section report and relevant clinical and operative information were available. Primary or recurrent tumours from any anatomical site for which margin assessment was clinically indicated. Exclusion criteria Cases in which frozen section was performed only to establish tumour diagnosis, determine tumour type or assess lymph nodes without evaluation of a surgical margin. Specimens without corresponding permanent-section material. Cases with incomplete, unavailable or damaged frozen-section or permanent-section records. Specimens that were inadequately labelled or for which orientation of the surgical margin could not be confirmed. Tissues unsuitable for frozen section because of extensive calcification, ossification or severe tissue autolysis. Non-neoplastic lesions and cases in which no tumour-margin assessment was requested. Repeat submission of the same margin unless it represented a separately labelled revised margin. Cases in which the final diagnosis remained inconclusive even after permanent-section examination. Procedure and methodology Before commencement, approval was obtained from the Institutional Ethics Committee. Permission was also obtained from the Department of Pathology and the concerned surgical departments. Patient confidentiality was maintained by assigning a unique study identification number to each case. The clinical diagnosis, anatomical site, type of operation, tumour type, number and location of submitted margins, surgeon’s orientation and any previous treatment were recorded. Each margin was identified according to the label provided by the surgeon, such as superior, inferior, medial, lateral, anterior, posterior, proximal, distal, mucosal or deep margin. Fresh tissue intended for frozen-section examination was transported immediately to the pathology laboratory in a clean, properly labelled container without formalin. The specimen was verified against the requisition form, and its orientation was confirmed with the operating surgeon whenever required. Gross examination was performed, and the size, appearance and orientation of the tissue were documented. Representative tissue was selected from the margin considered most likely to demonstrate residual tumour. When multiple separately labelled margins were submitted, each margin was examined and reported separately. The pathologist categorized the frozen-section result as: Positive margin: malignant tumour cells were present at the examined resection margin. Negative margin: no malignant tumour cells were identified at the examined margin. Deferred or indeterminate: a definitive opinion could not be given because of inadequate tissue, freezing artefact, suspicious morphology or another technical limitation. The frozen-section diagnosis was communicated immediately to the operating surgeon. Any additional tissue or revised margin submitted after a positive or suspicious result was processed in the same manner and documented separately. After frozen-section reporting, the remaining tissue and frozen-tissue block were fixed in 10% neutral-buffered formalin and processed for permanent sections. The frozen-section and permanent-section findings were subsequently compared. Any discrepancy was reviewed by at least two pathologists, wherever feasible, to determine its probable cause. Discordant cases were classified as being associated with: Sampling error: the tumour focus was absent from the tissue examined during frozen section but identified on deeper or additional permanent sections. Technical error: poor section quality, folding, crushing, freezing artefact or cautery artefact interfered with interpretation. Interpretative error: the relevant abnormality was present on the frozen slide but was misinterpreted. Communication or orientation error: the discrepancy resulted from incorrect identification, orientation or communication of the sampled margin. Unclassified error: the exact cause could not be established. Sample processing Fresh tissue selected for frozen section was placed on a metal chuck using an optimum cutting temperature embedding medium. The tissue was rapidly frozen in a cryostat maintained at approximately −20°C to −25°C. Sections measuring approximately 4-7 µm in thickness were cut, transferred to glass slides and rapidly stained with haematoxylin and eosin. The stained frozen-section slides were examined microscopically by the reporting pathologist. Deeper frozen sections were prepared when the initial section was inadequate or suspicious, provided sufficient tissue and operative time were available. Following intraoperative examination, the frozen tissue and remaining specimen were fixed in 10% neutral-buffered formalin for approximately 6-24 hours, depending on tissue size. The tissue was processed through graded alcohol, cleared in xylene, embedded in paraffin wax and sectioned at approximately 3-5 µm thickness. Sections were stained with routine haematoxylin and eosin. Additional levels and ancillary stains were performed when required. The permanent-section diagnosis was considered the final reference diagnosis. The margin was classified according to the institutional reporting protocol as involved or uninvolved by tumour. Where relevant, the measured distance of the tumour from the closest margin was also recorded. Data collection Data were collected using a structured case-record form. The following variables were recorded: Study identification number Age and sex Clinical diagnosis Anatomical site and organ involved Primary or recurrent tumour Type of surgical procedure Histological type and grade of tumour Neoadjuvant therapy, where applicable Number and type of margins examined Size and condition of the frozen-section tissue Presence of cautery, freezing or crushing artefacts Frozen-section diagnosis Turnaround time for the frozen-section report Intraoperative action taken following the report Permanent-section diagnosis Margin status and distance from the margin Concordance or discordance False-positive or false-negative result Deferred diagnosis Probable cause of discrepancy The completed forms were checked for completeness and consistency before data entry. Personal identifiers were removed from the analytical dataset. Operational definitions A true-positive result was defined as a margin reported positive on frozen section and confirmed positive on permanent section. A true-negative result was defined as a margin reported negative on frozen section and confirmed negative on permanent section. A false-positive result was defined as a margin reported positive on frozen section but found negative on permanent section. A false-negative result was defined as a margin reported negative on frozen section but found positive on permanent section. Concordance referred to agreement between frozen-section and permanent-section diagnoses, while discordance referred to disagreement between the two methods. Statistical methods Data were entered into Microsoft Excel and analysed using an appropriate statistical software package such as IBM SPSS Statistics. Continuous variables were summarized using mean and standard deviation when normally distributed and median with interquartile range when skewed. Categorical variables were presented as frequencies and percentages. A 2×2diagnostic table was prepared by comparing frozen-section findings with permanent-section results. The following diagnostic measures were calculated with their 95% confidence intervals: "Sensitivity"=TP/(TP+FN)×100 "Specificity"=TN/(TN+FP)×100 "Positive predictive value"=TP/(TP+FP)×100 "Negative predictive value"=TN/(TN+FN)×100 "Diagnostic accuracy"=(TP+TN)/(TP+TN+FP+FN)×100 Positive and negative likelihood ratios were also calculated. Agreement between frozen-section and permanent-section diagnoses was assessed using Cohen’s kappa coefficient. Kappa values were interpreted as poor (<0.20), fair (0.21-0.40), moderate (0.41-0.60), substantial (0.61-0.80) and almost perfect (>0.80) agreement. McNemar’s test was used to examine differences in paired positive and negative classifications. The chi-square test or Fisher’s exact test was used to evaluate associations between discordance and categorical variables. An independent-samples t-test or Mann-Whitney U test was used for continuous variables, depending on their distribution. Where an adequate number of discordant outcomes was available, binary logistic-regression analysis was performed to identify independent predictors of diagnostic discordance. Adjusted odds ratios with 95% confidence intervals were reported. All statistical tests were two-tailed, and a p value <0.05 was considered statistically significant.

Table 1: Clinicopathological characteristics and permanent-section margin status (N=200) Study parameter Overall (N=200) Positive margin (n=49) Negative margin (n=151) Effect estimate (95% CI) Test of significance P value Age, years, Mean (SD) 52.81 (13.72) 55.90 (12.60) 51.80 (13.90) MD: 4.10 (−0.12-8.32) Welch’s t=1.93 0.057 Male sex 107 (53.5) 30 (61.2) 77 (51.0) OR: 1.52 (0.79-2.93) Fisher’s exact test 0.250 Female sex 93 (46.5) 19 (38.8) 74 (49.0) Reference Primary tumour 169 (84.5) 36 (73.5) 133 (88.1) Reference Recurrent tumour 31 (15.5) 13 (26.5) 18 (11.9) OR: 2.67 (1.20-5.95) Fisher’s exact test 0.022 Tumour size, cm, Mean (SD) 3.98 (2.06) 4.82 (2.17) 3.71 (1.94) MD: 1.11 (0.42-1.80) Welch’s t=3.17 0.002 Tumour size ≤4 cm 122 (61.0) 22 (44.9) 100 (66.2) Reference Tumour size >4 cm 78 (39.0) 27 (55.1) 51 (33.8) OR: 2.41 (1.25-4.64) Fisher’s exact test 0.011 Low/intermediate-grade tumour 131 (65.5) 25 (51.0) 106 (70.2) Reference High-grade tumour 69 (34.5) 24 (49.0) 45 (29.8) OR: 2.26 (1.17-4.37) Fisher’s exact test 0.016 Tumour necrosis absent 154 (77.0) 31 (63.3) 123 (81.5) Reference Tumour necrosis present 46 (23.0) 18 (36.7) 28 (18.5) OR: 2.55 (1.25-5.19) Fisher’s exact test 0.011 Multiple margins submitted 91 (45.5) 26 (53.1) 65 (43.0) OR: 1.50 (0.79-2.88) Fisher’s exact test 0.246 Previous neoadjuvant treatment 37 (18.5) 12 (24.5) 25 (16.6) OR: 1.63 (0.75-3.54) Fisher’s exact test 0.225 Anatomical site χ²=4.05, df=5 0.542 Oral cavity/head and neck 57 (28.5) 19 (38.8) 38 (25.2) Breast 43 (21.5) 8 (16.3) 35 (23.2) Gastrointestinal/pancreatobiliary 36 (18.0) 9 (18.4) 27 (17.9) Gynaecological 29 (14.5) 6 (12.2) 23 (15.2) Thyroid 18 (9.0) 3 (6.1) 15 (9.9) Soft-tissue and other tumours 17 (8.5) 4 (8.2) 13 (8.6) Table 1 presents the clinicopathological characteristics of 200 cases according to permanent-section margin status. Positive margins were identified in 49 cases (24.5%), while 151 (75.5%) had negative margins. The mean age was slightly higher among patients with positive margins than among those with negative margins (55.90±12.60 versus 51.80±13.90 years); however, the mean difference of 4.10 years was not statistically significant (95% CI: −0.12-8.32; Welch’s t=1.93; p=0.057). Male sex was also not significantly associated with margin positivity (OR=1.52; 95% CI: 0.79-2.93; p=0.250). Recurrent tumours demonstrated significantly greater odds of positive margins than primary tumours (26.5% versus 11.9%; OR=2.67; 95% CI: 1.20-5.95; p=0.022). The mean tumour size was significantly greater in margin-positive cases (4.82±2.17 cm) than in margin-negative cases (3.71±1.94 cm), with a mean difference of 1.11 cm (95% CI: 0.42-1.80; p=0.002). Similarly, tumours larger than 4 cm were significantly associated with margin positivity (OR=2.41; 95% CI: 1.25-4.64; p=0.011). High-grade tumours had more than twice the odds of positive margins compared with low- or intermediate-grade tumours (OR=2.26; 95% CI: 1.17-4.37; p=0.016). Tumour necrosis was also significantly associated with margin positivity (OR=2.55; 95% CI: 1.25-5.19; p=0.011). No significant associations were observed for submission of multiple margins (p=0.246), previous neoadjuvant treatment (p=0.225), or anatomical tumour site (χ²=4.05; df=5; p=0.542). Table 2: Diagnostic performance of frozen-section examination against permanent-section histopathology Table 2A. A 2 × 2 comparison among cases with a definitive frozen-section diagnosis (n=193) Frozen-section result Permanent section positive Permanent section negative Total Frozen section positive 41, true positive 3, false positive 44 Frozen section negative 6, false negative 143, true negative 149 Total 47 146 193 Table 2B. Diagnostic-accuracy indices Diagnostic measure Calculation Estimate, % 95% CI Test of significance P value Sensitivity 41/47 87.2 74.3-95.2 Binomial test against 50% <0.001 Specificity 143/146 97.9 94.1-99.6 Binomial test against 50% <0.001 Positive predictive value 41/44 93.2 81.3-98.6 Binomial test against 50% <0.001 Negative predictive value 143/149 96.0 91.4-98.5 Binomial test against 50% <0.001 Overall diagnostic accuracy 184/193 95.3 91.3-97.8 Binomial test against 50% <0.001 Positive likelihood ratio Sensitivity/(1−specificity) 42.43 13.83-130.20 Negative likelihood ratio (1−sensitivity)/specificity 0.13 0.06-0.28 Diagnostic odds ratio TP×TN/FP×FN 325.72 76.46-1387.58 Wald test <0.001 Cohen’s kappa Agreement beyond chance 0.871 0.788-0.953 Z=20.63 <0.001 Difference between paired classifications FP=3 versus FN=6 Exact McNemar test 0.508 Table 2 demonstrates the diagnostic performance of frozen-section examination against permanent-section histopathology among the 193 cases with definitive frozen-section results. Frozen-section examination correctly identified 41 true-positive and 143 true-negative cases, while three cases were false positive and six were false negative. The sensitivity was 87.2% (95% CI: 74.3-95.2), indicating that frozen section correctly detected nearly nine out of every ten positive margins. Specificity was 97.9% (95% CI: 94.1-99.6), demonstrating excellent ability to identify negative margins. The positive predictive value was 93.2% (95% CI: 81.3-98.6), while the negative predictive value was 96.0% (95% CI: 91.4-98.5). Overall diagnostic accuracy was 95.3% (95% CI: 91.3-97.8), and all these diagnostic indices were significantly greater than 50% (p<0.001). The positive likelihood ratio was 42.43 (95% CI: 13.83-130.20), whereas the negative likelihood ratio was 0.13 (95% CI: 0.06-0.28), indicating a marked change in the probability of margin involvement following a positive or negative frozen-section result. The diagnostic odds ratio was 325.72 (95% CI: 76.46-1387.58; p<0.001). Cohen’s kappa coefficient was 0.871 (95% CI: 0.788-0.953; Z=20.63; p<0.001), demonstrating almost perfect agreement between frozen-section and permanent-section diagnoses. The exact McNemar test was not significant (p=0.508), indicating no systematic difference between the paired classifications. Table 3: Concordance, discordance and deferred diagnoses in the study cohort Diagnostic outcome Frequency Percentage 95% CI Test of significance P value Initial frozen-section result, N=200 Positive 44 22.0 16.5-28.4 Negative 149 74.5 67.9-80.4 Deferred/indeterminate 7 3.5 1.4-7.1 Binomial test versus 10% 0.003 Final permanent-section result, N=200 Positive margin 49 24.5 18.7-31.1 Negative margin 151 75.5 68.9-81.3 Overall outcome, N=200 Concordant definitive diagnosis 184 92.0 87.3-95.4 Binomial test versus 50% <0.001 Discordant definitive diagnosis 9 4.5 2.1-8.4 Deferred diagnosis 7 3.5 1.4-7.1 Outcome among definitive frozen sections, n=193 Exact diagnostic concordance 184 95.3 91.3-97.8 Cohen’s κ=0.871 <0.001 Diagnostic discordance 9 4.7 2.2-8.7 True-positive agreement 41 21.2 15.7-27.7 True-negative agreement 143 74.1 67.3-80.1 False-positive diagnosis 3 1.6 0.3-4.5 FP versus FN: exact McNemar test 0.508 False-negative diagnosis 6 3.1 1.1-6.6 Permanent results of deferred cases, n=7 Positive on permanent section 2 28.6 3.7-71.0 Negative on permanent section 5 71.4 29.0-96.3 Fisher’s exact test 0.286 Table 3 summarizes the concordance, discordance and deferred diagnoses. Of the 200 initial frozen-section examinations, 44 (22.0%; 95% CI: 16.5-28.4) were positive, 149 (74.5%; 95% CI: 67.9-80.4) were negative, and seven (3.5%; 95% CI: 1.4-7.1) were deferred or indeterminate. The observed deferral rate was significantly below the comparison value of 10% (p=0.003). On final permanent-section examination, 49 cases (24.5%; 95% CI: 18.7-31.1) had positive margins and 151 (75.5%; 95% CI: 68.9-81.3) had negative margins. In the entire cohort, 184 cases (92.0%; 95% CI: 87.3-95.4) had concordant definitive findings, nine (4.5%; 95% CI: 2.1-8.4) were discordant, and seven (3.5%) were deferred. Among the 193 cases with definitive frozen-section diagnoses, exact concordance was 95.3% (95% CI: 91.3-97.8), with a significant kappa value of 0.871 (p<0.001). This included 41 true-positive and 143 true-negative results. Nine cases were discordant, comprising three false-positive diagnoses (1.6%; 95% CI: 0.3-4.5) and six false-negative diagnoses (3.1%; 95% CI: 1.1-6.6). Although false-negative diagnoses were more frequent, the difference between false-positive and false-negative results was not statistically significant (McNemar p=0.508). Of the seven deferred cases, two were positive and five were negative on permanent-section examination; this distribution was not statistically significant (p=0.286). Table 4: Factors associated with discordance between frozen-section and permanent-section findings (n=193) Associated factor Overall (n=193) Discordant (n=9) Concordant (n=184) Odds ratio (95% CI) Test of significance P value Sampling limitation 28 (14.5) 6 (66.7) 22 (12.0) 14.73 (3.43-63.14) Fisher’s exact test <0.001 Freezing artefact 22 (11.4) 5 (55.6) 17 (9.2) 12.28 (3.01-50.11) Fisher’s exact test 0.001 Cautery artefact 23 (11.9) 4 (44.4) 19 (10.3) 6.95 (1.72-28.11) Fisher’s exact test 0.013 Small or fragmented specimen 35 (18.1) 4 (44.4) 31 (16.8) 3.95 (1.00-15.54) Fisher’s exact test 0.059 Uncommon/heterogeneous histology 24 (12.4) 3 (33.3) 21 (11.4) 3.88 (0.90-16.69) Fisher’s exact test 0.086 Multiple margins submitted 85 (44.0) 7 (77.8) 78 (42.4) 4.76 (0.96-23.52) Fisher’s exact test 0.045 Fatty, fibrotic or calcified tissue 17 (8.8) 3 (33.3) 14 (7.6) 6.07 (1.37-26.92) Fisher’s exact test 0.034 Tumour size >4 cm 74 (38.3) 6 (66.7) 68 (37.0) 3.41 (0.83-13.96) Fisher’s exact test 0.093 High-grade tumour 66 (34.2) 5 (55.6) 61 (33.2) 2.52 (0.65-9.78) Fisher’s exact test 0.279 Recurrent tumour 29 (15.0) 3 (33.3) 26 (14.1) 3.04 (0.72-12.85) Fisher’s exact test 0.144 Previous neoadjuvant treatment 34 (17.6) 3 (33.3) 31 (16.8) 2.47 (0.59-10.34) Fisher’s exact test 0.201 Margin tissue <10 mm 51 (26.4) 5 (55.6) 46 (25.0) 3.75 (0.96-14.60) Fisher’s exact test 0.056 Turnaround time, minutes, Mean (SD) 21.36 (6.42) 26.89 (7.31) 21.09 (6.27) MD: 5.80 (1.53-10.07) Welch’s t=2.67 Table 4 evaluates factors associated with discordance between frozen-section and permanent-section diagnoses among 193 cases after excluding seven deferred cases. Sampling limitation was the strongest factor associated with discordance and was present in 66.7% of discordant cases compared with 12.0% of concordant cases. It increased the odds of discordance nearly fifteen-fold (OR=14.73; 95% CI: 3.43-63.14; p<0.001). Freezing artefact was observed in 55.6% of discordant cases and 9.2% of concordant cases and was significantly associated with discordance (OR=12.28; 95% CI: 3.01-50.11; p=0.001). Cautery artefact was also a significant factor (OR=6.95; 95% CI: 1.72-28.11; p=0.013). Cases involving fatty, fibrotic or calcified tissue had approximately six times greater odds of discordance (OR=6.07; 95% CI: 1.37-26.92; p=0.034), while submission of multiple margins was associated with nearly five times greater odds (OR=4.76; 95% CI: 0.96-23.52; p=0.045). Small or fragmented specimens (p=0.059) and margin tissue measuring less than 10 mm (p=0.056) showed borderline associations with discordance. Uncommon or heterogeneous histology, tumour size greater than 4 cm, high-grade tumour, recurrent disease and previous neoadjuvant treatment were more frequent among discordant cases, but their associations did not reach statistical significance (p>0.05). The mean frozen-section turnaround time was significantly longer in discordant cases than in concordant cases (26.89±7.31 versus 21.09±6.27 minutes), with a mean difference of 5.80 minutes (95% CI: 1.53-10.07; Welch’s t=2.67; p=0.009).

The present study evaluated the diagnostic accuracy of intraoperative frozen-section examination for tumour-margin assessment using permanent-section histopathology as the reference standard. The study included 200 cases from different anatomical sites and demonstrated that frozen-section examination provided high diagnostic accuracy, almost perfect agreement with permanent sections and a relatively low frequency of deferred diagnoses. Nevertheless, sampling limitations and tissue-processing artefacts remained important causes of discordance. Clinicopathological characteristics and margin positivity Permanent-section examination identified positive tumour margins in 49 of the 200 cases, giving an overall margin-positivity rate of 24.5%. Margin positivity was not significantly associated with age or sex. The mean age was higher among margin-positive patients, but the difference was marginally nonsignificant. These findings suggest that local tumour characteristics and technical resectability may have a greater influence on margin status than basic demographic characteristics. Recurrent tumours had significantly greater odds of positive margins than primary tumours (OR=2.67; p=0.022). Recurrent tumours are often associated with fibrosis, altered anatomical planes, previous surgery or radiotherapy and infiltrative growth, all of which can make complete surgical excision difficult. Du et al. (2016) [1] emphasized that frozen-section assessment is particularly useful in anatomically complex head-and-neck resections, although discrepancies may occur when the tissue submitted does not adequately represent the true resection margin. Ettl et al. (2016) [2] further reported that positive frozen-section margins were associated with an increased risk of local recurrence, even when subsequent revision produced an apparently negative margin. Thus, identification of a positive margin may reflect aggressive tumour biology or technically difficult resection rather than an isolated microscopic finding. Tumour size was significantly associated with margin involvement. The mean tumour size was 4.82 cm in margin-positive cases compared with 3.71 cm in margin-negative cases, while tumours larger than 4 cm had approximately 2.4 times greater odds of margin positivity. Large tumours frequently have irregular or infiltrative borders and may involve multiple anatomical compartments, increasing the likelihood of incomplete excision. Long et al. (2022)[9] demonstrated that frozen-section assessment of tumour-bed margins in oral cavity squamous cell carcinoma was accurate for the sampled tissue but was less reliable in predicting final margin status of the entire surgical specimen. This distinction becomes increasingly relevant with large tumours because a small number of sampled margins may not fully represent the total resection surface. High-grade tumours and tumours showing necrosis were also significantly associated with positive margins. High-grade tumours had more than twice the odds of margin positivity, while the presence of necrosis increased the odds approximately 2.6 times. High-grade tumours are more likely to demonstrate infiltrative borders, cellular heterogeneity, satellite foci and lymphovascular or perineural extension. Necrosis may further complicate both gross identification and microscopic interpretation of the viable tumour boundary. Ramadan et al. (2025) [13] observed that positive final margins were associated with worse disease-specific survival and greater locoregional recurrence among patients with HPV-related oropharyngeal carcinoma, reinforcing the clinical significance of precise margin evaluation. Ramadan et al. study No significant association was observed between anatomical site and margin positivity. This may be attributed to the heterogeneous nature of the study population and the relatively small number of cases within each organ-specific subgroup. In addition, the definition and clinical implications of an adequate margin differ across breast, oral cavity, thyroid, gastrointestinal, gynaecological and soft-tissue tumours. Consequently, the overall results should be interpreted as reflecting institutional frozen-section performance across tumour types rather than the performance expected for any single organ. Diagnostic performance of frozen-section examination Among the 193 cases with definitive frozen-section diagnoses, the examination yielded 41 true-positive, 143 true-negative, three false-positive and six false-negative results. Frozen-section sensitivity was 87.2%, specificity was 97.9%, positive predictive value was 93.2%, negative predictive value was 96.0%, and overall diagnostic accuracy was 95.3%. These findings indicate that frozen-section examination was particularly effective in confirming negative margins while retaining a high capacity to detect involved margins. The present sensitivity and specificity closely correspond with those reported by Du et al. (2016) [1], who documented frozen-section accuracy of 96.7%, sensitivity of 83.1% and specificity of 97.9% in head-and-neck squamous cell carcinoma resections. The slightly higher sensitivity in the present study may reflect differences in case selection, tumour sites, sampling protocols and reporting thresholds. In a meta-analysis of intraoperative margin-assessment methods in head-and-neck squamous cell carcinoma, Higginson et al. (2023) [10] reported pooled sensitivity of 79.8%, specificity of 99.1%, diagnostic odds ratio of 309.8 and area under the receiver operating characteristic curve of 0.976 for frozen-section examination. The present sensitivity of 87.2% was higher, whereas the specificity of 97.9% was marginally lower. The diagnostic odds ratio of 325.72 in the present study was comparable with the pooled estimate reported by Higginson et al., supporting the strong discriminatory performance of frozen sections. Higginson et al. meta-analysis Ali et al. (2024) [11], in a study of frozen-section evaluation of oral cavity squamous cell carcinoma margins, reported sensitivity of 88.81%, specificity of 94.84%, positive predictive value of 95.20%, negative predictive value of 88.10% and diagnostic accuracy of 91.63%. The sensitivity reported by Ali et al. was nearly identical to the present value, while the present study showed higher specificity, negative predictive value and overall accuracy. Differences in tumour prevalence, anatomical site and whether margins were obtained from the specimen or tumour bed may explain these variations. Ali et al. study Demir et al. (2022) [8] reported sensitivity of 99%, specificity of 96%, positive predictive value of 93.3%, negative predictive value of 97.6% and overall accuracy of 96% for oral cavity cancers. The present study demonstrated comparable specificity, predictive values and accuracy but lower sensitivity. Demir et al. evaluated a relatively more homogeneous disease group, whereas the current study included tumours from multiple anatomical sites with different histological patterns and sampling requirements. Demir et al. study Garcia et al. (2021) [6] concluded from a systematic review and meta-analysis that frozen-section examination provided high accuracy, sensitivity and specificity during breast-conserving surgery and could reduce reoperation for positive margins. Dowling et al. (2024) [12] similarly found that frozen section remained among the most diagnostically accurate established intraoperative margin-assessment methods, although resource requirements and turnaround time limited its widespread use. Garcia et al. meta-analysis, Dowling et al. meta-analysis The positive likelihood ratio of 42.43 indicates that a positive frozen-section result markedly increased the probability of true margin involvement. Conversely, the negative likelihood ratio of 0.13 indicates that a negative result substantially reduced, but did not completely eliminate, the possibility of an involved permanent margin. This is clinically important because a negative frozen section must be interpreted in the context of sampling adequacy, orientation and the surgeon’s assessment. Cohen’s kappa coefficient was 0.871, demonstrating almost perfect agreement between frozen and permanent sections. The nonsignificant McNemar test indicated the absence of systematic directional disagreement. Nevertheless, the presence of six false-negative results indicates that permanent-section assessment remains necessary and that frozen section should not be regarded as an infallible replacement for formalin-fixed paraffin-embedded examination. Ramadan et al. (2025) [13] demonstrated this distinction clearly: frozen section achieved 97.1% accuracy, 72.2% sensitivity and 99.1% specificity for individual margins, but its sensitivity for predicting overall final surgical-margin status was only 21.7%. Thus, frozen section may accurately classify the tissue actually examined while failing to detect an involved area elsewhere along an incompletely sampled resection surface. Concordance, discordance and deferred diagnoses Of the 200 initial frozen-section examinations, 44 were positive, 149 were negative and seven were deferred. Concordant definitive results were obtained in 184 cases, while nine were discordant. Among the 193 cases with a definitive frozen-section result, concordance was 95.3% and discordance was 4.7%. Layfield et al. (2018) [4] examined 1,796 paired frozen and permanent sections from head-and-neck squamous cell carcinomas and reported 97% concordance, with discordance in 3.1% of pairs. False-negative diagnoses occurred in 1.9%, while false-positive diagnoses occurred in 1.1%. Their results are comparable with the present false-negative rate of 3.1% and false-positive rate of 1.6%, although the present overall discordance rate was slightly higher. In both studies, false-negative diagnoses were more frequent than false-positive diagnoses. Layfield et al. study Nayanar et al. (2019) [5] reported 15 discrepant frozen-section diagnoses among 265 head-and-neck cancer cases, including seven false-negative and eight false-positive findings. Their approximate discrepancy rate of 5.7% was slightly higher than the 4.7% observed in the current study. Differences may have arisen from tumour-site distribution, institutional experience, slide quality, specimen orientation and criteria used to define close or positive margins. Chavez et al. (2022) [7] evaluated 221 pancreatic frozen sections and found a specimen-level error rate of 1.8%. However, 7.2% of frozen sections were deferred, including eight margin specimens. The present deferral rate of 3.5% was lower, but its discordance rate was higher. Pancreatic tissue frequently presents specific interpretative challenges, particularly chronic pancreatitis, fibrosis and atypical ductal changes, which may encourage cautious deferral rather than immediate categorical diagnosis. Chavez et al. study Of the seven deferred cases in the present study, two were positive and five were negative on permanent-section examination. The relatively wide confidence intervals reflect the small number of deferred cases. Deferral should not automatically be regarded as diagnostic failure; it may represent appropriate recognition of inadequate tissue, severe artefact or equivocal morphology and may prevent an incorrect intraoperative decision. Factors associated with diagnostic discordance Sampling limitation was the strongest factor associated with discordance, increasing its odds approximately fifteen-fold. It was present in 66.7% of discordant cases compared with 12.0% of concordant cases. This finding is consistent with Layfield et al. (2018) [4], who reported that 21 of their 35 false-negative frozen-section discrepancies resulted from sampling error. Sampling error arises when the tissue selected for frozen section does not contain the microscopic focus subsequently identified in additional levels or permanent sections. The importance of sampling was also emphasized by Higginson et al. (2023) [10], who identified sampling limitations as a major weakness of frozen-section margin assessment despite its high specificity and overall diagnostic performance. Frozen section can only evaluate the tissue actually submitted and sectioned; it cannot exclude tumour in an unsampled portion of an extensive or irregular surgical margin. Freezing artefact was associated with approximately twelve-fold greater odds of discordance. Rapid freezing may produce ice-crystal artefact, cellular distortion, tissue tearing, folding or thick sections. These changes can obscure nuclear details and stromal invasion. Cautery artefact also significantly increased discordance, with an odds ratio of 6.95. Thermal damage may cause tissue coagulation, nuclear elongation, loss of cytological detail and separation of epithelial tissue from the underlying stroma, potentially simulating or concealing malignancy. Fatty, fibrotic or calcified tissues were significantly associated with discordance. Fatty breast tissue may be difficult to freeze and section uniformly, while fibrotic tissue may fragment or produce thick sections. Calcified specimens may damage the cryostat blade and yield incomplete sections. Chavez et al. (2022) [7] highlighted the challenges associated with fibrotic pancreatic tissue, while Garcia et al. (2021) [6] noted that technical and sampling factors contributed to variation in frozen-section performance during breast surgery. Submission of multiple margins was associated with greater diagnostic discordance. This may reflect more complex resections and an increased probability of orientation, labelling, communication or sampling error. It does not necessarily indicate that examining multiple margins is undesirable; rather, it highlights the need for standardized labelling, orientation diagrams and direct surgeon-pathologist communication. Small or fragmented specimens and margin tissues measuring less than 10 mm showed borderline associations with discordance. Small fragments are difficult to orient and may be exhausted during frozen-section processing. Similarly, uncommon or heterogeneous histology showed a nonsignificant trend towards greater discordance. The limited number of discordant cases reduced statistical power, as reflected by the wide confidence intervals surrounding several odds ratios. The mean turnaround time was significantly longer for discordant than concordant cases. Difficult frozen sections may require repeat cutting, deeper levels, additional tissue blocks or consultation with another pathologist, thereby prolonging reporting time. St John et al. (2017) [3] and Dowling et al. (2024) [12] recognized that although frozen section offers high diagnostic performance, the additional operative time, laboratory workload and resource requirements remain practical limitations.

Intraoperative frozen-section examination demonstrated high diagnostic performance for tumour-margin assessment when compared with permanent-section histopathology. It achieved a sensitivity of 87.2%, specificity of 97.9%, positive predictive value of 93.2%, negative predictive value of 96.0% and overall accuracy of 95.3%. The almost perfect agreement between the two methods (κ=0.871; p<0.001) supports frozen section as a reliable technique for guiding immediate surgical decisions and potentially reducing the need for repeat surgery. Nevertheless, false-negative, false-positive and deferred diagnoses occurred, predominantly because of sampling limitations, freezing and cautery artefacts, difficult tissue characteristics and complex multiple-margin submissions. Frozen-section findings should therefore be interpreted through close surgeon-pathologist collaboration and should not replace permanent-section histopathology, which remains the definitive reference standard for final tumour-margin assessment. Limitations of the study 1. The study was conducted at a single tertiary-care institution; therefore, the findings may not be directly generalizable to centres with different case profiles, laboratory resources or levels of pathology expertise. 2. Tumours from multiple anatomical sites and histological types were included. Differences in margin definitions, specimen consistency, sampling protocols and surgical techniques may have introduced clinical heterogeneity. 3. The number of cases within individual organ-specific subgroups was relatively small, limiting the reliability of site-specific diagnostic-accuracy estimates. 4. Seven deferred frozen-section diagnoses were excluded from the primary sensitivity and specificity calculations, which may have resulted in a slight overestimation of diagnostic performance. 5. Only nine definitive diagnoses were discordant. This small number resulted in wide confidence intervals and limited the statistical power to identify independent predictors of discordance. 6. The diagnostic accuracy of frozen section depended on the representativeness of the tissue submitted by the surgeon. Unsampled portions of an extensive margin could not be evaluated, and therefore sampling error could not be completely eliminated. 7. Frozen-section interpretation may have been influenced by pathologist experience, tissue orientation and technical quality. Interobserver and intraobserver variability were not formally assessed. 8. The effects of specimen-marginal versus tumour-bed sampling and perpendicular versus en face sectioning were not evaluated separately. 9. Turnaround time was measured, but a detailed cost-effectiveness analysis incorporating operative time, laboratory workload and avoidance of repeat surgery was not undertaken. 10. Long-term outcomes such as local recurrence, disease-free survival, overall survival and the rate of repeat surgery were not evaluated. Consequently, the study established diagnostic performance but not the long-term oncological benefits of frozen-section-guided margin revision.

1. Du E, Ow TJ, Lo YT, Gersten A, Schiff BA, Tassler AB, et al. Refining the utility and role of frozen section in head and neck squamous cell carcinoma resection. Laryngoscope. 2016;126(8):1768-75. doi:10.1002/lary.25899. 2. Ettl T, El-Gindi A, Hautmann M, Gosau M, Weber F, Rohrmeier C, et al. Positive frozen section margins predict local recurrence in R0-resected squamous cell carcinoma of the head and neck. Oral Oncol. 2016;55:17-23. doi:10.1016/j.oraloncology.2016.02.012. 3. St John ER, Al-Khudairi R, Ashrafian H, Athanasiou T, Takats Z, Hadjiminas DJ, et al. Diagnostic accuracy of intraoperative techniques for margin assessment in breast cancer surgery: a meta-analysis. Ann Surg. 2017;265(2):300-10. doi:10.1097/SLA.0000000000001897. 4. Layfield EM, Schmidt RL, Esebua M, Layfield LJ. Frozen section evaluation of margin status in primary squamous cell carcinomas of the head and neck: a correlation study of frozen section and final diagnoses. Head Neck Pathol. 2018;12(2):175-80. doi:10.1007/s12105-017-0846-6. 5. Nayanar SK, Krishnan M, KI M, Thavarool SB, Thiagarajan S. Frozen section evaluation in head and neck oncosurgery: an initial experience in a tertiary cancer center. Turk Patoloji Derg. 2019;35(1):46-51. doi:10.5146/tjpath.2018.01439. 6. Garcia MT, Mota BS, Cardoso N, Martimbianco ALC, Ricci MD, Carvalho FM, et al. Accuracy of frozen section in intraoperative margin assessment for breast-conserving surgery: a systematic review and meta-analysis. PLoS One. 2021;16(3):e0248768. doi:10.1371/journal.pone.0248768. 7. Chavez JA, Chen W, Freitag CE, Frankel WL. Pancreatic frozen section guides operative management with few deferrals and errors. Arch Pathol Lab Med. 2022;146(1):84-91. doi:10.5858/arpa.2020-0483-OA. 8. Demir B, Incaz S, Uckuyulu EI, Oysu C. Accuracy of frozen section examination in oral cavity cancers. Ear Nose Throat J. 2022;101. doi:10.1177/0145561320967334. 9. Long SM, McLean T, Valero Mayor C, Fitzgerald CWR, Feit NZ, Katabi N, et al. Use of intraoperative frozen section to assess final tumor margin status in patients undergoing surgery for oral cavity squamous cell carcinoma. JAMA Otolaryngol Head Neck Surg. 2022;148(10):911-7. doi:10.1001/jamaoto.2022.2131. 10. Higginson JA, Breik O, Thompson AH, Ashrafian H, Hardman JC, Takats Z, et al. Diagnostic accuracy of intraoperative margin assessment techniques in surgery for head and neck squamous cell carcinoma: a meta-analysis. Oral Oncol. 2023;142:106419. doi:10.1016/j.oraloncology.2023.106419. 11. Ali JP, Mallick BA, Rashid K, et al. Diagnostic accuracy of intraoperative frozen section for margin evaluation of oral cavity squamous cell carcinoma. BMC Res Notes. 2024;17:43. doi:10.1186/s13104-024-06698-8. 12. Dowling GP, Hehir CM, Daly GR, Hembrecht S, Keelan S, Giblin K, et al. Diagnostic accuracy of intraoperative methods for margin assessment in breast cancer surgery: a systematic review and meta-analysis. Breast. 2024;76:103749. doi:10.1016/j.breast.2024.103749. 13. Ramadan S, Bellas A, Al-Qurayshi Z, Chang K, Zolkind P, Pipkorn P, et al. Use of intraoperative frozen section to assess surgical margins in HPV-related oropharyngeal carcinoma. JAMA Otolaryngol Head Neck Surg. 2025;151(3):253-62. doi:10.1001/jamaoto.2024.4869.