Intraoperative frozen section in gynaecology cancers with special reference to ovarian tumours: time to “unfreeze” the pitfalls in the path of the Derby horse of Oncology

Abstract

Purpose

In an oncological set up the role of frozen section biopsy is undeniable. They serve as an important tool for surgeon’s intraoperative decision making but the diagnostic reliability of intraoperative frozen section may vary from institute to institute. The surgeon should be well aware of the accuracy of the frozen section reports in their setup to enable them to take decisions based on the report. This is why we had conducted a retrospective study at Dr B. Borooah Cancer Institute, Guwahati, Assam, India to find out our institutional frozen section accuracy.

Methods

The study was conducted from 1st January 2017 to 31st December 2022 (5 years). All gynaecology oncology patients who were operated on during the study period and had an intraoperative frozen section done were included in the study. Patients who had incomplete final histopathological report (HPR) or no final HPR were excluded from the study. Frozen section and final histopathology report were compared and analysed and discordant cases were analysed based on the degree of discordancy.

Results

For benign ovarian disease, the IFS accuracy, sensitivity and specificity are 96.7%, 100% and 93%, respectively. For borderline ovarian disease the IFS accuracy, sensitivity and specificity are 96.7%, 80% and 97.6%, respectively. For malignant ovarian disease the IFS accuracy, sensitivity and specificity are 95.4%, 89.1% and 100%, respectively. Sampling error was the most common cause of discordancy.

Conclusion

Intraoperative frozen section may not have 100% diagnostic accuracy but still it is the running horse of our oncological institute.

Introduction

The practical importance of intraoperative frozen section is well known in any surgical setup, especially in cancer speciality hospitals. Specimens from the gynaecologic organs account for a major chunk of specimens sent for intraoperative frozen section (IFS) analysis with the vast majority being ovarian masses. Ovarian cancer presents in women of various age strata including the reproductive age group in whom the incidence of malignant tumours is as high as 20% (Shen et al. 2021). This is of paramount importance considering the need for fertility preservation and the extent of surgery.

The gold standard for diagnosing a malignant tumour is the histo-pathological report (HPR). A variety of markers and investigations are used pre-operatively to predict the malignant nature of an adnexal mass which include tumour markers Cancer Antigen 125 (CA 125), Human epididymis 4 (HE 4), carcinoembryonic antigen CEA, Alfa feto-protein (AFP), CA 19-9, lactate dehydrogenase (LDH), beta-human chorionic gonadotropin (BHCG) levels and a combination of ultrasound, CA 125 and menopausal status, i.e. Risk of malignancy index (RMI score) (Al-Musalhi et al. 2015), risk of malignancy algorithm (ROMA) and use of computed tomography (CT) and magnetic resonance imaging (MRI) (Ledermann et al. 2013). However, most of these methods have not shown good accuracy in the pre-operative detection of malignancy.

Intraoperative frozen section (IFS) finds its place as an important diagnostic tool in this regard wherein a representative tissue is analysed, thus providing a preliminary diagnosis based on which the operating surgeon plans the extent of the surgery. IFS is also used in hysterectomy specimens to evaluate the endometrium for malignancy, in radical hysterectomy for cervical cancer to evaluate the nature of suspicious nodes and in vulval cancer to assess the margin status.

This study aims to determine the accuracy of diagnosis rendered on intraoperative frozen section (IFS) compared to the final diagnosis provided on formalin-fixed paraffin-embedded (FFPE) tissue sections. We analysed possible sources of discrepancy and the further management of such discordant cases. This enabled us to detect the pitfalls of IFS thereby serving as an internal audit to improve our health care.

Material and methods

Aim of study

1. Primary objective

To estimate the degree of accuracy of diagnosis rendered on intraoperative frozen sections (IFS) compared to final diagnosis provided on formalin-fixed paraffin-embedded histopathological tissue sections (FFPE-HPR) in gynaecological cancers with emphasis on ovarian masses in a tertiary cancer care institute.

• 2.
  Secondary objectives

  • A.
    To analyse the source of discrepancies with special emphasis on cases where a change in patient management had to be made.
  • B.
    To study the further management of discordant cases.

Inclusion criteria

All gynaecology oncology patients who were operated on during the study period and had an intraoperative frozen section done.

Exclusion criteria

All cases where a frozen section was done and final histopathology reports were incomplete or unavailable.

Study design

It is a retrospective observational study.

Study period

1st January 2017 to 31st December 2022 (5 years).

Site

Department of Gynaecology Oncology and Oncopathology, Dr Bhubaneswar Borooah Cancer Centre, Guwahati, Assam.

Prior medical reports of Gynaecology Oncology patients who have been operated on between 1st January 2017and 31st December 2022 in the institute and had an intraoperative frozen section done had been analysed.

Statistical methods

It is a descriptive study. All data analysis had been performed on computer using SPSS version 23. The descriptive statistics will be analysed as percentages and bar diagrams will be used. The results were compared to permanent sections to evaluate accuracy, sensitivity, specificity, positive predictive values and negative predictive value of frozen section.

The procedure of frozen section

The specimen for frozen section analysis was sent to the Oncopathology lab soon after removal in the operation theatre with normal saline and relevant clinical details. An immediate gross examination was done by our pathologists recording the size, volume and intactness/breach of the capsule. After gross examination, sections were obtained from representative areas at the discretion of the pathologist. A minimum of two sections and as many as six sections were taken depending upon the size or type of tumour. Sections of 4–6 microns were cut using Cryostat microtome (Leica CM1850) at temperatures − 22 to − 24 °C and using tissue freezing media from Leica. The sections were then stained by rapid haematoxylin and eosin stain. The average overall time for the procedure was 20 min which includes grossing and processing of 5–10 min and reporting which is of 8–10 min depending upon the number of sections and the nature of the specimen. The frozen sections were reported by trained senior pathologists and the report was communicated to the surgeon directly over the phone and entries were made into the frozen section register. The remaining tissue as well as original sections taken during IFS were then submitted for final formalin-based fixation and tissue processing.

We retrieved the data from archives of the Oncopathology Department frozen section register and electronic medical records from 1st January 2017 to 31st December 2022 which included FFPE-HPR reports as well as other clinical and demographical details of the cases included in the study. IFS and FFPE-HPR concordance data were recorded in tabular manner. Any discrepancy/discordancy between IFS and FFPE HPR reports was noted down.

We further divided discordancy into major and minor errors. Major discordancy referred to those cases where discordant results led to major differences in patient management in terms of the extent of radicality of surgery. Minor discordancy referred to those cases where the change was more of pathological significance rather than a change in the intraoperative management. In major discordant cases, the records were reviewed to analyse further management. The reasons for such discordancy were also reviewed.

Ethical clearance from the institutional review board had been obtained.

Results

The study included a total of 334 cases which included 307 ovarian masses to ascertain the nature of the tissue, s cervical cancers to assess suspicious lymph nodes, 15 vulval cancers for assessing surgical margin status and evaluation of nodes and 5 uterine cancers to assess the nature of endometrial tissue. Figure 1 shows the overall distribution of cases. The overall sensitivity of 91% and 100% specificity was recorded for IFS diagnosis overall.

Fig. 1.

Distribution of cases. A total of 17 major discordant cases had been reported all of which have been recorded in ovarian mass. The 27 other site cancers had 100% concordant reports

As there was 100% concordance in the evaluation of other gynaecological malignancies detailed analysis of IFS in ovarian mass was studied.

Table 1 shows the comparison of IFS and FFPE-HPR diagnoses with respect to ovarian masses. There were 276 (89.9%)concordant cases and 31 discordant cases(10.1%)of which 17 cases were major discordant(54.8%) and 14 cases were minor discordant(45.2%). All major discordancies were underdiagnosed cases which included seven benign cases and seven borderline cases which were later diagnosed to be malignant on FFPE-HPR.

Table 1.

Comparison of IFS and FFPE-HPR for ovarian cases (N = 307)

	Benign on final HPE	Borderline on final HPE	Malignant on final HPE	Concordancy (%)
Benign on IFS	163 (concordant)	3	7	94.2
Borderline on IFS	0	12 (concordant)	7	63.1
Malignant on IFS	0	0	115 (concordant)	100

IFS intraoperative frozen section, FFPE-HPR formalin-fixed paraffin-embedded histopathological tissue sections

Table 2 and Fig. 2 show the diagnostic values of accuracy, sensitivity, specificity, positive predictive value and negative predictive value of benign, borderline, malignant and combined borderline and malignant cases in IFS.

Table 2.

Diagnostic value of IFS for benign, borderline and malignant ovarian lesions N = 307

Diagnostic value % (95% CI)	Benign	Borderline	Malignant	Borderline and Malignant combined
Accuracy	96.7% (94–98.4)	96.7% (94–98.4)	95.4% (92.4–97.4)	96.7% (94–98.4)
Sensitivity	100% (97.7–100)	80% (51.9–95.6)	89.1% (82.4–93.9)	93% (87.6–96.6)
Specificity	93% (87.6–96.6)	97.6% (95.1–99)	100% (97.9–100)	100% (97.7–100)
Positive predictive value	94.2% (89.9–96.7)	63.1% (54.1–78.8)	100%	100%
Negative predictive value	100%	98.9% (97.1–99.6)	92.7% (88.5–95.4)	94.2% (89.9–96.7)

IFS intraoperative frozen section

Fig. 2.

Bar diagram showing the diagnostic value of IFS for benign, borderline and malignant ovarian lesions

For benign ovarian disease, the IFS accuracy, sensitivity, specificity, positive predictive value and negative predictive value are 96.7%, 100%, 93%, 94.2% and 100% respectively. For borderline ovarian disease the IFS accuracy, sensitivity, specificity, positive predictive value and negative predictive value are 96.7%, 80%, 97.6%, 63.1% and 98.9%, respectively. For malignant ovarian disease the IFS accuracy, sensitivity, specificity, positive predictive value and negative predictive value are 95.4%, 89.1%, 100%, 100% and 92.7%, respectively. When both borderline and malignant disease are clubbed together the IFS accuracy, sensitivity, specificity, positive predictive value and negative predictive value is 96.7%, 93%, 100%, 100% and 94.2%, respectively.

Tables 3, 4 and 5 shows the cases where major discordancy has been observed. There were no overdiagnosed cases. The 17 underdiagnosed cases have been further divided into 3 categories which include upstaging of a. Benign to borderline b. Benign to Malignant c. Borderline to malignant. Ten cases were planned for adjuvant chemotherapy; however, two patients defaulted chemotherapy. Two patients underwent completion surgery after chemotherapy.

Table 3.

Major discordancy between IFS and FFPE-HPR (benign to borderline) (1.7%)

S. no.	Age	IFS	HPE	Surgery done	Size of the lesion and Reason for discordancy	Management of discordant cases	Resurgery done
1.	24	Mucinous cystadenoma	Borderline mucinous	LSO Omental nodule excision	12 cm (1, 2)	Observation	No
2.	37	Mucinous cystadenoma	Borderline mucinous with micro invasion	TAH BSO ICO PLND PALNDa	30 cm (1, 2)	Observation	No
3.	47	Mucinous cystadenoma	Borderline with micro invasion	TAH BSO Omentectomy	15 cm (1, 2)	Observation	No

IFS intraoperative frozen section, HPR final histopathology report, LSO left salphingo-oophorectomy, TAH total abdominal hysterectomy, BSO bilateral salphingo-oophorectomy, ICO infracolic omentectomy, PLND pelvic lymphadenectomy, PALND paraaortic lymphadenectomy P-parity. Index of discordancy: 1 sampling error, 2 technical error

^aIn view of large size of tumour and the refusal of patient to undergo a second surgery an extensive surgery which included TAH BSO ICO PLND PALND was performed 

Table 4.

Major discordancy between IFS and FFPE-HPR (benign to malignant) (4%)

S. no.	Age	IFS	HPE	Surgery done	Size of the lesion and reason for discordancy	Management of discordant cases	Resurgery done
1.	30	Mucinous cystadenoma	Mucinous cystadenocarcinoma	TAH BSO ICOa	37 cm (1)	Adjuvant chemotherapy(defaulted)	No
2.	46	Benign serous cystadenoma	High-grade serous carcinoma	TAH BSO	15 cm (1, 3)	Adjuvant chemotherapy	No
3.	47	Endometriosis	Clear cell carcinoma	TAH BSO SCOb	12 cm (3, 4)	Adjuvant chemotherapy(defaulted)	No
4.	62	Benign serous cystadenoma	High-grade serous carcinoma	TAH BSO SCO PLNDc	14 cm (1, 2)	Adjuvant chemotherapy	No
5.	30	Sertoli Leydig tumour	High-grade serous carcinoma	RSO ICO	5 cm (3, 4, 5)	Adjuvant chemotherapy	TAH LSO SCO post chemotherapy
6.	32	Mucinous cystadenoma	High-grade serous carcinoma	LSO SCOa	12 cm (1, 3)	Observation	No
7.	22	Mucinous cystadenoma	Mucinous cystadenocarcinoma	LSO ICO multiple biopsy	26 cm (1, 2)	Adjuvant chemotherapy	No

IFS intraoperative frozen section, HPR final histopathology report, LSO left salphingo-oophorectomy, TAH total abdominal hysterectomy, BSO bilateral salphingo-oophorectomy, ICO infracolic omentectomy, SCo supracolic omentectomy, RSO right salphingo-oophorectomy, PLND pelvic lymphadenectomy, PALND paraaortic lymphadenectomy P-Parity

^aIn view of large size of tumour, positive peritoneal cytology and refusal of patient to undergo surgery at a second sitting TAH BSO ICO was done

^bIn view of long standing history of endometriosis in post-menopausal woman TAH BSO SCO was done

^In view of gross finding of solid cystic tumour with focal areas of surface papillary projections a thorough staging was undertaken. Index for discordancy: 1 sampling error (limited number of sections during IFS, 2 the technical difficulty in freezing tissues especially with high mucinous content, 3 frozen section artefacts, 4 inexperienced pathologist, 5 inherent nature of the disease with heterologous elements  

Table 5.

Major discordancy between IFS and FFPE-HPR (borderline to malignant) (36%)

S. no.	Age	IFS	HPE	Surgery done	Size of the lesion and Reason for discordancy	Management of discordant cases	Resurgery done
1.	58	Borderline mucinous	Mucinous cystadenocarcinoma	TAH BSO ICO	15 cm (1, 2)	Adjuvant chemotherapy	No
2.	47	Borderline micropapillary serous	Low-grade serous papillary carcinoma	TAH BSO ICO	8 cm (1, 5)	Adjuvant chemotherapy	No
3.	69	Borderline mucinous	Mucinous cystadenocarcinoma	TAH BSO ICO PLND	30 cm (1, 2)	Observation	No
4.	50	Borderline mucinous	Sero-mucinous cystadenocarcinoma	TAH BSO SCO	30 cm (1)	Adjuvant chemotherapy	No
5.	42	Borderline mucinous	Mucinous cystadenocarcinoma	TAH BSO PLND	10 cm (1, 2)	Observation	No
6.	58	Borderline micropapillary pattern	Low-grade serous cystadenocarcinoma	TAH BSO SCO	15 cm (1, 3)	Adjuvant chemotherapy	No
7.	39	Borderline mucinous	High-grade endometrioid adenocarcinoma	RSO SCO	7 cm (1)	Observation	TAH LSO 4 years later

Index for discordancy: 1 sampling error (limited number of sections during IFS, 2 the technical difficulty in freezing tissues especially with high mucinous content, 3 frozen section artefacts; inexperienced pathologist, 4 in experienced pathologist, 5 inherent nature of the disease with heterologous elements, 6 IHC requirements 

IFS intraoperative frozen section, HPR final histopathology report, LSO left salphingo-oophorectomy, TAH total abdominal hysterectomy, BSO bilateral salphingo-oophorectomy, ICO infracolic omentectomy, SCo supracolic omentectomy, RSO right salphingo-oophorectomy, PLND pelvic lymphadenectomy, PALND paraaortic lymphadenectomy P-parity

Table 6 shows cases having minor discordancy where there was a difference between the IFS and FFPE-HPR but the discordancy did not affect the surgery planned. It includes both the following: a. cases where a benign cyst had a different type of final HPE diagnosis and b. cases where a malignant cyst had a different type on final HPE diagnosis.

Table 6.

Minor discordant cases

S. no.	IFS	HPE	Reason
(a) Benign to benign
1.	Benign cyst	Sex cord-stromal tumour	1, 3
2.	Endometriotic cyst	Serous cystadenofibroma	1
3.	Benign serous cystadenoma	Haemorrhagic cyst	3, 4
4.	Fibrothecoma	Haemorrhagic cyst	2, 3
5.	Benign serous cystadenoma	Haemorrhagic cyst	3, 4
6.	Benign serous cystadenoma	Haemorrhagic cyst	3, 4
7.	Benign serous cystadenoma	Corpus luteal cyst	2, 3, 4
8.	Seromucinous cystadenoma	Endometriosis	1, 3
9.	Seromucinous cystadenoma	Endometriotic cyst	1, 3
10.	Mucinous tumour	Dermoid	1, 2, 5
(b) Malignant to malignant
1.	Malignant Brenner	High-grade serous carcinoma	1, 3, 4
2.	Clear cell carcinoma	Endometriod adenocarcinoma G1	1, 3
3.	Poorly differentiated adenocarcinoma	GIST	1, 6
4.	High-grade adenocarcinoma	Clear cell carcinoma	1, 3

Index for discordancy: 1 sampling error(Limited number of sections during IFS), 2 the technical difficulty in freezing tissue with high mucinous content, 3 frozen section artefacts, 4 inexperienced technician, 5 inherent nature of the disease with heterologous elements, 6 IHC requirements 

Figure 3 shows the causes of major and minor discordancy as discordant sources combined.

Fig. 3 .

Source of discordancy between IFS and FFPE-HPE

Discussion

Accurate classification of gynaecologic neoplasms, especially ovarian neoplasms as benign, borderline or invasive on frozen sections often is a real challenge in the practical world even for a well-trained pathologist. With the help of IFS diagnosis, the surgeon would have a provisional report of the nature of the specimen. This would translate into increased or decreased radicality of the surgery and advocate more aggressive follow-up of the patients diagnosed to have malignancy to ensure proper adherence to further follow-up visits in advanced disease or avoidance of chemotherapy/completion surgery in cases of well-differentiated stage 1A ovarian cancer disease(Maggioni et al. 2006). In addition to IFS application in ovarian neoplasms, it also finds its place in cervical and vulval carcinoma to see for lymph node metastasis and tumour margin analysis and assess prognostic markers in endometrial carcinoma (Scurry and Sumithran 1989). A study done by Swift et al. (2022) showed 97.2% PPV of IFS for lymph node assessment in vulval cancer. A similar study done by Santaro et al. (2019) showed 97.2% agreement for the evaluation of endometrium. A retrospective study done by Tanaka et al. (2020) showed 72% sensitivity in the evaluation of cervical cancer but this included the nature of mass as well. We found 100% concordance in our study for cervical, uterine and vulval cancer. This could be attributed to the lower number of cases in our study.

On analysis of IFS utility in ovarian masses, a recent systematic review by Meys et al. showed the average IFS sensitivity when borderline and malignant ovarian masses were clubbed together was 96.5% and specificity was 89.5% (Meys et al. 2017). In another study by Kung et al., the IFS sensitivity was 92.51% and the specificity was 100% (Kung et al. 2019). The sensitivity of IFS in other studies ranged between 73 and 98% (Cuello et al. 1999; Hamed et al. 1993; Buza 2019). In our study, the sensitivity and specificity of IFS for both borderline and malignant cases clubbed together were 93% and 100%, respectively. This implies while IFS can be criticized for underdiagnosing tumours resulting in a second surgery, it does not over-diagnose malignancy. The PPV in our study was 100% compared to a study by Kung et al. (2019) which showed PPV of 81.2% as we had also considered cases where the IFS diagnosis showed borderline with other features suggestive of malignancy as malignant cases in our analysis. This was proven to be a good decision as most of the patients had a malignant diagnosis on the final HPE. This cannot be taken as a discrepancy as considered in other studies as the surgeon would do a complete staging in these cases (Ureyen et al. 2014).

IFS demonstrates low performance in differentiating between borderline and malignant tumours. Various studies have shown the up gradation rate of borderline to malignant tumours was 20–25% (Medeiros et al. 2005; Ratnavelu et al. 2016). We found 63% upgradation of borderline tumours to malignant tumours. Five cases were mucinous in histology. This has been the largest source of discrepancy in gynaecology specimens as mucinous tumours are heterogeneous in nature (Medeiros et al. 2005; Tempfer et al. 2007). The mean size of mucinous tumours was 24.8 cm in a study by Subbian et al. (2013) while the average size of the tumours in our study was 17 cm. It is well known that the large size of mucinous tumours makes them pre-disposed to sampling errors. In a busy tertiary care centre, the logistic issue becomes a deterrent to ideal sectioning limits and it is impractical to do more than six frozen sections given the large size of the adnexal masses.

The accuracy of IFS is affected by various factors such as errors in sampling, technical errors and interpretation errors (Stewart et al. 2006). Heterogeneous tumours such as mucinous tumours and dermoid cysts are prone to sampling errors (Yoshida et al. 2021). In a study by Kung, the rate of sampling error was as high as 47% (Kung et al. 2019). We observed 64.5% of discordancy was due to sampling errors. It is important that proper sectioning of the cyst wall is done, especially if the ovarian tumour does not have solid areas. Proper orientation of the sample is also a key component during IFS as a properly oriented section can help to detect the infiltrative pattern better than the expansile variety in mucinous carcinomas which would have falsely been diagnosed as borderline in IFS (Baal et al. 2017). This could also explain the discordance reported in various studies of early ovarian cancer (Baal et al. 2017; Armstrong et al. 2021). In comparison to a study done by the College of American Pathologists which showed 40% interpretation error and 12.7% sectioning error (Bachner and Howanitz 1991), we found 25% of discordancy due to technical errors and 41% due to frozen section artefacts.

Serous and clear cell carcinoma are usually well identified on IFS compared to mucinous tumours (Kung et al. 2019). This is primarily due to the over-reliance of IFS on architectural pattern and cellular morphology making it difficult to pick up the correct diagnosis in mucinous tumours which are usually very large and have heterogeneous histology as described earlier.

A suggestion we would like to make is to have a peri-operative discussion between the operating surgeon, the pathologist and at times the medical oncologist along with the patient and her relatives. This is of paramount importance as the provision of relevant clinical details of the patient would enable the pathologist to give a better diagnosis. Another solution would be to prepare the frozen section slides in the operating room complex where the surgeons can provide their input thus leading to better sampling rates, especially in large tumours.

We observed 17 out of 307 cases (5%) to be underdiagnosed on IFS. Three cases were upstaged from benign to borderline and were observed. Of the remaining 14 patients four patients (28%) were observed whereas the remaining ten patients (72%) were given adjuvant chemotherapy. Completion surgery was done in 2 patients post-chemotherapy (20%). Patients with early-stage high-grade tumours were administered adjuvant chemotherapy. It worked in our favour that seven out of eight patients (87%) had a complete surgery upfront given their elderly age and completion of family. One patient refused further completion surgery as pre-operative imaging did not show any disease post-chemotherapy and only lymph nodes were to be removed. The upstaging of benign to malignant tumours was also seen in a study by Cross et al. (2012) who reported 19 similar cases and attributed interpretational error and sampling error to be the causative factors. It is always prudent to get a second opinion from another pathologist during a dicey report.

The main strength of the study is that of the heterogeneity of cases which gives an overall picture of a wide variety of cases an oncologist is bound to encounter. We also have explained the pitfalls of the pathology diagnosis and how to avoid it. The main drawback is being a retrospective study the possibility of review of the IFS slides is very much limited and since the pathologist was already aware of the IFS report, proper blinding could not be done.

There is something not found in literature called “surgeon gut feeling” which has helped a few of our patients where despite a diagnosis of benign/ borderline a complete surgery has been done due to the surgeon's clinical acumen and preoperative imaging of the patient. It is beneficial for patients to receive the provisional diagnosis based on IFS thus helping to counsel them for chemotherapy as the specificity was 100% in our study.

Conclusion

Frozen section despite its drawbacks still is the working horse of any oncology unit. It is up to the surgeon and pathologist as a team to correlate the clinical, radiological and intraoperative findings to keep the horse on track. Proper counselling of the patient is to be done to explain the pitfalls of such a procedure. We also encourage regular auditing of the IFS reports to improve the accuracy of reporting from time to time.

Author contributions

A, E and F main manuscript text B, C,D prepared figures and tables

Funding

The research had not received any grant from any funding agency in the public, commercial or not for profit sectors.

Declarations

Conflict of interest

There is no conflict of interest among the authors.