1. Dr Ollie Rupar Clinical Research Fellow 21st March 2017
Frozen Ovaries
Dr Ollie Rupar
Clinical Research Fellow
21st March 2017

2. Introduction
Frozen section (FS) used during an operation can help guide intraoperative management, allowing for a ‘quick and dirty’ assessment of the tumour characteristics.1
FS use for Ovarian neoplasms differ from other solid organs as the borderline disease entity exists.
Can spread but not necessarily invade tissue.
Retrospective audit of all Ovarian Frozen Sections taken in the 2 years to September 2016 were included.

3. Aims The use of frozen section in the diagnosis of ovarian neoplasms.
To record current accuracy rates.
Comparison
Historical data.
National datasets.

4. Current options for ovarian Masses: Cochrane 2
Surgically stage all women with suspicious ovarian masses.
2-stage procedure, depending on Paraffin Section histopathology.
Intraoperative assessment of the ovarian mass, with Frozen Section.
Accepting false negatives (and non-optimally stage, potentially requiring a second operation)
False Positives – resulting in overtreatment for some women.

5. A Rule of Thirds2 Stage everyone = unnecessary in 2/3’s of women.
Stage no-one = further operations in 1/3 of cases.
Use of Frozen section.

6. Screening tests – a Review
When there are three potential categories for both screening and diagnosis, to compress this 3x3 grid into a 2x2 grid - for purposes of sensitivity/specificity analysis.
Table 1, from Cross et al (2012) has FS/PS = benign = 769. This is the number of true negatives (TN) in table 2.
True Positives (TP) = = 601
False Positive (FP) = 8+3 = 11
False Negative (FN) = = 58.
Taken from Cross et al, 2012 (3).
Compare with Box 1.
Diagnosis at Paraffin Section
Non-Benign
Benign
Screening (Frozen Section)
601 11 58
769

7. Derriford: Methods. All frozen section reports were analysed.
2 years to September 2016.
Comparison was then made with Paraffin Section (recognised gold standard) to ascertain diagnostic accuracy.

8. Derriford Methods continued: What was included?
FS was separated into:
Benign
Borderline
Malignant.
Based on the wording of the FS report some stringent criteria for classification into the above categories.
“Cannot exclude borderline change” = borderline.
“Cannot exclude sarcomatous change.” = malignant.
This was an attempt to clarify the role of FS as a diagnostic test, requiring the sample to be definitively categorised.

9. Derriford Methods Continued.
Paraffin Section (PS):
Benign
Borderline
Malignant
These results were then compiled into a 3x3 grid, and subsequently into a 2x2 grid to give sensitivities, specificities, etc.

10. True Positives, True Negatives.
Tables one and two showing the variation of nomenclature in what constitutes a positive test.
Positive test = Malignant = Malignant only in table 1.
Positive test = Non-Benign in table 2.
Please note the importance of the ordering of the Malignant, Borderline and Benign in the farthest left column and the top rows in tables one and two.
This is not the same Format as other presentations of data. E.g. Cross et al have the reverse ordering in FS column and the PS row compared to the Cochrane evidence seen here. (2, 3).

11. Derriford Results 81 Frozen Sections in total. Average age: 59.
Range
Two methods of analysing results, depending on interpretation of Borderline Masses, and hence a ‘positive’ result.
Non-Benign vs. Benign
Malignant vs. Non-Malignant

12. Derriford Results, continued.
Current Data
n= 81
PS Benign
PS Borderline
PS Malignant
FS Benign 35 6
FS Borderline 10 5
FS Malignant 1 17

13. Derriford Non-Benign vs. Benign
New Data
Diagnostic
PS Non-Benign
PS Benign
Screening
FS Non-Benign 33 7
FS Benign 6 35 %
New
Historical
Sensitivity
84.6
68.4
Specificity
83.3
94.4
PPV
82.5
92.9
NPV
85.4
73.9
New
Historical
False Negatives
6 (7.4%)
6 (16.2%)
False Positives
7 (8.6%)
1 (2.7%)
Total Error
13 (16%)
7 (18.9%)

14. Derriford Malignant vs. Non-Malignant
New Data
Diagnostic
PS Malignant
PS Non- Malignant
Screening
FS Malignant 17 2
FS Non-Malignant 5 57 %
New
Historical
Sensitivity
77.3
69.2
Specificity
96.6
95.8
PPV
89.5
90.0
NPV
91.9
85.2
New
Historical
False Negatives
5 (6.2%)
4 (10.8%)
False Positives
2 (2.5%)
1 (2.7%)
Total Error
7 (8.6%)
5 (13.5%)

15. Ratnavelu et al, 2016 (2) Cochrane Review.
11181 patients from 38 studies.
Early stage disease (stage 1-2)
Risk of Malignancy Index
= USS x Menopausal status x CA125
>200
Per 1000 patients on average*
Accuracy varied with disease prevalence. *
Non-Benign/Benign
% (Range)
Sensitivity
96.5 (95.5 – 97.3)
Specificity
89.5 (58 – 99)
False Negatives*
n = 8 – 13
False Positives*
n = 67 – 81
Total Error
8 – 8.9%

16. Cochrane Review For 1000 hypothetical patients… 2
Borderline FS = positive test
280 women correctly diagnosed with cancer.
75 women incorrectly diagnosed with cancer (FP).
10 women missed (FN)
Borderline FS = negative test.
261 correctly diagnosed with cancer.
4 incorrectly diagnosed with cancer (FP).
29 would have cancer missed (FN).

17. Cross et al, 2012. (3) 1439 patients over 11 year period.
RMI also used.
Gynae-Path and General Pathologists report FS.
Non-Benign vs. Benign system adopted.
“The common theme is that an FS is only of value if it will actually alter the procedure that the surgeon performs at the time of the surgery. “
“Grouping of borderline tumours with malignant for clinical purposes is logical.”

18. Cross, cont’d. % Sensitivity 91.2 Specificity 98.6 PPV 98.2 NPV 93.0
Diagnostic
PS Non-Benign
PS Benign
Screening
FS Non-Benign
601 11
FS Benign 58
769
Cross et el, 2012. %
Sensitivity
91.2
Specificity
98.6
PPV
98.2
NPV
93.0
False Negatives
58 (4.0%)
False Positives
11 ( 0.8%)
Total Error
69 (4.8%)
Cross et al,

19. Comparison. Non-Benign vs. Benign 2016 Audit n = 81 2012 Audit. n = 37
Cross et al, n = 1439
Cochrane, 2016.
n = 11181
Sensitivity
84.6%
68.4%
91.2%
96.5% (86 –100%)
Specificity
83.3%
94.4%
98.6%
89.5 (58 – 99%)
False Negatives
7.4%
16.2% 4%
7.5%
False Positives
8.6%
2.7%
0.8% 1%
Total Error
16%
18.9%
4.8%
8.5%

20. Derriford 2016 - What happened? Non-Benign vs. Benign
False positives (n=7)
6 Underwent staging operation.
7th underwent cystectomy with the FS report:
“Single mitotic figure…convincing borderline changes are not identified.”
Of note regarding reporting bias: one FP case had previous borderline serous tumours in
False Negatives (n=6)
All 6 optimally staged.
“Thick walled endometrium and longstanding right ovarian mass.”
High grade adenocarcinoma (?incidental) in case with CGIN

21. Derriford 2016. Malignant vs Non-Malignant
False Positives (n=2)
One staging operation
One Radical (+LNs)
False Negatives (n=5)
3 were optimally staged
One had a radical operation.
One lady had a cystectomy, and then a staging operation 2 weeks later, following PS report.

22. Errors… The Epithelial cell lines are troublesome. Mucinous Tumours.
Literature states 36.6% error rate (3, 5)
46.1% (n=6) of errors in 2016 Audit
Serous Tumours
9% stated error rate (5)
38.5% (n =5) of errors involved in the audit.
Mixed Serous/Mucinous Cystadenofibroma
7.7% (n=1)

23. Frozen Section Cell Line make up.

24. Paraffin Section - Cell line

25. Frozen Section Accuracy
 Cell line/ Comment
Frozen Section (FS)
Number of Paraffin Section (PS) correctly diagnosed at Frozen Section.
Accuracy (%) of total PS diagnosed at FS.
Total Paraffin Section (PS)
PS as % of Total (81)
Serous 31 29
93.55% 41
50.6
Mucinous 13 10
76.92% 18
22.2
Germ 6 5
83.33%
7.4
Sex Chord/Stromal 3
60.00% 4
4.9
Brenner
75.00%
3.7
Clear
Endometroid 1
100.00% 2
2.5
Benign - unable to further classify 9
22.22%
Undifferentiated
1.2
Mixed
Borderline, no further comment.
0.0
Malignant - unable to further classify/comment
Total 81 53
65.4%
100.00
31 samples were categorised as Serous on FS however 29 of these were Serous on PS. Giving an accuracy of 93.5%. There were 41 PS sections in total (of 81) meaning 50.6% of samples were Serous.
53 of the 81 samples (65.4%) were correctly diagnosed at FS. These figures do not take into account of the benign/borderline/malignant status of the neoplasm.

26. Borderline. Borderline results account for 25.9% of FS reports (n=21)
Previously 10.8% (n=4)
Invasive Cancer (on PS) in 23.8% (n=5)
Borderline (PS) 47.6% (n=10)
Benign (PS) 28.6% (n = 6)
Cochrane found 21% of borderlines were invasive cancer on PS.

27. Discussion Comparison with a recent Cochrane Review proves useful.
Smaller numbers at Derriford.
Accuracy rates are comparable.
Non-Benign vs. Benign is the most commonly used method of dividing tissue into Benign and ‘Malignant.’
Error rates are higher than expected:
This could be the result of the design of this audit.
Size of sample.
However they are improving!
How can they continue to improve?

28. Conclusions
Communication between Surgeon and Pathologist is of paramount importance.
Verbal
Written
Accuracy rates are in line with nationally accepted limits, and improving on previous.
How could the FS service improve further?

29. Many thanks. Dr Tim Bracey for help with this audit.
Administrative team in Pathology.
And for your time!

30. References
Jo Morrison, Toby Lasserson. Finding time to make the right decision: using frozen section to inform intra- operative management of suspicious ovarian masses [editorial]. Cochrane Database of Systematic Reviews 2016;(3): DOI:  / ED000109
Ratnavelu NDG, Brown AP, Mallett S, Scholten RJPM, Patel A, Founta C, Galaal K, Cross P, Naik R.
Intraoperative frozen section analysis for the diagnosis of early stage ovarian cancer in suspicious pelvic masses. Cochrane Database of Systematic Reviews 2016, Issue 3. Art. No.: CD 
DOI: / CD pub2.
Cross P, Naik R, Patel A, Nayar A, Hemming J, Williamson S, Henry J, Edmondson R, Godfrey K, Galaal K, Kucukmetin A, Lopes A. Intra-operative frozen section analysis for suspected early-stage ovarian cancer: 11 years of Gateshead Cancer Centre experience. BJOG 2012; 119: 194–201. DOI: /j x
audit data, (Mr A Talaat, 2016, pers comm. 2/9/16).
Ganesan, R. Brown, L. J. R. Kehoe, S. McCluggage, W. G. El-Bahrawy, M. A. The role of frozen section sin gynaecological oncology: survery of practice in the United Kingdom. European Journal of Obstetrics & Gynaecology and Reproductive Biology. 2013, 166, 204 –

31. Historical data 4