Validating Intraoperative Pathology: Frozen Section
Validating Intraoperative Pathology: Frozen Section
Abstract & Commentary
Robert L. Coleman, MD, Associate Professor, University of Texas; M.D., Anderson Cancer Center, Houston Texas, is Associate Editor for OB/Gyn Clinical Alert
Synopsis: The accuracy of frozen section diagnosis for the assessment of the ovarian mass is good, with acceptable sensitivities for almost perfect specificities. Future studies on patient preferences for the different outcomes, as well as economic analysis, are needed for definite position of this diagnostic technique.
Source: Geomini P, et al. Gynecol Oncol. 2005;96:1-9.
The use of intraoperative frozen section for preliminary analysis of extirpated adnexal masses is ubiquitous and appears to benefit our surgical practice in providing some modicum of comfort in cases where no malignancy is found and guidance to further staging when malignancy is identified. However, how accurate is it? Geomini and colleagues performed a metaanalysis from published studies to determine the accuracy of frozen section when conducted in the context of adnexectomy. Manuscripts appearing in the medical literature between 1966 and 2003 in which intraoperative frozen section diagnosis was compared to final pathologic diagnosis were reviewed. For each study, prevalence of disease (malignant and borderline) and diagnostic sensitivity and specificity was calculated using the final histologic diagnosis as a reference. Two by two tables were constructed; one assigning borderline tumor as benign and one as malignant.
Eighteen studies were included in the analysis. Sensitivity and specificity were 65-97% and 97-100%, respectively when borderline tumors were classified as malignant and 71-100% and 98-100%, respectively, when considered benign. The pooled estimate for specificity was over 99%. Significant heterogeneity precluded an accurate pooled estimate for sensitivity, which varied widely among the surveyed trials. The authors concluded that, in general, the accuracy of frozen section diagnosis was good and characterized by almost perfect specificity. Given the range of prevalent malignancy, discriminate use of the technique could be practiced with likely little overall detrimental impact to patient care. Future work with economic analysis is warranted.
Comment by Robert L. Coleman, MD
I have to admit; the issue of validating the diagnostic accuracy of intraoperative frozen section is one I really never felt compelled to question scientifically. Since this is a diagnostic triage mechanism used by surgeons everyday, it would appear the limitations, risks and benefits are well understood, accepted, and validated in treatment planning. Indeed, part of the informed consent process entails a discussion of the anticipated surgical procedures surrounding the removal of an adnexal mass, which, are indirectly and directly based on the predictive capacity of the frozen section diagnosis. However, this paper represents another good example of questioning a routine practice in an effort to more clearly understand that which we base decisions upon.
It is reassuring that false-positive rates are quite low and reproducible across different centers; few patients will undergo unnecessary radical surgery based on over-diagnosing malignancy. However, it appears a more sizeable proportion of patients may inappropriately undergo an abbreviated adnexectomy in the presence of significant pathology (false negative). The rate at which this occurs was highly variable in the reviewed studies and may reflect significant biases of methodology and/or ranges in expertise for diagnosing ovarian pathology at frozen section. In either event, the impact of being wrong is an important consideration to quantify and scrutinize. Geomini et al approached this issue from a statistical standpoint; that is, using an interesting analytical probability model termed "regret." In this type of analysis, a relative value of being wrong in one’s prediction of malignancy intraoperatively is calculated by a ratio of the two adverse outcomes. For example, a regret ratio (false positive/false negative) of 10 indicates a false positive result at frozen section would generate 10 times the regret as a false negative result. Using this model in a fictive cohort of 1000 patients undergoing frozen section, the authors demonstrate that in most cases frozen section is of value and reduces regret.
One statistical challenge of note in the presented analysis was the assignment of borderline tumors. Since 2 × 2 tables were needed to render hypothesis testing, a decision as to whether this condition was "benign" or "malignant" was required. In the analysis presented both were considered, which led to relatively little variance in the calculated sensitivities and specificities. However, in practical terms, frozen section analysis suggesting borderline pathology may be associated with malignancy in up to 30% of cases—particularly with mucinous tumors. This relatively high rate of "up-diagnosing" at permanent section usually causes the gynecologic oncologist to consider only the "not normal" allocation, rather than a specific diagnosis. The downside of being wrong (understaging a true malignancy) is over-treatment with chemotherapy. However, about 60% of frozen section diagnoses suggesting a borderline lesion are truly associated with that pathology. The intrinsic value of complete surgical staging in this situation is debatable but is still preferred by most gynecologic oncologists. The economic impact of clinical practice based on these accuracy determinations may help to improve the precision of intraoperative diagnosis.
Additional Reading
- Houck K, et al. Obstet Gynecol. 2000;95(6 Pt 1):839-843.
- Pinto PB, et al. Gynecol Oncol. 2001;81(2):230-232.
- Rao GG, et al. Obstet Gynecol. 2004;104(2):261-266.
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.