Accuracy of Intraoperative Frozen Section in Detection of Acute Invasive Fungal Rhinosinusitis Open Access

After institutional review board approval was obtained, a retrospective search was conducted of the anatomic pathology information system at University Hospital to identify cases that were clinically suspicious for AIFRS from January 2009 to January 2019. Separate keyword searches were conducted, querying all report fields, using any of the terms “fungus,” “fungal,” “fungi,” “hyphae,” “ulcer,” “necrotic,” “Grocott methenamine silver stain,” and “Periodic acid–Schiff stain” in conjunction with any of the terms “sinonasal,” “sinus,” “nasal,” “sinusitis,” “palate,” or “orbit.” The resulting lists were manually merged and then culled to remove duplicates. After review of the preoperative clinical documentation, cases were excluded if AIFRS was not clinically suspected or if no intraoperative consultation was obtained.

Per procedure, frozen-section samples were received fresh from the operating room, embedded in optimal cutting temperature medium, and allowed to freeze within a cryostat at an approximate temperature of −20°C. Biopsies were entirely embedded when size permitted. Sections were prepared using a Leica Biosystems cryostat, at a thickness of approximately 5 μm, across 2 different levels. These were stained with hematoxylin and eosin (H&E) and interpreted by the general surgical pathologist on-call for frozen sections. After this, the frozen tissue was thawed, fixed in formalin, and processed for permanent sections. Permanent tissue sections were stained with H&E special stains for fungi were performed routinely if H&E-stained sections revealed no fungal organisms. Additional sterile specimens were sent for fungal culture based on the clinical team decision. All the specimens collect for fungal culture are handled aseptically by the microbiology lab. At the microbiology lab the tissue specimens are handled based on their size. If the tissue sample received larger than a pea size, aseptically transferred to a sterile, disposable Petrie dish, and by using a sterile scalpel, it is cut into 5 pea-sized portions. The portion that is selected for culture usually that is bordering and within the area of infection (ie, necrotic tissue, red, and purulent areas). Four portions usually used to inoculate the fungal culture media, whenever possible, whole pieces of tissue are used to inoculate fungal cultures.

For the purposes of calculating performance characteristics, only the original interpretations were used. We calculated the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) with the frozen section as the test method and the permanent section as the gold standard. For proportions, 95% CI were calculated according to the efficient score method and corrected for continuity. The sensitivity, specificity, PPV, and NPV were estimated with 95% CI by using descriptive measures from contingency table by using R-3.5.1 and R Studio-1.1.463.¹⁸ 

We then retrospectively reviewed frozen-section and permanent slides from all cases to identify background histopathologic features. In particular, we noted the presence or absence of neutrophil-mediated inflammation, fibrinopurulent debris, and tissue necrosis. Last, discrepant cases were reviewed to determine contributing factors.

Our search identified 53 cases of suspected AIFRS, 48 of which (90.6%) had at least 1 specimen sent for intraoperative consultation, resulting in 133 frozen-section specimens. Of 48 patients (Table), 30 (62.5%) had a final diagnosis of AIFRS on permanent section in at least 1 biopsied site (Figure 1), and 18 (37.5%) were negative for AIFRS. Of note, none of the 18 negative patients were diagnosed with AIFRS in the following 90 days. Of 30 patients with a final diagnosis of AIFRS, there were 12 (40%) females and 18 (60%) males, and the mean age was 56 years (range, 19–77). Of 18 patients whose final diagnosis was negative for AIFRS, 8 (44.4%) were females and 10 (55.6%) were males, with mean age of 57 years (range, 34–79).

Of 30 patients who had a final diagnosis of AIFRS, 28 (87.5%) had known predisposing conditions. One of the remaining 2 patients had a history of asthma with no immunosuppressive therapy, while the other had a history of rheumatoid arthritis with no immunosuppressive therapy. The most common underlying condition was diabetes mellitus, present in 19 patients (59%); others included myelodysplastic syndrome, acute leukemia, HIV/AIDS, iatrogenic immunosuppression, and cirrhosis with leukopenia.

The total number of frozen sections sent for pathologic interpretation per patient ranged from 1 to 8 (mean, 2.4). The mean number of frozen-section specimens in positive patients was 2.6 (range, 1–8), and the mean number in negative patients was 2.33 (range, 1–4). The most common subsites biopsied were the middle turbinates (25 specimens, 19%) and the maxillary sinus (24 specimens, 18%). The most commonly positive AIFRS subsite was the maxillary sinus, comprising 21% (13 specimens) of all positive samples, followed by the lateral nasal wall, comprising 18% (11 specimens).

Microbiologic findings included Rhizopus species in 12 of 30 positive cases, Aspergillus species in 5, Mucor species in 2, Candida species in 2, Bipolaris species in 1, Alternaria species in 1, and 7 showed no growth (Figure 2, A through D). Bipolaris and Alternaria species are commonly encountered as colonizers; however, each was isolated in association with concomitant histologic evidence of tissue invasion. It is possible that these represent culture contamination with colonizing agents, but invasive infection has been reported rarely in association with both agents.^19–26 

There were no cases in which fungal culture was positive after a negative histopathologic diagnosis. All Mucormycetes (formerly known as the Zygomycetes) were accurately classified histopathologically, as were all non-Mucormycetes. The cases of Candida were accurately classified as “morphologically consistent with Candida species.”

Of 18 patients who were negative for AIFRS, 13 (72.2%) had known predisposing conditions, including diabetes mellitus in 6 (33%) patients, iatrogenic immunosuppression, chronic myeloid leukemia, and acute leukemia. Of 18 patients, none were diagnosed with AIFRS in the next 90 days, while 2 were diagnosed as granulomatosis with polyangiitis (Wegener), 1 as sinonasal squamous cell carcinoma, and 1 with natural killer/T-cell lymphoma.

Among 30 patients who had a final diagnosis of AIFRS, 7 (23%) patients had at least 1 false negative at the time of intraoperative consultation. Four of 7 patients (patients #: 15, 16, 20, and 23) had multiple specimens and the diagnosis of AIFRS was ultimately made intraoperatively as a result of multiple samples from various sites for each patient. In the other 3 patients (patients #: 31, 47, and 48) with false-negative diagnoses, 1 underwent extensive debridement during the initial surgical procedure because of strong surgical suspicion for AIFRS; the other 2 underwent extensive debridement within 24 hours based upon final pathologic examination. All survived until discharge from the hospital, 20, 36, and 60 days later, respectively. Among 18 patients whose final diagnosis was negative for AIFRS, none had a frozen-section specimen interpreted as positive during intraoperative consultation.

Among the 133 individual specimens (see table in supplemental digital content at https://meridian.allenpress.com/aplm in the June 2021 table of contents), the final pathologic diagnosis was positive for AIFRS in 61 (45.8%) specimens and negative for AIFRS in 72 (54.2%). No false-positive frozen-section diagnoses were made (Figure 1), as all of the specimens receiving a positive diagnosis on frozen section had a positive diagnosis confirmed on permanent section. Of 61 positive cases on permanent sections, 54 (88.5%) were diagnosed with AIFRS on frozen section. Thus, individual frozen sections had a sensitivity of 88.5% (95% CI, 0.78–0.97) and specificity of 100% (95% CI, 0.94–1), with a 100% PPV (95% CI, 0.92–1), and 90.6% NPV (95% CI, 0.82–0.97).

Of 61 specimens that demonstrated AIFRS on permanent sections, 57 (93.4%) demonstrated tissue infiltration by neutrophils, 53 (86.6%) contained regions of necrosis, and fibrinopurulent debris was present in 37 (60.1%). At least 1 of these features was present in 58 positive specimens (95%), but in 3 (5%) positive specimens none of these features were present. These specimens demonstrated only mild chronic inflammation (Figure 3, A through F). Of 72 specimens that were negative for AIFRS on permanent sections, 27 (37.5%) had neutrophilic infiltration, 12 (16.7%) demonstrated regions of necrosis, 22 (30.5%) had fibrinopurulent debris, and 33 (45.8%) had at least 1 of these features.

The 7 false-negative specimens in our study included 5 with features of Mucormycetes and 2 with features of non-Mucormycetes (Figure 4, A through D, and Figure 5, A and B). In 1 of 5 false-negative Mucormycetes specimens, fungal hyphae were retrospectively identified on the frozen-section slide; in this case they were present only near the cauterized edge and somewhat obscured. In the other 4 of 5 false-negative Mucormycetes specimens, no fungi were identified in retrospective review of the frozen section slides. In 1 of 2 false-negative specimens with features of non-Mucormycetes, rare organisms were found after retrospective review. Thus, 2 of 7 false negatives could be classified as interpretive errors, while the other 5 reflected sampling variation. All of the false-negative specimens were submitted entirely for frozen section, so the number of frozen blocks was not a contributing factor.

The reduction in mortality associated with early diagnosis of AIFRS is thought to be related to early aggressive surgical debridement.^27–29  Such an intervention is rarely undertaken without definitive histopathologic diagnosis. Frozen section, as the fastest of the available options for histopathologic confirmation, is an attractive alternative. However, without clarity regarding the accuracy of frozen sections in this context, it is difficult to know how best to incorporate them into treatment protocols.

A few small studies have been published on this question. Ghadiali et al¹³  conducted a retrospective review of 20 patients with AIFRS at the University of Miami between 1993 and 2005. The study revealed overall sensitivity of 84%, specificity of 100%, PPV of 100%, and NPV of 72%. Of note, most of the false negatives (4 of 5) were in patients with Mucor. However, the study was designed to include only patients with a final pathologic diagnosis of AIFRS due to Mucor or Aspergillus species and did not include patients in whom the diagnosis was suspected but subsequently excluded or patients with other genera/organisms. Furthermore, the study was conducted by retrospective review of pathology reports and did not contain analysis of the false-negative specimens.

Papagiannopoulos et al¹⁴  retrospectively reviewed 18 patients and reported a sensitivity of 72.7%, specificity of 100%, PPV of 100%, and NPV of 64.7%. However, like the study from Ghadiali et al,¹³  this cohort consisted only of patients with a final pathologic diagnosis positive for AIFRS. Furthermore, all cases were examined by a head and neck pathologist, a circumstance that may not be feasible in most practices. Last, the study included predominantly patients with profound neutropenia secondary to underlying hematologic malignancies (14 of 18) and very few patients with diabetes mellitus (2 of 18), a context that may limit widespread applicability.

The study design employed in Melancon et al³⁰  was similar to ours in that it sought to include all patients with clinically suspected AIFRS. It included 28 patients in whom 21 had a final diagnosis of AIFRS. Frozen section demonstrated 87.5% sensitivity, 100% specificity, 100% PPV, and 70% NPV. Search terms used in this study may have significantly limited inclusion of negative cases. Moreover, like the Ghadiali et al¹³  study, this was a retrospective review of pathology reports in which pathologic factors affecting interpretation were not explored.

Last, a few very small studies are notable^17,31,32  primarily for the frequency with which they are cited in regard to the question of frozen section sensitivity. Adelson et al¹⁷  report data related to a single patient with invasive Mucor in whom 6 of 6 frozen sections were false negatives. This, together with the Ghadiali et al¹³  study, is sometimes touted as selectively low frozen-section sensitivity for Mucor. Taxy et al³²  identified 8 patients with confirmed cases of AIFRS in whom frozen sections were positive in only 5 (62.5%). The frozen sections were interpreted by rotating general pathologists; however, there was selective use of Romanowsky-stained touch imprints in these cases, and frozen section was performed in only 8 of 12 cases of AIFRS identified during the study period, indicating a possible patient selection bias. Likewise, Hofman et al³¹  studied only patients with final pathologic diagnosis of invasive mucormycosis in whom 6 of 7 (85.7%) had a positive frozen-section diagnosis. As in the study by Taxy et al,³²  frozen sections were performed in only 7 of 12 cases identified during the study period, and there was selective use of adjunctive stains intraoperatively, including Toluidine blue, Gomori Grocott's methenamine silver, and periodic acid-Schiff. Analysis of the false-negative frozen-section discrepancies is not offered in any of the above studies, including the 2—Hofman et al³¹  and Taxy et al³² —that were conducted by pathologists.

Our search identified 53 patients in whom tissue was obtained because of clinical suspicion for AIFRS, and among these 48 (90.5%) had frozen sections performed. We believe that this reflects a practice pattern in our hospital in favor of using frozen sections for this purpose and that their use was not substantially influenced by clinical parameters over and above those informing the decision whether to perform a biopsy at all.

Clinically, our patient group was substantially similar to those reported previously,^{13,14,17,30–32}  that is, patient demographics and underlying medical conditions were comparable, as was the average number of frozen sections performed and rate of AIFRS among the biopsied patients. As in prior studies, we found a preponderance of Mucormycetes and Aspergillus species, but these were included in our study cases of invasive Candida, Alternaria, and Bipolaris species.

As in prior studies, diabetes mellitus was the most common underlying medical condition in our patients with AIFRS.²  The most commonly implicated agents, based upon histomorphology and/or culture, was Mucormycete species. The most commonly biopsied sites in our study were the turbinates, especially the middle turbinate (19% of biopsies), and the most commonly involved (positive) site was the maxillary sinus (21%), both findings consistent with those previously reported.^9,33,34 

In our study, the sensitivity of frozen-section diagnosis was 88.5% per specimen and 90% per patient; the per-patient sensitivity was somewhat higher owing to multiple samplings. The specificity was 100% for both individual samples and patients. These performance characteristics were similar to those reported previously.

The 7 false-negative specimens in our study included 2 that were retrospectively deemed to be interpretive in nature in both cases because of a rarity of fungal hyphae. That is, the fungal structures were visible on H&E-stained frozen sections after thorough retrospective review in only 2 of 7 false-negative specimens. The false-negatives included 5 with Mucormycetes and 2 with Aspergillus species, and the 2 interpretive errors included 1 of each.

The final diagnosis in those patients who were negative for AIFRS is not reported in the prior studies. In our cohort, 2 of 18 negative patients were ultimately diagnosed as granulomatosis with polyangiitis (Wegener), 1 as sinonasal squamous cell carcinoma, and 1 as natural killer/T-cell lymphoma; the remaining 14 were classified as nonspecific rhinosinusitis, and none of these was diagnosed with AIFRS in the next 90 days.

Positive specimens were usually characterized by extensive necrosis, fibrinopurulent debris, and rare to abundant fungal hyphae. When we considered the background histopathologic findings, necrosis was frequently present in specimens positive for invasive fungi (86.6%), while fibrinopurulent debris was often present as well (60.1%). In fact, at least 1 of these findings was present in 95% of specimens containing invasive fungus; however, 5% of such specimens lacked both and demonstrated only the expected intensity of positive specimens. These specimens demonstrated viable tissue with only the expected intensity of chronic inflammation. Nevertheless, these features were significantly more common in the specimens with invasive fungus than those without.

The decision to undertake surgical debridement in patients with AIFRS requires definitive histopathologic diagnosis, for which intraoperative consultation with frozen section may be requested. Our analysis has demonstrated overall high accuracy of frozen sections in this scenario, and this information may be valuable in the surgical management of these patients.