Sinonasal carcinomas (SNC) are rare malignancies and are usually associated with poor outcomes. In the previous editions of the World Health Organization (WHO) Classification, sinonasal malignancies were mainly classified as conventional squamous cell carcinoma (SCC), non-keratinizing SCC, intestinal-type adenocarcinoma (ITAC), non-ITAC, neuroendocrine carcinoma (NEC), poorly differentiated carcinoma (PDCA), sinonasal undifferentiated carcinoma (SNUC), and other rare subtypes (1). The 2022 5^th edition of the WHO Classification of the Head and Neck has made significant classification revisions, with newly added molecular groups for SNC as compared to previous versions (1).

SWI/SNF complex-deficient carcinomas, defined by loss of one of the SWI/SNF complex genes, include two major subtypes: SMARCB1- and SMARCA4-deficient sinonasal carcinoma (2–4). Most of these cases were previously misdiagnosed as PDCA, SNUC, NEC, or teratocarcinosarcoma (TCS). Mutations in IDH2 have also been recently described in a subset of PDCA and SNUC (5, 6). Tumors with these mutations are generally associated with better outcomes relative to those without IDH2 mutations (7, 8); however, results to the contrary have also been reported (9). Because of the rarity of these new entities, we lack a detailed understanding of the clinicopathological features and prognoses between them. This meta-analysis aimed to investigate the clinicopathological characteristics and survival patterns of SWI/SNF-deficient and IDH2-mutant tumors in comparison to the previously described NUT midline carcinoma of the sinonasal tract.

After merging search results from three electronic databases and removing the duplicates, we had 340 studies for title and abstract screening. Following this step, 84 articles were selected for full-text reading. Sixty-one of them met inclusion criteria corresponding to 278 SNCs which were included for analysis (2–5, 9, 11–66) ( Figure 1 ). There were 25 IDH2-mutant, 41 NUT carcinoma, 187 SWI/SNF loss, and 25 triple negative SNCs. Among NUT carcinoma, BRD4:NUTM1 was the most common variant and only one case harbored BRD3:NUTM1 rearrangement. The R172 variant was the most predominant IDH2-mutant genotype. Regarding SNCs with inactivation of one of the SWI/SNF complex genes, SMARCB1-deficient carcinoma was the most frequent subtype followed by SMARCA4-deficient tumors. Loss of SMARCA1, SMARCA5, and SMARCE1 were also found in one SNC case each. IDH2 mutations, NUTM1 rearrangement, and inactivation of SWI/SNF complex were mutually exclusive with each other.

In recent years, new molecular profiles of SNCs have been further clarified and novel molecular groups have been incorporated into the latest WHO classification of SNCs (1–6). Prior to the molecular era, most IDH2-mutant, NUT midline, and SWI/SNF-deficient SNCs were categorized as SNUC, PDCA, TCS, or NEC (2, 3, 5, 33). In the latest WHO classification of head and neck tumors, NUT midline, SMARCA4-deficient, and SMARCB1-deficient SNCs have been recognized as separate entities. Given their distinct clinicopathological features and prognoses as compared to IDH2-wild type tumors (8), IDH2-mutant SNCs may nevertheless be regarded as a distinct molecular group in future WHO editions. Because of the rarity of SNCs, most data were presented as case reports or small- to medium-sized case series. The clinicopathological features and prognostic outcomes of new molecular groups of SNCs have been described. However, it is still controversial as to how these tumors are different from each other and in how clinicians can better assess patient outcomes. In this study, we integrated IPD of published studies into a meta-analysis to improve the statistical implication compared to cohort studies with limited sample size.

Our results showed that these tumors were uniformly high-grade and distributed in diverse histopathological spectrums with SNUC and PDCA being the most common variants. All molecular groups of SNCs had a relatively considerable risk for tumor metastases to distant organs with bone and lung being the most common sites. We found that IDH2-mutant SNCs were most likely to metastasize to liver and other rare organs compared to other groups. Like prior studies, our meta-analysis demonstrated improved survival of IDH2-mutant SNCs as compared to those without these mutations (5, 6, 8). The prognostic implication of IDH1/2 mutations in gliomas, chondrosarcoma, and cholangiocarcinoma have similarly been established (67–70). The discovery of IDH2 mutation in SNCs provides a promising opportunity for targeted therapy with IDH inhibitors. Most IDH2-mutant SNCs occur in codon 172 and can be diagnosed by immunohistochemistry assay which is an accessible, rapid, and inexpensive method (8, 31).

This meta-analysis also highlighted that SNCs usually present at older age except for NUT carcinoma, which is more commonly seen in young adults and pediatric patients. NUT carcinoma is exceedingly rare, typically occurs in the midline structures, and histopathologically resembles PDCA. This tumor is characterized by a chromosomal rearrangement involving NUTM1 gene (32). The availability of NUT immunohistochemistry antibody has improved the accuracy of NUT midline carcinoma diagnosis and differentiated them from other PDCA. NUT midline carcinoma is associated with high rates of mortality (71) and our study further confirmed the uniformly poor prognosis of these tumors compared to other genetic groups of SNCs.

The most common genetically defined group of SNCs involves the SWI/SNF complex genes with loss of SMARCB1 and SMARCA4 being the most common variants. It is still poorly understood regarding how these subtypes are different from each other. SMARCB1-loss SNC is associated with rhabdoid differentiation in SNUC (2), which is an important diagnostic parameter to differentiate them from other PDCA. On the other hand, recurrent loss of SMARCA4 is commonly observed in TCS (3) and SNCs with neuroendocrine differentiation (4). Our analyses further confirmed these histopathological associations. We also found that SMARCB1-deficient SNCs occur at a significantly older age and more likely to have radiotherapy administration in comparison to SMARCA4-deficient tumors. From our analysis, the DSS of these two new SNC entities were comparable. With the use of immunohistochemistry, it is easier to recognize these two rare entities and separate them from other sinonasal PDCAs.

This study is the first meta-analysis comparing the new molecular groups of SNCs in the new edition of WHO classification. It helps summarize and facilitate our current understanding about the clinicopathological behaviors and prognoses of these aggressive tumors. However, there are certain limitations. First, all included studies are retrospective cohort studies or case reports/series leading to inevitable selection bias. Next, we could not include other recently described molecular entities such as DEK::AFF2-rearranged non-keratinizing SCC and TP53-mutant ITAC due to limited data. In addition, we could not compare the effectiveness of treatment modalities in each molecularly defined SNC subgroup due to missing data. Finally, we could not assess PFS, an important prognostic value due to missing data in most included studies. Future large multicenter prospective studies are essential to validate the results of this study.

In summary, the evolution of molecular pathology alongside standard immunohistochemistry enables us to recognize and accurately diagnose novel molecular entities of SNCs. These tumors have distinct clinicopathological profiles and prognoses and should be distinguished from other SNCs to better understand their unique natural histories and treatment implications.

HV: conceptualization, data curation, formal analysis, investigation, methodology, project administration, software, validation, writing original, review, and editing. TL: data curation, formal analysis, investigation, methodology, review, and editing. TTBL: data curation, formal analysis, investigation, methodology, review, and editing. HL: data curation, formal analysis, investigation, methodology, review, and editing. EE-R: data curation, formal analysis, investigation, methodology, review, and editing. KM: data curation, formal analysis, investigation, methodology, review, and editing. ID: conceptualization, project administration, validation, review, editing, and supervisions. All authors contributed to the article and approved the submitted version.