Odontogenic Carcinosarcoma: Clinicopathologic and Molecular Features of Three Cases, a Literature Review and Nomenclature Proposal

Abstract

Background

Odontogenic carcinosarcoma (OCS) is a rare odontogenic malignancy with limited characterization and unexplored molecular features. We report clinicopathologic and molecular findings in 3 additional OCS and review the literature.

Methods

3 OCS (5.1%) were identified among 59 malignant odontogenic tumors (in our archives from 1992 to 2022). Clinical, radiologic, histopathologic, immunophenotypic, and molecular findings were reviewed. Data from prior case reports and systematic or non-systematic reviews were extracted for analysis.

Results

Three mandibular OCS (age range: 66 to 72 years; 1 male, 2 females) were identified. Case 1 had novel clear-cell morphology, multiple recurrences, and a lethal outcome 28 months after resection. EWSR1 rearrangements were negative, but the tumor showed focal nuclear β-catenin and strong LEF-1 immunoreactivity. Case 2 demonstrated ameloblastic and sclerosing features and encased the inferior alveolar nerve; the patient was disease-free 22 months after resection with adjuvant chemoradiation therapy. LEF-1 was again strongly positive, and next-generation sequencing demonstrated 9p region-(CDKN2A, CDKN2B) copy number loss, and 12q region-(MDM2, CDK4) copy number gain. Case 3 showed clear-cell and markedly sclerosing features; no follow-up information was available. Literature review along with the current cases yielded 20 cases. OCS showed a male predilection (1.5:1), mandibular predominance (80%, typically posterior), and a bimodal age distribution (modes: 27.7 years, 62.7 years). OCS presented as masses (100%), often with pain (55%), and paresthesia (45%). Tumors were typically radiolucent (88.9%), with bone destruction (61.1%), and/or tooth effacement (27.8%). Preoperative biopsy was sensitive for malignancy (85.7%). At least 45% show evidence for a precursor lesion. 3-year DSS and DFS were 58% and 35%, respectively. Regional and distant (usually lung) metastatic rates were 25% and 31.3%, respectively. Increased mitotic rates and presence of tumor necrosis trended toward worse DSS and DFS.

Conclusion

OCS is a rare but aggressive malignancy, often arising from precursor tumors and may represent a terminal phenotype rather than a distinct entity. We describe novel clear-cell and sclerosing morphologies. Wnt pathway alterations appear important. Mitotic rates and necrosis may be adverse prognosticators. In keeping with nomenclature trends in other sites, OCS may be more appropriately designated as “biphasic sarcomatoid odontogenic carcinomas.”

Supplementary Information

The online version contains supplementary material available at 10.1007/s12105-023-01569-3.

Introduction

Malignant odontogenic tumors are very rare, comprising 1–6.1% of all odontogenic tumors. Within this group, odontogenic carcinosarcoma (OCS) is even rarer, first recognized in the second edition of the World Health Organization (WHO) classification of odontogenic tumors in 1992 as “a tumor architecturally resembling an ameloblastic fibrosarcoma in which both the epithelial and the connective tissue components show cytologic evidence of malignancy”[1–3]. Although this entity disappeared in the 2005 WHO 3rd edition [4], it resurfaced in the WHO 4th edition [5], and is now defined in the WHO 5th edition more broadly as “a malignant mixed odontogenic tumor in which the epithelial component as well as the ectomesenchymal component show features of malignancy” [6].

Less than twenty cases of OCS are reported in the literature [3, 7–22]. Prototypically, OCS is expected to be a clinically and histologically aggressive lesion with a high recurrence and metastasis rate, but this understanding is hampered by the rarity of this entity. Furthermore, on limited sampling, both the epithelial and mesenchymal components may appear deceptively bland, posing a diagnostic challenge [2]. While molecular underpinnings for a variety of odontogenic lesions have been elucidated,[23] OCS remains uncharted in this regard. We report the clinicopathologic, immunohistochemical, and molecular features of three new cases of OCS retrieved from our files and perform an in-depth literature review.

Materials and Methods

This study was approved by our institutional review board (IRB: STUDY19030431).

Case Selection

The pathology archives of UPMC were queried for malignant central odontogenic tumors with attention to OCS from 1992 to 2022. Specific search terms used to identify OCS included “odontogenic carcinosarcoma,” “ameloblastic carcinosarcoma,” “carcinosarcoma,” and “sarcomatoid carcinoma.” Additionally, “ameloblastic fibrosarcoma,” and “intraosseous carcinoma, not otherwise specified (NOS)” were searched and reviewed for potential reclassification as OCS.

OCS cases selected for our study met the following inclusion criteria:

• Gnathic location.

• Odontogenic morphology.

• Malignant epithelial and spindled components in a biphasic distribution.

In addition, the following exclusion criteria were applied:

• Mucosal/cutaneous surface dyplasia.

• Salivary gland epicenter.

Immunohistochemistry

Immunohistochemistry for differentiation markers (i.e., epithelial, myoepithelial, and stromal markers) was performed by case contributors and/or our own laboratory using standard protocols. Select distinctive biomarker/molecular surrogate antibody characteristics are summarized in Table 1.

Table 1.

Select antibody characteristics

Antibody	Clone	Dilution
β-catenin	β-catenin-1	1:250
BRAF V600E	VE1	Pre-dilute
Ki-67	MIB1	Pre-dilute
LEF-1	EP310	Pre-dilute
p53	DO-7	Pre-dilute

High-Risk Human Papillomavirus (HPV) E6/E7 mRNA in situ hybridization

In situ hybridization was performed on Case 2 for high-risk HPV E6/E7 mRNA (Advanced Cell Diagnostics (RNAscope®), Hayward, CA). The probe, HPV-HR7, hybridizes to E6/E7 mRNA of seven highrisk HPV subtypes (HPV 16, 18, 31, 33, 35, 52, and 58).

Fluorescence in situ Hybridization (FISH)

Break-apart FISH for EWSR1 rearrangements was performed (LSI EWSR1 Dual-Color Break-apart Probe 22q12, Abbott Molecular, Des Plaines, IL) using Leica Biosystems (CytoVisionFISH Capture and Analysis Workstation, Buffalo Grove, IL) on Case 1. At least 60 nonoverlapping nuclei were included in the analysis. The translocation was denoted by a combination of separate/split orange and green signals and 1 intact EWSR1, indicated by juxtaposed orange and green probes and resulting in a yellow signal. Cases that showed disruption of EWS in > 12.57% of tumor cells were considered positive for a translocation.

Next-Generation Sequencing (NGS)

NGS was performed for Case 2 using a targeted Oncomine™ Comprehensive Assay v3 on an Ion Torrent™ NGS platform (both Thermo Fisher Scientific, Waltham, MA, USA) according to manufacturer’s instructions. A detailed list of genes tested is available online (https://mgp.upmc.com/Home/Print/Oncomine). This panel evaluates 161 cancer-relevant driver genes to include 760 fusion genes. Briefly, total DNA and mRNA are reverse transcribed into cDNA and subjected to multiplex PCR to amplify the regions of interest. Amplicons were barcoded, ligated with specific adapters, and purified. RNA library quantity and quality check were performed using the 4200 TapeStation (Agilent Technologies, Santa Clara, CA). The Ion Chef was used to prepare and enrich templates and enable testing via Ion Sphere Particles on a semiconductor chip. Massive parallel sequencing was carried out on an Ion GeneStudio S5 Prime System according to the manufacturer’s instructions (Thermo Fisher Scientific, Waltham, MA), and data was analyzed with variant explorer (UPMC) for single-nucelotide variants (SNV), insertions, deletions, copy number alterations, and RNA fusion genes. The limit of detection of this DNA/RNA assay was 3–5% allelic frequency, roughly translating to 10–20% neoplastic cellularity.

Literature Review

At least two systematic reviews [2, 16] and two semi-systematic reviews [19, 22] were previously performed. As such, the cases in these reviews were searched and crosschecked to identify all potential cases for inclusion in the present review. In addition, a PubMed search (conducted 4/4/2023) for cases subsequent to the most recent review [22] was performed. Non-English papers were translated directly or using web-based service https://smartcat.com. Inclusion requirements for our review were fulfillment of the criteria listed above (see “Case Selection” section) based on photomicrographs and pathologic description. Cases that did not meet these criteria were excluded even if included in prior reviews. Detailed epidemiologic, clinical, radiologic, surgical, morphologic, and immunohistochemical data were extracted and tabulated in a Masterfile (available on request). Inconsistencies, corrections, and clarifications were highlighted and annotated in this file. Extracted data were then summarized.

Key clinical features extracted were site, duration of symptoms, presence of mass, pain, paresthesia, prior procedure at site (including prior tumors), and other notable history (i.e., other malignancies). Key radiologic features extracted were imaging modality, radiographic density, presence of bone destruction, and tooth resorption. Incisional biopsy diagnoses were recorded descriptively. Definitive treatment details extracted were surgical procedure, including associated soft tissue removal, wide local excision/neck dissection, chemotherapy, and/or radiotherapy details.

Gross size and appearance were extracted. With regards to histologic description, OCS cases were subcategorized based on their resemblance to other malignant odontogenic tumors as follows:

• Ameloblastic carcinoma (AC)-like—These cases show malignant epithelial components with follicular, plexiform, or solid growth pattern. There is peripheral palisading with reverse nuclear polarization and central stellate reticulum-like differentiation. Acanthomatous change is acceptable.

• Ameloblastic fibrosarcoma (AFS)-like (i.e., WHO 1992 type[1])—These cases show malignant epithelial components composed of thin cords and trabeculae with peripheral palisading and reverse nuclear polarization, but limited or no stellate reticulum.

• Acanthomatous or squamous cell carcinoma (SCC)-like—These cases show a predominance of an acanthomatous malignant component that would be considered squamous cell carcinoma in a non-odontogenic context, with only limited ameloblastic epithelial areas.

• “Clear cell odontogenic carcinoma (CCOC)-like”—These cases show sheets, nests, and trabeculae composed of polygonal, clear, somewhat squamoid to occasionally eosinophilic cells with stromal sclerosis.

• “Sclerosing odontogenic carcinoma (SOC)-like.”—These cases show compressed strands and cords of malignant epithelial cells embedded in a dense fibrocollagenous stroma.

Based on reported descriptions, mitotic counts were categorized as “occasional,” “moderate,” or “numerous” as follows:

• Less than 5 per 2 mm²: occasional.

• 6–10 per 2 mm²: moderate.

• 11 or more per 2 mm²: numerous.

For the purpose of this review, 10 high-power fields were accepted as an equivalent to 2 mm².

Other extracted histologic parameters were necrosis, perineural invasion (PNI), lymphatic/vascular invasion (LVI), margin status, and evidence of a benign precursor lesion. Immunohistochemical findings were summarized by collapsing select individual markers into the following groups:

• Pan-cytokeratins and low molecular weight cytokeratins.

• High molecular weight keratins and the basal markers p63 and ∆Np63 (p40).

• Muscle markers (desmin, smooth muscle actin, all muscle actin, smooth muscle myosin heavy and light chains).

• S100 and SOX-10.

• Wnt pathway markers (β-catenin, LEF-1, E-cadherin).

Other markers were listed singly or under “other.”

Extracted outcome data included presence or absence of recurrence, recurrence type (local, regional, or distant), number of recurrences, time to first recurrence, time to distant metastasis, status at time of publication, and total follow-up time. Starting time (t₀) was defined as the date/time of definitive surgery for which a malignant diagnosis was confirmed. Additional therapy and secondary complications were summarized descriptively. Death from disease was defined as both deaths directly related to disease and consequent to disease (i.e. postoperative sequelae).

All extracted data were summarized using Microsoft Excel (Microsoft® Excel® 2016 MSO, Version 2302 Build 16.0.16130.20298, 32-bit) for basic descriptive statistics, simple arithmetic operations, and data manipulation. For histograms, density distribution curves, Clopper-Pearson exact binomial 95% confidence interval for proportions, bimodality testing, and Kaplan-Meier estimates with log rank tests, R programming language was used [R Core Team (2021), R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, https://www.R-project.org/, R version 4.2.1 (2022-06-23)]. The following key unique packages were used: LaPlacesDemon (v16.1.6, Statisticat, LLC 2021); survival (v3.5-5, TherneauT 2023); and ggsurvfit (v0.3.0, Sjoberg D et al. 2023).

Results

Three OCS were identified, comprising 5.1% [95% confidence interval (CI): 1.1–14.2%] of all confirmed malignant odontogenic tumors during the study period (Fig. 1).

Fig. 1.

Distribution of malignant odontogenic and central tumors at UPMC by category. Abbreviation: NOS not otherwise specified

Case 1

A 72-year-old male presented with a 2 cm anterior mandibular mass associated with the region of teeth 22–26, which included with failed implant #23. Over the course of 17 months, tumor size increased to 3 cm. On physical examination, the mass was red, ulcerated, and infiltrative. On panoramic radiograph, the lesion was expansile and radiolucent with cortical destruction. No lymphadenopathy was noted. An incisional biopsy was performed and was received by us in consultation.

The biopsy contained two fragments (Fig. 2A), one superficial with mucosa and one deep. The superficial mucosal component showed nests of somewhat monomorphic clear to squamoid cells reminiscent of clear cell odontogenic carcinoma (Fig. 2B). No surface dysplasia was noted. The deeper fragment showed more irregular nests and cords with central acanthomatous change and more profound nuclear enlargement, palisading, and hyperchromasia with central acanthomatous change (Fig. 2C); also present was a spindled component with a range of fusiform, ovoid, and polygonal tumor cells exhibiting moderate nuclear size variation (2D). Mitotic counts were up to 12 per 2 mm² and focal necrosis was noted.

Fig. 2.

Case 1 biopsy. A Surface and deep components on biopsy. B A superficial clear cell carcinoma-like component that C transitions to a more atypical squamoid/acanthomatous component. D Other areas are biphasic with a polygonal spindled component and necrosis. E The tumor shows nuclear β-catenin in spindled and scattered epithelial cells. F LEF-1 shows strong positivity in the spindled component with some epithelial cell staining

Immunohistochemistry performed by the contributor showed the nested epithelial component to be positive for cytokeratin AE1/AE3, CK19, and ∆Np63 (p40); PAX8, CK7, CK20, desmin, and smooth muscle actin were negative in both components. Ki-67 proliferation index ranged from 10 to 25% and was more prominent in the spindled component. Additional immunohistochemistry performed in our laboratory demonstrated nuclear as well as membranous/cytoplasmic β-catenin (Fig. 2E) in both components as well as LEF-1 reactivity (Fig. 2F, stronger in spindled component). Calretinin was focally positive. Interestingly, while moderate ERG reactivity was noted in the spindled component, CD31 and CD34 were negative. We observed no reactivity for SOX-10, STAT6, HMB-45, or MUC4. INI-1 was retained, and only cytoplasmic WT1 was noted in the spindled component. FISH for EWSR1 rearrangements was negative.

A diagnosis of “odontogenic carcinosarcoma, at least low grade” was rendered. The patient was then referred to the Head and Neck Surgery Service at the originating hospital approximately two months later. Computerized tomography (CT) corroborated the prior clinical and radiologic findings, and the mass was found to be fluorodeoxyglucose-¹⁸F-positron emission tomography (FDG-PET) avid with an indeterminate right tonsillar uptake. The patient underwent a segmental mandibulectomy with right tonsillectomy and bilateral modified radical neck dissection. The tumor grossly measured 4.0 × 2.3 × 1.8 cm and consisted of an ulcerated, irregular, tan-gray mass obliterating the mandible. Microscopic examination showed predominantly clear cell carcinoma morphology with a minor spindle cell component (Fig. 3A) and PNI but no LVI. All margins were negative, although the posterior soft tissue margin was close at 0.1 cm. All 105 dissected lymph nodes were negative, and the right tonsil was uninvolved.

Fig. 3.

Case 1 resection and recurrences/metastases. A Resection showing obliteration of bone and extension into submucosa. Inset: Biphasic morphology with a CCOC-like epithelial component. B Neck recurrence demonstrating biphasic morphology and obliteration of a lymph node. C Parotid recurrence is showing spindle cell predominance

The patient subsequently underwent radiotherapy (6600 cGy over 33 fractions spanning 6 weeks) and tolerated the treatment well. However, 13 months after definitive resection, the patient was noted to have two persistent left lower lobe lung masses. Biopsy demonstrated a spindle cell malignancy (not available for our review) for which he received radiotherapy (5000 cGy over 5 fractions). 20 months later, the patient was noted to have a regional recurrence in the right neck level II with a spindle cell predominance. As per the pathology report, the spindle cell component of the lung metastasis matched that of the resection and the right neck level II lesion (Fig. 3B), confirming distant and regional metastases. In less than a month, the patient developed a third recurrence in the right parotid (Fig. 3C). The patient unfortunately succumbed to his disease 28 months after resection.

Case 2

A 66-year-old female presented with a 12-month history of progressive left jaw and neck pain and with “electrical sensation”-type paresthesia along a V3 distribution. Subsequent examination demonstrated a left mandibular mass. On CT (Fig. 4A) a “ground glass” radiolucent/radiopaque expansile mass involving the left posterior mandible and extending from the angle to the mandibular foramen was noted; the radiographic impression was fibrous dysplasia. The lesion borders showed “focal bony fenestration and dehiscence” but no overt infiltration. The tumor was also FDG-PET avid (Fig. 4B). An incisional biopsy was performed and provisionally diagnosed as an “invasive carcinoma with squamous, spindle cell, and clear cell features,” raising the possibility of either a squamous cell carcinoma, salivary-type or odontogenic carcinoma with molecular testing ordered. The patient then underwent a left hemimandibulectomy and selective neck dissection (levels I–IV).

Fig. 4.

Case 2 radiologic and gross appearance. A CT demonstrates a ground glass mixed radiolucent-radioopaque expansile left posterior mandibular mass with cortical disruption. B PET-CT shows strong avidity. C Sagittal section through hemimandibulectomy specimen showing a firm ill-defined white lesion extending into soft tissue. Inset: Tumor effaces medullary cavity and encases the inferior alveolar nerve (arrow)

On gross examination, the tumor consisted of a 4.0 × 3.0 × 1.8 cm white-gray, firm, fibrotic tumor, effacing the entire mandible with focal extension through the cortex and adherence to the adjacent soft tissue. The tumor showed extensive involvement of the mandibular canal, surrounding the inferior alveolar nerve (Fig. 4C). A prior biopsy site scar was noted on the mucosa, which was otherwise uninvolved.

On microscopic examination, the tumor demonstrated a biphasic proliferation consisting of spindle cells with fascicular to storiform growth embedded in a fibrocollagenous stroma as well as compressed epithelial nests and cords with clear cell change and discrete ameloblastic nodules. The tumor showed extensive perineural invasion, enwrapping the inferior alveolar nerve as grossly noted, as well as extension through bone, impart a fibrous dysplasia like remodeling effect (Fig. 5A). Foci of myxoid ectomesenchymal-like stroma were noted as well (Fig. 5B). The spindled component was relatively monomorphic but still showed considerably enlarged and angulated nuclei with hyperchromasia and clear cell change (Fig. 5C). No tumor necrosis or lymphatic/vascular invasion were noted, and the maximal mitotic rate was 6 per 2 mm². Margins were free. All 50 lymph nodes were negative for this tumor. However, two lymph nodes (one level II, and one level IV) showed incidental metastatic papillary thyroid carcinoma.

Fig. 5.

Case 2 morphologic features. A Invasive biphasic tumor invading bone imparting a fibrous dysplasia like appearance and invading the inferior alveolar nerve (right) B Ameloblastic nodule juxtaposed to myxoid spindled component. C Bone invasion and more haphazard pattern and sclerosis with ropey collagen. D AE1/3 and E. P40 show stronger staining in epithelial components. F LEF-1 strong positivity throughout

Immunohistochemically, the epithelial components were strongly positive for cytokeratin AE1/3 (Fig. 5C), p40 (Fig. 5D), and CK 5/6. The spindled components also showed reactivity, albeit in a weaker and more heterogenous fashion. LEF-1 was diffusely and strongly positive in both components (Fig. 5E), and calretinin was focally positive. P16 was diffusely positive, but high-risk human papillomavirus (HPV) E6/E7 mRNA in situ hybridization was negative. Smooth muscle actin, SOX-10, calponin, smooth muscle myosin heavy chain, ALK-1, STAT6, PAX-8, and SATB2 were negative. The Ki-67 proliferation index was 10–15%.

Immunohistochemistry showed a non-clonal (wild-type) p53 expression pattern, and no reactivity for BRAF V600E. H3K27me3 and ATRX1 showed no loss. NGS performed on the biopsy showed no mutations or gene fusions but demonstrated the following copy number alterations: 9p region-(CDKN2A, CDKN2B) copy number loss, 12q region-(MDM2, CDK4) copy number gain (~ 4 copies).

A diagnosis of odontogenic carcinosarcoma was rendered. The patient subsequently received adjuvant chemotherapy (6 cycles of cisplatin, nanoparticle albumin–bound paclitaxel, and carboplatin) and radiotherapy (6000 cGy over 30 fractions spanning 6 weeks). She had no evidence of odontogenic carcinosarcoma 22 months after treatment. In addition, the patient’s thyroid was evaluated by ultrasound and demonstrated nodular disease. A right level V lymph node and right 1.6 cm thyroid nodule were aspirated and showed no evidence of thyroid malignancy (thyroid aspirate was diagnosed as atypia of undetermined significance, Bethesda III; molecular testing negative).

Case 3

A 73-year-old female presented with a left mandibular mass. CT demonstrated a radiolucency involving the roots of teeth 19–21. An incisional biopsy was performed and provisionally diagnosed as an odontogenic fibroma. Three months later, the tumor was enucleated along with peripheral ostectomy, and teeth 19–21 were extracted. The enucleation grossly consisted of three soft, tan-pink tissue fragments and an associated bone fragment, measuring 3.2 × 3.0 × 1.5 cm in aggregate.

Microscopically, the tumor was infiltrative, extending into skeletal muscle (Fig. 6A). It was again biphasic, demonstrating an epithelial component with corded, reticular anastomosing growth of hyperchromatic clear cells (Fig. 6B) and a juxtaposed spindled component with fascicular and storiform growth pattern, moderate nuclear size variation, and nuclear hyperchromasia (Fig. 6C). Stroma ranged from myxoid to mainly collagenous. Osteoid-like matrix was focally noted, surrounding epithelial cords (Fig. 6D), and focal epithelial cell dyshesion imparting a pseudovascular appearance (Fig. 6E) was also identified. Mitotic counts were up to 10 per 2 mm², and there was no necrosis.

Fig. 6.

Case 3 morphologic features. A A biphasic tumor with myxoid and sclerosing components invading extending into skeletal muscle. B The epithelial component shows a reticular clear cell anastomosing/plexiform growth. C The spindled component is embedded in a collagenous stroma with a paucicellular appearance and storiform pattern. D Primitive osteoid-like matrix is noted (right). E Epithelial elements focally demonstrate a pseudovascular pattern highlighted by F CAM 5.2, and G p63. Only rare spindled elements are positive for these markers

Immunohistochemical stains for CAM 5.2 (Fig. 6F), p63 (Fig. 6G), and CK 5/6 preferentially highlighted the epithelial components, including pseudovascular areas, and were only focally weakly positive in the spindled components. BCL-2 was positive in both components. CD34, ER, PR, and mammaglobin were negative.

Originally, a diagnosis of “spindle cell (sarcomatoid) carcinoma, favored to be of odontogenic type,” was rendered. (Of note, this diagnosis was made after the WHO 3rd edition, in which the diagnosis of OCS was removed, and prior to the WHO 4th edition, in which it was re-instated. Upon case review, we reclassified the tumor as OCS.) The patient was lost to follow-up.

Literature Review

Case Selection and Correction of Errata

Review of the literature, including crosscheck of two systematic reviews [2, 16] and two semi-systematic reviews [19, 22], in combination with our 3 cases resulted in 23 potential cases [3, 7–22, 24–26]. The following 3 cases were excluded from the present review:

• Despite being included in two of the reviews [2, 16], the case reported by Hosokawa et al., 1991 [24] was excluded. Review of the clinical, radiologic, and morphologic features suggested that this case represented an epithelioid osteosarcoma that had dedifferentiated from a lower-grade precursor. The tumor was not actually biphasic; although it showed aberrant keratin expression, this phenomenon has been well described in osteosarcomas [27].

• The “central pseudosarcomatous carcinoma” reported in Japanese by Matsuda et al., 1983 [25] showed morphology somewhat compatible with OCS. However, the translated report references parotid involvement and does not appear to adequately address the possibility of this being the site of origin.

• Finally, one potential OCS arising from a premature odontoma was identified in an open research publishing platform by Salem et al., 2020 [26]. On review, the tumor was felt to be better classified as an ameloblastic fibrosarcoma. Notably, the Ki-67 proliferation index was lower than that of other reported cases.

One case not traditionally included in reviews but flagged as biphasic by Slater, 1999 [3] was a tumor reported by Yoshida et al., 1989[7] as “odontogenic carcinoma with sarcomatous proliferation.” Our review of this paper suggested that it was sufficiently within the spectrum of other reported cases to be included. Notably, the first reported OCS appeared in the Turkish literature in 1981, was reprinted in 2013 (Tahsınoğlu et al. [21]), and was included in the recent report and literature review by Awang Hasyim et al. [22].

Overall, 20 OCS cases met our study criteria; the epidemiologic, clinical, radiologic, and pathologic findings of these cases are summarized in online resource Tables 1, 2, 3, and 4. Our review was concordant with prior reviews with a few key exceptions: (1) correction of outcome for one case (translated lost to “view” (i.e., follow-up) but not dead of disease at 36 months as stated in subsequent reviews) [10]; (2) correction of calculated follow-up/time of event for one case, (disease-specific survival: 76 months not 64 months) [11]; (3) conflation of misdiagnosis of ameloblastoma with a precursor of ameloblastoma in one case [13]; and (4) inconsistent classification of precursors for three cases [3, 10, 11].

Epidemiologic, Clinical-Radiologic, and Preoperative/Operative Characteristics

Epidemiologic, clinical, radiologic, preoperative, and operative characteristics are summarized in Table 2, derived from online resources 1 and 2. These are overall in line with prior reviews [2, 16, 19, 22], demonstrating a male predilection, overwhelming mandibular site predominance, and presentation as a mass lesion. Interestingly, age data demonstrate a somewhat bimodal distribution (modes: 27.7 years, 62.7 years, local minimum: 44.4 years). Pain (55%) and paresthesia (45%) were frequent findings, and a subset presented with ulceration (25%) and loosening or loss of teeth/implants (20%). Lymphadenopathy was rare (5%). Other notable findings included the following: one patient [12] was pregnant during tumor treatment and delivered without complication; one patient [14] also had a right tibial fibro-osseous lesion; one patient [17] had a prior ipsilateral parotid rhabdomyosarcoma treated with radiation 13 years prior to presentation; and one patient (our Case 2) had incidental metastatic papillary thyroid carcinoma with lymph node metastasis.

Table 2.

Clinical, radiologic, preoperative and operative characteristics of current and reported OCS

Clinical and demographic features
Median age (range) (yrs)	47 (9–79)*
M:F (ratio)	12:8 (1.5)
Laterality	Left:12; Right:8
Site
Mandible	17/20 (85.0%)
Posterior	15 (88.2%)
Anterior	2 (11.8%)
Maxilla	3/20 (15.0%)
Median duration of symptoms (range) (mo)	6 (0.3–144)
Signs and symptoms
Mass	20 (100%)
Pain	11/20 (55.0%)
Paresthesia	9/20 (45.0%)
Ulceration	5/20 (25.0%)
Bony deformation/loose or lost teeth/implants	4/20 (20.0%)
Lymphadenopathy	1/20 (5.0%)
Prior tumor	3 (15.0%)	Unknown (24 mo) Ameloblastic fibroma (120 mo) Ameloblastoma (120 mo)
Post-irradiation	1 (5.0%)	History of rhabdomyosarcoma

Radiologic findings
Imaging reported	18/20 (90.0%)
Panoramic radiograph	11 (61.1%)
CT	14 (77.8%)
MRI	2 (11.1%)
FDG-PET	2 (11.1%)
Other (plain film; specimen X-ray)	2 (11.1%)
Radiodensity
Radiolucent	16 (88.9%)
Mixed	2 (11.1%)
Aggression
Bone destruction	11 (61.1%)
Tooth resorption/destruction	5 (27.8%)

Preoperative and operative characteristics
Preoperative biopsy (presenting tumor)	14/20 (70.0%)
Malignant	12 (85.7%)
OCS or biphasic	5 (35.7%)
Resection type known	18/20 (90.0%)
Mandible (n = 16)
Hemimandibulectomy	6 (37.5%)	Skin/mucosa/soft tissue wide excision: 6/13 (46.1%)
Segmental or partial mandibulectomy	7 (43.8%)
Marginal/enucleation or currettings	2 (12.5%)
Debulking	1 (6.3%)
Neck dissection	5 (31.3%)
Maxilla (n = 2)
Hemimaxillectomy	1 (50.0%)
Endoscopic resection	1 (50.0%)
Chemotherapy	3/18 (16.7%)
Radiotherapy	3/18 (16.7%)

yrs years; M male; F female; mo months; CT computerized tomogram; MRI magnetic resonance imaging; FDG-PET fluorodeoxyglucose-¹⁸F -positron emission tomography; OCS odontogenic carcinosarcoma

*Bimodal distribution–modes: 27.7 years, 62.7 years, local minimum: 44.4 years

Radiologic findings were reported in 18/20 (90.0%) cases. CT (77.8%) and Panoramic radiograph (61.1%) were the most frequently reported radiologic studies. Most cases presented with destructive radiolucent lesions; only two cases (11.1%) showed a mixed radiolucent-radiopaque appearance. At least one preoperative incisional biopsy was performed in 70% of cases. Biopsy was modestly sensitive in detecting malignancy (85.7%), but a specific diagnosis of OCS or biphasic malignancy was rendered in only 35.7% of cases.

Surgical treatment was reported for 18/20 (90.0%) patients and represented the mainstay of treatment. Most patients with mandibular tumors underwent a hemimandibulectomy (37.5%) or partial/segmental mandibulectomy (43.8%), and of these almost half (46.1%) had an associated wide excision of skin, soft tissue, and/or mucosa. Slightly under a third of patients (31.3%) with mandibular tumors underwent neck dissection. For the two patients with maxillary tumors for whom surgical information was available, one underwent a hemimaxillectomy, and one underwent endoscopic resection. In addition to surgery, one sixth of patients received chemotherapy and/or radiotherapy (dosage range: 3000 cGy−66,000 cGy).

Pathologic Features

Pathologic features are summarized in Table 3 derived from Online Resource 3 and are overall similar to prior reviews [2, 16, 19, 22]. Median gross size was 5.0 cm (range 3.2–8.0 cm), and appearance ranged from soft and fleshy pink to firm and fibrous white. Gross infiltration and bone destruction were characteristic. Detailed histopathologic descriptions were extensively reported previously [2, 16, 19, 22]; collectively, tumors showed various combinations of ameloblastic, clear cell, and acanthomatous/squamous epithelial components, and spindled components that ranged from myxoid (ectomesenchymal-like) to heavily sclerotic. By subgroup, the majority (65%) could be subclassified as AC-like, a subset (23.1%) of which of which had acanthomatous elements. Three cases (15%) were AFS-like. Other less common morphologies included acanthomatous (SCC)-like, CCOC-like, and SOC-like (each noted at 10%). Temporal and/or histologic evidence for a benign precursor was noted in 45% of cases, with ameloblastoma being the most frequent precursor type.

Table 3.

Pathologic characteristics of current and reported OCS

Pathologic parameters
Median gross size (range) (cm)	5 (3.2–8)
Provisional subtypes (n = 20)
AC-like	13 (65.0%)
Acanthomatous	3/13 (23.1%)
AFS-like	3 (15.0%)
Acanthomatous (SCC-like)	2 (10.0%)
CCOC-like	2 (10.0%)
SOC-like	2 (10.0%)
Mitoses reported	18 (90.0%)
Occasional	4 (22.2%)
Moderate	4 (22.2%)
Numerous	10 (55.6%)
Atypical mitoses	6 (30.0%)
Necrosis	6 (30.0%)
Matrix
Dentinoid	1 (5.0%)
Equivocal (Osteoid)	3 (15.0%)
PNI, LVI reported	3 (15.0%)
PNI	2/3 (66.7%)	(1 case large nerve PNI)
LVI	0/3 (0%)
Margin status reported	7 (35.0%)
Free	6 (85.7%)
Close (1 mm)	1 (14.3%)
Temporal/histologic precursor	9 (45.0%)
Ameloblastoma	3 (33.3%)
Ameloblastic fibroma	2 (22.2%)
Ameloblastic fibrosarcoma	2 (22.2%)
Ameloblastic carcinoma	1 (11.1%)
Unknown	1 (11.1%)

Immunophenotype
Low MW and pancytokeratins
Epithelial	12/12 (100%)
Spindled	3/12 (25.0%)	Weaker and more focal
High MW cytokeratin, p63 and ∆Np63 (p40)
Epithelial	8/8 (100%)
Spindled	2/8 (25.0%)	Weaker and more focal
Vimentin
Epithelial	2/9 (22.2%)	Weaker and more focal
Spindled	9/9 (100%)
Muscle marker
Spindled only	3/6 (50%)
S100, SOX-10
Both	0/3 (0%)
p53 (n = 7)
Increased	5/7 (71.4%)
Extreme	1/7 (14.3%)
Ki-67
Epithelial	33% (10–93%)
Spindled	25% (10–45%)
Wnt pathway alterations	3 (15.0%)

cm centimeters; AC ameloblastic carcinoma; AFS ameloblastic fibrosarcoma; SCC squamous cell carcinoma; CCOC clear cell odontogenic carcinoma; SOC sclerosing odontogenic carcinoma; PNI perineural invasion; LVI lymphatic/vascular invasion; mm millimeters; MW molecular weight

Comments on mitoses were included in 18/20 (90%) cases. Although these were mostly qualitative, half of cases had “numerous” mitoses. Among cases with quantitative data, median mitotic count was 11 per 2 mm² (range: 6–15). Atypical mitoses were noted in a third of cases. Six cases (30%) showed tumor necrosis. Only one case had definitive matrix (dentinoid); three additional cases were equivocal for osteoid. Perineural invasion (PNI) and lymphatic/vascular invasion (LVI) were only addressed in our cases. Two of three cases (66.7%) showed PNI, one with named nerve PNI (our Case 2), and none showed LVI. All seven cases with margin status reported them to be free, although our Case 1 showed a close (1 mm) soft tissue margin.

Immunohistochemical findings were similar to those previously reported. Epithelial components were uniformly positive for cytokeratins and p63/p40, while the spindled components occasionally showed weaker staining. Conversely, vimentin was uniformly positive in the spindled components, and a subset showed weaker staining in the epithelial components. Muscle markers were more variable, and all tested cases were S100 and/or SOX-10 negative.

Wnt pathway proteins were surveyed in three cases with focal nuclear β-catenin in both tested cases and LEF-1 reactivity in both tested cases. The one case stained for E-cadherin showed no altered expression. Of the seven cases tested for p53, only one (14.3%) showed extreme staining that would be potentially predictive of a mutation. One case tested by immunohistochemistry for BRAF V600E was negative. Ten cases were tested for Ki-67 and were reported either qualitatively, as a single index or as two indices, one for each component. Six cases providing separate values for each component showed a median Ki-67 proliferation index of 33% (range: 10–93%) in the epithelial component, and 25% (range 10–45%) in the spindled component, with a median epithelial to spindled proliferation index ratio of 1.5 (range: 0.4–7.5). Two previously reported cases [22] and our Case 1 showed a stromal-predominant proliferation index. Molecular features were only assessed in two of our cases (Cases 1 and 2). Case 1 was negative for EWSR1 rearrangements by FISH. Case 2 demonstrated 9p21 (CDKN2A, CDKN2B) loss, and 12q (MDM2, CDK4) gain by NGS and was negative for fusions on our panel.

Outcome

Outcome data are summarized in online resource 4, and the disease course timeline for each patient is shown in Fig. 7A. Sixteen cases had postoperative follow-up; of these, 43.8% had recurrences and died of disease. Kaplan-Meier estimates of disease-specific survival (DSS) and disease-free survival (DFS) are shown in Fig. 7B, C. Median DSS was 46 months, while 3-year DSS was 58%. Median DFS was 25 months, and 3-year DFS was 35%. Number of recurrences per patient ranged from 1 to 3. First recurrences were most frequently local (5/7, 71.4%). The overall regional and distant metastatic rates were 25.0% and 31.3%, respectively. Distant metastases were most frequently to lung. Patients with recurrences were mainly treated with re-excision or resection of metastases. Four patients received additional chemotherapy and/or radiotherapy. Infectious complication rate was 18.8%, and one patient (6.3%) developed a secondary leukemia.

Fig. 7.

Outcomes for OCS reported to date. A Disease course by individual patient. Abbreviations: XRT Radiotherapy; NED no evidence of disease. DOD died of disease. B Disease-specific survival (DSS) and C Disease-free survival (DFS) for OCS. Shading represents 95% confidence intervals and dotted intercept represents 3-year survival. D DSS and E DFS stratified by mitotic counts. Abbreviations: HiMit Numerous Mitoses; LowMit Moderate and Occasional Mitoses. F DSS and G DFS stratified by necrosis

The limited number of cases and limited follow-up data precluded detailed assessment for prognostic factors. However, patients whose tumors had numerous mitoses trended toward a worse DSS (Fig. 7D, log rank p = 0.059) and exhibited a significantly worse DFS (Fig. 7E, log rank p = 0.045). Cases with necrosis showed a non-significant trend toward worse DSS (Fig. 7F, p = 0.067) and DFS (Fig. 7G, p = 0.15).

Discussion

While the bar is set low given the rarity of OCS, we report the largest series to date, and largest (by virtue of being the most recent) literature summary. Despite the limited number of cases, it does appear that OCS shows considerable morphologic and biologic variation. Several issues can be considered.

Epidemiology of OCS

OCS is rare among malignant odontogenic tumors, but not negligible at ~ 5% in our institutional experience. This proportion is higher than that reported (1.3%) in a recent systematic review of malignant odontogenic tumors [28]. Reasons for this difference are speculative, but our frequency data reflect the unique experience of our institution and busy consult service. When reviewing all 20 cases collectively, the slight male predominance is in line with other malignant odontogenic tumors, although not as pronounced as for many other odontogenic malignancies [28, 29]. The bimodal age distribution is of interest, but it is unclear whether this finding is an artifact of a low number of cases or a true biologic phenomenon.

Clinical Features

As is typical of most odontogenic malignancies [28], OCS was far more prevalent in the mandible, typically posteriorly. All patients presented with a gnathic mass lesion. Pain and paresthesia were also common findings. Ulceration, bony deformation, and tooth mobility were infrequent findings. A point of interest was the occurrence of one OCS in a post-irradiation setting [17], suggesting a potentially distinct pathogenesis for at least some OCS.

Radiologic Features, Preoperative Biopsy Findings, and Treatment

The majority of OCS are radiolucent destructive lesions with cortical perforation and/or tooth resorption in keeping with prior reviews [2, 16]. While FDG-PET was not frequently reported, both cases (our series) for which this was done showed avidity. Preoperative biopsy diagnoses were modestly sensitive (~ 85%) in establishing malignancy, and a diagnosis OCS was at least favored in ~ 30% of preoperative biopsies. Misclassification on biopsy often resulted from limited sampling of tumor components.

Aggressive surgical resection was the mainstay of therapy. Less than a third of patients received upfront neck dissection, which is in keeping with rarity of lymphadenopathy as a presenting finding. Furthermore, all neck dissections that were performed were pathologically negative, suggesting uncertain value for elective neck dissection for OCS patients with a clinically negative neck.

Morphologic and Biologic Variability

Although variability in the epithelial components, precursor lesion types, and clinical presentation are described, OCS is still generally framed as a single entity. However, based on our review, this may not be the case. At least one case (5%) had arisen in a post-irradiation setting as a secondary malignancy, and about 45% had histologic and/or temporal evidence of another odontogenic neoplastic precursor even with our more stringent criteria. While the literature is skewed toward ameloblastic-type epithelial components (AC-like, AFS-like), we feature cases that also show CCOC-like and SOC-like morphology, suggesting a greater diversity than previously reported.

The question then arises as to whether OCS is simply an end form of progression for a variety of odontogenic tumor types, akin to anaplastic thyroid carcinoma for a spectrum of differentiated thyroid carcinomas [30]. But while the thyroid progression model has extensive clinical, morphologic, and molecular substantiation, data to support this analogy for OCS are conspicuously absent. Three cases showed Wnt alterations by immunophenotype. Although this pathway has been implicated in many odontogenic neoplasms [23], most alterations are not necessarily specific. The notable exceptions are CTNNB1 mutations seen in the ghost cell family of tumors [31–33] (of interest, none of the current or previously reported OCS cases had ghost cell keratinization).

Our one case with NGS data shows copy number alterations in 9p21 (CDKN2A and CDKN2B) and 12q (MDM2, CDK4). CDKN2A and CDKN2B encode the cyclin dependent kinases: p14^ARF, p16^INK4A, and p15^INK4. These are among the most frequently deleted tumor suppressor genes across all cancer types and often show prognostic and predictive value [34–36]. MDM2 promotes oncogenesis via negative regulation of p53 [37], while CDK4 inactivates Rb and facilitates cell transition from G1 to S phase [38]. Amplification of 12q14-15 region encompassing MDM2 and CDK4 is noted in well differentiated and dedifferentiated liposarcomas; low grade osteosarcoma subtypes; and intimal sarcomas. While intimal sarcomas mainly involve great vessels (pulmonary trunk, aorta), dedifferentiated liposarcoma and osteosarcomas can occur in the head and neck region, and may show morphologic overlap with the spindled component of OCS suggesting caution when interpreting MDM2 or CDK4 gains in isolation or on limited sampling. Additionally, gains of MDM2 may also be noted as secondary events in other sarcoma types, gliomas, and carcinomas [39].

As one would expect, keratins and p63/p40 mainly were expressed by the epithelial components, and vimentin mainly was expressed by the spindled component. However, some overlap in expression among these components is possible. Interestingly, the concept of epithelial-mesenchymal transition (EMT) in the context of OCS was proposed and discussed previously, albeit in a limited fashion [18]. Muscle marker expression varies in the spindled components, and neither component expresses intercalated duct/myoepithelial salivary-type markers such as S100 and SOX-10.

Outcome

Even with key corrections of apparent errata, 3-year DSS and DFS of 58% and 35%, respectively support that OCS is an aggressive malignancy. To date, attempts at identifying prognostic features have not been made given the limited number of cases and insufficient annotation by modern oncologic standards [40]. However, with the understanding that numerous bias types may be introduced by pooling cases, transforming descriptors into discrete categories, and binning quantitative data, we have attempted some correlations. Cases with numerous mitoses and tumor necrosis did trend toward an adverse prognosis, suggesting these parameters may be useful in formulating a grading system for OCS.

Perineural invasion, lymphovascular invasion, and margin status were not consistently reported, although these parameters would be of interest for prognostic studies and were included in the International Collaboration on Cancer Reporting (ICCR) dataset for malignant odontogenic tumors [40]. In addition, there were insufficient data to comment on the prognostic value of tumor size. In a SEER registry query [29] of malignant odontogenic tumors, size, along with stage group (despite no official TNM for odontogenic malignancies), age, surgical therapy, and radiotherapy were univariate prognosticators.

Locoregional and distant recurrences were not uncommon and, when present, often multiple. A small subset of patients received additional chemotherapy and/or radiotherapy for their recurrent/metastatic disease, although disease progressed in these patients regardless. Other postoperative complications included infection, mucositis, and secondary leukemia.

Terminology/Nomenclature

OCS may have been affected by terminology shifts across the timeframe in which cases were reviewed. Prior to 1992, the term OCS did not officially exist (although apparently used at least once [21]). The concept of spindle cell change in malignant ameloblastic tumors appeared as early as the 1960’s in reports by Itoh, Toeda, and Kaneda [41] and Sugimura et al. [42]; however, these reports may reflect spindling of the stellate reticulum. Biphasic tumors first surfaced in the 1980 and 1990’s [7, 8, 21, 25]. and the term OCS was formalized in the WHO second edition [1]. Here the definition included resemblance to ameloblastic fibrosarcoma, and when resurrected in the WHO fourth edition in 2017, this component of the definition held [5]. As such, inclusion in the OCS category would depend on how strictly this definition was upheld. For instance, Slater [3] adopted a quite literal interpretation of the definition, though others obviously did not. In fact, only 3 AFS-like cases [3, 17, 20] would fit this interpretation. While the current definition [6] is broader, its impact has not likely been felt in the literature.

Why Not ‘Biphasic Sarcomatoid Odontogenic Carcinoma?’

A broader terminology issue is the word ‘carcinosarcoma’ itself. It is a simple, familiar, yet awe-inspiring term that has been attributed to Virchow [43] over 150 years ago, with the prototype being uterine carcinosarcoma. The historical subclassification that was originated by Meyer in 1920 [44] and followed for decades with some modifications [45, 46] is summarized as:

• A)Collision type: between independent carcinoma and sarcoma.
• B)Combination type: a single stem cell with divergent differentiation subs.
  • a
    Early.
  • b
    Late (secondary) (also historically ‘mutational’ type)[47].
• C)Composition type: concurrent evolution of dual cell types in the same topographic region.
• D)Pseudosarcomatous stromal reaction to the carcinoma.

However, the biologic accuracy and validity of this schema has been questioned for at least 3 decades [48]. Type 4 can be readily eliminated. Collision/composition types, while existent, are exceptionally rare. Essentially all tumors labeled carcinosarcoma have a monoclonal origin with later divergent differentiation (Type 2b) [45, 49–52]. This pathogenesis has been validated in salivary gland carcinosarcoma, which by modern criteria, almost always arises in the setting of a precursor pleomorphic adenoma as a late event; this observation has prompted a proposal to rename this entity “sarcomatoid carcinoma ex pleomorphic adenoma” [53].

The limited and circumstantial evidence to date suggests that OCS might also be subjected to a similar nomenclature overhaul. We propose the term “biphasic sarcomatoid odontogenic carcinoma.” Interestingly, OCS disappeared in the WHO 2005 3rd edition [4] with the claim that it could not be reliably separated from ’spindle cell ameloblastic carcinoma (SpAC)’[19]. However, reports of OCS continued to accumulate [12–14] within this timeframe, preceding the reintroduction in WHO 2017 4th edition and suggesting that SpAC was not universally accepted as an alternate [5]. During this hiatus between the WHO 3rd and 4th editions, SpAC usage in the literature showed only a slight uptick of reports toward the end of this timeframe [54–58]. All of this is difficult to read into too deeply as both entities are rare, but both terms simply might be unified.

The earliest description of spindling of ameloblastic carcinomas mentioned spindled cell change in the stellate reticulum [42]. However, many subsequent cases have described frank sarcomatoid transformation of ameloblastoma. In their comprehensive review, McLean-Holden et al. [59] define SpAC as follows: (1) a carcinoma with ameloblastic differentiation; and (2) intimately associated with a spindle cell population that is part of the neoplastic proliferation (distinct from the background stroma). While OCS and SpAC share similar epidemiologic and radiologic features, SpAC do seem more indolent with a lower recurrence rate (2/13, 15.3%), distant metastatic rate (1/13, 7.7%), and only one disease consequent death (post-operative complications) [59]. Thus, distinction between SpAC and OCS may simply resolve to extent of divergence, with the latter being more divergent (perhaps thus more aggressive). Of note, spindle cell variants of other odontogenic carcinomas (clear cell odontogenic carcinoma, ghost cell odontogenic carcinoma) have not been described to date.

In conclusion, OCS is rare among odontogenic malignancies. It is a locally aggressive malignancy with the capacity for regional and distant spread; necrosis and mitotic activity show potential as prognostic factors. OCS often arise from precursors and, thus, may represent a ‘terminal’ morphology. Molecular pathogenesis still requires clarification, but Wnt alterations seem important. For standardization purposes and to reflect modern understanding of biphasic spindled and epithelioid malignancies, we propose that OCS be renamed “biphasic sarcomatoid odontogenic carcinoma.”

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Conceptualization: GC, AC, DRC, MK, EAB, RRS; Methodology: GC, EAB, RRS; Formal analysis and investigation: GC, RRS; Writing—original draft preparation: GC; Writing—review and editing: GC, AC, DRC, MK, EAB, RRS; Funding acquisition: N/A; Resources: EAB, RRS; Supervision: EAB, RRS.

Funding

None.

Data Availability

Not Applicable.

Code Availability

Not Applicable.

Declarations

Conflicts of interest

The authors declare that they have no conflict of interest.

Ethics Approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Consent to Participate

This study has obtained IRB approval from (University of Pittsburgh Human Research Protection Office) and the need for informed consent was waived.

Consent for Publication

For this type of study consent for publication is not required.