The clinicopathologic and immunohistochemical features of 65 primary thymic carcinomas are reported (43 men and 22 women; 19–81 years old). Thymectomy was performed in all cases. Masaoka staging for 53 patients showed 3 patients in stage I, 14 in stage II, 17 in stage III, and 19 in stage IV. Histologic studies revealed 9 carcinoma subtypes. Immunohistochemically, the tumors showed high rates of expression for cytokeratin, Pax8, and FoxN1. Follow-up for 62 patients revealed that 36 patients were alive (mean follow-up, 51.1 months) and 26 had died (mean survival, 47.5 months). The 3- and 5-year overall survival rates were 76.6% and 65.7%, respectively. Our findings suggest that thymic carcinomas may behave less aggressively than commonly believed. Lymph node status and tumor size seem to be important prognostic factors. The Masaoka staging system does not seem to reliably predict outcome.
Thymic neoplasms are rare tumors with an incidence of less than 1% of all adult cancers.1 Thymic epithelial neoplasms are divided into thymomas and thymic carcinomas. Whereas thymomas are epithelial tumors that have retained the organotypical features (well-developed lobular architecture, presence of a dual cell population of neoplastic thymic epithelial cells and lymphocytes, dilated perivascular spaces, areas of medullary differentiation, lack of cytologic features of malignancy of neoplastic thymic epithelial cells) of the thymic gland, thymic carcinomas are overtly malignant epithelial tumors showing cytologic atypia, invasive margins, and loss of an organotypical appearance. However, it is important to clarify that thymoma does not represent the “benign” counterpart of thymic carcinoma, but rather exhibits at least low-grade malignant behavior with the potential for invasion and metastasis.
Thymic carcinoma represents a heterogeneous group of tumors with a wide morphologic spectrum. Only rarely have larger series of thymic carcinoma been reported in the literature, and these have often included material derived from small biopsy specimens.2–8 In addition, many of the studies included neuroendocrine carcinomas or salivary gland–type thymic tumors, which show different clinicopathologic features and outcome altogether.9,10
A discussion of the clinicopathologic and immunohistochemical features of 65 cases of primary thymic carcinoma derived from surgical resection material is presented.
Materials and Methods
During a review of 400 cases of thymic epithelial neoplasms from the surgical pathology files of The MD Anderson Cancer Center, Houston, TX, and the personal consultation files of one of us (C.A.M.) collected during a period from 1985 to 2011, 65 cases of thymic carcinoma were identified that fulfilled the criteria for inclusion in this study: Only cases in which material derived from total thymectomy specimens was available and with known clinical follow-up were included. Staging of the cases was performed according to the system proposed by Masaoka et al.11 This system has been widely adopted for the purpose of staging thymic epithelial tumors and is the one most often used. For 12 of our cases, operation notes were not available, precluding complete clinical staging. The tumors were histologically classified according to World Health Organization (WHO) criteria1 and previous reports12 as 38 squamous cell carcinomas (including 3 with basaloid features), 12 high-grade undifferentiated carcinomas, 4 lymphoepithelioma-like carcinomas, 3 spindle cell (sarcomatoid) carcinomas, 2 mucinous adenocarcinomas, 2 papillary carcinomas, 2 clear cell carcinomas, 1 rhabdoid carcinoma, and 1 anaplastic carcinoma.
H&E-stained sections were available for review in each case, and representative paraffin blocks were available in 31 cases for immunohistochemical studies. Deparaffinized tissue sections were incubated with antibodies directed against pancytokeratin (CK) (dilution 1:100; DAKO, Carpinteria, CA), p63 (dilution 1:200; Santa Cruz Biotechnology, Santa Cruz, CA), CK5/6 (dilution 1:50; DAKO), CK7 (dilution 1:300; DAKO), FoxN1 (dilution 1:50; Abcam, Cambridge, MA), Pax8 (dilution 1:100; ProteinTech Group, Chicago, IL), c-Kit (dilution 1:400; DAKO), CD5 (dilution 1:20; Thermo Fisher Scientific, Fremont, CA), CDX2 (dilution 1:50; Biogenex, San Ramon, CA), CD205 (dilution 1:50; Abcam), calretinin (dilution 1:40; Invitrogen, Carlsbad, CA), thyroid transcription factor-1 (TTF-1) (dilution 1:200; DAKO), napsin (dilution 1:200; DAKO), CD30 (dilution 1:100; Novocastra, Leica Microsystems, Bannockburn, IL), and Ki-67 (dilution 1:200; DAKO) using the polymeric biotin-free horseradish peroxidase method. Appropriate negative and positive controls were run for all antibodies tested. The immunostaining was scored on a sliding scale of 0 to 4+ according to the percentage of reactive cells (0, negative; 1+, 1%–25%; 2+, 26%–50%; 3+, 51%–75%; 4+, 76%–100%), and the staining intensity was graded as weak, intermediate, or strong. Clinical and follow-up information was obtained from the patients’ charts or referral information.
Statistical analysis of survival was performed using the Kaplan-Meier method. Statistical significance was defined as a P value less than .05. All statistical analyses were performed using Statistica software, version 6 (StatSoft, Tulsa, OK).
The clinical features are summarized in Table 1. Of the patients, 43 were men and 22 were women with an age range from 19 to 81 years (mean, 54.8 years). The most common symptoms included chest pain, cough, shortness of breath, and superior vena cava syndrome. In 11 patients without symptoms, the tumors were found incidentally on routine imaging. None of the patients had paraneoplastic syndromes, but 9 patients had a history of other malignancy. In these patients, metastatic disease to the mediastinum was ruled out by thorough clinical, radiologic, and immunohistochemical means. Two patients had HIV infection and/or viral hepatitis.
Grossly, the tumors were large lesions measuring 2 to 15 cm in maximum dimension (average, 7.3 cm). Macroscopically, 11 tumors appeared well circumscribed; the majority, however, showed gross invasion of adjacent structures, including the thymic fat, pleura, lung, and pericardium. On cut surface, most lesions were described as white to tan with a firm consistency. Areas of sclerotic change and foci of hemorrhage and necrosis could also be identified in some of the lesions. In 4 tumors, cystic change was found.
The tumors were classified according to the criteria set out by the WHO1 as 38 squamous cell carcinomas (including 3 basaloid type), 12 high-grade undifferentiated carcinomas, 4 lymphoepithelioma-like carcinomas, 3 spindle cell (sarcomatoid) carcinomas, 2 mucinous adenocarcinomas, 2 papillary carcinomas, 2 clear cell carcinomas, 1 rhabdoid carcinoma, and 1 anaplastic carcinoma. Common to all tumors was a characteristic dense fibrocollagenous stroma and an absence of any organotypical features in the carcinoma component. Necrosis was present most often in a comedo-like pattern but could involve as much as 30% of the tissue examined.
Squamous Cell Carcinoma
The 38 squamous cell carcinomas (58% of total) were characterized by a lobular growth pattern of large islands of polygonal tumor cells of varying size with vesicular nuclei, prominent eosinophilic nucleoli, and abundant eosinophilic glassy cytoplasm. In the periphery of the lesions, the architecture often changed to narrow strands, small nests, and trabeculae of tumor with tapered ends. Focal areas of keratinization were present, and in most cases, intercellular bridges could be identified Image 1A. Overall, 1 tumor was classified as well differentiated, 17 as moderately differentiated, and 20 as poorly differentiated. The well-differentiated tumor was associated with a multilocular thymic cyst (MTC) and also showed areas of transition with a thymoma component. In addition, 3 moderately and 2 poorly differentiated tumors showed additional thymoma elements. The mitotic activity ranged from 1 to 66 mitoses per 10 high-power fields (HPFs).
The 3 carcinomas with basaloid features were composed of a solid proliferation of small monotonous cells with a high nuclear/cytoplasmic ratio showing prominent peripheral palisading. Focal areas of squamous differentiation were noted in all 3 cases. Of the tumors, 2 were associated with MTCs. Mitotic activity in these tumors was high, ranging from 21 to 32 mitoses per 10 HPFs.
High-Grade Undifferentiated Carcinoma
A solid-nodular growth pattern was observed for the high-grade undifferentiated carcinomas (12/65 [18%]). These were composed of large tumor cells with basophilic cytoplasm. Keratinization, intercellular bridges, and other defining features were not identified Image 1B. Of the high-grade carcinomas, 2 showed associated MTCs and 1 was associated with a thymoma. Up to 86 mitoses per 10 HPFs were identified in these tumors.
The 4 lymphoepithelioma-like carcinomas (6%) showed a conspicuous syncytial growth pattern. Individual tumor cells were large with ill-defined cell borders, vesicular nuclei, and prominent nucleoli. Infiltrating this epithelial proliferation were numerous lymphoid cells Image 1C. In 1 of these tumors, thymomatous components were identified. Mitotic activity was variable and ranged from as low as 2 to up to 28 mitoses per 10 HPFs.
Spindle Cell (Sarcomatoid) Carcinoma
Of the 65 tumors, 3 (5%) were histologically classified as spindle cell or sarcomatoid carcinomas Image 1D. These tumors demonstrated cellular and, in places, fascicular proliferations of fusiform cells with elongated nuclei, prominent nucleoli, and occasional nuclear inclusions. Most cells had a pale, wispy cytoplasm. Heterologous elements were not identified in any of these tumors. Two spindle cell carcinomas showed thymomatous and MTC components, respectively. Spindle cell carcinomas were mitotically active tumors containing up to 32 mitoses per 10 HPFs. All 3 tumors were entirely composed of fusiform cells with no recognizable other carcinoma subtypes identified.
The 2 mucinous adenocarcinomas (3%) were composed of islands, tubules, and nests of cuboidal to columnar tumor cells floating in pools of extracellular mucin Image 1E. This mucinous change was variable in the 2 tumors, affecting 20% in one and up to 70% of the tissue in the other lesion. Mild to moderate nuclear pleomorphism and various amounts of intracytoplasmic mucin could be identified. The mitotic activity ranged from 45 to 51 mitoses per 10 HPFs.
Alternating solid and papillary growth patterns were seen in the 2 tumors (3%) classified as papillary carcinomas. Papillary structures constituted approximately 30% of the tumor volume Image 1F. These papillae were lined by a single layer of cells arranged around fibrovascular cores. The individual tumor cells had round to oval nuclei, clear to eosinophilic cytoplasm, and small nucleoli. Two tumors were associated with a thymoma and an MTC, respectively. Up to 10 mitoses per 10 HPFs could be identified.
Clear Cell Carcinoma
The 2 clear cell carcinomas (3%) were composed of sheet-like proliferations of cells with abundant clear cytoplasm, prominent cell borders, mild nuclear atypia, and inconspicuous nucleoli Image 1G. In some areas, the cells had a more eosinophilic appearance, but clear cell change in both tumors was prominent (∼80%). Overall, the tumors had a lobulated arrangement and showed only moderate proliferative activity (5 mitoses/10 HPFs).
In the 1 tumor diagnosed as rhabdoid carcinoma (2%), diffuse sheets of medium-sized to large cells were seen. The cells were moderately pleomorphic and had a vesicular nuclear chromatin pattern and prominent nucleoli. The most characteristic feature was the presence of intracytoplasmic paranuclear inclusions displacing the nuclei to the periphery of the cells Image 1H. This phenomenon was seen in the majority of the tumor cells. The mitotic activity was 8 mitoses per 10 HPFs.
In 1 case (2%), the classification was anaplastic carcinoma. This tumor was composed of islands of large polygonal cells with distinct cell borders, basophilic cytoplasm, prominent nuclear pleomorphism, and conspicuous nucleoli. Scattered bizarre-appearing multinucleate tumor giant cells were present dispersed throughout the entire tumor; many of them contained nuclear inclusions Image 1I. We counted 4 mitoses per 10 HPFs.
Immunohistochemical studies were performed on 31 of the tumors, including 24 squamous cell carcinomas, 4 high-grade undifferentiated carcinomas, 1 mucinous adenocarcinoma, 1 clear cell carcinoma, and 1 rhabdoid carcinoma. In combination, all of the tumors were positive for CK (31/31 [100%]) Image 2A, 25 showed reactivity for p63 (81%), 24 expressed CK5/6 and Pax8 (77%) Image 2B, 21 were positive for CK7 and FoxN1 (68%) Image 2C, 20 for c-Kit (65%) Image 2D, 12 for CD5 (39%), 3 for CD205 and calretinin (10%), and 1 for CDX2 (3%). None of the tumors showed staining for TTF-1, napsin, or CD30. Ki-67 demonstrated proliferative activity ranging from 0% to 60%.
The immunohistochemical results, including staining by subtype, are summarized in Table 2.
All tumors were surgically resected, and completeness of resection could be assessed in 47 cases. In 21 cases (45%), complete surgical resection was achieved. In these cases, 17 patients (81%) were alive 4 to 126 months after diagnosis and 4 (19%) had died at 18 to 123 months after diagnosis. In 26 cases, resection was incomplete (55%); 12 (46%) of the patients were alive (follow-up period, 3–114 months) and 14 (54%) had died 1 to 128 months after diagnosis. Completeness of resection could not be assessed in 18 cases (28%). Although statistical significance could not be demonstrated (P = .08658), there seemed to be a trend toward better survival in patients who had undergone complete resection Figure 1.
Staging information was available for 53 patients. In the Masaoka staging system, 3 cases were stage I (6%), 14 were stage II (26%), 17 were stage III (32%), 4 were stage IVa (8%), and 15 were stage IVb (28%) at the time of resection. Staging was not possible in 12 cases. Of the 3 patients with stage I tumors, 2 (67%) were alive with a follow-up period ranging from 8 to 73 months, and 1 patient (33%) had died after 123 months. Among the 14 patients with Masaoka stage II tumors, 11 (79%) were alive (follow-up period, 4–72 months) and 3 patients (21%) had died 31 to 165 months after diagnosis. Of 17 patients with stage III tumors, 8 (47%) were alive (6–126 months), and 9 (53%) had died after 12 to 76 months. Only 4 patients had stage IVa tumors; all of them (100%) were alive 3 to 53 months after diagnosis. Last, 15 patients had stage IVb tumors; 6 (40%) of them were alive (11–107 months); 9 patients (60%) had died after 1 to 128 months.
No significant association between Masaoka stage and survival could be demonstrated when comparing the survival of patients with stage I + II tumors with the survival of patients with stage III and IV tumors for 3- and 5-year overall survival (P = .05545 and P = .08082, respectively) Figure 2. A significant association, however, was identified when patients with tumors classified as Masaoka stages I, II, and III were combined into a single group and compared with patients with stage IV tumors for 3- and 5-year overall survival (P = .02207 and P = .03577, respectively) Figure 3.
Tumor size was assessed in 42 cases with known follow-up. Because the average tumor size was approximately 7 cm, a cutoff at 7 cm was used to correlate follow-up information. Among the 19 patients with tumors smaller than 7 cm, 15 (79%) were alive 4 to 91 months after diagnosis and 4 (21%) had died after 6 to 165 months. Of the 23 patients with tumors 7 cm or larger, 12 (52%) were alive (3–126 months) and 11 (48%) had died 2 to 128 months after diagnosis. Tumor size proved to be a significant prognostic factor for survival (P = .00453) Figure 4.
Lymph nodes were removed during thymectomy in 28 patients, 27 of whom had follow-up information available. In 16 patients (59%), the sampled lymph nodes showed no evidence of metastasis; 11 of these patients (69%) were alive 3 to 126 months after diagnosis, and 5 (31%) had died after 21 to 76 months. Among the 11 patients (41%) with positive lymph node status, 3 (27%) were alive 11 to 18 months after diagnosis and 8 (73%) had died 1 to 128 months after diagnosis. Positive lymph node status was associated with significantly worse survival than negative lymph node status (P = .01070) Figure 5.
Histologic Findings and Follow-up
Comparison of the histologic subtypes of thymic carcinoma with follow-up information was possible in 62 cases with the following results: 25 (66%) of 38 patients with squamous cell carcinoma were alive with a follow-up interval ranging from 4 to 126 months, and 13 patients (34%) had died 1 to 128 months after diagnosis. Among the 10 patients with high-grade undifferentiated carcinoma for whom follow-up was available, 3 (30%) were alive (6–51 months) and 7 (70%) had died (6–76 months). Of the 4 patients with lymphoepithelioma-like carcinomas, 1 (25%) was alive 18 months after diagnosis and 3 (75%) had died after 12 to 165 months. Follow-up was available for 2 patients with spindle cell carcinoma, and both (100%) were alive 11 and 34 months after diagnosis. The 2 patients (100%) with mucinous adenocarcinoma were alive 50 months after diagnosis, and among the 2 patients with papillary carcinoma, 1 (50%) was alive (49 months) and the other (50%) had died 3 months after diagnosis. Of 2 patients with clear cell carcinoma, 1 (50%) was alive (8 months) and 1 (50%) had died after 2 months. The 1 patient (100%) with rhabdoid carcinoma died 22 months after diagnosis, and the 1 patient (100%) with anaplastic carcinoma was alive after a limited follow-up period of 3 months.
In addition to surgical resection, a total of 41 patients (63%) received further treatment in the form of chemotherapy or radiation. Adjuvant cytotoxic or radiation therapy was given to 34 patients; 7 patients received neoadjuvant therapy only. For 2 patients (3%) with a low Masaoka stage and complete resection, surgical therapy was the only treatment. For 22 patients (34%), no information about additional therapy was available.
Of the 34 patients who received adjuvant therapy, 20 (59%) were alive after 6 and 114 months, and 14 (41%) had died after 12 to 128 months. Among the patients who had received neoadjuvant therapy only and patients with no additional treatment, follow-up was available for 8 patients. Of the 8 patients, 6 (75%) were alive at 3 to 126 months, and 2 (25%) had died 2 to 6 months after diagnosis. Overall, there was no statistically different survival between patients who received adjuvant therapy and patients who did not (P = .54459).
Follow-up information available for 62 patients demonstrated that 36 (58%) were alive with a follow-up period ranging from 3 to 126 months (mean follow-up, 51.1 months). There were 26 patients (42%) who had died 1 to 165 months after diagnosis (mean survival, 47.5 months). The overall 3- and 5-year survival rates were 76.6% and 65.7%, respectively Figure 6.
Thymic epithelial neoplasms comprise a rare group of tumors that encompass thymomas and thymic carcinomas. The latter are exceedingly rare, representing only about 15% to 20% of all thymic epithelial tumors. Their rarity reflects the lack of larger series studying these tumors, thus precluding implementation of uniform treatment strategies and the collection of more reliable prognostic data.
Thymic carcinomas can show diverse differentiation and are histologically undistinguishable from various types of carcinoma of other organ systems. Thymic carcinoma subtypes include squamous cell carcinoma, lymphoepithelioma-like carcinoma, basaloid carcinoma, mucoepidermoid carcinoma, sarcomatoid carcinoma, clear cell carcinoma, papillary adenocarcinoma, nonpapillary adenocarcinoma, carcinoma with t(15;19) translocation, anaplastic carcinoma, and neuroendocrine carcinoma.1,12 Our study confirms the histologic spectrum of these tumors, reflected by the 9 different histologic types found in our tumors. Among these, squamous cell carcinoma was by far the most common subtype.
The clinical features and manifestations in our patients are similar to those reported in previous studies: a peak age at diagnosis in the sixth decade and main symptoms of chest pain, shortness of breath, and superior vena cava syndrome. None of our patients had any associated paraneoplastic syndrome; however, well-known associations with thymoma and MTC were confirmed in 10 and 7 cases, respectively.2–8,13–15
Immunohistochemically, thymic carcinoma demonstrates an epithelial phenotype. In this respect, our results confirm the findings of previous studies by demonstrating diffuse immunoreactivity for CK, CK5/6, CK7, and p63.7,8,13,16–19 In addition, data from prior studies suggesting reactivity for c-Kit (73%–86%), FoxN1 (76%), and calretinin (36%) were supported by our series, showing expression for these markers in 65%, 68%, and 10% of cases, respectively.18,20–23
It is important to note that CDX2 is often positive in mucinous adenocarcinomas of the thymus,24 and this was also true in our study. Furthermore, a lack of expression of TTF-1 and CD30 has been noted,16–18,25–27 and none of our tumors showed staining for these markers. Contrary to other studies that showed high rates of immunoreactivity for CD5 (30%–70%) and CD205 (59%), our results for these markers were less convincing (<10% each), casting some doubt on the diagnostic value of these antibodies in this context.18,21,23,26,28,29 Besides, CD5 is also expressed by a range of other neoplasms, including mesothelioma, adenocarcinoma, and atypical thymoma (WHO type B3), limiting its use in the differential diagnosis of mediastinal neoplasms.21,30 An interesting finding was the strong and diffuse expression of Pax8 in 77% of our thymic carcinomas,31 supporting similar, albeit preliminary, results on a smaller number of thymic epithelial neoplasms.32,33 Another intriguing observation was that Pax8 and FoxN1 staining overlapped in all but 3 tumors. FoxN1 was described as a relatively sensitive and specific marker for thymic carcinoma23; taking into account the even higher rate of Pax8 expression of these tumors, their combination markers may prove to be a valuable diagnostic tool in the differential diagnosis of malignant anterior mediastinal neoplasms.
The expression of napsin, to the best of our knowledge, has not been described in thymic epithelial neoplasms and was absent in all of our tumors.
Some of the most compelling observations of our study pertain to the following aspects:
The limitation of the Masaoka staging system for thymic carcinomas
In the absence of a more refined staging system for thymic carcinoma, it is common for tumors to be staged using the Masaoka staging system, which was originally devised as a clinical staging system for thymomas.11 In our series, 53 cases were staged accordingly. By using this staging system, we could not find any statistically significant differences between the different stages. In contrast, a significant difference was identified only when the patients with Masaoka stages I, II, and III tumors were considered in a single group (86.4% and 73.1%, respectively) and compared with stage IV (65.9% and 52.5%, respectively) (P = .02207 and P = .03577, respectively). Our results are in keeping with several other studies that have suggested that the usefulness of the Masaoka staging system is limited to thymoma and does not seem to predict outcome in patients with thymic carcinoma.34,35 As a consequence, the introduction of a new staging system for thymic carcinoma seems warranted.
Lymph node status
It seems that lymphogenous spread is not an uncommon occurrence in thymic carcinoma; in fact, lymph node involvement was seen in approximately 40% of our cases in which lymph nodes were sampled. In addition, we were able to demonstrate significant statistical correlation between lymph node status and survival, emphasizing the need for systematic lymph node dissection in these tumors. Based on our experience, lymph node sampling is not performed in all thymic carcinomas, nor is it performed in any systematic way when done, often yielding only small numbers of randomly sampled nodes. The high importance of lymph node sampling has also been highlighted by other authors in an attempt to develop more standardized operative procedures.14,35
Our data suggest that larger tumors most likely correspond to advanced stage disease, as they are more likely to compromise adjacent structures.
Limited importance of histologic subtype as a prognostic factor
This point should not be surprising given that the tumors included in our series were all high-grade lesions except for 1 well-differentiated squamous cell carcinoma. It is also important to note that most histologic subtypes encountered in our series were represented only in small numbers, precluding more meaningful assessment.
Limited impact of adjuvant therapy
This factor did not seem to have a dramatic benefit on the survival of our patients. On that issue, there are conflicting views in the literature with regard to the efficacy of adjuvant treatment; while some reports advocate the use of postoperative cytotoxic or radiation therapy, others failed to prove the effectiveness of such a multimodal approach.2,4,5,14,36 This seems to indicate that a surgical approach is still the most crucial step in the treatment of patients with thymic carcinoma. In this context, although statistical significance could not be demonstrated, we observed a trend toward an apparent survival benefit in patients whose tumors were completely resected, confirming the results of several previous studies.3–5,35
Follow-up information for our patients revealed a mean survival of 47.5 months and 3- and 5-year overall survival rates of 76.6% and 65.7%, respectively. Whereas previous reports have labeled thymic carcinomas as aggressive neoplasms with a mean survival of only 18 to 35.4 months and 5-year overall survival rates ranging from 14.5% to 50%,2–7,37 our data seem to indicate that thymic carcinomas may run a less aggressive course. Because our study findings do not seem to differ significantly from those of previous studies in terms of clinical patient characteristics (male predominance, peak age),2–7 tumor size,6–8 tumor stage distribution,2,3,5 or percentage of completely resected tumors,3–5 we consider that the potential factor for better survival in our series is the fact that neuroendocrine thymic neoplasms were not included in our study. These tumors are known to behave more aggressively when originating in the anterior mediastinum,9 and inclusion of these tumors in most of the other series may account in part for the less favorable clinical course observed in those studies.
Finally, in the evaluation of thymic carcinoma, it is important to consider and rule out other neoplasms. The differential diagnosis of thymic carcinoma primarily includes distinction from metastatic disease, direct invasion from neighboring organs, and primary nonepithelial thymic neoplasms. Owing to a lack of a specific immunohistochemical marker for thymic carcinoma and its morphologic variability, exclusion of metastatic disease to the thymus still heavily relies on clinical and radiologic investigations. The presence of an isolated anterior mediastinal mass with absence of tumor elsewhere would point toward primary thymic carcinoma. In addition, direct invasion from neighboring organs, like tumors of primary lung or thyroid origin, often have to be excluded by clinical or imaging criteria.
Thymic carcinomas with a large cell component or glandular differentiation may be mistaken for primary mediastinal germ cell tumors, eg, embryonal carcinoma or yolk sac tumor. These tumors generally occur in a younger age group with heavy male predominance and show a different immunohistochemical phenotype, being positive for CD30 and α-fetoprotein.38,39 Primary mediastinal lymphomas will also have to be considered in the differential diagnosis. These neoplasms demonstrate a discohesive growth pattern and lack immunoreactivity for CKs, which distinguish these lesions from thymic carcinoma.40 Last, primary sarcomas, like synovial sarcoma (primarily the monophasic type), may enter the differential diagnosis, especially when dealing with a tumor with spindle cell morphologic features. The spindle cells in thymic carcinoma, however, show a greater degree of cytologic atypia and a different immunohistochemical staining pattern displaying more diffuse reactivity for CK and lacking the expression of vimentin, bcl-2, and CD99 associated with synovial sarcoma.41
We have described the clinicopathologic and immunohistochemical characteristics of 65 thymic carcinomas and confirmed previous data regarding histologic variability and clinical manifestations. The importance of tumor size, lymph node sampling, and complete surgical resection in these tumors is emphasized. Contrary to other reports that describe thymic carcinomas as highly aggressive tumors with low overall survival, our data seem to suggest better biologic behavior. In addition, we were able to demonstrate that the Masaoka staging system is insufficient to predict the clinical outcome of these neoplasms, making the introduction of a new staging system highly desirable.
We are grateful to Carmen Behrens, MD, for help with statistical analysis.