Comparison of Patterns of Relapse in Thymic Carcinoma and Thymoma

Abstract

Objective

Thymic carcinomas (TC) are considered to be more aggressive than thymomas and carry a worse prognosis. We reviewed our recent experience with the surgical management of thymic tumors and compared the outcomes and patterns of relapse between TC and thymoma.

Methods

Single institution retrospective cohort study. Data included patient demographics, stage, treatment, pathologic findings, and postoperative outcomes.

Results

During the period 1995–2006, 120 patients with thymic tumors underwent surgery, including 23 patients with TC and 97 patients with thymoma by the WHO 2004 histologic classification. The overall 5 year survival was significantly different between TC and thymoma (TC 53%, thymoma 89%, p=0.01). Data on relapse was available for 112 patients. The progression-free 5 year survival was also significantly different between TC and thymoma (TC 36%, thymoma 75%, p<0.01). By multivariate analysis, thymic carcinoma and incomplete resection were found to be independent predictors of progression-free survival. Relapses in TC tended to occur earlier, and occurred signficantly more frequently at distant sites than in thymoma (60% vs. 13%, p=0.01).

Conclusions

Patterns of relapse differ significantly between TC and thymoma with lower progression-free survival, earlier onset and more distant relapses in TC. Given the greater propensity for distant failures, the inclusion of systemic therapy in the treatment of TC may take on greater importance. Despite significantly higher rates of distant relapse, good overall survival in TC can be achieved.

Introduction

Surgery is considered the mainstay of treatment for thymic tumors, but despite resection, relapses are common and the optimal treatment approaches remain unclear. Disease progression after treatment tends to occur in the mediastinum or the pleural cavity, but some patients return with distant metastases. [1] Given that long-term survival is achievable with resection, freedom from disease progression may be a more useful metric [2]. However, prior studies on recurrence can be difficult to compare in the setting of differing histological classifications, study periods spanning multiple decades, inadequate statistical analyses, and a focus on overall survival.

A more complete understanding of relapse patterns may lead to improved strategies for the prevention, surveillance and treatment of recurrence. For thymoma, disease stage, as defined by the Masaoka system and completeness of surgical resection are considered the most consistent predictors of survival.[3, 4] Our knowledge of thymic carcinoma (TC) has been limited due to their rarity, but TC is considered to be more aggressive and carry a worse prognosis. The number of conflicting classification systems reflects the persistent confusion surrounding thymic tumors, and even the original World Health Organization (WHO) classification labeled thymic carcinomas as Thymoma (Type C).[5] Although the recent update of the WHO classification [6] officially recognizes the distinction between TC and thymoma as separate and distinct histological entities, they tend to be treated fairly similarly. We hypothesized that the patterns of relapse differ significantly between TC and thymoma, reflecting their different behavior. We reviewed our recent experience with the surgical management of thymic tumors in order to define patterns and predictors of disease progression.

Methods

We reviewed all patients undergoing resection for thymic epithelial tumors at Memorial Sloan-Kettering Cancer Center between January 1995 and December 2006. The Institutional Review Board granted approval for this study on August 30, 2005. In this retrospective cohort study, we included all patients with a pathologic diagnosis of thymoma or TC under the WHO (2004) histological classification. All resected specimens were re-reviewed by one reference pathologist (W.T.) for the purposes of this study in order to confirm the diagnosis and the WHO classification. Patients who had undergone prior resection, patients with a diagnosis of thymic carcinoid, and patients presenting with evidence of distant disease (Masaoka stage IVB) were excluded from this analysis.

Patient characteristics and outcomes were abstracted from medical records. The independent variables analyzed are listed in Table 1. Demographic variables included age as continuous and gender and race as categorical variables. Race categories were defined according to the Surveillance Epidemiology and End Results (SEER) database definitions.[7] Patient characteristics included pathologic Masaoka stage as a categorical variable, radiographic tumor size (longest diameter) as a continuous variable, and administration of preoperative therapy and postoperative therapy as dichotomous variables. Resection status was determined from the operative notes and the pathology reports. Resection status was dichotomized as complete if R0 with microscopically negative margins, and incomplete if microscopically or grossly incomplete.

Table 1.

Patient Characteristics

Variable	Thymic Carcinoma (n=23)	Thymoma (n=97)	p-value
Age
Mean	58.1±2.6	58.1±1.5	0.99
Gender
Male	15 (65)	46(47)	0.13
Female	8 (35)	51(53)
Race
White	20 (87)	82 (85)	0.96
Black	1 (4)	5 (5)
Asian	2 (9)	9 (9)
Other	0 (0)	1 (1)
Sizea
Mean	7.0±0.6	7.1±0.4	0.94
Stage
I	1 (4)	24 (25)	0.01
II	5 (22)	32 (33)
III	12 (52)	20 (20)
IVA	5 (22)	21 (22)
Resection
Complete	12 (52)	76 (78)	0.01
Incomplete	11 (48)	21 (22)
Preop Therapy	18 (78)	42 (43)	0.03
Postop Therapyb	13 (59)	36 (38)	0.06

Continuous variables reported as mean ± standard error. Numbers in parentheses represent percentages.

^an=17 for thymic carcinoma, n=94 for thymoma

^bn=22 for thymic carcinoma, n=96 for thymoma

Outcome variables included time from operation to disease progression and site of progression. As there were a high proportion of incomplete resections, we used the term ‘disease progression’ to encompass the occurrence of any treatment failure after resection. Disease progression was timed to the date of the first postoperative radiographic study that demonstrated progression or recurrence, or was censored at the date of the most recent radiographic study documenting absence of disease. Site of progression was dichotomized as locoregional (confined to mediastinum or pleural space), or distant (hematogenous) disease.

Patient characteristics were compared using two-tailed t-tests for continuous variables, and chi-square and Fisher’s exact tests for categorical variables. Relapse was analyzed as a time to event rather than a binary outcome. Survival was estimated using the Kaplan-Meier method. As not all patients had complete resections, relapse was described in this study in terms of progression-free rather than disease-free survival. Log-rank tests were used for univariate analysis of survival. Multivariate analysis was conducted using the Cox proportional hazards method, incorporating clinically relevant and significant variables (p<0.10) by univariate analysis into the model. Statistical analyses were conducted using STATA/IC 10.0 software (StataCorp, College Station, TX), and hypothesis tests were considered significant at p<0.05.

Results

One hundred and twenty patients underwent resection during the study period and served as the basis for analysis of overall survival. 18 of these patients formed the basis for a prior study.[8] Of the 120 patients, 3 patients underwent exploration only and 5 patients were missing data on recurrence status resulting in 112 patients for the analysis of disease progression. Median follow-up for death status was 41 months (mean, 53) and 26 months (mean, 34) for disease-progression.

Patient characteristics are summarized in Table 1. The average age at the time of operation of all patients in this study was 58 years and the ratio of males to females was even overall. Whites (non-Hispanic and Hispanic) formed the predominant category in the distribution of race. In the 111 patients with recorded tumor dimensions, the average tumor size was 7.1 cm in the longest diameter (range 1.2–19). There was no significant difference in age, gender ratio, race distribution, or tumor size between patients with TC or thymoma (p>0.10). With regards to Masaoka stage, there was a significant difference in the distribution, with TC having an appreciably greater proportion of higher stages (III and IVA). Overall the proportion of complete resection (R0) was 73%, but patients with TC had a significantly lower proportion of R0 resections compared to thymoma (52% vs. 76%, p=0.01). Consistent with the higher stage and lower resection rate, a higher proportion of patients with TC underwent preoperative therapy (78% vs. 43%, p=0.003) and postoperative therapy (59% vs. 38%, p=0.06). Identical findings were obtained when comparing the 112 patients with data on disease progression, again with only stage, resection rate, preoperative, and postoperative therapy significantly different (p<0.05).

Pre- and postoperative therapy regimens are detailed in Table 2. The majority of patients who underwent preoperative therapy received chemotherapy alone in both groups. Two patients underwent concurrent chemoradiation in both groups. All chemotherapy regimens were platinum-based. The most common regimens used in TC were cisplatin-docetaxel (4), cisplatin-doxorubicin-cyclophosphamide (3), carboplatin-paclitaxel (3). The most common regimens used in thymoma were cisplatin-etoposide (16) and cisplatin-doxorubicin-cyclophosphamide (15). Of the patients who underwent postoperative therapy, the majority underwent radiation therapy alone in both groups, at a median dose of 5040 cGy (range 2500–6600 cGy).

Table 2.

Perioperative Therapy

Regimen	Thymic Carcinoma	Thymoma
Preoperative	18	42
Chemo only	16 (89)	40 (95)
ChemoRT	2 (11)	2 (5)
Postoperative	13	36
RT only	8 (62)	26 (72)
Chemo only	0 (0)	5 (14)
ChemoRT	5 (38)	5 (14)

Numbers in parentheses represent percentages.

Overall survival was significantly lower at 5 years for TC (53%) as compared to thymoma (89%, p=0.01). Progression-free survival (PFS) was significantly different between TC and thymoma (p=0.0002). The PFS at 5 years for TC was 36%, and 75% for thymoma. The median PFS was 29 months for TC and has not yet been reached for thymoma. Univariate analysis of other additional factors predictive of PFS included Masaoka stage (p<0.0001), resection status (p<0.0001), preoperative therapy (p<0.0001) and postoperative therapy (p=0.01). Age, gender, race and size were not found to be significant predictors (p>0.10).

Multivariate analysis demonstrated that TC consistently predicted PFS after adjusting for Masaoka stage, resection status, preoperative and postoperative therapy in the model (Table 3). TC (hazard ratio 4.1, p=0.006) and resection status (hazard ratio 22.9, p<0.0001) remained as the only significant predictors of progression-free survival in multivariate analysis. Stratifying by the these two independent predictors, complete resection resulted in significantly greater PFS in both TC (p=0.0001, 70% vs. 0%) and thymoma (p<0.0001, 87% vs. 28%).

Table 3.

Multivariate Analysis of Progression-free Survival

Variable	Hazard Ratio	95% CI	p-value
Thymic Carcinoma	4.1	1.5 – 11.3	0.006
Incomplete Resection	22.9	4.0 – 133	<0.001
Stage
III	1.4	0.2 – 10.5	0.77
IVA	1.0	0.1 – 8.2	0.97
Preop Therapy	0.5	0.1 – 3.9	0.52
Postop Therapy	0.5	0.1 – 1.6	0.22

All stage I and II tumors were completely resected (R0). The extent of disease precluded resection of all gross tumor in 19 of 58 stage III and IVA cases. The most common reason for a grossly incomplete resection was invasion of the aorta (7), followed by invasion of the heart (5), invasion of the SVC in conjunction with other structures (3), unresectable pleural disease (3) and bilateral diaphragmatic involvement (1).

Out of 112 patients, 25 patients were found to have disease progression (22%). The distribution of sites of progression is listed in Table 4. Disease progression occurred significantly more often at distant sites in TC. In TC, the majority of relapses (6, 60%) occurred at distant sites, compared to only 2 (13%) in thymoma (p=0.01). The locations of disease progression are summarized in Table 5. Relapses in thymoma were overwhelmingly intrathoracic with the majority pleural based. The only distant site of progression occurred in lung (distinct from pleural based metastases). In TC, distant sites of progression included lung, bone, brain and liver. In the 25 patients who progressed, relapse occurred earlier in TC with a median time from operation to progression of 9 months and 20 months in thymoma. (Table 6). The time to progression in those who relapsed was shorter in those who had incomplete resections (8 months) than those with complete resections (29 months). Among all patients who progressed, the majority (80%) of incomplete resections led to relapse in less than 3 years, as did all relapses in TC regardless of resection status.

Table 4.

Comparison of Sites of Progression

Site	Thymic Carcinoma (n=10)	Thymoma (n=15)	p-value*
Distant	6 (60)	2 (13)	0.01
Locoregional	4 (40)	13 (87)

Table 5.

Location of Progression

Location	Thymic Carcinoma (n=10)	Thymoma (n=15)
Distant
Lung	1	2
Bone	1	-
Bone, lung	1	-
Brain, bone, lung	1	-
Liver	1	-
Liver, lung	1
Locoregional
Pleural	2	8
Mediastinum	2	5

Table 6.

Time to Progression in Patients Who Progressed

Type	Time to Progression (months)
	Overall	Complete	Incomplete
Thymic Carcinoma (n=10)
Median	9.4 (4.5–33.4)	22.7 (16.2–29.2)	7.8 (4.5–33.4)
Thymoma (n=15)
Median	19.5 (3.5–95.0)	45.7 (5.8–64.6)	13.2 (3.5–95.0)

Discussion

The objective of this study was to characterize and compare the patterns of relapse between TC and thymoma and identify independent predictors of relapse. The principle findings are that TC and thymoma exhibit significantly different behavior, with higher stage, lower resectability, lower overall and progression-free survival, more distant metastases and earlier relapses. Our data supports the notion that these two thymic tumors are clinically distinct entities with different biologies, and add to our understanding of the patterns and natural history of relapse in thymic tumors. The clinical relevance of this study is strengthened by a relatively large sample drawn from a recent time period more reflective of current classifications and treatments.

Consistent with previous reports, overall and progression-free survival was significantly worse in TC in our population.[3, 9–11] Multivariate analysis demonstrated that TC and incomplete resection, but not stage, were strong independent predictors of progression-free survival. These results are consistent with reports that WHO histology is independently predictive of recurrence and that Masaoka stage fails to predict outcomes in TC.[10, 11] In those reports size was noted to be an important prognosticator, but we were not able to confirm those findings in our population, possibly due to fewer tumors that reached their size cutoff (>8cm).[10]

The extent of surgical resection is an important consideration given that complete resection is one of the primary determinants of outcome. Incomplete resections occurred only in Masaoka stage III and IVA thymic tumors. All stage I and II tumors were completely resected. The extent of disease precluded resection of all gross tumor in one-third of stage III and IVA cases, most commonly due to invasion of the aorta or the heart in the majority of these cases. Although resection of the superior vena cava (SVC) was performed whenever that would permit a complete resection, the associated morbidity of an SVC reconstruction in conjunction with extrapleural pneumonectomy or chest wall resection was deemed unacceptable in three cases. Distribution of relapse sites differed significantly between TC and thymoma, with disease progression occurring at distant sites in the majority of relapses in TC. These distant sites included brain, liver, bone and lung. The only thymoma distant metastasis occurred in lung in our series, and others have noted lung as the most common site of distant hematogenous metastases as well.[12] As with previous observations, pleural (locoregional) metastases appear to be most common mode of recurrence in thymoma.[10, 12–15] Other extrathoracic sites in thymoma have been reported in the literature, including 3 patients who had brain and liver metastases.[1, 16] The one thymoma patient with a lung metastasis originally presented with what is considered a low-grade histologic subtype (WHO A), serving as a cautionary tale that no patient with a thymic tumor is immune from the potential for distant spread.

Our results should be interpreted in the light of several limitations. The analysis of a single-institution experience raises the issue of external validity. However, this study population more accurately reflects contemporary management and may be more generalizable than prior study populations dating back multiple decades. The duration of follow-up was relatively limited, given the potential for recurrence years after treatment, but is a direct consequence of selecting a more contemporary study population. Nevertheless, significant differences in progression-free survival were clearly evident early on between tumor types and resection status. The limited sample population is a function of the rarity of these tumors, but this study represents one of the larger experiences with TC in the past decade.[17] Although surgery is the mainstay of treatment, including only patients undergoing surgical management necessarily injects selection bias, and fails to take into account the outcomes of those who might have been treated non-surgically. Although patients were often followed with annual CT scans, not all patients had standardized follow-up, and recurrences may have come to light only when symptomatic, or on serendipitous discovery. Conversely, patients who died may have had undetected recurrences. This along with the average yearlong detection interval will tend to overestimate the true time to progression. However, this is likely more reflective of the day to day reality of medical practice conditions.

Intuitively, relapses would most likely be expected to occur locally after incomplete resections. While this appears to be the case with thymoma, the majority of patients with TC who relapsed presented with progression at distant sites despite best attempts at curative resection. This lends support to the idea that TC is a tumor with a more aggressive biology, and may behave more like a systemic disease. The early time course of disease progression raises the possibility that distant micrometastases may already have been present at the time of treatment. The significantly higher frequency of distant failure in TC raises the question of whether greater consideration should be given to the inclusion of systemic therapy when the diagnosis of TC is established.

The notion that different biologies are at work is reflected in reports observing differential expression patterns of tyrosine kinase receptors in thymic tumors. Thymic carcinomas tend to be c-KIT positive and EGFR negative; while thymomas tend to be EGFR positive and c-KIT negative.[18–20] The overexpression of c-KIT in thymic carcinoma points to potential new avenues for targeted therapy, and at least one report has documented clinical response to imatinib, a tyrosine kinase inhibitor directed against c-KIT.[21]

An alternative explanation might be that patients were understaged at time of treatment, but with the exception of brain metastases, most of the other observed sites of distant relapse in these patients would have been encompassed by initial chest CT scan that included the upper abdomen. Several reports on the utility of PET in thymoma suggest that this modality may be helpful in staging, and may be particularly appropriate in cases of TC given the higher potential for systemic disease.[22–24]

The high proportion of distant metastases raises further questions about the value of postoperative radiation in the treatment of thymic tumors.[25–27] Postoperative therapy was not found to be an independent predictor of progression-free relapse, although selection bias must be acknowledged, and all distant relapses occurred in areas outside the radiation field. One hypothesis that has been advanced is whether concurrent induction chemoradiation may decrease the likelihood of recurrence.[28] By aiming for a pathological complete response and sterilizing the tumor prior to surgery, the possibility of shedding viable tumor cells through surgical manipulation may be lessened. These questions can only be answered through prospective controlled studies.

Our approach to pre-operative biopsy has generally been to obtain a core biopsy when there was suspicion for lymphoma or germ cell tumor, or when the clinical appearance of a locally invasive thymic tumor raised consideration for preoperative chemotherapy. Well-circumscribed tumors typically proceed directly to resection. Distinguishing between thymoma and thymic carcinoma prior to resection would likely not alter the treatment course initially, as locally advanced tumors would be treated with neoadjuvant chemotherapy in either case in our practice.

Patients with TC who progressed exhibited a tendency towards earlier relapse, as did patients with incomplete resections. In this series, all patients with TC who progressed have done so within the first 3 years, along with 80% of all patients who progressed after R1 or R2 resection. While it has been our practice to have annual follow-up, more frequent surveillance may be appropriate in the first few years in the setting of thymic carcinoma or incomplete resection. With our latest recurrence occurring at 95 months, we concur with others in advocating prolonged, if not lifelong follow-up.[14, 29]

In conclusion, thymic carcinoma is a distinct entity from thymoma, and as a more aggressive tumor, has a higher propensity for distant spread, early relapse, and lower overall and progression-free survival. Our results reiterate the importance of complete resection and the need for accurate pathological diagnosis and vigilant surveillance post-therapy. The propensity for distant failure places an emphasis on the need for better systemic therapies, and prospective efforts to improve our understanding of the biology and the optimal management of these tumors is imperative.