Chondrosarcoma is a rare disease, with an estimated incidence of 1 in 200,000 per year. There is inconsistency in the literature with regard to the factors affecting the survival rate because many reports have focused on relatively small single institutional experiences. Although tumor grade and surgical stage are assumed to be the most significant prognostic variables in modern clinical practice, there has been disagreement in the literature regarding the validity of this assumption.
A recently published study determined the relative five-year survival rate to be 75.2% on the basis of a national cancer registry of 9606 cases1. Although unprecedented in magnitude, that study did not present demographic or other clinical information about the patients, making it impossible to derive prognostic information from the data. Furthermore, the registry is not population-based and thus is not representative of the average cross-section of patients who receive treatment for chondrosarcoma.
To our knowledge, the current study includes the largest number of cases with both demographic and survival information from a validated, population-based national registry. The SEER (Surveillance, Epidemiology and End Results) program database is the most frequently used and best estimate of cancer incidence in the United States2-8. Several hundred studies based on the SEER database have been performed in the last several years to evaluate the outcomes of breast, colorectal, prostate, lung, and ovarian cancer4. We present for the first time the SEER data on chondrosarcoma.
Chondrosarcoma is a uniquely suited disease process to be evaluated with a long-term database because the treatment paradigm has remained largely unchanged over the study period. Chondrosarcoma has been deemed a surgical disease for many decades, and surgery remained the primary treatment modality over the entire data collection period. Chemotherapy and radiation have not been routinely utilized for the treatment of this disease, as they have for bone tumors such as Ewing sarcoma and osteosarcoma, and thus there are fewer confounding variables to complicate the statistical analyses. These characteristics of the treatment of chondrosarcoma in the United States make this disease process ideally suited for evaluation with use of the SEER database.
The Surveillance, Epidemiology and End Results (SEER) Program
The Surveillance, Epidemiology and End Results (SEER) Program of the National Cancer Institute is the largest registry source of information on cancer incidence and survival in the United States9. SEER currently collects and publishes data on cancer incidence and survival from seventeen population-based cancer registries encompassing approximately 26% of the U.S. population. These specific local registries were chosen for their completeness and their adequate representation of minority populations. An epidemiological study comparing SEER areas with non-SEER areas in the United States concluded that the age and sex distributions of these areas were comparable, although the SEER areas tended to be more affluent and more urban than non-SEER areas10. With just over 6,100,000 incident cancer cases diagnosed from January 1, 1973 to June 1, 2003, SEER represents the largest cancer database in the country. The SEER program registries routinely collect data on demographic characteristics of the patients, the primary sites of tumors, the morphological characteristics of tumors, the stage at the time of diagnosis, the first course of treatment (occurring within four months after the diagnosis), and follow-up for survival status (that is, the presence or absence of disease at the time of the latest follow-up). The SEER program is the only comprehensive source of population-based information in the United States that includes the stage of cancer at the time of diagnosis as well as data on patient survival. National Cancer Institute staff members work with the North American Association of Central Cancer Registries to guide all state registries to achieve data content and compatibility that are acceptable for pooling data and improving national estimates. The SEER program's standard for case completeness is 98%. The SEER program is considered to be the prevailing standard for quality among cancer registries around the world7.
The latest SEER data, based on the April 2005 release, were used to identify all incident cases of chondrosarcoma that were diagnosed from 1973 to 2003 with use of the ICD-O-3 (International Classification of Diseases for Oncology, Third Edition) morphology codes 9220-9221, 9230-9231, 9240, and 9242-924311. Data on a total of 2890 patients were extracted from the database. There were no duplicate cases in the study sample. Information extracted from the database included the demographic characteristics of the patients, the histological characteristics and grade of the tumors, the location of the tumors, the surgical stage, the use of surgery and radiation treatment (if provided within four months after the diagnosis), and the survival time (in months) until death or loss to follow-up. Because of reporting omissions, data on tumor grade and tumor size at the time of the initial presentation were not available for 684 cases (23.7%) and 771 cases (26.7%), respectively. Only the percentages based on available data for each individual variable are given. Patients with missing data were excluded from the respective univariate and multivariate analyses.
Patient age was arbitrarily converted to a categorical variable (fifty years or less, or more than fifty years) to simplify the univariate and multivariate analyses. Tumor location is not uniformly described in the database; thus, categorical variables of appendicular, axial, and soft tissues were used on the basis of available descriptors. Tumors were classified as appendicular if they involved the arm or leg and as axial if they involved the pelvis, spine, or scapula. Low-grade tumors included well differentiated and moderately differentiated lesions (ICD-O-3 Grades 1 and 2), whereas high-grade tumors included poorly differentiated, undifferentiated, and anaplastic lesions (ICD-O-3 Grades 3 and 4)11. Staging was determined with use of the American Joint Committee on Cancer (AJCC) staging system for bone sarcomas, and the tumors were categorized as local, regional, or distant disease (designated as M0, M1, and M2, respectively)12.
SEER Data Confidentiality
All available data in the SEER database are retrospective in nature. All personal identifiers are absent from the database. In addition, all variables that might lead to reidentification such as the date of birth have been removed or transposed. Any remaining risk of reidentification has been minimized by the governing agency in not allowing the data to be available as public-use information. Investigators wishing to use the data must sign a legally binding data-use agreement with the Centers for Medicare and Medicaid Services and SEER, and all projects using the data must be preapproved to ensure no confidentiality breach13. The results reported in the present study are in compliance with the Health Insurance Portability and Accountability Act of 1996.
Statistical Methods
SEER*Stat software (version 6.3.6, National Cancer Institute, Bethesda, Maryland) was used to analyze the incidence rates and trends from 1973 to 2003. All data on incidence were age-adjusted and normalized to the 2000 United States Standard Population. The annual percentage change was calculated with use of the weighted least squares method. The level of significance was set at p < 0.05. Correlations between categorical variables were made with use of the chi-square test. Disease-specific survival reflects death as a consequence of the malignant disease, and overall survival reflects death from any cause. One, two, five, and twenty-year overall and disease-specific survival rates were calculated with the Kaplan-Meier method. Survival was calculated from the time of the initial diagnosis to the date of last contact (or the date of death if the patient had died). The effects of demographic, clinical, pathologic, and treatment variables on survival were tested with use of the log-rank test for categorical values. A multivariate analysis with use of the Cox proportional-hazards model was used to further test prognostic factors that were found to be significant in the univariate analysis of 2890 patients. Specifically, age, sex, race, tumor location, stage, year of diagnosis, and the implementation of either surgical resection or radiation therapy were included in the multivariate analysis. Because of concern regarding competing risks of death, the data were analyzed with the Gray method14-16. Analysis was performed with the R Foundation statistical package17.
Source of Funding
No external funding was received for this study.
Two thousand eight hundred and ninety cases of chondrosarcoma were identified from the SEER database. The demographic data on the study population are summarized in Table I. The mean age at the time of diagnosis was fifty-one years (range, one to 102 years). The sex distribution demonstrated a slight male predominance (55%). The tumor location was most commonly classified as appendicular (44.5%), followed by axial (31.1%), and then by soft tissue (9.6%). The anatomic location was not specified for 14.9% of the patients.
Low-grade chondrosarcomas, comprising Grade-1 tumors (910 patients; 31.5%) and Grade-2 tumors (926 patients; 32.0%), were more frequent than high-grade chondrosarcomas, comprising Grade-3 tumors (207 patients; 7.2%) and Grade-4 tumors (163 patients; 5.7%). Grade-1 and 2 tumors (1836 patients) were grouped as low-grade lesions and Grade-3 and 4 tumors (370 patients) were grouped as high-grade lesions to permit statistical comparison.
In the first four months after the diagnosis, 2468 patients (85.4%) were managed operatively and 373 patients (12.9%) were managed nonoperatively. The rationale for nonoperative or delayed operative treatment was not available. Radiation therapy was received by 346 (12%) of the patients. Whether the radiation therapy was used for primary treatment or for palliation could not be determined.
Table II demonstrates the histological profile of tumors and their associated five-year survival rates. Most tumors (2387 patients; 82.6%) were classified as chondrosarcoma NOS (not otherwise specified). Other reported subtypes, in order of decreasing frequency, were myxoid (283 patients; 9.8%), mesenchymal (126 patients; 4.4%), dedifferentiated (forty patients; 1.4%), juxtacortical (twenty-one patients; 0.7%), and clear cell (thirteen patients; 0.4%) variants. Malignant chondroblastoma (twenty patients; 0.7%) was also included in the study. The five-year survival rates according to histological type varied widely, from 0% (dedifferentiated subtype) to 100% (clear cell subtype). In the group of patients with the dedifferentiated subtype, there were no survivors after three years (p < 0.05). Patients with chondrosarcoma NOS had a five-year survival rate of 70%, similar to the rate for those with the myxoid subtype (71%). Juxtacortical chondrosarcoma and malignant chondroblastoma were associated with slightly better survival rates (93% and 85%, respectively), and mesenchymal chondrosarcoma was associated with a slightly worse survival rate (48%). Subtypes with a worse survival rate were associated with a higher percentage of high-grade tumors (85.3% for dedifferentiated chondrosarcoma and 77.8% for mesenchymal chondrosarcoma; p < 0.05). The survival rate for all other histological types was not significantly different (p > 0.05).
The disease-specific and overall survival rates for all patients with chondrosarcoma are depicted in Figure 1. The disease-specific survival rate stabilized at about ten years after the diagnosis and remained at 72.80% at thirty years. The overall survival rate, however, continued to decline steadily after diagnosis and reached 33.30% at thirty years. These data indicate that patients who survive the first ten years after diagnosis are more likely to die of other causes than they are to die from chondrosarcoma-related events.
With regard to sex, women had a slightly better survival rate than men did at thirty years of follow-up (78.60% compared with 68.70%), which was found to be significant in the univariate analysis (p < 0.05). The primary difference in survival occurred within the first ten years, and the difference between the sexes remained constant thereafter.
The disease-specific survival rate, stratified according to histological grade (low or high), is depicted in Figure 2. The rate of survival at thirty years was 76% for patients with low-grade tumors, compared with 50% for those with high-grade tumors (p < 0.05). In the group of patients with high-grade tumors, there was an accelerated mortality rate in the first ten years, after which the survival rate stabilized. The group of patients with low-grade tumors had a more gradual attenuation in the survival rate over time.
The prognostic significance of surgical stage can be interpreted from the Kaplan-Meier plot shown in Figure 3. Patients were grouped according to whether the disease status was localized, regional, or distant, which reflect the M0, M1, and M2 classifications of the AJCC staging system. Patients with localized disease (M0) had twice the thirty-year survival rate of those with regional disease (M1) (43% compared with 22.30%). The latter patients had twice the thirty-year survival rate of patients with metastatic disease (M2) (22.30% compared with <10%). These differences were significant (p < 0.05).
With regard to the effect of anatomic location on the overall survival rate, disease-specific survival plots could not be obtained because of insufficient data; therefore, the overall survival rate was analyzed as an alternative means of addressing this question. Appendicular lesions were associated with a better overall survival rate than axial lesions were (82.7% compared with 61.6%) (p < 0.05).
Figure 4 depicts the overall survival rate stratified according to age at the time of diagnosis. Patients who were fifty years old or less had a significantly better disease-free survival rate (data not shown) as well as a significantly better overall survival rate (>60%) after thirty years than did patients who were more than fifty years old (<20%) (p < 0.05).
Disease-specific survival rates are presented in Figure 5 for patients who were diagnosed in each of the three decades during this data-collection period (1973 to 1983, 1984 to 1993, 1994 to 2003). These data show that there has been no significant change in survival rates for chondrosarcoma over the course of thirty years.
The results of the multivariate analysis are shown in Table III. When the significant univariate variables of histological grade, surgical stage, sex, and tumor site were evaluated, only histological grade and surgical stage were shown to be independent predictors of survival (p < 0.001). High-grade tumors had a hazard ratio of 3.4 when compared with low-grade tumors. When compared with local disease, regional and distant disease had hazard ratios of 1.9 and 5.8, respectively (p < 0.001). Sex and site were not significant independent predictors of survival, in contrast to what was determined in the univariate analysis.
To evaluate the reason why the sex and site variables lost significance in the multivariate analysis, these two variables were compared. The analysis showed that male patients had a nearly equal proportion of appendicular (46.7%) and axial (40.4%) chondrosarcomas, whereas female patients had a disproportionate amount of appendicular disease (60.9%) as compared with axial disease (30.3%). This difference was significant (p < 0.001).
Finally, we extended the Cox regression model to evaluate for competing risks of death. The data were reanalyzed to account for other causes of failure, and each of the significant variables was considered. While relatively small changes in the numerical value of the p value were observed in some variables, in no instance did the competing risk analysis fundamentally change the prognostic significance of the variables.
Our findings regarding the overall five-year survival rate according to histological subtype (Table II) substantiate the findings of another large-size database-derived descriptive study1. The relative five-year survival rates for conventional, myxoid, juxtacortical, and mesenchymal chondrosarcoma were 70%, 71%, 87%, and 52%, respectively, in the study by Damron et al.1, compared with 70%, 71%, 93%, and 48%, respectively, in the current study. The survival rates associated with the clear cell and dedifferentiated varieties (100% and 0%, respectively), are only available for the current study.
When analyzing the time-to-event data on death from chondrosarcoma, it is obvious that some individuals may die from other causes18. Death from other causes is termed competing risk and can either hinder the observation of an event or alter the probability of occurrence19. The influence of competing risk was analyzed by completing a Cox regression that ignored competing risk and secondly by analyzing the data by grouping all competing risks as one variable. The alternative statistical analysis did not substantially change the prognostic significance of any variable.
Interestingly, previous reviews of chondrosarcoma involving relatively large numbers of patients have failed to consistently reveal a prognostic value of histological grade20 or surgical stage21-23. Table IV lists the previously published major studies of chondrosarcoma that have included an evaluation of prognostic variables. One of the possible explanations for this inconsistency is the subjectivity of the current grading system, which leads to individual and institutional differences in the assignment of histological grades. The subjectivity in the histological grading of chondrosarcoma was recently shown to be the case in a blinded study of histological grading by multiple treating specialists24. The large sample size across many institutions in the current study may diminish this bias. Other explanations for variability in tumor grading include tumor heterogeneity or nonrepresentative biopsy tissue. It also has been postulated that occasional cases of chondrosarcoma can have an unpredictable behavior despite the best efforts to appropriately classify the disease25,26. Last, missing data in the SEER database may be responsible for some or all of the inconsistencies between the current report and previous ones.
Sex-related differences in the incidence of chondrosarcoma have not previously been described. Surprisingly, the present series did demonstrate a survival advantage for women in the univariate analysis; however, this difference was found not to be independent in the multivariate analysis. To investigate these findings, the sex distribution for axial and appendicular chondrosarcoma was further studied. We found a higher proportion of appendicular chondrosarcomas in women (60.9%) than in men (46.7%) and a lower proportion of axial chondrosarcomas in women (30.3%) than in men (40.4%). Appendicular chondrosarcomas have been associated with a better survival rate22 and a lower local recurrence rate27 in comparison with axial tumors, and this difference might account for the apparently better survival rate in women in the present series. It was an unexpected finding in itself that women had proportionately fewer axial chondrosarcomas than men did. The finding that the location of disease was not reported for 14.9% of the patients in the present study may account for the apparently higher proportion of appendicular disease in women.
The presumption that the treatment paradigm for chondrosarcoma did not change over the study period is also reflected in the results shown in the disease-specific survival plot in Figure 5. There appears to be no significant difference between the five-year survival rates; however, there is a small trend toward slightly better survival for the most recently diagnosed group. This finding also holds true for the ten-year survival rates, with a small trend toward better survival for the more recently diagnosed group. Because there were no twenty to thirty-year long-term survival data on the recently diagnosed groups, it is unknown whether this trend continues or becomes significant.
One of the benefits of the SEER database is that it is population-based and includes a large number of cases. The database has been extensively evaluated for validity, and age, sex, and vital status can be reliably determined7,28. Another potential benefit is its linkage to Medicare data since 1991 and the resultant billing information that it provides. It is possible, with these linked data, to create a non-cancer control group on the basis of selected patient geographic and demographic characteristics. The Medicare data also offer a longitudinal perspective, making available services before, during, and after a cancer diagnosis. Medicare data provide information on chemotherapy, which is not included in the SEER data. Future applications of the Medicare-linked data can include issues such as quality of care, disparities in health care, access to health care, and cost of treatment.
The limitations of the present study are related to the SEER database and include the lack of information on how the cancer was detected, patient comorbidities, and treatment provided more than four months after the diagnosis7, although this information can be extracted from a database derived from Medicare, which is a health-care insurance program administered by the U.S. government. Lack of knowledge regarding treatment provided after four months eliminates any potential information on patients who have had multiple operations, which would be clinically useful. Specific information on the type of surgery performed and the adequacy of surgical margins is not included. These factors have been shown to have significance in terms of recurrence, metastasis, and survival in many other studies of chondrosarcoma25,29-38. The variable of race or ethnicity, although not evaluated in the present study, can be misleading because of the changing definitions of race since the inception of data collection in 197328. Correlation between medical records and claims information is usually accurate, but diagnosis codes can sometimes vary. These are observational data, and thus patients are not randomly assigned to treatment groups. This might be more important in a study evaluating different treatment outcomes. The date-of-death variable is derived from state death certificates and from periodic uploads from the Social Security Administration, which have a small inherent inaccuracy. The Cause of Death (COD) is determined by the ICD-9 (International Classification of Diseases, Ninth Revision) diagnosis code for cause of death and is categorized as Alive; Dead, Cause not Cancer; Dead, Cause Cancer; and Dead, Cause Unknown. The literature on the validity of the use of this cause-of-death determination is extensive and controversial28,39. Another limitation is the incomplete data on several important variables in the present study. The anatomic location of disease was not reported for 14.9% of the patients. As previously discussed, it is impossible to determine if female sex actually confers a survival benefit or if this benefit was due to the fact that women in the present series had more appendicular disease. Neither the survival benefit of the female sex nor the increased incidence of appendicular disease in women has ever been reported, to our knowledge. The other limitation is the fact that the data on tumor grade and tumor size at the time of the initial presentation were available for only 684 cases (23.7%) and 771 cases (26.7%), respectively. We were still able to show that tumor grade and stage had a significant effect on survival, despite the high percentage of incomplete data.
The greatest advantage of the present study is its size and the breadth of available patient demographic information and the availability of long-term follow-up data, which has permitted the determination of prognostic variables for survival. This is especially important because previous studies were limited by small size, and thus many were contradictory in their findings. Although there were no new findings in this review, it is equally as important to confirm the previously inconsistent findings that surgical stage and histological grade are independent prognosticators for survival. Clinically, it is also helpful for both patients and their treating specialists to know that after surviving ten years of this disease, it is unlikely that chondrosarcoma will be the ultimate cause of the patient's death. New advances in the treatment of this disease will be a welcome change in the current treatment paradigm, for we have reached a stable, but disappointing, maximum survival with surgical treatment alone over the last thirty years. 