Myxofibrosarcoma (fibromyxosarcoma): A population-based analysis of clinicopathological features, treatment patterns, and survival outcomes from the SEER database

Abstract

Myxofibrosarcoma (fibromyxosarcoma) is one of the most common soft tissue sarcomas in elderly patients, yet population-based data characterizing its epidemiology and outcomes remain limited. This study aimed to characterize the clinicopathological features, treatment patterns, and survival outcomes of fibromyxosarcoma using the Surveillance, Epidemiology, and End Results database. Patients diagnosed with myxofibrosarcoma between 1992 and 2022 were identified from the Surveillance, Epidemiology, and End Results database. Overall survival and cancer-specific survival were estimated using Kaplan–Meier methods. Cox proportional hazards regression identified independent prognostic factors. Competing risks analysis using cumulative incidence functions and Fine–Gray models accounted for non-cancer mortality. A total of 1879 patients were included, with a median age of 64 years and male predominance (53.7%). Age-adjusted incidence increased significantly (annual percent change: 6.4%; 95% confidence interval [CI]: 5.4%–7.5%). The lower extremity was the most common primary site (45.4%), with 51.0% presenting with localized disease. Surgical treatment was performed in 90.5% of patients, and 48.9% received radiotherapy. Five-year overall survival was 74.3% (95% CI: 72.2%–76.5%) and cancer-specific survival was 85.2% (95% CI: 83.4%–87.0%). In multivariate analysis, age was the strongest independent prognostic factor (age 75+ vs <40: hazard ratio: 19.77, 95% CI: 12.10–32.30, P < .001), followed by stage (distant vs localized: hazard ratio: 5.43, 95% CI: 3.66–8.05, P < .001) and grade. Competing risks analysis confirmed these associations with attenuated effect sizes, particularly for age. Myxofibrosarcoma demonstrates favorable intermediate-term survival, particularly for localized extremity tumors in younger patients. Age is the dominant prognostic factor, reflecting both cancer-specific and competing mortality risks. Surgical resection remains the treatment cornerstone; radiotherapy was associated with improved survival in adjusted analyses, though unmeasured confounders (surgical margin status, radiation field and dose, and performance status) preclude causal interpretation. The competing risks framework offers clinically relevant estimates for counseling elderly patients with significant non-cancer mortality risks.

1. Introduction

Myxofibrosarcoma (MFS), previously classified as the myxoid variant of malignant fibrous histiocytoma, represents one of the most common soft tissue sarcomas in elderly patients, with a predilection for the 6th to 8th decades of life.^[1,2] This malignant fibroblastic neoplasm is characterized by variable myxoid stroma, cellular pleomorphism, and a distinctive curvilinear vascular pattern with perivascular condensation of tumor cells.^[3,4] The histological grade of MFS ranges from low to high, with grade serving as a critical determinant of metastatic potential: low-grade lesions exhibit primarily local aggressiveness, whereas intermediate- and high-grade tumors demonstrate substantially elevated risks of distant metastasis.^[5,6] The foundational clinicopathological characterization of myxofibrosarcoma was described by Mentzel et al,^[7] with subsequent comprehensive reviews by Roland et al^[8] and van der Horst et al.^[9]

A defining feature of MFS is its propensity for local recurrence, which has been attributed to the tumor’s characteristic infiltrative growth pattern along fascial planes and its tendency to extend beyond macroscopically apparent margins.^[10,11] Local recurrence rates ranging from 16% to 47% have been reported in the literature, with recurrences frequently demonstrating progression to higher histological grades, suggesting a multistep tumor progression model driven by increasing genetic instability.^[12,13] The achievement of adequate surgical margins remains paramount for local control; however, the definition of what constitutes an adequate margin in MFS remains controversial, with some studies suggesting that margins >1 mm may be sufficient while others have found no significant association between margin status and prognosis in high-grade disease.^[14,15]

Treatment of MFS centers on surgical resection, with limb-sparing surgery representing the standard of care when feasible.^[16] The role of adjuvant radiation therapy (RT) has been investigated in several institutional series, with some evidence suggesting improved local control, particularly when adequate surgical margins cannot be achieved.^[17,18] However, the effect of radiation on overall survival (OS) remains incompletely characterized, and out-of-field recurrences have been observed more frequently in MFS compared to other extremity sarcomas, raising questions about appropriate radiation field margins.^[19] The role of systemic chemotherapy in MFS is even less well-defined, with no consensus regarding its impact on local recurrence, distant metastasis, or survival outcomes.^[20]

Despite growing recognition of MFS as a distinct clinicopathological entity, population-based data characterizing its epidemiology, treatment patterns, and outcomes remain limited. Previous studies have been predominantly single-institution retrospective analyses constrained by small sample sizes, typically including fewer than 100 patients.^[1,3,21] The increasing incidence of MFS observed in recent decades, which may reflect both true epidemiological changes and improved diagnostic recognition, underscores the need for large-scale population-based investigations to better understand this malignancy.^[22] Furthermore, the evaluation of prognostic factors has yielded inconsistent results across studies, and the competing risk of non-cancer mortality in this typically elderly population has rarely been addressed in survival analyses. Accordingly, this study aimed to characterize the clinicopathological features, treatment patterns, and survival outcomes of fibromyxosarcoma using the Surveillance, Epidemiology, and End Results (SEER) database, with particular attention to identifying independent prognostic factors and accounting for competing risks of mortality.

2. Methods

This retrospective cohort study utilized data from the SEER program of the National Cancer Institute. The SEER database represents approximately 12.2% of the United States population and serves as the authoritative source for cancer statistics in the country. Data were extracted from the SEER 12 Registries Research Data (1992–2022).

Patients diagnosed with myxofibrosarcoma were identified using the International Classification of Diseases for Oncology, Third Edition (ICD-O-3) histology code 8811/3, which designates myxofibrosarcoma (fibromyxosarcoma). Of note, low-grade fibromyxoid sarcoma is a separate entity classified under ICD-O-3 code 8840/3. Inclusion criteria were histologically confirmed fibromyxosarcoma (ICD-O-3 code 8811/3); malignant behavior code (ICD-O-3 behavior code 3); diagnosis between January 1, 1992 and December 31, 2022; and available survival data. Exclusion criteria included cases identified by death certificate only or autopsy only, nonmalignant behavior codes, and missing survival or vital status information.

Age at diagnosis was extracted from the SEER variable and categorized into clinically relevant groups: <40, 40 to 59, 60 to 74, and 75+ years. Race was derived from the SEER race recode variable and categorized as White, Black, Asian/Pacific Islander, American Indian/Alaska Native, or unknown. Marital status at diagnosis was categorized as married, single, divorced/separated, widowed, or unknown.

Primary site was extracted using ICD-O-3 topography codes and classified into: lower extremity (C49.2), upper extremity (C49.1), trunk (C49.3, C49.4, C49.6), pelvis (C49.5), head/neck (C49.0), retroperitoneum, and other. A binary site classification (extremity vs non-extremity) was created for subgroup analyses. Due to changes in SEER staging systems over time, a unified stage variable was created using summary stage 2000 (1998–2017) as the primary source, SEER historic stage A (1973–2015) as a secondary source, and derived extent of disease 2018 stage group for cases diagnosed 2018 and later. Final unified stage categories were localized, regional, distant, and unknown. Tumor grade was harmonized from multiple SEER variables, with final categories of grade I through IV and unknown.

Surgical treatment was extracted from the SEER surgery variable and harmonized as: no surgery, local excision (codes 10–29), wide excision (code 30), extended surgery (codes 40–79), and amputation (codes 80–90). A binary surgery variable (yes/no) was created for modeling. RT was classified as yes or no based on the SEER radiation recode. Chemotherapy was classified as yes or no/unknown based on the chemotherapy recode. A combined treatment modality variable was created with categories: surgery alone, surgery plus RT, surgery plus chemotherapy, surgery plus RT plus chemotherapy, nonsurgical treatment, no treatment, and unknown. Diagnosis year was categorized into 3 eras to account for changes in diagnostic criteria, staging systems, and treatment paradigms: 1992 to 2000, 2001 to 2010, and 2011 to 2022. The SEER chemotherapy variable is coded as “Yes” or “No/Unknown,” meaning that some patients classified as not receiving chemotherapy may have received it outside the reporting institution. Similarly, the radiation variable includes “None/Unknown” categories. Accordingly, treatment variables should be interpreted as receipt versus no documented receipt.

OS was defined as time from diagnosis to death from any cause or last follow-up. Cancer-specific survival (CSS) was defined using the SEER cause-specific death classification variable as time from diagnosis to death attributable to myxofibrosarcoma, with deaths from other causes censored. For competing risks analysis, the event indicator was coded as censored, cancer-specific death, or death from other causes.

Baseline patient demographics, tumor characteristics, and treatment patterns were summarized using frequencies and percentages for categorical variables. Characteristics were compared between survivors and non-survivors using chi-squared tests. Annual age-adjusted incidence rates were computed by direct standardization to the 2000 U.S. Standard Population (19 five-year age groups). The annual percent change (APC) in age-adjusted incidence was estimated by weighted log-linear regression of the log-transformed rates on year of diagnosis (weights = annual case counts), with APC and its 95% confidence interval derived from the regression slope as (exp(beta) − 1) × 100. A Poisson generalized linear model with log (population) offset was fit as a sensitivity analysis and yielded a concordant estimate.

OS and CSS probabilities were estimated using the Kaplan–Meier product-limit method. Survival rates were reported at 1, 3, 5, and 10 years. Differences in survival between subgroups were assessed using the log-rank test, with stratified analyses performed by stage, grade, primary site, surgery type, RT, age group, era of diagnosis, and sex. Each potential prognostic factor was evaluated individually using univariate Cox proportional hazards regression, and hazard ratios (HR) with 95% confidence intervals (CI) were calculated. Variables with clinical significance or P < .20 in univariate analysis were included in the multivariate Cox model. The final multivariate model included age group, sex, race, stage, grade, primary site (binary: extremity vs non-extremity), surgery (performed vs not), RT, chemotherapy, and era of diagnosis. The proportional hazards assumption was tested using Schoenfeld residuals.

To account for deaths from causes other than fibromyxosarcoma, competing risks analysis was performed using the cumulative incidence function method. cumulative incidence functions were calculated for cancer-specific death and death from other causes. Differences in cumulative incidence between groups were tested using Gray test. The Fine–Gray subdistribution hazard model was used to estimate subdistribution hazard ratios (sHR) for cancer-specific death while accounting for the competing risk of non-cancer death.

To examine prognostic factors within clinically relevant subpopulations, separate Cox regression models were fit for patients with localized disease only, patients with extremity tumors, and patients who received surgical treatment. Sensitivity analyses were performed restricting to patients for whom fibromyxosarcoma was their 1st malignant primary tumor, and to patients diagnosed 1998 or later when more complete treatment and staging data became available. Patients with unknown values for key variables were retained in the primary analysis but excluded from multivariate models requiring complete covariate data. No imputation was performed. A 2-sided P value of <.05 was considered statistically significant for all analyses.

All analyses were performed using R (version 4.3.2; R Foundation for Statistical Computing) with the survival, survminer, cmprsk, and gtsummary packages. The SEER database contains de-identified data that is publicly available for research purposes. This study was exempt from Institutional Review Board approval as it utilized existing publicly available data without patient identifiers and was conducted in accordance with the principles of the Declaration of Helsinki.

3. Results

A total of 1879 patients with histologically confirmed fibromyxosarcoma were identified from the SEER database between 1992 and 2022. No cases were excluded for nonmalignant behavior, death certificate or autopsy-only ascertainment, or missing survival data. The median patient age at diagnosis was 64 years (mean 61.3 years), with a slight male predominance (53.7%). The majority of patients were White (79.1%), followed by Asian/Pacific Islander (11.9%) and Black (6.5%). Most patients were married at diagnosis (75.7%). Baseline characteristics presented for the overall cohort (Table S1, Supplemental Digital Content 1 presents characteristics stratified by survival status) are presented in Table 1.

Table 1.

Overall MFS patient characteristics and demographics.

Characteristic	N = 1879*
Age at diagnosis
<40	264 (14%)
40–59	496 (26%)
60–74	595 (32%)
75+	524 (28%)
Unknown	0 (0%)
Sex
Male	1009 (54%)
Female	870 (46%)
Race
White	1486 (79%)
Black	122 (6.5%)
Asian/Pacific Islander	223 (12%)
American Indian/Alaska Native	17 (0.9%)
Unknown	31 (1.6%)
Marital status
Married	1055 (56%)
Single	367 (20%)
Divorced/separated	154 (8.2%)
Widowed	191 (10%)
Unknown	112 (6.0%)
Era of diagnosis
1992–2000	138 (7.3%)
2001–2010	515 (27%)
2011–2022	1226 (65%)
Primary site
Lower extremity	853 (45%)
Upper extremity	477 (25%)
Trunk	283 (15%)
Pelvis	101 (5.4%)
Head/neck	68 (3.6%)
Retroperitoneum	13 (0.7%)
Other	84 (4.5%)
Stage
Localized	959 (51%)
Regional	228 (12%)
Distant	54 (2.9%)
Unknown	638 (34%)
Grade
Grade I	296 (16%)
Grade II	672 (36%)
Grade III	329 (18%)
Grade IV	273 (15%)
Unknown	309 (16%)
Surgery type
No surgery	104 (5.5%)
Local excision	734 (39%)
Wide excision	875 (47%)
Extended surgery	81 (4.3%)
Amputation	10 (0.5%)
Unknown	75 (4.0%)
Radiation therapy
Yes	919 (49%)
No/unknown	960 (51%)
Chemotherapy
Yes	139 (7.4%)
No/unknown	1740 (93%)
Treatment modality
Surgery alone	820 (44%)
Surgery + RT	769 (41%)
Surgery + chemo	21 (1.1%)
Surgery + RT + chemo	90 (4.8%)
Nonsurgical treatment	47 (2.5%)
No treatment	57 (3.0%)
Unknown	75 (4.0%)

RT = radiation therapy.

^*n (%).

Annual case counts of fibromyxosarcoma increased substantially over the study period, from 3 cases in 1992 to 115 cases in 2022 (Fig. 1). The age-adjusted incidence rose from approximately 0.01 to 0.29 per 100,000 over the period, with a statistically significant APC of 6.4% (95% CI: 5.4%–7.5%; P < .001) estimated by weighted log-linear regression; the Poisson model with population offset gave a concordant 7.2% (95% CI: 6.5%–7.8%). When stratified by era of diagnosis, 138 cases (7.3%) were diagnosed during 1992 to 2000, 515 cases (27.4%) during 2001 to 2010, and 1226 cases (65.2%) during 2011 to 2022, reflecting both the rising incidence and the longer period captured in the most recent era.

Figure 1.

Age-adjusted incidence trends of myxofibrosarcoma from 1992 to 2022 in the SEER 12 registries. Age-adjusted rates (per 100,000 population, standardized to the 2000 US Standard Population) demonstrate a statistically significant increasing trend over the 30-year study period (APC: 6.4%, 95% CI: 5.4%–7.5%). APC = annual percent change, SEER = Surveillance, Epidemiology, and End Results.

The most common primary site was the lower extremity (45.4%), followed by upper extremity (25.4%), trunk (15.1%), pelvis (5.4%), head/neck (3.6%), and retroperitoneum (0.7%). Regarding disease extent, 51.0% of patients presented with localized disease, 12.1% with regional disease, and 2.9% with distant metastases; stage was unknown in 34.0% of cases. Tumor grade was distributed as follows: grade I (15.8%), grade II (35.8%), grade III (17.5%), grade IV (14.5%), and unknown (16.4%).

The majority of patients (90.5%) underwent surgical treatment. Surgical approaches included wide excision (46.6%), local excision (39.1%), extended surgery (4.3%), and amputation (0.5%). RT was administered to 48.9% of patients, while chemotherapy was received by only 7.4% of the cohort. The most common treatment modality was surgery alone (43.6%), followed by surgery combined with RT (40.9%), surgery with RT and chemotherapy (4.8%), nonsurgical treatment (2.5%), and no treatment (3.0%). Treatment patterns evolved over time, with the proportion of patients receiving wide excision increasing from 39.1% in 1992 to 2000 to 52.8% in 2011 to 2022, while chemotherapy use decreased from 15.9% to 6.4% over the same period (Fig. 2).

Figure 2.

Treatment patterns for myxofibrosarcoma over 3 eras (1992–2000, 2001–2010, 2011–2022). The proportion of patients receiving wide excision increased over time, while chemotherapy use remained limited across all eras.

During a median follow-up of 62 months, 631 deaths (33.6%) occurred. The Kaplan–Meier estimated OS rates were 93.0% (95% CI: 91.9%–94.2%) at 1 year, 81.6% (95% CI: 79.8%–83.5%) at 3 years, 74.3% (95% CI: 72.2%–76.5%) at 5 years, and 62.1% (95% CI: 59.5%–64.9%) at 10 years (Fig. 3). The CSS rates were consistently higher: 95.6% (95% CI: 94.6%–96.5%) at 1 year, 89.2% (95% CI: 87.7%–90.7%) at 3 years, 85.2% (95% CI: 83.4%–87.0%) at 5 years, and 80.1% (95% CI: 77.9%–82.3%) at 10 years.

Figure 3.

Kaplan–Meier overall survival curve for patients with myxofibrosarcoma. The median follow-up was 105 months (reverse Kaplan–Meier method). Overall survival rates were 93.0% at 1 year, 81.6% at 3 years, 74.3% at 5 years, and 62.1% at 10 years. The x-axis extends to 240 months to display the full follow-up range including 10-year estimates.

Significant differences in OS were observed across multiple clinical and demographic factors. Log-rank tests revealed statistically significant survival differences by age group (chi-squared = 498.1, P < .001), stage (chi-squared = 131.2, P < .001), grade (chi-squared = 110.6, P < .001), primary site (P < .001), surgery type (chi-squared = 148.3, P < .001), and era of diagnosis (P < .001). Sex (P = .4) and RT (P = .09) did not demonstrate statistically significant associations with survival in univariate analysis.

Survival decreased markedly with advancing age. Patients aged 75 years and older had substantially worse survival compared to those under 40 years (univariate HR 13.07, 95% CI: 9.07–18.82, P < .001). Disease stage was strongly prognostic, with 5-year OS rates declining from localized to regional to distant disease (Fig. 4). Patients with distant metastases had nearly 5-fold higher mortality risk compared to those with localized disease (univariate HR: 4.90, 95% CI: 3.56–6.73, P < .001). Higher tumor grade was associated with progressively worse outcomes, with grade III (HR: 3.23, 95% CI: 2.36–4.42, P < .001) and grade IV (HR: 3.26, 95% CI: 2.40–4.42, P < .001) tumors showing substantially elevated mortality compared to grade I tumors (Fig. 5).

Figure 4.

Kaplan–Meier survival curves stratified by disease stage with 95% confidence intervals. Patients with localized disease demonstrated significantly better survival compared to those with regional or distant disease (log-rank P < .001). Confidence intervals are displayed for all strata, including the distant stage group (n = 54), to illustrate the precision of estimates.

Figure 5.

Kaplan–Meier survival curves stratified by tumor grade (grades I–IV only; patients with unknown grade excluded from this figure). Higher tumor grade was associated with progressively worse survival outcomes (log-rank P < .001).

Anatomical location influenced prognosis, with retroperitoneal tumors demonstrating the worst survival (HR: 3.31, 95% CI: 1.70–6.43, P < .001) and trunk tumors showing modestly elevated risk (HR: 1.26, 95% CI: 1.01–1.57, P = .04) compared to lower extremity tumors. Patients who did not receive surgery had a 4-fold higher mortality risk (HR 4.05, 95% CI: 3.12–5.26, P < .001) (Fig. 6). Survival outcomes improved for patients diagnosed in more recent eras, with the 2011 to 2022 cohort showing improved survival compared to earlier periods (log-rank P < .001).

Figure 6.

Kaplan–Meier overall survival curves stratified by treatment modality, with categories matching Table 1 (patients with unknown treatment excluded, n = 1804). Six strata are shown: surgery alone (n = 820), surgery + RT (n = 769), surgery + chemotherapy (n = 21), surgery + RT + chemotherapy (n = 90), nonsurgical treatment (n = 47), and no treatment (n = 57). The log-rank test indicated significant heterogeneity across groups (P < .001). RT = radiation therapy.

The multivariate analysis included 1066 patients with complete data for all covariates (429 events; events-per-variable ratio = 22.6). Comparison of patients included versus excluded from the multivariate model revealed no significant differences in key covariates (Table S2, Supplemental Digital Content 2). The multivariate Cox proportional hazards model included age, sex, race, stage, grade, primary site (extremity vs non-extremity), surgery, RT, chemotherapy, and era of diagnosis. Testing of the proportional hazards assumption using Schoenfeld residuals revealed violations for stage (P = .005), surgery (P < .001), and chemotherapy (P = .02), with a significant global test (P < .001). The reported HRs represent time-averaged effects over the follow-up period. The global Schoenfeld test was significant (chi-squared = 48.48, df = 19, P < .001), with individual violations detected for stage (P = .005), surgery (P < .001), and chemotherapy (P = .019). As sensitivity analyses, a stratified Cox model (stratifying on stage and surgery) was constructed. Additionally, restricted mean survival time analyses at 60 months were performed for treatment variables that violated the proportional hazards assumption, providing a valid summary measure without requiring the proportional hazard assumption.

After adjustment for covariates, age remained the strongest independent prognostic factor. Compared to patients younger than 40 years, those aged 40 to 59 had a 2-fold increased mortality risk (HR: 2.05, 95% CI: 1.23–3.42, P = .006), those aged 60-74 had a 6.6-fold increased risk (HR 6.62, 95% CI: 4.08–10.75, P < .001), and those 75 years and older had nearly a 20-fold increased risk (HR: 19.77, 95% CI: 12.10–32.30, P < .001).

Stage retained strong prognostic significance in the multivariate model. Regional disease was associated with an 81% increased mortality risk (HR: 1.81, 95% CI: 1.44–2.28, P < .001), while distant disease conferred a 5.4-fold increase (HR: 5.43, 95% CI: 3.66–8.05, P < .001) compared to localized disease. High-grade tumors (grades III and IV) demonstrated approximately 80% to 90% higher mortality compared to grade I tumors (grade III: HR: 1.87, 95% CI: 1.25–2.78, P = .002; grade IV: HR: 1.79, 95% CI: 1.23–2.61, P = .002). A non-extremity primary site was associated with a 43% increased mortality risk (HR: 1.43, 95% CI: 1.15–1.77, P = .001).

Regarding treatment factors, patients who did not undergo surgery were associated with 82% higher mortality (HR: 1.82, 95% CI: 1.15–2.88, P = .011). The absence of documented RT was associated with 30% higher mortality (HR: 1.30, 95% CI: 1.05–1.60, P = .015). The apparent inverse association between documented chemotherapy receipt and mortality (HR: 0.49, 95% CI: 0.35–0.68, P < .001) reflects confounding by indication, as chemotherapy was selectively administered to patients with more aggressive disease. Sex (P = .29), race (all categories P > .05), and era of diagnosis (P > .17) were not independently associated with survival after multivariable adjustment (Fig. 7).

Figure 7.

Forest plot showing multivariate Cox proportional hazards regression results for overall survival. Hazard ratios with 95% confidence intervals are displayed for each prognostic factor. Note: Hazard ratios represent time-averaged effects. The proportional hazards assumption was violated for stage (P = .005), surgery (P < .001), and chemotherapy (P = .019).

Given the substantial proportion of deaths from causes other than fibromyxosarcoma, competing risks analysis was performed to estimate the cumulative incidence of cancer-specific death while accounting for the competing risk of death from other causes (Fig. 8). Gray test confirmed significant differences in cancer-specific mortality by stage (P < .001), while no significant difference was observed for death from other causes across stage groups (P = .47).

Figure 8.

Cumulative incidence functions for competing risks analysis by stage. Both cancer-specific death (solid lines) and other-cause death (dashed lines) are displayed. Gray test confirmed significant differences in cancer-specific mortality across disease stages.

The Fine–Gray subdistribution hazard model demonstrated that after accounting for competing risks, advancing age remained associated with increased cancer-specific mortality, though with attenuated HRs compared to the standard Cox model (age 60–74: sHR: 2.90, 95% CI: 1.50–5.60, P = .002; age 75+: sHR: 4.29, 95% CI: 2.18–8.44, P < .001). Stage retained strong prognostic significance for cancer-specific death (regional: sHR: 1.91, 95% CI: 1.34–2.74, P < .001; distant: sHR: 4.64, 95% CI: 2.64–8.18, P < .001). Notably, grade was not independently associated with cancer-specific mortality after accounting for competing risks (grade III: sHR: 1.89, 95% CI: 0.99–3.64, P = .055; grade IV: sHR: 1.74, 95% CI: 0.93–3.26, P = .081). Chemotherapy receipt was associated with increased cancer-specific mortality risk (sHR: 2.51, 95% CI: 1.57–4.00, P < .001 for chemotherapy vs no chemotherapy), consistent with its selective use in higher-risk patients.

To examine prognostic factors within clinically relevant subpopulations, separate multivariate Cox models were constructed for patients with localized disease, extremity tumors, and those who underwent surgical treatment.

Among patients with localized disease (n = 959), age remained the dominant prognostic factor, with patients 75 years and older having a 25-fold increased mortality risk compared to those under 40 (HR: 24.63, 95% CI: 12.57–48.26, P < .001). Female sex was independently associated with improved survival in this subgroup (HR: 0.73, 95% CI: 0.57–0.93, P = .011). Grade IV tumors showed elevated mortality (HR: 1.65, 95% CI: 1.07–2.55, P = .023), while primary site and surgery did not achieve statistical significance in this subgroup.

Among patients with extremity tumors (n = 1330), the prognostic impact of age was even more pronounced (age 75+: HR: 32.29, 95% CI: 14.43–72.25, P < .001). Stage and grade retained significance, with distant disease (HR: 3.84, 95% CI: 2.29–6.44, P < .001) and grade IV tumors (HR: 2.04, 95% CI: 1.24–3.34, P = .005) conferring substantially elevated risk. Female sex was associated with improved survival (HR: 0.77, 95% CI: 0.61–0.98, P = .034). Not receiving RT was associated with worse outcomes (HR: 1.37, 95% CI: 1.06–1.77, P = .015).

In the surgically treated subgroup (n = 1700), age, stage, grade, and primary site all retained independent prognostic significance. Notably, within this subgroup, surgery type (local excision, wide excision, and extended surgery) did not significantly impact survival after adjustment for other factors (all P > .45), suggesting that adequate surgical resection, rather than the extent of surgery, is the key determinant. Non-extremity site was associated with a 54% increased mortality (HR: 1.54, 95% CI: 1.23–1.92, P < .001). Grade demonstrated a clear dose-response relationship in this subgroup (grade II: HR: 1.62, 95% CI: 1.09–2.41, P = .016; grade III: HR: 2.45, 95% CI: 1.58–3.78, P < .001; grade IV: HR: 2.43, 95% CI: 1.61–3.66, P < .001).

Two sensitivity analyses were performed to assess the robustness of findings. First, analysis restricted to patients for whom fibromyxosarcoma was their 1st malignant primary tumor yielded similar results, with age, stage, site, and chemotherapy status retaining significance. The effect of surgery was attenuated and no longer reached statistical significance (HR: 1.73, 95% CI: 0.99–3.04, P = .056), and radiation was no longer significant (HR: 1.11, 95% CI: 0.86–1.42, P = .44). Second, analysis restricted to cases diagnosed in 1998 and later (when more complete staging and treatment data became available) produced results virtually identical to the main analysis, confirming the stability of effect estimates across the study period.

4. Discussion

This population-based study represents the largest comprehensive analysis of fibromyxosarcoma to date, encompassing nearly 2000 patients from the SEER database over a 30-year period. Our findings demonstrate that fibromyxosarcoma is a heterogeneous malignancy with generally favorable survival outcomes, particularly for localized disease, but with substantial mortality risk associated with advanced age, metastatic disease, and high histological grade. The survival rates observed in our cohort are consistent with the intermediate prognosis typically attributed to this soft tissue sarcoma subtype.

A striking finding of this study is the marked increase in fibromyxosarcoma incidence over the study period. This rising incidence likely reflects improved diagnostic recognition and reclassification of tumors previously categorized as other soft tissue sarcoma subtypes rather than a true epidemiological increase. The 2013 World Health Organization classification clarified the distinction between MFS and related entities, which may have contributed to more consistent pathological diagnoses in recent years.^[23] Similar trends have been observed in European registry data, with Amadeo and colleagues reporting increasing age-standardized incidence rates for MFS in France between 2000 and 2013.^[24]

Age emerged as the strongest independent prognostic factor in our analysis, with elderly patients demonstrating dramatically increased mortality risk compared to younger individuals. This finding aligns with the established predilection of MFS for elderly patients and underscores the challenges of managing this malignancy in an aging population. The pronounced effect of age persisted across all subgroup analyses, including patients with localized disease and those with extremity tumors. Recent investigations have begun to explore the biological and physiological mechanisms underlying this age-associated mortality. Kylies and colleagues demonstrated that sarcopenia, measured by computed tomography-morphometric analysis, independently predicts survival in MFS patients, with substantial decreases in skeletal muscle index associated with markedly shorter OS.^[25] These findings suggest that frailty markers may refine prognostication beyond chronological age alone.

The competing risks analysis provided additional insight into age-related mortality patterns. While age remained significantly associated with cancer-specific death in the Fine–Gray model, the effect was substantially attenuated compared to the standard Cox model. This attenuation reflects the considerable competing risk of non-cancer mortality in elderly patients and highlights the importance of considering all-cause mortality when counseling patients regarding treatment decisions. For older patients with indolent, localized disease, the risk of dying from other causes may exceed the risk of cancer-specific mortality, a consideration that should inform shared decision-making regarding treatment intensity.

Disease stage demonstrated the expected strong association with survival outcomes, with patients presenting with distant metastases experiencing markedly worse prognoses compared to those with localized disease. Regional disease also conferred elevated risk, though to a lesser degree. These findings are consistent with prior institutional series and population-based studies.^[21,26] Tumor grade similarly retained independent prognostic significance, with high-grade tumors demonstrating substantially elevated mortality compared to low-grade lesions. Sambri and colleagues reported similar grade-dependent outcomes in their institutional cohort, with pronounced differences in 5-year survival between grade I (100%), grade II (91.9%), and grade III (73.1%) tumors.^[27] The well-established association between histological grade and metastatic potential in MFS underscores the importance of accurate pathological assessment.

Primary tumor site influenced prognosis, with non-extremity locations associated with increased mortality risk compared to extremity tumors. Retroperitoneal tumors demonstrated particularly poor outcomes, consistent with the technical challenges of achieving adequate surgical margins in this location and the tendency for later presentation. The extremities represented the most common primary sites in our cohort, consistent with the established distribution pattern of MFS.^[28]

Surgical resection remained the cornerstone of treatment, with the vast majority of patients undergoing operative management. The absence of surgical treatment was associated with substantially increased mortality, emphasizing the critical importance of resection when feasible. However, among surgically treated patients, the specific surgical approach did not significantly impact survival after adjustment for other factors. This finding suggests that the achievement of tumor resection, rather than the extensiveness of the procedure per se, is the key determinant of outcome. Recent studies have emphasized the importance of surgical margins in MFS, given its propensity for infiltrative growth along fascial planes. Hoang and colleagues reported that margin status was the foremost determinant of local recurrence, with cumulative recurrence rates at 3.5 years of 80% for intralesional margins, 40% for marginal margins, and only 10% for wide margins.^[29] Tomas and colleagues similarly found that clear margins impacted recurrence-free survival, with R1/R2 resection conferring more than double the risk of recurrence.^[30]

Treatment patterns evolved substantially over the study period. The proportion of patients receiving wide excision increased over time, reflecting growing recognition of the importance of adequate margins in this infiltrative tumor. RT was administered to approximately half of patients, and its use was associated with improved survival in the multivariate model. The role of radiotherapy in MFS has been examined in several recent studies. Fahad and colleagues demonstrated that neoadjuvant RT was more effective than adjuvant therapy in reducing local recurrence, with particularly pronounced benefit among patients with negative surgical margins.^[31] Abdou and colleagues reported that preoperative radiation followed by surgery with frozen section margin analysis achieved excellent local control rates and uniformly negative margins.^[32]

Chemotherapy was received by a small minority of the cohort, reflecting the limited role of systemic therapy in soft tissue sarcomas and the predominantly localized nature of the disease at presentation. The apparent inverse association between documented chemotherapy receipt and mortality, likely reflecting confounding by indication (patients receiving chemotherapy had more advanced disease) in our multivariate model, represents confounding by indication rather than a true treatment effect. Chemotherapy was selectively administered to patients with more aggressive disease characteristics, which were not fully captured by the covariates included in the model. This phenomenon was confirmed in the competing risks analysis, where chemotherapy receipt was associated with increased cancer-specific mortality, again reflecting its use in higher-risk patients. These findings should not be interpreted as evidence against the use of chemotherapy in appropriate clinical contexts.

The competing risks analysis represents an important methodological contribution of this study. The substantial gap between OS and CSS rates highlights the significant burden of non-cancer mortality in this patient population. Gray test confirmed that stage significantly influenced cancer-specific mortality but not death from other causes, supporting the validity of our cause-of-death coding. The Fine–Gray model appropriately accounts for the competing risk structure and provides estimates that are directly interpretable as effects on the cumulative incidence of cancer-specific death. For clinicians counseling patients, these competing risks estimates may be more relevant than traditional HRs, particularly for elderly patients in whom non-cancer mortality is substantial.

Several limitations warrant consideration. The SEER database does not capture detailed information on chemotherapy regimens, radiation doses and fields, surgical margin status, tumor size in a standardized format across all years, performance status, or comorbidities. These unmeasured confounders may influence survival outcomes and limit causal inference regarding treatment effects. The diagnosis period spans 3 decades, during which diagnostic criteria, staging systems, and treatment approaches evolved substantially. Although era was included as a covariate, residual temporal confounding may exist. Grade and stage information was missing in a substantial proportion of cases, potentially introducing selection bias in the multivariate models. The proportional hazards assumption was violated for several variables, indicating that the reported HRs represent time-averaged effects rather than constant effects over the follow-up period. Finally, local recurrence data are not captured in the SEER database, precluding analysis of this important outcome in MFS. Additionally, the SEER chemotherapy and radiation variables include “No/Unknown” categories, meaning that some patients classified as not receiving treatment may have received it at non-reporting facilities. This misclassification would bias treatment effect estimates toward the null for true nonreceipt and introduce confounding for the unknown group.

In conclusion, this large population-based study demonstrates that fibromyxosarcoma is characterized by favorable intermediate-term survival, particularly for localized extremity tumors in younger patients. Age emerged as the dominant prognostic factor, with elderly patients facing substantially elevated mortality risk from both cancer and non-cancer causes. Disease stage, histological grade, and primary site provide additional prognostic information. Surgical resection remains the cornerstone of treatment, with RT being associated with improved survival in adjusted analyses, though unmeasured confounders including margin status, tumor size, and performance status preclude causal inference. The competing risks framework employed in this study provides clinically relevant estimates of cause-specific mortality that may inform shared decision-making, particularly for elderly patients with significant competing mortality risks. Future studies should focus on molecular characterization to identify patients who may benefit from systemic therapy and on optimization of local treatment strategies to minimize the substantial local recurrence risk inherent to this infiltrative malignancy.

Author contributions

Conceptualization: Akef Obeidat.

Data curation: Akef Obeidat.

Formal analysis: Akef Obeidat.

Methodology: Akef Obeidat.

Writing – original draft: Akef Obeidat.

Writing – review & editing: Akef Obeidat.