Risk factors for lymph node metastasis of soft tissue sarcomas of the head, neck, and extremities, and the clinical significance of negative lymph node dissection

Background This study sought to define the risk factors for lymph node metastasis (LNM) of soft tissue sarcomas (STS) of the head, neck, and extremities, and the clinical significance of negative lymph node dissection (NLND). Methods STS patient data in the Surveillance, Epidemiology, and End Results (SEER) database from 1988 to 2015 were extracted and pooled. Logistics regression analysis was used to identify risk factors for LNM, Cox proportional hazards and Fine–Grey’s models were used for survival analysis, and Propensity score matching analysis (PSM) was used to assess the impact of NLND on patient prognosis. Results A total of 3276 patients were enrolled in the study, of whom 283 (8.6%) developed LNM. Rhabdomyosarcoma had the highest rate of LNM (25.3%), followed by clear cell sarcoma (16.8%) and epithelioid sarcoma (12.4%), while leiomyosarcoma had the lowest rate of LNM (1.3%). Sex, tumor size, grade, histology, and site were significantly associated with LNM. For specific histologic subtypes of STS, NLND significantly improves overall survival (HR: 0.718, 95%CI 0.535–0.962; P = 0.026) and cancer-specific survival (HR: 0.699, 95%CI 0.506–0.967; P = 0.031) and reduces cancer-specific mortality (Gray’s test, P = 0.017). However, NLND did not improve overall survival (P = 0.46) or reduce cancer-specific mortality (Gray’s test, P = 0.772) of patients with leiomyosarcoma. Conclusions Histology is an independent risk factor for LNM in STS of the head, neck, and extremities. Prophylactic NLND treatment was necessary and had a clinical benefit for patients with STS who were at high risk for LNM but had no significant impact on the prognosis of patients with leiomyosarcoma. Supplementary Information The online version contains supplementary material available at 10.1186/s13018-022-03050-3.

It is generally believed that LNM is associated with a poorer prognosis [3,10,11]. Crettenand et al. found that LNM reduces both overall (median survival: 15.1 vs. 73.9 months, respectively; p = 0.002) and diseasefree survival (median disease-free survival: 8.0 vs. 33.0 months, respectively; p = 0.006) [12]. Metastatic disease is also the first sign of many occult malignancies, and metastases can cause complications that impair quality of life [13,14]. The 8th edition of the American Joint Committee on Cancer (AJCC) staging system defines lymph node involvement as stage IV disease in sarcomas of the trunk and extremities. Thus, identifying patients who are at risk for LNM and defining LNM-specific risk factors is important for clinical decision-making and patient prognosis. However, identifying these patients has always been a challenge. Identifying at-risk patients has been a challenge, however, and the prognostic factors of STS patients without LNM who have undergone surgical treatment are not well-defined. At present, there are no studies on the prognostic impact of NLND treatment on patients with STS of the head, neck, and extremities. Thus, the present study sought to identify the risk factors for LNM in STS patients with six histological subtypes, the prognostic factors of patients without LNM, and the impact of NLND treatment on patient prognosis.

Data sources and patient selection
Clinical data for this retrospective study was obtained from the Surveillance Epidemiology and End Results (SEER) database. The SEER database is free and publicly accessible, and individual consent for this retrospective analysis was waived because the patient information is anonymous. The study was conducted according to the Declaration of Helsinki (as revised in 2013). Patient data were downloaded using the SEER*Stat software (version 8.3.6). The final study population was selected based on the following inclusion criteria: (1) diagnosis based on positive pathology, (2) tumor sites in the head, neck, and extremities, (3) histological codes: 8890-1, 8893-6,   8900-8902, 8910, 8912, 8920, 9120, 9130, 9133, 9150,  9170, 9260, 9364, 9473, 8804, 8005, 9044, 4) specified tumor was the first primary tumor at the time of diagnosis, and 5) available data on lymph node status and survival month. The detailed inclusion and exclusion process is shown in Fig. 1.

Variables and outcomes
The following information was collected for each patient: (1) demographic variables including age, sex, race, marital status, and year of diagnosis, (2) oncological variables including tumor size, site, laterality, stage, grade, lymph node status, and histology, and (3) therapeutic variables including surgery and lymph node dissection. Prior research has shown that the oncologic features of STS patients differ by age group. For this study, age groups included those < 19 years of age (children and adolescents) and ≥ 19 years of age (adults). Since SEER began including information on lymph node dissection in 1988, data were only collected for cases recorded from 1988 to 2015. The years of diagnosis were divided into two categories: 1988-2005 and 2006-2015. Tumor histological subtypes were determined using the "ICD-O-3 Hist/ Behav" field in the SEER database. Six STS types with the following histological codes were included in the study: 8890-1, 8893-6, 8900-8902, 8910, 8912, 8920, 9120, 9130, 9133, 9150, 9170, 9260, 9364, 9473, 8804, 8005, and 9044. Tumor sites were divided into three groups according to the "Primary Site-labeled" field in the SEER database. Staging information was obtained according to the SEER historic stage A (1973-2015) field. Lymph node status information was obtained from the "EOD 10-nodes (1988-2003)" and "CS-lymph nodes (2004-2015)" fields for different periods, and further divided into negative (N0) and positive (N1). Regional lymph node dissection information was obtained from the "Regional nodes examined (1988+)" fields. Lymph node "dissection" was defined as the removal of most or all of the nodes in the lymph node chains that drain the area around the primary tumor, include lymphadenectomy, radical node dissection, and lymph node stripping. Patients with a history of lymph node aspiration or core biopsy, sentinel node procedures, or an uncertain number of removed lymph nodes were excluded. The patients were divided into non-NLND and NLND groups based on examined lymph node (ELN) data. Optimal tumor size cutoff values were defined using X-tile software (version 3.6.1) and divided into the following groups: < 4, 4-10, and ≥ 10 cm (Fig. 2). The primary outcomes were overall survival, defined as the time from diagnosis to death of any cause, and cancer-specific survival, defined as the time from diagnosis to death resulting from the primary STS.

Statistical analysis
Categorical variates are presented as frequencies and percentages, and continuous variates are presented as the median and interquartile range (IQR). For categorical variables, the Chi-square test was used to analyze between-group differences and for continuous variables, the Wilcoxon rank-sum tests was used. For baseline variables, univariate analysis using the Kaplan-Meier curve and the log-rank test was performed. When the P value was < 0.1, variables were included in multivariate Cox regression analysis to identify independent prognostic factors and estimate the hazard ratio (HR) and 95% confidence interval (CI) of each covariate. Risk factors for LNM in STS of the head, neck, and extremities were identified using univariate and multivariate logistic regression analyses. Then, we further analyzed the effect of NLND on the prognosis of patients with STS. Taking non-cancer-specific death as a competing risk to cancerspecific death, cumulative incidence function curves for patients with NLND were plotted, and Gray's test was performed to compare fatality rates. Multivariate analysis was performed on the cohort data to identify independent prognostic factors by using the Cox proportional hazards regression model and the Fine and Gray's regression model. In the Cox analysis, patients who were alive at the last follow-up were considered as censored cases, while in the Fine and Gray's regression model, non-cancer-specific death was considered a competing risk, and we used the sub-distribution hazards ratio (SHR) was used to represent the contribution of each variable to cancer-specific death. The effect of NLND on patient prognosis was also investigated. To eliminate the impact of other factors and minimize the selection bias between the NLND and non-NLND groups, patients from each group were matched 1:1 using propensity score matching (PSM). Chi-square was used to compare the clinicopathological features of the NLND and non-NLND groups. Variables that differed between the two groups as well as those thought to influence treatment for negative lymph nodes were included in the matching analysis. After PSM, Kaplan-Meier and cumulative incidence function curves were created. All statistical analyses were performed using R software (version 3.6.2). The R "survival" package was used for the Kaplan-Meier curve and the Cox regression analysis, the "Matchit" package was used for the PSM analysis, and the "timereg" and "cmprsk" packages were used for the competing risk analysis. A two-sided P value of < 0.05 was considered statistically significant.

Demographics and clinical characteristics of patients with STS of the head, neck, and extremities
A total of 3276 eligible cases were included in this study. The median age (IQR range) of the entire cohort population was 53 years (31-69 years). The lymph node positivity rate for the entire cohort was 8.6% and varied significantly between tumor types ( Fig. 3A, P < 0.001). Lymph node positivity was the highest for rhabdomyosarcoma (25.3%), followed by clear cell sarcoma (16.8%) and epithelioid sarcoma (12.4%), and was the lowest for leiomyosarcoma (1.3%). The cohort was divided into a negative lymph node group (NLN) and a positive lymph node group (PLN). Patients in the PLN group were significantly younger than those in the NLN group [median (IQR), 24 years (10-51 years) vs. 54 years (35-70 years), respectively; P < 0.001] ( Table 1).
In the PLN group, tumors were mostly located in the head, neck, and lower extremities, while in the NLN group, they were primarily found in the lower extremities (73.1% vs. 55.7%, respectively; P < 0.001). Rhabdomyosarcoma was predominant in the PLN group (51.9%), while leiomyosarcoma was predominant in the NLN group (52.7%). The rate of NLND was higher in the PLN group than in the NLN group (56.5% vs. 12.6%, respectively; P < 0.001). There were also significant differences in race, marital status and tumor grade, stage, size, laterality, and surgery between the two groups (P < 0.05). The median survival time was significantly shorter in the PLN than NLN groups [median (IQR), 23 months (11.5-74.5 months) vs. 55 months (22-112 months), respectively; P < 0.001].

Risk factors for LNM of STS in head and neck and extremities
Our results suggested that patients with LNM have a poorer prognosis (Fig. 3B), reinforcing the need to identify risk factors. Univariate logistics and multivariate cox regression analyses were used on the target cohort to identify risk factors for LNM. Variables with P values < 0.1 in the univariate analysis were included in the multivariate analysis to adjust for potential confounding ( ; P < 0.001) were associated with increased risk of cancer-specific death. However, NLND was not associated with cancer-specific mortality (sHR: 0.838, 95% CI 0.662-1.061; P = 0.14). Sex, race, year of diagnosis, tumor site, and laterality had no statistical effect on cancer-specific death ( Table 3).

The effect of NLND on the prognosis of patients with different soft tissue sarcomas
The effect of NLND on the prognosis of patients with different histological STS subtypes was assessed. NLND was an independent risk factor for the prognosis of five subtypes of STS at high risk for LNM: rhabdomyosarcomas, angiosarcomas, Ewing sarcomas, epithelioid sarcomas, and clear cell sarcomas. In all cases, the NLND group had significantly higher overall and cancer-specific survival rates (Fig. 4A, B). Cumulative incidence function curves were also plotted to compare the differences in cancer-specific mortality between the two groups. As shown in Fig. 6A, the cumulative incidence of cancer-specific death was lower in the NLND group than in the non-NLND group (Gray's test, P = 0.001).  However, NLND did not improve cancer-specific survival in patients with leiomyosarcoma (Fig. 5B). To further clarify the impact of NLND on patient prognosis, PSM analysis was performed. First, Chi-square was used to compare the baseline characteristics of patients in the NLND and non-NLND groups. Results showed significant differences in age, marital status and tumor site, laterality, histology, stage, and size, between the two groups (p < 0.05, Table 4). The covariables were then matched to create a new cohort. After PSM, the covariates were balanced between the two groups, and the results further confirmed that NLND significantly improved both overall (HR: 0.718, 95%CI 0.535-0.962; P = 0.026) and cancer-specific survival (HR: 0.699, 95%CI 0.506-0.967; P = 0.031) ( Table 5 and Fig. 4C, D). Patients in the NLND group had significantly lower cancer-specific mortality (Gray's test, P = 0.017) (Fig. 6B), and significantly longer median survival time than patients in the non-NLND group, both before (median: 69 vs. 47 months, respectively; p < 0.001) and after PSM (median: 69.5 vs. 57.5 months, respectively; p = 0.080). However, NLND did not improve overall survival (P = 0.46) or reduce cancer-specific mortality (Gray's test, P = 0.772) of patients with leiomyosarcoma (Figs. 5C, D, 6D).

Discussion
The prognosis for STS patients with metastasis is poor. The primary STS metastasis sites include the lung and bone, with LNM being relatively rare. However, according to the latest AJCC staging guidelines, sarcomas of the trunk and extremities that involve the lymph nodes are defined as stage IV disease [1].  [15]. The current study included six types of STS of the head, neck, and extremities that are associated with the highest risk of LNM. The prognosis of 2756 STS patients without LNM who received surgical treatment was quantified. This study also clarified the impact of NLND on STS patient prognosis. In our cohort, rhabdomyosarcoma (25.3%), clear cell sarcoma (16.8%), and epithelioid sarcoma (12.4%) patients had the highest rates of LNM, while leiomyosarcoma patients had the lowest rate of LNM (1.3%). This is consistent with previously reported rates of lymph nodepositivity in different STS types (26.7% for rhabdomyosarcoma, [16][17][18].8% for clear cell sarcoma, and 13-14.5% for epithelioid sarcoma) from several large cohort studies [8,11,16]. In children and adolescents (< 19 years of age), rhabdomyosarcoma had the highest lymph node-positivity rate, while in adults (≥ 19 years), clear cell sarcoma had the highest positivity rate. These results are similar to those found in previous studies [8,15]. This may be the result of differences in the histological subtypes of STS by age. Indeed, rhabdomyosarcoma was the most common STS in children and adolescents and accounts for onehalf of pediatric STS [17]. Independent risk factors for LNM were identified using the multivariate Cox proportional risk model. Patients who were male or had STS in the head and neck, high-grade (III + IV), tumors > 4 cm in size, or non-leiomyosarcomas were more likely to have LNM. Several studies have reported similar results. Miccio et al. found that patients with high-grade STS and histology including clear cell, angiosarcoma, rhabdomyosarcoma, and epithelioid (CARE) sarcomas were associated with LNM [7]. Behranwala et al. examined   [15]. Other studies have shown that LNM also correlates with age and primary site [10,18], but none have reported an association between gender and LNM. Our results suggest that males are at higher risk of LNM than females (OR: 1.291, 95% CI 1.012-1.646; P = 0.040). Studies show that males have a higher predisposition for STS than females, but the extent to which gender affects LNM has not been established [19]. This study also performed a prognostic analysis of 2756 patients without LNM who had undergone surgery. Age, grade, stage, size, histology and marital status were found to be independent prognostic factors for cancer-specific survival. This result is similar to previous studies. A study with patient data also from the SEER database showed that for the historically high-risk extremity, STS, age,  grade, size, surgery, and regional lymph node status were independent disease-specific prognostic factors [11].
Another study found that for epithelial sarcoma, tumor site was a prognostic factor for event-free survival and overall survival, and extremities site had a better prognosis than proximal-type variant [20]. Understanding these STS characteristics can inform clinical counseling and personalized treatment for patients. The current study further investigated the association between NLND and disease prognosis. After PSM, we found that NLND was an independent prognosis factor for patients with a high risk of LNM such as rhabdomyosarcomas, angiosarcomas, Ewing sarcomas, epithelioid sarcomas, and clear cell sarcomas. Surprisingly, NLND did not improve the prognosis of leiomyosarcoma before and after PSM. Further analysis showed that patients with leiomyosarcoma were older than the rest five types of sarcomas (median age: 61 vs. 41 years, respectively; P < 0.001), and only a small proportion of leiomyosarcoma patients (6.81%) received NLND than those with other tumors (21.2%). Unfortunately, no information about why these patients did not receive a lymph node dissection is available in the SEER database. This may be because these patients are older or have comorbidities or treatment contraindications. In addition, the SEER database does not record information about adjuvant treatments such as radiotherapy and chemotherapy.
The absence of these data may affect the accuracy of the results and cause confounding bias so further clinical validation is needed.
Currently, there is no consensus guidelines on how to assess lymph node status in high-risk STS patients. Common strategies for lymph node status assessment include physical examination, ultrasound, CT, MRI, and sentinel lymph node biopsy (SLNB). Radical lymphadenectomy or systemic chemotherapy are often recommended for the treatment of regional lymph node metastases, but treatment modalities are mostly clinician-and centerdependent [19]. Some studies have reported the prognosis of lymph node examination/dissection for STS with LNM, but the result remains controversial. Al-Refaie et al. suggest that regional lymph node dissection may prolong survival time [21]. Ecker et al. support regional lymph node examination for patients with epithelioid and possibly clear cell sarcoma [22]. Brady et al. found that lymph node sampling was associated with improved disease-specific survival in patients with extremity rhabdomyosarcoma (64% vs. 49%, P = 0.005) [23]. Riad et al. showed that resection of involved lymph nodes had an estimated 5-year survival of 57%, whereas nine patients treated without surgery all died within 30 months [24]. In another study, NLND was proved to be an independent risk factor for cancer-specific survival of non-metastatic colorectal sarcomas patients [25]. However, some studies found that lymph node examination/dissection had no impact on prognosis. One study of epithelioid sarcomas found that lymphadenectomy did not improve the overall survival of patients with LNM [26]. Another study found that resection of the metastatic lymph node had better survival at 1.5 years, but did not improve the long-term survival of patients with STS [9]. Some studies argued that the management of positive lymph nodes remains uncertain and that more research is required to assess the impact of lymphadenectomy on the overall survival of STS patients with LNM [3,10]. The current study suggests that NLND is an appropriate treatment for specific STS patients such as those at high risk for LNM. Nevertheless, we still recommend that patients' treatment decisions should be based on the clinical reality of the patient, because lymph node dissection may have some acute and chronic complications, such as lymphorrhea, chylous ascites, seroma, delayed wound healing, and chronic lymphedema [27]. This study has some limitations which need to be considered. First, this is a retrospective study and may have inherent limitations, so the results must be validated using prospective studies. Second, the SEER database does not record detailed information about chemotherapy, radiotherapy, comorbidities, complications, and recurrence, which may have a potential impact on the results. Despite these limitations, our findings are of significance.

Conclusion
This study identified the rate of LNM in patients with six STS subtypes of the head, neck, and extremities. In addition, risk factors for LNM and the prognostic factors for STS patients without LNM were further clarified. Most importantly, our study suggests that prophylactic lymph node dissection was necessary and had a clinically beneficial for STS patients at high risk for LNM in the head, neck, and extremities. However, for leiomyosarcoma, NLND did not improve the prognosis and prophylactic lymph node dissection needs to be more carefully evaluated.
Additional file 1: Table S1. Univariate logistics and Multivariate Cox analysis of cancer-specific survival for 2756 patients without LNM.