A novel preoperative risk score to guide patient selection for resection of soft tissue sarcoma lung metastases: An analysis from the United States Sarcoma Collaborative

Abstract

Background:

Surgical resection for sarcoma lung metastases has been associated with improved overall survival (OS).

Methods:

Patients who underwent curative-intent resection of sarcoma lung metastases (2000–2016) were identified from the US Sarcoma Collaborative. Patients with extrapulmonary metastatic disease or R2 resections of primary tumor or metastases were excluded. Primary endpoint was OS.

Results:

Three hundred and fifty-two patients met inclusion criteria. Location of primary tumor was truncal/extremity in 85% (n = 270) and retroperitoneal in 15% (n = 49). Forty-nine percent (n = 171) of patients had solitary and 51% (n = 180) had multiple lung metastasis. Median OS was 49 months; 5-year OS 42%. Age ≥55 (HR 1.77), retroperitoneal primary (HR 1.67), R1 resection of primary (HR 1.72), and multiple (≥2) lung metastases (HR 1.77) were associated with decreased OS(all p < 0.05). Assigning one point for each factor, we developed a risk score from 0 to 4. Patients were then divided into two risk groups: low (0–1 factor) and high (2–4 factors). The low-risk group (n = 159) had significantly better 5-year OS compared to the high-risk group (n = 108) (51% vs. 16%, p < 0.001).

Conclusion:

We identified four characteristics that in aggregate portend a worse OS and created a novel prognostic risk score for patients with sarcoma lung metastases. Given that patients in the high-risk group have a projected OS of <20% at 5 years, this risk score, after external validation, will be an important tool to aid in preoperative counseling and consideration for multimodal therapy.

1 ∣. INTRODUCTION

The lungs are the most frequent site of metastasis for soft tissue sarcoma (STS).^1,2 Risk of metastasis is largely driven by tumor grade; approximately 60% of patients with high-grade tumors will develop lung metastases, with the vast majority occurring within two years of primary tumor resection.^3,4 Pulmonary metastasectomy for STS was first described in the 1970s by Huang and colleagues, who presented fifty cases performed between 1956 and 1977, reported a median overall survival (OS) of 22.2 months and cited a shorter tumor doubling time as an indicator of poor prognosis.⁵ Since this report, pulmonary metastasectomy for STS has been increasingly performed, though there are no randomized controlled trials demonstrating its efficacy. Generally accepted selection criteria for patients considered for metastasectomy include control of the primary tumor, absence of unresectable extrapulmonary metastasis, technical resectability of the pulmonary metastases, and good functional status.^6,7

Modern series have reported 5-year OS after pulmonary metastasectomy between 15% and 52%, representing meaningful improvement compared to best supportive care and nonoperative management, and, given the lack of novel, effective systemic therapies for patients with advanced STS, surgical resection remains the treatment of choice for resectable lung metastases.^2,6,8-11 However, it stands to reason that not all patients would benefit equally from pulmonary metastasectomy. Various prognostic factors have been identified, including tumor doubling time, disease-free interval before metastasis, completeness of metastasectomy, number and size of pulmonary nodules, and presence of unilateral versus bilateral disease, though the relative influence and impact of each of these factors remains controversial.^1,2 Additionally, the majority of studies identifying prognostic factors were conducted in single-institutions with relatively small sample sizes, given the rare nature of the disease.^1,2,5,12

Thus, our aims for this study were to identify preoperative prognostic factors associated with survival and to develop a risk score to stratify patients being considered for lung metastasectomy using the United States Sarcoma Collaborative (USSC) large, multi-institutional database.

2 ∣. METHODS

The USSC is a collaboration of high-volume academic tertiary and quaternary referral centers including Emory University, Stanford University, The Ohio State University, Medical College of Wisconsin, University of Chicago, Wake Forest University, University of Wisconsin, and Washington University in St Louis. Institutional Review Board (IRB) approval was obtained at each study site before data collection. Participating institutions determined a series of objectives and designed a standardized data collection tool as well as a data dictionary. A master database was created with variables of interest and distributed to all sites. A thorough retrospective review of the electronic medical record was then conducted. All adult patients undergoing resection of lung metastases were evaluated. Pertinent baseline intraoperative, pathologic, and postoperative outcome data were collected. Staging was based on the American Joint Committee on Cancer 7th edition. Data regarding disease recurrence and survival were also recorded. Primary and metastatic tumors were entered separately in the database, and thus patients potentially underwent treatment of their primary tumor at a different facility. Additionally, because primary and metastatic tumors were entered separately, patients receiving chemotherapy in our analysis received it specifically for their metastatic disease. Tumor grade was reported using either the FNCLCC or AJCC criteria; these criteria were merged to form two categories, “low” and “high” grades, for the purpose of analysis. Disease status was determined at time of last follow up at each respective institution.

Patients who underwent curative-intent resection of STS lung metastases between 2000 and 2016 were included. Patients with extrapulmonary metastatic disease or R2 resections of either primary tumor or lung metastases were excluded. Primary endpoint was OS.

2.1 ∣. Statistical analysis

Statistical analysis was conducted using SPSS 26.0 software (IBM Inc.). Descriptive analyses were performed for the entire cohort. OS was defined as the time from surgery to date of death due to any cause or date of last available follow up. Age was defined as <55 and ≥55 using a receiver operating characteristic curve with the outcome of overall survival. χ² analysis was used to compare categorical variables, and Student's t test or one-way analysis of variance was used for continuous variables, where indicated. The univariate associations between each covariate including study cohorts and primary study outcome of OS were assessed using Kaplan-Meier curves. Multivariable models were created using Cox survival analyses. For risk score development, covariates associated with OS (p ≤ 0.1) on univariate Cox regression were included in a final multivariable model. Covariates significantly associated with OS (p < 0.05) on multivariable analysis were used to create the final risk score.

3 ∣. RESULTS

3.1 ∣. Patient characteristics and outcome

Of the 4153 patients in the USSC, 352 underwent resection of lung metastasis that met inclusion criteria (Figure 1). Average age was 53 years and 57% (n = 200) were male. Eighty-five percent (n = 270) had primary tumors of the trunk or extremity and 15% (n = 49) of the retroperitoneum. Ninety-two percent (n = 322) of patients had metachronous lung metastases. Most common tumor histologies included undifferentiated pleomorphic sarcoma (26%, n = 92), leiomyosarcoma (18%, n = 63), and synovial sarcoma (15%, n = 54) and the majority (94%, n = 236) of tumors were high-grade. Median number of lung metastases was 2 (interquartile range [IQR] 1–3, range 1–11). Twenty-eight percent (n = 97) of patients received chemotherapy. Sixty percent (n = 209) of patients had a thoracotomy with wedge resection, 39% (n = 135) a video-assisted thoracoscopic (VATS) wedge resection, and 1% (n = 6) a thoracotomy with lobectomy to remove their lung metastases. Over time, patients were more likely to undergo VATS wedge resection; 22% (n = 43) between 2000 and 2009 had a VATS procedure compared to 64% (n = 92) from 2010 to 2016, p < 0.001. The median follow-up for survivors was 26 months. The median OS was 49 months with a 5-year OS of 42% (Table 1).

FIGURE 1.

Patient selection

TABLE 1.

Baseline characteristics

Variable	N (%)
Age mean ± SD	53.4 ±16.2
Gender
Male	200 (56.8)
Female	152 (43.2)
Smoking history	44 (12.8)
Severe COPD	4 (1.2)
Dyspnea on presentation	6 (1.7)
Primary tumor location
Truncal/extremity	270 (84.6)
Retroperitoneal	49 (15.4)
Primary tumor resection margin
R0	236 (84.3)
R1	44 (15.7)
Timing of lung metastasis
Synchronous	30 (8.5)
Metachronous	322 (91.5)
Tumor histology
UPS	92 (26.1)
LMS	63 (17.9)
Synovial sarcoma	54 (15.3)
Other histology	143 (40.6)
Tumor grade
Low grade	18 (6.4)
High grade	264 (93.6)
Number of lung metastases, median (IQR)	2 (1–3)
Chemotherapy	97 (28.2)
Metastasectomy type
Open wedge resection	209 (59.7)
Thoracoscopic wedge resection	135 (38.6)
Open lobectomy	6 (1.7)
Thoracoscopic lobectomy	0(0)
Disease status
No evidence of disease	90 (25.7)
Alive with disease	105 (30.0)
Dead of disease	137 (39.1)
Dead of other causes	3 (0.9)
Dead of unknown causes	15 (4.3)
Median OS (months)	49
5-year OS	42%

Abbreviations: COPD, chronic obstructive pulmonary disease; IQR, interquartile range; LMS, leiomyosarcoma; OS, overall survival; UPS, undifferentiated pleomorphic sarcoma.

3.2 ∣. Factors associated with overall survival

On univariate analysis, age ≥55, retroperitoneal primary tumor location, R1 resection of primary tumor, high tumor grade, multiple (≥2) lung lesions, and receipt of chemotherapy were associated with decreased OS in patients who underwent pulmonary metastasectomy (Table 2). Timing of metastatic disease (synchronous vs. metachronous) and metastasectomy type (VATS vs open resection) were not associated with OS (Table 2). On multivariable analysis, age ≥55 (HR 1.872, 95% CI 1.242–2.821, p = 0.003), retroperitoneal primary tumor location (HR 1.683, 95% CI 1.079–2.626, p = 0.022), R1 resection of primary tumor (HR 1.689, 95% CI 1.026–2.780, p = 0.039) and multiple (≥2) lung lesions (HR 1.704, 95% CI 1.109–2.617, p = 0.015) were independently associated with decreased OS. Chemotherapy was not associated with OS on multivariable analysis (Table 2).

TABLE 2.

Univariate and multivariable Cox regression analysis for overall survival

Variable	Univariate analysis		Multivariable analysis
	HR (95% CI)	p value	HR (95% CI)	p value
Age
<55	Reference	<0.001	Reference	0.003
≥55	2.353 (1.675–3.305)		1.872 (1.242–2.821)
Gender			-	-
Female	Reference	0.465
Male	0.888 (0.645–1.222)
Smoking history	0.729 (0.439–1.211)	0.222	-	-
Primary tumor location
Truncal/extremity	Reference	0.014	Reference	0.022
Retroperitoneal	1.638 (1.106–2.427)		1.683 (1.079–2.626)
Primary tumor resection margin
R0	Reference	0.056	Reference	0.039
R1	1.544 (0.989–2.409)		1.689 (1.026–2.780)
Tumor grade
Low	Reference	0.006	Reference	0.145
High	4.094 (1.510–11.105)		2.184 (0.765–6.239)
Final metastasectomy margin status			-	-
R0	Reference	0.124
R1	1.486 (0.897–2.463)
# Lung lesions
Single	Reference	0.003	Reference	0.015
Multiple	1.626 (1.180–2.241)		1.704 (1.109–2.617)
Metastasectomy type			-	-
VATS	Reference	0.163
Open	0.789 (0.565–1.101)
Timing of metastases			-	-
Metachronous	Reference	0.164
Synchronous	0.633 (0.333–1.205)
Chemotherapy	1.456 (1.038–2.045)	0.029	1.314 (0.833–2.073)	0.240

Note: Bold values indicate statistical significance.

Abbreviation: VATS, video-assisted thoracoscopic.

3.3 ∣. Risk score for patients undergoing lung metastasectomy

On final multivariable analysis, age ≥55 (HR 1.767, 95% CI 1.171–2.668, p = 0.007), retroperitoneal primary tumor location (HR 1.665, 95% CI 1.071–2.588, p = 0.023), R1 resection of primary tumor (HR 1.715, 95% CI 1.042–2.824, p = 0.034), and multiple (≥2) lung lesions (HR 1.766, 95% CI 1.174–2.656, p = 0.006) were independently associated with decreased OS (Table 3). Based on similar hazard ratios, each risk factor was assigned a score of 1. Patients with 0 risk factors (n = 32) had a 5-year OS of 60.9%, 1 risk factor (n = 127) 48.1%, 2 risk factors (n = 83) 16.9%, 3 risk factors (n = 24) 9.8%, and all 4 risk factors (n = 3) 0.0% (Figure 2). Based on similar OS for patients with 0 or 1 risk factor and patients with 2, 3, and 4 risk factors, these groups were combined to form low- and high-risk groups. Patients in the low-risk group had a median OS of 62 months and 5-year OS of 50.7% compared to a median OS of 25 months and 5-year OS of 16.2% seen in the high-risk group (p < 0.001) (Figure 3). Patients in the high-risk group were more likely to have leiomyosarcoma histology and more likely to receive chemotherapy than patients in the low-risk group (Table 4). Gender, surgery year (2000–2009 vs. 2010–2016), metastasectomy type (VATS vs. open), final margin status, and major postoperative complication rates were similar between the two groups (Table 4).

TABLE 3.

Final multivariable Cox regression analysis for risk score formation

	HR (95% CI)	p value
Age
<55	Reference	0.007
≥55	1.767 (1.171–2.668)
Primary tumor location
Truncal/extremity	Reference	0.023
Retroperitoneal	1.665 (1.071–2.588)
Primary tumor resection margin
R0	Reference	0.034
R1	1.715 (1.042–2.824)
Tumor grade
Low	Reference	0.132
High	2.237 (0.785–6.373)
# Lung lesions
Single	Reference	0.006
Multiple	1.766 (1.174–2.656)

Note: Bold values indicate statistical significance.

Abbreviations: CI, confidence interval; HR, hazard ratio.

FIGURE 2.

Stratification and survival for patients who underwent lung metastasectomy by number of risk factors (risk factors included: age ≥ 55, retroperitoneal primary tumor, R1 primary tumor resection, and multiple lung metastasis)

FIGURE 3.

Stratification and survival for patients who underwent lung metastasectomy by risk factor group (risk factors included: age ≥ 55, retroperitoneal primary tumor, R1 primary tumor resection, and multiple lung metastasis)

TABLE 4.

Patient characteristics by risk score group

	Low risk n = 159	High risk n = 110	p value
Gender			0.399
Male	99 (62.3%)	62 (56.4%)
Female	60 (37.7%)	48 (43.6%)
Surgery year			0.597
2000–2009	84 (52.8%)	62 (56.9%)
2010–2016	75 (47.2%)	47 (43.1%)
Histology			<0.001
UPS	47 (29.6%)	32 (29.1%)
LMS	14 (8.8%)	31 (28.2%)
Synovial sarcoma	24 (15.1%)	15 (13.6%)
Other	74 (46.5%)	32 (29.1%)
Metastasectomy type			0.515
VATS	73 (46.5%)	45 (41.7%)
Open	84 (53.5%)	63 (58.3%)
Final metastasectomy margin			0.075
R0	149 (94.9%)	97 (88.2%)
R1	8 (5.1%)	13 (11.8%)
Chemotherapy	33 (21.0%)	35 (33.3%)	0.037
Major complication	3 (1.9%)	2 (1.8%)	1.000

Note: Bold values indicate statistical significance.

Abbreviation: VATS, video-assisted thoracoscopic.

4 ∣. DISCUSSION

In a modern, large, multi-institutional cohort of patients, we sought to determine preoperative factors associated with decreased survival after resection of soft tissue sarcoma lung metastases and create a simple risk score to aid in patient selection and informed consent before metastasectomy. Median survival in this cohort was 49 months, with a 5-year OS of 42%, however we found that survival varied widely amongst groups of patients with specific characteristics. Age ≥ 55, R1 resection of primary tumor, retroperitoneal primary tumor location, and multiple (≥2) lung metastases were clinicopathological factors associated with decreased OS in patients undergoing pulmonary metastasectomy. Further, the presence of multiple of these factors portends worse survival; patients with 0–1 factors had a 5-year OS of 50.7%, while patients with 2–4 of these factors had a significantly worse OS of 16.2%.

Pulmonary metastases are the most common cause of death in patients with soft tissue sarcomas and metastatic disease carries a poor prognosis; patients with pulmonary lesions have a median survival of approximately 11–18 months.^1,10 Metastasectomy offers a potentially curative option for select patients with completely resectable disease, with a range of reported 5-year OS after complete resection between 25% and 53%, similar to the 5-year OS of 42% seen in our cohort.^13,14 While there have not been any randomized trials comparing surgical and nonsurgical management of pulmonary metastases, Shimizu and colleagues¹⁴ found significantly improved survival in patients who underwent metastasectomy; 53% compared to 0% at 5 years and median OS of 101 compared to 11 months (p = 0.029). Currently, metastasectomy is the standard of care for patients with resectable disease; however, it is important to reiterate that patient selection is paramount in achieving these outcomes.

Though the majority of factors present in our risk score including control of the primary tumor and number of lung metastases are echoed in the literature, increasing age was not found to be a significant predictor of decreased OS in other studies. Cheung and colleagues found that age >58 (the median age of their cohort) was not associated with decreased OS on univariate analysis.¹⁵ Similarly, Chudgar and colleagues¹⁶ found that patients aged ≥ 50 had similar median disease-free and OS compared to younger patients. The association of increasing age with decreased survival in our cohort is likely multifactorial, with increasing age potentially representing increasing rate of comorbidities and decreased tolerance of loss of lung parenchyma. This difference in results could also be due to selection bias in single institution compared to multi-institutional studies. Increasing age has been associated with decreased survival in patients with soft tissue sarcoma in other large, multi-institutional studies; for example, Stahl and colleagues noted this association using the National Cancer Database.¹⁷ Our results indicate differences in patient selection in large, high-volume centers across the country compared to single-institution experiences.

Historically, control of the primary tumor and complete resection have been defined as no gross tumor remaining, in other words, an R0 or R1 resection.¹⁷ While an R1 resection is believed to increase the risk of local recurrence, there is certain evidence that it also portends worse survival. Stahl and colleagues¹⁷ found that patients who underwent an R1 resection of a retroperitoneal sarcoma had worse survival compared to patients with an R0 resection in a propensity score-matched cohort. Similarly, in an analysis of 2084 patients, Stojadinovic and colleagues¹⁸ found inferior distant recurrence-free and disease-specific survival in patients with R1 resections of soft tissue sarcomas from any primary tumor site. These findings are echoed in our analysis. Interestingly, microscopic margin status of the metastasectomy does not appear to have the same prognostic impact. This finding is both contradicted and supported in the literature. For example, Yamamoto and colleagues² found a significant survival difference in patients with complete (defined as R0) compared to incomplete metastasectomy, while an analysis by Chudgar and colleagues¹⁶ did not find metastasectomy margin status to be independently associated with survival.

The fact that patients with retroperitoneal primary sarcomas should have worse survival than those with truncal/extremity primary tumors in a population with uniformly metastatic disease is likely due to differences in histology. In our lung metastasectomy cohort, retroperitoneal primary tumors were most commonly leiomyosarcomas (non-uterine: 35%, uterine: 18%) followed by dedifferentiated liposarcomas (14%) and undifferentiated pleomorphic sarcomas (10%). In contrast, truncal/extremity primary tumors were most commonly undifferentiated pleomorphic sarcomas (31%), followed by synovial sarcomas (16%), and leiomyosarcomas (11%). Truncal/extremity primary tumors also exhibited a wider variety of histologic subtypes, with 25 unique histologies represented, compared to 10 in patients with retroperitoneal primary tumors. Additionally, patients in the high-risk group using our risk score were significantly more likely to have leiomyosarcoma histology compared to those in the low-risk group (28% vs. 9%, p < 0.001, Table 4). Historically, soft tissue sarcomas have been considered a single entity, however more recently there is evidence that histology can not only impact treatment effectiveness, but also disease-free survival and recurrence patterns, and this appears to be particularly true of retroperitoneal sarcomas.^19,20 These findings in our cohort add to the impetus for additional, prospective, histology-specific investigation moving forward.

Disease-free interval is widely cited as an important prognostic factor in patients undergoing resection for lung metastases, with longer intervals, generally defined as a cut point between 12 and 36 months, associated with improved survival.^10,15,21,22 Interestingly, patients with synchronous lesions, defined in our database as metastatic lesions diagnosed within 6 months of diagnosis of the primary tumor, did not have significantly worse survival after metastasectomy, compared to patients with metachronous lesions. However, this is likely due to careful patient selection and utilization of prior prognostic evidence, as the vast majority (92%) of patients in our cohort had metachronous metastases.

Finally, surgical approach to metastasectomy has been a point of some controversy. Historically, posterolateral thoracotomy was the approach for choice, and proponents of continuing this practice emphasize the benefit of manual palpation of the lungs for occult metastases not seen on imaging.^23,24 However, with improvements in computed tomography in detecting sub-centimeter lesions, use of a thoracoscopic approach has increased. Additionally, open and VATS metastasectomies have not been shown to have different rates of recurrence or survival in the literature, which is confirmed in our findings.²³

Here we present a novel risk score, using solely preoperative factors, to stratify patients being considered for metastasectomy into low and high-risk groups, with striking survival differences. This easily calculable risk score, after prospective validation, will aid in patient selection and improve informed consent and patient education. Additionally, the 5-year OS of 16.2% in the high-risk group begs the question as to whether these patients would gain additional benefit from multi-modality therapy. Though a significantly greater percentage of patients in the high-risk group received chemotherapy compared to patients in the low-risk group, those receiving chemotherapy were the minority in both groups (low risk: 21%; high risk: 33%, p = 0.037, Table 4). Further studies are needed to determine the optimal treatment regimen to improve outcomes in these high-risk patients.

Limitations to this study include its retrospective design and thus an inherent risk of selection bias. The USSC allowed for analysis of patients from geographically diverse high-volume centers, however these centers are highly specialized tertiary and quaternary referral centers, potentially limiting the generalizability of our results. Additionally, our sample size did not allow for internal validation of our risk score; external and prospective validation should be performed before implementation in clinical practice. We also were unable to include response to chemotherapy and the sequence of multimodal treatment, limiting comment on the impact of multimodal treatment and its inclusion in the risk score. Finally, pathology was not re-reviewed before data analysis and thus our results reflect data entered into the electronic medical record by individual institutions.

5 ∣. CONCLUSION

Complete resection is the standard of care for select patients with sarcoma lung metastases. However, tools are needed to aid in patient selection and risk stratification. A simple risk score including age (<55 vs. ≥55), primary tumor location (truncal/extremity vs. retroperitoneal), primary tumor resection margin (R0 vs. R1), and number of lung lesions (single vs. multiple) that differentiates patients into two groups based on the number of factors present (0–1 vs. 2–4) is associated with significantly poorer survival in patients with increased risk factors. Though prospective and external validation is needed, this risk score will facilitate operative decisionmaking and may identify patients who will benefit from multi-modality therapy.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.