Neoadjuvant radiotherapy followed by surgery compared with surgery alone in the treatment of retroperitoneal sarcoma: a population-based comparison

Abstract

Introduction

Retroperitoneal sarcoma (rps) encompasses a heterogeneous group of malignancies with a high recurrence rate after resection. Neoadjuvant radiotherapy (nrt) is often used in the hope of sterilizing margins and decreasing local recurrence after excision. We set out to compare local recurrence-free survival (lrfs) and overall survival (os) in patients treated with or without nrt before resection.

Methods

Patients diagnosed with rps from February 1990 to October 2014 were identified in the Alberta Cancer Registry. Patients with complete gross resection of rps and no distant disease were included. Patient, tumour, treatment, and outcomes data were abstracted in a primary chart review. Baseline characteristics were compared using the Wilcoxon nonparametric test for continuous data and the Fisher exact test for dichotomous and categorical data. Survival was analyzed using Kaplan–Meier curves with log-rank test. Cox regression was performed to control for age, sex, tumour size, tumour grade, date of diagnosis, multivisceral resection, and intraoperative rupture.

Results

Resection alone was performed in 62 patients, and resection after nrt, in 40. Use of nrt was associated with multivisceral resection and negative microscopic margins. On univariate analysis, nrt was associated with superior median lrfs (89.3 months vs. 28.4 months, p = 0.04) and os (119.4 months vs. 75.9 months, p = 0.04). On multivariate analysis, nrt, younger age, and lower tumour grade predicted improved lrfs and os; sex, tumour size, date of diagnosis, multivisceral resection, and tumour rupture did not.

Conclusions

In this population-based study, nrt was associated with superior lrfs and os on both univariate and multivariate analysis. When feasible, nrt should be considered until a randomized controlled trial is completed.

INTRODUCTION

Retroperitoneal sarcoma (rps) is a rare, but often lethal disease, affecting 2.7 individuals per 1,000,000 population annually¹. The cornerstone of curative-intent treatment is en bloc resection of an intact tumour^2,3. Local recurrence affects more than 60% of patients, who can experience such a recurrence up to 15 years after treatment⁴. The significance of microscopic margins is unclear, although there is some evidence for better outcomes with R0 than with R1 resection³. Chemotherapy is minimally effective in most settings^5,6, but it potentially has roles in downsizing borderline resectable lesions and in treating the small proportion of chemosensitive subtypes of rps such as dedifferentiated liposarcoma, undifferentiated pleomorphic sarcoma, and leiomyosarcoma⁷. The addition of chemotherapy to preoperative radiation therapy (rt) regimens as a radiosensitizer is a topic of new interest, but is as yet poorly understood⁸.

Neoadjuvant radiotherapy (nrt) has been adopted for rps in some centres, because it is thought to sterilize resection margins to increase the likelihood of an R0 resection³. However, the utility of nrt has been difficult to establish clinically⁹. Several retrospective studies have suggested a lower rate of local recurrence with nrt^10–14. Some studies have suggested an effect on overall survival (os)^10,14, but more have not^3,11–13. The Multi-institutional Collaborative RPS Working Group demonstrated a lower rate of local recurrence, but no os benefit with perioperative (pre- or post-resection) rt¹⁵.

The Alberta Cancer Registry is an ideal resource for investigating rps. In the province (population 4,067,175)¹⁶, most patients with rps are treated at 2 tertiary referral centres, the Cross Cancer Institute in Edmonton and the Tom Baker Cancer Centre in Calgary, allowing for a detailed retrospective review over a long period of time.

We therefore set out to investigate the effect of age, tumour grade, tumour size, nrt, and intraoperative tumour rupture on local recurrence-free survival (lrfs) and os.

METHODS

Patients

All patients more than 18 years of age with a pathology diagnosis of sarcoma and site code c48.0 (“retroperitoneum”) within the Alberta Cancer Registry from February 1990 to October 2014 were identified. Patients undergoing complete gross resection of their rps as determined by the operative surgeon were included. Exclusion criteria were treatment for a retroperitoneal sarcoma before the specified date range, metastatic disease at presentation, unavailability of the index operative details, receipt of postoperative adjuvant radiation, and a tumour thought to be of gastrointestinal or gynecologic origin.

Data were then abstracted by chart review. Patient demographics included sex and date of birth. Tumour characteristics included date of diagnosis, tumour maximum dimension, tumour grade and histologic type, and presence of distant metastases at diagnosis.

Treatment

The timing and dose of nrt were recorded. Resections were classified as multivisceral (including involved adjacent organs), tumour-only, or debulking (gross residual disease). Tumour rupture, including intraoperative biopsy and inadvertent capsule breach, was noted. Microscopic margin status was classified as positive, negative, or not reported, based on the original pathology report.

Outcomes

Patients were considered to have experienced a recurrence when they converted from no evidence of disease to evidence of disease based on biopsy or convincing radiographic findings. Recurrence was classified as local (retroperitoneal), distant (extra-abdominal or sarcomatosis), or both.

Statistical Analysis

Baseline characteristics of the groups receiving and not receiving nrt were compared using the Wilcoxon nonparametric test for continuous variables and the Fisher exact test for categorical variables.

Descriptive statistics are used to characterize the sites of recurrence in the nrt and resection-only groups. The proportions of patients experiencing local recurrence (with or without distant recurrence) and distant recurrence only were compared by Fisher exact test.

The lrfs and os in the nrt and resection-only groups were compared using Kaplan–Meier analysis and the log-rank test, with an a priori p value for significance of 0.05.

Patient age and sex, date of diagnosis, multivisceral resection, tumour maximum diameter, tumour grade, rupture, and nrt were included in a Cox regression analysis for lrfs and os.

RESULTS

Screening criteria identified 173 patients in the Registry. Patients were excluded for non-target pathology [n = 2 (1 with gastrointestinal stromal tumour, 1 with uterine leiomyosarcoma)], diagnosis at autopsy (n = 1), unavailability of index resection details (n = 5), postoperative adjuvant rt (n = 2), metastatic disease at presentation (n = 23), and primary not resected or gross residual disease (n = 38). Of the 102 remaining patients, 62 underwent resection alone, and 40 received nrt followed by resection.

The median dose of nrt was 49 Gy in 25 fractions. Surgery was generally performed 4–6 weeks after completion of nrt. Receipt of nrt was associated with multivisceral resection (87.5% vs. 66.1%, p = 0.02), negative margins (72.5% vs. 30.6%, p < 0.001), and later date of diagnosis (80.0% in 2003–2014 vs. 43.5% before 2003, p < 0.001). No statistically significant associations with age, sex, histology (well-differentiated liposarcoma vs. other), tumour grade, tumour size, or rupture were observed (Table I). No patient received neoadjuvant chemotherapy. Two patients received adjuvant chemotherapy, both of them in the resection-only group. Median follow-up was 90 months.

TABLE I.

Clinicopathologic characteristics of 102 patients receiving neoadjuvant radiation therapy (RT) before resection or resection only for retroperitoneal sarcoma

Variable	Neoadjuvant RT		p Value
	No	Yes
Patients (n)	62	40
Mean age (years)	60.0	56.7	0.16a
Sex [n (%)]
Men	24 (38.7)	17 (42.5)	0.84b
Women	38 (61.3)	23 (57.5)
Tumour grade [n (%)]
1	23 (39.7)	14 (40.0)	0.24b
2	6 (10.3)	8 (22.9)
3	29 (50.0)	13 (37.1)
Histology [n (%)]
WDLS	17 (27.4)	13 (32.5)	0.66b
Other	45 (72.6)	27 (67.5)
Maximum diameter
Median (cm)	184.0	136.5
<20 cm [n (%)]	33 (53)	27 (67.5.0)	0.22b
>20 cm [n (%)]	29 (47)	13 (32.5)
Resection type [n (%)]
Multivisceral	41 (66.1)	35 (87.5)	0.02b
Tumour-only	21 (33.7)	5 (12.5)
Tumour rupture or spillage [n (%)]
No	42 (70.0)	32 (80.0)	0.35b
Yes	18 (30.0)	8 (20.0)
Margins [n (%)]
Positive	29 (46.8)	9 (22.5)	<0.001b
Negative	19 (30.6)	29 (72.5)
Not stated	14 (22.6)	2 (5.0)
Diagnosis year [n (%)]
1990–2002	35 (56.5)	8 (20.0)	<0.001b
2003–2014	27 (43.5)	32 (80.0)

^aBy the Wilcoxon nonparametric test.

^bBy the Fisher exact test.

WDLS = well-differentiated liposarcoma.

Recurrence was documented in 17 patients in the nrt group (43%). Of those 17 patients, 6 experienced local recurrence only; 4, distant recurrence only; and 7, both. In the resection-only group, recurrence was documented in 39 patients (63%): 24 local, 4 distant, and 11 both. As a proportion of all patients with recurrence, those with local recurrence, whether associated with distant recurrence or not, therefore constituted 76% of the nrt group and 90% of the resection-only group (p = 0.23).

On univariate analysis, nrt was associated with superior median lrfs (89.3 months vs. 28.4 months, p = 0.04) and os (119.4 months vs. 75.9 months, p = 0.04). On multivariate analysis, age, grade, and receipt of nrt were statistically significant predictors of lrfs and os; sex, tumour size, histologic type, extent of surgery, rupture, and date of diagnosis were not (Tables II and III).

TABLE II.

Multivariate associations between clinico-demographic factors and local recurrence-free survival

Factor	HR	95% CI	p Value
nRT	0.43	0.24 to 0.79	0.01
Age	1.03	1.01 to 1.05	0.01
Sex
Women		1
Men	1.22	0.69 to 2.12	0.49
Tumour size
<20 cm		1
≥20 cm	0.58	0.31 to 1.07	0.08
Grade
1		1
2	5.46	1.97 to 15.14	0.001
3	4.44	1.66 to 11.89	0.003
Histology
Other		1
WDLS	1.24	0.44 to 3.50	0.68
Rupture
No		1
Yes	1.39	0.75 to 2.58	0.30
Resection type
Multivisceral		1
Tumour-only	0.82	0.42 to 1.60	0.57
Diagnosis year
1990–2002		1
2003–2014	0.77	0.42 to 1.39	0.38

HR = hazard ratio; CI = confidence interval; nRT = neoadjuvant radiation therapy; WDLS = well-differentiated liposarcoma.

TABLE III.

Multivariate associations between clinico-demographic factors and overall survival

Factor	HR	95% CI	p Value
nRT	0.42	0.19 to 0.90	0.03
Age	1.03	1.00 to 1.05	0.02
Sex
Women		1
Men	1.00	0.52 to 1.95	0.99
Size
<20 cm		1
≥20 cm	0.96	0.48 to 1.93	0.91
Grade
1		1
2	4.00	1.21 to 13.17	0.02
3	3.30	1.09 to 9.96	0.03
Histology
Other		1
WDLS	0.81	0.24 to 2.70	0.73
Rupture
No		1
Yes	1.02	0.49 to 2.10	0.96
Resection type
Multivisceral		1
Tumour-only	1.01	0.47 to 2.16	0.99
Diagnosis year
1990–2002		1
2003–2014	0.88	0.42 to 1.83	0.73

HR = hazard ratio; CI=confidence interval; nRT = neoadjuvant radiation therapy; WDLS = well-differentiated liposarcoma.

Kaplan–Meier survival analysis with log-rank test demonstrated a statistically significant improvement in both lrfs and os associated with nrt (p = 0.04 in both cases, Figure 1).

FIGURE 1.

Kaplan–Meier plots of (A) local recurrence-free survival and (B) overall survival with or without the use of neoadjuvant radiation therapy (nRT).

DISCUSSION

In our cohort, nrt was associated with superior lrfs and os in both univariate and multivariate analysis. Overall, on multivariate analysis, age, tumour grade, and nrt were found to be statistically significant predictors of lrfs and os. A trend toward less local recurrence as a fraction of all recurrences was evident in the nrt group, but was not statistically significant.

Comparing our results with those reported in the literature, the data appear to be mixed. That inconsistency was well demonstrated by Nussbaum et al.¹⁷ in their recent publication of a U.S. National Cancer Database study of rt for rps. Those authors suggested that the lack of clearly demonstrable treatment effect in the literature to date is partly attributable to the inherent biases of small sample size.

Current ly, other than the ongoing randomized phase iii strass trial (European Organisation for Research and Treatment of Cancer 62092, NCT01344018 at https://ClinicalTrials.gov/), no published randomized study has compared nrt with resection alone for rps.

In two prospective trials of nrt for high-risk lesions involving a total of 72 patients, median survival was not reached after a follow-up of more than 60 months. The 5-year lrfs was 60% compared with 20%–50% in historical controls¹⁸.

Several recent studies using the U.S. Surveillance, Epidemiology, and End Results Program and the U.S. National Comprehensive Cancer Network databases have investigated the effect of rt on os in rps, with conflicting results (Table IV). Four studies demonstrated no effect on os^13,19,21,22, and four demonstrated a benefit^17,20,23,24. The relevant datasets were limited in that they lacked specific recurrence data for an evaluation of lrfs or disease-free survival. Only two of the studies^17,23 compared resection alone with nrt; the rest included other radiation modalities in one of the arms. In studies comparing any rt with no rt, most patients receiving rt did so postoperatively.

TABLE IV.

Studies in administrative databases of neoadjuvant radiation therapy (nRT) in retroperitoneal sarcoma

Reference	Database	Comparison	OSa	p Value
Nathan et al., 200919	SEER	RTb vs. none	HR: 0.95	0.6
Zhou et al., 201020	SEER	RTc vs. none	HR: 0.78	0.01
Tseng et al., 201121	SEER	RTd vs. none	82 vs. 87 months	0.11
Choi et al., 201222	SEER	RTe vs. none	Kaplan–Meier curve only	0.10
Nussbaum et al., 201513	NCDB	nRT vs. no nRTf	5-Year: 53.2% vs. 54.2%	0.695
Ecker et al., 201623	NCDB	nRT vs. no RT	129.2 vs. 84.3 months	0.046
Nussbaum et al., 201617	NCDB	nRT vs. no RT	HR: 0.70	<0.0001
		Postoperative RT vs. no RT	HR: 0.78	<0.0001
Bates et al., 201824	SEER	Postoperative RT vs. no RT	HR: 0.8	0.029

^aTreatment compared with control.

^bPostoperative in 83.1%.

^cPreoperative and postoperative proportions not specified.

^dPostoperative in 80.4%.

^ePostoperative in 80.2% of patients after propensity score matching.

^fOf the no-nRT group, 32.9% received postoperative RT.

OS= overall survival; SEER = U.S. Surveillance, Epidemiology, and End Results program; HR= hazard ratio; NCDB = U.S. National Cancer Database;

Some of the more robust literature is derived from multi-institutional cooperative groups. The Multi-institutional Collaborative RPS Working Group, an effort of 9 high-volume centres, demonstrated improved lrfs with perioperative rt [hazard ratio (hr): 0.58; p = 0.001], but no effect on os (hr: 0.98; p = 0.864)¹⁵. In a report of the Trans-Atlantic Retroperitoneal Sarcoma Working Group, restricted to liposarcoma (stratified into 3 cohorts according to grade), perioperative rt was, on univariate analysis, associated with improved local control in all cohorts; that association disappeared after control for confounders was introduced. No effect of rt on distant metastasis or os was observed²⁵.

In retrospective studies, results have been conflicting. Some studies have found no benefit from perioperative rt^26,27; several have demonstrated improved lrfs^11,28–31; and several have demonstrated benefits in both lrfs and os^10,14,32,33 (Table V).

TABLE V.

Summary of retrospective studies of perioperative radiation therapy in retroperitoneal sarcoma

Reference	Patients		OSe	p Value	RFSe	p Value
	(N)	(n w/nRT)
Stoeckle et al., 200130	165	3	—	—	RR: 0.31b	<0.001
Hassan et al., 200427	97	16	—	—	41% vs. 45%c	0.76
Bonvalot et al., 200926	249	36	—	—	HR: 0.52b	0.116
Gronchi et al., 200932	288	—d	HR: 0.55	0.008	HR: 0.65b	0.057
Sampath et al., 201029	261	6	—	—	HR: 0.42b	<0.05
Gronchi et al., 201233	331	—d	HR: 0.64	0.05	HR: 0.57b	0.02
Trovik et al., 201414	97	5	71% vs. 52%e	0.019	77% vs. 39%e	<0.001
Kelly et al., 201511	204	30	—	—	HR: 0.26b	0.03
Lane et al., 201510	74	9	HR: 0.30	0.02	HR: 0.34c	<0.01
Snow et al., 201831,f	88	57	HR: 1.0	0.93	HR: 0.33b	0.014

^aIntervention compared with control.

^bFor local recurrence.

^cFor local or metastatic recurrence.

^dNot stratified by preoperative compared with postoperative delivery.

^e5-Year results.

^fOnly study specific for nRT.

w/nRT = with neoadjuvant radiation therapy; OS = overall survival; RFS = recurrence-free survival; RR = relative risk; HR = hazard ratio.

There are several possible reasons for the conflicting results. Sample size is a concern, given that all the retrospective studies include a relatively small number of patients. Treatment regimens are not always uniform within study groups; many studies include patients with either preoperative or postoperative rt in a single treatment arm. Finally, there are arguments (discussed shortly) for the superiority of preoperative compared with postoperative rt, and the preoperative approach has been relatively poorly represented in retrospective studies. With the exception of Kelly et al.¹¹ and Snow et al.³¹, studies delivered rt to patients postoperatively.

There are good theoretical reasons to propose the superiority of nrt over postoperative rt. In the treatment of rps, nrt is justified in part by extrapolation from the success of rt in sarcoma of the extremities. When a limb-sparing approach is selected, local control is significantly improved with adjunctive rt^34–36. However, microscopically negative margins are more difficult to obtain in rps than in sarcoma of the extremities, because of the typically greater size and more prohibitive anatomic barriers to wide resection³⁷. Indeed, some authors argue that nearly all rps resections have at least a focally positive microscopic margin, which can be missed on pathologic section because of the difficulty of assessing a large lesion^38,39. Radiation therapy might therefore be expected to have a greater oncologic benefit in the preoperative setting, by sterilizing the margins of the eventual resection and reducing the opportunity for viable tumour to be left in the operative bed. Those effects ought, in turn, to improve os, given that repeat resection is usually difficult or impossible, and options for systemic therapy are limited. Further, delivering rt with tumour in situ allows for more precise targeting of the lesion and limits the exposure of adjacent organs, possibly preventing treatment interruption for toxicity. Those considerations appear to be supported by a comparison of preoperative compared with postoperative rt in the U.S. National Cancer Database, which demonstrated improved os (hr: 0.72; p < 0.01) and cancer-specific mortality (hr: 0.64; p < 0.01) in the preoperative group⁴⁰. Finally, postoperative rt is, in many cases, simply not feasible because of the large field sizes and higher doses required when rt is delivered postoperatively—especially with the small bowel or other dose-limiting structures potentially within the field.

Our study demonstrated, on univariate and multivariate analysis alike, improved lrfs and os with nrt, which accords with some published retrospective series and contrasts with others. It seems plausible that, in some prior negative studies, the benefits of rt were blunted by the inclusion of patients who received rt postoperatively. Without directly addressing that question, our study adds to the evidence that nrt might be superior to resection alone.

We believe that our study design provides certain advantages over some of the studies discussed earlier. Although our study was population-based, the concentration of cases in 2 centres allowed for the collection of data about local recurrence, which has not been possible in administrative databases. At the same time, the study population was larger than that in many single-centre series and likely more representative of the population at large. To our knowledge, ours is only the second study (Snow et al.³¹ being the first) with institution-level data in which the intervention arm was limited to nrt rather than to perioperative rt. It is the first such study to apply multivariate analysis. Snow et al. did not include a multivariate analysis, and the authors note that nrt was strongly associated with treatment at a high-volume centre and was therefore likely confounded by greater extent of surgery, use of core biopsy, and greater experience of the treating team.

The retrospective nature of our study imposes certain limitations. Most importantly, the nrt and resection-only groups differed with respect to extent of surgery, margin positivity, and date of diagnosis. It is therefore possible that some of the apparent treatment effect is a result of changing practice patterns such as multidisciplinary referral and oncologically superior biopsy and resection techniques. The rationale for nrt compared with resection alone, discussion at a multidisciplinary tumour board, and resection at a high-volume centre were not available as data points. That lack is somewhat mitigated by our inclusion of multivisceral resection in the multivariate analysis, because that technique would be expected to be more prevalent at high-volume centres. Treatment at a high-volume centre would also be expected to have become more universal in recent years, and our multivariate therefore controlled for date of diagnosis.

The number of tumours with unreported margins (15 of 102) prevented inclusion of margin status in the multivariate analysis. That lack is somewhat mitigated by the inclusion of tumour rupture, which would be associated with positive margins.

CONCLUSIONS

In a population-based study, nrt was associated with improved lrfs and os on both univariate and multivariate analysis. Where tumour and patient factors permit, nrt should be considered for rps until a randomized controlled trial is completed.