Utilizing the Radiosensitivity Index (RSI) to Predict Pelvic Failure in Endometrial Cancer Treated with Adjuvant Radiotherapy

Abstract

Background:

Variability exists in the adjuvant treatment for endometrial cancer (EC) based on surgical pathology and institutional preference. The radiosensitivity index (RSI) is a previously validated multigene expression index that estimates tumor radiosensitivity. We evaluate RSI as a genomic predictor for pelvic failure (PF) in EC patients treated with adjuvant radiotherapy.

Patients and Methods:

Using our institutional tissue biorepository, we identified EC patients treated between January 1999 and April 2011 of primarily endometrioid histology (n=176; 86%) with receipt of various adjuvant therapies. The RSI 10-gene signature was calculated for each sample using the previously published algorithm. Radiophenotype was determined using the previously identified cutpoint where RSI ≥0.375 denotes radioresistance (RR) and RSI <0.375 describes radiosensitivity (RS).

Results:

A total of 204 patients were identified of which 83 (41%) were treated with adjuvant radiotherapy. Median follow-up was 38.5 months. All patients underwent hysterectomy with bilateral salpingo-oophorectomy with the majority undergoing lymph node dissection (n=181; 88%). In patients treated with radiation, RR tumors were more likely to experience PF (3-year pelvic control 84% vs 100%; P=0.02) with worse PF-free survival (PFFS) (3-year PFFS 65% vs. 89%; P=0.04). Furthermore, in the patients who did not receive RT, there was no difference in PF (P=0.87) or PFFS (P=0.57) between the RR/RS tumors. On multivariable analysis (MVA), factors that continued to predict for PF included the RR phenotype (HR=12.2, P=0.003), lymph node involvement (HR=4.4, P=0.02), and serosal and/or adnexal involvement (HR=5.3, P=0.01).

Conclusions:

On MVA, RSI was found to be a significant predictor of PF in patients treated with adjuvant radiotherapy. We propose utilizing RSI to predict which patients are at higher risk for failing in the pelvis and may be candidates for treatment escalation in the adjuvant setting.

INTRODUCTION:

Endometrial cancer (EC) is the most common gynecological cancer.(1) It generally carries a favorable prognosis, as it is often detected at an early stage and responds well to treatment.(2) The preliminary intervention for EC, regardless of histological type, is a total hysterectomy with bilateral salpingo-oophorectomy.(3) Currently, adjuvant treatment decisions are selected based on the clinicohistological characteristics of the primary tumor.(4) The standard adjuvant treatment for patients with Stage 1A Grade 3 EC and above is radiation therapy (RT).(5,6) Multiple randomized controlled trials have demonstrated that adjuvant RT, including external beam (EBRT) and vaginal brachytherapy (VBT), reduce the absolute risk of locoregional recurrence.(6-9) These studies evaluated various adjuvant treatment paradigms which are reflected as options recommended by National Comprehensive Cancer Network (NCCN) guidelines(10). As a result of these heterogenous trials and recommendations, practice patterns vary amongst different institutions and physicians.(11)

Considering the local control (LC) benefit, the rate of grade 1 and 2 acute toxicity from EBRT has been reported to be as high as 53.8%.(5) Additionally, there is significant long-term health-related quality of life evidence demonstrating persistence of bowel symptoms from EBRT for these patients.(12) Receipt of intensity modulated radiation therapy (IMRT) may reduce these risks. RTOG 1203 reported 52% of women who received standard RT and 34% who received IMRT reported frequent or almost constant diarrhea (P = .01)(13). Using individual patient and tumor characteristics may optimize the selection of patients for adjuvant RT.(14) There are currently no routinely used genomic signatures to help select EC patients most likely to benefit from RT administration or RT dose selection.

Our group had previously developed the radiosensitivity index (RSI) using a multigene expression model that is directly proportional to tumor radioresistance (high RSI = radioresistance).(15,16) The model was initially developed in a panel of 48 human cancer cell lines and has since been validated in multiple independent clinical cohorts including breast, colon, esophageal, head and neck, glioblastoma, liver, pancreas, prostate, and rectal malignancies.(16-23) Given the variation in practice patterns for adjuvant RT selection in EC, we sought to use the RSI to determine which patients are most prone to failure and candidates for adjuvant therapy.

PATIENTS AND METHODS:

Patient data

Between January 1999 and April 2011, a total of 204 consecutively treated EC patients were identified from the IRB-approved Total Cancer Care (TCC) protocol at Moffitt Cancer Center..(24) Patient follow-up was conducted on a per-provider dependent basis and were followed until January 2019.

RNA Preparation and Gene Expression Profiling

We assayed endometrial tumor tissue samples that were snap frozen in the TCC protocol on Affymetrix Hu-RSTA-2a520709 (Affymetrix; Santa Clara, CA, USA), which contains approximately 60, 000 probesets representing 25,000 genes. (25)These were then normalized using the iterative rank order normalization method used across samples in the TCC database.(22) From the resulting normalized data, a RNA-quality related batch-effect was identified and removed by training a partial least squares model(26) to the RNA integrity number, and then subtracting the first partial least squares component.

Radiosensitivity signature

RSI was calculated by running the previously tested(15,16,21,23) 10-hub-gene assay on tissue samples and ranking the genes according to expression (with the most highly expressed gene being 10 and the lowest being 1). The RSI value was then determined using the previously published linear regression algorithm:

$$\begin{gathered} \text{RSI}=-.0098009{}^{\ast }\text{AR}+0.0128283{}^{\ast }\text{cJun}+0.0254552{}^{\ast }\text{STAT1}-0.0017589{}^{\ast }\text{PKC}- \\ 0.0038171{}^{\ast }\text{Re1A}+0.1070213{}^{\ast }\text{cABL}-0.0002509{}^{\ast }\text{SUMO1}-0.0092431{}^{\ast }\text{PAK}2- \\ 0.0204469{}^{\ast }\text{HDAC1}-0.0441683{}^{\ast }\text{IRF1}. \end{gathered}$$

As previously published, patients were dichotomized into radioresistant and radiosensitive groups using the 50^th percentile RSI cutpoint as established across all available samples in the TCC.(19,27) The dichotomization was set at 0.375, with ≥0.375 = radioresistant (RR) and <0.375 = radiosensitive (RS).

Statistical analyses

Descriptive statistics were used to summarize the cohort including median and range values for continuous variables and percentages for categorical variables. Baseline patient characteristics were compared using the Wilcoxon and Pearson chi-squared tests. Pelvic failure (PF) and survival data were calculated from the date of pathologic diagnosis to last follow-up or death from any cause. Pelvic failure, PF free survival (PFFS), disease free survival (DFS) and overall survival (OS) were estimated by using the Kaplan-Meier method and the log-rank test was used to complete intergroup comparisons. Multivariable analyses (MVA) were performed with Cox regression analyses for any variables with P ≤ 0.10 on univariable analyses (UVA). A P ≤ 0.05 was considered statistically significant. All statistical analyses were performed using JMP 13 (SAS Institute Inc, Cary, NC, USA).

RESULTS:

Patient and Tumor Characteristics

Table 1 shows all patient characteristics and differences between RS and RR tumors. A total of 204 patients were identified for analysis. The median age of patients was 62 (range: 29 to 90 years). All patients had undergone primary tumor treatment with hysterectomy and bilateral salpingo-oophorectomy (TAH-BSO) and 181 (88%) patients underwent a lymph node dissection (LND). The median RSI was 0.42 (range: 0.11 to 0.70). A total of 83 (41%) patients were treated with adjuvant radiotherapy with vaginal brachytherapy (N=19; 23%), pelvic radiotherapy (N =26; 31%), or both (N =38; 46%). Sequential chemotherapy was received by 55 (27%) of patients.

Table 1:

Patient and Tumor Characteristics

Variable	N (%)	RS (RSI <0.375)	RR (RSI ≥ 0.375)	P-Value
No of Patients	204	68 (33%)	136 (67%)
Follow-Up From Date of Surgery (Months)
Median	38.5	38	39.5
Range	0.2-216	0.3-135.6	0.2-216
Age
Median	62	62	61	0.99
Range	29-90	31-82	29-90
Grade
1	54 (26%)	15 (22%)	39 (29%)	0.03
2	98 (48%)	28 (41%)	70 (51%)
3	52 (25%)	25 (37%)	27 (20%)
Histology
Endometrioid Adenocarcinoma	176 (86%)	56 (82%)	120 (82%)	0.25
Carcinosarcoma	5 (2%)	2 (3%)	3 (2%)
Clear Cell	9 (4%)	6 (9%)	3 (2%)
Mucinous	2 (1%)	1 (1%)	1 (1%)
Serous	12 (6%)	3 (4%)	9 (7%)
FIGO Staging
1A	94 (46%)	31 (46%)	63 (46%)	0.54
1B	28 (13%)	12 (18%)	16 (12%)
II	25 (12%)	6 (9%)	19 (14%)
IIIA	13 (6%)	3 (4%)	10 (7%)
IIIB	9 (4%)	5 (7%)	4 (3%)
IIIC1	27 (13%)	9 (13%)	18 (13%)
IIIC2	8 (4%)	2 (3%)	6 (4%)
MMI
<50%	130 (64%)	44 (65%)	86 (64%)	0.88
≥50%	75 (36%)	24 (35%)	49 (36%)
Cervical Stromal Invasion
Y	62 (30%)	19 (28%)	43 ( 32%)	0.59
N	142 (70%)	49 (72%)	93 (68%)
Invasion of Serosa and/or Adnexa
Y	27 (13%)	7 (10%)	20 (15%)	0.38
N	177 (87%)	61 (90%)	116 (85%)
Vaginal or Parametrial Involvement
Y	17 (8%)	7 (10%)	10 (7%)	0.47
N	187 (92%)	61 (90%)	126 (93%)
LVSI
Y	64 (31%)	24 (35%)	40 (29%)	0.39
N	140 (69%)	44 (65%)	96 (71%)
Lymph Node Positive
Y	35(17%)	11 (16%)	24 (18%)	0.79
N	169 (83%)	57 (84%)	112 (82%)
Lymph Node Dissection
Y	180 (88%)	61 (90%)	119 (88%)	0.65
N	24 (12%)	7 (10%)	17 (13%)
Adjuvant Chemotherapy
Y	55 (27%)	23 (34%)	32 (24%)	0.12
N	149 (73%)	45 (66%)	104 (76%)
Adjuvant Radiation Therapy
EBRT	19 (9%)	7 (10%)	12 (9%)	0.62
VBT	26 (13%)	10 (15%)	16 (12%)
Both EBRT and VBT	38 (19%)	15 (22%)	23 (17%)
None	121 (59%)	36 (53%)	85 (63%)

^*RS = Radiosensitive

^**RR= Radioresistant

^***MMI= Myometrial Invasion

^****LVSI= Lymphovascular Space Invasion

^*****EBRT= External Beam Radiation Therapy

^*****VBT= Vaginal Brachytherapy

Characteristics of RS and RR Patients

There were no significant differences in RS and RR tumors in age (P=0.99), serosa and/or adnexa involvement (P=0.38), vaginal and/or parametrial involvement (P=0.47), cervical stromal invasion (P=0.59), receipt of adjuvant chemotherapy (P=0.12), and node positivity (P=0.79). RS tumors were more likely to be grade 3 (P=0.03). Characteristics of RS and RR tumors treated with and without RT are detailed in Supplemental Tables 1 and 2, respectively.

Pelvic Failure

Median follow-up for the whole cohort was 38.5 months (range: 0.2 to 216). In patients treated with radiation, those with RR tumors were more likely to experience PF (3 year pelvic control [PC] 84% vs 100%; P=0.02), Figure 1 with worse PFFS (3 year PFFS 65% vs. 89%; P=0.04). However, since RSI is a radiation specific signature it did not predict PF (P=0.87), Figure 2, or PFFS (P=0.57) in patients not treated with radiation. The characteristics of the 19 patients experiencing pelvic failure are detailed in Supplemental Table 3.

Figure 1:

Kaplan-Meier Pelvic Control between RS and RR patients treated with RT

Figure 2:

Kaplan-Meier Pelvic Control between RS and RR patients treated without RT

PF UVA and MVA results for patients treated with radiation are detailed in Table 2. Factors found to predict PF on UVA included grade 3/1-2 (HR 3.8; 95% CI 1.2 to 14.8, P=0.03), serosa and/or adnexal involvement (5.9; 95% CI 1.6 to 20.2, P=0.008), lymph node involvement (4.9; 95% CI 1.5 to 18.8, P=0.009), and RR/RS (7.7; 95% CI 1.5 to 140.9, P=0.01). On MVA, factors that continued to predict for PF included RR/RS (12.2; 95% CI 2.1 to 232.6, P=0.003), lymph node involvement (4.4; 95% CI 1.2 to 18.1, P=0.02), and serosa and/or adnexal involvement (5.3; 95% CI 1.4 to 20.1, P=0.01).

Table 2:

Univariable and Multivariable Pelvic Control Analysis

	Univariable Analysis			Multivariable Analysis
Variable	HR	95% CI	P-Value	HR	95% CI	P-Value
Grade
3/1–2	3.8	1.2 to 14.8	0.03	3.1	0.86 to 12.2	0.08
MMI
≥ 50%/<50%	1.2	0.36 to 4.6	0.76
LVSI
Y/N	3.1	0.9 to 14.3	0.07
Cervical Stromal Invasion
Y/N	1.2	0.36 to 4.2	0.75
Serosa and Adnexa Invasion
Y/N	5.9	1.6 to 20.2	0.008	5.3	1.4 to 20.1	0.01
Vaginal/Parametrial Invasion
Y/N	3.5	0.91 to 11.5	0.07
Lymph Node Positive
Y/N	4.9	1.5 to 18.8	0.009	4.4	1.2 to 18.1	0.02
RSI ≥0.375
Y/N	7.7	1.5 to 140.9	0.01	12.2	2.1 to 232.6	0.003

^*MMI= Myometrial Invasion

^**LVSI= Lymphovascular Space Invasion

^***EBRT= External Beam Radiation Therapy

^****VBT= Vaginal Brachytherapy

Distant Control, Disease Free Survival and Overall Survival

A total of 20 patients (24%) treated with RT experienced distant failure (24%). No significant difference was noted in distant control in patients with RS and RR tumors treated with RT (3 year 89% vs. 74%; p=0.27). A total of 36 patients (43%) treated with RT died at the time of study analysis. In patients treated with RT, there was a trend towards patients with RS tumors having improved survival (3 year OS 93% vs. 73%; P=0.06) and DFS (3 year 85% vs. 65%; p=0.09).

DISCUSSION:

In this study we present the first genomic signature to help predict PF in EC patients treated with adjuvant RT. We found RR tumors were more prone to pelvic recurrence compared to RS tumors in patients treated with adjuvant RT. This remained significant on MVA. Since RSI is a RT specific signature, we found no difference in the rates of PF between RS and RR tumors not treated with RT. For patients with EC, RSI may be used to develop a personalized treatment regimen and could guide clinicians to both toxicity reduction and improved locoregional control.

There is growing evidence to demonstrate the value of RSI in guiding radiotherapy decisions in a number of malignancies(16-21,23,25). Torres-Roca et al. have previously used RSI to identify RR triple negative breast cancer tumors were at higher risk for local recurrence (LR) compared to RS triple negative tumors (HR 0.37, P=0.02).(20) Strom et al. evaluated the role of RSI in stratifying pancreatic cancer patients with high risk features (positive post-operative margin, positive lymph nodes, or post-operative CA19-9 > 90) who received adjuvant RT. They demonstrated that patients with RR tumors had significantly worse outcomes compared to RS tumors with a median OS of 13.2 compared to 31.2 months, respectively (P=0.04).(17) The principle of these findings and those of our study were also demonstrated in studies regarding lung primaries and metastases, colon primaries and metastases, and glioblastoma.(18,19,25,27)

Adjuvant treatment escalation may improve outcomes for certain high risk subpopulations of endometrial cancer. As demonstrated by the updated results of Post Operative Radiation Therapy in Endometrial Carcinoma (PORTEC)-3 prospective study, patients may benefit from adjuvant chemoradiotherapy with concurrent cisplatin and adjuvant carboplatin and paclitaxel vs. EBRT alone.(28) The 5-year OS was 81·4% with chemoradiotherapy versus 76·1% with EBRT alone (P=0.03) and 5-year failure-free survival was 76·5% versus 69·1% (P=0.02). Chemoradiotherapy was well tolerated and at 5 years, reported grade 3 adverse events did not differ significantly between the two groups.

Our study demonstrates significant differences in 3-year PC rates between patients with RR and RS tumors. The pelvic failures in this analysis all received adjuvant EBRT prescribed to a dose of 45 – 50.4 Gy. Therefore, these recorded pelvic recurrences represent in-field treatment failures. Dose escalation of the entire pelvic field has previously demonstrated unacceptable rates of small and large bowel toxicity and is not recommended for the treatment of any genitourinary or gynecologic malignancy.(29) Therefore, patients with RR tumors would be ideal candidates for treatment escalation with the addition of radiosensitizing agents. Given the updated results of PORTEC-3, the results of our study are particularly intriguing.

There are efforts to use molecular information to inform higher risk endometrial cancer subpopulations. (30-32) The pilot PORTEC-4a multicenter ongoing phase 3 study randomizes high-intermediate risk patients to either VBT or adjuvant radiotherapy treatment based on the molecular integrated risk profile.(33) We await the results of this ongoing study and believe RSI may be another means to stratify adjuvant radiotherapy decisions.

The role of adjuvant RT in EC is well established(7,9,34,35) but it is important to recognize that treatment related toxicities exist and have a significant impact on patient reported QOL outcomes.(36-39) Acute toxicities are primarily driven by gastrointestinal events and overall rates of any grade 2 toxicity has been reported to be as high as 31% using modern treatment approaches.(40) De Boer et al. report on the QOL data from the long-term survivors of the PORTEC-2 trial. At 7 years they demonstrate that up to 42% of patients continue to experience mild gastrointestinal or urinary symptoms. Additionally, 39.3% and 23.3% of patients experience moderate/severe urinary and bowel urgency, respectively.(12) Our study identified a subset of patients with RS tumors who demonstrate a 100% 3-year PC rate. In an attempt to mitigate long-term toxicities for these patients, dose de-escalation could prove beneficial without a loss in PC rate.

Limitations of this study include the retrospective collection of clinical data from the tissue cohort. In addition, patients were treated over a 12 year span of time and while the majority of patients were treated at our institution this led to a variety of treatment techniques and adjuvant therapy strategies that were employed. Since not all patients received adjuvant RT at our institution, we were unable to review the complete treatment planning record for all patients. In addition, RSI was validated in a relatively small sample of RT treated patients.

In conclusion, using a previously validated genomically-derived radiation specific signature, we have identified a subset of EC patients with RR tumors that are at higher risk for PF. We propose RSI may be a means to identify patients at higher risk for PF and candidates for adjuvant treatment intensification.