KEYWORDS: health disparities, National Cancer Data Base (NCDB), neoadjuvant chemoradiation, neoadjuvant chemotherapy, rectal cancer.

Transcription

1 Can We Eliminate Neoadjuvant Chemoradiotherapy in Favor of Neoadjuvant Multiagent Chemotherapy for Select Stage II/III Rectal Adenocarcinomas: Analysis of the National Cancer Data Base Richard J. Cassidy, MD 1,2 ; Yuan Liu, PhD 3 ; Kirtesh Patel, MD 1,2 ; Jim Zhong, MD 1,2 ; Conor E. Steuer, MD 2,4 ; David A. Kooby, MD 2,5 ; Maria C. Russell, MD 2,5 ; Theresa W. Gillespie, PhD 2,5 ; and Jerome C. Landry, MBA, MD 1,2 BACKGROUND: Stage II and III rectal cancers have been effectively treated with neoadjuvant chemoradiotherapy (NCRT) followed by definitive resection. Advancements in surgical technique and systemic therapy have prompted investigation of neoadjuvant multiagent chemotherapy (NMAC) regimens with the elimination of radiation (RT). The objective of the current study was to investigate factors that predict for the use of NCRT versus NMAC and compare outcomes using the National Cancer Data Base (NCDB) for select stage II and III rectal cancers. METHODS: In the NCDB, 21,707 patients from 2004 through 2012 with clinical T2N1 (ct2n1), ct3n0, or ct3n1 rectal cancers were identified who had received NCRT or NMAC followed by low anterior resection. Kaplan-Meier analyses, log-rank tests, and Cox-proportional hazards regression analyses were conducted along with propensity score matching analysis to reduce treatment selection bias. RESULTS: The 5-year actuarial overall survival (OS) rate was 75% for patients who received NCRT versus 67.2% for those who received NMAC (P <.01). On MVA, those who received NCRT had improved OS (hazard ratio, P <.01), and this effect was confirmed on propensity score matching analysis (hazard ratio, 0.72; P 5.01). In the same model, the following variables improved OS: age < 65 years, having private insurance, treatment at an academic center, living in an affluent zip code, a low comorbidity score, receipt of adjuvant chemotherapy, and a shorter interval before surgery (all P <.05). African Americans, men, patients with highgrade tumors, those with ct3n1 tumors, and those who underwent incomplete (R1) resection had worse OS (all P <.05). CONCLU- SIONS: In this series, the elimination of neoadjuvant RT for select patients with stage II and III rectal adenocarcinoma was associated with worse OS and should not be recommended outside of a clinical trial. Cancer 2017;123: VC 2016 American Cancer Society. KEYWORDS: health disparities, National Cancer Data Base (NCDB), neoadjuvant chemoradiation, neoadjuvant chemotherapy, rectal cancer. INTRODUCTION An estimated 40,000 new cases of rectal cancer are diagnosed in the United States annually. 1 Since 1990, the standard of care for stage II (T3 and T4 tumors) and stage III (N1 or N2 lymph node status) rectal cancer (with stage defined according to the American Joint Committee on Cancer staging system 2 ) has been trimodality therapy with radiation (RT), chemotherapy, and surgery. 3 Historical studies demonstrated that pelvic RT reduced local rectal cancer recurrence and increased the chance for sphincter preservation with surgery. 4-6 The sequencing of chemoradiotherapy (CRT) for stage II and III tumors was defined based on a landmark German randomized trial indicating that neoadjuvant CRT reduced local recurrence, improved sphincter preservation, and had an improved side-effect profile. 7,8 These results were confirmed in the United States by the National Surgical and Adjuvant Breast and Bowel Project trial NSABP R Consequently, the standard of care for locally advanced rectal cancers is now preoperative CRT with 5-fluorouracil (5-FU) followed by surgical resection and additional adjuvant chemotherapy. Corresponding author: Richard J. Cassidy, MD, Radiation Oncology, Winship Cancer Institute of Emory University, Emory University School of Medicine, 1365 Clifton Road NE, Atlanta, GA 30322; Fax: (888) ; richardjcassidy@emory.edu 1 Department of Radiation Oncology, Rollins School of Public Health, Emory University, Atlanta, Georgia; 2 Winship Cancer Institute, Emory University, Atlanta, Georgia; 3 Department of Biostatistics and Bioinformatics, Rollins School of Public Health, Emory University, Atlanta, Georgia; 4 Department of Medical Oncology, Emory University, Atlanta, Georgia; 5 Department of Surgery, Emory University, Atlanta, Georgia. The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Data used in the study are derived from a de-identified National Cancer Data Base file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology used or the conclusions drawn from these data by the investigators. Additional supporting information may be found in the online version of this article. DOI: /cncr.30410, Received: July 21, 2016; Revised: September 21, 2016; Accepted: September 23, 2016, Published online October 25, 2016 in Wiley Online Library (wileyonlinelibrary.com) Cancer March 1,

2 Recently, with the incorporation of total mesorectal excision (TME) surgical principles 10,11 and improved systemic therapy, 12,13 the benefit of pelvic radiation is being challenged. CRT is associated with short-term toxicity in 50% of patients as well as long-term effects, including fibrosis, fecal incontinence, bladder and sexual dysfunction, and myelosuppression From a chemotherapy perspective, less toxicity has been reported with preoperative versus postoperative chemotherapy. 18 This rationale prompted the investigational use of neoadjuvant multiagent chemotherapy (NMAC) followed by excision, eliminating pelvic radiation, in a pilot trial that produced promising results. 19 The accruing PROSPECT trial is implementing this paradigm nationally for patients who have clinically staged T3N0 (ct3n0), ct3n1, and ct2n1 rectal cancer. 20 Given the small number of patients studied with this evolving regimen, we used the National Cancer Data Base (NCDB) to identify patients who appeared to meet eligibility criteria for the PROS- PECT trial to determine the impact of NMAC versus neoadjuvant chemoradiotherapy (NCRT) on overall survival (OS) and to determine variables that predicted for patients receiving NMAC versus NCRT. MATERIALS AND METHODS NCDB Criteria The NCDB is a large, prospectively acquired database of patients from Commission on Cancer-accredited institutions across the United States that captures approximately 70% of all newly diagnosed malignancies in the United States. The data set includes detailed information on patient, disease, and treatment characteristics as well as survival outcomes. The NCDB includes detailed RT information regarding treatment site, radiation source, and dose as well as detailed surgical and pathologic information, including type of surgery, margin status, and number of lymph nodes removed. The rectal NCDB was queried for patients aged >18 years who diagnosed during 2004 through Patients with ct2n1, ct3n0, and ct3n1 rectal adenocarcinomas who received NCRT or NMAC followed by a low anterior resection (LAR) were included. Patients who underwent abdominoperineal resection, did not receive neoadjuvant therapy, received nonmegavoltage photonbeam RT, or received single-agent chemotherapy alone and those who had medical reasons to not receive RT or had a lack of survival information were excluded (Fig. 1). The following patient characteristics were examined: age (<65 or 65 years), race (white, African American, other), sex, treatment facility type (community cancer center, comprehensive community cancer center, or academic/research program), treatment facility geographic region, insurance coverage (not insured, private insurance, government insurance [including Medicaid and US Veterans Affairs Administration (VA) coverage], and Medicare insurance), patient county of residence (urban, rural, or metropolitan), distance from patient residence zip code to treatment facility (<10 miles, miles, or > 50 miles), year of diagnosis ( , , or ), patient education and income based on county of residence (determined by the 2012 American Community Survey administered by the US Census), and comorbidities (as quantified by the Charlson-Deyo score). 21,22 The following disease characteristics were evaluated: tumor grade (low, moderately, or poorly differentiated), lymphovascular invasion, serum carcinoembryonic antigen (positive or negative, as defined by values above the upper limit of normal), tumor stage, pathologic down-staging (complete response, defined as posttreatment classification T0N0 [ypt0n0] disease; partial response, defined as any tumor [T] or lymph node [N] regression; or no response), and tumor regression grade (based on pathologic assessment, ranging from a complete response [no viable cancer cells] to a poor response [extensive residual cancer cells]). Finally, the following treatment characteristics were also evaluated: receipt of NMAC or NCRT, time from the end of neoadjuvant therapy to surgery (<1 month, 1-3 months, or > 3 months), extent of resection (complete [R0] or incomplete [R1]), the number of lymph nodes surgically examined (0-7, 8-12, 12-17, or 18 lymph nodes), and receipt of adjuvant chemotherapy (yes or no). Statistical Methods Statistical analyses were conducted using SAS version 9.4 (SAS Inc, Cary NC) and SAS macros developed by the Biostatistics and Bioinformatics Shared Resource at Winship Cancer Institute. 23 Descriptive statistics for each variable were determined. Follow-up was assessed using the reverse Kaplan-Meier method. 24 The univariate association of each covariate with receipt of NCRT versus NMAC was assessed using the chi-square test for categorical covariates and an analyses of variance for numerical covariates. The primary endpoint of OS was measured in months from LAR to death from any cause or last followup if the patient remained alive. The univariate association (univariate analysis [UVA]) of each variable with OS was tested using Cox proportional hazards models and log-rank tests. A multivariable Cox proportional hazards 784 Cancer March 1, 2017

3 Neoadjuvant Therapy in Rectal Cancer/Cassidy et al Figure 1. This is a Consolidated Standards of Reporting Trials (CONSORT) diagram of the current study population. NCDB indicates National Cancer Data Base. model (multivariate analysis [MVA]) was fit by using a backward elimination method applying a P 5.20 removal criterion. The effect of CRT versus NMAC on OS stratified by different variables was estimated by testing the interaction effect between each modality and each relevant variable. It is noteworthy that the variables lymphovascular space invasion, serum carcinoembryonic antigen, and tumor regression grade were eliminated from MVA because many patients were missing these data. Kaplan- Meier plots were produced to compare the survival curves by subgroups along with log-rank P values, and P <.05 was considered statistically significant. Propensity score matching (PSM) was then used to address potential treatment selection bias. A logistic regression model predicting receipt of CRT versus NMAC was carried out to predict the propensity score by age, Cancer March 1,

4 TABLE 1. Summary of Patient Demographic, Socioeconomic, Tumor, and Treatment Characteristics, N 5 21,707 Variable No. of Patients (%) Patient characteristics Age at diagnosis. y <65 13,508 (62.2) (37.8) Sex Men 13,616 (62.7) Women 8091 (37.3) Race White 18,932 (87.2) African American 1669 (7.7) Other/missing 1106 (5.1) Facility type Community cancer program 3728 (17.2) Comprehensive community cancer program 9583 (44.1) Academic/research program 7452 (34.3) Unknown 944 (4.3) Facility location Northeast 3996 (18.4) South 7234 (33.3) Midwest 6582 (30.3) West 2951 (13.6) Unknown 944 (4.3) Distance from home zip code to treatment facility, miles < (46) (41.7) > (12.3) Missing 244 Insurance status Not insured 959 (4.4) Private insurance 11,250 (51.8) Government insurance: 1464 (6.7) Medicaid and VA Medicare 7618 (35.1) Missing 416 (1.9) County of residence Rural 527 (2.4) Urban 3958 (18.2) Metropolitan 16,617 (76.6) Missing 605 (2.8) Median annual income quartile of county of residence <$30, (12) $30,000-$35, (18.4) $36,000-$45, (27.7) $46, (38.4) Missing 755 (3.5) Percentage with no high school degree in patient county of residence 29% 3275 (15.1) 20%-28.9% 4912 (22.6) 14%-19.9% 5328 (24.5) <14% 7435 (34.3) Missing 757 (3.5) Year of diagnosis (24.4) (33.4) (42.2) Charlson-Deyo Score 0 17,480 (80.5) (15.8) (3.7) TABLE 1. Continued Variable No. of Patients (%) Tumor characteristics Stage T2N (4.9) T3N0 11,366 (52.4) T3N (42.8) Grade Well differentiated 1522 (7) Moderately differentiated 14,969 (69) Poorly differentiated 2265 (10.4) Not determined/missing 2951 (13.6) Lymphovascular invasion Yes 891 (4.1) No 6059 (86.8) Missing 14,726 Carcinoembryonic antigen status Positive 5517 (40.9) Negative 7985 (59.1) Missing 8205 Treatment characteristics Type of neoadjuvant therapy Multiagent chemotherapy 274 (1.3) Concurrent chemoradiotherapy 21,433 (98.7) Tumor regression grade based on pathologic response Complete response: No viable cancer cells 1793 (52.8) Moderate response: Single cells or 1051 (31) small groups Minimal response: Residual cancer 550 (16.2) Poor response: Extensive residual cancer 43 Missing 16,866 Pathologic down-staging Complete: ypt0n (13.2) Partial response 6176 (36.5) No response 8517 (50.3) Missing 4778 No. of regional lymph nodes examined (28.7) (24.8) (24.7) (21.8) Missing 565 Extent of resection R0 20,061 (92.4) R1 891 (4.1) Missing 755 (3.5) Time from end of neoadjuvant therapy to surgery, mo <1 226 (1) ,512 (48.4) ,969 (50.5) Adjuvant chemotherapy Yes 6047 (33) No 12,251 (67) Missing 3409 Abbreviations: VA, US Veterans Affairs Administration; yp, posttreatment classification. race, facility type and location, insurance status, county of residence, distance to treatment facility from home zip code, year of diagnosis, education and income level of home zip code, Charlson-Deyo score, lymph node status, 786 Cancer March 1, 2017

5 Neoadjuvant Therapy in Rectal Cancer/Cassidy et al TABLE 2. Multivariable Analysis of Factors Predicting for the Use of Neoadjuvant Chemoradiotherapy Over Neoadjuvant Multiagent Chemotherapy a Variable OR (95% CI) OR P P Patient characteristics Insurance status Unknown 0.29 ( ).02 b Medicare 0.46 ( ).05 <.01 b Government insurance: Medicaid and VA 0.60 ( ).25 Private insurance 0.68 ( ).33 Not insured Distance from home zip code to treatment facility, miles < ( ) <.01 b ( ).01 b <.01 b >50 Facility location Unknown 1.11 ( ).79 West 0.68 ( ) Midwest 1.17 ( ).42 South 0.85 ( ).40 Northeast County of residence Unknown 1.04 ( ).93 Rural 2.07 ( ) Urban 1.46 ( ).04 b Metropolitan Tumor characteristics Grade Well differentiated 0.74 ( ).23 Moderately differentiated 1.17 ( ) Poorly differentiated 1.20 ( ).49 Not determined/missing Stage T2N ( ).59 T3N ( ).01 b.01 b T3N1 Treatment characteristics No. of regional lymph nodes examined ( ) ( ) ( ) ( ).18 Missing Time from the end of neoadjuvant therapy to surgery, mo < ( ) ( ).01 b.01 b 3-6 Abbreviations: CI, confidence interval; OR, odds ratio; VA, US Veterans Affairs Administration. a Backward selection with an a level for removal of.2 was used. The following variables were removed from the model: age at diagnosis, Charlson-Deyo score, facility type, median income of county of residence, percentage of county of residence without a high school degree, race, surgical margin status, sex, receipt of adjuvant chemotherapy, and year of diagnosis. b This P value indicates a statistically significant difference. T2N1 versus T3N0 versus T3N1 tumors, tumor grade, surgical margin, number of regional lymph nodes examined, time from the start of neoadjuvant therapy to surgery, and receipt of adjuvant chemotherapy. Patients who received NMAC were matched with those who received NCRT at a 1:5 ratio based on the propensity score using a greedy algorithm and an SAS macro provided by the Mayo Clinic. 25 The matching caliper was set at a 0.2 standard deviation of logit of the propensity score. After matching, the balance of covariates between 2 cohorts was evaluated using standardized differences, and a value of <0.1 was considered a negligible imbalance. 26 The effects were estimated in the matched sample using a Cox model with a robust variance estimator 27 for OS. RESULTS Patient Characteristics In total, 21,707 patients met study entry criteria (Fig. 1); and, of these, 274 (1.3%) received NMAC. The median Cancer March 1,

6 Figure 2. Overall survival is compared between patients who received neoadjuvant chemoradiotherapy (NCRT) and those who received neoadjuvant multiagent chemotherapy (NMAC). CI indicates confidence interval. age at diagnosis was 60 years (range, years). Overall, the median follow-up was months. The median distance from a patient s zip code to the treatment facility was 11.2 miles. The median time from the end of neoadjuvant therapy to LAR was 62 days. For patients in the NCRT group, the median dose of RT was 50.4 grays (Gy) (range, Gy). The median daily RT dose was 1.8 Gy. Table 1 summarizes the study population. Predictors of Receipt of Neoadjuvant Treatment Type On UVA, age < 65 years, Northeast and Midwest practice locations, having government insurance or a lack of insurance, patient s zip code within 50 miles of treatment facility, and treatment at community-based programs were demographic factors predictive of receiving NCRT. Lymph node-positive tumors (T2N1 or T3N1), attaining a complete pathologic response, and an interval from the end of neoadjuvant therapy to surgery of 1 to 3 months were tumor and treatment factors associated with the receipt of NCRT versus NMAC (all P <.05) (Supporting Table 1; see online supporting information). On MVA, patients who had a distance from home zip code to treatment facility <50 miles, a lack of insurance, and T3N1 tumors were more likely to receive NCRT versus NMAC (all P.01) (see Table 2). Impact of Neoadjuvant Treatment Type on OS The actuarial 5-year OS rate was 75% for patients who received NCRT compared with 67.2% for those who received NMAC (P <.01) (Fig. 2). NCRT was associated with improved OS on UVA (hazard ratio [HR], 0.74; 95% confidence interval [CI], ; P <.01) (Supporting Table 2; see online supporting information) and MVA (HR, 0.77; 95% CI, ; P 5.03). On MVA, additional variables associated with improved OS included age <65 years, treatment at an academic cancer program, having private insurance, living in a zip code with a median income >$46,000, Charlson-Deyo score of 0 or 1, having at least 12 lymph nodes examined during surgery, and LAR within 1 to 3 months from the end of neoadjuvant therapy (all P <.05) (see Table 3). Additional variables associated with worse OS on MVA included African American race, male gender, diagnosis between 2010 and 2012, poorly differentiated tumors, ct3n1 tumors, R1 resection, and not receiving adjuvant 788 Cancer March 1, 2017

7 Neoadjuvant Therapy in Rectal Cancer/Cassidy et al chemotherapy (all P <.05) (see Table 3). The receipt of NCRT did not have a significant interaction effect with tumor classification, facility type, or time from the end of neoadjuvant therapy to surgery on OS (all P >.10), indicating a benefit from NCRT across all levels of these variables. Impact of Neoadjuvant Therapy on OS With PSM Analysis After matching, PSM analysis identified 1343 patients who received NCRT and 274 who received NMAC, and the 2 groups were balanced on demographic, tumor, and treatment characteristics (Supporting Table 3; see online supporting information). On PSM analysis, the 5-year actuarial OS rate was 73.4 % for patients who received NCRT compared with 65.8% for those who received NMAC (P 5.02) (Fig. 3). In Cox analysis of the PSM patient cohorts, there was a statistically significant improvement in OS with NCRT (HR, 0.72; 95% CI, ; P 5.01). DISCUSSION In this analysis of the NCDB, we observed that, for patients with ct2n1, ct3n0, and ct3n1 rectal adenocarcinomas who underwent LAR, the receipt of NCRT versus NMAC was associated with improved OS on UVA, MVA, and interaction effect testing. This survival benefit persisted when PSM was implemented to balance the 2 cohorts for patient, tumor, and treatment characteristics. Our study is the first in the literature to compare these 2 neoadjuvant treatment strategies head-to-head and to examine factors that predict for treatment choice. We observed that patients who lived in a zip code within 50 miles of the treatment facility, who lacked insurance, and who had T3N1 tumors were more likely to receive NCRT. Our study also confirmed known poor prognostic factors in rectal cancer, including inadequate lymph node dissection, delay to surgery after neoadjuvant therapy, 28 incomplete resection, poorly differentiated tumors, and higher stage tumors. There is a movement to eliminate or reduce the use of NCRT in favor of NMAC in select patients, as evidenced by the ongoing PROSPECT trial, to decrease RTinduced toxicity ,29 The basis of this accruing trial is a pilot, single-arm, phase 2 study from Memorial Sloan- Kettering Cancer Center that enrolled 32 patients. 19 Patients on that trial had clinically staged (on magnetic resonance imaging and endoscopic rectal ultrasound) T3 tumors (N0 or N1) that were deemed resectable and did not require temporary diverting ostomy or endorectal stent. Each patient received 4 cycles of modified 5-FU, leucovorin, and oxaliplatin (FOLFOX6) 30 with bevacizumab followed by 2 cycles of modified FOLFOX alone. If there was any progression, then patients instead received concurrent 5-FU and RT. From 3 to 6 weeks after they completed neoadjuvant chemotherapy, patients underwent surgery with TME. Adjuvant chemotherapy was personalized, and FOLFOX was recommended by the investigators. All patients underwent R0 resection, and 25% had a pathologic complete response. Only 2 patients required neoadjuvant RT because of FOLFOX intolerance, and 1 required adjuvant RT. The 4-year results were extremely promising, demonstrating 0% local recurrence and 91% OS 19 and prompting creation of the PROS- PECT trial. The PROSPECT trial randomizes patients with ct2n1, ct3n0, and ct3n1 rectal adenocarcinomas to receive 6 cycles of FOLFOX versus the standard, concurrent 5-FU and RT to 50.4 Gy. Patients in the FOL- FOX group are clinically and radiographically assessed on magnetic resonance imaging or endoscopic rectal ultrasound studies; and, if there is no progression and they have an estimated regression of 20%, then they proceed to LAR with TME. Patients who progress or have <20% regression receive standard, concurrent 5-FU and RT. Systemic chemotherapy is delivered after surgery, and adjuvant CRT is recommended for patients who do not achieve an R0 resection and did not undergo preoperative RT. The trial is currently enrolling patients at multiple sites in the United States. 20 There are several reasons for the interest in elimination of NCRT from the treatment paradigm in rectal cancer. Justification for the elimination of upfront RT in all patients with stage II and III rectal cancers comes from the low local recurrence rates observed with modern TME techniques, with overall 5-year estimates of less than 10%. 10,11,31 Moreover, randomized evidence indicates that RT improves local control in the TME era with no improvement in OS. 32 Improved systemic therapies, 12,13,33 with increasing pathologic complete responses and longer disease-free survival intervals, have also been used to rationalize the elimination of RT. However, beyond the above-mentioned single-arm study of 32 patients, there is a lack of prospective evidence for the elimination of RT in patients with stage II and III rectal cancers. A recently published Canadian review examining 566 patients who would meet eligibility for the PROS- PECT trial demonstrated that these patients had improved recurrence-free survival (HR, 0.75) compared with patients who did not meet enrollment criteria, suggesting that the elimination of RT could be considered. 29 Cancer March 1,

8 TABLE 3. Multivariable Analysis of Patient, Tumor, and Treatment Factors and Their Association With Overall Survival a Variable HR (95% CI) HR P P Patient characteristics Age at diagnosis, y ( ) <.01 b <.01 b <65 Facility type Comprehensive community cancer program 1.02 ( ).74 <.01 b Academic/research program 0.85 ( ).01 b Unknown 0.74 ( ).01 b Community cancer program Race Other/missing 0.93 ( ) b African American 1.22 ( ).01 b White Sex Men 1.22 ( ) <.01 b <.01 b Women Insurance status Missing/unknown 0.77 ( ).16 Private insurance 0.71 ( ) <.01 b <.01 b Government insurance: Medicaid and VA 0.98 ( ).83 Medicare 0.98 ( ).83 Not insured Median income quartile of county of residence Missing/unknown 1.37 ( ) <.01 b <$30, ( ) <.01 b <.01 b $30,000-$35, ( ) <.01 b $36,000-$45, ( ) <.01 b $46,000 Charlson-Deyo score ( ) <.01 b <.01 b ( ) <.01 b 2 Year of diagnosis ( ).01 b.01 b ( ) <.01 b County of residence Missing/unknown 1.20 ( ).04 b Rural 0.90 ( ) Urban 0.97 ( ).40 Metropolitan Tumor characteristics Grade Well differentiated 1.15 ( ).07 <.01 b Moderately differentiated 1.11 ( ).06 Poorly differentiated 1.70 ( ) <.01 b Not determined/missing Stage T2N ( ) <.01 b T3N ( ) <.01 b <.01 b T3N1 Treatment characteristics Type of neoadjuvant therapy Concurrent chemoradiotherapy 0.72 ( ).01 b.01 b Multiagent chemotherapy Surgical margin status Missing/unknown 1.04 ( ).68 <.01 b Positive 2.53 ( ) <.01 b Negative No. of regional lymph nodes examined Missing/unknown 0.84 ( ) b ( ).01 b ( ) <.01 b ( ) Cancer March 1, 2017

9 Neoadjuvant Therapy in Rectal Cancer/Cassidy et al TABLE 3. Continued Variable HR (95% CI) HR P P Time from end of neoadjuvant therapy to surgery, mo < ( ).01 b ( ) <.01 b <.01 b 3-6 Adjuvant chemotherapy No 1.38 ( ) <.01 b <.01 b Yes Abbreviations: CI, confidence interval; HR, hazard ratio; VA, US Veterans Affairs Administration. a Backward selection with an a level for removal of.2 was used. The following variables were removed from the model: distance from patient zip code to treatment facility, facility location, and percentage of zip code without a high school degree in the patient s county of residence. b This P value indicates a statistically significant difference. Figure 3. Overall survival is compared in a propensity score-matched analysis between patients who received neoadjuvant chemoradiotherapy (NCRT) and those who received neoadjuvant multiagent chemotherapy (NMAC). CI indicates confidence interval. The current study has several strengths and limitations. The strengths include that our series serves as the largest assessment of the receipt of NMAC in patients with ct2n1, ct3n0, and ct3n1 disease and serves as the only comparison of these patients with those who received the current standard of care (NCRT). In addition, Cancer March 1,

10 to our knowledge, we are the first to summarize these practice patterns in the United States and to report on factors that potentially influence clinicians choice of treatment modality. The limitations of our study are the lack of uncaptured variables by the NCDB, like other cancer registries, leading to a potential inherent imbalance between the NCRT and NMAC groups. We performed stratified analysis, PSM analysis, and interaction variable testing to account for these confounding variables, and there was a persistent OS benefit from NCRT; however, even with these multiple statistical methods to balance the cohorts, there may be confounders that cannot be accounted for in the NCDB. We excluded patients who were coded as not being able to receive RT for medical reasons, but there may be confounding, uncaptured reasons why a patient did not receive RT that could impact OS. In addition, we could not determine specifically whether all patients in the NMAC would meet eligibility criteria for the PROSPECT trial. We can conclude that they are a similar cohort based on clinical stage, surgery modality, and their OS outcomes, but we do not have specific information on the methods used to stage these patients or on their clinical response before surgery. In addition, we could not capture specific chemotherapy regimens in the NCDB and could only determine whether patients had received multiagent chemotherapy in the NMAC group and single-agent chemotherapy in the NCRT group. Toxicity information and patient-reported outcomes are not reported, so we could not comment on the morbidity of the treatment regimens. Finally, the NCDB does not provide information on cause of death, and it is possible that the OS benefit from NCRT on stratified analysis was because of competing risks of death. CONCLUSION In summary, our series indicates that the receipt of NCRT is associated with improved OS compared with NMAC on stratified analysis for select patients with stage II and III rectal adenocarcinomas who undergo LAR. This benefit persisted when demographic, tumor, and treatment variables were balanced. Patients who lived in a zip code within 50 miles of their treatment facility, had more advanced disease, and lacked insurance were more likely to receive NCRT instead of NMAC. Until results from the accruing PROSPECT trial are published, our series suggests that patients who have ct2n1, ct3n0, and ct3n1 rectal adenocarcinomas should receive NCRT, and NMAC should be reserved for clinical trials. FUNDING SUPPORT The research reported in this publication was supported in part by the Biostatistics and Bioinformatics Shared Resource of Winship Cancer Institute of Emory University and the National Institutes of Health/National Cancer Institute under award P30CA CONFLICT OF INTEREST DISCLOSURES The authors made no disclosures. AUTHOR CONTRIBUTIONS Richard J. Cassidy: Study design, data collection/analysis, writing initial draft, and validation. Yuan Liu: Study design, data collection/analysis, and validation. Kirtesh Patel: Data collection/ analysis and validation. Jim Zhong: Data collection/analysis and validation. Conor E. Steuer: Validation. David A. Kooby: Validation. Maria C. Russell: Validation. Theresa W. Gillespie: Study design, data collection/analysis, and validation. Jerome C. Landry: Study design and validation. REFERENCES 1. Siegel RL, Miller KD, Jemal A. Cancer statistics, CA Cancer J Clin. 2016;66: Edge S, Byrd DR, Compton CC, Fritz AG, Greene FL, Trotti A, eds. AJCC Cancer Staging Manual. 7th ed. New York: Springer; NIH consensus conference. Adjuvant therapy for patients with colon and rectal cancer. JAMA. 1990;264: Fisher B, Wolmark N, Rockette H, et al. Postoperative adjuvant chemotherapy or radiation therapy for rectal cancer: results from NSABP protocol R-01. J Natl Cancer Inst. 1988;80: Krook JE, Moertel CG, Gunderson LL, et al. Effective surgical adjuvant therapy for high-risk rectal carcinoma. N Engl J Med. 1991; 324: Tepper JE, O Connell MJ, Petroni GR, et al. Adjuvant postoperative fluorouracil-modulated chemotherapy combined with pelvic radiation therapy for rectal cancer: initial results of Intergroup J Clin Oncol. 1997;15: Sauer R, Becker H, Hohenberger W, et al. Preoperative versus postoperative chemoradiotherapy for rectal cancer. N Engl J Med. 2004; 351: Sauer R, Liersch T, Merkel S, et al. Preoperative versus postoperative chemoradiotherapy for locally advanced rectal cancer: results of the German CAO/ARO/AIO-94 randomized phase III trial after a median follow-up of 11 years. J Clin Oncol. 2012;30: Roh MS, Colangelo LH, O Connell MJ, et al. Preoperative multimodality therapy improves disease-free survival in patients with carcinoma of the rectum: NSABP R-03. J Clin Oncol. 2009;27: Merchant NB, Guillem JG, Paty PB, et al. T3N0 rectal cancer: results following sharp mesorectal excision and no adjuvant therapy. J Gastrointest Surg. 1999;3: Carne PW, Nelson H. Are we overtreating rectal cancer: time for another trial? Ann Surg Oncol. 2004;11: Grothey A, Sargent D, Goldberg RM, Schmoll HJ. Survival of patients with advanced colorectal cancer improves with the availability of fluorouracil-leucovorin, irinotecan, and oxaliplatin in the course of treatment. J Clin Oncol. 2004;22: Goldberg RM, Gill S. Recent phase III trials of fluorouracil, irinotecan, and oxaliplatin as chemotherapy for metastatic colorectal cancer. Cancer Chemother Pharmacol. 2004;54(suppl 1):S57-S Minsky BD, Conti JA, Huang Y, Knopf K. Relationship of acute gastrointestinal toxicity and the volume of irradiated small bowel in patients receiving combined modality therapy for rectal cancer. J Clin Oncol. 1995;13: Cancer March 1, 2017

11 Neoadjuvant Therapy in Rectal Cancer/Cassidy et al 15. Matzel KE, Bittorf B, Gunther K, Stadelmaier U, Hohenberger W. Rectal resection with low anastomosis: functional outcome. Colorectal Dis. 2003;5: Temple LK, Wong WD, Minsky B. The impact of radiation on functional outcomes in patients with rectal cancer and sphincter preservation. Semin Radiat Oncol. 2003;13: Prabhu RS, Cassidy RJ, Landry JC. Radiation therapy and neutropenia. Curr Probl Cancer. 2015;39: Fernandez-Martos C, Pericay C, Aparicio J, et al. Phase II, randomized study of concomitant chemoradiotherapy followed by surgery and adjuvant capecitabine plus oxaliplatin (CAPOX) compared with induction CAPOX followed by concomitant chemoradiotherapy and surgery in magnetic resonance imaging-defined, locally advanced rectal cancer: Grupo Cancer de Recto 3 study. J Clin Oncol. 2010;28: Schrag D, Weiser MR, Goodman KA, et al. Neoadjuvant chemotherapy without routine use of radiation therapy for patients with locally advanced rectal cancer: a pilot trial. J Clin Oncol. 2014;32: Alliance for Clinical Trials in Oncology. PROSPECT: chemotherapy alone or chemotherapy plus radiation therapy in treating patients with locally advanced rectal cancer undergoing surgery. Available at: cancer&rank=1. Accessed August 15, Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45: Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40: Nickleach D, Liu Y, Shrewsberry A, Ogan K, Kim S, Wang Z. SAS macros to conduct common biostatistical analyses and generate reports. SESUG 2013: Proceeding of the South East SAS User Group. Available at: Altman DG, McShane LM, Sauerbrei W, Taube SE. Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK): explanation and elaboration [serial online]. PLoS Med. 2012;9: e Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques [paper ]. Presented at SAS Users Group International (SUGI) 26; April 22-25, 2001; Long Beach, CA. 26. Austin PC, Grootendorst P, Anderson GM. A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: a Monte Carlo study. Stat Med. 2007;26: Lin DY, Wei LJ. The robust inference for the Cox proportional hazards model. J Am Stat Assoc. 1989;84: Foster JD, Jones EL, Falk S, Cooper EJ, Francis NK. Timing of surgery after long-course neoadjuvant chemoradiotherapy for rectal cancer: a systematic review of the literature. Dis Colon Rectum. 2013;56: Bosse D, Mercer J, Raissouni S, et al. PROSPECT eligibility and clinical outcomes: results from the Pan-Canadian Rectal Cancer Consortium. Clin Colorectal Cancer. 2016;15: Allegra CJ, Yothers G, O Connell MJ, et al. Phase III trial assessing bevacizumab in stages II and III carcinoma of the colon: results of NSABP protocol C-08. J Clin Oncol. 2011;29: Kusters M, Marijnen CA, van de Velde CJ, et al. Patterns of local recurrence in rectal cancer; a study of the Dutch TME trial. Eur J Surg Oncol. 2010;36: Kapiteijn E, Marijnen CA, Nagtegaal ID, et al. Preoperative radiotherapy combined with total mesorectal excision for resectable rectal cancer. N Engl J Med. 2001;345: Rodel C, Liersch T, Becker H, et al. Preoperative chemoradiotherapy and postoperative chemotherapy with fluorouracil and oxaliplatin versus fluorouracil alone in locally advanced rectal cancer: initial results of the German CAO/ARO/AIO-04 randomised phase 3 trial. Lancet Oncol. 2012;13: Cancer March 1,