Colorectal cancer adjuvant chemotherapy trends among a nonelderly veteran cohort at a southern veterans health administration

Abstract Background For patients with high‐risk stage II or stage III colorectal cancer (CRC), adjuvant chemotherapy (AC) improves survival, yet use varies substantially across medical oncology settings. Aim Utilization of guideline concordant CRC AC was assessed at a Veterans Health Administration (VHA) facility to determine quality improvement initiatives. Methods and Results The study was a retrospective review of CRC surgeries from January 1, 2000 to December 31, 2015 at a South Regional VHA. Inclusion criteria consisted of pathologic high‐risk stage II or stage III CRC, with exclusion for age ≥80, age ≥75 hospitalized with major co‐morbidity in the prior year, and death or discharge to hospice within 30 days of the index surgery. The primary predictor was year‐group; partitioned 2000–2005, 2006–2010, 2011–2015 to account for changes in NCCN high risk stage II definitions. Primary outcome was AC receipt. Secondary outcome was reason for chemotherapy omission. Among 180 eligible surgeries (121 colon and 59 rectal cancers), patients were mostly male (96%), white (79%) and with median age 64 years. Overall, 117 (65%) received AC. Compared to 2000–2005, patients undergoing surgery between 2011 and 2015 were less likely to receive AC (odds ratio 0.35; 95% confidence interval [CI] 0.14–0.82), due to more patients declining AC (27% vs. 6%, p < .01) in the NCCN eligible cohort (N = 180), and (32% vs. 8%, p < .01) in an analysis of patients who completed appointments and had AC recommended by providers (N = 146). Conclusions Survival benefitting AC decreased over time among a nonelderly Veteran cohort eligible for AC. Evaluating care decisions and trends within other VHA facilities and outside the VHA are warranted.


| INTRODUCTION
Approximately half of the 145 600 patients diagnosed with colorectal cancer (CRC) each year will present with high-risk stage II or stage III disease. 1 For these patients, 5-fluorouracil (5FU) based adjuvant chemotherapy (AC) improves disease-free survival (DFS) and overall survival (OS). [2][3][4] Compared to observation alone, 5FU/leucovorin (LV) demonstrated a 12% absolute risk reduction [ARR] in overall survival yielding a Number Needed to Treat [NNT] = 8.4. 5 Adding oxaliplatin to 5FU/LV (FOLFOX) improved overall survival compared to 5FU/LV, however, the magnitude of benefit was substantially smaller; ARR 4.2% and NNT = 24 for stage III disease and no difference for stage II disease. 6 Despite these benefits, substantial variation (39%-98%) exists across medical oncology settings in the receipt of AC for eligible CRC patients. [7][8][9] Patient factors associated with lowers odds of receiving of AC include older age, increased co-morbidities, female sex, and nonwhite race. 9 Healthcare-system factors include nonprivate insurance, postoperative complications, increased distance to medical oncology facilities, and medical oncology facilities separate from the surgical facility or with low CRC volumes. [10][11][12] How these factors influence decisions surrounding chemotherapy workflows remains unknown but may manifest as lower rates of medical oncology referrals, physicians recommending against AC, and patients declining AC. 13 These associations suggest that the barriers and solutions to achieving quality CRC care may be highly specific to the system of care. For example, from 2003 to 2006, 77.5% of Veterans treated at a Veterans Health Administration (VHA) facility were referred to medical oncology and received timely AC. 8 Compared to the north, west, and central regions, however, Veterans treated within a south regional VHA were significantly less likely to be referred to medical oncology or receive chemotherapy. 14 Reasons for this regional difference remain uncertain. Additionally, few studies have assessed VHA AC utilization patterns over time, specifically with respect to the types of chemotherapy regimens prescribed. Given the differences in routes of administration and toxicities between regimens, assessing regimenlevel data may reveal important sub trends in prescribing patterns and chemotherapy acceptance. We evaluated receipt of AC within a south regional VHA among patients with National Comprehensive Cancer Network (NCCN) guideline-eligible high-risk stage II and stage III CRC.
The aims of this study were (1) to describe temporal trends in the receipt of NCCN-guideline-concordant AC, 15,16 and (2) to identify specific workflow processes associated with omissions or delays in achieving guideline-appropriate AC.

| Design and data sources
The study was a retrospective secondary data review cohort of Veterans undergoing colorectal resections at a south regional VHA and microscopic positive margins were added to the above criteria. 16 For rectal cancers, eligibility criteria consisted of having any of the following: T3/T4 tumor, clinical or pathologic node positivity, or neoadjuvant chemoradiation.
Due to equivocal evidence of benefit in elderly patients and older patients with significant co-morbidities, 19,20 we excluded patient's age ≥80 or ≥75 with an active co-morbidity, defined as at least one organ dysfunctional co-morbidity and an inpatient hospitalization related to that co-morbidity in the prior 365 days. Patients without an opportunity for AC consideration due to death or hospice enrollment within 30 days of index surgery were excluded.

| Outcome variables
The primary outcome was receipt of any chemotherapy (yes/no). Secondary outcomes included the type of AC regimen received (5FU/LV, capecitabine, FOLFOX, CAPEOX). Exploratory outcomes regarding patient and provider reasons for omission of AC were abstracted from chart-level information contained within documented notes.

| Data collection and measurements
Study data were collected and managed using REDCap electronic data capture tools hosted through the VHA. 21  To ensure data accuracy, 8% of charts were independently abstracted by both reviewers, and two randomly chosen variables (cancer recurrence, time to treatment) were assessed for inter-reviewer agreement, yielding very-good agreement (kappa 0.79 and 0.92).

| Receipt of adjuvant chemotherapy
Of the 180 patients eligible to receive AC, 117 (65%) received chemotherapy (    primarily by patient refusal and age, especially with respect to oxaliplatin. 28 Three additional studies documented patient refusal and age as major reasons for AC omission. 13,28,29 Finally, Ko et al. demonstrated decreasing rates of AC acceptance from 2008 to 2010 among patients ≥75 years of age or with multiple co-morbidities. 29 Given the equivocal benefit in this study population (75+ with active co-morbidity or 80+) our study specifically excluded this group yet still identified increasing refusal of AC among those without the above conditions.

| Decisions and workflows surrounding chemotherapy
Our study has limitations that should be noted. First, our study is retrospective and observational. As a cohort of predominantly white men, our VHA sample was underpowered to test for associations with sex and race and has limits on generalizability. The higher rates of comorbidities and rurality in our population likely biased our findings toward lower overall receipt of AC, however, these factors remained stable across the study period, and are unlikely to explain the temporal reduction in AC.
It is possible that unmeasured confounders impacted our findings.
For example, our study did not account for co-insurance. Prior to the 2008 recession, approximately 77% of Veterans were co-insured with non-VHA insurance, however, the number of Veterans enrolling or utilizing VA services increased during 2009-2013 (particularly in non-Medicaid expanding states). 30 While our study attempted to address numerous local VHA practice changes directly pertinent to CRC AC such as pre-operative tumor boards and intra-operative lymph node assessments. It is possible that other unmeasured practice level changes such as oncology personnel, visit time, or quality of risk/benefit discussions contributed to greater AC refusal among patients. It is possible that 5FU and LV drug shortages contributed to the reduction in AC. 5FU and LV are two of the more common chemotherapies with shortages, and oncology providers often respond to drug shortages by substituting, delaying, or even omitting therapy. 31 National VHA drug shortages may have contributed to our findings in the 2011-2015 year-group, however, patients who declined AC between 2011 and 2015 were not specific to any periods of drug shortages.
Our findings raise the possibility that declining rates of CRC AC may not simply be a phenomenon of frail and elderly patients, but a more general trend occurring across ages driven primarily by factors influencing patient attitudes toward AC. While these specific factors remain unknown, they do not appear to be related to worsening Veteran health status nor do they impact other CRC postoperative care such as surgical follow up or colonoscopies. 24 Examples of such patient factors might include diminished trust in medical professionals and experts, increased concerns over costs and toxicities of chemotherapy, greater reliance on alternative therapies or sources of information, and an inadequate understanding of risks and benefits.
Similarly, changes in patient/provider shared decision-making conversations regarding the risks and benefits of AC may be impacting acceptance. Factors influencing these conversations could involve clinic time, personnel, setting, and regimen selection. Specifically, physician conveyance of concerns related to oxaliplatin toxicity may change the acceptability and logistics of chemotherapy for patients.
With recent studies observing noninferiority for 3 versus 6 months CAPEOX in subsets of patients with HR Stage II/IIIA disease, 32 assessing duration and completion of chemotherapy will be important for future observational studies.
In summary, our study provides several novel insights on how quality care patterns can change over time and how reliance on national-level data can lead to underestimation of important local barriers and facilitators that are critical to improving high quality cancer care. Future work should investigate underlying reasons why patients chose not to receive AC despite meeting NCCN eligibility and whether our observed reduction in AC is specific to our institution and/or CRC or reflects a broader trend of patients declining potentially curative chemotherapy.