Determinants of adjuvant oxaliplatin receipt among older stage II and III colorectal cancer patients
Jennifer L. Lund PhD,
Department of Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
Department of Clinical Epidemiology, Aarhus University, Aarhus, Denmark
Corresponding author: Jennifer L. Lund, PhD, Department of Clinical Epidemiology, Aarhus University Hospital, Olof Palmes Allé 43-45, 8200 Aarhus N, Denmark; Fax: (011) 45 871 67215; Jennifer.Lund@dce.au.dk
Controversy exists regarding adjuvant oxaliplatin treatment among older patients with stage II and III colorectal cancer (CRC). This study sought to identify patient/tumor, physician, hospital, and geographic factors associated with oxaliplatin use among older patients.
Individuals diagnosed at age > 65 with stage II or III CRC from 2004 through 2007 undergoing surgical resection and receiving adjuvant chemotherapy were identified using the Surveillance, Epidemiology and End Results program (SEER)-Medicare database, which includes patient/tumor and hospital characteristics. Physician information was obtained from the American Medical Association. Poisson regression was used to identify independent predictors of oxaliplatin receipt. The discriminatory ability of each category of characteristics to predict oxaliplatin receipt was assessed by comparing the area under the receiver operating curve from logistic regression models.
We identified 4388 individuals who underwent surgical resection at 773 hospitals and received chemotherapy from 1517 physicians. Adjuvant oxaliplatin use was higher among stage III (colon = 56%, rectum = 51%) compared to stage II patients (colon = 37%, rectum = 35%). Overall, patients who were older; diagnosed before 2006; separated, divorced, or widowed; living in a higher poverty census tract or in the East or Midwest; or with higher levels of comorbidity were less likely to receive oxaliplatin. Patient factors and calendar year accounted for most of the variation in oxaliplatin receipt (area under the curve = 75.8%).
In 2011, an estimated 141,210 patients were diagnosed with colorectal cancer (CRC) in the United States, with 49,380 deaths. Health care spending for CRC was estimated at $14.1 billion in 2010.[2, 3] Because the median age at CRC diagnosis is 69 years, older patients account for a substantial portion of the overall CRC disease burden in the United States.
Adjuvant chemotherapy after surgical resection improves overall survival among older patients with stage III colon cancer. Three adjuvant chemotherapies are available and include 5-fluorouracil (FU), capecitabine, or the combination of 5-FU or capecitabine with oxaliplatin; no other agents have been shown to improve outcomes. Adjuvant chemotherapy with 5-FU compared with surgery alone reduces the risk of death by 24% among older patients with stage III colon cancer. Randomized controlled trials (RCTs) demonstrated that adding adjuvant oxaliplatin to 5-FU leads to an incremental 4.2% reduction in death among patients with stage III colon cancer. However, individuals enrolled in these RCTs had a median age at diagnosis of 60 to 63 years, and only 17% were ≥ 70 years, limiting the generalizability of these findings to older patients. Recent studies have shown that the addition of adjuvant oxaliplatin to 5-FU or capecitabine results in minimal, if any, incremental survival benefit for older patients with stage III colon cancer and average risk for patients with stage II colon cancer. On the basis of these findings, the National Comprehensive Cancer Network guidelines now state that there is no demonstrated benefit to the addition of adjuvant oxaliplatin to 5-FU in average-risk stage II colon cancer or in patients > 70 years. The role of adjuvant oxaliplatin in rectal cancer is not yet defined. In light of the growing concerns regarding the incremental benefits of oxaliplatin in addition to 5-FU in these subgroups and the lack of RCT evidence regarding its role in rectal cancer, we sought to examine the dissemination of adjuvant oxaliplatin among older patients and factors influencing its use in routine clinical practice.
MATERIALS AND METHODS
The Surveillance, Epidemiology and End Results (SEER)-Medicare database is a linkage of 2 large population-based data sources providing detailed clinical and health care utilization information on Medicare beneficiaries diagnosed with cancer. The SEER registries collect demographic, clinical, and tumor characteristics, vital status, and cause of death for all incident cancers reported for individuals residing within one of the registry areas, covering approximately 28% of the United States. Patients in the SEER database are linked to their Medicare Part A, B, and D (only 2007 and after) claims. Nearly all Medicare beneficiaries are eligible for Part A and almost 93% opt to enroll in the Part B. The SEER-Medicare Hospital File reports descriptive information for hospitals that are part of the SEER-Medicare database, including whether hospitals were National Cancer Institute (NCI)-designated cancer centers or participated in cooperative groups for clinical trials. Medicare claims were linked to the Hospital file using a unique number. The American Medical Association (AMA) Physician Masterfile data contain information on the characteristics of >1 million physicians in the United States, which are linked to Medicare claims by each physician's Unique Physician Identification Numbers.
We identified all patients in the SEER data diagnosed at age ≥65 years with their first primary stage II or III cancer of the colon or rectum. Exclusions included: diagnoses at autopsy or death; a missing month of diagnosis; and those without continuous Medicare Part A and B fee-for-service enrollment for the 12 months before and 8 months after diagnosis (because health care use and treatment information would not be complete for all patients).
To identify characteristics of the hospital where cancer surgery was performed, we restricted this cohort to individuals with a surgical claim (ie, colectomy or proctectomy) in the 6 months following diagnosis. If a patient had surgical claims from multiple hospitals, the first hospital was retained for analysis. We linked the cohort to the SEER-Medicare Hospital File by the provider number and year of diagnosis for each patient. Hospitals that did not match were excluded.
To examine characteristics of physicians providing chemotherapy, eligible patients had to have at least one claim for a specific chemotherapeutic agent during the initial chemotherapy treatment period (described below). The physician with the most chemotherapy-related claims during the initial chemotherapy treatment period was considered the treating physician. Unique Physician Identification Numbers that did not match to the AMA Physician Masterfile were excluded.
Measurement of Adjuvant Oxaliplatin
We categorized patients as receiving adjuvant chemotherapy with or without oxaliplatin, by using the algorithm in Figure 1. Because adjuvant chemoradiotherapy for rectal cancer may be delivered pre- or postoperatively, a different exposure window was required for patients receiving postoperative adjuvant radiotherapy to distinguish between radiosensitizing chemotherapy and systemic adjuvant chemotherapy. We defined adjuvant chemotherapy using a 120-day window to avoid misclassifying chemotherapy for recurrent cancer as adjuvant therapy. Patients were divided into an oxaliplatin group, those having ≥ 1 claim for oxaliplatin within 60 days from the first chemotherapy claim, and a nonoxaliplatin group, those without any oxaliplatin claims within 60 days from the first chemotherapy. The validity of Medicare claims to identify chemotherapy administration and oxaliplatin has been previously reported to be high.[33, 34]
Characteristics including year of diagnosis, sex, age at diagnosis (66-69, 70-74, 75-79, 80-84, or ≥ 85 years), race/ethnicity (white non-Hispanic, black non-Hispanic, other non-Hispanic, Hispanic, or unknown), marital status (married, single, other [divorced, separated, widowed], or unknown), and region of residence (Northeast, South, Midwest, or West) came from SEER. County-level metropolitan area was defined as metropolitan or nonmetropolitan. Census tract level information about the percentage of residents living below the federal poverty level served as an aggregate measure of socioeconomic status and was categorized into quartiles: ≤ 4%, 4.01 - ≤ 8%, 8.01 - ≤ 15%, and > 15.[35, 36] Each patient's preexisting health conditions in the 365 days before diagnosis were measured using the Klabunde adaptation of the Charlson Comorbidity Index.
Hospital and Physician Characteristics
Hospital characteristics included NCI center designation (none, clinical, or comprehensive), NCI cooperative membership group count (0 or ≥ 1), teaching hospital status (yes/no), type of hospital (nonprofit, private, or government), and total bed size, measured in quartiles (< 204, 204-343, 344-487, or ≥ 488). Physician characteristics included medical degree (Medical Doctor [MD] or Doctor of Osteopathy [DO]), whether the physician was trained in the United States (yes/no), year of medical school graduation (< 1981 or ≥ 1981), primary specialty (oncology, hematology/oncology, hematology, internal medicine, or other), and sex.
We estimated the proportion of patients receiving adjuvant oxaliplatin by year, cancer site, stage, and other multilevel characteristics. Patient observations were clustered within hospitals and physicians in a non-nested manner. Our analysis accounted for this correlation and produced estimates of marginal (population-averaged) associations. We estimated adjusted prevalence ratios (aPRs) for all variables using multivariable Poisson models with a log link and generalized estimating equations with an independent working correlation matrix. Stratified analyses for colon and rectal cancer were performed. Finally, we assessed the discriminatory ability of a variety of logistic regression models to determine adjuvant oxaliplatin receipt by comparing the area under the receiver operating curve (AUC). Models included combinations of calendar year and patient/tumor, physician, hospital, and geographic characteristics. All analyses were conducted using SAS, version 9.2 (SAS Institute, Inc., Cary, NC).
We identified 4388 patients diagnosed with stage II or III CRC who received chemotherapy and met all eligibility criteria for inclusion (Fig. 2). The study population was primarily over the age of 70 years at diagnosis (73%), white non-Hispanic (81%), married (61%), living in a metropolitan area (84%), and free from comorbidities (68%). Most tumors were located in the colon (85%) and diagnosed at stage III (74%) (Table 1).
Table 1. Characteristics of Older Patients With Stage II or III Colorectal Cancer by Chemotherapy Treatment Groupa
Cases obtained from the Surveillance, Epidemiology and End Results Program 17 registries were included in this analysis.
aA total of 94% of the individuals who received treatment without oxaliplatin (n = 2473) received 5-FU or capecitabine alone. All remaining individuals (n = 163) received another combination of chemotherapy agents.
bPercentage of census tract living below the poverty line and county of residence in metro area size are linked from 2000 Census data.
Abbreviations: AJCC, American Joint Committee on Cancer; SD, standard deviation.
The majority of the 1517 physicians were male (81%), MDs (97%), US-trained (67%), medical school graduates from 1981 or later (56%) with a primary specialty of oncology or hematology/oncology (76%) (Table 2). Three percent of the 773 hospitals had a NCI clinical or comprehensive cancer center designation, and 48% had at least one cooperative group membership. Almost 40% were teaching hospitals and 62% were nonprofit entities.
Table 2. Characteristics of Physicians and Hospitals Providing Care for Older Patients With Stage II and III Colorectal Cancer
aA total of 94% of the individuals who received treatment without oxaliplatin (n = 2473) received 5-fluorouracil or capecitabine alone. All remaining individuals (n = 163) received another combination of chemotherapy agents.
bAll hospital information was obtained from the year of patient diagnosis with exception of the National Cancer Institute (NCI) cancer center designation and cooperative group count, which are reported for the year 2002. The total number of hospitals reported here (n = 874) is greater than the total number of unique hospitals because some of the hospital characteristics changed over time and are reported here according to year of the patient's cancer diagnosis.
The prevalence of adjuvant oxaliplatin treatment over the 4-year study period was 52% and 46% for patients diagnosed with colon and rectal cancer, respectively. There was a steady increase in the prevalence of adjuvant oxaliplatin use over time, which was similar for all site and stage subgroups (Fig. 3). By 2007, 60% and 73% of stage II and III colon and 52% and 68% of stage II and III rectal cancer patients, respectively, received oxaliplatin.
Patients diagnosed in earlier years were less likely to receive adjuvant oxaliplatin than those diagnosed in 2007 (eg, 2004: aPR = 0.42, 95% confidence interval [CI] = 0.37, 0.47; Fig. 4). Patients with colon versus rectal cancer were more likely to receive oxaliplatin (aPR = 1.15, 95% CI = 1.05, 1.26), whereas those with stage II versus III disease were less likely (aPR = 0.65, 95% CI = 0.60, 0.71). Increasing age was associated with a gradual monotonic decrease in the likelihood of oxaliplatin receipt (eg, patients aged ≥ 85 years versus 66-69 years: aPR = 0.22, 95% CI = 0.14, 0.34) and patients with increased comorbidities were less likely to receive oxaliplatin compared to those with no comorbidity (eg, Charlson score of 1 versus 0: aPR = 0.92, 95% CI = 0.87, 0.98). In addition, patients who were separated, divorced, or widowed (versus married), living in the East or Midwest (versus West), residing in a nonmetropolitan area (versus metropolitan area), or in a census tract with a higher proportion of individuals living under the poverty level were less likely to receive oxaliplatin. No physician- or hospital-level variables were strongly associated with adjuvant oxaliplatin receipt in the overall analysis. Results from the cancer site–stratified analyses were similar, although they were less precise for rectal cancer due the smaller number of patients (data not shown). Two exceptions were that among rectal cancer patients, surgical treatment at an NCI comprehensive cancer center was associated with increased adjuvant oxaliplatin use (aPR = 1.48, 95% CI = 1.15, 1.90), whereas Hispanic ethnicity was associated with decreased use (aPR = 0.52, 95% CI = 0.32, 0.86). Among patients with stage II colon cancer, more patients with T4 tumors (40%) received oxaliplatin compared with those who had T3 tumors (36%).
The category headings in Figure 4 denote the AUC for components of the logistic regression model (calendar year, patient/tumor, physician, hospital, and geographic). The AUC reports the ability of each model to accurately distinguish between those patients who received adjuvant oxaliplatin and those who did not. When calendar year was included alone, the model had fair ability to discriminate oxaliplatin users (AUC = 68%). The addition of patient/tumor factors enhanced the model's discriminatory ability (AUC = 75.8%). However, when physician, hospital, and geographic factors were added separately to calendar year, the AUC only increased to 68.9%, 68.9%, and 69.1%, respectively. The full model AUC was 76.6%; therefore, among patients who received chemotherapy, physician and hospital characteristics contributed very little to the determination of oxaliplatin receipt.
This population-based analysis demonstrates substantial increases in oxaliplatin use among older patients with stage II and III CRC who received chemotherapy from 2004 through 2007. By 2007, the prevalence of oxaliplatin treatment was 60% and 73% among stage II and III colon cancer patients, respectively, and 52% and 68% among stage II and III rectal cancer patients, respectively. Patient- and tumor-level characteristics, together with calendar year, accounted for most of the discriminatory power in determining oxaliplatin use in older patients with CRC, whereas physician and hospital characteristics contributed little.
Patients treated by providers within the Community Clinical Oncology Program, a network connecting NCI cooperative groups to community physicians treating cancer patients, were recently found to be more likely than those who did not to receive adjuvant oxaliplatin. Community Clinical Oncology Program data was not available for our analysis; however, when restricted to stage III colon cancer patients, we did not find an association between hospital cooperative group participation and adjuvant oxaliplatin use (aPR = 0.98, 95% CI = 0.91, 1.05). This discrepancy may reflect that our study examined a slightly later time period, potentially weakening the impact of cooperative group membership on oxaliplatin diffusion. By 2007, 73% of all older patients with stage III colon cancer in our study received adjuvant chemotherapy with oxaliplatin, suggesting that its adoption in routine practice was rapid and widespread across all physicians. However, our finding that rectal cancer patients undergoing surgery at an NCI comprehensive cancer center were more likely to receive adjuvant oxaliplatin compared with similar patients at hospitals without an NCI designation may be due to the unique timing and coordination of surgery, radiation, and chemotherapy required for rectal cancer treatment. The sex and race/ethnicity of patients were not predictive of adjuvant oxaliplatin receipt, which may reflect limited variation in access to care due to the uniform insurance setting of our study.
Among patient subgroups where RCT evidence is entirely lacking (ie, stage II and III rectal cancer) or has shown no benefit (ie, stage II colon cancer), we found widespread adjuvant oxaliplatin use by 2007 in older patients with CRC. Well over half of all stage II colon and stage II and III rectal cancer patients undergoing adjuvant chemotherapy received oxaliplatin. Rectal cancer patients have not been included in RCTs examining adjuvant oxaliplatin and as such, there is no phase III trial evidence supporting its use. It is possible that higher local and systemic recurrence rates lead physicians and older patients to choose adjuvant oxaliplatin treatment[40, 41]; however, the added toxicity of (neo)adjuvant radiotherapy makes tolerance of adjuvant chemotherapy more challenging. The question of whether this practice is appropriate will hopefully be answered by ongoing trials examining the role of oxaliplatin in adjuvant chemotherapy of rectal cancer (PETACC 6 and German Rectal Cancer Study Group CAO/ARO/AIO-04). However, these trials are unlikely to include adequate numbers of older rectal cancer patients for subanalysis, and therefore some uncertainty will remain.
Even in colon cancer, the evidence to support the addition of adjuvant oxaliplatin to 5-FU in older patients is weak. An initial pooled analysis of 4 RCTs demonstrated that the efficacy of adding oxaliplatin to 5-FU was similar in older and younger patients with stage III and metastatic CRC. However, 3 recent RCT analyses relying on subgroup or pooled data have reported conflicting results regarding the efficacy of adjuvant oxaliplatin in older patients.[19, 20, 42] Two analyses of multiple observational databases found that the addition of adjuvant oxaliplatin treatment may benefit those younger than 75 years old, but for those aged ≥ 80 years, any benefit is likely modest. Therefore, because controversy remains regarding adjuvant oxaliplatin in older patients with stage III colon cancer, any potential benefits must be carefully considered alongside patient preferences and the cost and risk of adverse events.
Our study has multiple strengths. The cohort included a large sample derived from population-based cancer registries linked to longitudinal Medicare claims. These data were augmented by information from the AMA and SEER-Medicare Hospital File. The aggregation of data resulted in a rich data source to examine the influence of multilevel characteristics on the receipt of oxaliplatin. Our analysis reflects patterns of chemotherapy use in the community setting.
However, our study is limited by the constraints of the data sources. SEER-Medicare contains information on many patient and tumor characteristics that may be associated with treatment patterns, but unobserved factors such as patient preferences, financial considerations, or functional status may also influence treatment decisions. Further linkage efforts to data sources, including electronic medical records or patient surveys, may improve measurement of these factors for future research. Because clinical stage was not available in the SEER data, we relied on pathological staging and as a result may have selected patients with rectal cancer treated neoadjuvantly who had a poorer prognosis. Our analysis did not examine within-physician variation in prescribing using a random effects framework, which may further illuminate patterns of physician preference regarding the use of oxaliplatin. Finally, a number of exclusions were made in the creation of our study population and thus the findings from our analysis may not be generalizable to individuals aged < 66 years or those who were managed care enrollees.
In conclusion, following its approval in 2004, adjuvant oxaliplatin was increasingly used to treat patients with stage II and III CRC, and its use was influenced strongly by patient, as opposed to physician or hospital, characteristics. By 2007, 70% of all chemotherapy-treated patients with stage II and III CRC received adjuvant oxaliplatin, despite no evidence to support its use in rectal cancer and mounting evidence against its use in patients with stage II colon cancer and older patients with stage III colon cancer. The widespread dissemination of oxaliplatin into the older population highlights the critical importance of studying the benefits of anticancer therapies in older patient populations. Future studies should focus on the comparative effectiveness and safety of adjuvant chemotherapy with oxaliplatin in older patients with stage II and III rectal cancer, because utilization rates within these patient groups were substantial and the evidence for benefit and harm are unknown.
This research was supported in part by a grant from the National Institutes of Health (NIH) (T32 DK07634). Other coauthor funding includes NIH RO1AG023178 (to T.S.), the Agency for Healthcare Research and Quality (to T.S.), National Cancer Institute (NCI) grant K07CA16077 (to H.K.S.), and unrestricted research grants from Merck and Sanofi (to T.S.).