Opportunity lost: Adjuvant chemotherapy in patients with stage III colon cancer remains underused Adan Z. Becerra, BA,a Christian P. Probst, MD,a Mohamedtaki A. Tejani, MD,b Christopher T. Aquina, MD,a Maynor G. Gonz alez, PhD,a Bradley J. Hensley, MD, MBA,a a Katia Noyes, PhD, MPH, John R. T. Monson, MD,a and Fergal J. Fleming, MD,a Rochester, NY
Introduction. There is strong evidence supporting the efficacy of adjuvant chemotherapy for patients with pathologic, stage III colon cancer. This study examines differences in adherence to evidence-based adjuvant chemotherapy guidelines for pathologic, stage III colon cancer cases across hospital and patient subgroups. Methods. Patients with stage III colon cancer were identified from the 2003 to 2011 National Cancer Data Base (NCDB). A logistic regression model was used to estimate the odds of receipt of adjuvant chemotherapy across varying hospital and patient characteristics. A multivariable Cox proportional hazards model was used to estimate the association between receipt of adjuvant chemotherapy and 5-year survival. Risk adjusted observed/expected (O/E) outcome ratios were calculated for each hospital to compare hospital-specific quality of care during the study period. Results. A total of 124,008 patients met the inclusion criteria. Adjuvant chemotherapy was not administered to 34%. The rates of adjuvant chemotherapy have shown little improvement over time (63% in 2003 vs 66% in 2011). The Cox model indicates that patients receiving adjuvant chemotherapy had better survival (hazard ratio = 0.48, 95% confidence interval 0.47–0.49). Analysis of risk adjusted O/E ratios indicated no consistent pattern as to which hospitals were performing optimally or subopitmally over time. Conclusion. There has been no meaningful improvement in receipt of chemotherapy in patients with stage III colon cancer. The fact that chemotherapy is not being considered or offered to more than 20% of patients with node-positive colon cancer suggests that there are substantial process failures across many institutions and regions in the United States. (Surgery 2015;158:692-9.) From the Surgical Health Outcomes & Research Enterprise, Department of Surgery,a and Department of Medicine, Hematology/Oncology,b University of Rochester Medical Center, Rochester, NY
IT HAS BEEN MORE THAN 10 YEARS SINCE THE NATIONAL COMPREHENSIVE CANCER NETWORK and the American Society of Clinical Oncology published guidelines recommending the use of adjuvant chemotherapy in patients with stage III colon cancer.1-3 These guidelines were established in part on evidencebased clinical trials that demonstrated improved survival in patients with node-positive or stage III Presented at the 10th Annual Academic Surgical Congress in Las Vegas, NV, February 3–5, 2015. Accepted for publication March 17, 2015. Reprint requests: Adan Z. Becerra, BA, Surgical Health Outcomes & Research Enterprise, Department of Surgery, University of Rochester Medical Center, 601 Elmwood Avenue, Box SURG, Rochester, NY 14642. E-mail: adan_becerra@urmc. rochester.edu. 0039-6060/$ - see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.surg.2015.03.057
692 SURGERY
colon cancer receiving adjuvant chemotherapy.4-6 As cited by a recent Institute of Medicine report,7 however, the national cancer system remains in a state of crisis, with care not being patientcentered and often being implemented without scientific basis. This failure to practice evidencebased care has led to suboptimal patient outcomes and unequal treatment across many patient subpopulations. Variation in adherence to treatment guidelines among patients with stage III colon cancer has been reported previously.8-10 Data on this matter in recent years, however, are limited. It is well established by National Comprehensive Cancer Network and American Society of Clinical Oncology that adjuvant chemotherapy should be administered for 6 months after primary surgical treatment.11,12 A recent analysis of the 2003–2007 National Cancer Data Base (NCDB) reported that only 71% of patients with stage III disease were
Surgery Volume 158, Number 3
Becerra et al 693
recommended adjuvant chemotherapy with marked variation across specific patient groups.8 Given the known benefits of chemotherapy on 5year survival rates,13,14 understanding whether these patterns have persisted in more recent years is crucial to promote effective delivery of cancer care. No study, however, has updated this analysis using data from more recent years. To address this gap in the literature, the 2003– 2011 NCDB was queried for patients with pathologic, stage III colon cancer. The purpose of this study was to estimate variation in receipt of adjuvant chemotherapy for patients with stage III colon cancer across specific patient and hospital characteristics and to examine temporal trends in hospital quality of care. Understanding current trends in adherence to evidence-based guidelines in patients with stage III colon cancer will help to identify vulnerable patients at risk of not receiving optimal care as well continued systematic deficiencies in the quality of cancer care in the United States. METHODS Study population. Data for this study were retrieved from NCDB, a joint project of the American Cancer Society and the Commission on Cancer of the American College of Surgeons.15,16 NCDB is a hospital-based cancer registry that collects annual oncology and demographic data from more than 1,500 hospitals nationwide. The data are coded via a standardized protocol coordinated by the Facility Oncology Registry Data Standards.17 A total of 124,008 patients with colon cancer were identified from the 2003 to 2011 NCDB Participant User File Report. The sample initially included cases of clinical stage I III adenocarcinoma, mucinous adenocarcinoma, and signet ring cell carcinoma. The sample was further restricted to patients with a documented surgical resection who were coded as pathologic stage III. Figure 1 shows a flow diagram of this inclusion process. Predictor variables. NCDB captures a variety of patient and hospital characteristics that were considered as potential predictors of receipt of adjuvant chemotherapy and 5-year survival. Patient age was treated as continuous. Sex was stratified into male and female. Race was categorized as white, black, Asian, American Indian, and other. Education and income strata were derived from US Census data that couples area measures with zip codes. Insurance status was categorized into uninsured, private, Medicaid, Medicare, and military/VA government insurance. Urban/rural
Fig 1. Inclusion diagram.
classification is provided via the typology published by the US Department of Agriculture. The Charlson/Deyo Score18 was used to capture comorbid conditions. Year of diagnosis and histology type also were considered as predictors. Hospital characteristics that were considered in these analyses were facility type, facility location, and hospital volume. Facility location was classified using the following categories from the US Census Report: Northeast (CT, MA, ME, NH, RI, VT), Atlantic (NJ, NY, PA), Southeast (DC, DE, FL, GA, MD, NC, SC, VA, WV), Great Lakes (IL, IN, MI, OH, WI), South(AL, KY, MS, TN), Midwest (IA, KS, MN, MO, ND, SD, NE), West (AR, LA, OK, TX), Mountain (AZ, CO, ID, MT, NM, NV, UT, WY), and
694 Becerra et al
Pacific (AK, CA, HI, OR, WA). The facility types examined were academic, community cancer programs, comprehensive community programs, and other facilities. Hospital volume was classified into tertiles. Measurement of adjuvant chemotherapy. NCDB contains an explicit variable for receipt of different kinds of chemotherapy. Patients were categorized into a dichotomous measure of receipt of adjuvant chemotherapy (yes/no) regardless of whether or not it was recommended. This variable also captures the reason why chemotherapy was not administered when it was recommended by the health care provider. Measurement of survival. NCDB captures information on survival. The vital status of each patient is recorded as the date of last contact or death as well as the number of months between the date of diagnosis and this date of last contact or death. NCDB only captures survival for patients diagnosed before 2006, thus the analysis for this outcome was restricted to the sample from 2003 to 2006. Statistical analysis. v2 tests and independent 2sample t tests were used to compare covariate distributions across the 2 outcome groups. It was decided a priori that factors associated with receipt of chemotherapy at P value < .20 in these bivariate analyses would be included in multivariable analyses. A multivariable logistic regression model was used to assess variation in the odds of receiving chemotherapy across patient and hospital characteristics. In addition, Kaplan-Meier curves and a multivariable Cox proportional hazards model were used to estimate the association between receipt of chemotherapy and survival for patients diagnosed from 2003 to 2006. In addition, the population-attributable risk of death was calculated to estimate the number of deaths per year that could have been avoided had everyone in the sample received adjuvant chemotherapy as recommended by the current guidelines. Risk-adjusted observed/expected (O/E) outcome ratios were calculated for each hospital to compare hospital-specific quality of care in 2 periods: 2003–2007 versus 2008–2011. The numerator captures the observed number of patients that did not receive adjuvant chemotherapy in each period, whereas the denominator captures the expected number of patients that did not receive adjuvant chemotherapy. The expected number of patients was calculated using the sum of all patients’ estimated probabilities of not receiving chemotherapy. These probabilities were estimated using a multivariable logistic regression model that
Surgery September 2015
was fitted on the data of each study period. The corresponding exact Poisson 95% confidence interval was calculated for each O/E outcome ratio. Hospitals with a statistically significant O/E ratio < 1 were better than average performers. Hospitals with a statistically significant O/E ratio > 1 were worse than average performers. All other hospitals were categorized as average performers. The O/E outcome ratios were compared across the 2 periods to judge if the same hospitals were performing suboptimally over time. All analyses used SAS software (version 9.3; SAS Institute Inc., Cary, NC). RESULTS Adjuvant chemotherapy was not administered to 34% of the cohort. Of these, 66% did not have a reason as to why chemotherapy was not offered as part of the planned first course of therapy. The most common reasons why the chemotherapy was not administered were patient refusal (15%) and comorbidities (12%). The rates of adjuvant chemotherapy receipt have not improved much over time (63% in 2003 vs 66% in 2011). Table I displays patient and hospital characteristics for the sample stratified by receipt of adjuvant chemotherapy. Patients who did not receive adjuvant chemotherapy were more likely to be older, of lower income and education, uninsured, and have more comorbidities. Table II presents results from the multivariable logistic regression model used to estimate variation in the odds of receiving adjuvant chemotherapy across multiple characteristics. Factors associated with lower odds of receiving adjuvant chemotherapy include no insurance, lower income, increased comorbidity score, black race, and older age. The Kaplan-Meier curves (Fig 2) indicate that patients receiving chemotherapy had better survival (log rank test P < .001). The median survival in the group of patients who received adjuvant chemotherapy was 114 months versus 38 months for patients who did not receive adjuvant chemotherapy. This effect persisted in the multivariable analysis, which estimated a 52% reduction in the hazard of death (HR = 0.48, 95% confidence interval 0.47–0.49) in patients who received adjuvant chemotherapy compared with those who did not. The population-attributable risk is 21%, which indicates that more than 1,400 deaths per year in the entire US population could be avoided if all stage III patients received adjuvant chemotherapy. A total of 1,364 O/E outcome ratios were calculated for each study period (2003–2007 vs 2008–2011). From 2003 to 2007, 68% of the hospitals were average performers, 27%
Becerra et al 695
Surgery Volume 158, Number 3
Table I. Bivariate analyses that compare covariates across chemotherapy use status Variable
Chemotherapy (N = 81,637)
No chemotherapy (N = 42,371)
P value
Age Sex Male Female Race White Black Native American Asian Other Primary payer Not insured Private Medicaid Medicare Other government Income
Introduction. There is strong evidence supporting the efficacy of adjuvant chemotherapy for patients with pathologic, stage III colon cancer. This study examines differences in adherence to evidence-based adjuvant chemotherapy guidelines for pathologic, stage III colon cancer cases across hospital and patient subgroups. Methods. Patients with stage III colon cancer were identified from the 2003 to 2011 National Cancer Data Base (NCDB). A logistic regression model was used to estimate the odds of receipt of adjuvant chemotherapy across varying hospital and patient characteristics. A multivariable Cox proportional hazards model was used to estimate the association between receipt of adjuvant chemotherapy and 5-year survival. Risk adjusted observed/expected (O/E) outcome ratios were calculated for each hospital to compare hospital-specific quality of care during the study period. Results. A total of 124,008 patients met the inclusion criteria. Adjuvant chemotherapy was not administered to 34%. The rates of adjuvant chemotherapy have shown little improvement over time (63% in 2003 vs 66% in 2011). The Cox model indicates that patients receiving adjuvant chemotherapy had better survival (hazard ratio = 0.48, 95% confidence interval 0.47–0.49). Analysis of risk adjusted O/E ratios indicated no consistent pattern as to which hospitals were performing optimally or subopitmally over time. Conclusion. There has been no meaningful improvement in receipt of chemotherapy in patients with stage III colon cancer. The fact that chemotherapy is not being considered or offered to more than 20% of patients with node-positive colon cancer suggests that there are substantial process failures across many institutions and regions in the United States. (Surgery 2015;158:692-9.) From the Surgical Health Outcomes & Research Enterprise, Department of Surgery,a and Department of Medicine, Hematology/Oncology,b University of Rochester Medical Center, Rochester, NY
IT HAS BEEN MORE THAN 10 YEARS SINCE THE NATIONAL COMPREHENSIVE CANCER NETWORK and the American Society of Clinical Oncology published guidelines recommending the use of adjuvant chemotherapy in patients with stage III colon cancer.1-3 These guidelines were established in part on evidencebased clinical trials that demonstrated improved survival in patients with node-positive or stage III Presented at the 10th Annual Academic Surgical Congress in Las Vegas, NV, February 3–5, 2015. Accepted for publication March 17, 2015. Reprint requests: Adan Z. Becerra, BA, Surgical Health Outcomes & Research Enterprise, Department of Surgery, University of Rochester Medical Center, 601 Elmwood Avenue, Box SURG, Rochester, NY 14642. E-mail: adan_becerra@urmc. rochester.edu. 0039-6060/$ - see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.surg.2015.03.057
692 SURGERY
colon cancer receiving adjuvant chemotherapy.4-6 As cited by a recent Institute of Medicine report,7 however, the national cancer system remains in a state of crisis, with care not being patientcentered and often being implemented without scientific basis. This failure to practice evidencebased care has led to suboptimal patient outcomes and unequal treatment across many patient subpopulations. Variation in adherence to treatment guidelines among patients with stage III colon cancer has been reported previously.8-10 Data on this matter in recent years, however, are limited. It is well established by National Comprehensive Cancer Network and American Society of Clinical Oncology that adjuvant chemotherapy should be administered for 6 months after primary surgical treatment.11,12 A recent analysis of the 2003–2007 National Cancer Data Base (NCDB) reported that only 71% of patients with stage III disease were
Surgery Volume 158, Number 3
Becerra et al 693
recommended adjuvant chemotherapy with marked variation across specific patient groups.8 Given the known benefits of chemotherapy on 5year survival rates,13,14 understanding whether these patterns have persisted in more recent years is crucial to promote effective delivery of cancer care. No study, however, has updated this analysis using data from more recent years. To address this gap in the literature, the 2003– 2011 NCDB was queried for patients with pathologic, stage III colon cancer. The purpose of this study was to estimate variation in receipt of adjuvant chemotherapy for patients with stage III colon cancer across specific patient and hospital characteristics and to examine temporal trends in hospital quality of care. Understanding current trends in adherence to evidence-based guidelines in patients with stage III colon cancer will help to identify vulnerable patients at risk of not receiving optimal care as well continued systematic deficiencies in the quality of cancer care in the United States. METHODS Study population. Data for this study were retrieved from NCDB, a joint project of the American Cancer Society and the Commission on Cancer of the American College of Surgeons.15,16 NCDB is a hospital-based cancer registry that collects annual oncology and demographic data from more than 1,500 hospitals nationwide. The data are coded via a standardized protocol coordinated by the Facility Oncology Registry Data Standards.17 A total of 124,008 patients with colon cancer were identified from the 2003 to 2011 NCDB Participant User File Report. The sample initially included cases of clinical stage I III adenocarcinoma, mucinous adenocarcinoma, and signet ring cell carcinoma. The sample was further restricted to patients with a documented surgical resection who were coded as pathologic stage III. Figure 1 shows a flow diagram of this inclusion process. Predictor variables. NCDB captures a variety of patient and hospital characteristics that were considered as potential predictors of receipt of adjuvant chemotherapy and 5-year survival. Patient age was treated as continuous. Sex was stratified into male and female. Race was categorized as white, black, Asian, American Indian, and other. Education and income strata were derived from US Census data that couples area measures with zip codes. Insurance status was categorized into uninsured, private, Medicaid, Medicare, and military/VA government insurance. Urban/rural
Fig 1. Inclusion diagram.
classification is provided via the typology published by the US Department of Agriculture. The Charlson/Deyo Score18 was used to capture comorbid conditions. Year of diagnosis and histology type also were considered as predictors. Hospital characteristics that were considered in these analyses were facility type, facility location, and hospital volume. Facility location was classified using the following categories from the US Census Report: Northeast (CT, MA, ME, NH, RI, VT), Atlantic (NJ, NY, PA), Southeast (DC, DE, FL, GA, MD, NC, SC, VA, WV), Great Lakes (IL, IN, MI, OH, WI), South(AL, KY, MS, TN), Midwest (IA, KS, MN, MO, ND, SD, NE), West (AR, LA, OK, TX), Mountain (AZ, CO, ID, MT, NM, NV, UT, WY), and
694 Becerra et al
Pacific (AK, CA, HI, OR, WA). The facility types examined were academic, community cancer programs, comprehensive community programs, and other facilities. Hospital volume was classified into tertiles. Measurement of adjuvant chemotherapy. NCDB contains an explicit variable for receipt of different kinds of chemotherapy. Patients were categorized into a dichotomous measure of receipt of adjuvant chemotherapy (yes/no) regardless of whether or not it was recommended. This variable also captures the reason why chemotherapy was not administered when it was recommended by the health care provider. Measurement of survival. NCDB captures information on survival. The vital status of each patient is recorded as the date of last contact or death as well as the number of months between the date of diagnosis and this date of last contact or death. NCDB only captures survival for patients diagnosed before 2006, thus the analysis for this outcome was restricted to the sample from 2003 to 2006. Statistical analysis. v2 tests and independent 2sample t tests were used to compare covariate distributions across the 2 outcome groups. It was decided a priori that factors associated with receipt of chemotherapy at P value < .20 in these bivariate analyses would be included in multivariable analyses. A multivariable logistic regression model was used to assess variation in the odds of receiving chemotherapy across patient and hospital characteristics. In addition, Kaplan-Meier curves and a multivariable Cox proportional hazards model were used to estimate the association between receipt of chemotherapy and survival for patients diagnosed from 2003 to 2006. In addition, the population-attributable risk of death was calculated to estimate the number of deaths per year that could have been avoided had everyone in the sample received adjuvant chemotherapy as recommended by the current guidelines. Risk-adjusted observed/expected (O/E) outcome ratios were calculated for each hospital to compare hospital-specific quality of care in 2 periods: 2003–2007 versus 2008–2011. The numerator captures the observed number of patients that did not receive adjuvant chemotherapy in each period, whereas the denominator captures the expected number of patients that did not receive adjuvant chemotherapy. The expected number of patients was calculated using the sum of all patients’ estimated probabilities of not receiving chemotherapy. These probabilities were estimated using a multivariable logistic regression model that
Surgery September 2015
was fitted on the data of each study period. The corresponding exact Poisson 95% confidence interval was calculated for each O/E outcome ratio. Hospitals with a statistically significant O/E ratio < 1 were better than average performers. Hospitals with a statistically significant O/E ratio > 1 were worse than average performers. All other hospitals were categorized as average performers. The O/E outcome ratios were compared across the 2 periods to judge if the same hospitals were performing suboptimally over time. All analyses used SAS software (version 9.3; SAS Institute Inc., Cary, NC). RESULTS Adjuvant chemotherapy was not administered to 34% of the cohort. Of these, 66% did not have a reason as to why chemotherapy was not offered as part of the planned first course of therapy. The most common reasons why the chemotherapy was not administered were patient refusal (15%) and comorbidities (12%). The rates of adjuvant chemotherapy receipt have not improved much over time (63% in 2003 vs 66% in 2011). Table I displays patient and hospital characteristics for the sample stratified by receipt of adjuvant chemotherapy. Patients who did not receive adjuvant chemotherapy were more likely to be older, of lower income and education, uninsured, and have more comorbidities. Table II presents results from the multivariable logistic regression model used to estimate variation in the odds of receiving adjuvant chemotherapy across multiple characteristics. Factors associated with lower odds of receiving adjuvant chemotherapy include no insurance, lower income, increased comorbidity score, black race, and older age. The Kaplan-Meier curves (Fig 2) indicate that patients receiving chemotherapy had better survival (log rank test P < .001). The median survival in the group of patients who received adjuvant chemotherapy was 114 months versus 38 months for patients who did not receive adjuvant chemotherapy. This effect persisted in the multivariable analysis, which estimated a 52% reduction in the hazard of death (HR = 0.48, 95% confidence interval 0.47–0.49) in patients who received adjuvant chemotherapy compared with those who did not. The population-attributable risk is 21%, which indicates that more than 1,400 deaths per year in the entire US population could be avoided if all stage III patients received adjuvant chemotherapy. A total of 1,364 O/E outcome ratios were calculated for each study period (2003–2007 vs 2008–2011). From 2003 to 2007, 68% of the hospitals were average performers, 27%
Becerra et al 695
Surgery Volume 158, Number 3
Table I. Bivariate analyses that compare covariates across chemotherapy use status Variable
Chemotherapy (N = 81,637)
No chemotherapy (N = 42,371)
P value
Age Sex Male Female Race White Black Native American Asian Other Primary payer Not insured Private Medicaid Medicare Other government Income
