Fax: (323) 865-3671
Rates and predictors of chemotherapy use for stage III colon cancer
A systematic review
Article first published online: 24 OCT 2008
Copyright © 2008 American Cancer Society
Volume 113, Issue 12, pages 3279–3289, 15 December 2008
How to Cite
Etzioni, D. A., El-Khoueiry, A. B. and Beart, R. W. (2008), Rates and predictors of chemotherapy use for stage III colon cancer. Cancer, 113: 3279–3289. doi: 10.1002/cncr.23958
- Issue published online: 4 DEC 2008
- Article first published online: 24 OCT 2008
- Manuscript Accepted: 15 JUL 2008
- Manuscript Revised: 10 JUL 2008
- Manuscript Received: 6 FEB 2008
- colonic neoplasms;
- adjuvant chemotherapy;
- quality of healthcare;
- systematic review;
- health disparities
Despite consensus regarding the benefits of chemotherapy for stage III colon cancer, multiple reports have found significant variations in rates of use. In the current study, the authors attempted to systematically review reports of the community rates at which chemotherapy is administered for stage III colon cancer in the US, and in so doing plan strategies for improving rates of use. A systematic search strategy was undertaken using MEDLINE, Web of Science, and bibliographies to find reports of the rates at which patients with stage III colon cancer receive chemotherapy. A total of 22 studies published since 1990 were identified, with rates of chemotherapy use ranging from 39% to 71%. Age and comorbidity were found to be the most significant patient factors, but studies also found racial/ethnic and socioeconomic disparities in the rates of chemotherapy. Patients treated at teaching hospitals did not clearly receive chemotherapy more often. Oncologists and surgeons who treat a higher volume of colorectal cancer patients were more likely to have chemotherapy initiated in their patients. The authors developed a conceptual model of the process pathway experienced by patients with stage III colon cancer to demonstrate areas of potential underuse of chemotherapy. Nearly half of patients with stage III chemotherapy in the US do not receive chemotherapy. Although many patients are too old or frail to benefit appropriately, for many patients chemotherapy is simply not initiated. Attention needs to be focused on systematic approaches to prevent systems failures that result in underuse. Guidelines regarding chemotherapy use in elderly patients are especially important. Cancer 2008. © 2008 American Cancer Society.
Carcinoma of the colon and rectum is the second leading cause of cancer-related mortality in the US. In 2007, an estimated 150,000 individuals will have been diagnosed with the disease and 52,000 will have died of causes related to colon and rectal cancer.1
For colon cancer, the primary treatment is surgical, and the need for adjuvant therapy depends largely on the extent of the disease as determined at the time of surgical resection. The National Institutes of Health (NIH) consensus guidelines published in 1990 advocated the use of postoperative adjuvant chemotherapy for patients with stage III colon cancer.2 More recent findings have confirmed the clinical efficacy and cost-effectiveness of a 6-month regimen of 5-fluorouracil-based chemotherapy in improving rates of disease recurrence (by 34%) and overall survival (by 24%).3, 4 Over the last 20 years, significant advances have been made in the efficacy of chemotherapeutics and more therapeutic options are available, but the core recommendations are unchanged. Recent professional recommendations from the National Comprehensive Cancer Network maintain the same practice guidelines initially coalesced by the NIH in 1990.5
Unfortunately, recent studies have shown that a significant proportion of patients with stage III colon cancer do not receive standard adjuvant therapy.6, 7 In this study, we sought to systematically review existing knowledge regarding the use of chemotherapy for patients with stage III colon cancer in the community. Our specific objectives were to 1) obtain an estimate of the rates of adjuvant therapy use across different populations, 2) catalog the factors that are associated with use, and 3) identify directions in which health services research may be directed to improve appropriate use of adjuvant therapy.
MATERIALS AND METHODS
A systematic search strategy was constructed to identify existing literature. Broadly, we sought reports of the rates at which patients with stage III colon cancer receive postoperative adjuvant chemotherapy.
Only studies analyzing the use of chemotherapy in patients diagnosed with stage III colon cancer in the US were included for review. At a minimum, studies needed to define 1) a cohort of patients (including sample size) with stage III colon cancer based on established cancer staging practices, 2) the population from which the cohort was identified, and 3) the proportion of patients in the cohort that either received or was offered chemotherapy. In an initial review, we found several analyses of the rates at which patients are evaluated by medical oncologists, which is a subtly different endpoint. To augment understanding of potential barriers to care, we allowed our search strategy to encompass analyses using this auxiliary endpoint. The results of these analyses are reported separately.
We included only studies analyzing the care of patients treated since the publication of the 1990 NIH consensus guidelines. Studies that did not differentiate between colon and rectal cancer were excluded whenever this lack of distinction precluded the assessment of chemotherapy use for patients with stage III colon cancer (separate from stage II or III rectal cancer). If possible, data present in studies were used to calculate rates of standard adjuvant therapy, even when it was not explicitly reported. For example, to report rates of chemotherapy use in patients aged ≥75 years, we combined patient counts from groups of patients ages 75 to 79 years, 80 to 84 years, and ≥85 years. Studies in which the cohort of interest was developed based on survey response were excluded. Investigations published in letter or abstract form were not included.
A comprehensive literature search was performed to identify any relevant publications. This search was comprised of a 3-staged approach. First, we deployed a keyword search strategy with the search terms (“colonic neoplasms” OR “colorectal neoplasms” OR “antineoplastic agents/therapeutic use” AND ”chemotherapy, adjuvant” OR “combined modality therapy” OR “colonic neoplasms/therapy” OR “colorectal neoplasms/therapy”). This search strategy was applied to the PubMed database, with limits including English language studies, human studies, and studies published since 1990 (4468 unique references). Of these 4468 references, we identified 15 studies that met our inclusion criteria. Second, the bibliographies of each of these 15 studies were collected and combined into a single list of references. This yielded 98 studies, of which 1 additional study met inclusion criteria. Third, using Web of Science, we performed a reverse literature search to identify studies that had cited any of the 16 studies identified in the previous 2 search strategies. This final step yielded 340 studies, and 5 additional studies meeting our inclusion criteria. Our search therefore yielded a total of 22 publications. We used PubMed/MEDLINE and library resources to access abstracts and, when necessary, the full text of any study that was published in 1990 or later. Searches were performed during the week of September 24 through September 28, 2007.
We used a systematic review method instead of a meta-analytic approach because of the heterogeneity of the existing literature. Many of the studies we analyzed used the Surveillance, Epidemiology, and End Results-Medicare (SEER-Medicare) data source. These studies varied somewhat in terms of the time periods of analysis, and with regard to the details of their multivariate models and inclusion/exclusion criteria, but several studies used the same source data. Meta-analytic methods would be inappropriate in the context of these overlapping analyses. In addition to analyzing rates of adjuvant therapy use, we gathered information regarding patient, provider, and environmental factors that are associated with use. We only reported those factors that were analyzed within the context of a multivariate model predicting use of chemotherapy. Studies in which analyses did not distinguish between stage III colon cancer and stage II or III rectal cancer were excluded. We did not specifically analyze the type or duration of chemotherapy administered because this was not reported in the majority of studies analyzed.
Of the 22 studies that met our criteria, 20 were population-based6–25 and 2 were hospital-based.26, 27 The majority of the population-based studies used data from either SEER9–13, 15, 16, 18–23 or the National Cancer Database6, 14; 6 other studies used specific regional cancer registries as the basis for identifying a study cohort.7, 8, 17, 24, 25, 27 Many of these studies analyzed overlapping populations of patients, especially the studies based on SEER-Medicare datasets. We did not, therefore, calculate the overall size of the cohort of patients considered in the studies we reviewed.
The quality of the 22 studies was generally good, with the endpoint of interest clearly stated in each. In most of the studies, the stated endpoint was defined as the initiation of any adjuvant chemotherapeutic regimen. For 3 of the 22 studies, the outcome variable of interest was defined as whether chemotherapy was offered, recommended, refused, or administered.19, 24, 27 Population inclusion and exclusion criteria were clearly defined within each study. Several of the studies conducted multivariate analyses on populations that included patients with either colon or rectal cancer.8, 10, 19 To minimize heterogeneity, these studies were not included in our review of predictors of chemotherapy use. However, each of these studies did report absolute rates of chemotherapy use in patients with stage III colon cancer, and these results were included for analysis.
The statistical techniques used within these studies were appropriate. None of the 11 studies reporting multivariate analyses was based on data sources with stratified sampling methods and therefore weighted analytic techniques (eg, SUDAAN) were not used.7, 9, 13, 15–17, 21, 22, 24–27 With the exception of 2 studies, each of the studies used multivariate logistic regression for significance testing. One study used Poisson regression,15 and another used a hierarchical generalized linear regression model to analyze the nested effects of patient and surgeon characteristics.16
The hypotheses driving each of the 11 studies were not always clearly defined. In 6 of these studies, the specific aims of the studies were primarily exploratory, seeking to analyze a broad range of factors that might be associated with rates of chemotherapy use9, 17, 22, 24, 25, 27 The remaining 5 studies were targeted more at specific testable hypotheses.7, 13, 15, 16, 21
The rates of chemotherapy use varied from 39% to 71% (Table 1).6–27 A total of 12 multivariate analyses were present in 11 different studies. Of these 12 analyses, 10 examined receipt (initiation) of chemotherapy as the outcome of interest7, 9, 13, 15, 17, 21, 22, 24, 25, 27 and 2 examined evaluation by a medical oncologist as the designated outcome.15, 16 The results of these different multivariate analyses are summarized in Tables 2 and 3.
|Reference||Data Source||Cohort Year(s)||Sample Size||Received Chemotherapy,%||Comments|
|Jessup 199614||National Cancer Data Base||1985-1993||9509||46.3%||Some patients diagnosed/treated before 1990|
|Tropman 199927||NCI Physician Query Database Rural NC/SC||1991||52||56%|
|Mahoney 200017||NJ Hospital Cancer Registry||1989-1996||69||49%|
|Sundararajan 200122 and Sundararajan 200223||SEER-Medicare||1992||1015-1082||48%-51%||Studies used nearly identical methodologies. Discrepancies in sample sizes and results are due to differences in inclusion criteria.|
|VanEenwyk 200224||Washington State discharge records, cancer registry, census records||1996-1997||632||61.9%*|
|Potosky 200219||SEER-Based Patterns of Care Study||1990-1991, 1995||779||56%*|
|Ayanian 20038||California Cancer Registry, California discharge records||1996-1997||1421||66.6%|
|Wu 200425||Louisiana Tumor Registry||2001||303||66%|
|Jessup 20056||National Cancer Data Base||1990||6207||39%|
|Abraham 200626||Michael E. DeBakey, VA Medical Center, Houston, TX||1999-2003||62||71%|
|Cronin 200610||SEER Patterns of Care Study||2000||263||65.5%*|
|McGory 20067||California Cancer Registry, California Discharge Records||1994-2001||13,231||47.8%|
|Luo 200615 and Luo 200716||SEER-Medicare||1992-1999||6158-7569||58.5%-60.2%||Studies used nearly identical methodologies. Discrepancies in sample sizes and results are due to differences in inclusion criteria.|
|Reference||Significant Predictors||Nonsignificant Predictors||Comments|
|Tropman 199927||Age* and race (white vs nonwhite)†||Tumor size, comorbidity, sex, region, and y of diagnosis (1991 vs 1996)||Small sample size (n = 70)|
|Mahoney 200017||Sex (female*) and age‡||None||Small sample size (n = 69)|
|Schrag 200121||Age*, race (black vs white)*, no. of positive lymph nodes†, comorbidity*, and y of diagnosis†||Sex, race (other vs white), and income|
|Sundararajan 200122||Age*, year of diagnosis†, race (black vs white)*, no. of positive lymph nodes†, and comorbidity*||Sex and race (Hispanic vs white)|
|VanEenwyk 2002§24||Age*, low income*, and insurance (Medicare vs private)*||Insurance (Medicaid/other vs private)|
|Wu 2004§25||Age* and hospital status (COC-approved vs not)†||Race/sex‖, insurance (private vs public/none), residence (urban vs rural), poverty, and comorbidity|
|Baldwin 20059||Age*, race (black vs white)*, comorbidity*, married†, LOS for resection*, rehospitalization*, region, MOVOLYR†, y medical oncologist in practice†, medical oncologist admissions to hospital in which resection was performed†, hospital ownership (nonprofit/for-profit vs government)†, and hospital status (teaching vs nonteaching)*||Sex, oncologist board certification, census tract education level, census tract median income, y of diagnosis, medical oncologist admissions to hospital in which resection was performed, and MOVOLYR2|
|McGory 20067||Sex (male)†, age*, comorbidity*, poverty*, and insurance (Medicaid/private/other vs Medicare)*||Race (black/Hispanic/Asian/other vs white) and y of diagnosis|
|Luo 200615||Age*, race (black vs white), married†, y of diagnosis†, no. of positive lymph nodes†, and comorbidity*||Race (Asian/Hispanic vs white), sex, region, and poverty|
|Gross 200713||Comorbidities (CHF/COPD/diabetes)*|
|Reference||Sample Size||% Evaluated||Significant Predictors||Nonsignificant Predictors||Comments|
|Baldwin 20059||5294||78.7%||No multivariate analysis was performed||Main point of the study was analysis of black/white differences in use among patients who were evaluated by oncologists|
|Abraham 200626||62||92%||No multivariate analysis was performed|
|Luo 200615||7569||78.1%||Age*, race (black vs white)*, comorbidity*, sex (female†), married†, y of diagnosis†, no. of positive lymph nodes†, and region||Race (Asian/Hispanic vs white), poverty, and tumor grade|
|Luo 200716||6158||82.3%||Age*, surgeon factors‡, race (other vs white)*, married†, comorbidity*, state insurance*,§, comorbidity*, and no. of positive lymph nodes†||Sex, zip code median income, and race (black/Hispanic vs white)|
In the studies we reviewed, older patients received adjuvant chemotherapy significantly less often than younger patients. Unadjusted rates for patients aged 65 to 75 years ranged from 70% to 76% but dropped sharply in older age groups, to 8% to 10% in patients aged ≥85 years.11, 18, 21–23 Within the reviewed literature, age was modeled both as a categoric variable9, 16, 21, 22, 24, 25 and as a continuous variable7, 15, 27 in multivariate analyses. In each of these models, advanced age was found to be a significant predictor of failure to be evaluated by medical oncologists and/or failure to initiate chemotherapy.
Ten studies analyzed the role of sex in predicting the receipt of chemotherapy, and 3 of these studies reported statistically significant findings.7, 15, 17 Mahoney et al, using a single-institution study (n = 69 patients), found women were less likely to receive adjuvant treatment than men (adjusted odds ratio of 5.8) based on a statistical model that adjusted only for age.17 In a larger study using California Cancer Registry data and hospital discharge records (n = 13,231 patients), McGory et al found male patients were more likely to receive chemotherapy than females (adjusted odds ratio of 1.11).7 Luo et al found that women with stage III colon cancer were slightly more likely than men (adjusted relative risk of 1.042) to be evaluated by a medical oncologist.15
Potential racial/ethnic disparities in the use of chemotherapy were investigated in most of the studies we reviewed. Of 9 studies that analyzed race using multivariate models, 7 found some evidence of disparity. When present, disparity was present when comparing black, African-American, or nonwhite patients with white patients.9, 15, 16, 21, 22, 25, 27 The most focused of these studies was published by Baldwin et al in 2005, who found that among patients ages 66 to 70 years, black patients received chemotherapy at a rate that was slightly less than that of white patients (adjusted relative risk of 0.88).9 Their model adjusted for a broad range of patient factors, social factors, and environmental factors and still found a statistically significant effect based on race. Two studies found race/ethnicity to be nonpredictive of the use of chemotherapy.7, 25 McGory et al examined a large cohort of patients (n = 13,231 patients), but did not control for as diverse a range of environmental factors (eg, characteristics of the treating oncologist and hospital).7 Wu et al used a variable classification that categorized race and sex into 4 categories (white males, white females, black males, and black females), with nonsignificant results.25
In 3 studies, married patients demonstrated higher rates of chemotherapy use than nonmarried patients. Studies analyzing rates of evaluation by oncologists15, 16 and initiation of chemotherapy9, 15 found marital status to be a significant predictor of chemotherapy use. The impact of marital status on chemotherapy use was small, however, with an adjusted relative risk of 1.16 (for married vs unmarried patients) in the largest of these studies.15
The effect of comorbidity on the likelihood of receiving chemotherapy was analyzed in 9 studies.7, 9, 13, 15, 16, 21, 22, 25, 27 The majority of studies modeled comorbidity based on the Charlson system, stratifying patients into 3 or 4 levels based on historical diagnoses. Seven of these 9 studies found increasing comorbidity to be predictive of a lower likelihood of receiving adjuvant chemotherapy.7, 9, 13, 15, 16, 21, 22 The 2 studies that did not find a correlation were likely limited by small sample size.25, 27 The magnitude of the effect of comorbidity was large. For the most comorbid patients (Charlson level 2+ or 3+), the odds ratios (relative to a Charlson level of 0) for receiving chemotherapy were 0.39 in the analysis of the California Cancer Registry by McGory et al7 and 0.38 to 0.44 in SEER-Medicare data.21, 22
Each of the 9 studies controlled for age, but none investigated an interaction term between age and comorbidity. Length of inpatient hospitalization after surgical resection and postoperative rehospitalization were found to be negative predictors of the use of chemotherapy.9 A recent study by Gross et al found that specific chronic illnesses (congestive heart failure, chronic obstructive pulmonary disease, and diabetes) were associated with a decreased likelihood of initiating adjuvant chemotherapy.13 Of these specific comorbidities, congestive heart failure was the most powerful negative predictor for initiating chemotherapy, with an adjusted odds ratio of 0.49, compared with 0.81 and 0.83, respectively, for chronic obstructive pulmonary disease and diabetes.
The impact of socioeconomic status (SES) on the use of chemotherapy was examined in 7 studies.7, 9, 15, 16, 21, 24, 25 Each of these studies was based on cancer registry data, linked either to Medicare or hospital discharge data. SES variables were constructed by linking to a median income within a patient's census tract, and stratified either by quartile, as poverty versus nonpoverty area (if ≥20% of residences were below the poverty level), or as a percentage of patients in the census tract living below 200% of the poverty level. Three studies found disparities based on these constructed variables,7, 16, 24 whereas other authors found no significant impact of SES on the use of chemotherapy.9, 15, 21, 25 The magnitude of the effect of SES was quite small in the study by McGory et al (adjusted odds ratio of 0.991 for likelihood of receiving chemotherapy among patients living below 200% poverty threshold),7 but was much larger in the study by VanEenwyk et al (adjusted odds ratio of 2.3 for failure to receive adjuvant treatment among patients in the lowest income quartile).24 One study used enrollment in state insurance plans as a proxy for low income, and found enrollees to be less likely to be evaluated by oncologists (adjusted odds ratio of 1.54 for those not on state insurance plans).16
Number of positive lymph nodes
Four studies found that patients with more positive lymph nodes had a greater overall use of chemotherapy.15, 16, 21, 22 These studies found a higher rate of evaluation by oncologists as well as the more frequent initiation of chemotherapy among patients with a higher number of positive lymph nodes.
Only 1 study, by Luo et al, analyzed tumor grade, and this study found it to be a nonsignificant predictor of evaluation by an oncologist.15 Tropman et al investigated whether tumor size (stratified as >3 or <3 cm) was predictive of the use of chemotherapy.27 In their analysis, which controlled for no other tumor-related factors (except for stage III colon cancer as an inclusion criteria), tumor size was not found to be a significant predictor.
Luo et al analyzed variations in the use of chemotherapy related to the patient's surgeon.16 They found that 20% of the variation in rates of evaluation could be attributed to the surgeon, and that patients treated by surgeons who had graduated more recently, higher volume surgeons, and those practicing within a major teaching hospital were more likely to be evaluated for chemotherapy. This study did not estimate the magnitude of this effect.
The role of the oncologist in driving chemotherapy use was analyzed in 1 study by Baldwin et al.9 In their study, they found that patients treated by oncologists who had been in practice for a longer period (>10 years), oncologists with a higher annual volume of consultations for colon/rectal cancer, and oncologists who practiced in the hospital in which the surgical resection took place were all more likely to receive chemotherapy. Other factors related to the medical oncologist, including board certification in internal medicine and whether the medical oncologist admits patients to the hospital in which the surgical resection was performed, were both found to be nonpredictive of the use of chemotherapy.
Hospital factors were analyzed in 3 of the studies we reviewed.9, 16, 25 Baldwin et al found that patients treated at teaching hospitals were less likely to receive chemotherapy (adjusted odds ratio of 0.77).9 In contrast, Luo et al found that patients treated by surgeons affiliated with a teaching hospital were more likely to be evaluated by an oncologist (adjusted odds ratio of 1.3).16 Wu et al found that patients treated in hospitals with cancer programs approved by the American College of Surgeons' Commission on Cancer were more likely to receive chemotherapy (adjusted odds ratio of 2.0).25
Half of the studies reviewed were based on SEER-Medicare data, and in these studies insurance coverage was homogenous. The type of insurance was specifically analyzed in 3 studies, each of which was based on state cancer registries with information regarding insurance coverage obtained through a linkage to hospital discharge data.7, 24, 25 McGory et al7 found a lower use of chemotherapy in patients covered by Medicaid or private plans compared with patients covered by Medicare (adjusted odds ratio of 0.65), whereas VanEenwynk et al24 found higher rates of use in patients covered by private plans (adjusted odds ratio of 2.2). Wu et al noted no significant differences between patients with public insurance or who were uninsured versus those with private insurance.25
Year of Diagnosis
Five studies used multivariate techniques to determine whether chemotherapy use increased over time since the introduction of the NIH consensus guidelines.7, 15, 21, 22, 27 In 3 of these studies, SEER-Medicare data were used to analyze patients diagnosed between 1992 and 1999, and in each of these analyses chemotherapy use was found to be more frequent in more recent years.15, 21, 22 Luo et al investigated temporal trends in the rates at which patients with stage III colon cancer are evaluated by oncologists, and found evaluation to be more likely to occur in the more recent years of their study period.15 McGory et al used California Cancer Registry data, and stratified the year of diagnosis into 1997 through 2001 versus 1994 through 1996.7 Based on this demarcation, they found no significant difference in receipt of chemotherapy. Tropman et al, with a sample size of 80 patients, found the use of chemotherapy to be no different in 1996 compared with 1991.27 Several other studies analyzed temporal trends in the use of chemotherapy without multivariate analysis.6, 10, 19 Based on data from the National Cancer Data Base, Jessup et al noted an increase in chemotherapy use from 39% in 1990 to 64% in 2002.6 Potosky et al19 and Cronin et al10 reported improvements across 3 time periods (1991, 1995, and 2000) using data from a SEER patterns of care study, although they did not specifically state the magnitude of these changes. Unadjusted rates of use for all studies reporting rates of chemotherapy are displayed in a bubble plot shown in Figure 1 (Fig. 1).
Regional differences in the use of chemotherapy were analyzed in 2 studies. Tropman et al examined colon cancer treatment in rural North and South Carolina, and found no differences in patterns of care between these 2 states with regard to receipt of chemotherapy.27 Luo et al found no differences in the rates of use of chemotherapy, but did find that 1 SEER area (Detroit) had a higher rate of medical oncology evaluation compared with the majority of other SEER areas.15 The magnitude of this regional effect is difficult to interpret because the Detroit region was the reference group in the multivariate analysis.
Our review found that the use of adjuvant chemotherapy for patients with stage III colon cancer has been well-studied. We identified 22 studies analyzing the care provided to different populations of patients, and estimated that rates of chemotherapy use vary from 39% to 71%, with the most recent reports showing rates of 48% to 62%. Within these studies, 12 separate multivariate analyses were performed, identifying a broad range of factors that are at least somewhat responsible for variations and significant rates of nonuse.
The appropriate use of adjuvant therapy is the end result of a process pathway involving multiple steps. Luo et al proposed a 2-step pathway consisting of 1) referral to medical oncology followed by 2) use (or nonuse) of chemotherapy.16 Based on our analysis, we sought to expand on this model in a way that would add structure to future efforts aimed at improving the rates of appropriate use of adjuvant therapy (Fig. 2).
In the model proposed by Luo et al, the first step in the pathway for the adjuvant treatment of a patient with stage III colon cancer is consultation with a medical oncologist.16 It is important to realize, however, that this first step encompasses 2 distinct processes of patient care: the initiation of a referral and the propensity of a patient to accept and follow through with an evaluation. Our review found evidence that problems may exist in both of these processes. Mahoney et al17 examined a population of 35 patients with stage III colon cancer who did not receive chemotherapy. For 12 of these patients (34%), they found no evidence that chemotherapy was offered or discussed. They also found that 11 patients (31%) either refused a referral or failed to follow up with an appointment.17 In a similar study, Wu et al found that of 101 patients who did not receive chemotherapy, nearly 40% had no identifiable reason.25 Using a SEER-Medicare dataset, Luo et al examined whether the surgeon plays a role in determining whether patients are evaluated by an oncologist.16 They found that surgeons bear responsibility for approximately 20% of the variations in whether patients are seen by a medical oncologist. Taken together, these studies hint at what is potentially significant underuse.
Despite these issues, the vast majority of patients with stage III colon cancer undergo oncologic evaluation (Table 2). Based on SEER-Medicare data, Baldwin et al and Luo et al found that 78% to 82% of patients with stage III colon cancer are evaluated by medical oncologists.9, 15, 16 However, such an evaluation is not a guarantee that a patient with stage III colon cancer will receive standard adjuvant therapy, because these same studies reported that 27% to 31% of patients evaluated by an oncologist did not receive chemotherapy.9, 15 To our knowledge, the most important study examining reasons for the nonuse of standard adjuvant therapy was conducted by Ayanian et al using data from the California Cancer Registry.8 In this study, provider surveys were used to identify the reasons why adjuvant chemotherapy was not offered to eligible patients with colorectal cancer. Patient refusal accounted for 29.6% of nonuse, comorbid illness for 21.6%, and advanced age for 11.4%. The most illuminating finding from this study, however, was that for 14.8% of patients, ‘no reason’ was listed, and ‘not clinically indicated’ was the reason given for another 21.6% of cases of nonuse. Similar research into the care of patients with breast cancer has shown that 32% of cases of failure to use appropriate adjuvant therapy have no definable reason other than system failure.28
The last step in the process pathway for patients with stage III colon cancer is the continuation and completion of adjuvant treatment. Although this was not a focus of the current review, 2 recent analyses used SEER-Medicare data to analyze the duration of adjuvant therapy received by patients with stage III colon cancer. These studies found that 22% to 31% of patients were unable to complete a course of treatment, and that the mortality rate was 20% to 47% lower among patients who completed therapy versus those who were unable. In addition to age and frailty, no clear reasons for these failures could be determined.11, 18
In summary, our conceptual model categorizes 4 specific processes within which patients are at risk for failing to receive appropriate adjuvant therapy: 1) referral to an oncologist, 2) acceptance and follow through with evaluation, 3) oncologic consultation/evaluation, and 4) the ability of patients to maintain a regimen of chemotherapy. We have found evidence pointing to issues regarding the nonuse of standard adjuvant therapy related to each of these processes. In what directions should efforts be made to address these shortcomings?
Quality assurance efforts need to be aimed at identifying and promoting key process measures. On several fronts, this is already happening. The American College of Surgeons has worked through its own Commission on Cancer to develop quality-of-care measures for breast and colorectal cancer, and these measures were endorsed by the National Quality Forum in April 2007. One of these measures is that postoperative chemotherapy be considered or administered within 4 months of diagnosis in patients aged <80 years with colon cancer that involves regional lymph nodes. In a similar vein, the Medicare Quality Improvement Community, through its Physician's Quality Reporting Initiative, is advocating a quality measure, namely the proportion of patients with stage III colon cancer who were ages 18 to 80 years and were prescribed chemotherapy. These efforts, although important, need to be expanded in a way that improves their usefulness. Simply knowing that a patient did not receive adjuvant therapy does not necessarily indicate substandard care. Our model classifies failure to receive appropriate adjuvant therapy into 1 of several domains. Patient factors, such as noncompliance and refusal of care, could have a dramatic impact on rates of evaluation by oncologists and the use of chemotherapy. Similarly, a patient who is evaluated for adjuvant chemotherapy and documented as refusing care may actually have received inadequate communication regarding the role of therapy. The next generation of quality indicators should evolve to consider these possibilities, and report them in a way that enables targeted efforts at quality improvement.
Process-oriented approaches to assessing patterns of treatment are clearly important, but attention also needs to be paid to environmental conditions that foster evidence-based quality of care. Environments in which surgeons, oncologists, and other physicians work collaboratively are intuitively more likely to result in an improved quality of care. The existing literature has touched on this concept. Two of the studies we reviewed analyzed whether patients treated in teaching hospitals were more likely to be evaluated by a medical oncologist, and these studies yielded conflicting results.9, 16 Baldwin et al found that patients treated at teaching hospitals were less likely to initiate chemotherapy, whereas Luo et al found that patients treated by surgeons affiliated with a teaching hospital were more likely to be evaluated by an oncologist.9, 16 Wu et al found that treatment in a hospital approved by the American College of Surgeons' Commission on Cancer was associated with a greater use of chemotherapy.25 Collaborative models for the delivery of care have great appeal, but have not yet achieved their potential. Furthermore, although a multidisciplinary model of care may be an important step, it is unlikely that all patients with colon cancer will be cared for in these types of systems. Alternative approaches should be investigated, with an eye toward empowering hospitals, cancer registries, and payor systems to assist in coordinating services. Research should also be directed toward those barriers faced by patients to identify and address problems with access to care. To our knowledge, relatively little is known regarding the impact of financial, personal, and pragmatic (eg, transportation, scheduling) considerations on uptake of care. One novel approach is to identify dedicated ‘navigators’ to advocate for patients during their cancer treatment program.29, 30 These types of systems have the potential to significantly improve the quality of care delivered to all patients, especially to vulnerable populations and those with cultural/language barriers.
Finally, focused attention should be directed toward fostering evidence-based standards of care regarding the use of adjuvant chemotherapy in elderly patients. A significant body of research suggests that chemotherapy is well-tolerated and efficacious in an elderly patient population.4, 31–36 Undoubtedly, there are patients who are poor candidates for adjuvant therapy, either because of frailty or comorbidity. Of the 22 articles we reviewed, none included an assessment of functional status, and this is a significant shortcoming of the research in this area. In considering the appropriateness of chemotherapy, physicians are asked to weigh factors related to age, comorbidity, and functional status. The extent to which these factors determine the relative risks and benefits of adjuvant therapy needs to be examined closely. Advances in chemotherapy medications, especially the introduction of oral chemotherapeutics, make ongoing research in this area even more important.
Nearly 20 years have passed since the publication of the 1990 NIH consensus guidelines. Although much has changed, much has remained the same. Approximately 40% of patients with stage III colon cancer fail to receive standard adjuvant therapy, even though the weight of evidence strongly supports its use. Earlier work by Hodgson et al reviewed the knowledge base regarding the relation between provider factors and a broad range of treatment-related processes and outcomes.37 They found significant variations in terms of the patterns of treatment experienced by patients with colorectal cancer. In the current study, we chose to focus on 1 specific process of care-the receipt of chemotherapy for stage III colon cancer-and found significant problems in terms of its uniform use for eligible patients. Despite the impressive amount of research performed on this topic to date, there is still significant room for improvement in assuring that patients with colon cancer receive appropriate adjuvant therapy.
- 5National Comprehensive Cancer Network. Clinical Practice Guidelines in Oncology: Colon Cancer. Fort Washington, Penn: National Comprehensive Cancer Network; 2007.
- 10Patterns of care for adjuvant therapy in a random population-based sample of patients diagnosed with colorectal cancer. Am J Gastroenterol. 2006; 101: 2308–2318., , , , , .Direct Link:
- 25Treatment patterns for stage III colon cancer and factors related to receipt of postoperative chemotherapy in Louisiana. J La State Med Soc. 2004; 156: 255–261., , , , , .
- 26Receipt of recommended therapy by patients with advanced colorectal cancer. Am J Gastroenterol. 2006; 101: 1320–1328., , , , .Direct Link:
- 31Adjuvant treatment of colorectal cancer in the elderly. Cancer Control. 1995; 2: 36–38., , , et al.
- 32Systematic review of management of colorectal cancer in elderly patients. Clin Colorectal Cancer. 2003; 3: 165–171., , .