Receipt of guideline-recommended follow-up in older colorectal cancer survivors

A Population-based Analysis

Abstract

BACKGROUND

After curative resection for colorectal cancer, routine follow-up with office visits, carcinoembryonic antigen (CEA), and colonoscopy is recommended. The actual adherence to these guidelines as well as the potential overuse of testing in routine practice has not been well studied.

METHODS

The authors identified 9426 eligible patients aged ≥66 years in a linked tumor registry-claims database who were diagnosed with adenocarcinoma of the colon or rectum from 2000 to 2001. Patients were observed to 3 years after diagnosis. Receipt of ≥2 office visits per year, ≥2 CEA tests per year (years 1 and 2), and ≥1 colonoscopy within 3 years constituted guideline fulfillment.

RESULTS

Guidelines for office visits, colonoscopy, and CEA testing were met in 92.3%, 73.6%, and 46.7% of patients, respectively. In addition, receipt of 2 nonrecommended procedures, abdominal/pelvic computed tomography scans and positron emission tomography scans, was documented in 47.7% and 6.8%, respectively. Overall, 60.2% received testing below recommended levels, 17.1% at recommended frequency, and 22.7% above guideline recommendations. In a multivariate analysis, factors associated with meeting guidelines included younger age group, white race, regional stage cancers, and poorly differentiated tumors. Considerable geographic variation in meeting guidelines was also observed.

CONCLUSIONS

Many older colorectal cancer survivors in this population-based cohort underwent testing below a minimum frequency specified by clinical practice guidelines, especially with regard to CEA. Further studies should ascertain the reasons for poor compliance and the effect on patient outcome. Cancer 2008. © 2008 American Cancer Society.

Professional society guidelines typically recommend routine postoperative surveillance for patients with colorectal cancer who undergo potentially curative resection.1–8 Surveillance guidelines typically specify a combination of regularly scheduled office visits with a provider involved in cancer care coordination, carcinoembryonic antigen (CEA) testing, and colonoscopy. The goals of surveillance are to detect recurrent cancer in the colon or rectum (colonoscopy) or metastatic sites (CEA) before the onset of other symptoms or signs, as well as to screen for cancers and polyps (colonoscopy). In addition to lack of adherence to guidelines, a subset of patients may receive imaging procedures such as computerized tomography (CT) and positron emission tomography (PET) scans, which are generally not recommended by guidelines. Thus, patients may receive care in excess of guidelines as well as not meeting guideline-based standards. Although there are data about the utilization of specific procedures,9, 10 as well as published survey data about physician-reported surveillance practices,11, 12 the actual compliance with follow-up guidelines has not been well studied.

We therefore conducted the present study to determine the compliance with guideline-based surveillance recommendations as well as to describe the potential overuse of follow-up testing in colorectal survivors. By using the population-based linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database, we measured the overall adherence as well as differences across patient subgroups. The current study is the first known national, population-based study in the US to examine actual adherence to published comprehensive guidelines.

MATERIALS AND METHODS

Data Sources

The study cohort was identified from the linked SEER-Medicare database, which includes patients aged 65 years and older who are diagnosed with cancer and reside in 1 of the geographic areas contained in the SEER registries.13, 14 The files are then created from a linkage of tumor registry data contained in SEER with claims for medical services billed through Medicare. All patient identifiers such as social security number are then replaced by the SEER record number, which is anonymous, but allows linkage of patients over time.

The Patient Entitlement and Diagnosis Summary Files (PEDSF) contain the data elements collected by the SEER registries, as well as information pertaining to Medicare eligibility and enrollment. Cancer-related variables in PEDSF include demographic characteristics, previous cancer diagnoses, date of cancer diagnosis, stage at diagnosis, the most invasive treatment within 4 months of diagnosis, and date of death, if applicable. Medicare-related variables include reason for entitlement, enrollment history in Medicare, and health maintenance organization (HMO) and Parts A and B coverage.

Procedure use after cancer diagnosis was measured using data from Medicare claims. The National Claims History (NCH) files include claims from physicians, as well as physician assistants, nurse practitioners, independent clinical laboratories, and free-standing ambulatory surgery centers, among other providers. The Outpatient files include claims from institutional outpatient providers, including hospital outpatient departments, rural health clinics, and renal dialysis facilities. The Medicare Provider Analysis and Review (MEDPAR) files include all hospitalizations, as well as short stay and skilled nursing facility claims.

Cohort Identification

The cohort was obtained from patients aged 66 years and older who were included in the SEER-Medicare linked database and diagnosed with adenocarcinoma of the colon or rectum during years in which oncology society guidelines were continuously available, 2000 to June 2001. Patients were excluded if they did not undergo surgical resection per PEDSF or Medicare claims, or were diagnosed with carcinoma in situ, distant stage cancer, or unstaged cancer. Thus, the cohort was limited to patients who underwent treatment with curative intent. In addition, because of the high likelihood of incompletely reported Medicare claims, patients were excluded if they were not enrolled in Medicare Part B or were members of a Medicare-sponsored HMO during 2000 to 2004. Patients aged 65 years were excluded because of the need to obtain Medicare claims for 1 year before diagnosis to calculate comorbidity scores. Patients younger than age 65 years were enrolled in Medicare because of end stage renal disease or chronic disability and were excluded because they were not representative of the general US population. Finally, because moribund patients would presumably be less likely to undergo routine testing, patients who died within the 3.5-year period after cancer diagnosis were excluded. However, we did subsequently compare procedure use in patients who died versus those who survived.

Measures

The follow-up period of interest was from 6 months after diagnosis through 42 months after diagnosis. Procedures within 6 months were excluded to avoid including tests to evaluate possible postoperative complications and routine postoperative visits. Thus, the first year of follow-up included months 7 through 18, the second year included months 19 through 30, and the third year included months 31 through 42. Medicare claims data were examined for receipt of procedures of interest using relevant Current Procedural Terminology, Fourth Edition (CPT-4) or International Classification of Diseases, Ninth Revision Clinical Modification (ICD-9-CM) codes. Procedures included office visits (CPT-4 99,201-99,215, 99,241-99,245), carcinoembryonic antigen (CPT-4 82,378), colonoscopy (CPT-4 44,388, 44,389, 44,392, 44,393, 44,394, 45,378, 45,380, 45,382, 45,383, 45,384, 45,385, G0105; ICD-9-CM 45.23, 45.25, 45.41, 45.42, 45.43, 48.36), CT scan (CPT-4 72,191-72,194, 74,150, 74,160, 74,170, 74,175, 75,635; ICD-9-CM 88.01, 88.38), and PET scan (CPT-4 78,811-78,816, G0213, G0214, G0215, G0163, G0231). The NCH, Outpatient, and MEDPAR files were combined, and records were unduplicated by the beneficiary identification code, procedure, and date of service, such that a procedure documented in both files for the same individual and date of service was counted only once.

Stage was coded according to the historic stage, categorized as local (confined to the bowel wall) or regional (spread to adjacent organs or regional lymph nodes), as American Joint Committee on Cancer staging was not routinely available. The presence and severity of comorbid illnesses was measured using a previously validated claims data-based algorithm.15 Diagnoses included myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, peptic ulcer disease, liver disease, diabetes, hemiplegia or paraplegia, renal disease, and AIDS. The relevant MEDPAR, Outpatient, and NCH files were searched for a diagnosis code of interest, according to ICD-9-CM during the 365-day to 30-day interval (total 11 months) before the diagnosis date. To maximize the true-positive rate of a listed diagnosis, as previously suggested,16 we only included diagnoses that were either present in MEDPAR or appeared more than once in the Outpatient or NCH files. A weighted index was used to assign a comorbidity score to each patient, excluding cancer diagnoses.15

The principal outcome of interest was adherence to professional society guidelines for routine surveillance, such as those of the American Society of Clinical Oncology (ASCO)2, 3 and the National Comprehensive Care Network6 (Table 1). The American Gastroenterological Association guidelines were published in 19891 and included procedures not recommended by others, such as fecal occult blood testing and liver enzymes (Table 1). Guidelines from the American Society of Colon and Rectal Surgeons5 were published in 2004, and revised ASCO guidelines were published in 2005,4 both after the study period. Although the American Cancer Society, US Multisociety Task Force (USMSTF), and American Society for Gastrointestinal Endoscopy (ASGE) have also proposed guidelines, these are restricted to the use of colonoscopy.7, 8 Because we could not ascertain which individual set of guidelines were most likely referenced in practice, we selected a composite minimum frequency of service and procedure receipt to measure guideline adherence:

• Guidelines met: ≥2 office visits per year, ≥2 CEA tests per year in years 1 and 2, ≥1 colonoscopy within 3 years.

• Excess of guidelines: Patient met guidelines and received ≥1 CT scan for cancers not poorly differentiated and/or ≥1 PET scan.

Table 1. Recommended Colorectal Cancer Surveillance Guidelines

Guideline (Year)	Office Visits	CEA	Colonoscopy
CEA indicates carcinoembryonic antigen. * Also recommend fecal occult blood test and liver enzymes every 3 to 6 months for 2 years, then every 6 to 12 months for 2 years; chest x-ray every 6-12 months for 2 years, then yearly. † Also recommend annual computed tomography (CT) scans for patients at higher risk of recurrence. ‡ Also recommend considering annual CT scans for patients at high risk for recurrence (ie, lymphatic or venous invasion or poorly differentiated tumors).
American Gastroenterological Association (1999)1*	Every 3-6 mo for 2 y, then 6-12 mo for 2 y	Every 2 mo for 2 y then 4 mo for 2 y	1 y and then every 3 y
American Society of Clinical Oncology (1999, 2000)2, 3	Every 6 mo for 3 y	Every 2-3 mo for at least 2 y in stage II and III	Every 3-5 y
American Society of Clinical Oncology (2005)4†	Every 3-6 mo for 3 y	Every 3 mo for at least 3 y in stage II and III	3 y; if normal, every 5 y
American Society of Colon and Rectal Surgeons (2004)5	At least 3 times yearly for 2 y	At least 3 times yearly for 2 y	Every 3 y
National Comprehensive Cancer Network (updated yearly)6‡	Every 3-6 mo for 2 y then every 6 mo	Every 3-6 mo for 2 y then every 6 mo	1 y, 3 years later, 5 y later

All others were considered to have failed to have met guidelines.

Analysis

Differences in the proportion of eligible patients who met guidelines for individual procedures (office visits, CEA, colonoscopy) as well as overall guidelines were compared according to patient and clinical characteristics using chi-square analysis or chi-square for trend. In addition, a multivariable logistic regression model was used to determine the independent association of factors with meeting or exceeding guidelines (vs not meeting guidelines). Variables in the models included clinical (age, sex, race, comorbidity scores, SEER registry) and cancer-related (cancer site, stage, differentiation) factors. The multivariate analysis was repeated to determine independent predictors of exceeding guidelines (vs not meeting or meeting guidelines). Finally, we considered guideline adherence in the entire cohort, including those who died from 6 months through 3.5 years after diagnosis. We excluded those patients who died within 6 months, as they likely had an inherently poor prognosis and would not be subject to routine surveillance.

The study was approved by the University Hospitals Case Medical Center Institutional Review Board.

RESULTS

We identified a total of 9426 patients who met the study eligibility criteria (Table 2). The mean age of the cohort was 76.9 ± 6.6 years, 54.5% were women, and 86.7% were white. Cancers were located in the colon in 76.1% and rectum or rectosigmoid in 23.9%. Tumors were considered to be local stage in 59.5% and poorly differentiated in 14.6%. Most patients (82.0%) had inpatient comorbidity scores of 0, and a significant proportion of patients (45.4%) had no major comorbidities as determined by outpatient claims.

Table 2. Adherence to Individual Components and Overall Surveillance Guidelines

	n	Office Visits		CEA		Colonoscopy		Guideline Adherence
		% Met	P	% Met	P	% Met	P	Less Than Recommended	Recommended	Greater Than Recommended	P
CEA indicates carcinoembryonic antigen; SEER, Surveillance, Epidemiology, and End Results.
Age group, y			.001		.001		.001				.001
66-69	1546	92.4		56.3		82.7		48.5	19.7	31.8
70-74	2379	93.3		51.9		81.3		54.2	18.5	27.3
75-79	2583	93.7		50.8		76.4		56.2	19.2	24.6
80-84	1709	82.2		38.9		68.9		69.1	14.7	16.2
≥85	1209	87.6		26.5		47.3		83.1	9.3	7.6
Sex			.001		.036		.267				.001
Men	4287	90.9		45.5		74.2		60.9	15.5	23.6
Women	5139	93.5		47.7		73.2		59.7	18.3	22.0
Race			.002		.001		.001				.003
White	8169	92.5		46.7		74.6		59.9	17.3	22.7
Black	606	89.1		40.8		65.4		67.8	12.2	20.0
Hispanic	317	94.3		55.2		72.9		53.9	20.2	25.9
Asian	136	87.5		50.0		61.8		61.0	14.0	25.0
Other	198	94.4		48.9		68.7		58.6	17.2	24.2
Cancer type			.142		.016		.838				.001
Colon	7171	92.1		46.0		73.6		60.7	17.7	21.6
Rectum	2255	93.0		48.9		73.8		58.8	14.9	26.3
Stage			.269		.001		.006				.001
Local	5611	92.1		35.3		73.9		69.7	13.3	17.0
Regional	3815	92.7		63.5		73.2		46.3	22.5	31.2
Poorly differentiated			.418		.001		.204				.001
No	8054	92.2		44.9		73.7		61.6	12.7	25.6
Yes	1372	92.9		57.2		72.2		52.0	42.3	5.7
Comorbidity score			.001		.001		.001				.001
0	4111	90.2		48.9		75.4		57.5	18.7	23.8
1	2676	94.5		45.6		74.3		60.8	16.9	22.3
2	1256	93.3		46.3		71.3		62.6	16.4	21.0
≥3	1383	93.6		42.7		69.1		60.2	17.0	22.8
SEER registry location			.001		.001		.001				.001
Connecticut	562	91.5		40.2		71.2		66.4	15.5	18.1
Michigan	663	91.9		49.5		77.7		55.4	17.2	27.5
Hawaii	137	95.6		47.5		78.8		56.9	18.3	24.8
Iowa	872	86.0		37.2		77.6		67.4	19.0	13.5
New Mexico	179	87.7		41.3		67.6		67.6	15.6	16.8
Seattle	406	95.1		44.8		73.9		63.1	19.7	17.2
Utah	226	89.4		44.3		70.8		64.2	16.8	19.0
Georgia	256	93.0		44.9		73.4		60.9	22.3	16.8
California	1600	93.2		49.8		70.9		58.7	17.7	23.6
Kentucky	864	91.4		44.9		76.7		60.3	15.6	24.1
Louisiana	660	94.1		48.2		74.7		57.7	15.2	27.1
New Jersey	1775	94.1		49.4		73.9		58.1	15.6	26.3

The majority of patients (92.3%) fulfilled surveillance guidelines for office visits, with at least 2 visits in each year of follow-up. Whereas statistically significant differences were observed across certain patient subgroups (Table 2), with the exception of age ≥80 years, the differences in general were not of large magnitude. In addition, across the SEER registries, guidelines for office visits were fulfilled in 86.0% (Iowa) to 95.6% (Hawaii) of patients.

Only 46.7% of patients in the cohort met the guideline-based recommendations for CEA testing, with lower rates associated with age ≥80 years, African American race, and increased comorbidity scores (Table 2). Although certain cancer-specific characteristics were associated with greater use of CEA (ie, regional vs local stage, poorly differentiated vs others), no more than 63.5% of a specific subgroup fulfilled the testing criteria. There was also considerable geographic variation in the use of CEA testing, ranging from 37.2% (Iowa) to 49.8% (California) across SEER sites.

Guideline-specified criteria for colonoscopy (at least 1 examination within 3 years of diagnosis) were met by 73.6% of patients. Decreased use of colonoscopy was associated with advancing age, African American race, and increased comorbidity (Table 2). Differences in colonoscopy use according to other patient characteristics were not large, but there was variation observed across SEER sites, with a range of 67.6% (New Mexico) to 78.8% (Hawaii).

We also examined the use of 2 procedures that are not routinely recommended for cancer surveillance, abdominal/pelvic CT scans and PET scans. One or more CT scans were received by 47.7% of patients. Although guidelines recommend consideration of CT for tumors that are poorly differentiated, 51.4% of patients with poorly differentiated cancer compared with 47.1% of others underwent CT scans. Other factors associated with CT use included regional stage cancer (57.1% vs 41.3% for local stage) and younger age (53.1% for 66-69 years vs 36.2% for 85 years and older). PET scans were performed in only 6.8% of patients, and thus subgroup analyses were not done.

Overall, guideline-based recommendations were met in 17.1% and exceeded in 22.7%, with 60.2% receiving testing below the minimum frequency. Factors most strongly associated with lack of guideline adherence included local stage and nonpoorly differentiated cancers (Table 2). Other clinical characteristics associated with less than recommended surveillance included advancing age, African American race, and increased comorbidity. Geographic variation was also observed with frequency of less than recommended testing ranging from 55.4% (Michigan) to 67.6% (New Mexico).

In a multivariate logistic regression model, several factors were associated with meeting or exceeding guideline-recommended surveillance care (Table 3). The variables that were most strongly associated with meeting or exceeding guidelines were younger age group and regional stage cancers. We found that patients who were non-African American, and with lower comorbidity scores and poorly differentiated cancers, were more likely to undergo testing. As in the univariate analyses, differences were also observed across geographic location, with odds ratios (ORs) ranging from 0.81 (New Mexico) to 1.89 (Michigan).

Table 3. Factors Associated With Adherence to or Excess of Surveillance Guidelines: Multivariate Analysis

Variable	Odds Ratio	95% CI	P
CI, confidence interval; SEER, Surveillance, Epidemiology, and End Results.
Age group, y
66-69	1.0
70-74	0.77	0.67-0.88	.001
75-79	0.71	0.62-0.81	.001
80-84	0.39	0.33-0.45	.001
≥85	0.16	0.13-0.20	.001
Sex
Men	1.0
Women	1.19	1.09-1.30	.001
Race
White	1.0
Black	0.57	0.47-0.70	.001
Hispanic	1.10	0.85-1.42	.48
Asian	0.93	0.64-1.35	.70
Other	0.98	0.71-1.35	.90
Cancer type
Colon	1.0
Rectum	1.08	0.98-1.20	.13
Stage
Local	1.0
Regional	2.81	2.57-3.08	.001
Poorly differentiated
No	1.0
Yes	1.32	1.17-1.50	.001
Comorbidity score
0	1.0
1	0.93	0.83-1.03	.16
2	0.87	0.76-1.00	.05
≥3	0.75	0.66-0.86	.001
SEER registry location
Connecticut	1.00
Michigan	1.89	1.47-2.42	.001
Hawaii	1.40	0.90-2.16	.13
Iowa	0.97	0.77-1.23	.82
New Mexico	0.81	0.55-1.18	.27
Washington	1.02	0.77-1.36	.87
Utah	1.10	0.78-1.55	.60
Georgia	1.42	1.02-1.97	.04
California	1.46	1.19-1.79	.001
Kentucky	1.35	1.07-1.71	.01
Louisiana	1.63	1.27-2.10	.001
New Jersey	1.61	1.30-1.99	.001

Predictors of potential overuse of testing were also identified in a multivariate analysis. These also included younger age group (age 70-74 years: OR, 0.79; confidence interval [CI], 0.68-0.91; age 75-79 years: OR, 0.69; CI, 0.60-0.80; age 80-84 years: OR, 0.40; CI, 0.33-0.48; age ≥85 years: OR, 0.17; CI, 0.13-0.21), and regional stage cancer (OR, 2.84; CI, 2.56-3.15). There was a lower likelihood of exceeding guidelines in African Americans (OR, 0.67; CI, 0.54-0.84), but there was no association with comorbidity scores. We also observed differences across geographic areas, with odds ratios ranging from 0.77 (Iowa) to 1.88 (Michigan).

In addition to the procedure use in colorectal cancer survivors, we considered all patients who survived at least 6 months after diagnosis, including those who died during the follow-up period. Among these 12,885 patients, as may be expected given the inclusion of patients with less opportunity to undergo testing, fulfillment of guideline-recommended testing for office visits, CEA testing, and colonoscopy was somewhat lower (88.1%, 45.1%, and 62.8%, respectively). In addition, overall guideline adherence was met or exceeded in 13.6% and 20.2% of patients, respectively, with 66.2% receiving less than recommended intensity of care. In multivariate analyses, the ORs for predictors of recommended care or greater were of similar magnitude to the analysis that was limited to survivors (data not shown).

DISCUSSION

The estimated incidence of colorectal cancer is 149,000 cases in 2008, and approximately 75% of cases are diagnosed as local or regional stage tumors.17 Thus, assuming that these patients receive treatment with curative intent, there is an extremely large population of colorectal cancer survivors in whom routine surveillance testing would be recommended. The goals of surveillance testing are to discover a recurrence that is potentially resectable, identify metachronous neoplasms at an early stage, and provide reassurance to the patient.

Recommendations for routine testing have been formalized in the context of professional society practice guidelines (Table 1).1–8 In the present study, which included a large cohort of patients treated in routine clinical practice, we found that a significant number of patients did not receive testing according to practice guidelines for cancer surveillance. Surveys of colorectal surgeons11 and surgical oncologists12 in which respondents were asked to state their reported surveillance practices indicated substantial agreement with schedules outlined in guidelines. However, despite including the lowest extreme of the recommended range for testing from the above guidelines (2 office visits per year, 2 CEA tests per year in years 1 and 2, 1 colonoscopy in 3 years), in our population-based study of a large cohort of cancer survivors, we found that less than half of patients achieved compliance.

In contrast, a subset of patients received procedures not routinely recommended by guidelines, such as CT and PET scans. An unknown proportion of these nonrecommended tests were likely performed because of signs or symptoms suggestive of recurrence and/or abnormal results of routine testing such as CEA. However, we suspect that many were obtained for routine follow-up. Factors associated with potential overuse of testing were markers of poorer prognosis such as regional stage disease as well as younger age. These findings could indicate risk stratification by providers with more aggressive monitoring of patients with a greater chance of recurrence but also an inherently better survival. However, these procedures are costly and have potential false-positive findings, which may result in additional procedures. Thus, the cost-effectiveness even in this subgroup is questionable.

Whereas some of the differences in procedure receipt may be explained by clinical factors such as stage of disease or perceived longevity as measured by age or comorbidity, we also found important differences across racial groups and geographic sites. The generally lower use of testing in African Americans is likely a contributing factor to the known poorer stage-specific survival compared with Caucasians.17 The results, especially the geographic differences across SEER sites, also suggest that patient and physician preferences may influence choice of testing.

In addition to the comprehensive guidelines described in Table 1, other professional guidelines have been developed for colonoscopic surveillance after resection. Guidelines from the USMSTF changed from a 1-year follow-up interval in 199718 to 3 years in 200319 and 1 year in 2006.7 The most recent iteration was based on data from clinical trials reporting a relatively high rate of metachronous colorectal cancer within the first 2 years of follow-up.20, 21 The ASGE also published guidelines in 2006 with a recommended 1-year follow-up colonoscopy.8 Our study criteria used a conservative minimum 3-year follow-up interval, as a 1-year recommended follow-up was only present during part of the study period.18 We found that approximately 70% underwent colonoscopy during the 3-year interval.

The relative benefit of routine testing as opposed to clinical follow-up for development of signs and symptoms has been debated. Although a meta-analysis of randomized trials22 did confirm an overall survival benefit with intensive surveillance (OR, 0.73), the studies varied considerably in terms of what procedures and schedule constituted intensive follow-up. In addition, it is unclear which specific procedures accounted for the greatest survival differences. A more recently published meta-analysis23 demonstrated an improved overall survival associated with the regular use of CEA and colonoscopy, but there was no association with cancer-specific mortality.

Previous studies of population-based cohorts have examined the use of surveillance procedures. In a study that used the linked SEER-Medicare database for patients diagnosed with nonmetastatic colorectal cancer in 1991 and observed through 1994, we reported 1 or more CEA test in 35% and 1 or more colonoscopy in 52%.10 In general, procedures were more commonly used in younger patients and those with fewer comorbidities, and CEA testing was more frequent in regional stage than local stage cancer. The use of office visits and compliance with practice guidelines were not examined. In another study that included patients in SEER-Medicare diagnosed between 1986 and 1996 and observed through 1998, 47% of patients underwent 1 or more bowel surveillance procedure (colonoscopy, sigmoidoscopy, and/or barium enema) within the initial year of follow-up and 48% during the 1 to 4 year interval.9 Guideline compliance or subgroup differences in procedure use were not specifically examined.

Other studies have used administrative data from health plans to describe surveillance practices. Among patients diagnosed between 1990 and 1995, 55% received 1 or more colon examinations and 71% received at least 1 CEA test within 18 months of follow-up.24 Despite similar health insurance, lower rates of testing were observed in African Americans and residents of more impoverished geographic areas. Lower use of colon examinations in African Americans and patients aged 80 and older were also documented in a study that examined patients treated in 2 large managed care organizations.25 Racial differences in receipt of guideline-recommended surveillance care were found in our study as well. In a study of survivors of colorectal cancer from 1 health plan, adherence to guideline recommendations for individual procedures was specifically addressed.26 Compliance with 2 physical examinations and 4 CEA tests within 18 months of diagnosis was 78% and 17%, respectively, and 61% received a complete colon examination within this time interval. Our findings were similar, in that compliance with guidelines for office visits and colonoscopy were significantly higher than CEA testing.

There are several inherent limitations to the data that are used for this study. First, the accuracy of procedure coding in the Medicare population has not been formally studied, although we have evaluated its accuracy in a similar population from a large health plan.27 In our study,27 the concordance of claims data and medical cords for CEA and colonoscopy was excellent (kappa 0.80 and 0.78, respectively) and the concordance for office visits was moderate (kappa 0.48). Second, the study did not measure the indication for procedures, which could have been performed for diagnostic purposes as well as routine surveillance, particularly for procedures used to detect metastatic disease such as CEA and CT and PET scans. Other factors that may have predicted the use of testing could not be measured. Data on cancer recurrence are not available in SEER, and the accuracy of using claims data as indirect markers for recurrence (ie, second surgery or chemotherapy) has not been evaluated. If patients with recurrent disease could have been excluded, they would have likely decreased the overall frequency of routine testing. Similarly, the indications for office visits may have been for routine care as well as symptoms. Although indications could have been imputed using associated diagnosis codes, a recent analysis reported a low sensitivity for identifying diagnostic examinations.28 In addition, in a study of a large medical group practice, most procedures in colorectal cancer survivors were performed for routine indications.29 Third, based on the study criteria, a large number of patients with colorectal cancer were excluded. Because most procedures are performed in the outpatient setting, exclusion of patients without Medicare Part B insurance, which pays for outpatient services, likely increased the overall procedure use. It is unknown whether procedure use would be higher or lower in members of Medicare-sponsored HMOs, which was another exclusion criterion. The primary analysis also excluded patients who died within the follow-up period, as these individuals would have less opportunity to undergo recommended levels of testing, especially a colonoscopy within 3 years. Fourth, in this database, TNM staging was not available, and a subset of patients with stage I tumors may have appropriately not received CEA testing per American Society of Clinical Oncology guidelines.3, 4 Fifth, because the sample size was large, statistically significant differences that were observed were not necessarily clinically relevant. Sixth, the database was limited to older patients (≥66 years) with colorectal cancer, and thus adherence in younger individuals could not be measured. Seventh, because physician identifiers or specialty were not available in this database, we could not measure potential differences in practice according to provider type nor clustering by individual providers. Finally, we did not measure other factors that could affect procedure use such as access to care and socioeconomic status, which cannot be ascertained from this database. In a previous analysis, we found that census tract measures of socioeconomic status had little association with the use of follow-up testing.10

In summary, in this population-based cohort of older colorectal cancer survivors, we found that most patients underwent testing below a minimum frequency specified by clinical practice guidelines. In addition, a significant number of patients received procedures not recommended by clinical practice guidelines, suggesting potential overuse of surveillance tests. Further studies should ascertain the reasons for poor compliance and the effect on patient outcome.