Receipt of colorectal testing prior to colorectal carcinoma diagnosis†
A population-based study
Presented in part at Digestive Disease Week, New Orleans, Louisiana, May 17, 2004, and published in abstract form (Cooper GS, Payes JD. Receipt of colorectal testing prior to colorectal cancer diagnosis: a population-based study. Gastroenterology 2004;126:A–345).
Among patients with established colorectal carcinoma, the prior use of procedures for colorectal carcinoma screening is unknown. In addition, the association of disease stage at the time of diagnosis with previous procedure use has not been studied previously at a population level.
The Surveillance, Epidemiology, and End Results tumor registry files identified 5806 patients age ≥ 70 years with an initial colorectal carcinoma diagnosis in 1999. Medicare claims data from 1995 through diagnosis were extracted and identified receipt of fecal occult blood testing (FOBT), sigmoidoscopy, colonoscopy, and barium enema. Time intervals were divided into > 6 months prior to diagnosis and ≤ 6 months of diagnosis as probable screening versus diagnostic indications, respectively.
The most commonly performed procedure was colonoscopy (63%), but most colonoscopic examinations were within 6 months of diagnosis. Only 44% of patients received ≥ 1 procedures > 6 months before diagnosis, with FOBT the most frequent procedure (36%). Colonoscopy was performed in only 6% of patients prior to the 6-month peridiagnostic period. Compared with patients who underwent no procedures or who underwent procedures only within 6 months of diagnosis, those who underwent procedures > 6 months earlier presented with earlier stages of carcinoma.
The receipt of colorectal procedures in the interval > 6 months prior to a diagnosis of colorectal carcinoma was low, suggesting the under use of screening in this population. The association of procedure use with earlier stage at presentation as well as unmeasured benefits in reducing carcinoma incidence suggest beneficial effects of this testing. Cancer 2005. © 2005 American Cancer Society.
Population-based screening for colorectal carcinoma with fecal occult blood testing (FOBT), sigmoidoscopy, colonoscopy, and/or barium enema currently is recommended by all relevant professional societies, including the American Cancer Society, the United States Preventive Services Task Force, the American Gastroenterological Association, and the American College of Gastroenterology.1–4 The goals of screening are to detect premalignant adenomatous polyps as well as presymptomatic carcinomas that presumably are at an earlier stage than neoplasms that produce gastrointestinal or constitutional symptoms.
With the exception of FOBT, which t has been studied in randomized clinical trials,5–9 the data supporting the benefit of screening are obtained from case–control studies.10–13 Those studies generally compared the receipt of a given procedure between patients who had terminal colorectal carcinoma with matched controls who did not have terminal disease but did not evaluate the relative merits of one of the several alternative procedures that are available. In addition, many of the existing studies were not population-based.
Compared with previous studies that included patients who died from colorectal carcinoma and controls without disease, the current study was conducted in a large cohort of patients with colorectal carcinoma who were identified through the population-based Surveillance, Epidemiology, and End Results (SEER) tumor registry. To provide a measure of procedure use prior to the colorectal carcinoma diagnosis, the tumor registry files were linked with health claims data. The objectives of the study were to describe the use of prior procedures in this patient population and to evaluate the association of procedure use with disease stage. In addition, we differentiated procedures that were performed immediately prior to diagnosis from procedures that ere performed earlier and, thus, more likely, were for screening purposes.
MATERIALS AND METHODS
The study utilized the linked SEER-Medicare data base, which was developed in the 1990s as a large, population-based source of information for cancer-related epidemiological and health services research.14, 15 The files represent patients age ≥ 65 years who were diagnosed with cancer and reside in 1 of the geographic areas contained in the SEER registries. Within each registry, approximately 93% of patients age > 64 years are included, with the 7% of patients who are not captured consisting of those not enrolled in Medicare (3%) or patients who receive care exclusively in the Department of Veterans Affairs. The SEER files and Medicare data are linked based on an algorithm involving a match of Social Security number, name, gender, and date of birth. All patient identifiers, such as Social Security number, then are 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, information pertaining to Medicare eligibility and enrollment, and Census tract level sociodemographic factors. Cancer-related variables in PEDSF include demographic characteristics, previous cancer diagnoses, date of cancer diagnosis, and relevant tumor-related data, including disease stage at diagnosis. Medicare-related variables include reason for entitlement, enrollment history in Medicare, and health maintenance organization (HMO) and Medicare Parts A and B coverage. Because individual level socioeconomic status is not available, Census tract level per capita income is used as a surrogate measure.
Procedure use prior to cancer diagnosis was measured using data from Medicare claims. The National Claims History (NCH) files include physician/supplier 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.
The cohort was obtained from patients age ≥ 70 years who were included in the SEER-Medicare linked data base and who were diagnosed with adenocarcinoma of the colon or rectum during 1999. The age restriction was used to limit the study to patients who would have been Medicare-eligible during the 5-year period prior to cancer diagnosis. 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 if they were members of a Medicare-sponsored HMO.
For each patient in the cohort, the following data elements were obtained from the PEDSF: age, gender, race, Census tract per capita income, date of cancer diagnosis, stage of cancer (categorized as carcinoma in situ, local, regional, distant, and unstaged), and location of tumor (categorized as distal colon [rectum, rectosigmoid junction, sigmoid colon, or descending colon] or proximal colon [splenic flexure, transverse colon, hepatic flexure, ascending colon. or cecum]). Medicare claims from 1995 through the date of diagnosis were examined for receipt of procedures of interest. These included FOBT (Current Procedural Terminology, fourth edition [CPT-4] codes 82270, 82273, and G0107); flexible sigmoidoscopy (CPT-4 codes 45330, 45331, 45333, 45338, 45339, and G0104; International Classification of Diseases, ninth revision-clinical modification [ICD-9-CM] codes 45.22, 45.24, 48.22, and 48.24), colonoscopy (CPT-4 codes 44388, 44389, 44392, 44393, 44394, 45378, 45380, 45382, 45383, 45384, 45385, and G0105; ICD-9-CM codes 45.23, 45.25, 45.41, 45.42, 45.43, and 48.36), and barium enema (CPT-4 codes 74270, 74280, G0106, G0120, and G0122; ICD-9-CM code 87.64) The NCH Outpatient and MEDPAR files were combined, and records were unduplicated using the beneficiary identification code, procedure, and date of service, such that a procedure that was documented in both files for the same individual and date of service was counted only once.
The procedure date also was compared relative to the date of diagnosis. To differentiate procedures with a greater likelihood of being performed for evaluation of suspected cancer versus other indications, the dates were divided into before and after 6 months prior to diagnosis. For each procedure, a total of 4 categories were developed: no procedure, procedure only < 6 months prior to diagnosis, procedure only > 6 months prior to diagnosis, and procedures both < 6 months and > 6 months prior to diagnosis.
Differences in disease stage according to patient characteristics was evaluated using chi-square analysis. The association between receipt of procedures and disease stage at diagnosis also was determined. Finally, to determine the independent association of procedure use and disease stage, multivariable logistic regression analysis was used. The dependent variable in the models was advanced-stage carcinoma (regional or distant stage) versus local stage or carcinoma in situ. All analyses were performed using the Statistical Analysis System (version 9.0; SAS Institute, Inc., Cary, NC). The study was approved by the Institutional Review Board at University Hospitals of Cleveland.
In total, 13,165 patients with an initial diagnosis of colorectal carcinoma in 1999 were identified from the linked SEER-Medicare data base, including 9388 patients age ≥ 70 years. Among this group, 2766 patients were excluded because of HMO enrollment, and 396 patients were excluded because of lack of Medicare Part B enrollment.
The mean ± standard deviation age of the cohort was 79 ± 6 years, 55% of patients were female, 84.5% of patients were Caucasian, and 7.2% of patients were African American. The mean Census tract per capita income was $25,993 and was available for 5148 patients. Data on tumor location were available for 5966 patients and included the distal colon or rectum and proximal colon in 50% each. Staging was specified for 5806 patients and included in situ carcinoma (n = 322 patients), local-stage disease (n = 2358 patients), regional-stage disease (n = 2173 patients), and distant-stage disease (n = 953 patients). These 5806 patients were included in the remaining analyses.
The association between patient characteristics and disease stage at presentation is shown in Table 1. African Americans (P < 0.0001) and, to some degree, residents of less affluent Census tracts (P = 0.01) were less likely to present with early-stage carcinoma. Patients with distal colorectal carcinoma rather than proximal colorectal carcinomas also tended to present at a less advanced tumor stage (P < 0.0001). There was no significant association between age or gender and tumor stage.
Table 1. Association of Patient Characteristics and Disease Stage in Patients with Colorectal Carcinoma
|Age (yrs)a|| || || || || |
|Raceb|| || || || || |
|Genderc|| || || || || |
|Per capita incomed|| || || || || |
| First quartile||1176||4.5||39.2||37.8||18.5|
| Second quartile||1214||3.9||40.1||39.4||16.6|
| Third quartile||1212||6.4||42.3||36.5||14.8|
| Fourth quartile||1217||6.4||41.3||36.9||15.5|
|Tumor siteb|| || || || || |
The receipt of procedures prior to a diagnosis of colorectal carcinoma was examined using Medicare claims data (Table 2). One or more procedures were performed in 87.5% of patients. The overall use of colonoscopy was the highest (63%), followed by FOBT (49.1%), sigmoidoscopy (28%), and barium enema (23.1%). However, many of the procedures were performed within 6 months of the diagnosis date, suggesting that the test indication may have been for the diagnosis itself. Only 44% of patients had any test performed at least 6 months prior to diagnosis. The procedure most frequently used in the time interval preceding the 6-month peridiagnostic period was FOBT (35.9%), followed by sigmoidoscopy (9.2%), and barium enema (8.2%). Colonoscopy was the least frequently performed procedure during this period (6.1% of patients).
Table 2. Association of Procedure Use and Disease Stage in Patients with Colorectal Carcinoma
|FOBTa|| || || || || || |
| < 6 mos||765||13.2||5.6||41.7||38.6||14.1|
| > 6 mos||1276||22.0||5.3||43.9||36.3||14.5|
| > 6 mos and < 6 mos||809||13.9||7.4||44.1||37.2||11.3|
|Sigmoidoscopyb|| || || || || || |
| < 6 mos||1091||18.7||6.2||42.2||36.8||14.9|
| > 6 mos||421||7.3||5.0||44.9||36.1||14.0|
| > 6 mos and < 6 mos||115||2.0||6.1||53.0||32.2||8.7|
|Colonoscopya|| || || || || || |
| < 6 mos||3303||56.9||3.8||42.6||40.3||13.3|
| > 6 mos||127||2.2||10.2||42.5||29.9||17.3|
| > 6 mos and < 6 mos||227||3.9||2.2||43.6||42.3||11.9|
|Barium enemaa|| || || || || || |
| < 6 mos||872||15.0||3.6||33.7||43.9||18.8|
| > 6 mos||380||6.5||6.3||46.3||34.0||13.4|
| > 6 mos and < 6 mos||94||1.6||4.3||42.6||31.9||21.3|
|Any procedurea|| || || || || || |
| < 6 mos||2521||43.4||4.4||39.4||39.6||16.6|
| > 6 mos||381||6.6||8.9||39.6||29.9||21.5|
| > 6 mos and < 6 mos||2179||37.5||5.3||44.4||38.0||12.4|
The stage of colorectal carcinoma at diagnosis was compared between patients with and without each of the procedures of interest stratified according to the date of the procedure in relation to carcinoma diagnosis date (Table 2). Compared with patients who had no previous procedures, receipt of each of the four procedures of interest was associated significantly with an earlier stage of disease. Although differences in staging appeared to be present regardless of the timing of the procedure, the greatest proportion of local-stage and in situ carcinomas were observed in patients who underwent the procedure > 6 months prior to diagnosis (with or without a procedure < 6 months from diagnosis as well).
The independent association of procedure use and disease stage was determined using a series of multivariable logistic regression models. Presentation with regional or distant-stage carcinoma was associated inversely with the receipt of FOBT > 6 months before diagnosis (odds ratio [OR] 0.78; 95% confidence interval [95% CI], 0.70–0.87; P < 0.0001), flexible sigmoidoscopy > 6 months prior to diagnosis (OR, 0.74; 95% CI, 0.62–0.89; P = 0.001), or barium enema > 6 months prior to diagnosis (OR, 0.78; 95% CI, 0.64–0.94; P = 0.01). The association of colonoscopy > 6 months before diagnosis with regional or distant-stage disease did not achieve statistical significance (OR, 0.86; 95% CI, 0.69–1.07; P = 0.18). The only other factor in multivariable models that was associated consistently with regional or distant-stage tumors was disease site, with patients who had distal carcinoma less likely to present with advanced-stage disease (OR, 0.74 in all models; 95% CI, 0.67–0.84; P < 0.0001). The other variables, including age, race, and per capita income, were not associated with disease stage.
Although colorectal carcinoma screening in the asymptomatic population is recommended widely, data supporting its efficacy generally are obtained from case–control studies utilizing health plan data,10–12 administrative data,13 or from studies with historic controls.16 In addition, epidemiologic data that report a reduction in colorectal carcinoma mortality rates since the late 1980s are consistent with the increased use of screening.17 Using a linked tumor registry and a health claims data base, we identified patients with incident colorectal carcinomas and studied the use of potential screening procedures in the prediagnosis period. Consistent with the under use of screening procedures in the general population,18 we found that < 50% of patients received a procedure > 6 months prior to diagnosis. Most of the procedures that preceded the 6-month prediagnosis period were FOBT, which is the least invasive and least costly screening test but also the least sensitive screening test.2 In contrast, colonoscopy, which is the most accurate screening test, ostensibly was performed during a potential screening period in only a minority of patients. Thus, almost 94% of patients with colorectal carcinoma did not undergo colonoscopy at all or only underwent a procedure that led to the diagnosis of colorectal carcinoma. Conversely, it is unclear whether the approximately 6% of patients who did undergo a prior colonoscopy had lesions that were missed on the index examination, had incompletely removed lesions, or had rapidly growing tumors.
However, compared with patients who had no testing or who had testing within 6 months of diagnosis, those patients who underwent procedures prior to 6 months before diagnosis were more likely to present with earlier stage tumors. The data also indicate a lower risk of advanced-stage carcinoma with distal disease, as opposed to proximal disease, and may reflect the increased likelihood that patients with these tumors will present with symptoms such as rectal bleeding and changes in bowel pattern.
The current findings are consistent with randomized clinical trials of FOBT that reported an earlier stage at diagnosis in patients with colorectal carcinoma who underwent regular screening.5–9 In addition, in case–control studies of patients with terminal colorectal carcinoma and at-risk control patients without terminal colorectal carcinoma, a lower prior use of screening procedures was noted (both FOBT and sigmoidoscopy) in patients with carcinoma10–13; and, for sigmoidoscopy, an association was noted that was limited to carcinomas within reach of a sigmoidoscope.11, 12 In another study that used claims data, the use of endoscopic procedures (sigmoidoscopy, colonoscopy, and polypectomy) was compared between patients with colorectal carcinoma and disease-free controls who received care through the Department of Veterans Affairs.13 Patients with carcinoma were approximately 50% less likely than controls to receive a procedure prior to their diagnosis of carcinoma. In all case–control studies, the reported overall use of procedures was low, even in disease-free controls, with 10-year rates no higher than 29%.10–13 In the current study, the most frequently recorded procedure in the interval > 6 months prior to diagnosis was FOBT, but it was performed in only 36% of patients. These findings emphasize the relative under use of colorectal carcinoma screening in the United States.18
We recognize several relevant limitations of the current study. The most important limitation relates to the selection of patients for the study. In contrast to case–control studies that included patients with colorectal carcinoma and disease-free controls,10–13 the current analysis was limited to patients with colorectal carcinoma. In fact, the earlier disease stage at diagnosis in patients who had undergone previous procedures may have been a result of screening,19, 20 which is analogous to a lead-time bias. Future studies should include a control group of Medicare beneficiaries without cancer, which may provide a more accurate assessment of the potential benefit of screening. In addition, the receipt of procedures was measured through Medicare claims data, which are collected for billing (and not research) purposes. Although the accuracy of all of the codes have not been evaluated, in general, it is believed that procedural coding is more reliable than diagnosis coding.21–23 Moreover, prior to 1998, reimbursement for routine colorectal carcinoma screening in Medicare beneficiaries was not provided. In the current analysis, we did not differentiate procedures according to indication (diagnostic vs. screening); thus, large numbers of tests may have been ordered for the evaluation of signs or symptoms. Because of a lack of reimbursement, a subset of procedures actually performed may not have been billed, especially for inexpensive tests, such as FOBT. It is unlikely, however, that significant numbers of more costly procedures, such as colonoscopy, would have been performed without claims submitted. An analysis using more recent claims (when available) could be used to study procedure use in the period of reimbursement for routine colorectal screening. Another limitation is that certain screening procedures, especially colonoscopy, are designed primarily to remove precursor lesions rather than to detect cancer. Thus, the potential benefits of procedures in reducing the incidence of colorectal carcinoma cannot be quantitated with the current data set. Although the sample size was relatively large, the power to detect significant differences in certain subsets nonetheless was limited. For example, because only a small number of patients underwent colonoscopy > 6 months prior to diagnosis, the use of colonoscopy and disease stage typically did not achieve statistical significance. Another limitation is that unmeasured demographic or clinical factors may have confounded the association between disease stage and receipt of testing. Finally, the generalizability of the current findings to younger patients and/or to non-Medicare beneficiaries is unknown.
The results of the current study have highlighted the under use of colorectal screening procedures in clinical practice, which is relevant given the high incidence and mortality of colorectal carcinoma in the United States.24 In addition, the association between colorectal screening procedure use and earlier disease stage at diagnosis is consistent with the potential benefits of screening.