SEARCH

SEARCH BY CITATION

Keywords:

  • Barrett esophagus;
  • endoscopy;
  • esophageal neoplasms;
  • stomach neoplasms

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

BACKGROUND

Barrett esophagus, a consequence of chronic gastroesophageal reflux disease, is a premalignant condition for adenocarcinoma of the esophagus and, possibly, the gastric cardia. However, the actual use and clinical impact of upper gastrointestinal endoscopy in screening and surveillance for Barrett esophagus are unknown.

METHODS

A cohort included 1633 patients with adenocarcinoma (777 esophagus, 856 cardia) who were 70 years or older. They were diagnosed between 1993 and 1996 and were identified from the Surveillance, Epidemiology and End Results program registry. All claims for upper endoscopy and a diagnosis of Barrett esophagus from 1991 through 1 year before diagnosis were identified from linked Medicare files.

RESULTS

One or more upper endoscopies before diagnosis were performed in 9.7% of patients (13.0% esophagus, 6.8% cardia) and a diagnosis of Barrett esophagus was present in only 3.7% of patients. A shift toward earlier stage at diagnosis was observed in patients with previous endoscopy or Barrett diagnosis. For example, 62% of patients with esophageal and 49% of patients with cardia tumors who underwent previous endoscopy presented with in situ or local stage carcinoma, compared with 35% and 27% of other patients, respectively. Receipt of endoscopy was also associated with a reduced risk of death for esophageal adenocarcinoma (relative hazard 0.73, 95% confidence interval 0.57–0.93; P = 0.01), but not for adenocarcinoma of the cardia.

CONCLUSIONS

Receipt of upper endoscopy at least 1 year before diagnosis of adenocarcinoma, which may reflect prediagnosis screening, was associated with an earlier tumor stage and improved survival. These data support the role of endoscopic screening and surveillance for Barrett esophagus and highlight the underdiagnosis of populations at risk. Cancer 2002;95:32–8. © 2002 American Cancer Society.

DOI 10.1002/cncr.10646

Adenocarcinoma of the esophagus is a neoplasm that is increasing in incidence, particularly among middle-aged Caucasian males, and is believed to represent a consequence of chronic gastroesophageal reflux disease (GERD).1–4 Specifically, it is hypothesized that development of intestinal metaplasia in the tubular esophagus, or Barrett esophagus, is a premalignant condition that results from chronic acid exposure.4 The relative risk of development of esophageal adenocarcinoma in patients with Barrett metaplasia has been estimated to be 30–40-fold,4–6 but this estimation has been questioned because most cases of Barrett esophagus are not detected clinically.7 In addition, a population-based study from Scandanavia established an association between frequency and severity of GERD symptoms and the development of esophageal adenocarcinoma.8 In addition to esophageal carcinoma, there is also a potential association of Barrett metaplasia with adenocarcinoma of the gastric cardia, another neoplasm that has been increasing in incidence.2, 3, 9

Given the rising incidence of carcinoma, increasing attention has been devoted to the role of population-based surveillance programs among patients with known Barrett metaplasia. Patients with chronic GERD are also being screened for the presence of Barrett epithelium. There are approximately 10 million persons in the United States older than 50 with chronic GERD10 and approximately 700,000 adults are estimated to have Barrett esophagus.11 Therefore, the clinical and economic implications are enormous.

Current recommendations from professional societies specify that regular endoscopic surveillance should generally be performed in patients with known Barrett metaplasia.12 However, despite the rationale for endoscopic biopsies to detect dysplasia or unrecognized carcinoma in patients with known Barrett metaplasia, there are no prospective controlled studies that support the benefit of surveillance in improving life expectancy. Rather, the data that support routine surveillance are from studies that compare stage of disease or survival in patients with esophageal adenocarcinoma that is detected by screening versus symptoms.13–15 A major limitation of these studies, however, is that data were typically obtained from single centers that often served as referral sites. Similar data from patients treated in community practice are lacking. These data would be important to better estimate the adherence to published guidelines and quantitate the economic impact of this diagnosis. Effectiveness data are also needed to better justify the enormous potential costs and morbidity associated with targeted surveillance.

The goals of this study were to define the frequency of screening or surveillance endoscopy before a diagnosis of adenocarcinoma of the esophagus or gastric cardia in a Medicare enrolled population, as well as to compare outcomes after carcinoma diagnosis of patients who received endoscopy versus other patients. The study used data from the population-based Surveillance, Epidemiology and End Results (SEER) tumor registries and included Medicare health claims data to capture procedures performed in the prediagnosis phase.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Patients

The cohort for the study was obtained from a unique merged Medicare-SEER database that was developed jointly by the National Cancer Institute and the Centers for Medicare and Medicaid Services (formerly the Health Care Financing Administration) to conduct cancer-related health services research.16 As previously described, the joint database includes all patients older than 64 years who were diagnosed with cancer and resided in one of the ten SEER areas (Atlanta, Detroit, Los Angeles, Seattle-Puget Sound, San Francisco-Oakland, Connecticut, Hawaii, Iowa, New Mexico, Utah).16 Attempts were made to link all patients in the SEER database with Medicare data using patient identifiers (social security number, name, gender, and dates of birth and death). Ninety-four percent of the SEER cases were successfully matched.16 The primary sources of SEER data are medical records from hospitals, with additional cases ascertained from pathologists, oncologists, and radiotherapists.

Medicare Part B claims include physician and outpatient data and are contained in two sources: 1) the Outpatient Standard Analytic Files (SAF) for outpatient hospital procedures and 2) the National Claims History (NCH) files, which contain all physician and supplier billed procedures. Part B files include procedures coded by the Current Procedure Terminology, 4th edition (CPT-4).

The cohort included all patients aged 70 and older in the SEER-Medicare database with adenocarcinoma of the esophagus or gastric cardia diagnosed between 1993 and 1996 according to the SEER registry records. Patients were excluded for the following reasons: enrollment in a Medicare Health Maintenance Organization, which accounted for less than 5% of enrollees during that time interval (because of incomplete claims); lack of enrollment in Medicare Part B (because outpatient claims were required); and histology not specified or other than adenocarcinoma. In addition, for comparisons of stage at diagnosis or survival time, patients with unstaged tumors were also excluded.

Measures

All claims from the Outpatient SAF and NCH files from 1991 through 1 year before diagnosis were examined and all claims with procedure codes for diagnostic endoscopic procedures (CPT-4 43200, 43202, 43235, 43239) were extracted. In addition, we obtained claims with a diagnosis consistent with Barrett esophagus (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM] code 530.2, which can indicate esophageal ulcer as well as Barrett).

Data from the SEER files included date of diagnosis, histology, TNM stage and grade of tumor, the most invasive treatment given within the first 4 months after diagnosis (i.e., patients who received both endoscopic prosthesis and surgical resection were coded as resection), and survival through the end of 1998. The SEER files did not include data on comorbid conditions, which may be independently associated with treatment decisions and outcomes. Thus, the Medicare claims were searched for ICD-9-CM codes that indicated comorbidities, including chronic ischemic heart disease (ICD-9-CM 411–414.9), congestive heart failure (ICD-9-CM 398.91, 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, 428), chronic pulmonary disease (ICD-9-CM 490–496.99), chronic liver disease (ICD-9-CM 571–572.99), chronic renal failure (ICD-9-CM 581–583.9, 585, 586), and cerebrovascular disease (ICD-9-CM 430–438.99).

Analysis

The Outpatient SAF and NCH files from 1991 through 1 year before carcinoma diagnosis were examined and the proportion of patients with one or more claims for upper gastrointestinal endoscopy as well as those for whom a potential diagnosis of Barrett esophagus was listed was determined. The 1-year window was selected to avoid including procedures that were performed for the diagnosis of adenocarcinoma rather than for screening or surveillance as well as symptom assessment. All analyses were stratified by site of tumor (esophagus vs. cardia).

Using data from the SEER files, comparisons of stage at diagnosis were determined for patients with one or more Medicare codes for upper endoscopy at 1 or more years before diagnosis and for patients without endoscopy. In addition, stage of disease was compared between patients with a diagnosis code consistent with Barrett esophagus and other patients. All comparisons used chi-square statistics, including the Cochran-Mantel-Haenszel chi-square statistics for trend.

Survival time from date of diagnosis to date of death or last follow-up was determined through the SEER database. The comparisons of survival time between patients with and without previous endoscopy, as well as patients with a previous diagnosis consistent with Barrett esophagus and others, were carried out using a series of multivariable Cox proportional hazards models adjusted for potentially confounding factors including age, gender, and comorbid conditions.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

A total of 1633 patients aged 70 years and older with adenocarcinoma of the esophagus (n = 777) and cardia (n = 856) were identified from the linked SEER-Medicare database. The median age of the participants was 76 years, 76% of patients were male, and 5.1% were African American. The most common comorbidities were chronic ischemic heart disease (38%), chronic pulmonary disease (33%), and cerebrovascular disease (19%). Chronic renal failure (3.6%), chronic liver disease (1.8%), and congestive heart failure (1.2%) were noted less frequently. Forty percent of patients had no comorbidities listed, 32% had one major comorbidity, 19% had two major comorbidities, and 9% had three or more major comorbidities. Staging was performed in 1256 patients and included carcinoma in situ (1.9%), local stage (31%), regional stage (32%), and distant stage (34%) carcinomas.

In the cohort, one or more upper endoscopies were performed at least 1 year before carcinoma diagnosis in only 9.7% of patients, including 13% of patients with esophageal adenocarcinoma and 6.8% with adenocarcinoma of the cardia (P < 0.001 for comparison of esophageal and cardia tumors). A diagnosis code consistent with Barrett esophagus was present at least 1 year before cancer diagnosis in only 3.7% of patients, including 7.0% with esophageal and 0.7% with cardia adenocarcinoma (P < 0.001). Both a previous endoscopy and a diagnosis consistent with Barrett esophagus were found in only 3.7% of patients. The rates of upper endoscopy and a diagnosis consistent with Barrett in the time period 6 months or more before carcinoma diagnosis were only 11.5% and 4.2%, respectively.

The tumor stage according to receipt of endoscopy and diagnosis consistent with Barrett esophagus 1 year or more before carcinoma diagnosis is shown in Figure 1. For both adenocarcinoma of the esophagus and cardia, an earlier stage at diagnosis was observed in patients with previous endoscopy or Barrett diagnosis. For example, 62% of patients with esophageal and 49% of patients with cardia tumors who underwent previous endoscopy presented with in situ or local stage carcinoma, compared with 35% and 27% of other patients, respectively. The differences were all statistically significant with the exception of a previous diagnosis consistent with Barrett esophagus in patients with adenocarcinoma of the cardia, which was present in only sixpatients.

thumbnail image

Figure 1. The association of tumor stage at presentation with upper endoscopy (esophagogastroduodenoscopy disease [EGD]) performed 1 year or more before carcinoma diagnosis and a diagnosis consistent with Barrett esophagus (BE) 1 year or more before carcinoma diagnosis. For esophageal adenocarcinoma, both previous endoscopy (A) and Barrett diagnosis (B) were associated with earlier stage at presentation (P < 0.001 for both). For adenocarcinoma of the cardia, previous endoscopy (C; P = 0.009) but not Barrrett diagnosis (D; P = 0.60) was associated with earlier stage at presentation.

Download figure to PowerPoint

Surgical resection was performed in 28% of patients with esophageal adenocarcinoma and in 43% of patients with adenocarcinoma of the cardia. Surgical resection was more common among esophageal carcinoma patients with a previous diagnosis consistent with Barrett esophagus than in patients without the diagnosis (44% vs. 26%, P < 0.005). The frequency of surgery did not differ between patients who underwent previous endoscopy and other patients (34% vs. 27%, P = 0.14 for esophagus; 53% vs. 43%, P = 0.11 for cardia).

The overall survival rate was 16.3%, including 15.3% of esophageal carcinoma patients and 17.2% of patients with cardia carcinomas. Among patients with esophageal adenocarcinoma, the median survival time was 5 months for patients without previous endoscopy and 7 months for patients with endoscopy. The corresponding survival times for cardia tumors were 6 months and 8 months, respectively. Esophageal adenocarcinoma patients without a previous diagnosis consistent with Barrett esophagus had a median survival time of 5 months, compared with 7 months for patients with a Barrett diagnosis. These differences, although modest, were all statistically significant (P < 0.01). There was no difference in survival time for patients with adenocarcinoma of the cardia when stratified by previous Barrett diagnosis (6 months in both groups).

In a multivariable Cox proportional hazards model that adjusted for age at diagnosis, gender, race, and number of comorbidities, the association of previous endoscopy and previous Barrett diagnosis with survival time was assessed. Receipt of endoscopy 1 year or more before diagnosis was associated with a reduced risk of death for esophageal adenocarcinoma (relative hazard [RH] = 0.73; 95% confidence interval [CI] = 0.57–0.93; P = 0.01), but not for adenocarcinoma of the cardia (RH = 0.87; 95% CI = 0.63–1.20; P = 0.87). Similarly, a previous diagnosis consistent with Barrett esophagus was associated with a lower risk of death for esophageal carcinoma (RH = 0.61; 95% CI = 0.43–0.85; P = 0.004), but not for adenocarcinoma of the cardia (RH = 0.79; 95% CI = 0.29–2.11; P = 0.63). In an analysis that was restricted to patients who underwent surgical resection for combined esophageal and cardia tumors, previous endoscopy (RH = 0.70; 95% CI = 0.48–0.99; P = 0.05) but not a previous diagnosis consistent with Barrett esophagus (RH = 0.66; 95% CI = 0.38–1.12; P = 0.12) was associated with a decreased risk of death.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Because the incidence of adenocarcinoma of both the esophagus and gastric cardia has increased in the United States over the past several years,1–3 the role of screening and surveillance for the premalignant condition, Barrett esophagus, has been of great interest. In the current study, which was based on data from the population-based SEER registries and linked Medicare claims, patients who underwent previous endoscopy presented at an earlier stage of cancer. In addition, for esophageal, but not cardia carcinomas, a previous diagnosis consistent with Barrett esophagus, and presumably recognition of increased cancer risk, was also associated with earlier stage of disease. Among esophageal carcinoma patients with previous endoscopy and/or Barrett diagnosis, both unadjusted and adjusted survival rates were also improved. The data confirmed findings from smaller, single-center studies13–15 and highlight the potential importance of endoscopic screening and surveillance.

The association of endoscopy with improved cancer staging and survival was more significant for esophageal than for cardia carcinomas. Only six patients with adenocarcinoma of the cardia had diagnosis codes consistent with Barrett esophagus. Other investigators have also found the association of chronic GERD symptoms and adenocarcinoma of the cardia to be less pronounced than with carcinomas of the esophagus.8, 9 It is likely that other factors besides GERD and Barrett esophagus are involved in the pathogenesis of these tumors. However, if a patient with known chronic GERD and/or Barrett esophagus undergoes upper endoscopy, visualization of the cardia and any suspicious lesions would expected to be undertaken as well.

In addition to the current study, two other recent studies have also examined the association of endoscopic screening and surveillance with cancer outcome. Corley et al.17 studied a cohort of 589 patients with adenocarcinoma of the esophagus or cardia who were diagnosed between 1990 and 1997 and who were members of a large health plan. As in the current study, the frequency of previous endoscopy (10.9% at least 6 months before diagnosis) and Barrett diagnosis (3.9%) was low. Through medical record review, the investigators also determined that although only 15 patients were enrolled in an endoscopic surveillance program, all carcinomas detected in this group were early stage and, in contrast to our findings, the long-term survival was significantly improved (odds ratio = 30). Alternatively, if a patient with known Barrett esophagus was not surveyed until symptoms developed, the carcinomas were detected at later stages. Kearney et al.18 conducted a case-control study of the Veterans Administration population. They compared 245 patients who died of adenocarcinoma of the esophagus or gastric cardia with 980 matched controls with uncomplicated GERD. Cases with carcinoma were 34% less likely to have undergone upper endoscopy when compared with controls and the protective association was present for endoscopies performed up to 8 years before diagnosis. These two studies, although conducted in different populations, help to substantiate our findings.

Although there is increasing evidence about the potential benefits of endoscopic screening, current efforts are directed toward development of less invasive and/or more accurate means of performing Barrett screening and surveillance. The use of ultrathin endoscopes that can be passed transnasally and without sedation may represent a more feasible screening method.19, 20 Methylene blue staining of the esophagus,21 use of balloon cytology,22 and flow cytometric analysis of DNA from biopsy specimens23 may also provide more accurate assessment of dysplastic tissue. In addition, the optimal frequency of endoscopic surveillance is also debated, with economic models suggesting longer intervals11 and reports of unrecognized carcinoma and high-grade dysplasia favoring shorter intervals.24

There are some relevant limitations to the study. First, because the ICD-9-CM code for Barrett esophagus is also used for esophageal ulcer, the clinical or pathologic recognition of this diagnosis before the carcinoma diagnosis, as well as the presence or absence of dysplasia, could not be definitively measured. However, the use of these codes suggested the recognition of at least severe esophagitis. Second, we were unable to ascertain the indication for upper gastrointestinal endoscopy and it is likely that a subset of examinations were performed for symptom assessment rather than Barrett screening or surveillance. Third, the accuracy of diagnostic and procedural coding of outpatient Medicare claims has been studied on only a limited basis.25, 26 However, where studied, the completeness of procedural claims, such as endoscopy, was typically very high.27–30 The accuracy of the SEER registry data for case ascertainment and tumor staging has been extremely high. Fourth, patients who have undergone previous endoscopic procedures may have appeared to have better survival because of length time bias or lead time bias.31 However, the exclusion of procedures performed within 1 year of diagnosis would be expected to minimize these potential biases. Fifth, because of the databases that were used, the study was limited to an older population. It is likely that larger differences in survival would have been observed in younger patients who underwent endoscopic screening or surveillance.

In summary, this study has provided additional evidence about the potential benefits of endoscopic screening in the early detection of esophageal adenocarcinoma and to a lesser degree, adenocarcinoma of the cardia. The findings also emphasize that the majority of patients at risk do not undergo endoscopy. Further studies should define other factors that will be useful in selecting patients at highest risk for development of carcinoma based on the nature of their GERD symptoms, and other recognized risk factors, and who would derive the maximum benefit from screening and surveillance.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES
  • 1
    RiesLAG, EisnerMP, KosaryCL, et al., editors. SEER cancer statistics review, 1973–1997. Bethesda: National Cancer Institute, 2000.
  • 2
    Blot WJ, Devesa SS, Kneller RW, Fraumeni JF, Jr. Rising incidence of adenocarcinoma of the esophagus and gastric cardia. J Am Med Assoc. 1991; 265: 12871289.
  • 3
    Pera M, Cameron AJ, Trastek VF, Carpenter HA, Zinsmeister AR. Increasing incidence of adenocarcinoma of the esophagus and esophagogastric junction. Gastroenterology. 1993; 104: 510513.
  • 4
    Spechler SJ. Esophageal columnar metaplasia (Barrett's esophagus). Gastrointest Endosc Clin North Am. 1997; 7: 118.
  • 5
    Spechler SJ, Goyal RK. Barrett's esophagus. N Engl J Med. 1986; 315: 362371.
  • 6
    Cameron AJ. Epidemiology of columnar-lined esophagus and adenocarcinoma. Gastroenterol Clin North Am. 1997; 26: 487494.
  • 7
    Cameron AJ, Zinsmeister AR, Ballard DJ, Carney JA. Prevalence of columnar-lined (Barrett's esophagus): comparison of population-based clinical and autopsy findings. Gastroenterology. 1990; 99: 918922.
  • 8
    Lagergren J, Bergstrom R, Lindgren A, Nyren O. Symptomatic gastroesophageal reflux as a risk factor for esophageal adenocarcinoma. N Engl J Med. 1999; 340: 825831.
  • 9
    Weinstein WM. Precursor lesions for cancer of the cardia. Gastrointest Endosc Clin North Am. 1997; 7: 1928.
  • 10
    Provenzale D, Homan RK. Screening for Barrett's esophagus in patients with GERD: can we afford the cost? [abstract]. Am J Gastroenterol. 2001; 96: S278.
  • 11
    Provenzale D, Kemp JA, Arora S, Wong JB. A guide for surveillance of patients with Barrett's esophagus. Am J Gastroenterol. 1994; 89: 670680.
  • 12
    American Society for Gastrointestinal Endoscopy. The role of endoscopy in the surveillance of premalignant conditions of the upper gastrointestinal tract. Gastrointest Endosc. 1998; 48: 663668.
  • 13
    Streitz JM, Jr., Andrews CW, Jr., Ellis FH, Jr., Endoscopic surveillance of Barrett's esophagus: does it help? J Thorac Cardiovasc Surg. 1993; 105: 383388.
  • 14
    Roberston CS, Mayberry JF, Nicholson DA, James PD, Atkinson M. Value of endoscopic surveillance in the detection of neoplastic change in Barrett's oesophagus. Br J Surg. 1988; 75: 760763.
  • 15
    Altorki NK, Skinner DB. Adenocarcinoma in Barrett's esophagus. Semin Surg Oncol. 1990; 6: 274278.
  • 16
    Potosky AL, Riley GF, Lubitz JD, Mentnech RM, Kessler LG. Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care. 1993; 31: 732748.
  • 17
    Corley DA, Levin TR, Habel L, Weiss NS, Buffler P. Surveillance and survival in Barrett's adenocarcinoma: a population-based study. Gastroenterology. 2002; 122: 633640.
  • 18
    Kearney DJ, Crump C, Maynard C, Boyko E. A case-control study of endoscopy and mortality from adenocarcinoma of the esophagus or gastric cardia for persons with chronic gastroesophageal reflux disease [abstract]. Gastroenterology. 2001; 120: A-414.
  • 19
    Catanzaro AT, Faulx AL, Pfau PR, et al. Accuracy of a narrow diameter battery powered endoscope in sedated and unsedated patients. Gastrointest Endosc. 2002; 55: 484487.
  • 20
    Faulx AL, Catanzaro AT, Zyzanski S, et al. Patient tolerance and acceptability of unsedated ultrathin esophagoscopy. Gastrointest Endosc. 2002; 55: 620623.
  • 21
    Canto MIF, Setrakian S, Petras RE, Blades E, Chak A, Sivak MV Jr. Methylene blue selectively stains intestinal metaplasia in Barrett's esophagus. Gastrointest Endosc. 1996; 44: 17.
  • 22
    Falk GW, Chittajallu R, Goldblum JR, et al. Surveillance of patients with Barrett's esophagus for dysplasia and cancer with balloon cytology. Gastroenterology. 1997; 112: 17871797.
  • 23
    Reid BJ, Blount PL, Rubin CE, Levine DS, Haggitt RC, Rabinovitch PS. Flow-cytometric and histological progression to malignancy in Barrett's esophagus: prospective endoscopic surveillance of a cohort. Gastroenterology. 1992; 102: 12121219.
  • 24
    Sharma P, Weston A, Falk G, Johnston M, Reker D, Sampliner R. Can two upper endoscopies negative for dysplasia eliminate the need for future surveillance in patients with Barrett's esophagus [abstract]. Am J Gastroenterol. 2001; 96: S36.
  • 25
    Cooper GS, Yuan Z, Stange KC, Dennis LK, Amini SB, Rimm AA. The sensitivity of Medicare claims data for case ascertainment of six common cancers. Med Care. 1999; 37: 436444.
  • 26
    Javitt JC, McBean AM, Sastry SS, DiPaolo F, Jr. Accuracy of coding in Medicare Part B claims: cataract as a case study. Arch Ophthalmol. 1993; 111: 605607.
  • 27
    Fisher ES, Whaley FS, Krushat WM, et al. The accuracy of Medicare's hospital claims data: progress has been made, but problems remain. Am J Public Health. 1992; 82: 243248.
  • 28
    Romano PS, Luft HS. Getting the most out of messy data: problems and approaches for dealing with large administrative data sets. In: GradyML, SchwartzHA. Medical effectiveness research data methods. Rockville, MD: U.S. Public Health Service, 1992.
  • 29
    Cooper GS, Chak A, Lloyd LE, Hammar PJ, Harper DL, Rosenthal GE. The accuracy of diagnosis and procedural codes for patients with upper gastrointestinal hemorrhage. Gastrointest Endosc. 2000; 51: 423426.
  • 30
    Cooper GS, Yuan Z, Stange KC, Dennis LK, Amini SB, Rimm AA. Agreement of Medicare claims and tumor registry data for assessment of cancer related treatment. Med Care. 2000; 38: 411421.
  • 31
    Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology: a basic science for clinical medicine (2nd edition). Boston: Little Brown, 1991.