Review: Fecal occult blood test screening reduces colorectal cancer mortality
ACP J Club. 1999 Jan-Feb;130:13. doi:10.7326/ACPJC-1999-130-1-013
Towler BP, Irwig L, Glasziou P, Weller D, Kewenter J. Screening for colorectal cancer using the faecal occult blood test, Hemoccult. Cochrane Review, latest version 16 Jan 1998. In: The Cochrane Library. Oxford: Update Software.
Towler B, Irwig L, Glasziou P, et al. A systematic review of the effects of screening for colorectal cancer using the faecal occult blood test, Hemoccult. BMJ. 1998 Aug 29;317: 559-65.
Does screening using Hemoccult (a fecal occult blood test) reduce colorectal cancer mortality? How often do adverse effects occur?
Studies were identified from the references of a 1995 review done by some of the authors and a MEDLINE search (1993 to January 1997) to identify studies by the authors of trials included in the 1995 review. Additional searches were done using MEDLINE (1990 to 1996) with the terms colorectal neoplasms, screening, randomized controlled clinical trial, and controlled clinical trial; Current Contents (January 1996 to January 1997) with the term colorectal; and the Cochrane Controlled Trials Register (January 1997) with the term colorectal. Bibliographies of relevant papers were also reviewed.
Studies were selected if they were controlled trials of screening for colorectal cancer using Hemoccult.
Data were extracted on study methods, participants, interventions, and outcomes. The main outcome was colorectal cancer mortality.
4 randomized and 2 nonrandomized controlled trials were identified. They involved annual or biennial screening of approximately 442 000 persons ≥ 40 years of age from Denmark, England, France, Sweden, and the United States. Follow-up ranged from 7.8 to 13 years. Meta-analysis of intention-to-treat data from 5 trials using both random- and fixed-effects models showed that patients who were screened had lower colorectal cancer mortality than those who were not screened (Table). Meta-analysis of data from only the 4 randomized controlled trials showed similar reductions. The rate of major complications from colonoscopy ranged from 12 to 30 per 10 000 procedures.
Fecal occult blood test screening using the Hemoccult test reduces colorectal cancer mortality.
Source of funding: No external funding.
For correspondence: Dr. B.P. Towler, Department of Public Health and Tropical Medicine, James Cook University, Townsville, 23 Forth Street, Mackay, Queensland, 4740 Australia. FAX 61-7-4957-5851.
Table. Hemoccult test screening vs no screening for colorectal cancer at 7.8 to 13 years*
|Outcome||Hemoccult||No screening||RRR (95% CI)||NNT (CI)|
|Colorectal cancer mortality||0.50%||0.58%||16% (8 to 23)||1237 (782 to 2961)|
*Abbreviations defined in Glossary; RRR, NNT, and CI calculated from data in article.
This systematic review by Towler and colleagues displays many of the strengths of the Cochrane Library and ACP Journal Club. The best evidence on an important question is brought together where everyone in the world can find it. The authors have carefully searched the world's literature, reassuring readers that all strong evidence is included. They take a modern view of screening by summarizing both benefits and harms.
But the conclusion, that fecal occult blood test screening prevents colorectal cancer deaths, is no surprise. There are, after all, only 3 completed randomized controlled trials to keep track of, each showing effectiveness beyond chance, and they are the basis of current guidelines. Moreover, pooling results into a summary effect size of 16%, although technically possible, takes us in the wrong direction. The studies differ from each other in many ways that would be expected to affect the results: the frequency of screening, whether specimens were rehydrated, the duration of follow-up, and even the nature of the comparison group (in the New York study, controls had screening sigmoidoscopies). Thus, the studies address substantially different questions. It is not reassuring that the results of a statistical test for heterogeneity were negative because the test has little statistical power, especially with only these few trials. Use of a random-effects model does not correct the problem, it only acknowledges that the study results might differ for reasons other than chance.
Thus, the meta-analysis part of this systematic review does not add value. We can learn more about the important clinical policy issues, such as how often to screen and whether to rehydrate, by trying to understand the differences in results, not by pooling them.
Robert H. Fletcher, MD, MSc
Harvard UniversityBoston, Massachusetts, USA