Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Occult blood tests for colorectal cancer

Meta-analysis
Results
Comment

People get hot under the collar when it comes to screening, and screening for cancer in particular. Two general criticisms are often made of screening trials. First that the design of many studies was compromised, resulting in possible bias, with better studies giving less encouraging results (as for breast cancer screening in Bandolier 72). The other is that results of screening are provided in terms of death reduction for the cancer being screened, not all cause mortality.

For instance, a Cochrane review [1] of occult blood testing for colorectal cancer screening found that biannual occult blood screening reduced colorectal cancer deaths by about 20%, preventing about one death per year per 10,000 people. Comments on that review include the criticism of the failure to analyse overall deaths, but that has now been done [2], and provides interesting reading.


Meta-analysis

The original Cochrane review included four randomised trials, but did not report overall deaths. Two have now published more follow up, allowing the analysis to be done.


Results

Three of the four trials in the original review provided data, on 245,000 people, with 2.8 million years of follow up, and using biannual screening. There were 2,148 colorectal cancer deaths, and 65,000 deaths in total.

The death rate from colorectal cancer was about 1 in 100 people over the whole period, or 1 in 1,250 per year. As in the Cochrane review, colorectal cancer deaths were reduced with screening, though the absolute effect was small, almost 10,000 people needing to be screened for one year to prevent a single colorectal cancer death. Table 1 shows the analysis as per patient, and per patient year.



Table 1: Meta-analysis of colorectal cancer deaths and death from all causes, with biannual occult blood screening



Number of
Percent deaths
Analysis
Deaths
Patients/
patient years
Screening
Control
Relative Risk
(95% CI)
NNTp
Colorectal cancer deaths
Per patient
2,148
245,217
0.82
0.94
0.87 (0.80 to 0.95)
830 (520 to 2,200)
Per patient per year
2,148
2,757,795
0.083
0.073
0.87 (0.80 to 0.95)
9,400 (5,800 to 25,000)
Deaths from all causes
Per patient
64,949
245,217
26.5
26.5
1.00 (0.99 to 1.01)
not calculated
Per patient per year
64,949
2,757,795
2.36
2.35
1.00 (0.98 to 1.02)
not calculated


The death rate from all causes was 1 in 4 over the whole period, about 1 in 40 per year. Neither analysis by patient nor by per patient year showed any difference between the screened and the control population in terms of overall mortality.


Comment

The corollary of all this was that screened persons died more often from other causes, significantly so. How could such a result be possible? It is unlikely that biannual occult blood testing would, in itself, be a cause of death.

The most obvious point is that only 1 death in 30 was a colorectal cancer death. Moreover, the difference between screened and non-screened people was only 1 death in every 300 total deaths. How likely is it, then, that a difference this small would be seen in an analysis of overall mortality. The answer is that it is vanishingly small, even with large numbers of patients observed over many years; it would be washed out by the random play of chance.

It may also just be possible that the fact of screening could give a false sense of health security, with a greater tendency to less healthy lifestyle. Another possibility would be that these open trials could be open to bias, with more intensive investigation for people being screened.

The final word, though, should be on the balance between benefit and risk. We know that over 80% of positive tests were false; the tests were positive but patients did not have cancer [1]. Those patients had the stress of receiving a positive test, and underwent further examination, which is not entirely benign. In 10,000 people an estimated 60-280 would have at least one colonoscopy, with 2-4 perforations or haemorrhages.

Some of these will be fatal. So for occult blood screening for one year, the chance of avoiding dying from colon cancer is 1 in 1,200, while the risk of a perforation or haemorrhage is 1 in 3,000. Maybe it is better and more productive to get people to eat more fibre, especially when we can be pretty sure that screening in practice is unlikely to be as thorough as screening in trials.

Reference:

  1. BP Towler et al. Screening for colorectal cancer using the faecal occult blood test, Hemoccult. Cochrane Database of Systematic Reviews issue 2, 1998.
  2. P Moayyedi, E Achkar. Does fecal occult blood testing really reduce mortality? A reanalysis of systematic review data. American Journal of Gastroenterology 2006 101: 380-384.

previous or next story