All screening interventions have the potential to do harm and screening by faecal occult blood testing (FOBT) is no exception. The harm arises not from the test itself, although the psychological impact is not negligible, but mainly from the investigation and treatment of test positives. The benefits of FOBT have been studied extensively and it seems that among those who comply with biennial testing, colorectal cancer mortality is significantly reduced1-3 and that the reduction is greater when the test is used annually.4 But there is a price to pay. Firstly, there is the anxiety engendered by performing the test itself and the greater anxiety induced in those who are found to be positive requiring further investigation. Although this has been shown to be short-lived, lasting mainly until the completion of diagnostic investigations, cumulatively it might be expected to affect around 10% of those complying with the test biennially over 20 years.

Secondly, there are the well documented complications associated with colonoscopy, which is considered the gold standard for the diagnostic workup of test positives.5 Overall mortality rates are reported to average around 0.02%, varying between 0 and 0.06%. Perforation can be expected in 0.3% and haemorrhage in around 1% of people undergoing polypectomy. A variety of other complications have been reported, the most serious of which are cardiovascular.6 Finally, there is the significant morbidity associated with the surgical treatment of colorectal cancer and large adenomas, with published 30-day mortality rates varying between 1 and 8%.7

Based on the promising findings of the randomised trials, pilot studies of FOBT screening are about to start in two sites in the UK in preparation for possible implementation in a national screening programme. These studies aim to examine feasibility and acceptability rather than efficacy and adverse effects. Therefore the paper in this issue reporting on the adverse effects of screening in the Nottingham study is timely. Robinson et al (see page588) specifically tackle three issues raised by Ahlquist6in a paper questioning the wisdom of implementing screening by FOBT. Ahlquist argues: (1) that false negative results falsely reassure patients, leading to delayed diagnosis and poorer outcome; (2) that colonoscopic mortality compounded over several years could counter any gains in reducing deaths from colorectal cancer; (3) that true positive FOBT results may lead to overdiagnosis and treatment of asymptomatic colorectal cancers which might have an innocuous natural history, perhaps leading to a shortened lifespan. To counter the first point, Robinson et al show that there is no difference in the stage at diagnosis in interval and control cancers, suggesting that false negative FOBTs probably do not delay diagnosis. In response to the second point, they show that there were no colonoscopy deaths although their perforation rate (5 of 1474, 0.3%) was exactly as that reported by Ahlquist. Moreover, they restrict their analysis only to baseline colonoscopies even though those who were found to have adenomas were entered into a colonoscopic surveillance programme. If we assume that the 40% of patients undergoing colonoscopy who were found to have adenomas underwent an average of two more colonoscopies, this approximately doubles the exposure of the cohort to colonoscopy.

One point not tackled by Robinson et al was the observation by Ahlquist that, in the Minnesota study,4 the reduction in colorectal mortality was precisely offset by an increase in mortality from cardiac ischaemia. An increase in cardiovascular deaths, of similar magnitude to the reduction in colorectal cancer mortality, was also observed in the Danish study.2 The non-colorectal cancer mortality rate was also increased in Nottingham,1 more than offsetting any reduction in deaths from colorectal cancer, therefore it would have been of interest to have examined deaths from ischaemic heart disease in Nottingham (table 1).

So what should we make of this? If there is an increase in deaths from coronaries, when do they occur and what is causing them? Robinsonet al state that none of 1778 patients who were not being treated for detected colorectal neoplasia, died within 30 days of colonic investigation. It is not clear how many suffered a cardiac event within that time which might have led to death at a later time. Myocardial infarction duringcolonoscopy seems to be rare5 with a large US survey reporting three (0.012%) events in 25 298 colonoscopies, but the incidence of infarction within a few hours or days of colonoscopy is unknown.

The possibility of an increase in coronary mortality observed in the published randomised trials of colorectal screening suggests that lower gastrointestinal endoscopy and surgery should be undertaken judiciously, balancing in each patient the risks and benefits. Haemoccult, the FOBT used in the European randomised trials, when used in the unrehydrated form is a specific test with a predictive value for cancer of a positive test of around 10%. There is no other indication for colonoscopy in which the yield of cancers is so high. However around 40% of those undergoing colonoscopy as a result of a positive FOBT are found to have adenomas and it is customary to enter them into an endoscopic surveillance programme where the benefits are less clearly defined and, perhaps more importantly, the harm has yet to be quantified.