In response

In response

Our points of view [1] mainly deviate from those in the letter by Ekelund and Janzon as to the degree of evidence and magnitude of effect in screening for colorectal cancer.

First the evidence. Ekelund and Janzon call for absolute effects (their percentages are still measures of relative effect). Take an example in absolute figures, if an expected number of 100 colorectal cancer deaths can be reduced to 80, there is a 20% and statistically significant reduction in mortality. At the same time about a total of 2 500 deaths will take place in that population. However, because of statistical random variation, the range of deaths we observe is between 2 400 and 2 600. The effect of 20 deaths can not be distinguished from the random noise of 200 deaths. The observation of Ekelund and Janzon of no statistically significant reduction in total mortality is correct but their inference or conclusion is simply wrong. The disappearance of statistical significance is not necessarily because of disappearance of the effect due to the misclassification of deaths but, even if the effect remained the same, the random error would increase making the observation not statistically significant.

Next the magnitude. It is true that the number needed to screen (NNS) is large. The above example of reduction of 20 deaths compared to the total of 2 500 deaths is just another way to explain this basic fact. Screening for colorectal cancer is effective but the effect is small. The prolongation of life due to screening is an alternative measure with public health importance for NNS. For mammography with proven effectiveness the prolongation due to single screening is about two days [2], and the order of magnitude is the same in screening for colorectal cancer. Again the observation of Ekelund and Janzon is perfectly correct but their inference is wrong. The small magnitude is relevant not only for colorectal cancer screening but it is a general fact for most medical interventions including clinical treatment.

Furthermore, health services activities are based only seldom on direct evidence of final outcome but they are based mainly on intermediate end points or beliefs. Medicine based on direct evidence of small effect will ultimately result in better human health than one based on those inconclusive alternatives. At present and with the evidence available we believe that screening for colorectal cancer is more effective and also offers better use of resources than many established health services activities.

We agree with the comment by Ekelund and Janzon on application of the scientific results on routine screening. Their attitude, however, would prevent any public health activity. The implementation of a routine screening programme, sometimes called feasibility study, can meet even the preconditions by Ekelund and Janzon. It should be an organised one and it should follow rigorous scientific principles as randomisation during the time window of implementation. Routine screening with mammoraphy [3] and with the FOB test [4] in Finland shows that experimental design can be applied in a routine service and it is the best, and often the only, means to evaluate for its effectiveness.

Because any activity in the Nordic countries has only a small effect on length of life, research is urgently needed on other final end points related to quality of biological, mental and social life, to provide evidence on routine health services activities.