Researchers from Dartmouth-Hitchcock Medical Center in New Hampshire analysed 12 studies of randomised cancer-screening trials for which data on all-cause and disease-specific mortality were available. Seven were mammography trials, three were faecal occult blood testing and two were chest X-ray screening for lung cancer.

To test the accuracy of the trial findings, the scientists subtracted the disease-specific death rates in the screened populations from those in the controls, and did the same for all-cause mortality rates. They then compared the two different rates of deaths.

Such trials should reveal a greater number of deaths due to specific cancers than from other causes. But, in seven of the 12 trials, there were important inconsistencies. In five of the trials, the differences in the two death rates went in opposite directions, suggesting both positive and negative effects of screening. In four trials, the disease-specific deaths were lower in the screened group than in the controls, but the all-cause mortality was the same or higher in those screened compared with the controls.

In two other trials, while the difference in death rates was more consistent, nevertheless, the difference in all-cause mortality exceeded the disease-specific mortality in the controls.

It is suggested that these differences are mainly the result of biases that led the researchers to misclassify the causes of death.

They recommend that screening trials target those at highest risk because then the ratio of disease-specific to all-cause deaths would be greater than found in the general population. They add that the selection of a high-risk population would also help to avoid the misinterpretation of statistical significance that could make a harmful screening intervention appear to be beneficial or vice versa (J Natl Cancer Inst, 2002; 94: 167-73).