ABSTRACT
BACKGROUND: Cancer screening trials have traditionally focused on disease-specific mortality, the number of subjects whose death is attributed to the screened disease. This end point is generally easier to study than all-cause mortality (the overall death rate), because fewer subjects are needed to achieve a statistically significant result. However, this approach has many potential biases, and it neglects the possibility that screening may lead to potentially fatal complications. The authors of this study compared disease-specific mortality changes to all-cause mortality changes in a collection of cancer screening trials.
POPULATION STUDIED: This study examined 12 published randomized trials of cancer screening. Of 16 initial trials identified, the 12 chosen for study were those in which disease-specific and all-cause mortality could be determined. The 12 chosen studies included 7 of mammography, 3 of fecal occult blood testing, and 2 of chest x-rays for lung cancer.
STUDY DESIGN AND VALIDITY: The researchers used a list published in a text on cancer screening to identify randomized trials for inclusion in this study. Updated information from each of the trials was obtained by performing a PubMed search of authors’ names and other relevant terms. This was not an exhaustive, systematic review of the literature. A more extensive literature search would have used multiple databases, evidence-based search methods, and possibly unpublished data. Very little information is given on the search terms used in PubMed. However, since this was a comparison of different outcome measures rather than a meta-analysis, a systematic review is not necessarily required.
OUTCOMES MEASURED: For each study, the difference in mortality between screened and unscreened (control) groups was reported as the screening benefit. The screening benefits from both disease-specific mortality and all-cause mortality data were then compared in terms of number of deaths per 10,000 person-years of observation.
RESULTS: One would expect that if a screening program decreased mortality related to the disease, overall mortality would be less as well. The authors found that this correlation did not occur in most of these studies. Five of the studies found that disease-related mortality and overall mortality went in different directions. Three of these 5 studies reported a statistically significant benefit in disease-specific mortality, but the all-cause mortality was either not affected or was worse. Two trials showed no benefit in disease-specific mortality but a trend in a positive or negative direction in all-cause mortality.
Although disease-specific mortality has been the standard for reporting mortality benefit in cancer screening, it does not necessarily correlate with significant benefits in all-cause mortality. In other words, some cancer screening may decrease deaths due to the screened disease, but patients still die at the same (or even higher) rate despite the screening. Inconsistent results are evident in trials studying mammography screening for breast cancer, fecal occult blood testing for colon cancer, and chest x-ray screening for lung cancer. When deciding whether a screening intervention is potentially beneficial, we may be misled by reports of disease-specific mortality.