Gains in life expectancy after elimination of major causes of death: revised estimates taking into account the eVect of competing causes

Background—It is generally acknowledged that conventional estimates of the potential number of life years to be gained by elimination of causes of death are too generous. This is because these estimates fail to take into account the fact that those who are saved from the cause are likely to have one or more other conditions (“competing” causes of death), which may increase their risks of dying. It is unknown to what extent this introduces bias in comparisons of life years to be gained between underlying causes of death. The purpose of the study was to assess this bias. Data and methods—A sample of 5975 death certificates from the Netherlands, 1990, was coded for the presence of diseases that, according to a set of explicit rules, could be regarded as potential causes of death “competing” with the underlying cause. Logistic regression analysis was used to quantify age and sex adjusted diVerences between four main underlying causes of death (neoplasms, cardiovascular diseases, respiratory diseases, all other diseases) in prevalence of the six most frequent competing causes of death (neoplasms, ischaemic heart disease, cerebrovascular disease, other cardiovascular diseases, chronic obstructive lung disease, all other diseases). These prevalence diVerences were then used to revise conventional calculations of gains in life expectancy, by taking them to indicate diVerences in risk of dying from these competing causes after the underlying cause has been eliminated. Results—The prevalence of competing causes of death is relatively low among persons dying from neoplasms as the underlying cause, about average among persons dying from cardiovascular diseases, and relatively high among persons dying from respiratory diseases. Taking this into account results in substantial decreases of potential life years to be gained by elimination of cardiovascular diseases and respiratory diseases, relative to the number of years to be gained by elimination of neoplasms. Specifically, while according to the conventional calculations the gain in life expectancy by elimination of cardiovascular diseases exceeds that for neoplasms by more than one year, in the revised calculations the number of life years to be gained is approximately equal. Conclusions—Despite its limitations, mainly relating to reliance on death certificate data, this study suggests that conventional estimates of diVerences between underlying causes of death in life years to be gained by elimination are seriously biased by ignoring the eVects of competing causes. Specifically, the relative impacts of eliminating cardiovascular diseases and respiratory diseases, as compared with eliminating neoplasms, seem to be overestimated. The implications are discussed. (J Epidemiol Community Health 1999;53:32–37) Estimates of the number of life years to be gained by elimination of specific causes of death provide an easy to grasp and powerful summary of the relative importance of these causes of death, as well as of the potential benefits of intervention programmes. Unfortunately, however, conventionally derived estimates are likely to be biased. The main assumption on which cause-elimination life table calculations rest is, that those who are saved from dying because of the eliminated cause (for example, ischaemic heart disease) have risks of dying from non-eliminated causes that are equal to the average risks of dying from these causes as observed in the total population. 6–8 This assumption is far from realistic: because of common risk factors many persons dying from ischaemic heart disease will be at a higher than average risk of dying from various other causes, such as lung cancer (smoking), cerebrovascular disease (hypertension, arteriosclerosis), and chronic obstructive lung disease (smoking). Gains in life expectancy after elimination of ischaemic heart disease are therefore likely to be too optimistic. Moreover, this element of overestimation may diVer by underlying cause of death, thereby introducing bias in the comparison between the eVects of eliminating diVerent causes of death. If among patients saved from dying caused by lung cancer or breast cancer, the remnant mortality risk would be less (or more) increased than it is among patients saved from dying because of ischaemic heart disease, the diVerence between these causes in life years gained would be smaller (or larger). Although this problem of “competing” causes of death, and its potential implications for cause-elimination life tables, has frequently been discussed in the epidemiologic and demographic literature, a lack of empirical data on associations between causes of death has until now prevented a solution from being J Epidemiol Community Health 1999;53:32–37 32 Department of Public Health, Erasmus University, Rotterdam, the Netherlands