Time-dependent covariates are often both confounders and intermediate variables.
In the presence of such covariates, standard approaches for adjustment for confounding are biased.
The method of G-estimation allows for appropriate adjustment.
Previous studies applying the G-estimation method have addressed effects on all-cause mortality rather than on specific causes of death.
In the present study, a method to adjust for censoring by competing risks is presented.
The authors used the approach to estimate the causal effect of isolated systolic hypertension on cardiovascular mortality in the Framingham Heart Study, with a 10-year follow-up using data from 1956 to 1970.
Arterial rigidity is a major determinant of isolated systolic hypertension and may be a confounder of the relation between isolated systolic hypertension and cardiovascular death.
Conversely, isolated systolic hypertension may by itself contribute to stiffening of the vessel wall, and arterial rigidity may therefore also be an intermediate variable in the causal pathway from isolated systolic hypertension to cardiovascular death.
While controlling for arterial rigidity and other baseline and time-dependent covariates, isolated systolic hypertension decreased the time to cardiovascular death by 45% (95% confidence interval 3-69).
Mots-clés Pascal : Hypertension artérielle, Appareil circulatoire pathologie, Mortalité, Analyse statistique, Modèle statistique, Cause, Epidémiologie, Méthodologie, Biais méthodologique, Homme, Etude cohorte
Mots-clés Pascal anglais : Hypertension, Cardiovascular disease, Mortality, Statistical analysis, Statistical model, Cause, Epidemiology, Methodology, Methodological bias, Human, Cohort study
Notice produite par :
Inist-CNRS - Institut de l'Information Scientifique et Technique
Cote : 98-0426840
Code Inist : 002B30A01A1. Création : 25/01/1999.