BACKGROUND: The prognostic impact of nutrition and chronic kidney disease (CKD) complications has already been described in elderly haemodialysis patients but their relative weights on risk of death remain uncertain. Using structural equation models (SEMs), we aimed to model a single variable for nutrition, each CKD complication and cardiovascular comorbidities to compare their relative impact on elderly haemodialysis patients' survival. METHODS: This prospective study recruited 3165 incident haemodialysis patients \textgreater/=75 years of age from 178 French dialysis units. Using SEMs, the following variables were computed: nutritional status, anaemia, mineral and bone disorder and cardiovascular comorbidities. Systolic blood pressure was also used in the analysis. Survival analyses used Poisson models. RESULTS: The population average age was 81.9 years (median follow-up 1.51 years, 35.5% deaths). All variables were significantly associated with mortality by univariate analysis. Nutritional status was the variable most strongly associated with mortality in the multivariate analysis, with a negative prognostic impact of low nutritional markers \incidence rate ratio [IRR] 1.42 per 1 standard deviation [SD] decrement [95% confidence interval (CI) 1.32-1.53]\. The 'cardiovascular comorbidities' variable was the second variable associated with mortality [IRR 1.19 per 1 SD increment (95% CI 1.11-1.27)]. A trend towards low intact parathyroid hormone and high serum calcium and low values of systolic blood pressure were also associated with poor survival. The variable 'anaemia' was not associated with survival. CONCLUSIONS: These findings should help physicians prioritize care in elderly haemodialysis patients with CKD complications, with special focus on nutritional status.