No survival benefit from early-start dialysis in a population-based, inception cohort study of Swedish patients with chronic kidney disease

Abstract

To investigate how the timing of dialysis initiation is associated with mortality. Population-based, prospective, observational cohort study. Clinical laboratories (n = 69) provided information on all patients in Sweden whose serum creatinine level for the first time and exceeded 3.4 mg dL(-1) (men) or 2.8 mg dL(-1) (women) between 20 May 1996 and 31 May 1998. All patients (n = 901), aged 18-74 years, in whom the cause of serum creatinine elevation was chronic kidney disease, were included in the study; participants were interviewed and followed for 5-7 years. Information on date of death was obtained from a national Swedish population register. Early-start dialysis [estimated glomerular filtration rate from serum creatinine (eGFR) ≥7.5 mL min(-1) per 1.73 m(2)] was compared to late start of dialysis (eGFR <7.5 mL min(-1) per 1.73 m(2)), and no dialysis. Relative risk [hazard ratio (HR)] of death was modelled with time-dependent multivariate Cox proportional hazards regression. Mean eGFR was 16.1 mL min(-1) per 1.73 m(2) at inclusion and 7.6 mL min(-1) per 1.73 m(2) at the start of dialysis. Among the 385 patients who started dialysis late, 36% died during follow-up compared to 52% of 323 who started early. The adjusted HR for death was 0.84 [95% confidence interval (CI) 0.64, 1.10] among late versus early starters. The mortality among nondialysed patients increased significantly at eGFR below 7.5 mL min(-1) per 1.73 m(2) (HR 4.65; 95% CI 2.28, 9.49; compared to eGFR 7.5-10 mL min(-1) per 1.73 m(2)). After the start of dialysis, the mortality rate further increased. Compared to nondialysed patients with eGFR ≤15 mL min(-1) per 1.73 m(2), adjusted HR was 2.65 (95% CI 1.80, 3.89) for patients receiving dialysis. We found no survival benefit from early initiation of dialysis.