Superior survival of incident patients on high-volume online hemodiafiltration compared to high-flux hemodialysis

Introduction and Aims: Hemodiafiltration (HDF) has become the preferred treatment modality for dialysis patients in many countries. Randomised controlled clinical studies and observational studies have indicated a survival advantage with high volumes of substitution fluid, while modality treatment crossover during follow-up remains problematic. The aim of this large international epidemiological cohort study was to compare patient survival on high-volume HDF (HV-HDF) with high-flux hemodialysis (HD) applying inverse probability of censoring weighting (IPCW) to take treatment modality crossover into consideration. HV-HDF was defined as HDF with a substitution fluid volume of ≥ 21 liters per session. Methods: The study population was extracted from a database of 47979 patients on dialysis treatment in 369 NephroCare centers throughout Europe between January 1, 2007 and December 31, 2012. Main predictor was treatment modality. Patients were termed either HV-HDF or HD patients if on that treatment modality for ≥75% of the time within the enrollment period, defined as the 3 months after baseline. Patients were followed for a maximum of 5 years. Other analyzed parameters were country, age, gender, BMI, systolic blood pressure, Charlson comorbidity index, diabetes mellitus, other comorbidities, hemoglobin, albumin, C-reactive protein, phosphate and Kt/V. Propensity score matching and IPCW analyses were applied to reduce bias by indication and to take treatment modality crossover into consideration, respectively. Results: Following exclusion criteria and propensity score matching, 1590 incident patients remained in the study population (795 each on HV-HDF and HD). Kaplan-Meier and proportional Cox regression analyses revealed no significant advantage of HV-HDF. Results were biased by modality crossover as many patients switched treatment modality during the subsequent 5-year study period: 7% of HV-HDF patients switched to HD and 55% of HD patients switched to HV-HDF. Application of IPCW to account for these switches revealed a statistically significant survival advantage of HV-HDF over HD (OR 0.501; CI 0.366-0.684; p=0.00001). Subgroup analysis indicated a higher benefit of HV-HDF for some subgroups than for the complete cohort, e.g. females, non-diabetics, patients aged between 65 and 74, obese patients and patients with high blood pressure (Figure). Conclusions: Failure to account for modality switches when comparing survival of patients treated by HV-HDF and HF can lead to inaccurate results. Application of advanced statistical analysis tools unveiled a significant 0.5 odds ratio for HV-HDF compared to HD. Sub-analysis showed that some sub-cohorts appear to benefit more from HV-HDF than others. (Figure Presented).