BACKGROUND AND AIMS : Vascular access choice for patients with high risk of arteriovenous (AV) access failure has been sparking growing controversy as recent studies show similar survival and morbidity across patients receiving arteriovenous (AV) fistula or graft. We assessed hospitalization and mortality risks associated with access type in patients who started hemodialysis with a catheter without previous AV access creation in France, overall and by subgroups of age, sex and comorbidities. METHOD : Longitudinal study of 18 800 incident hemodialysis patients from 2010 through 2018, based on the linkage of the French REIN registry of kidney replacement therapy (KRT) with the national health administrative database (SNDS). First-line AV access (fistula or graft) was ascertained from SNDS procedures codes. Hospitalizations were also identified through the SNDS, whereas mortality data was obtained from the REIN Registry. We used joint frailty models to estimate hazard ratios (HR) and 95% confidence intervals (CI) of recurrent hospitalization and death associated with AV grafts, compared with AV fistulæ. These models accounted for dependence between hospitalization and death. We further estimated propensity scores for first-line AV graft placement and used inverse probability weighting (IPW) to obtain weighted HR (wHR), accounting for potential indication bias. RESULTS : Among studied patients, 35% were women, 45% had diabetes, 26% had history of heart failure and 19% had history of peripheral artery disease. More than half started dialysis urgently (52%). Patients with first-line AV graft (5%) were older than those with AV fistula (72 ± 14 versus 68 ± 15 years, respectively), and required more frequently assistance to walk (29% versus 17%). IPW resulted in covariate balance (absolute standardized difference <10%) within the overall population and the subgroups of interest (except for the timing of AV access creation in patients aged <70). Over a median follow-up of 48 months (IQR 27–48), hospitalization rates were 334 and 310 per 100 patient-years in the AV graft and fistula groups, respectively; mortality rates were 16 and 13 per 100 patient-years. Patients with AV graft had a 14% higher hazard of all-cause hospitalization (HR 1.14, 95% CI 1.08–1.20), which was only slightly attenuated in IPW analysis (wHR 1.11, 1.09–1.13). AV access type was not associated with mortality—HR 1.03 (0.89–1.19), wHR 1.11 (0.85–1.46). Results were consistent for most subgroups, except that the highest hazard of hospitalization with AV grafts compared to fistulæ was much attenuated in patients with diabetes, heart failure or peripheral artery disease with respect to patients without these comorbidities (Figure 1). CONCLUSION : In patients starting hemodialysis with a catheter without previous AV access creation, the fistula first approach is associated with similar mortality, but lower risk of hospitalization compared to first-line AV graft. This may, however, not be the case for patients with a poor vascular condition, i.e. those with diabetes or peripheral artery disease, who have a similar hospitalization risk with either graft or fistula.< Réduire