This retrospective cohort study demonstrates similar vascular access survival, adverse events, interventions and hospitalisations during a 12-month period of conventional daytime in-centre haemodialysis and a subsequent 12-month period of thrice-weekly extended-hours INHD. There were no deaths due to vascular access complications and no confirmed episodes of needle dislodgement. There appears to be an association between the use of an AVG and progression to a vascular access adverse event (intervention period), and an association between regular vitamin K antagonist prescription and progression to a vascular access adverse event (intervention and control periods).

This is the first study, to the authors’ knowledge, to compare vascular access survival, adverse events and safety between conventional in-centre haemodialysis and thrice-weekly extended-hours INHD. Other studies which have explored vascular access events in patients receiving extended-hours haemodialysis have involved intensive (i.e., > 3 times per week haemodialysis) regimens during the day or night in home and in-centre locations [4, 6]. A small, randomised crossover trial comparing conventional daytime in-centre haemodialysis to thrice-weekly, six to eight hours per session, of home haemodialysis provides a potentially comparable population [7], but there is insufficient data to explore vascular access complications [8].

There were no significant differences in the primary outcome between the control and intervention period for this retrospective cohort study. There were more episodes of post-dialysis AVF/AVG bleed, thrombosis, stenosis and medical thrombolysis in the intervention period than the control period. These secondary outcome measures did not reach statistical significance. However, there are a number of limitations (see below), and the lack of statistical significance may be a question of power. Despite not reaching statistical significance, increased episodes of such complications could be clinically significant; the nursing time and associated cost in managing these episodes, alongside the impact on the patient, have not been explored. A large proportion of the events occurred in the same patients for both the control and intervention periods. Similarly, findings from the Dialysis Outcomes and Practice Patterns Study (DOPPS) suggest that vascular access events tended to recur in the same vascular access, and that the longer a vascular access remained complication-free, the lower the risk of complications occurring [9].

A proposed mechanism for the similarity across treatment periods may relate to the frequency of the haemodialysis sessions, with patients always receiving thrice-weekly haemodialysis for both the control and intervention period. In the FHN Daily trial, participants were randomised to conventional daytime in-centre haemodialysis or to six times a week haemodialysis for 1.5–2.75 h per session. Allocation to the intervention was associated with earlier vascular access intervention compared to thrice-weekly haemodialysis (hazard ratio 1.71, 95% CI 1.08–2.73) [10]. A similar trend was observed with the FHN Nocturnal trial, with the intervention arm receiving extended-hours nocturnal HHD six times a week compared to a control arm of conventional in-centre daytime haemodialysis, but this did not reach significance (hazard ratio 1.88, 95% CI 0.97–3.64) [3]. In a case series of over 250 patients receiving extended-hours haemodialysis at home or on a nocturnal schedule in Australia, the frequency of the haemodialysis sessions was a significant predictor of vascular access event incidence, with those dialysing ≤ 3.5 times per week experiencing significantly greater vascular access survival than those dialysing > 3.5 sessions per week (log rank P < 0.001) [6].

It is important to acknowledge that all centres principally employed a rope ladder technique for AVF/AVG cannulation; this may have contributed to the similarity in events for the control and intervention periods. Parisotto et al. have demonstrated that the cannulation technique is associated with vascular access survival; area cannulation associates with a significantly higher risk of access failure than the rope ladder and button hole techniques [11].

There were no significant safety events such as death due to vascular access adverse events and needle dislodgement during haemodialysis for both the control and intervention periods. Needle dislodgement from AVF and AVG during haemodialysis presents a potentially life-threatening event; complete venous needle dislodgement results in blood loss at ≥ 300 mL/min (i.e., the blood flow rate) until the dialysis pump is stopped, with the potential for haemorrhagic shock and death within minutes [12]. Similar to the findings of this study, fatal needle dislodgement events are reportedly rare in the literature. Jose et al. evaluated all reported episodes of death from vascular access bleeding in Australia and New Zealand over a 14-year period; 79 people died of fatal vascular access haemorrhage between 2000 and 2013, with an incidence of 1 death for every 1250 person-years of haemodialysis [13]. Despite these low event rates, there is concern that events are underreported [14] and near-misses are not captured in routine dialysis practice [15]. As a result, it is important to acknowledge that this retrospective cohort study may not reflect all the safety events that have occurred at each participating site due to potential underreporting in routine clinical practice and a lack of measures to capture near-misses.

Arteriovenous grafts had a significant association with reduced vascular access survival probability during the intervention period, and vitamin K antagonist prescription had a significant association with reduced vascular access survival during the intervention and the control periods. These findings need to be interpreted with caution due to the low numbers within the study population. However, similar observations have been reported in the literature. Arteriovenous grafts have poorer long-term patency [16] and an increased risk of venous thromboembolism [17] compared to AVF. In a retrospective cohort study of individuals receiving nocturnal home haemodialysis, the use of an AVG was a significant predictor of the occurrence of vascular access adverse events and was associated with an increased frequency of events [18]. The UK Kidney Association guidelines recommend favouring AVF formation over AVG whenever possible due to the increased complication rate and poorer outcomes associated with AVGs [19]. With regard to regular vitamin K antagonist prescription, observational studies have demonstrated that warfarin use is associated with poorer primary graft patency [20] and anticoagulation use is associated with higher wound infection rates for a study population utilising both AVF and AVGs [21]. In a placebo-controlled RCT, warfarin use did not significantly impact graft survival and resulted in significantly more major bleeding events [22].

There was also a significant association between age and the 12-month vascular access survival probability during the control period, whereby younger patients (< 52 years) had a lower vascular access survival probability than those ≥ 52 years. This association did not persist for the intervention period. These findings are not reflected in the literature, whereby increasing age has been identified as a risk factor for vascular access dysfunction [23,24,25]. It also seems implausible that the impact of age would reduce with time as the absence of a significant association in the intervention period suggests. Thus, the significant association between vascular access survival probability and age observed in the control period may reflect the impact of an unaccounted for confounding factor(s), alongside the limitations of the dataset and study design.

There are a number of limitations that must be considered. The findings are limited to adult prevalent haemodialysis populations. The retrospective design introduces bias, including: potential confounding factors, e.g., patients with well-established vascular access may be more likely to be offered INHD in routine clinical practice; information bias due to the collection of data from clinical records; selection bias favouring INHD, with the study cohort being substantially younger (median age of 52 years) than the typical age for people starting dialysis in the UK (median age of 63 years) [26]. However, it is not uncommon for INHD cohorts to be younger, with average ages reported at less than 60 years old [27, 28]. All the demographic data were collected when the patient commenced INHD, and changes in routine patient care are unknown. Vascular access vintage was not an included covariate. Data collection from retrospective review of medical records was limited by the clinical information available, e.g., causes of death outside of the primary outcome, and individual cannulation techniques were not retrievable.

The study design, whereby all participants received conventional daytime in-centre haemodialysis followed by INHD, introduces bias, including: period effect, e.g., there may be an overall decline in a patient’s condition and vascular ageing with time; carryover effect, whereby the impact of receiving one type of dialysis has lasting influence on the occurrence of events in the intervention period; and sequence effect, whereby there may be a specific interaction of receiving conventional in-centre daytime dialysis prior to INHD [29]. There was an increased number of events for post-dialysis AVF/AVG bleed, thrombosis and stenosis during the intervention (INHD) period, with a substantial contribution from the same patients in both follow-up periods. This may suggest an impact of both a period and a carryover effect. The observational design of the study also introduces immortal time bias as all the participants were required to survive a period of time on haemodialysis without experiencing an event before being eligible for the data collection control and intervention periods.

Nonetheless, the results fill an important gap in understanding vascular access complications in an extended-hours, thrice-weekly INHD cohort. Data collection involved three dialysis centres across two continents, increasing the generalisability of the findings. Although the findings are not definitive, they demonstrate that concerns regarding vascular access survival and safety events should not inhibit dialysis centres from offering an INHD programme. The findings could also be used as a preliminary dataset regarding INHD safety when counselling patients about dialysis choices. Further work is needed, primarily the collection of prospective data of vascular access survival, adverse events and safety, comparing INHD to daytime in-centre haemodialysis. The evaluation of safety needs to capture near-misses alongside confirmed events.