Back‐up arteriovenous fistula in peritoneal dialysis patients: a retrospective cohort study of long‐term meaningful outcomes Authors: Hicham I. Cheikh Hassan, Karumathil M. Murali, Jenny H. C. Chen, Judy Mullan Published: 2026 ABSTRACT ABSTRACTBackgroundBack‐up arteriovenous fistula (bAVF) is used in peritoneal dialysis (PD) to avoid haemodialysis transfer via a central venous catheter (CVC). However, no previous study examined mortality or complications associated with bAVF creation.AimsComparing mortality, haemodialysis transfer overall or with a CVC and vascular access complications (bAVF or CVC) between patients with a bVF and those with no bAVF.MethodsA retrospective cohort study of adult PD patients within an Australian Local Health District (2017–2023). Cox regression models were used for hazard ratios (HR) and 95% confidence intervals (95% CI).ResultsOf 115 patents followed up (median of 1.9 years (interquartile range 0.9, 3.5)), 50 (44%) had a bAVF of which 31 (62%) matured. The bAVF group had 56 procedures (1.1/patient), including 32 assisted interventions (0.6/patient). There was no associated difference in mortality between bAVF and no bAVF group (HR 0.8, 95% CI 0.4–1.6, P = 0.6) or haemodialysis transfer (HR 1.0, 95% CI 0.5–2.0, P = 0.9). A bAVF was associated with a lower haemodialysis transfer risk with a CVC (HR 0.4, 95% CI 0.1–0.9, P = 0.04). Complications between the bAVF and no bAVF group were similar (28% vs 14%, P = 0.06).ConclusionbAVF were associated with lower risk of haemodialysis transfer to haemodialysis with a CVC, compared to those with no bAVF. However, there was no difference in mortality, risk of transferring to haemodialysis or complications. Introduction Peritoneal dialysis (PD) is a commonly preferred treatment for patients with kidney failure. It helps preserve remaining kidney function, 1 , 2 offers flexibility in scheduling, supports quality of life and may be associated with better survival rates compared to in‐centre haemodialysis. 3 , 4 However, PD is not without its challenges, one of which is transfer to haemodialysis. Haemodialysis transfer can occur in 20% of PD patients within the first year of starting PD 5 , 6 and can ultimately affect 40%–50% of patients. 7 , 8 , 9 Haemodialysis transfer can be sudden, which is anticipated given that infections are the most common cause of haemodialysis transfer, with the peritoneum being the leading site of infection. 6 , 7 , 8 Given the sudden nature of some haemodialysis transfers, many patients require a central venous catheter (CVC) rather than an arteriovenous fistula (AVF) for vascular access. According to data from the Australian and New Zealand patients (ANZDATA), 65% of patients who transfer from PD to haemodialysis required a CVC. 9 For haemodialysis patients an AVF is the preferred vascular access, supported by initiatives such as the ‘fistula first approach’. 10 Starting haemodialysis with an AVF is associated with a lower risk of mortality, fewer bloodstream infections and better dialysis efficiency compared to CVCs. 10 However, AVFs require time to mature, making them less suitable for acute transfer from PD. Additionally, AVF maturation failure rates are high, ranging from 30% to 60% in haemodialysis cohorts. 11 , 12 The optimal approach to vascular access in PD patients, particularly given the high rates of haemodialysis transfer, remains unclear. While initiating haemodialysis with an AVF offers advantages, it also presents practical challenges. To avoid reliance on CVCs, some centres have a policy of creating a back‐up AVF (bAVF) in PD patients. A bAVF offers potential benefits, especially in reducing the likelihood of haemodialysis transfer with a CVC as reported in a recent meta‐analysis. 13 However, caution must also be applied when considering a bAVF, with the same meta‐analysis also reporting high rates of bAVF non‐use (reaching 70%). 13 Decision making around bAVF creation is further complicated by the nature of the available evidence. None of the bAVF studies examined hard clinical outcomes, such as mortality or vascular access‐related infection risks. Most focused solely on the mode of haemodialysis transfer, bAVF versus CVC. To address the gap in the evidence, we undertook an analysis of a cohort of PD patients within an Australian Local Health District (LHD). Our primary aim was to examine mortality comparing patients with a bAVF to those with no bAVF. We also examined (1) risk of haemodialysis transfer overall and (2) with a CVC, (3) bAVF maturation rates, (4) interventions and procedures in patients with a bAVF and (5) complication rates between the bAVF and the no‐bAVF groups. Methods Study design and participants A retrospective cohort study of adult patients (≥18 years of age) in a LHD in Australia, who commenced PD or commenced haemodialysis and transferred to PD (2017–2023). We excluded patients who recovered kidney function, transferred from another health service or were on haemodialysis for >3 months (those likely to have had an AVF created for haemodialysis access rather than a planned bAVF). This study was approved by the LHD and Medical Human Research Ethics Committee (ISLHDLNR2023/207) and reported in accordance with the STROBE guidelines. 14 Data collection Data collected included baseline sociodemographic and clinical information, including dates for dialysis start, PD catheter insertion, modality change, access creation and PD complications, in addition to arterial and venous mapping for vascular access creation planning, vascular access assessments and complications. Australia and New Zealand Dialysis and Transplant Registry (ANZDATA) information was also obtained and linked to supplement our dataset and check for discrepancies or data coding errors. Urbanity and socio‐economic indices were determined from the postcode of the patients’ residential addresses, using census data provided by the Australian Bureau of Statistics. 15 We captured peritonitis episodes (categorised into 0, 1 or ≥ 2), mortality, cause of death and transplantation. All vascular access procedures, including AVF or CVC, were created by the local vascular surgical team, with vascular access sites monitored for complications. Exposure The exposure was the creation of a bAVF, so that those patients could be compared to those for whom no bAVF was created. A bAVF was defined as an AVF created or maintained to facilitate haemodialysis transfer from PD, with the aim of avoiding the use of a CVC. The decision to create a bAVF was based on the recommendation of the treating nephrologist. Generally, in our Local Health District Renal Unit, patients deemed at risk of being transferred to haemodialysis or those deemed likely to benefit from a bAVF were offered the option after discussing possible risks and benefits. Outcomes The primary outcome was mortality, and secondary outcomes were transfer to haemodialysis overall and with a CVC and vascular access complications (bAVF or CVC). A vascular access complication was defined as infectious or non‐infectious, including procedures performed for bAVF or CVC (e.g. removal due to poor flow). bAVF maturation was defined by surgical assessment (based on physical examination), ultrasound assessment following the rule of 6s (flow >600 mL/min, vein diameter > 6 mm and vein depth < 6 mm) or adequate flow rates on first dialysis use. bAVF‐assisted maturation was defined as maturation achieved through an intervention to promote bAVF maturation, including surgical or percutaneous procedures: open revision of the anastomosis, angiogram with drug eluding stent (DES), ligation of tributary, superficialisation or mobilisation. Procedures such as angiogram without angioplasty or procedures for complications were not considered assisted interventions but were included in the total number of procedures. Follow‐up time was from commencement of PD until death, transplantation, loss of follow‐up or end of the study period. Statistical analysis Categorical data were expressed as numbers (percentages) and continuous data as means with standard deviation (SD) or median with interquartile ranges (IQR 25th to 75th percentile) as per distribution. Baseline characteristics were compared using chi‐squared test, Fishers's exact test, t test or Mann–Whitney U test as appropriate. We used an intention‐to‐treat approach, with our exposure (bAVF creation) assigned regardless of whether the bAVF matured or was ever used. Cox regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs), with time to event defined as time of commencing PD until primary or secondary outcome or censoring. A multivariate model was created using backward selection of variables (variables having a P value in the primary outcome of >0.1 excluded). We incorporated age, gender, diabetes, body mass index (BMI) and peritonitis in all models regardless of significance. We included initial kidney replacement therapy (KRT) modality (PD or haemodialysis) in the analysis examining haemodialysis transfer and haemodialysis transfer with a CVC. Cause of kidney failure was not included due to collinearity with diabetes. Statistical analysis was performed using STATA 15.1 (StatCorp, College Station, TX, USA). P‐values of <0.05 were considered statistically significant. Results We included 115 patients (Fig. 1), with a median follow‐up time of 1.9 years (IQR 0.9, 3.5). Study cohort characteristics are outlined in Table 1. Of the 115 patients, 50 (44%) had a bAVF and 65 (56%) did not. Median time from bAVF creation to PD start was 1.8 months (IQR 0.4 months before PD start to 10.4 months after PD start). BMI was the only baseline variable significantly different between the bAVF and the no‐bAVF group (30.7 kg/m2 vs 27.6 kg/m2, P = 0.003). During follow‐up, peritonitis occurred in 26 (23%) patients, with 0 in 89 (77%), 1 in 20 (17%), and ≥2 in 6 patients (5%). Patient flowchart for study cohort. Table 1Baseline characteristics of all study groups and by back‐up AVF statusAll study group n = 115Back‐up AVFYes n = 50 (44%)No n = 65 (56%)Gender (female, (n, %))43 (38%)21 (42%)22 (34%)Age (mean years, SD)61.0 (16.3)62.1 (14.7)60.2 (17.5)Cause of kidney failure (n, %)Diabetic39 (34%)15 (30%)24 (37%)Hypertension12 (18%)8 (16%)13 (20%)Cystic10 (9%)4 (8%)6 (9%)Glomerulonephritis28 (24%)16 (32%)12 (19%)Other17 (15%)7 (14%)10 (15%) Smoking status (n, %) Non‐smoker 62 (54%) 22 (44%) 40 (62%) Ex‐smoker41 (36%)21 (42%)20 (31%)Current smoker12 (10%)7 (14%)5 (8%)BMI (mean kg/m2, SD)BMI category (n, %)28.9 (5.6)30.7 (5.7)27.6 (5.2)Underweight3 (3%)1 (2%)2 (3%)Normal26 (23%)6 (12%)20 (31%)Overweight47 (41%)17 (34%)30 (46%)Obese39 (34%)26 (52%)13 (20%)Comorbidities (n, %)Chronic lung disease11 (10%)3 (6%)8 (12%)Coronary artery disease37 (32%)13 (26%)24 (37%)Peripheral vascular disease21 (18%)10 (20%)11 (17%)Cerebrovascular disease13 (11%)6 (12%)7 (10%)Diabetes mellitus50 (44%)23 (46%)27 (42%)Cancer24 (21%)14 (28%)10 (15%)Urbanity (n, %)Major city78 (68%)33 (66%)45 (69%)Regional37 (32%)17 (34%)20 (31%) Socio economic Index (n, %) High 38 (33%) 15 (30%) 23 (35%) Middle48 (42%)21 (42%)27 (42%)Low29 (25%)14 (28%)23 (35%)Initial KRT (n, %)PD104 (90%)46 (92%)58 (89%)Haemodialysis11 (10%)4 (8%)7 (11%) Note: Data expressed as number (percentage) or mean (standard deviation). P‐value for significance was >0.05 except for BMI where P‐value was 0.003 for both continuous and categorical measures. AVF, arteriovenous fistula; BMI, body mass index; KRT, kidney replacement therapy; PD, peritoneal dialysis; SD, standard deviation. Baseline characteristics of all study groups and by back‐up AVF status Smoking status (n, %) Non‐smoker 62 (54%) 22 (44%) 40 (62%) Socio economic Index (n, %) High 38 (33%) 15 (30%) 23 (35%) Note: Data expressed as number (percentage) or mean (standard deviation). P‐value for significance was >0.05 except for BMI where P‐value was 0.003 for both continuous and categorical measures. AVF, arteriovenous fistula; BMI, body mass index; KRT, kidney replacement therapy; PD, peritoneal dialysis; SD, standard deviation. Characteristics of bAVF group Of the 50 bAVFs created, 31 (62%) matured, of which 14 (28%) were unassisted maturation. In the bAVF group, 56 procedures were undertaken, averaging 1.1 procedures/patient. An assisted intervention occurred in 23 (46%) of the bAVF patients, with time to first intervention 6 months (IQR 2.4, 14.4). In total 32 assisted interventions were required, averaging 0.6 assisted interventions/patient. Among the bAVF group, 27 (54%) patients received no interventions, 17 (34%) one intervention and six (12%) two or more. The most common assisted intervention was an angiogram with DES, performed in 26 (81%) patients, followed by ligation of tributaries in four (13%) patients and superficialisation or mobilisation of the bAVF in two (6%) patients. There were 24 procedures performed for reasons other than to assist bAVF maturation, including angiogram with no intervention (n = 8), open thrombectomy (n = 4), open aneurysm resection (n = 2), open revision (n = 2), distal revascularisation with interval ligation (n = 2) and proximalisation of anastomosis (n = 2), bAVF ligation for heart failure (n = 1) and bAVF ligation for vascular steal syndrome (n = 3). Mortality Death occurred in 27 (24%) patients, transplant 18 (15%), end of study period 69 (60%) and relocation in one (1%). Death was due to cardiovascular causes in 12 (44%) patients, infection in 10 patients (37%), withdrawal in four patients (15%) and cancer in one patient (4%). There was no difference in the proportion of patients who died in the bAVF group compared to the no‐bAVF group (26% vs 22%, P = 0.6), and survival analysis showed no difference in univariate analysis (HR 0.8, 95% CI 0.4–1.7, P = 0.6) and adjusted analysis (aHR 0.8, 95% CI 0.4–1.6, P = 0.6) (Tables S1, 2 and Fig. 2). Table 2Risk factors for mortality and transfer to haemodialysisMortalityTransfer haemodialysisTransfer haemodialysis with a CVCUnivariableMultivariableUnivariableMultivariableUnivariableMultivariablebAVF0.8 (0.4–1.7)0.8 (0.3–1.8)1.2 (0.6–2.3)1.0 (0.5–2.0)0.4 (0.1–0.9)0.4 (0.1–0.9)Gender (Male)0.9 (0.4–1.9)0.4 (0.2–1.1)1.5 (0.7–3.2)1.6 (0.7–3.5)2.0 (0.8–5.6)1.9 (0.7–5.5)Age (decade)2.1 (1.3–3.2)2.3 (1.4–3.8)0.9 (0.8–1.1)0.9 (0.7–1.1)0.9 (0.7–1.2)0.8 (0.6–1.1)Cause of kidney failureDiabeticRef–Ref–Ref–Hypertension1.6 (0.6–3.9)1.0 (0.4–2.5)0.8 (0.3–2.7)Cystic1.3 (0.4–4.8)0.8 (0.2–2.8)0.4 (0.1–2.9)Glomerulonephritis0.3 (0.1–1.2)0.7 (0.3–1.7)0.7 (0.2–1.9)Other0.7 (0.1–3.0)1.0 (0.3–3.1)1.0 (0.3–3.8)Smoking statusNon‐smokerRef–Ref–Ref–Ex‐smoker1.9 (0.8–4.4)1.2 (0.6–2.5)0.7 (0.3–1.8)Current smoker2.3 (0.7–7.3)1.4 (0.5–3.7)1.0 (0.3–3.7)BMI0.9 (0.9–1.0)0.9 (0.8–1.0)1.0 (0.9–1.1)1.1 (0.9–1.1)1.0 (0.9–1.1)1.0 (0.9–1.1)ComorbiditiesChronic Lung disease2.0 (0.7–5.9)–0.8 (0.2–2.6)–1.4 (0.4–4.8)–Coronary artery disease2.3 (1.1–5.0)–1.7 (0.9–3.4)–2.3 (0.9–5.3)–PVD2.1 (0.9–4.7)2.3 (1.0–5.4)1.6 (0.7–3.6)–2.8 (1.0–6.4)–Cerebrovascular disease3.1 (1.2–8.2)–1.4 (0.5–3.9)‐–1.7 (0.5–5.9)–Diabetes mellitus2.0 (0.9–4.4)3.2 (1.2–7.9)0.9 (0.5–1.9)0.7 (0.3–1.7)1.3 (0.6–2.9)1.3 (0.5–3.4)Cancer1.0 (0.5–2.4)–1.2 (0.6–2.6)–1.4 (0.5–3.5)–UrbanityMajor CityRef–Ref‐Ref–Regional1.1 (0.5–2.4)1.2 (0.6–2.3)1.1 (0.4–2.7)Socio‐economic IndexHighRef–Ref‐Ref–Middle0.9 (0.4–2.2)1.1. (0.5–2.4)0.8 (0.3–2.1)Low1.0 (0.4–2.7)0.6 (0.2–1.6)0.7 (0.2–2.0)Peritonitis1.1 (0.5–2.4)0.7 (0.3–1.7)0.6 (0.3–1.3)0.7 (0.3–1.7)0.8 (0.4–2.3)0.8 (0.3–2.3)0RefRefRef10.9 (0.4–2.4)0.3 (0.1–0.9)0.3 (0.1–1.5)≥ 21.3 (0.4–4.6)2.1 (0.7–6.0)3.5 (1.2–10)Initial KRT (haemodialysis)1.9 (0.4–8.1)–3.2 (1.2–8.4)3.6 (1.3–9.9)1.8 (0.4–7.9)2.2 (0.5–10) Risk factors for mortality and transfer to haemodialysis Survival curves showing time to mortality according to back‐up AVF status. bAVF, Back‐up arteriovenous fistula. Transfer to haemodialysis overall and with a CVC There were 36 (31%) patients who transferred to haemodialysis after a median of 13.2 months (IQR 5–12), with no significant difference in the proportion of patients in the bAVF compared to the no‐bAVF group (40% vs 24%, P = 0.08, Table S1). This was similar in the univariate (HR 1.2, 95% CI 0.6–2.3, P = 0.6) and the adjusted analysis (aHR 1.0, 95% CI 0.5–2.0, P = 0.9) (Table 2, Fig. 3). Only initial KRT with haemodialysis predicted the risk of haemodialysis transfer (aHR 3.6, 95% CI 1.3–9.9, P = 0.01) (Table 2). Survival curves showing time to haemodialysis transfer according to back‐up AVF status. bAVF, Back‐up arteriovenous fistula. In patients with a bAVF, 30 (60%) did not transfer to haemodialysis, 14 (28%) transferred with a bAVF and six (12%) transferred with a CVC. In patients without a bAVF, 49 (75%) did not transfer to haemodialysis and for the remainder, 16 (25%) transferred with a CVC. A bAVF significantly reduced the risk of haemodialysis transfer with a CVC in the univariate analysis (HR 0.4, 95% CI 0.1–0.9, P = 0.04) and adjusted analysis (aHR 0.4, 95% CI 0.1–0.9, P = 0.04) (Table S1, Table 2, and Fig. S1). Complications Twenty‐three (20%) patients experienced a vascular access complication, 14 (28%) in the bAVF group and nine (14%) in the no‐bAVF group (P = 0.06). A breakdown of the complications is provided in Table 3. Table 3Number of complications in the study cohort and stratified by bAVF statusBack‐up AVFYes (n = 50)No (n = 65)Vascular access infectionCVC17AVF30Fistula complicationsAneurysm repair20Open thrombectomy40Open revision20Ligation for high output10Ligation for steal30Distal revascularisation20Proximalisation20Removal CVAD for poor flow15 Number of complications in the study cohort and stratified by bAVF status Twenty‐two (19%) patients had a CVC inserted, with 50 CVCs inserted over the study period. There were no CVCs inserted in 93 (81%) patients, one in 10 (9%) patients, and two or more insertions in 12 (10%) patients. In the bAVF group, there were six (12%) patients who required a CVC insertion compared to 16 (25%) in the no‐bAVF group (P = 0.08). In total, 41 CVCs were inserted in the no‐bAVF group (with 10 patients having more than one CVC insertion), and nine were inserted in the bAVF group (with one patient having more than one CVC insertion). Discussion We examined outcomes for PD patients with a bAVF compared to those with no bAVF, showing no association in the risk of mortality, haemodialysis transfer or complications. However, a bAVF was associated with a 60% lower risk of haemodialysis transfer with a CVC. A successful bAVF creation relies on predicting patients at risk of haemodialysis transfer, with most studies, including ours, indicating that bAVFs are selected for patients deemed at risk of transfer. 16 , 17 , 18 It would therefore be anticipated that haemodialysis transfer rates would be higher in the bAVF group. However, our results and the published literature 13 , 17 , 18 , 19 show this is not the case, implying that the prediction of haemodialysis transfer is poor. Variables shown to be associated with haemodialysis transfer include centre size, diabetes and a high BMI, 20 , 21 but there is considerable heterogeneity and inconsistencies across studies. It could therefore be argued that before a successful bAVF programme can be implemented, a model or method to accurately and reliably predict and identify PD patients at risk of transfer is first needed. Without this, the utility of a bAVF strategy for patients who commence PD remains fundamentally constrained. In our study, 44% of patients on PD had a bAVF created, which is comparable to the average of 32% reported in the literature (range 25%–73%). 13 However, the proportion of centres that routinely adopt a bAVF approach remains unclear, with only one survey from the United Kingdom in the 1990s examining the prevalence of this practice and reporting that 42% of 70 renal units surveyed employed a bAVF approach. 22 There is also a notable and significant absence of patient and consumer perspective, including quality‐of‐life measures. Patient input on the burden of requiring more vascular access procedures (with possible complications) balanced against ease of mind in having a functioning bAVF for use if needed is important in assessing the rationale for bAVF creation. Additionally, there are no qualitative studies on physician and care provider perspective on creating a bAVF, including views on risks and benefits as well as evidence required to re‐evaluate their practice. A bAVF is associated with a lower risk of transfer to haemodialysis with a CVC. This aligns with previous studies and our meta‐analysis, which showed that bAVFs were associated with lower CVC‐dependent transfers by almost 60%. However, our study also did not find a mortality benefit, an outcome classified as a key core clinical outcome of importance among PD patients and care providers. 23 We also showed that, on average, 1.1 procedures were required per patient (not including the initial fistula creation procedure). This is consistent with the limited studies examining procedure rates and reporting a mean of 1.3 and 2.7 procedures/patient. 11 , 22 Additionally, our bAVF maturation rate was 60%, which is similar to the rates reported in the literature for the haemodialysis population. 24 , 25 Finally, the complication and procedure rates in the bAVF group may appear high but remain no different when compared to the complication rate in the no‐bAVF group (which included CVC line removal for poor flows and CVC line infections). The combined benefits and risks of a bAVF creation again highlight the importance of patient and consumer perspective on this topic as well as emphasising the need for clear communication to patients during the informed consent process. The strengths of our study are in the development of a comprehensive district wide PD database with independent review of patient charts to verify data, ensuring a high level of data accuracy and reliability. This may not be possible in larger databases. We also examined clinically meaningful outcomes such as mortality, and to our knowledge, ours is the first study to do so in this cohort of patients. We focused and emphasised the risks of a bAVF, including the number of procedures required to maintain patency and the risk of non‐use. These insights are important for informing clinical decision‐making and guiding patient discussions around bAVF placement. We acknowledge the limitations of our study, which is retrospective and cannot prove the direction of causality or the effect of unmeasured confounders. Our study may also be underpowered in some outcomes (such as mortality and adverse events), despite showing statistically significant outcomes in other measures. Confounding was likely present, such as selection bias for when a bAVF is created in older and more frail patients who are deemed more likely to fail PD but are also more likely to die. In summary, the use of a bAVF in the PD population may offer benefits, such as an association with a lower transfer to haemodialysis with a CVC, consistent with what was previously reported in meta‐analyses. However, there is no association with a survival advantage or reduction in complication rates. We recommend that future studies examining the utility of a bAVF in patients on PD incorporate clinically meaningful outcomes, such as mortality, and be designed as a randomised controlled study. Only then can the role of bAVF in PD patients be definitively supported or refuted. Supporting information Table S1: Events per patient year stratified by back‐up AVF status for the complete cohort and the sensitivity analysis. Figure S1: Survival curves showing time to haemodialysis transfer with a CVC according to back‐up AVF status.