Arteriovenous fistula for haemodialysis as a predictor of de novo heart failure in kidney transplant recipients

ABSTRACT Background The haemodynamic effects of a functioning haemodialysis arteriovenous fistula (AVF) can cause or exacerbate heart failure (HF). We investigated whether the presence of an AVF at the time of kidney transplant (KT) is associated with de novo HF. Methods This was an observational cohort study including adult patients who received a KT in the West of Scotland between 2010 and 2020. We evaluated the risk and associations of pretransplant factors with de novo HF, alone and as a composite cardiovascular (CV) outcome (including non-fatal myocardial infarction, non-fatal stroke, de novo HF and CV death). Multivariable proportional hazards regression and sensitivity analyses were used to identify independent correlates of the outcomes. Results Among 1330 included patients, the incident rate of de novo HF after transplantation was 58/1000 person-years [95% confidence interval (CI) 50–67] in AVF patients (n = 716) compared with 33/1000 person-years (95% CI 27–41) in non-AVF patients (n = 614). De novo HF was associated with the presence of an AVF [adjusted hazard ratio (aHR) 2.14 (95% CI 1.40–3.26)], duration of dialysis [aHR 1.03/year increase (95% CI 1.01–1.04)], age at transplant [aHR 1.03/year increase (95% CI 1.02–1.05)], female sex [aHR 1.93 (95% CI 1.40–2.65)] and pretransplant diabetes [aHR 2.43 (95% CI 1.48–4.01)]. The presence of an AVF was also associated with the composite CV outcome [aHR 1.91 (95% CI 1.31–2.78)]. Conclusions The presence of an AVF may be an underrecognized modifiable predictor of de novo HF posttransplantation.


INTRODUCTION
Although kidney transplantation considerably improves survival and quality of life when compared with chronic dialysis treatment [1 ], cardiovascular disease ( CVD) remains the major cause of death among kidney transplant ( KT) recipients [2 ].Heart failure ( HF) whether pre-existing or newly diagnosed posttransplant, is associated with increased mortality after transplantation [3 , 4 ].Studies have shown that the prevalence of pretransplant HF with reduced ejection fraction ( HFrEF) in patients referred or wait-listed for transplantation may be as high as 25% [3 , 5 , 6 ].Data on the prevalence of de novo HF posttransplant are outdated and limited due to universal lack of echocardiographic data in reported studies, which use clinical criteria or coded diagnoses for HF [4 , 7 , 8 ].Notably, more recent studies have observed a reduction in de novo HF posttransplantation over time, most likely as a result of improvements in pretransplantation cardiac risk stratification and modification [9 , 10 ], dialysis care [11 ] and contemporaneous improvements in posttransplantation kidney care [8 , 12 , 13 ].This occurs despite the aging and increased comorbidity burdens among KT recipients [14 ].
Traditional risk factors associated with de novo HF posttransplant include age, female, African American, smoking, obesity, atherosclerotic disease, diabetes and hypertension [4 , 7 , 15 ].Non-traditional risk factors include delayed graft function, suboptimal KT function and donor-related factors [4 , 7 , 15 ].The haemodynamic effects of a functioning haemodialysis ( HD) arteriovenous fistula ( AVF) can cause or exacerbate HF in patients with end-stage kidney disease ( ESKD) [16 -18 ] and studies advocate AVF ligation posttransplantation to mitigate these effects [19 -22 ].The effect of a functioning AVF pretransplantation in the development of de novo HF posttransplantation has not been investigated.
The goal of this study was to evaluate the risk of de novo HF posttransplantation using echocardiographic data and determine pretransplant demographic and clinical factors predictive of de novo HF.This study sought to specifically determine whether receiving dialysis via an AVF before transplantation is associated with new-onset HF posttransplantation.

Data sources
Data were obtained from the West of Scotland Electronic Renal Patient Record ( SERPR) to identify all KT recipients and linked with the Scottish Electronic Health Records using the Community Health Index number.The West of Scotland KT program is a regional service including five health boards (

Study population
We included adult ( ≥18 years of age) recipients of a KT who underwent transplantation from 1 January 2010 to 31 March 2020.During this era, all patients were treated with the SYM-PHONY immunosuppressive regime of induction with an anti-interleukin-2 receptor monoclonal antibody ( or antithymocyte globulin in case of high immunological risk) , an antiproliferative, tacrolimus and prednisolone, except where clinical circumstances dictated alternative therapy [23 ].Patients with pre-existing HF ( before undergoing kidney transplantation) were excluded.We also excluded patients with missing data on the dialysis modality and type of vascular access used prior to transplantation and those receiving dialysis via a synthetic graft.We ended observation on 1 April 2021, to allow at least 1 year for detection of the primary outcome.We restricted our study in the pre-coronavirus disease 2019 era to avoid the impact of the pandemic on our transplant program and healthcare services.As we analysed routine care data, institutional ethical approval was waived but the study was approved by the data protection officer on behalf of the NHS Greater Glasgow & Clyde health board ( Caldicott Guardian approval reference number NHSGGC/1061/11Jun21) .

Baseline variables
Information on age, gender, blood pressure, body mass index ( BMI) , cause of ESKD, pretransplantation dialysis modality and vascular access, duration of dialysis, diabetes, ischaemic heart disease, atrial fibrillation, cerebrovascular disease, peripheral arterial disease and HF diagnoses before and after transplantation was obtained through the SERPR.The most recent blood results up to 90 days pretransplantation, medications ( including immunosuppressants) up to 90 days posttransplantation, mortality and KT outcomes were also retrieved.All echocardiography and chest X-ray reports and N-terminal prohormone of brain natriuretic peptide ( NT-proBNP) levels were retrieved from the electronic patient records.The last echocardiography before transplantation ( performed within 2 years) and the most recent posttransplantation ( performed within 5 years) were used to define HF outcomes.Structured Query Language ( SQL) and visual coding strategies were used to capture relevant 'text strings' from echocardiography and chest X-ray reports and physician reported diagnoses.

Outcome definitions
Primary outcomes included de novo HF, alone and as a composite cardiovascular ( CV) outcome.HF was defined using a combination of echocardiography criteria, as per the 2021 European Society of Cardiology ( ESC) guidelines for the diagnosis and treatment of acute and chronic HF [24 ], physician-reported diagnoses and radiological criteria ( Table 1 ) .The same criteria were used to define HF pre-and posttransplantation.We included patients with both HFrEF and HF with preserved ejection fraction ( HFpEF) .
To define HFpEF we used markers of structural heart disease [left ventricular hypertrophy ( LVH) or left atrium ( LA) enlargement] or diastolic dysfunction.We have used measures of the LA area as a marker of LA enlargement, as LA volumes were not available in our database to estimate the LA volume index ( LAVI) [25 ].The CV outcome included a composite of non-fatal myocardial infarction ( MI) , non-fatal stroke, de novo HF and CV death ( death from MI, stroke, HF or arrhythmia) .Secondary outcomes included death, KT failure and death-censored KT failure.Observations were censored at lost to follow-up, death, KT failure or end of observation.

Patient allocation
We divided the patients into two groups based on the dialysis modality and type of vascular access used prior to transplantation: patients on HD via an AVF versus all the rest [patients on HD via a central venous catheter ( CVC) , on peritoneal dialysis ( PD) , with a prior KT or low clearance].In a subgroup analysis, we only included patients on HD ( dialysis via an AVF versus CVC) , as patients with pre-emptive transplantation or on PD tend to have better outcomes and lower mortality after KT.The last renal replacement therapy modality and last dialysis access entries before the date of transplantation were used to ascertain the use of an AVF ( or not) immediately prior to transplantation.As a verification measure, the first dialysis access used within the first 7 days of transplantation ( for patients with delayed graft function) was also retrieved.

Statistical analyses
Continuous variables were presented as either mean with standard deviation ( SD) or median with interquartile range ( IQR) .Categorical variables were expressed as percentages.
Due to the different durations of follow-up, incidence rates of de novo HF were estimated using person-time ( years) of the population at risk.
Values were missing in < 20 individuals, except for primary renal disease We plotted survival curves using Kaplan-Meier estimators and applied proportional hazards ( Cox) regression models to assess the effect of a functioning AVF at the time of transplantation compared with no AVF on the risk for de novo HF, alone and as a composite CV outcome.In a sensitivity analysis, we used Fine-Gray models [26 ] to assess the effect of AVFs on de Receiver operating characteristics ( ROC) curves were generated to determine the predictive value of AVF, age and duration of dialysis in de novo HF.
For all analyses, P -values of .05 were considered statistically significant.All analyses were performed using Stata 16.1 ( StataCorp, College Station, TX, USA) .

Baseline patient characteristics
We identified 1670 patients who underwent kidney transplantation during the study period.From this cohort, we excluded 295 patients with pretransplantation HF and 24 patients who were receiving HD via an AV graft.Another 21 patients were excluded due to missing data ( Fig. 1 ) .The remaining 1330 patients served as the study cohort.The demographic and clinical characteristics of study participants by AVF status at the time of transplantation are summarized in Table 2 .Males were more likely to have an AVF compared with females and patients with an AVF had been receiving dialysis for longer.In the non-AVF group, 220 ( 35.8%) patients received a pre-emptive KT.The median time between the last echocardiography and transplantation was 13.6 months ( IQR 6.  and the median time between transplantation and the most recent echocardiography was 39.8 months ( IQR 15.6-60) .The main echocardiographic characteristics and NT-proBNP levels before and after transplantation are shown in Table 3 .Regression of LVH and a reduction in LV volumes were observed post-transplantation despite no differences in left ventricular ejection fraction ( LVEF) , and these changes were more pronounced in the no-AVF group.Furthermore, subtle changes in measures of elevated left ventricular filling pressures were noted post-transplantation in both groups [elevated LA area, E:e ( early filling velocity on transmitral Doppler/early relaxation velocity on tissue Doppler) ratio and TR peak velocity].

Incidence of de novo HF after transplantation
Over a median follow-up of 4.2 years, de novo HF occurred in 175/716 ( 24.4%) patients with an AVF at the time of transplantation compared with 91/614 ( 14.8%; P < .001)patients with no AVF.Similarly, the composite CV outcome occurred in 201/716 ( 28.1%) patients with an AVF compared with 107/614 ( 17.4%; P < .001)patients with no AVF.Secondary outcomes were more frequent in the AVF group ( Table 4 ) .
The incident rate of de novo HF after transplantation was 46/1000 person-years ( 58/1000 person-years in patients with an AVF versus 33/1000 person-years in patients with no AVF; Table 5 ) .For the composite CV outcome, the incident rate was 55/1000 person-years ( 68/1000 person-years in patients with an AVF versus 40/1000 person-years in patients with no AVF; Supplementary Table S1) .
The cumulative incidences of de novo HF and the composite CV outcome using Kaplan-Meier failure estimates are shown in Fig. 2 .

Independent predictors for de novo HF
Multivariable analysis demonstrated five factors predicting de novo HF: the presence of an AVF, the duration of dialysis, increasing age, female sex and pretransplant diabetes.When adjusted for demographic data, duration of dialysis, primary renal diagnosis, comorbid conditions, haemoglobin and albumin levels and medications ( angiotensin-converting enzyme inhibitors or angiotensin receptor blockers and immunosuppressants) , the risk for developing de novo HF was higher for patients with an AVF at the time of transplantation compared with those with no AVF {adjusted hazard ratio [aHR] 2.14 [95% confidence interval ( CI) 1.40-3.26]}when death and KT failure were treated as censoring events ( Table 5 ) .In sensitivity analysis that treated death and KT failure as competing events, the risk for de novo HF was also higher in patients with an AVF [aHR 1.91 ( 95% CI 1.28-2.86)] ( Table 5 ) .The presence of an AVF was also predictive of the composite CV outcome following adjustment of covariates [aHR 1.91 ( 95% CI 1.31-2.78)] ( Supplementary Table S1) .

Sensitivity analyses
The presence of an AVF was independently associated with both de novo HFpEF [ n = 171; aHR 2.13 ( 95% CI 1.34-3.38)] and HFrEF [ n = 95; aHR 2.21 ( 95% CI 1.10-4.46)] ( Supplementary data, Tables S5 and S6) .Subgroup analysis in HD patients ( n = 923) showed a higher adjusted risk of de novo HF in patients dialysing with an AVF at the time of transplantation compared with those dialysing via CVC [aHR 2.13 ( 95% CI 1.40-3.25)] ( Supplementary data, Table S7) .Further subgr oup anal ysis in HD patients dialysing via an AVF ( n = 716) showed no difference in the adjusted risk of de novo HF by AVF type ( Supplementary data, Table S8) .ROC curves assessing the presence of an AVF pretransplant, age at transplantation and duration of dialysis showed that each of these individual factors was mildly predictive of de novo HF.When all three factors were considered collectively, however, the model was fairly predictive of de novo HF, with an area under the curve of 0.66 ( 95% CI 0.63-0.70)( Fig. 3 ) .

DISCUSSION
This study assessed the prevalence of de novo HF after kidney transplantation and the predictive value of a functioning AVF in de novo HF in a contemporary transplant population, with several noteworthy findings.First, integrating echocardiographic criteria, we reported outcomes of HF including HFpEF, which is known to be the most prevalent form of HF in CKD patients [27 ].Second, using multivariable analysis, the presence of an AVF was strongly associated with de novo HF and adverse cardiovascular outcomes posttransplantation and the effect was independent of exposure to dialysis.Finally, in the multivariable model, age, female sex, duration of dialysis and the presence of pretransplant diabetes were also predictive of de novo HF.These observations have important implications and challenge current practices where AVFs are traditionally maintained in asymptomatic patients after successful kidney transplantation.
Data on the prevalence of HF before kidney transplantation are sparse and mostly derived from the US registry and single-centre studies.The reported prevalence of HF at the time of KT varies between studies and ranges from 5.8 to 11.9% [5 , 6 ].Lentine et al. [4 ] reported cumulative incidences of HF on the transplant waiting list of 6.5%, 12% and 32% at 6, 12 and 36 months, respectively.In another study, the prevalence of physician-reported HF in patients on the transplant waiting list was 46% at 36 months [28 ].In a study using single-photon emission computed tomography ( SPECT) myocardial perfusion scans for pretransplant assessment, 18% of patients had LV systolic dysfunction ( defined as an LVEF ≤45%) [3 ].In studies reporting long-term outcomes, HF at the time of transplantation was independently associated with a higher risk of death, cardiovascular events and KT failure [3 , 4 , 6 ].
In patients with a KT, HF has been most commonly defined in administrative and clinical databases and there are no studies using diagnostic testing, such as echocardiography.In a study including KT recipients between 1995 and 2001, Lentine et al. [4 ] reported cumulative incidences of de novo HF of 10.2% and 18.3% at 12 and 36 months, respectively, which is slightly higher than reported in our study ( 6.8% and 12.0%, respectively) .This may be explained by potential misclassification of HF diagnoses when using registry International Classification of Diseases, Ninth Revision-based claims or it may represent a true decline in HF as demonstrated in a recent study, where the adjusted incidence of de novo HF after KT significantly declined between 1998 and 2010 [8 ].Rigatto et al. [7 ] reported a cumulative incidence of de novo HF of 3.6% at 5 years ( versus 17.2% in our study) , but they only included patients who were alive and free of cardiac disease at 1-year posttransplant, thereby eliminating the highest risk interval.Studies investigating the risk for HF in specific populations have shown an increased age-adjusted risk for HF hospitalizations in African American compared with Caucasian KT recipients [15 ].In a study investigating the cardiac implications of obesity in KT recipients, the 5-year cumulative incidence of HF increased significantly from 3.6% to 18.4% from the first to fourth BMI quartile [28 ].
Traditional risk factors associated with HF after transplantation [4 , 7 , 15 ] are also present in the general population with HF and share common pathophysiologic mechanisms [29 , 30 , 31 ].In our study, older age ( > 50 years) , female sex and the presence of diabetes were independently associated with do novo HF posttransplant.Non-traditional risk factors pertinent to the transplant population are related to chronic kidney disease ( CKD) , immunosuppressive agents and donor-related factors and include increased duration of dialysis before transplant, deceased donor kidney, increased donor age, delayed graft function, allograft rejection and KT failure [4 , 7 , 15 ].This is the first study to report that the presence of a functioning AVF at the time of transplantation is a predictor of de novo HF, showing that the haemodynamic demands of an HD AVF are long-lasting and may precipitate HF even when not used for dialysis.A striking finding in our adjusted analysis was that the effect of an AVF was independent of dialysis duration, which was also a predictor of de novo HF.The presence of an AVF was also independently associated with a composite of non-fatal MI, non-fatal stroke, de novo HF and CV death.
We observed a higher incidence of de novo HF in patients receiving dialysis via an AVF compared with a CVC.This is suggestive of a deleterious effect of an AVF on the cardiovascular system rather than HD per se.Although this is an important finding, we should emphasize that patients receiving dialysis via an AVF are inherently different from those dialysing via a CVC.There are  often clinical reasons why patients are more likely to have a CVC than an AVF, including old age and frailty, but also younger patients with a live donor who are expected to receive dialysis for a short period.One small cohort study suggested that patients with an AV access had an increased risk of HF compared with a CVC, even allowing for CVC patients been older [32 ], but this is not necessarily a consistent finding [33 ].Although patients with a fistula blood flow > 2 l/min or an upper arm AVF are traditionally at increased risk for the development of HF [34 ], in our study an AVF configuration was not associated with the risk of de novo of HF.However, data on fistula flows were not available for comparisons.
Studies have reported the reversal of LV remodelling and clinical cardiac dysfunction after kidney transplantation [35 -39 ] driven by correction of the uraemic state and abolishment of the adverse effects of prolonged dialysis on myocardial function.KT recipients have a lower CV death risk and survive longer than patients on the transplant waiting list [40 , 41 ], suggesting that the progression of CVD can be ameliorated by restoring kidney function with a transplant.In contrast, persistence of a functioning AVF after kidney transplantation has been associated with increased LV mass and LV dimensions [42 ], while closure of the fis-tula results in significant reductions in LV mass, LV end-diastolic diameter and atrial sizes [19 -21 ].
This study is limited by its retrospective design, and as is the case with all observational studies, residual confounding is a possibility.Although we have used objective criteria of cardiac structural, functional and serological abnormalities to define HFpEF, we were unable to obtain data on LA volumes, which is a better marker of LA enlargement.Another limitation is that echocardiography was not routinely performed in all KT recipients but only in patients with symptoms suggestive of HF ( by indication) and, as such, the exact time of HF diagnosis could not be accurately determined.Although AVFs were functioning at the time of transplantation, whether these fistulas were still functioning months or years posttransplant could not be verified.Despite its limitations, this study is strengthened by the use of echocardiographic data rather than clinically coded diagnoses for HF, the quality of data in our prospectively maintained database, which may approach that obtainable in a prospective cohort study, and the nearly complete follow-up.We also avoided misclassification of pre-existing HF as ' de novo ' because of underascertainment of pretransplantation status, which is common in registry data.Finally we have reported on measures of LV diastolic dysfunction beyond the LVEF and, in fact, LVEF does not reflect the complex haemodynamic adaptations of the heart to the AVF [34 , 43 , 44 ].
Our study suggests that the presence of a functioning AVF at the time of transplantation may be a modifiable underrecognized risk factor for de novo HF after transplantation.Older age, female sex, duration of dialysis and pretransplant diabetes also predict de novo HF.The risk of de novo HF in the presence of an AVF should be carefully evaluated as part of routine pretransplant assessment.Towards this, a systematic evaluation has recently been proposed suggesting prophylactic AVF ligation starting at 12 months after transplantation if risk factors for HF are present [45 ].Whether closure of an 'unnecessary' AVF in stable KT recipients would prevent HF could be an area for future research and must be weighed against its generalizability to diverse transplant populations, the risk of losing a dialysis access site and potentially high implementation costs.

Figure 1 :
Figure 1: Creation of the study cohort.

Figure 2 :
Figure 2: Survival curves were plotted using Kaplan-Meier estimators.HRs and 95% CIs were estimated using Cox proportional hazards regression models.The models were adjusted for baseline factors at the time of transplant.( A) HF. ( B) Composite cardiovascular outcome ( non-fatal MI, non-fatal stroke, de novo HF and CV death) .

Figure 3 :
Figure 3: Combining age at transplant, duration of dialysis and the presence of an AVF at the time of transplantation in the model was fairly predictive of de novo HF ( AUC 0.66) .AUC: area under the curve.
NHS Greater Glasgow & Clyde, Forth Valley, Dumfries & Galloway, Lanarkshire, Ayrshire & Arran) and occasionally other parts of Scotland covering a population of 1.7 million.Since the year 2000, > 3000 KTs were performed in the Glasgow Transplant Unit and 150-200 patients are transplanted each year.This study is reported according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines ( see Supplementary data) .