Prognostic relevance of hemodialysis for short-term survival in patients after LVAD implantation

End-stage heart failure (HF) is associated with renal failure (RF). This study aimed to determine the prognostic influence of RF and post-operative hemodialysis on short-term survival following left ventricular assist device (LVAD) implantation. This retrospective study includes 68 patients undergoing LVAD treatment. Kidney function was recorded prior to LVAD implantation, immediately afterwards and after 30 days, noting the need for hemodialysis. Median pre-operative Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) classification was 3.47 ± 1.08. 30 days after implantation there was a significant improvement of estimated glomerular filtration rate (eGFR) and reduction of blood urea nitrogen (BUN). Of pre-operative RF parameters, BUN was associated with increased mortality and need for early post-operative hemodialysis. Post-operative hemodialysis was associated with significantly lower short-term survival, while pre-operative hemodialysis did not impact mortality. Post-operative acute kidney injury (AKI) requiring hemodialysis can be regarded as a strong negative prognostic marker for short-term survival. The absence of a clear correlation between most routine RF parameters and survival or the need for early post-operative hemodialysis calls into question the predictive value of pre-operative RF. The negative association of only post-operative hemodialysis on short-term survival emphasises the impact of the occurrence of AKI.

Chronic kidney disease (CKD), however, often occurs in the context of CRS type 2 (CRS 2) 6 . After LVAD implantation, an initial transient recovery of eGFR is described in literature, but is followed by a late decline of eGFR 7 . Pre-implantation CKD as well as post-implantation acute renal failure are proposed to be a significant predictors of mortality [8][9][10][11] . Though an influence on RF has been hypothesised, existing studies do not consider different INTERMACS classes, which support stratification of patients with advanced HF receiving permanent MCS 12 . The present study assessed the influence of LVAD therapy on RF in patients with critical INTERMACS classes as well as the impact of hemodialysis on short-term mortality after LVAD implantation.

Materials and Methods
This retrospective, single centre observational study evaluates institutional data. Between April 2010 and January 2017, a total of 68 patients with symptomatic end-stage HF were enrolled under the following conditions: • New York Heart Association (NYHA) classification ≥3 • INTERMACS classification ≤5 • LVAD indication as bridge to transplantation or destination therapy Before initiating LVAD, cardiologists systematically investigated all patients and conservative heart failure therapy was optimized according European Society of Cardiology guidelines 13 .
Analysis of the patients' clinical course included full assessment of patient history, laboratory measurements, exertion capacity and need for hemodialysis. The key time-points for data were (i) before LVAD implantation, (ii) immediately following LVAD and (iii) at 30 days post-implantation. eGFR was calculated using the Modification of Diet in Renal Disease (MDRD) equation. Intraoperative data was also recorded. Indication for hemodialysis was confirmed by both, intensivists and nephrologists in every individual case. The study was approved by the ethical committee of the medical faculty of the University Heidelberg (S-663/2017). No specific tissue and/ or blood samples were taken at any time deviating from the routine therapy pre-, peri-and postoperatively. According to the ethical approval, no specific individual written consent was necessary for this retrospective analysis of data, which were already present by the routine medical follow up. No experiments on humans were performed, all medication as well as medical devices were completely legally approved by the time of application.
Statistical analysis was processed using the Statistical Package for Social Sciences version 24 (SPSS Inc., Chicago, IL, USA). Variables are given as continuous or categorical variables. Statistical analysis included Kolmogorov-Smirnov test to determine normality of the distribution. Continuous data were shown as mean ± standard deviation and analysed with the Student t-test for paired variables. Wilcoxon signed-rank test was utilized for categorical variables. Kaplan-Meier estimator was used to estimate survival function and log-rank test to compare the survival distributions. Cox-regression analysis and binary regression analysis, respectively, was performed for estimating the relationships among variables. Level of significance was α = 5%.
Timing of hemodialysis as risk factor for short-term mortality. Overall distribution of peri-operative hemodialysis is shown in Table 3. Within the 30 day-survival group, 37.0% (n = 20) needed early post-operative hemodialysis, while in the non-survival group (n = 14), 85.7% (n = 12) required early post-operative hemodialysis. At 30 days after LVAD implantation, 15 5, 30.4]) in the non-hemodialysis group (n = 38); (log-rank test P = 0.004), as illustrated in Fig. 2A. Of the patients that required hemodylasis pre-operatively (n = 6), one patient (16.7%) had recovery of RF immediately following LVAD. 5 of these patients received early post-operative hemodialysis, but 4 (66.7%) n %     Table 3. Distribution of hemodialysis peri-operatively (30 days period). Need for RVAD and higher BUN as additional risk factors. Pre-operative levels of creatinine and the eGFR also did not influence mortality (P = 0.600 and P = 0.222), but higher BUN was associated with increased risk of short-term mortality (HR 1.013, 95% CI: 1.000;1.027, P = 0.046). Lower pre-operative albumin levels also did not impact mortality (P = 0.569), neither did age (P = 0.183), gender (P = 0.239) or NYHA classification (P = 0.678). Intra-operative change of the right atrial pressure also did not influence mortality (P = 0.572).
Notably, different INTERMACS classifications did not significantly influence early mortality (P = 0.173). Simultaneous RVAD implantation significantly increased short-term mortality (HR 0.062, 95% CI: 0.07;0.544, P = 0.012), while overall concomitant heart surgery (P = 0.475) or re-operation (P = 0.455) had no effect on the mortality. Neither did the general intention of LVAD support (BTT vs. DT; P = 0.388). In a binary logistic regression analysis, the incidence of post-operative hemodialysis was also independent from patient's

Discussion
End-stage HF is associated with very high one-year mortality 14 , with a one-year survival described as low as 25% in the only medical therapy group 15 . Though heart transplantation remains the gold standard for the definite treatment of severe HF, the shortage of suitable donor organs results in the growing importance for LVAD support as a bridge to heart transplantation strategy 16 . Additionally, in those patients who are unsuitable for a heart transplantation, permanent LVAD is increasingly used as a support for the so called "destination therapy" 17,18 . In fact, destination therapy is growing and accounts for about 38.2% of implants 8 . Emerging research demonstrates the growing potential of LVAD therapy with described benefits even in patients already suffering from severe end-organ dysfunction 19 .
As the number of patients receiving LVAD therapy increases, it is imperative for all specialists involved to further explore the opportunities and benefits of using LVAD as well as better understand the limitations and risks of this therapy.
The annual INTERMACS report states an overall survival in adults with about 80% at 1 year and 70% at 2 years using LVAD with continuous-flow technology (>2,000 implants per year), approaching favourable outcomes after heart transplantation. However, within INTERMACS, up to 30% of LVAD-patients underwent heart transplantation within the first year after LVAD-implantation. Notably, 1-year survival is about 60% in patients still on LVAD support, according to the INTERMACS registry. In addition, Survival is significantly worse using LVAD in lower INTERMACS classes, especially in the short-term (HR 1,69 for INTERMACS 1 and HR 1,44 for INTERMACS 2). Other short-term risk factors identified by the INTERMACS registry are right heart failure (HR 2.57), history of cardiac surgery (HR 1.24) and concomitant cardiac surgery (HR 1.26) as well as unfavourable demographics (older age HR 1.03, female HR 1.32, higher BMI HR 1.10) 8 . Kirklin et al. also described the relevance of end-organ function on LVAD outcome. In particular, RF is known to be closely associated with survival after LVAD implantation. The 6 th INTERMACS report identified lower albumin levels (HR 0.90, short-term), higher BUN (HR 1.06, short-term and long-term) and higher creatinine (HR 1.05, long-term) as short and long-term negative predictive value for survival. Pre-operative hemodialysis is regarded as strongest negative predictor for survival in the short term (HR 2.37) 8 . However, the results of our study indicate that it is of crucial importance to distinguish between pre-and post-operative hemodialysis, with a strong impact of post-operative hemodialysis on short term mortality. In fact, to the best of our knowledge, the influence of the timing of hemodialysis after LVAD has not been addressed before and requires further investigation.
In the present study, we additionally identified pre-operative hypoalbuminemia and elevated BUN, as clinical markers for a deranged volume status, as independent risk factors for post-operative hemodialysis and in turn, the need for post-operative hemodialysis as a risk factor for short-term mortality following LVAD implantation.
It is difficult to identify authoritative cut-off lab values of renal parameters predicting post-operative kidney function. The MDRD equation tends to underestimate the eGFR for overweight patients and overestimate eGFR for underweight people, since there is no adjustment for body mass.
In the present study, RF monitored by eGFR and BUN significantly improved initially after LVAD implantation. It may be speculated that renal impairment due to CRS 1 (caused by poor renal perfusion) improves with a stabilisation of hemodynamic condition, as stated by Hasin et al. 20 . However, creatinine levels decreased post-operatively, failing to reach statistical significance. This might be explained by a better nutrition status and improved muscle mass after LVAD implantation. Nevertheless, in our cohort elevated pre-operative creatinine levels were not associated with a significantly higher mortality. These findings are reflect the results of the annual INTERMACS report, where creatinine was not associated with a higher short-term mortality, whereas BUN did influence short-term outcome 8 . Relevant literature, however, often includes both, serum creatinine levels and BUN, as variables in temporary LVAD risk scores 21,22 . Nevertheless, we could only confirm pre-operative BUN as relevant short-term marker for survival and would encourage re-evaluating about the value of peri-operative retention parameters.
Kirklin et al. have described the unfavourable prognostic relevance of renal failure before LVAD implantation 8 . In this study, we could confirm only that post-operative hemodialysis was a negative prognostic marker for short-term survival after LVAD. Conceivably, post-operative renal failure with the need for hemodialysis serves SCIEntIfIC REPoRTS | (2018) 8:8546 | DOI:10.1038/s41598-018-26515-0 as an early indicator for an insufficient organ perfusion and the subsequent development of multiple organ failure, the timing of hemodialysis seems to be essential to determine prognosis.
Regarding the connection between an unfavourable pre-operative cardiac condition and the need for immediate post-operative hemodialysis, patients were divided into INTERMACS subgroups and analysed according to need for post-operative hemodialysis in a binary logistic regression model. In fact, no statistical significant differences in incidence of post-operative hemodialysis could be demonstrated between critical INTERMACS classifications. Thus, in our patient population, we conclude that poor pre-operative INTERMACS classification does not give any further distinct information about potential renal failure requiring hemodialysis after LVAD implantation. It is of great significance that different low INTERMACS classifications did also not predict short-term mortality sufficiently. It could be suspected, that in critically unwell patients the INTERMACS differentiation might be blurred and INTERMACS classification is only a part of the total clinical picture. Interestingly, concomitant surgery, which is associated with longer procedure time due to the more complex surgery, did not result in more post-operative hemodialysis. Aiming for optimization of cardiovascular status (e.g. by treating intracardiac shunts or valve disease) results in a better hemodynamical status and might preserve kidney function, thus cardio-pulmonary bypass and procedure time (including hypothermic period) might cause damage to the kidneys. A larger study population would be required to explore this further.
Our data suggest that in the case of CRS 1 there is a clear potential for significant recovery of RF according to kidney function parameters after LVAD. This result is in concordance to the literature [23][24][25] . Within our cohort, even those patients receiving pre-operative hemodialysis recovered within 30 days after LVAD, suggesting better potential for kidney recovery contradicting previous findings 26 . However, within our analysis, pre-operative renal retention parameters such as creatinine and eGFR did not determine the need for hemodialysis in the early post-operative course. In terms of INTERMACS classifications, no difference in renal outcome was monitored between subgroups. In contrast, increased pre-operative BUN and impaired pre-operative albumin are the only predictors for post-operative need for renal replacement therapy. Post-operative hemodialysis, on the other hand, must be regarded as a strong negative predictor for mortality even in the short-term. Kidney function serves as an early indicator for insufficient cardiac output, hence, early detection might prevent early-mortality, especially in patients undergoing LVAD implantation in low INTERMACS levels. As a result, acute post-operative kidney injury is of great prognostic value in the early post-operative period. Notably, none of the patients included in this study suffered from a genuine chronic kidney disease pre-operatively other than CRS. It remains crucial to clearly determine the origin of kidney dysfunction prior to LVAD implantation, as the nature of CKD alters prognosis for recovers sustainably.
There is a strong case to be argued that closer cooperation between cardiac surgeons and cardiologists with specialization in nephrology could help preserve kidney function, particularly as it remains difficult to determine the prognosis of kidney function after LVAD implantation. In addition, a pre-operative and intra-operative approach to optimize volume status is an area that the authors believe must receive increased attention due to its potentially important role in the post-operative course. Most importantly, the diagnosis of pre-operative AKI does not influence outcome after LVAD. Instead, the significant influence of post-operative AKI with subsequent need for hemodialysis should be emphasised. To clearly distinguish the course of AKI and timing of hemodialysis is a key factor to understand the impact of renal function on short-term outcome after LVAD. Additional clinical trials are needed to investigate the impact of different circulatory support strategies on prognosis of end-stage heart failure patients and to help improve short-term and long-term outcomes for these critically ill patients.

Limitations
The present study reflects results of single-centre retrospective study with a limited number of patients involved. Furthermore, we were not able to conduct power calculations at the beginning of the study to estimate a sample size with sufficient power. However, post-hoc power analysis yielded a power of 0.94, according to our observation of 85.7% hemodialysis requirement early postoperatively in 30-day non-survivors compared with 37.0% postoperative hemodialysis need among 30-day survivors. Therefore, our investigated cohort of 68 patients was sufficient to address our hypotheses. However, our data impose the "real-life" constraint of early LVAD success limited by end-organ failure.