Non-recovery from dialysis-requiring acute kidney injury and short-term mortality and cardiovascular risk: a cohort study

Patients with end-stage renal disease (ESRD) experience a high burden of early death and cardiovascular disease (CVD), particularly within the first year after dialysis initiation (189 to 382 per 1000 person-years) [1‐7]. Furthermore, CVD contributes to > 50% of all ESRD deaths with known causes [8].

Recent literature suggests that acute kidney injury may also be associated with increased risk of CVD events [9‐16]. Acute kidney injury is associated with increased inflammatory cytokines [17], which may increase risk of subsequent CVD events [18] via endothelial dysfunction or plaque rupture [19]. Of note, abrupt decline in kidney function during the three months before initiating hemodialysis is a strong, independent risk factor for early death after ESRD onset [20]. Dialysis-requiring acute kidney injury (AKI-D) is not an uncommon precipitant to ESRD [21, 22] and may be an important contributor to the high early mortality rates and excess CVD risk in incident ESRD patients. However, few studies have examined outcomes in this ESRD subgroup.

In a diverse, community-based cohort, we examined the association between AKI-D and short-term death and CVD outcomes. We hypothesized that patients with incident ESRD due to AKI-D are at higher short-term risk for death and CVD events compared to patients with incident ESRD that is not attributable to AKI-D, even after controlling for known vascular risk factors. Furthermore, we hypothesized that patients who recover adequate kidney function after AKI-D to discontinue dialysis would experience better outcomes. Our goal was to provide insight into the clinical implications of AKI-D as well as recovery from AKI-D to improve risk stratification and guide development of more tailored preventative strategies.

Healthy People 2020 includes as key objectives to reduce all-cause and CVD-related mortality in ESRD patients [34]. While the U.S. Renal Data System tracks long-term mortality in ESRD patients, ascertainment of both early death and CVD events, as well as understanding potentially modifiable risk factors, has been much more challenging. Our study based within a large, community-based population helps to fill this important knowledge gap.

Our study has several key findings. First, a large fraction (one third) of incident ESRD patients in our cohort developed ESRD via AKI-D. Because we censored patients who died within 90 days of RRT initiation, one caveat is that the numbers of incident ESRD patients in both the AKI-D and non-AKI-D groups were reduced. It can also be clinically challenging to determine whether patients with chronic kidney disease who initiate dialysis during a hospitalization do so because of progression of underlying disease versus superimposed acute injury. However, prior single-center and VA data have reported similar proportions of incident ESRD patients initiating dialysis in the hospital [35‐37]. Therefore, our finding that patients with incident ESRD due to AKI-D have significantly higher adjusted all-cause death rates compared to incident ESRD patients without AKI-D indicates that the large number of patients who reach ESRD via AKI-D may explain, in part, the high early mortality rate in the overall ESRD population.

Second, we found that patients with incident ESRD due to AKI-D have higher rates of heart failure hospitalization after RRT initiation than ESRD patients without AKI-D. These findings extend prior observations that patients with abrupt decline in renal function before ESRD have higher rates of heart failure events in the three months before dialysis initiation compared to those who did not experience an abrupt decline [20], and patients who initiate dialysis due to volume overload have higher mortality rates [38]. In our cohort, patients with ESRD due to AKI-D were more likely to have pre-existing heart failure compared to patients with ESRD not due to AKI-D. Taken together, these data suggest that the higher mortality rates among those who experienced AKI-D leading to ESRD may be driven, in part, by a greater burden of heart failure. This finding suggests that better fluid management and more systematic use of evidence-based therapies for heart failure may potentially improve outcomes in this vulnerable population.

Third, our finding that AKI-D patients who recovered had lower adjusted mortality compared to those who did not recover highlights yet another reason we urgently need to devise treatment strategies to enhance or hasten recovery from AKI-D. Most prior studies evaluating long-term sequelae after recovery from AKI-D compared AKI-D patients to matched controls without AKI-D [13, 14], while we were able to distinguish outcome differences between those who did and did not recover. Although one small study reported that patients who recovered after AKI-D had lower 2-year mortality compared to patients who did not recover after AKI-D, our cohort was substantially larger and able to account for a larger set of potential confounders [39].

Finally, although we hypothesized that patients who recovered after AKI-D would have better outcomes, we found that even those who are able to discontinue dialysis remain at high risk for cardiovascular events, particularly hospitalizations for heart failure. We examined early outcomes after RRT initiation because mortality is particularly high during this time [8] and because AKI-D may be linked to elevated short-term CVD risk [16]. Several studies in the general population have suggested a transient increase in risk of stroke or myocardial infarction after an episode of infection or systemic inflammation [40, 41], and another study found that septicemia is a CVD risk factor in ESRD patients [42]. However, we did not find significant differences in outcomes that may be related to arterial plaque rupture or occlusion (e.g., ACS, stroke/TIA). Instead, AKI-D leading to ESRD was most strongly associated with all-cause death and excess heart failure events, which is also linked to aberrations in multiple signaling pathways related to inflammation, fluid overload, myocardial injury, and ventricular stress [43, 44].

Strengths of our study included the large number of AKI-D cases, which allowed us to examine the impact of renal recovery on clinical outcomes. In contrast to the majority of studies examining highly selected populations (e.g., post-operative acute kidney injury, contrast-induced nephropathy) [9], our study included a broad mix of AKI-D etiologies, which increases generalizability. We were also able to examine the pre-ESRD disease course (prior medication use, medical history, and other clinical features) to characterize incident ESRD subgroups and add important new insights on the transition to ESRD via AKI-D as well as the impact of recovery from AKI-D on CVD outcomes [8, 37, 39, 45, 46]. Furthermore, our data were much more rigorous than data collected via the Medical Evidence Report that is used for registration into the national U.S. Renal Data System [47‐50].

Limitations include that some patients classified in our study as having AKI-D may have initiated dialysis as inpatients because of progression of advanced chronic kidney disease rather than true AKI-D. However, for the comparison of patients with ESRD due to vs. not due to AKI-D, this type of misclassification would bias results toward the null, which bolsters our confidence in the differences we detected. We did not have details on inpatient dialysis prescriptions (e.g., modality, dose, or anticoagulation use), or etiology of AKI-D, including underlying cardiorenal physiology. Only insured patients from Northern California enrolled in an integrated health care delivery system were included, but the KPNC membership is highly representative of the local surrounding and statewide population, which argues for greater generalizability [23]. While we relied on electronic medical records for key data elements, our data sources are much more comprehensive than administrative databases and include clinical data elements (instead of primarily relying on diagnostic or procedural codes) [21, 28, 30‐32]. Because we wanted to explore how recovery from AKI-D influences outcomes, we were only able to compare outcomes that occurred between 90 days after RRT initiation and up to 1 year thereafter. Conditioning on survival to 90 days (in all three subgroups examined) may introduce immortal time bias, but this potential bias would likely decrease the observed effect size because the sickest AKI-D patients in the non-recovery group who died early would be excluded. We also had to exclude a small number (n = 182) of patients who were hospitalized for ≥ 90 days during their AKI-D hospitalization, but again these patients were likely the sickest AKI-D patients who would have been at highest risk for early death and CVD events. However, our design did preclude addressing potential differences in death and CVD event rates during the first 90 days, which may be higher as compared to the subsequent 1-year follow-up period.

In conclusion, we found that patients who transition to ESRD via AKI-D are at higher risk for short-term death and heart failure hospitalization compared to those who start chronic dialysis without experiencing AKI-D. Our findings indicate that the high short-term mortality in incident ESRD patients may be explained, at least in part, by the subset of patients who developed ESRD via AKI-D. Furthermore, our findings suggest that this subset of incident ESRD patients may have worse outcomes because of pre-existing heart failure. Recovery from AKI-D was also independently associated with lower short-term mortality. Collectively, these findings suggest that aggressive surveillance for and medical management of heart failure in ESRD patients may be a potential strategy for reducing early mortality after AKI-D. Further studies are warranted to evaluate whether promotion of renal recovery from AKI-D and systematic surveillance and intervention for heart failure can improve outcomes in this vulnerable patient population.