Nephrotoxin Exposure in the 3 Years following Hospital Discharge Predicts Development or Worsening of Chronic Kidney Disease among Acute Kidney Injury Survivors

Acute kidney injury (AKI) affects up to 40% of hospitalized patients and carries a 3.5-fold increased risk of mortality [1-3]. Beyond short-term consequences of AKI, including volume overload and electrolyte abnormalities, a single episode of AKI independently predicts recurrence of AKI, progression to chronic kidney disease (CKD), dialysis dependence, and death [4-7]. Increasing evidence also indicates that AKI heightens the risk of cardiovascular events [8]. Given the myriad of potential AKI sequelae, it is imperative that adequate post-AKI care is provided [9].

Despite the presence of strong evidence supporting the association between AKI and worsened outcomes, little is known about the appropriate short- and long-term management of post-AKI patients [10]. During the course of hospitalization, AKI is often viewed as a transient event with even the most robust of decision support sequences ending abruptly upon resolution [11, 12]. With such circumstances, suspension of directed care may allow for additional avoidable kidney insults to occur, as there is a potential lack of perceived risk when kidney function parameters return to baseline. In many cases, AKI resolves before discharge and is no longer considered an active problem requiring additional intervention [13, 14]. Yet it is established that even when kidney function markers, such as serum creatinine (SCr), return to baseline before discharge, patients still experience a heightened risk for adverse kidney outcomes [15]. In the hospital setting, clinical and laboratory data are often evaluated on a daily basis, but upon discharge, a marked gap occurs in the care continuum for AKI patients who may receive no kidney-specific follow-up.

One aspect of AKI follow-up care is nephrotoxin management [16, 17]. Nephrotoxic medication use is ubiquitous among patients hospitalized with AKI. Nearly two-thirds of all medications are excreted extensively by the kidneys, and 23% are potentially nephrotoxic [18-20]. The Acute Disease Quality Initiative group recommended that patients with at least stage 1 AKI that persists for more than 7 days should receive comprehensive medication reconciliation and follow-up at discharge, yet available data suggest that medication optimization in the post-AKI period remains an unmet need [10, 21]. Therefore, the purpose of this study was to assess the impact of nephrotoxin exposure post-hospitalization for AKI to determine if an opportunity for care optimization exists.

Using data from the Rochester Epidemiology Project (REP), we performed a population-based cohort study of AKI survivors in Olmsted County, Minnesota, to determine the long-term risk of adverse kidney outcomes after exposure to nephrotoxins. IRB approval was obtained from both Mayo Clinic and Olmsted Medical Center. Included individuals were adult residents of Olmsted County, MN (≥18 years) with an episode of AKI during a hospitalization detected by electronic surveillance (see below) between January 1, 2006, and December 31, 2014. Longitudinal follow-up data from the REP was used to comprehensively capture outcomes similar to our previously described approach²⁶.

The REP is a medical record linkage system established in the 1960s that amasses data from all of the main healthcare providers in Olmsted County, allowing for virtually complete ascertainment of medical data for local residents. The REP includes demographic data and comprehensive information about medical diagnoses, hospital admissions, surgical procedures, outpatient follow-up care, and prescription data from all participating institutions regardless of patient characteristics, including insurance status.

AKI cases were identified and staged with an electronic surveillance tool developed and validated against a manually collected reference standard that uses both SCr and urine output criteria using the Acute Kidney Injury Network criteria [22-24]. Baseline creatinine was calculated as the median of all creatinine values in the 6 months to 7 days before the index admission or if unavailable, estimated from back-calculation using the modification of diet in renal disease formula and an assumed estimated glomerular filtration rate (eGFR) of 60 mL/min/1.73 m² [25]. We excluded any individual who died in the 90 days after AKI development because the focus of the study was de novo or progressive CKD, which requires sustained evidence of kidney decline benchmarked at 3 months [26]. Patients with persistent dialysis dependence in the 3-month post-discharge interval or those who underwent kidney transplantation before, during, or in the 3 months after AKI admission were excluded. We also excluded patients who declined the use of their medical record for research, lacked any medical follow-up in the 3 years after their hospitalization, or did not remain an Olmsted County resident during the study period [27].

The included cohort was followed for 3 years after discharge to characterize time-dependent use of potentially nephrotoxic medications. Potentially nephrotoxic systemic medications included proton pump inhibitors (PPIs), select antibiotics, antifungals, antivirals, nonsteroidal anti-inflammatory drugs (NSAIDs), chemotherapeutic agents, and immunosuppressants (online suppl. Table S1; see www.karger.com/doi/10.1159/000522139 for all online suppl. material). The medication list was validated and reconciled by study investigators, including pharmacists with expertise in evaluating drug-associated kidney injury, as well as with primary and tertiary references [28, 29]. Angiotensin converting enzyme inhibitors/angiotensin receptor blockers (ACEI/ARBs), were omitted from this medication list due to recent literature which suggests no relationship between use and worse kidney outcomes in the post-AKI period [30-32]. Further, use of ACEI/ARB is confounded by indication as the conditions they are associated with also affect kidney outcomes (i.e., heart failure, hypertension). For similar reasons, diuretics were omitted from this list [33]. Renally eliminated medications that are not known to be nephrotoxic were not included in the analyses.

Nephrotoxin use was identified through an electronic data pull from the REP. In cases of discrepancy, such as two prescriptions for the same medication class during a time period, discharge medication lists, prescription records, and clinical notes were manually reviewed to assess use. The total fill period of a prescription was used to calculate days of therapy. In cases where an order was placed to discontinue a prescription medication before completion of the total fill days, the duration was truncated to the medication discontinuation date. Medication use was analyzed in a time-dependent manner. Patients were classified in the “use” group only during the time periods that they were prescribed one or more potentially nephrotoxic mediations. If this period did not last the entirety of the 3-year follow-up, patients were dynamically switched into the “no-use” group as medications were stopped and started (Fig. 1). These formed the two comparators, “use” and “no-use” of nephrotoxic agents. Patients without any potentially nephrotoxic medication use throughout the duration of follow-up were kept in the “no-use” group throughout. By using the same patients but allowing their exposure status to change over time, we reduced the risk that differences found among groups were due to factors outside of nephrotoxin use (i.e., disease state or severity).

We captured patient demographics at baseline and comorbid conditions (e.g., cardiovascular disease, heart failure, peripheral vascular disease) using ICD 9/10 codes (online suppl. Table S2). Hospital length of stay, maximum AKI severity (Stage 1, 2 or 3), duration of AKI, urine protein, kidney function based on SCr, eGFR, rehospitalization, and death were assessed during the study period. We also determined whether a patient developed transient or persistent AKI (<48 h or ≥48 h, respectively) [10].

The primary outcome of de novo or progressive CKD was identified using multiple approaches to account for baseline preexisting kidney dysfunction and potential detection bias from variable follow-up in the outpatient setting [34]. Given that the diagnostic criteria for CKD requires persistent kidney dysfunction for at least 90 days, we reviewed medical records for the study end point from 90 days to 3 years following the AKI event. Individuals that did not survive to 90 days after the AKI event were excluded, as they would not be able to fulfill the criteria for the diagnosis of CKD. An additional motivation for this strategy is dismissal medication lists for individuals with an expected lifespan of <90 days are often designed to focus on comfort and quality of life. Therefore, they may not be reflective of the study aims. Patients met the end point of de novo or progressive CKD on the first date that one or both of the following criteria were true:

First, incident CKD, end-stage renal disease, or kidney transplantation was identified using an a priori defined list of diagnosis codes (online suppl. Table S3). As these diagnosis codes rarely accurately stage the severity of kidney dysfunction, the focus with this approach was to capture incident CKD as a binary event. For this aspect of the analysis, any patient with preexisting kidney dysfunction based on these codes or eGFR <60 mL/min/1.73 m² prior to hospital admission was excluded.

Second, calculated eGFR was used to determine new or worsening kidney dysfunction from baseline. All patients with follow-up creatinine concentrations could be considered in this analysis. Preadmission eGFR was calculated using the median baseline creatinine in the 6-months to 7 days prior to hospital admission and the Modification of Diet in Renal Disease equation. Upon dismissal, SCr values were serially evaluated and eGFR was calculated. Patients with an eGFR <60 mL/min/1.73 m² were staged based on standardized criteria (Stage 3a: 45–59 mL/min/1.73 m²; Stage 3b: 30–44 mL/min/1.73 m²; Stage 4: 15–29 mL/min/1.73 m²; Stage 5: <15 mL/min/1.73 m²). SCr values were only included if drawn in the outpatient setting, >7 days before or after an inpatient encounter to avoid classifying transient changes in kidney function as CKD. Medical records were reviewed for the 3-year follow-up interval or until the last available clinic visit or SCr concentration in that timeframe, whichever was sooner. New or worsening kidney dysfunction was defined as at least one measurement with a 30% decline in eGFR from baseline; which was necessary even for those with a new eGFR < 60 mL/min/1.73 m². This criterion was established to avoid classification of individuals with very small changes in eGFR as positive for the outcome of interest (i.e., 61 mL/min/1.73 m² to 59 mL/min/1.73 m²).

Univariable Cox proportional hazards regression models were fit to evaluate the time-dependent incidence of de novo or progressive CKD during the 3 years of follow-up. A composite model was also developed to assess the incidence of de novo or progressive CKD, readmission with AKI, or death. Multivariable models were fit to evaluate the association of nephrotoxin use with long-term kidney outcomes after adjusting for baseline risk [8, 15, 35, 36]. Included risk factors were nephrotoxin use as a time-dependent variable, age, sex, body mass index, baseline eGFR, history of cardiovascular disease, heart failure, or peripheral vascular disease, maximum AKI stage, presence of persistent AKI, and change in eGFR during index hospitalization.

Of the 4,551 patients screened for eligibility, 2,461 were included (Fig. 2). The primary reasons for exclusion were death within 90 days of AKI development (n = 1,184) or a lack of Olmsted County, MN residence for the duration of the study period (n = 815). All patients included had clinical follow-up at some point during the 3-year follow-up period. Over the course of this follow-up, 2,140 (87%) patients were prescribed at least one potentially nephrotoxic medication.

Baseline SCr and eGFR at the time of the index hospital admission were 1.0 (IQR 0.9, 1.3) mg/dL and 59 (IQR 53, 74) mL/min/1.73 m², respectively. Patients had a median of 1 (0, 2) outpatient SCr values in the 6 months to 7 days before admission (online suppl. Table S4). Cardiovascular disease, congestive heart failure, and peripheral vascular disease were present in 61%, 47%, and 57% of patients, respectively (Table 1). During the index hospitalization, maximum AKI severity was most often stage 1 (77%) and the median duration was approximately 1 day (IQR 0.5, 5.5). Discharge SCr values increased by approximately 0.4 mg/dL from admission SCr values, and the corresponding mean eGFR fell by approximately 10 mL/min/1.73 m² by discharge.

Among the 2,140 patients prescribed potentially nephrotoxic medications, PPIs, antibiotics and NSAIDs were the most commonly prescribed classes of medications (Table 1; online suppl. Table S5). 24% of all patients (n = 600) had an NSAID on their prescribed medication list at some point over the 3 year follow-up period.

For the primary end point, periods when patients were prescribed potentially nephrotoxic medications were associated with a significantly higher risk of de novo or progressive CKD during 3 years of follow-up (HR 1.38: 95% CI: 1.24–1.54, p < 0.001; Table 2). This association was preserved in analyses adjusted for baseline CKD risk (Adjusted HR 1.39: 95% CI: 1.24–1.56, p < 0.001; Table 3).

In multivariable analyses, we observed similar findings that indicated a risk with use of potential nephrotoxins (online suppl. Table S6). When we used only the diagnosis code aspect of the CKD definition, we saw a significant increase in the incidence of CKD during periods of nephrotoxin exposure compared to periods of no use (unadjusted HR 1.50: 95% CI: 1.17–1.91, p < 0.001). When we used only the eGFR aspect of the definition (at minimum, a 30% decline in eGFR from baseline) there was a significantly higher risk of de novo or progressive CKD during potentially nephrotoxic medication use (unadjusted HR 1.35: 95% CI: 1.20–1.51). 18 patients (0.7%) were excluded from this analysis due to absent SCr measurement in the 90-day to 3-year follow-up period. Seven of these excluded individuals were prescribed a potentially nephrotoxic medication at some point during follow-up. Active use of potentially nephrotoxic medications also increased the risk of the composite secondary end point of de novo or progressive CKD, AKI readmission, or death during the 3-year follow-up (HR 1.41; 95% CI: 1.28–1.56; Table 2). This outcome was primarily driven by CKD development (n = 1,223, 73.9%) followed by death (n = 360, 21.8%), AKI readmission (n = 60, 3.6%) and combined CKD and AKI readmission on the same day (n = 11, 0.7%).

Several a priori specified sensitivity analyses and subgroup evaluations were performed to evaluate the robustness of the study findings (Table 3). Findings were generally consistent including with truncated follow-up at 1 year, in the subgroup with renal recovery by discharge, in patients with and without heart failure, and in patients with a baseline eGFR <60 mL/min/1.73 m². The relative contribution to CKD development of each queried medication class was similar when ranked by c-statistic (online suppl. Table S7).

In AKI survivors, nephrotoxic medications are one of the few modifiable risk factors that may impact the development or progression of CKD. In this large population-based cohort study, we observed that nephrotoxic medication use following a hospitalization complicated by AKI increases the risk for de novo or progressive CKD in the 3 years after discharge. Increasing nephrotoxin burden [37] was associated with a stronger risk for long-term kidney dysfunction.

The implications of our findings are significant. Nearly two-thirds of medications are eliminated extensively by the kidney, and 20–40% are directly or indirectly nephrotoxic [18-20]. Our study has confirmed that AKI survivors are frequently prescribed potentially nephrotoxic medications at hospital discharge, and those medications are a modifiable risk factor for progression to CKD. We observed a step-wise increase in risk of de novo or progressive CKD during time periods when patients were prescribed increasing counts of nephrotoxic medications. This association remained significant even in patients that recovered from their AKI episode prior to discharge. These data demonstrate that a clear opportunity exists for clinician-guided nephrotoxin stewardship programs both at and after hospital discharge in AKI survivors. Our results could be used to support de-prescription of potentially nephrotoxic medications that are either unnecessary or can be interchanged for therapeutically similar alternative medications that carry less risk. Omeprazole and pantoprazole, PPIs that may be able to be exchanged for other acid-suppressive therapies during the at-risk period (e.g., histamine 2-receptor antagonists), were among the most commonly used potentially nephrotoxic medications in this cohort. In patients where nephrotoxic medications must be utilized for compelling indications such as transplantation or cancer care, stewardship initiatives could focus on an evaluation of the cumulative nephrotoxin burden and use of alternative non-nephrotoxic treatments for pain, heartburn, and some infections [37]. Further investigation is critical to test real-world feasibility and impact of nephrotoxin stewardship interventions.

In addition to nephrotoxin stewardship interventions, an opportunity exists to support patient education initiatives that explain the role that nephrotoxic medications can have on long-term kidney outcomes. Unfortunately, most adults are unaware of the function of their kidneys and their role in processing medications. In a large survey of community-dwelling adults, only 12% of individuals were aware that the kidneys processed medications [38]. As a consequence, patients may inadvertently take over the counter, potentially nephrotoxic medications after an episode of AKI because they are not aware of the risk. In the Southern Community Cohort Study, NSAID use was common among AKI survivors with as many as 19% of patients regularly using prescription or nonprescription NSAIDs [39]. These findings suggest that in addition to de-prescribing potentially nephrotoxic medications at discharge, educational programs are also imperative to prevent additional renal insults from over the counter medications such as NSAIDs [40]. We found that 24% of included individuals had prescriptions for NSAIDs during the follow-up period, likely only a subset of total use given the nonprescription status of such drugs like ibuprofen and naproxen sodium.

In our heart failure sub-analysis, we found a consistent risk of CKD development with nephrotoxin exposure. Previous studies have demonstrated that patients with heart failure are at greater risk for adverse kidney outcomes including repeat AKI episodes and CKD development [33, 41]. Our study suggests that despite the overall increased risk of CKD development, nephrotoxins remain a potentially modifiable factor for CKD development in this subset of patients. Importantly, ACEI/ARBs and diuretics were not considered nephrotoxic in this study, a pertinent difference from previous research [42].

The current study is not without limitations. Due to the observational design, there is a potential that residual confounding from unmeasured or unknown factors that influence long-term kidney function could have played a role in the findings. To the extent possible, this risk was minimized. We conducted multivariable analyses including numerous known risk factors for long-term kidney dysfunction in AKI survivors and the results were consistent. Several additional analyses were performed to evaluate the findings in different subgroups of AKI survivors (e.g., those with renal recovery by discharge, those with heart failure, those with baseline CKD). As previously acknowledged, some nephrotoxic medications may be unavoidable. Further investigation within this subset of patients is necessary to determine the real-world opportunity for modification of these therapies and the relative impact of disease processes on renal outcomes. We did observe that cumulative nephrotoxin burden was an important risk factor. In these cases where select nephrotoxic drugs are mandatory, the cumulative burden should be evaluated for other potential modifiable agents. It must also be acknowledged that nephrotoxins exhibit unique patterns and onset of injury which may be obscured with our aggregated approach [43]. For example, NSAIDs have multiple mechanisms by which they can contribute to nephrotoxicity. One such presentation is acute (<7 days) due to altered intraglomerular hemodynamics. In contrast, PPIs may present with acute interstitial nephritis which may occur over several months. Duration of exposure as it relates to the potential mechanism of injury was not included in this analysis. Given the available outpatient data in this historical study, we are also unable to ascertain whether patients actually utilized the medications as prescribed, or account for nonprescription medication use, herbals, and supplements. To minimize the impact of this limitation, prescription data were reconciled from multiple sources to approximate true medication administration as closely as possible. Finally, although all of the patients included in our study had some form of medical follow-up in the 3 years after their AKI event, not all patients had creatinine values obtained in the follow-up period, potentially leading to underestimation of the true incidence of CKD in this cohort.

In conclusion, data from this cohort of patients discharged after a hospitalization complicated by AKI suggests that prescription of potentially nephrotoxic medications at or after dismissal increases the risk for new or worsening CKD, suggesting that an opportunity for care optimization may exist. Further research is needed to fully elucidate the risk associated with individual medication classes and the extent to which an intervention to adjust medication regimens would improve outcomes.

This study protocol was reviewed and approved by Institutional Review. Boards from both Mayo Clinic (No. 18-007889) and Olmsted Medical Center (No. 032-OMC-18). All included individuals consented to the use of their medical records for research. In accordance with IRB recommendations, informed consent was waived due to the retrospective nature of this study.

The funding sources had no role in the study design; data collection, analysis, or interpretation; writing the report; or the decision to submit the report for publication. The contents of this work are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health or the American Society of Health System Pharmacy. The authors have no other conflicts of interest to disclose.

This project was supported in part by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Number K23AI143882 (PI; E.F.B.), the American Society of Health System Pharmacy Resident Research Grant (PI; D.J.S.), and the Agency for Healthcare Research and Quality HS028060-01 (PI; E.F.B.). Additional funding was provided by the Mayo Clinic Department of Pharmacy. This study was made possible using the resources of the Rochester Epidemiology Project, which is supported by the National Institute on Aging of the National Institutes of Health under Award Number R01AG034676.

D.J.S., A.D.R., K.B.K., K.C.M., S.L.K., J.C.L., A.M.C., and E.F.B. contributed to the design of the research, the analysis of the results, and the writing of the manuscript. K.C.M. performed the statistical analysis of the research findings.

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants, but are available from the corresponding author EFB upon reasonable request.