Introduction

Acute kidney injury (AKI) is a common complication of critical illness, increasing the risk of morbidity and mortality [1,2,3], yet there remain no direct therapeutic options. Sepsis-associated AKI occurs in 40–50% of patients with sepsis [1] and increases mortality three to five-fold [4]. Although critical care nephrology research has advanced significantly in the last ten years with the implementation of a standardized definition using the Kidney Disease Global Outcomes (KDIGO) staging criteria [5], there remains difficulty in identifying AKI before there is a loss of organ function. Much work has focused on developing biomarkers of kidney stress or damage. Notably, kidney injury molecule 1 (KIM-1) has been extensively studied as a marker of kidney injury [6]. It is a type I membrane glycoprotein that is upregulated in response to renal tubular injury, and has been shown to be a sensitive and specific marker for kidney damage [6, 7]. Tissue inhibitor of metalloproteinase-2 (TIMP-2) and insulin-like growth factor binding protein 7 (IGFBP7) are markers of G1 cell cycle arrest that act as kidney biomarkers of cellular stress [8]. The NephroCheck Test, reporting the product of TIMP-2 and IGFBP7 ([TIMP-2]•[IGFBP7]) provides a quantitative risk index for AKI [9]. Multiple study results have supported its use for assessing AKI risk [9]. Notably, previous studies have shown that [TIMP-2]•[IGFBP7] maintains its performance despite the presence of chronic kidney disease [10].

Most drug discovery programs begin with patients suffering from a disease of interest. When a disease is common, studying people who appear to have natural resistance may be possible [11]. We hypothesized that there are individuals who are less likely to develop AKI despite a potent exposure (i.e. septic shock). Therefore, for this investigation, using data from the Protocolized Care for Early Septic Shock (ProCESS) trial [12] and leveraging the kidney biomarkers KIM-1 and [TIMP-2]•[IGFBP7] for kidney injury and stress, we sought to identify a unique subset of patients resistant to AKI. We define patients with evidence of kidney stress by high levels of [TIMP-2]•[IGFBP7] but the absence of AKI based on either KDIGO clinical criteria or an increase in KIM-1 (to identify subclinical injury) as meeting criteria for AKI resistance. Our primary study objective was to evaluate a framework of AKI resistance by comparing baseline susceptibilities of patients with AKI resistance to those with AKI and those at reduced risk for AKI. We also aimed to compare patient-centered outcomes in critically ill patients with septic shock meeting our proposed criteria for AKI resistance to patients with AKI and patients at reduced risk of AKI defined by both clinical and kidney biomarker criteria.

Methods

Study design

For this retrospective cohort study, we used existing data from the ProCESS trial. The ProCESS trial was a multicenter, randomized clinical trial of three different resuscitation strategies in patients with septic shock that enrolled 1341 patients from March of 2008 to May of 2013 at 31 academic and community emergency departments and intensive care units [12]. The methods and primary results of the ProCESS trial have been detailed previously [4, 12]. The study was performed in accordance with the 1964 Declaration of Helsinki and approved by the Office of Human Research Protection at the University of Pittsburgh (study number 19040099). The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline (Supplementary Table S1).

Data acquisition and biomarker testing

Patients were excluded with end-stage renal disease (ESRD), a baseline serum creatinine > 4 mg/dl at the time of enrollment [4] or missing enrollment serum creatinine values. For this study, we selected urine samples for biomarker testing at 6 h after the start of resuscitation, based on prior research indicating that [TIMP-2]•[IGFBP7] levels are most informative at this time point in patients with sepsis [13]. As shown in Fig. 1, a total of 999 patients in the ProCESS cohort had urine available for analysis at the 6-hour time period. Both [TIMP-2]•[IGFBP7] and [KIM-1] were obtained in the 573 patients with urine sample volumes adequate for analysis. Samples were centrifuged immediately after collection, and the supernatant was frozen and stored at temperatures below − 70 °C. Prior to testing for [KIM-1], the supernatant was thawed and analyzed using enzyme-linked immunosorbent assay kits from EKF Diagnostics (Cardiff, UK), following the manufacturer’s instructions. Similarly, the supernatant was thawed before measuring the [TIMP-2]•[IGFBP7] levels with the NephroCheck Test, which was also conducted in accordance with the manufacturer’s guidelines.

Fig. 1
figure 1

Flow diagram of the study population. The diagram illustrates the exclusion criteria and stratification of patients from the Protocolized Care for Early Septic Shock (ProCESS) study (n = 1,341) based on biomarker and clinical criteria. Patients were excluded with end-stage renal disease (ESRD), a baseline serum creatinine > 4 mg/dl, or missing admission creatinine. A total of 573 patients were included in the analysis. Patients were stratified into three groups based on their product of tissue inhibitor of metalloproteinase-2 and insulin-like growth factor binding protein 7 ([TIMP-2]*[IGFBP7]) levels, Kidney Disease Improving Global Outcomes Acute Kidney Injury (AKI) classification, and kidney injury molecule-1 (KIM-1) classification: 1. Reduced Risk for AKI ([TIMP-2]*[IGFBP7] ≤ 1): This group included 339 patients, further subdivided into those with no AKI by KDIGO (n = 312) and those with AKI by KDIGO (n = 27). 2. AKI Resistant ([TIMP-2]*[IGFBP7] > 1, n = 40): These patients did not develop AKI by KDIGO or show evidence of subclinical AKI with a [KIM-1] level ≤ 2. 3. AKI ([TIMP-2]*[IGFBP7] > 1,n = 194): These patients developed AKI by KDIGO or had a [KIM-1] level > 2

Full size image

We collected patient demographics and health history data at the time of study enrollment. Severity of illness was defined using the Sequential Organ Failure Assessment (SOFA) score [14]. The kidney component of the score, which includes creatinine and urine output, was removed for the analysis to avoid collinearity. The Charlson Comorbidity Index was used to quantify the baseline multimorbidity of the study cohort [15].

AKI resistance definition and outcomes

Serum creatinine values (up to 12 months before, at enrollment, and then daily until discharge from the hospital or truncated at seven days) and urine output (hourly for the first 72 h or until discharge from the ICU) were obtained as previously reported [12]. We classified AKI according to the KDIGO classification using both serum creatinine level and urine output [5] between 0 and 7 days after enrollment. As shown in Table 1, for comparison, we divided the study cohort into 3 groups: (1) reduced risk for AKI (lack of biomarker evidence for greatly elevated kidney stress), (2) AKI-resistant (biomarker evidence of kidney stress yet no evidence of AKI based on KDIGO and biomarker testing), and (3) AKI (biomarker evidence of stress and damage or AKI based on KDIGO criteria). To determine evidence of kidney stress, we chose a 1.0 (ng/mL)2/1000 cutoff for [TIMP-2]•[IGFBP7]. This cutoff has been previously validated in patients with sepsis and represents the upper bound of 95% CI for patients without AKI [16]. Evidence of subclinical kidney injury not evident by KDIGO clinical criteria was determined by KIM-1 levels. For this assessment, we used a [KIM-1] cutoff of 2 ng/mL as this threshold provides high specificity [17, 18]. In order to meet criteria for AKI resistance, patients needed to: (1) have a [TIMP-2]•[IGFBP7] > 1.0 (ng/mL)2/1000, (2) not meet KDIGO urine output or serum creatinine criteria for AKI, and (3) have a [KIM-1] value ≤ 2 ng/mL. Survival 30 days after enrollment and ICU length of stay were the primary endpoints.

Table 1 Proposed framework for defining acute kidney injury resistance
Full size table

Statistical analysis

Baseline characteristics were presented as medians with interquartile ranges (IQR) for continuous data or as frequencies and percentages (n, %) for categorical data. To account for potential imbalances in baseline characteristics between groups, standardized differences were calculated for each variable. For continuous variables, the standardized difference was defined as the absolute difference in means between groups divided by the pooled standard deviation. For categorical variables, the standardized difference was calculated by dividing the difference in proportions by the square root of the average of the group-specific variances. A threshold of greater than 0.1 was considered a meaningful difference between groups. The Chi-square test was used to assess significant differences between those meeting the criteria for reduced risk for AKI, AKI resistance, and AKI. For continuous data, the Mann-Whitney U test was used to compare differences between groups after the results of the Shapiro-Wilk test indicated that the data were not normally distributed. Logistic regression was done adjusting for non-renal SOFA score to assess the association of reduced risk of AKI, AKI resistance, and AKI with mortality. A negative binomial regression model was used to evaluate associations of reduced risk of AKI, AKI resistance, and AKI with ICU length of stay, adjusting for non-renal SOFA score. We completed a sensitivity analysis for the length of stay outcome, excluding patients who died before hospital day 7. We also performed two sensitivity analyses in order to validate our chosen criteria for reduced risk of AKI, AKI resistance, and AKI with mortality. First, we moved any patients classified as having a reduced risk of AKI who met KDIGO criteria for AKI into the AKI cohort and repeated all analyses. Second, we used a [KIM-1] cutoff of ≤1 ng/ml instead of ≤2 ng/ml to define AKI-resistant patients. A p-value < 0.05 was considered statistically significant for all analyses. All statistical analyses were performed using STATA 16.1 (Stata Corp, College Station, TX).

Results

After excluding patients with ESRD (n = 83), a baseline serum creatinine > 4 mg/dl (n = 8), and missing serum creatinine at study enrollment (n = 7), we measured [KIM-1] and [TIMP-2]•[IGFBP7] on 573 patients (Fig. 1). Of the 339 (59%) patients with reduced risk for AKI, 27 (8%) met KDIGO criteria for AKI. The maximum KDIGO AKI stage of patients classified as reduced risk for AKI between 0 and 7 days after enrollment was: 93% Stage 1 (n = 25) and 7% Stage 2 (n = 2). No patients in this cohort met the criteria for Stage 3 AKI.

Of the 234 (41%) patients with a [TIMP-2]•[IGFBP7] > 1 (ng/mL)2/1000, 179 (76.5%) met clinical KDIGO AKI criteria and an additional 15 (6.4%) patients had [KIM-1] > 2 ng/mL for a total of 194 (33.9%) patients. Among the 179 patients who met KDIGO criteria for AKI, 26 (14.5%) were classified as Stage 1, 73 (40.8%) as Stage 2, and 80 (44.7%) as Stage 3. Eight patients who met KDIGO criteria for AKI received dialysis within 7 days of enrollment. Finally, 40 (7.0%) of the 573 patients had a [TIMP-2]•[IGFBP7] > 1 (ng/mL)2/1000, showing evidence of kidney stress, but did not meet KDKIGO AKI criteria or a [KIM-1] > 2 mg/mL. These patients were considered to have AKI resistance. Notably, the median [TIMP-2]•[IGFBP7] value for the patients with AKI resistance was 2.4 (IQR: 1.5–4.5) and 2.9 (IQR: 1.5–6.1) (ng/mL)2/1000 for patients with AKI. This difference was not statistically significant (p = 0.28). Twenty-one (52.5%) of patients with AKI resistance had a [TIMP-2]•[IGFBP7] > 2 (ng/mL)2/1000. During 28 days of follow-up after hospital admission, no patients who met criteria for AKI resistance subsequently met KDIGO criteria for AKI.

Table 2 presents the baseline characteristics of the study participants. The median age was 61 years (IQR 50–73), with 59% male. Arterial hypertension was the most common comorbidity (58.3%), followed by diabetes (34.2%) and chronic respiratory disease (21.8%). There were no differences in baseline characteristics when comparing patients with AKI resistance to those with AKI, including co-morbidities, need for mechanical ventilation, and severity of illness based on non-renal SOFA scores, based on P-values. However, both patients with AKI and AKI resistance had an increased need for mechanical ventilation and higher non-renal SOFA scores when compared to patients with reduced AKI risk. A greater proportion of patients meeting criteria for AKI were male when compared to those with reduced risk of AKI and those considered to have AKI resistance based on p-values and/or standardized differences (66% vs. 55.5% and 55.0%). Based on standardized differences, a history of myocardial infarction was more common in the reduced risk group when compared to those with AKI (11.5% vs. 6.7%). Cirrhosis was more common in patients with AKI when compared to patients with reduced risk for AKI (9.3% vs. 5.3%).

Table 2 Baseline characteristics and severity of illness
Full size table

As shown in Tables 3 and 122 (21.3%) patients died within 30 days of enrollment. The 30-day mortality and length of stay were similar when comparing patients with AKI resistance to patients with reduced risk for AKI. Mortality was higher in AKI patients (32%) compared to AKI-resistant patients (10%), with an adjusted odds ratio of 3.79 (95% CI: 1.24–11.54, P = 0.02). AKI patients experienced a significantly longer ICU stay, with a median of 4 (IQR: 2–7) days when compared to 2 (IQR: 1.5–4.5) days in AKI-resistant patients (incidence rate ratio [IRR]: 1.43, 95% CI: 1.07–1.90, P = 0.01). Mortality was higher in patients with AKI compared to those with reduced risk for AKI (16.5%) with an adjusted odds ratio of 1.62 (95% CI: 1.03–2.56, P = 0.04). The length of stay was shorter for those with a reduced risk of AKI, with a median of 3 (IQR 2–5) days when compared to those with AKI (IRR: 1.20, 95% CI: 1.04–1.40, P = 0.01). After excluding the 59 patients who died within the first 7 days of admission, our findings remained consistent. ICU length of stay did not differ between patients with AKI resistance and those with reduced AKI risk. As in the original analysis, patients with AKI had significantly longer ICU stays (median 5 days, IQR 3–7) compared to those with AKI resistance (median 2 days, IQR 2–5; IRR: 1.55, 95% CI: 1.17–2.06, P < 0.01). Similarly, patients with reduced AKI risk had shorter ICU stays (median 3 days, IQR 2–5) than those with AKI (IRR: 1.29, 95% CI: 1.11–1.50, P < 0.01).

Table 3 Comparison of clinical endpoints
Full size table

We performed two sensitivity analyses to validate our selected criteria for reduced risk of AKI, AKI resistance, and AKI. For the first analysis, the 27 patients (8%) with reduced risk for AKI who met KDIGO AKI criteria were included in the AKI cohort rather than the reduced risk for AKI cohort. As shown in Supplementary Table S2 the significance of baseline characteristics and severity of illness when comparing the three categories were unchanged. Supplementary Table S3 shows that the significance of the comparisons of the groups based on clinical outcomes were also unchanged. We performed a second sensitivity analysis using a [KIM-1] cutoff of 1 ng/mL rather than 2 ng/mL, therefore providing a more stringent definition for AKI resistance and a more sensitive definition for AKI. The cohort of patients with AKI resistance (n = 26) decreased while the cohort with AKI increased (n = 214) by 14 patients. Once again, the significance of the results was unchanged when compared to the primary analyses, except there was no difference in the need for mechanical ventilation when comparing those who were AKI resistant to those with a reduced risk (Supplementary Tables S4 and S5).

Discussion

Among critically ill patients with septic shock, AKI occurs frequently. The majority (339, 89%) of the 379 patients not developing AKI had reduced levels of kidney stress as assessed by [TIMP-2]•[IGFBP7] ≤ 1.0 (ng/mL)2/1000. Although these patients suffered a potent exposure, namely septic shock, they may not have engaged one or more of the underlying mechanisms responsible for sepsis-induced AKI [19]. Therefore, these patients may simply have been low risk for AKI. 7% of our patients were high-risk for AKI not only because they had septic shock, but also they experienced high levels of kidney stress. These patients, whom we classified as AKI-resistant, were clinically indistinguishable from those with AKI, in that they had a similar severity of illness, age, and comorbidity status when compared to those with AKI. Despite sharing similar baseline susceptibility to the AKI sufferers, the AKI-resistant group exhibited significantly better clinical outcomes, including lower 30-day mortality and shorter ICU stays when compared to those with AKI. Importantly, the AKI-resistant patients were not just negative for AKI by clinical criteria, they were also negative by [KIM-1]. In a sensitivity analysis that lowered the threshold to define AKI by KIM − 1 concentration from 2 ng/mL to 1 ng/mL, the cohort of patients with AKI resistance decreased by 35% (n = 26). Nonetheless, this smaller group of AKI-resistant patients still had no difference in age, comorbidities, or severity of illness when compared to those with AKI. Moreover, this group of patients still continued to demonstrate significantly shorter ICU stays and decreased mortality when compared to those with AKI. Validating [TIMP-2]•[IGFBP7] > 1(ng/mL)2/1000 as a cutoff for defining kidney stress in this cohort, only 8% of patients with [TIMP-2]•[IGFBP7] ≤ 1 met KDIGO criteria for AKI. This aligns with findings in sepsis from Honore and colleagues, where [TIMP-2]•[IGFBP7] ≤1 (ng/mL)2/1000 had a negative predictive value of 94% for AKI [16].

The greater than three-fold increase in the odds of death among the AKI group when compared to those with AKI resistance highlights the profound impact that AKI has on critically ill patients. Patients with AKI resistance, having the same baseline sensitivities and exposures as those with AKI, fared significantly better regarding outcomes than those with AKI. This finding underscores the importance of early identification and management of AKI while also raising important questions about the mechanisms underlying AKI resistance. The concept of AKI resistance is intriguing, and several potential explanations warrant consideration. One possibility is that this resistance may be attributable to chance, as suggested by chaos theory [20]. This theory suggests that complex systems, such as human physiology during critical illness, can behave unpredictably, and in some cases, apparent resilience to injury may result from random or unforeseen factors. Another explanation could be related to renal functional reserve, which represents the kidney’s capacity to increase its function in response to stress or injury [21]. However, it is more likely that AKI resistance is due to a combination of physiological and molecular mechanisms that have yet to be fully elucidated. Intriguingly, there was a greater proportion of men with AKI when compared to those with reduced risk for AKI and with AKI resistance, consistent with the concept that men may be more susceptible to sepsis-associated AKI when compared to women [22]. Importantly, this finding should also be considered in the context of diagnostic bias, as AKI can be more commonly missed in females, potentially leading to underdiagnosis and underreporting in this population [23].

Much like the discovery of the CCR5-delta 32 mutation in individuals resistant to HIV, we question if genetic factors may play a role in AKI resistance [11]. This discovery led to the development of CCR5 receptor antagonists, such as Maraviroc, which became a significant therapeutic advancement in HIV treatment [24]. Similarly, genetic variants or polymorphisms could exist in the AKI-resistant cohort that confer some protection against AKI, and identifying these genetic factors could offer insights into novel therapeutic targets. One promising avenue for future research is identifying differentially expressed proteins in AKI-resistant patients compared to those who develop AKI. The use of proteomic technology has proven successful in identifying therapeutic targets in other diseases, such as cancer, Alzheimer’s disease, and cardiovascular disease [25,26,27,28]. Applying similar proteomic techniques to the study of AKI resistance could provide information about the molecular pathways involved in kidney protection. Identifying proteins that are upregulated in response to kidney stress but do not result in injury could help uncover potential targets for novel therapies aimed at enhancing renal resilience.

The identification of a cohort of patients who are exposed to the risk factors for AKI yet do not develop the condition provides a unique opportunity to study potential protective mechanisms in greater detail. Future research could focus on understanding the upregulation of protective proteins and other molecular factors that differentiate AKI-resistant patients from those who succumb to AKI. Such studies could ultimately lead to the development of actionable targets that could be leveraged to create therapies designed to bolster kidney protection in patients at risk for AKI or with AKI. We know from previous analyses of the ProCESS cohort that 75% of AKI episodes occurred either shortly after or immediately following the presentation of the patient to the emergency department [4]. This suggests that early interventions aimed at preventing AKI will be inherently challenging due to the rapid onset of kidney injury. Therefore, therapies for AKI must be effective after the initiation of kidney injury, underscoring the need for treatments that can halt the progression of AKI or promote kidney recovery once the injury has occurred.

There are limitations to this study that need to be acknowledged. First, it was a retrospective analysis of data collected from a randomized clinical trial, and the study was not originally designed to test our specific hypothesis. While we could access a large set of clinical data with timely sample collection and standardized definitions of AKI and clinical outcomes, the dataset did not include biomarker results for all patients, which may have introduced selection bias. However, we ran biomarker assays using all remaining urine samples from the ProCESS cohort for the purposes of this study. Additionally, we did not have access to data regarding nephrotoxin exposure, which could have been a critical factor influencing AKI development. We note that although KIM-1 is a validated marker for kidney damage, our use of the marker to help exclude subclinical AKI while [TIMP-2]•[IGFBP7] is positive is debatable.

In conclusion, this investigation identified a small subset of critically ill septic shock patients who exhibited kidney stress and were similarly critically ill when compared to those with AKI but did not progress to AKI. These AKI-resistant patients experience significantly better outcomes, including lower mortality and shorter ICU stays, highlighting the importance of further investigation into the underlying mechanisms of AKI resistance. By leveraging proteomic technologies, future research may identify proteins and pathways responsible for kidney protection, leading to the development of novel, targeted therapies for patients at risk of AKI. Ultimately, studying AKI resistance may provide the key to identifying new, effective treatments for AKI, much like the discovery of CCR5 antagonism in HIV led to transformative advancements in therapy.