The Road to Precision Medicine for Acute Kidney Injury.

OBJECTIVES: Acute kidney injury (AKI) is a common form of organ dysfunction in the ICU. AKI is associated with adverse short- and long-term outcomes, including high mortality rates, which have not measurably improved over the past decade. This review summarizes the available literature examining the evidence of the need for precision medicine in AKI in critical illness, highlights the current evidence for heterogeneity in the field of AKI, discusses the progress made in advancing precision in AKI, and provides a roadmap for studying precision-guided care in AKI. DATA SOURCES: Medical literature regarding topics relevant to precision medicine in AKI, including AKI definitions, epidemiology, and outcomes, novel AKI biomarkers, studies of electronic health records (EHRs), clinical trial design, and observational studies of kidney biopsies in patients with AKI. STUDY SELECTION: English language observational studies, randomized clinical trials, reviews, professional society recommendations, and guidelines on areas related to precision medicine in AKI. DATA EXTRACTION: Relevant study results, statements, and guidelines were qualitatively assessed and narratively synthesized. DATA SYNTHESIS: We synthesized relevant study results, professional society recommendations, and guidelines in this discussion. CONCLUSIONS: AKI is a syndrome that encompasses a wide range of underlying pathologies, and this heterogeneity has hindered the development of novel therapeutics for AKI. Wide-ranging efforts to improve precision in AKI have included the validation of novel biomarkers of AKI, leveraging EHRs for disease classification, and phenotyping of tubular secretory clearance. Ongoing efforts such as the Kidney Precision Medicine Project, identifying subphenotypes in AKI, and optimizing clinical trials and endpoints all have great promise in advancing precision medicine in AKI.

Assessing the role of Chemokine (C-C motif) ligand 14 in AKI: a European consensus meeting.

BACKGROUND: Urinary Chemokine (C-C motif) ligand 14 (CCL14) is a biomarker associated with persistent severe acute kidney injury (AKI). There is limited data to support the implementation of this AKI biomarker to guide therapeutic actions. METHODS: Sixteen AKI experts with clinical CCL14 experience participated in a Delphi-based method to reach consensus on when and how to potentially use CCL14. Consensus was defined as ≥ 80% agreement (participants answered with 'Yes', or three to four points on a five-point Likert Scale). RESULTS: Key consensus areas for CCL14 test implementation were: identifying challenges and mitigations, developing a comprehensive protocol and pairing it with a treatment plan, and defining the target population. The majority agreed that CCL14 results can help to prioritize AKI management decisions. CCL14 levels above the high cutoff (> 13 ng/mL) significantly changed the level of concern for modifying the AKI treatment plan (p < 0.001). The highest level of concern to modify the treatment plan was for discussions on renal replacement therapy (RRT) initiation for CCL14 levels > 13 ng/mL. The level of concern for discussion on RRT initiation between High and Low, and between Medium and Low CCL14 levels, showed significant differences. CONCLUSION: Real world urinary CCL14 use appears to provide improved care options to patients at risk for persistent severe AKI. Experts believe there is a role for CCL14 in AKI management and it may potentially reduce AKI-disease burden. There is, however, an urgent need for evidence on treatment decisions and adjustments based on CCL14 results.

Exploring the Cost-Utility of a Biomarker Predicting Persistent Severe Acute Kidney Injury: The Case of C-C Motif Chemokine Ligand 14 (CCL14).

Background: Approximately 24% of hospitalized stage 2-3 acute kidney injury (AKI) patients will develop persistent severe AKI (PS-AKI), defined as KDIGO stage 3 AKI lasting ≥3 days or with death in ≤3 days or stage 2 or 3 AKI with dialysis in ≤3 days, leading to worse outcomes and higher costs. There is currently no consensus on an intervention that effectively reverts the course of AKI and prevents PS-AKI in the population with stage 2-3 AKI. This study explores the cost-utility of biomarkers predicting PS-AKI, under the assumption that such intervention exists by comparing C-C motif chemokine ligand 14 (CCL14) to hospital standard of care (SOC) alone. Methods: The analysis combined a 90-day decision tree using CCL14 operating characteristics to predict PS-AKI and clinical outcomes in 66-year-old patients, and a Markov cohort estimating lifetime costs and quality-adjusted life years (QALYs). Cost and QALYs from admission, 30-day readmission, intensive care, dialysis, and death were compared. Clinical and cost inputs were informed by a large retrospective cohort of US hospitals in the PINC AI Healthcare Database. Inputs and assumptions were challenged in deterministic and probabilistic sensitivity analyses. Two-way analyses were used to explore the efficacy and costs of an intervention preventing PS-AKI. Results: Depending on selected costs and early intervention efficacy, CCL14-directed care led to lower costs and more QALYs (dominating) or was cost-effective at the $50,000/QALY threshold. Assuming the intervention would avoid 10% of PS-AKI complications in AKI stage 2-3 patients identified as true positive resulted in 0.066 additional QALYs and $486 reduced costs. Results were robust to substantial parameter variation. Conclusion: The analysis suggests that in the presence of an efficacious intervention preventing PS-AKI, identifying people at risk using CCL14 in addition to SOC is likely to represent a cost-effective use of resources.

Initial renal replacement therapy (RRT) modality associates with 90-day postdischarge RRT dependence in critically ill AKI survivors.

PURPOSE: Real-world comparison of RRT modality on RRT dependence at 90 days postdischarge among ICU patients discharged alive after RRT for acute kidney injury (AKI). METHODS: Using claims-linked to US hospital discharge data (Premier PINC AI Healthcare Database [PHD]), we compared continuous renal replacement therapy (CRRT) vs. intermittent hemodialysis (IHD) for AKI in adult ICU patients discharged alive from January 1, 2018 to June 30, 2021. RRT dependence at 90 days postdischarge was defined as ≥2 RRT treatments in the last 8 days. Between-group differences were balanced using inverse probability treatment weighting (IPTW). RESULTS: Of 34,804 patients, 3804 patients (from 382 hospitals) had claims coverage for days 83-90 postdischarge. Compared to IHD-treated patients (n = 2740), CRRT-treated patients (n = 1064) were younger; had more admission to large teaching hospitals, surgery, sepsis, shock, mechanical ventilation, but lower prevalence of comorbidities (p < 0.05 for all). Compared to IHD-treated patients, CRRT-treated patients had lower RRT dependence at hospital discharge (26.5% vs. 29.8%, p = 0.04) and lower RRT dependence at 90 days postdischarge (4.9% vs. 7.4% p = 0.006) with weighted adjusted OR (95% CI): 0.68 (0.47-0.97), p = 0.03. Results persisted in sensitivity analyses including patients who died during days 1-90 postdischarge (n = 112) or excluding patients from hospitals with IHD patients only (n = 335), or when excluding patients who switched RRT modalities (n = 451). CONCLUSIONS: Adjusted for potential confounders, the odds of RRT dependence at 90 days postdischarge among survivors of RRT for AKI was 30% lower for those treated first with CRRT vs. IHD, overall and in several sensitivity analyses. SUMMARY: Critically ill patients in intensive care units (ICU) may develop acute kidney injury (AKI) that requires renal replacement therapy (RRT) to temporarily replace the injured kidney function of cleaning the blood. Two main types of RRT in the ICU are called continuous renal replacement therapy (CRRT), which is performed almost continuously, i.e., for >18 h per day, and intermittent hemodialysis (IHD), which is a more rapid RRT that is usually completed in a little bit over 6 h, several times per week. The slower CRRT may be gentler on the kidneys and is more likely to be used in the sickest patients, who may not be able to tolerate IHD. We conducted a data-analysis study to evaluate whether long-term effects on kidney function (assessed by ongoing need for RRT, i.e., RRT dependence) differ depending on use of CRRT vs. IHD. In a very large US linked hospital-discharge/claims database we found that among ICU patients discharge alive after RRT for AKI, fewer CRRT-treated patients had RRT dependence at hospital discharge (26.5% vs. 29.8%, p = 0.04) and at 90 days after discharge (4.9% vs. 7.4% p = 0.006). In adjusted models, RRT dependence at 90 days postdischarge was >30% lower for CRRT than IHD-treated patients. These results from a non-randomized study suggest that among survivors of RRT for AKI, CRRT may result in less RRT dependence 90 days after hospital discharge.

Development and external validation of deep learning clinical prediction models using variable-length time series data.

OBJECTIVES: To compare and externally validate popular deep learning model architectures and data transformation methods for variable-length time series data in 3 clinical tasks (clinical deterioration, severe acute kidney injury [AKI], and suspected infection). MATERIALS AND METHODS: This multicenter retrospective study included admissions at 2 medical centers that spanned 2007-2022. Distinct datasets were created for each clinical task, with 1 site used for training and the other for testing. Three feature engineering methods (normalization, standardization, and piece-wise linear encoding with decision trees [PLE-DTs]) and 3 architectures (long short-term memory/gated recurrent unit [LSTM/GRU], temporal convolutional network, and time-distributed wrapper with convolutional neural network [TDW-CNN]) were compared in each clinical task. Model discrimination was evaluated using the area under the precision-recall curve (AUPRC) and the area under the receiver operating characteristic curve (AUROC). RESULTS: The study comprised 373 825 admissions for training and 256 128 admissions for testing. LSTM/GRU models tied with TDW-CNN models with both obtaining the highest mean AUPRC in 2 tasks, and LSTM/GRU had the highest mean AUROC across all tasks (deterioration: 0.81, AKI: 0.92, infection: 0.87). PLE-DT with LSTM/GRU achieved the highest AUPRC in all tasks. DISCUSSION: When externally validated in 3 clinical tasks, the LSTM/GRU model architecture with PLE-DT transformed data demonstrated the highest AUPRC in all tasks. Multiple models achieved similar performance when evaluated using AUROC. CONCLUSION: The LSTM architecture performs as well or better than some newer architectures, and PLE-DT may enhance the AUPRC in variable-length time series data for predicting clinical outcomes during external validation.

CCL14 testing to guide clinical practice in patients with AKI: Results from an international expert panel.

PURPOSE: Urinary C-C motif chemokine ligand 14 (CCL14) is a strong predictor of persistent stage 3 acute kidney injury (AKI). Multiple clinical actions are recommended for AKI but how these are applied in individual patients and how the CCL14 test results may impact their application is unknown. METHODS: We assembled an international panel of 12 experts and conducted a modified Delphi process to evaluate patients at risk for persistent stage 3 AKI (lasting 72 hours or longer). Using a Likert scale, we rated 11 clinical actions based on international guidelines applied to each case before and after CCL14 testing and analyzed the association between the strength and direction of recommendations and CCL14 results. RESULTS: The strength and direction of clinical recommendations were strongly influenced by CCL14 results (P < 0.001 for the interaction). Nine (82%) recommendations for clinical actions were significantly impacted by CCL14 results (P < 0.001 comparing low to highest CCL14 risk category). CONCLUSIONS: Most recommendations for care of patients with stage 2-3 by an international panel of experts were strongly modified by CCL14 test results. This work should set the stage for clinical practice protocols and studies to determine the effects of recommended actions informed by CCL14.

Recommendations for clinical trial design in acute kidney injury from the 31st acute disease quality initiative consensus conference. A consensus statement.

PURPOSE: Novel interventions for the prevention or treatment of acute kidney injury (AKI) are currently lacking. To facilitate the evaluation and adoption of new treatments, the use of the most appropriate design and endpoints for clinical trials in AKI is critical and yet there is little consensus regarding these issues. We aimed to develop recommendations on endpoints and trial design for studies of AKI prevention and treatment interventions based on existing data and expert consensus. METHODS: At the 31st Acute Disease Quality Initiative (ADQI) meeting, international experts in critical care, nephrology, involving adults and pediatrics, biostatistics and people with lived experience (PWLE) were assembled. We focused on four main areas: (1) patient enrichment strategies, (2) prevention and attenuation studies, (3) treatment studies, and (4) innovative trial designs of studies other than traditional (parallel arm or cluster) randomized controlled trials. Using a modified Delphi process, recommendations and consensus statements were developed based on existing data, with > 90% agreement among panel members required for final adoption. RESULTS: The panel developed 12 consensus statements for clinical trial endpoints, application of enrichment strategies where appropriate, and inclusion of PWLE to inform trial designs. Innovative trial designs were also considered. CONCLUSION: The current lack of specific therapy for prevention or treatment of AKI demands refinement of future clinical trial design. Here we report the consensus findings of the 31st ADQI group meeting which has attempted to address these issues including the use of predictive and prognostic enrichment strategies to enable appropriate patient selection.

Development and external validation of multimodal postoperative acute kidney injury risk machine learning models.

Objectives: To develop and externally validate machine learning models using structured and unstructured electronic health record data to predict postoperative acute kidney injury (AKI) across inpatient settings. Materials and Methods: Data for adult postoperative admissions to the Loyola University Medical Center (2009-2017) were used for model development and admissions to the University of Wisconsin-Madison (2009-2020) were used for validation. Structured features included demographics, vital signs, laboratory results, and nurse-documented scores. Unstructured text from clinical notes were converted into concept unique identifiers (CUIs) using the clinical Text Analysis and Knowledge Extraction System. The primary outcome was the development of Kidney Disease Improvement Global Outcomes stage 2 AKI within 7 days after leaving the operating room. We derived unimodal extreme gradient boosting machines (XGBoost) and elastic net logistic regression (GLMNET) models using structured-only data and multimodal models combining structured data with CUI features. Model comparison was performed using the receiver operating characteristic curve (AUROC), with Delong's test for statistical differences. Results: The study cohort included 138 389 adult patient admissions (mean [SD] age 58 [16] years; 11 506 [8%] African-American; and 70 826 [51%] female) across the 2 sites. Of those, 2959 (2.1%) developed stage 2 AKI or higher. Across all data types, XGBoost outperformed GLMNET (mean AUROC 0.81 [95% confidence interval (CI), 0.80-0.82] vs 0.78 [95% CI, 0.77-0.79]). The multimodal XGBoost model incorporating CUIs parameterized as term frequency-inverse document frequency (TF-IDF) showed the highest discrimination performance (AUROC 0.82 [95% CI, 0.81-0.83]) over unimodal models (AUROC 0.79 [95% CI, 0.78-0.80]). Discussion: A multimodality approach with structured data and TF-IDF weighting of CUIs increased model performance over structured data-only models. Conclusion: These findings highlight the predictive power of CUIs when merged with structured data for clinical prediction models, which may improve the detection of postoperative AKI.