Machine learning for prediction of chronic kidney disease progression: Validation of the Klinrisk model in the CANVAS Program and CREDENCE trial.

AIM: To validate the Klinrisk machine learning model for prediction of chronic kidney disease (CKD) progression in patients with type 2 diabetes in the pooled CANVAS/CREDENCE trials. MATERIALS AND METHODS: We externally validated the Klinrisk model for prediction of CKD progression, defined as 40% or higher decline in estimated glomerular filtration rate (eGFR) or kidney failure. Model performance was assessed for prediction up to 3 years with the area under the receiver operating characteristic curve (AUC), Brier scores and calibration plots of observed and predicted risks. We compared performance of the model with standard of care using eGFR (G1-G4) and urine albumin-creatinine ratio (A1-A3) Kidney Disease Improving Global Outcomes (KDIGO) heatmap categories. RESULTS: The Klinrisk model achieved an AUC of 0.81 (95% confidence interval [CI] 0.78-0.83) at 1 year, and 0.88 (95% CI 0.86-0.89) at 3 years. The Brier scores were 0.020 (0.018-0.022) and 0.056 (0.052-0.059) at 1 and 3 years, respectively. Compared with the KDIGO heatmap, the Klinrisk model had improved performance at every interval (P < .01). CONCLUSIONS: The Klinrisk machine learning model, using routinely collected laboratory data, was highly accurate in its prediction of CKD progression in the CANVAS/CREDENCE trials. Integration of the model in electronic medical records or laboratory information systems can facilitate risk-based care.

Expanding Access to High KDPI Kidney Transplant for Recipients Aged 60 y and Older: Cost Utility and Survival.

Background: Modern organ allocation systems are tasked with equitably maximizing the utility of transplanted organs. Increasing the use of deceased donor organs at risk of discard may be a cost-effective strategy to improve overall transplant benefit. We determined the survival implications and cost utility of increasing the use of marginal kidneys in an older adult Canadian population of patients with end-stage kidney disease. Methods: We constructed a cost-utility model with microsimulation from the perspective of the Canadian single-payer health system for incident transplant waitlisted patients aged 60 y and older. A kidney donor profile index score of ≥86 was considered a marginal kidney. Donor- and recipient-level characteristics encompassed in the kidney donor profile index and estimated posttransplant survival scores were used to derive survival posttransplant. Patients were followed up for 10 y from the date of waitlist initiation. Our analysis compared the routine use of marginal kidneys (marginal kidney scenario) with the current practice of limited use (status quo scenario). Results: The 10-y mean cost and quality-adjusted life-years per patient in the marginal kidney scenario were estimated at $379 485.33 (SD: $156 872.49) and 4.77 (SD: 1.87). In the status quo scenario, the mean cost and quality-adjusted life-years per patient were $402 937.68 (SD: $168 508.85) and 4.37 (SD: 1.87); thus, the intervention was considered dominant. At 10 y, 62.8% and 57.0% of the respective cohorts in the marginal kidney and status quo scenarios remained alive. Conclusions: Increasing the use of marginal kidneys in patients with end-stage kidney disease aged 60 y and older may offer cost savings, improved quality of life, and greater patient survival in comparison with usual care.

Validation of the Klinrisk chronic kidney disease progression model in the FIDELITY population.

Background: Chronic kidney disease (CKD) affects >800 million individuals worldwide and is often underrecognized. Early detection, identification and treatment can delay disease progression. Klinrisk is a proprietary CKD progression risk prediction model based on common laboratory data to predict CKD progression. We aimed to externally validate the Klinrisk model for prediction of CKD progression in FIDELITY (a prespecified pooled analysis of two finerenone phase III trials in patients with CKD and type 2 diabetes). In addition, we sought to identify evidence of an interaction between treatment and risk. Methods: The validation cohort included all participants in FIDELITY up to 4 years. The primary and secondary composite outcomes included a ≥40% decrease in estimated glomerular filtration rate (eGFR) or kidney failure, and a ≥57% decrease in eGFR or kidney failure. Prediction discrimination was calculated using area under the receiver operating characteristic curve (AUC). Calibration plots were calculated by decile comparing observed with predicted risk. Results: At time horizons of 2 and 4 years, 993 and 1795 patients experienced a primary outcome event, respectively. The model predicted the primary outcome accurately with an AUC of 0.81 for 2 years and 0.86 for 4 years. Calibration was appropriate at both 2 and 4 years, with Brier scores of 0.067 and 0.115, respectively. No evidence of interaction between treatment and risk was identified for the primary composite outcome (P = .31). Conclusions: Our findings demonstrate the accuracy and utility of a laboratory-based prediction model for early identification of patients at the highest risk of CKD progression.

The SUB-urothelial DUrvalumab InjEction-1 (SUBDUE-1) trial: first-in-human trial in patients with bladder cancer.

OBJECTIVES: To assess the safety of sub-urothelial injection of durvalumab and examine the impact on tissue and circulating immune cell populations. PATIENTS AND METHODS: The patients were chemotherapy and immunotherapy naïve (bacille Calmette-Guérin allowed) with non-metastatic muscle-invasive bladder cancer or non-muscle-invasive bladder cancer planned for radical cystectomy (RC). The study was a Phase Ib 3 + 3 dose-escalation design with sub-urothelial injection of durvalumab at three pre-determined doses (25, 75, 150 mg) diluted in 25 mL normal saline, injected at 25 locations (25 × 1 mL injections), at least 2 weeks before RC. RESULTS: A total of 11 patients were recruited (10 male, one female). No significant changes were reported on American Urological Association Symptom Score or O'Leary Interstitial Cystitis Scale. In all, 14 adverse events (AEs) were reported (10 Grade 1, three Grade 2, one Grade 3), none considered immune-related. No Grade 4 or 5 AEs were recorded. All the patients underwent RC. Tissue immune populations changed following durvalumab injection (P = 0.012), with a statistically significant increase in M2-macrophage (CD163) when comparing the 25-150 mg dose (P = 0.021). Basal/mixed cancers showed a larger CD163 increase than luminal cancers (P = 0.033). CONCLUSION: Sub-urothelial injection of durvalumab is feasible and safe without immune-related AEs and shows local immunological effects.

Patient-Reported Outcomes in KEYNOTE-564: Adjuvant Pembrolizumab Versus Placebo for Renal Cell Carcinoma.

BACKGROUND: In patients with renal cell carcinoma (RCC) enrolled in the phase III KEYNOTE-564 trial (NCT03142334), disease-free survival (DFS) following nephrectomy was prolonged with use of adjuvant pembrolizumab therapy versus placebo. Patient-reported outcomes (PROs) provide an important measure of health-related quality of life (HRQoL) and can complement efficacy and safety results. PATIENTS AND METHODS: In KEYNOTE-564, 994 patients were randomly assigned to receive pembrolizumab 200 mg (n = 496) or placebo (n = 498) intravenously every 3 weeks for ≤17 cycles. Patients who received ≥1 dose of treatment and completed ≥1 HRQoL assessment were included in this analysis. HRQoL end points were assessed using the EORTC QLQ-C30, FKSI-DRS, and EQ VAS. Prespecified and exploratory PRO end points were mean change from baseline in EORTC QLQ-C30 GHS/QoL score, EORTC QLQ-C30 physical function subscale score, and FKSI-DRS score. RESULTS: No clinically meaningful difference in least squares mean scores for pembrolizumab versus placebo were observed at week 52 for EORTC QLQ-C30 GHS/QoL (-2.5; 95% CI -5.2 to 0.1), EORTC QLQ-C30 physical functioning (-0.87; 95% CI -2.7 to 1.0), and FKSI-DRS (-0.7; 95% CI -1.2 to -0.1). Most PRO scores remained stable or improved for the EORTC QLQ-C30 GHS/QoL (pembrolizumab, 54.3%; placebo, 67.5%), EORTC QLQ-C30 physical functioning (pembrolizumab, 64.7%; placebo, 68.8%), and FKSI-DRS (pembrolizumab, 58.2%; placebo, 66.3%). CONCLUSIONS: Adjuvant treatment with pembrolizumab did not result in deterioration of HRQoL. These findings together with the safety and efficacy findings support adjuvant pembrolizumab treatment following nephrectomy. TRIAL REGISTRATION: Clinicaltrials.gov Identifier: NCT03142334.

Association of body mass index with the development of metabolic acidosis in patients with chronic kidney disease.

Aims: Higher body mass index (BMI) is associated with higher bone mass and bone serves as a buffer during the development of metabolic acidosis. The authors sought to examine the relationship between BMI and metabolic acidosis in patients with chronic kidney disease (CKD). Materials and Methods: The study utilized a large US longitudinal data repository including over 103 million patients from healthcare provider organizations to evaluate the relationship between the exposure variable (BMI) and the prevalence and incidence of metabolic acidosis among patients with estimated glomerular filtration rate <60 ml/min/1.73 m2. Incident metabolic acidosis was identified at the first of two consecutive post-index serum bicarbonate values, 10-365 days apart, between 12 and <22 mEq/L in patients with normal index serum bicarbonate. Cox proportional hazard models were adjusted for multiple variables including demographics, comorbidities, income, education, and kidney function. Results: 103,766 patients qualified for this study; 6472 (6.2%) had metabolic acidosis at index. An inverse association between BMI category and metabolic acidosis was observed for both baseline (prevalence) and new-onset (incidence) metabolic acidosis. Compared to BMI category of 18.5 to <25 kg/m2, each category of incrementally higher BMI was associated with a decreasing risk of incident metabolic acidosis; the adjusted hazard ratios (95% confidence interval) were 0.866 (0.824-0.911), 0.770 (0.729-0.813), 0.664 (0.622-0.709), and 0.612 (0.571-0.655) for BMI 25 to <30, 30 to <35, 35 to <40, and 40+ kg/m2, respectively. Conclusions: Among patients with CKD, an incremental increase in BMI was inversely associated with both the prevalence and incidence of metabolic acidosis. These associations suggest that increased body weight may protect against the development of metabolic acidosis, a risk factor for progressive loss of kidney function.

Pembrolizumab plus lenvatinib as first-line therapy for advanced non-clear-cell renal cell carcinoma (KEYNOTE-B61): a single-arm, multicentre, phase 2 trial.

BACKGROUND: Immunotherapy-based combinations including pembrolizumab plus lenvatinib are the standard of care for patients with first-line clear-cell renal cell carcinoma, but these combinations are not well characterised in non-clear-cell renal cell carcinoma. We aimed to assess the activity and safety of pembrolizumab plus lenvatinib as a first-line treatment for patients with advanced non-clear-cell renal cell carcinoma. METHODS: KEYNOTE-B61 is a single-arm, phase 2 trial being conducted at 48 sites (hospitals and cancer centres) in 14 countries (Australia, Canada, France, Hungary, Ireland, Italy, Poland, South Korea, Russia, Spain, Türkiye, Ukraine, the UK, and the USA). Adult patients (aged ≥18 years) with previously untreated stage IV non-clear-cell renal cell carcinoma and a Karnofsky performance status of 70% or higher were eligible for enrolment. All enrolled patients received pembrolizumab 400 mg intravenously every 6 weeks for up to 18 cycles (2 years) plus lenvatinib 20 mg orally once daily or until disease progression, unacceptable toxicity, or withdrawal; lenvatinib could be continued beyond 2 years. The primary endpoint was the proportion of patients with a confirmed objective response as per adjusted Response Evaluation Criteria in Solid Tumours (version 1.1) assessed by independent central review. Activity and safety were analysed in all patients who received at least one dose of study treatment (the as-treated population). This trial is registered with ClinicalTrials.gov (NCT04704219) and is no longer recruiting participants but is ongoing. FINDINGS: Between Feb 23, 2021, and Jan 21, 2022, 215 patients were screened; 158 were enrolled and received treatment. Median age at baseline was 60 years (IQR 52-69), 112 (71%) of 158 patients were male, 46 (29%) were female, 128 (81%) were White, 12 (8%) were Asian, three (2%) were Black or African American, and 15 (9%) were missing data on race. As of data cutoff (Nov 7, 2022), median study follow-up was 14·9 months (IQR 11·1-17·4). 78 of 158 patients had a confirmed objective response (49%; 95% CI 41-57), including nine (6%) patients with a confirmed complete response and 69 (44%) with a confirmed partial response. Grade 3-4 treatment-related adverse events occurred in 81 (51%) of 158 patients, the most common of which were hypertension (37 [23%] of 158), proteinuria (seven [4%]), and stomatitis (six [4%]). Serious treatment-related adverse events occurred in 31 (20%) of 158 patients. Eight (5%) patients died due to adverse events, none of which was considered related to the treatment by the investigators (one each of cardiac failure, peritonitis, pneumonia, sepsis, cerebrovascular accident, suicide, pneumothorax, and pulmonary embolism). INTERPRETATION: Pembrolizumab plus lenvatinib has durable antitumour activity in patients with previously untreated advanced non-clear-cell renal cell carcinoma, with a safety profile consistent with that of previous studies. Results from KEYNOTE-B61 support the use of pembrolizumab plus lenvatinib as a first-line treatment option for these patients. FUNDING: Merck Sharp & Dohme (a subsidiary of Merck & Co, NJ, USA), and Eisai.

A Predictive Model for Kidney Failure After Nephrectomy for Localized Kidney Cancer: The Kidney Cancer Risk Equation.

RATIONALE & OBJECTIVE: Nephrectomy is the mainstay of treatment for individuals with localized kidney cancer. However, surgery can potentially result in the loss of kidney function or in kidney failure requiring dialysis/kidney transplantation. There are currently no clinical tools available to preoperatively identify which patients are at risk of kidney failure over the long term. Our study developed and validated a prediction equation for kidney failure after nephrectomy for localized kidney cancer. STUDY DESIGN: Population-level cohort study. SETTING & PARTICIPANTS: Adults (n=1,026) from Manitoba, Canada, with non-metastatic kidney cancer diagnosed between January 1, 2004, and December 31, 2016, who were treated with either a partial or radical nephrectomy and had at least 1 estimated glomerular filtration rate (eGFR) measurement before and after nephrectomy. A validation cohort included individuals in Ontario (n=12,043) with a diagnosis of localized kidney cancer between October 1, 2008, and September 30, 2018, who received a partial or radical nephrectomy and had at least 1 eGFR measurement before and after surgery. NEW PREDICTORS & ESTABLISHED PREDICTORS: Age, sex, eGFR, urinary albumin-creatinine ratio, history of diabetes mellitus, and nephrectomy type (partial/radical). OUTCOME: The primary outcome was a composite of dialysis, transplantation, or an eGFR<15mL/min/1.73m2 during the follow-up period. ANALYTICAL APPROACH: Cox proportional hazards regression models evaluated for accuracy using area under the receiver operating characteristic curve (AUC), Brier scores, calibration plots, and continuous net reclassification improvement. We also implemented decision curve analysis. Models developed in the Manitoba cohort were validated in the Ontario cohort. RESULTS: In the development cohort, 10.3% reached kidney failure after nephrectomy. The final model resulted in a 5-year area under the curve of 0.85 (95% CI, 0.78-0.92) in the development cohort and 0.86 (95% CI, 0.84-0.88) in the validation cohort. LIMITATIONS: Further external validation needed in diverse cohorts. CONCLUSIONS: Our externally validated model can be easily applied in clinical practice to inform preoperative discussions about kidney failure risk in patients facing surgical options for localized kidney cancer. PLAIN-LANGUAGE SUMMARY: Patients with localized kidney cancer often experience a lot of worry about whether their kidney function will remain stable or will decline if they choose to undergo surgery for treatment. To help patients make an informed treatment decision, we developed a simple equation that incorporates 6 easily accessible pieces of patient information to predict the risk of reaching kidney failure 5 years after kidney cancer surgery. We expect that this tool has the potential to inform patient-centered discussions tailored around individualized risk, helping ensure that patients receive the most appropriate risk-based care.

Subject-matter and intensional operators I: conditional-agnostic analytic implication.

Although logical settings are typically concerned with tracking alethic considerations, frameworks exist in which topic-theoretic considerations-e.g., tracking subject-matter or topic-are given equal importance. Intuitions about extending topic through a propositional language are generally straightforward for extensional cases. For a number of reasons, arriving at a compelling account of the subject-matter of intensional operators-such as intensional conditionals-is a more difficult task. In particular, the framework of topic-sensitive intentional modals (TSIMs) championed by Francesco Berto and his collaborators leave the topics of intensional formulae undefined, which artificially constricts the expressivity of the theory. This paper proposes an approach to fill in this lacuna, emphasizing an analogous problem in Parry-style containment logics. In this setting, the approach receives a proof-of-concept through the introduction of a natural and general family of subsystems of Parry's PAI-with sound and complete axiomatizations-that allow a fine degree of control over the topics of intensional conditionals.

Practical Utilization of Prediction Equations in Chronic Kidney Disease.

Chronic kidney disease (CKD) is common and can lead to kidney failure, cardiovascular complications, and early mortality. While nephrologists can provide valuable insights for patients at all stages of CKD, these scarce resources should be targeted at patients with the highest risk of progression and adverse outcomes. Prediction models are tools that can help providers risk stratify patients if they are effectively implemented into the clinical workflow. We believe these equations should demonstrate (1) clinical utility: where they can provide useful information to the physician and patients; and (2) clinical usability: where they are able to be easily integrated into clinical workflow and do not result in unnecessary costs or visits. CKD often remains unrecognized until later stages when a large window of opportunity to delay progression has already passed. Models to determine progression of CKD using thresholds such as a 40% decline in eGFR can provide clinical utility in risk stratifying patients at all stages of CKD, an endpoint that has been recommended by the FDA for the evaluation of drug approvals for disease-modifying therapies. For patients at more advanced stages of CKD with a greater risk of kidney failure, tools such as the kidney failure risk equation can be implemented to help guide most costly decisions, such as referral to multidisciplinary care, commencing dialysis modality education, or planning for vascular access placement surgery. In addition, models focused on determining outcomes following dialysis initiation can help inform shared decision-making between patient and provider to better inform decisions around conservative care. To ensure widespread adoption of these tools, it is important to ensure that they are broadly generalizable to many health settings and easily implemented into existing clinic workflows with minimum disruption.

Association of the Kidney Failure Risk Equation With High Health Care Costs.

Introduction: The Kidney Failure Risk Equations (KFRE) are accurate and validated to predict the risk of kidney failure in individuals with chronic kidney disease (CKD), but their potential to predict health care costs in the US health care system is unknown. We assessed the association of kidney failure risk from the 4-variable and 8-variable 2-year KFRE models with monthly health care costs in US patients with CKD stages G3 and G4. Methods: This was an ancillary study to a larger observational, retrospective cohort study examining the association between serum bicarbonate and adverse kidney outcomes. Monthly medical costs were calculated from individual health care insurance claims. Generalized linear regression models were used to examine the association of KFRE score with health care costs. Results: A total of 1721 patients qualified for the study (1475 and 246 with CKD stages G3 and G4, respectively). For 8-variable KFRE, each 1% (absolute) increase in risk was associated with 13.5% (P < 0.001) and 4.1% (P < 0.001) higher monthly costs for patients with CKD stage G3 and G4, respectively. For 4-variable KFRE, a 1% increase in risk was associated with 6.7% (P = 0.016) and 2.9% (P= 0.014) increase in monthly costs for patients with CKD stage G3 and G4, respectively. Conclusion: Higher risks of kidney failure as predicted by the 4-variable or 8-variable KFRE were associated with higher 2-year medical costs for patients with CKD stages G3 and G4. The KFRE may be a useful tool to anticipate medical costs and target cost-reducing interventions for patients at risk of kidney failure.

Evaluation of administrative case definitions for hypertension in Canadian children.

Hypertension is increasing in children and warrants disease surveillance. We therefore sought to evaluate the validity of case definitions to identify pediatric hypertension in administrative healthcare data. Cases of hypertension in children 3-18 years of age were identified utilizing blood pressures recorded in the Manitoba Primary Care Research Network (MaPCReN) electronic medical record from 2014 to 2016. Prevalence of hypertension and associated clinical characteristics were determined. We then evaluated the validity of 18 case definitions combining outpatient physician visits (ICD9CM codes), hospital claims (ICD9CM/ICD10 codes) and antihypertensive use within 1-3 years of data housed at the Manitoba Centre for Health Policy. The MaPCReN database identified 241 children with hypertension and 4090 without (prevalence = 5.6%). The sensitivity of algorithms ranged between 0.18 and 0.51 and the specificity between 0.98 and 1.00. Pharmaceutical use increased the sensitivity of algorithms significantly. The algorithms with the highest sensitivity and area under the ROC curve were 1 or more hospitalization OR 1 or more physician claim OR 1 or more pharmaceutical record. Evaluating 2 years of data is recommended. Administrative data alone reflects diagnosis of hypertension with high specificity, but underestimate the true prevalence of this disease. Alternative data sources are therefore required for disease surveillance.

AUTOPHAGY IN THE EYE: FROM PHYSIOLOGY TO PATHOPHYSOLOGY.

Autophagy is a catabolic self-degradative pathway that promotes the degradation and recycling of intracellular material through the lysosomal compartment. Although first believed to function in conditions of nutritional stress, autophagy is emerging as a critical cellular pathway, involved in a variety of physiological and pathophysiological processes. Autophagy dysregulation is associated with an increasing number of diseases, including ocular diseases. On one hand, mutations in autophagy-related genes have been linked to cataracts, glaucoma, and corneal dystrophy; on the other hand, alterations in autophagy and lysosomal pathways are a common finding in essentially all diseases of the eye. Moreover, LC3-associated phagocytosis, a form of non-canonical autophagy, is critical in promoting visual cycle function. This review collects the latest understanding of autophagy in the context of the eye. We will review and discuss the respective roles of autophagy in the physiology and/or pathophysiology of each of the ocular tissues, its diurnal/circadian variation, as well as its involvement in diseases of the eye.

Increasing Serum Bicarbonate is Associated With Reduced Risk of Adverse Kidney Outcomes in Patients with CKD and Metabolic Acidosis.

Introduction: Low serum bicarbonate at a single point in time is associated with accelerated kidney decline in patients with chronic kidney disease (CKD). We modeled how changes in serum bicarbonate over time affect incidence of adverse kidney outcomes. Methods: We analyzed data from Optum's deidentified Integrated Claims-Clinical data set of US patients (2007-2019) with ≥1 year of prior medical record data, CKD stages G3 to G5, and metabolic acidosis (i.e., index serum bicarbonate 12 to <22 mmol/l). The primary predictor of interest was the change in serum bicarbonate, evaluated at each postindex outpatient serum bicarbonate test as a time-dependent continuous variable. The primary outcome was a composite of either a ≥40% decline in estimated glomerular filtration rate (eGFR) from index or evidence of dialysis or transplantation, evaluated using Cox proportional hazards models. Results: A total of 24,384 patients were included in the cohort with median follow-up of 3.7 years. A within-patient increase in serum bicarbonate over time was associated with a lower risk of the composite kidney outcome. The unadjusted hazard ratio (HR) per 1-mmol/l increase in serum bicarbonate was 0.911 (95% confidence interval [CI]: 0.905-0.917; P < 0.001). After adjustment for baseline eGFR and serum bicarbonate, the time-adjusted effect of baseline eGFR and other covariates, the HR per 1-mmol/l increase in serum bicarbonate was largely unchanged (0.916 [95% CI: 0.910-0.922; P < 0.001]). Conclusion: In a real-world population of US patients with CKD and metabolic acidosis, a within-patient increase in serum bicarbonate over time independent of changes in eGFR, was associated with a lower risk of CKD progression.

A Machine Learning Model to Predict Diuretic Resistance.

BACKGROUND: Volume overload is a common complication encountered in hospitalized patients, and the mainstay of therapy is diuresis. Unfortunately, the diuretic response in some individuals is inadequate despite a typical dose of loop diuretics, a phenomenon called diuretic resistance. An accurate prediction model that predicts diuretic resistance using predosing variables could inform the right diuretic dose for a prospective patient. METHODS: Two large, deidentified, publicly available, and independent intensive care unit (ICU) databases from the United States were used-the Medical Information Mart for Intensive Care III (MIMIC) and the Philips eICU databases. Loop diuretic resistance was defined as <1400 ml of urine per 40 mg of diuretic dose in 24 hours. Using 24-hour windows throughout admission, commonly accessible variables were obtained and incorporated into the model. Data imputation was performed using a highly accurate machine learning method. Using XGBoost, several models were created using train and test datasets from the eICU database. These were then combined into an ensemble model optimized for increased specificity and then externally validated on the MIMIC database. RESULTS: The final ensemble model was composed of four separate models, each using 21 commonly available variables. The ensemble model outperformed individual models during validation. Higher serum creatinine, lower systolic blood pressure, lower serum chloride, higher age, and female sex were the most important predictors of diuretic resistance (in that order). The specificity of the model on external validation was 92%, yielding a positive likelihood ratio of 3.46 while maintaining overall discrimination (C-statistic 0.69). CONCLUSIONS: A diuretic resistance prediction model was created using machine learning and was externally validated in ICU populations. The model is easy to use, would provide actionable information at the bedside, and would be ready for implementation in existing electronic medical records. This study also provides a framework for the development of future machine learning models.

Metabolic acidosis is undertreated and underdiagnosed: a retrospective cohort study.

BACKGROUND: Guidelines recommend treatment of metabolic acidosis (MA) in patients with chronic kidney disease (CKD), but the diagnosis and treatment rates in real-world settings are unknown. We investigated the frequency of MA treatment and diagnosis in patients with CKD. METHODS: In this retrospective cohort study, we examined administrative health data from two US databases [Optum's de-identified Integrated Claims + Clinical Electronic Health Record Database (US EMR cohort; 1 January 2007 to 30 June 2019) and Symphony Health Solutions IDV® (US claims cohort; 1 May 2016 to 30 April 2019)] and population-level databases from Manitoba, Canada (1 April 2006 to 31 March 2018). Patients who met laboratory criteria indicative of CKD and chronic MA were included: two consecutive estimated glomerular filtration results <60 mL/min/1.73 m2 and two serum bicarbonate results 12 to <22 mEq/L over 28-365 days. Outcomes included treatment of MA (defined as a prescription for oral sodium bicarbonate) and a diagnosis of MA (defined using administrative records). Outcomes were assessed over a 3-year period (1 year pre-index, 2 years post-index). RESULTS: A total of 96 184 patients were included: US EMR, 6179; Manitoba, 3223; US Claims, 86 782. Sodium bicarbonate treatment was prescribed for 17.6%, 8.7% and 15.3% of patients, and a diagnosis was found for 44.7%, 20.9% and 20.9% of patients, for the US EMR, Manitoba and US Claims cohorts, respectively. CONCLUSIONS: This analysis of 96 184 patients with laboratory-confirmed MA from three independent cohorts of patients with CKD and MA highlights an important diagnosis and treatment gap for this disease-modifying complication.

Metabolic Acidosis is Associated With Acute Kidney Injury in Patients With CKD.

Introduction: Metabolic acidosis in patients with chronic kidney disease (CKD) results from a loss of kidney function. It has been associated with CKD progression, all-cause mortality, and other adverse outcomes. We aimed to determine whether metabolic acidosis is associated with a higher risk of acute kidney injury (AKI). Methods: This was a retrospective cohort study. Using electronic health records and administrative data, we enrolled 2 North American cohorts of patients with CKD Stages G3-G5 as follows: (i) 136,067 patients in the US electronic medical record (EMR) based cohort; and (ii) 34,957 patients in the Manitoba claims-based cohort. The primary exposure was metabolic acidosis (serum bicarbonate between 12 mEq/l and <22 mEq/l). The primary outcome was the development of AKI (defined using ICD-9 and 10 codes at hospital admission or a laboratory-based definition based on Kidney Disease: Improving Global Outcomes guidelines). We applied Cox proportional hazards regression models adjusting for relevant demographic and clinical characteristics. Results: In both cohorts, metabolic acidosis was associated with AKI: hazard ratio (HR) 1.57 (95% confidence interval [CI] 1.52-1.61) in the US EMR cohort, and HR 1.65 (95% CI 1.58-1.73) in the Manitoba claims cohort. The association was consistent when serum bicarbonate was treated as a continuous variable, and in multiple subgroups, and sensitivity analyses including those adjusting for albuminuria. Conclusion: Metabolic acidosis is associated with a higher risk of AKI in patients with CKD. AKI should be considered as an outcome in studies of treatments for patients with metabolic acidosis.

Pembrolizumab versus placebo as post-nephrectomy adjuvant therapy for clear cell renal cell carcinoma (KEYNOTE-564): 30-month follow-up analysis of a multicentre, randomised, double-blind, placebo-controlled, phase 3 trial.

BACKGROUND: The first interim analysis of the KEYNOTE-564 study showed improved disease-free survival with adjuvant pembrolizumab compared with placebo after surgery in patients with clear cell renal cell carcinoma at an increased risk of recurrence. The analysis reported here, with an additional 6 months of follow-up, was designed to assess longer-term efficacy and safety of pembrolizumab versus placebo, as well as additional secondary and exploratory endpoints. METHODS: In the multicentre, randomised, double-blind, placebo-controlled, phase 3 KEYNOTE-564 trial, adults aged 18 years or older with clear cell renal cell carcinoma with an increased risk of recurrence were enrolled at 213 hospitals and cancer centres in North America, South America, Europe, Asia, and Australia. Eligible participants had an Eastern Cooperative Oncology Group performance status of 0 or 1, had undergone nephrectomy 12 weeks or less before randomisation, and had not received previous systemic therapy for advanced renal cell carcinoma. Participants were randomly assigned (1:1) via central permuted block randomisation (block size of four) to receive pembrolizumab 200 mg or placebo intravenously every 3 weeks for up to 17 cycles. Randomisation was stratified by metastatic disease status (M0 vs M1), and the M0 group was further stratified by ECOG performance status and geographical region. All participants and investigators involved in study treatment administration were masked to the treatment group assignment. The primary endpoint was disease-free survival by investigator assessment in the intention-to-treat population (all participants randomly assigned to a treatment). Safety was assessed in the safety population, comprising all participants who received at least one dose of pembrolizumab or placebo. As the primary endpoint was met at the first interim analysis, updated data are reported without p values. This study is ongoing, but no longer recruiting, and is registered with ClinicalTrials.gov, NCT03142334. FINDINGS: Between June 30, 2017, and Sept 20, 2019, 994 participants were assigned to receive pembrolizumab (n=496) or placebo (n=498). Median follow-up, defined as the time from randomisation to data cutoff (June 14, 2021), was 30·1 months (IQR 25·7-36·7). Disease-free survival was better with pembrolizumab compared with placebo (HR 0·63 [95% CI 0·50-0·80]). Median disease-free survival was not reached in either group. The most common all-cause grade 3-4 adverse events were hypertension (in 14 [3%] of 496 participants) and increased alanine aminotransferase (in 11 [2%]) in the pembrolizumab group, and hypertension (in 13 [3%] of 498 participants) in the placebo group. Serious adverse events attributed to study treatment occurred in 59 (12%) participants in the pembrolizumab group and one (<1%) participant in the placebo group. No deaths were attributed to pembrolizumab. INTERPRETATION: Updated results from KEYNOTE-564 support the use of adjuvant pembrolizumab monotherapy as a standard of care for participants with renal cell carcinoma with an increased risk of recurrence after nephrectomy. FUNDING: Merck Sharp & Dohme LLC, a subsidiary of Merck & Co, Inc, Rahway, NJ, USA.

Development and External Validation of a Machine Learning Model for Progression of CKD.

Introduction: Prediction of disease progression at all stages of chronic kidney disease (CKD) may help improve patient outcomes. As such, we aimed to develop and externally validate a random forest model to predict progression of CKD using demographics and laboratory data. Methods: The model was developed in a population-based cohort from Manitoba, Canada, between April 1, 2006, and December 31, 2016, with external validation in Alberta, Canada. A total of 77,196 individuals with an estimated glomerular filtration rate (eGFR) > 10 ml/min per 1.73 m2 and a urine albumin-to-creatinine ratio (ACR) available were included from Manitoba and 107,097 from Alberta. We considered >80 laboratory features, including analytes from complete blood cell counts, chemistry panels, liver enzymes, urine analysis, and quantification of urine albumin and protein. The primary outcome in our study was a 40% decline in eGFR or kidney failure. We assessed model discrimination using the area under the receiver operating characteristic curve (AUC) and calibration using plots of observed and predicted risks. Results: The final model achieved an AUC of 0.88 (95% CI 0.87-0.89) at 2 years and 0.84 (0.83-0.85) at 5 years in internal testing. Discrimination and calibration were preserved in the external validation data set with AUC scores of 0.87 (0.86-0.88) at 2 years and 0.84 (0.84-0.86) at 5 years. The top 30% of individuals predicted as high risk and intermediate risk represent 87% of CKD progression events in 2 years and 77% of progression events in 5 years. Conclusion: A machine learning model that leverages routinely collected laboratory data can predict eGFR decline or kidney failure with accuracy.