Medical Policy

This document addresses the use of a machine learning derived probability score (i.e., artificial intelligence) which may combine a variety of clinical characteristics such as, biomarkers, genetics, gender or race, to generate prognostic information with the end-goal of facilitating a more personalized approach to the management of chronic kidney disease (e.g., KidneyIntelX,^™ Renalytix AI, Inc., London, England). This document does not address the standard use of blood-based biomarkers, estimated glomerular filtration rate (eGFR) or urinary albumin and creatinine levels in the prognostic evaluation of newly diagnosed kidney disease.

Investigational and Not Medically Necessary:

Use of a machine learning derived probability score (e.g., KidneyIntelX) to predict rapid kidney function decline in chronic kidney disease is considered investigational and not medically necessary for all indications.

Chronic kidney disease (CKD) is defined by the Kidney Disease Improving Global Outcomes (KDIGO) organization as abnormalities of kidney structure or function, present for > 3 months. In the KDIGO Clinical Practice Guidelines for the Evaluation and Management of Chronic Kidney Disease, factors associated with CKD progression to inform prognosis include the etiology of CKD (e.g., diabetes, hypertension, etc.), level of GFR, level of albuminuria, age, sex, race/ethnicity, elevated blood pressure, hyperglycemia, dyslipidemia, smoking, obesity, history of cardiovascular disease and ongoing exposure to nephrotoxic agents (ungraded recommendation; KDIGO, 2024). A standardized system for integrating sociodemographic risk factors with clinically relevant biomarkers to accurately identify those most at risk for progression is not yet available in most practice settings, potentially hampering clinicians’ timely intervention in CKD management.

Recently, the use of machine learning approaches that can combine biomarkers and electronic health record data to produce prognostic risk scores have been explored. One such approach is the KidneyIntelX, a proprietary artificial intelligence-enabled algorithm which combines blood-based biomarkers, genetics and personalized data from electronic health records to generate a unique risk score which is then used to develop a prediction of progressive kidney function decline in diabetes-related CKD. On June 29, 2023 KidneyIntelX was granted Food and Drug Administration (FDA) De Novo marketing authorization to aid in the assessment of risk of progressive kidney function decline in adults diagnosed with Type 2 diabetes and existing chronic kidney disease. The FDA notes that KidneyIntelX is not intended for screening or as a stand-alone diagnostic test. KidneyIntelX is classified by the FDA as a

Prognostic test for assessment of chronic kidney disease progression. A prognostic test for assessment of chronic kidney disease progression is an in vitro diagnostic device intended to measure one or more analytes obtained from human samples as an aid in assessing the risk for progression of chronic kidney disease. This device is not intended for diagnosis of any disease, for serial monitoring of kidney disease progression, or for monitoring the effect of any therapeutic product.

In 2020, Chan and colleagues published results of a study evaluating the clinical utility of KidneyIntelX. This retrospective cohort enrolled 1146 individuals with diabetes-related CKD age 21-81. During the follow-up period (median of 4.3 years) 241 study enrollees (21%) experienced progressive decline in kidney function. KidneyIntelX stratified 46%, 37% and 16.5% of validation cohort (n=460) into low-, intermediate- and high-risk groups, respectively, with a positive predictive value (PPV) of 62% (PPV of 37% for the clinical model and 40% for KDIGO; p<0.001) in the high-risk group and a negative predictive value (NPV) of 91% in the low-risk group. The net reclassification index for events into the high-risk group was 41% (p<0.05). In this retrospective, exploratory validation study, KidneyIntelX scores accurately classified more cases into the KidneyIntelX-defined low, intermediate and high-risk strata (p-value<0.05) relative to KDIGO risk strata. The study authors conclude, “A machine learned model combining plasma biomarkers and EHR [electronic health record] data improved prediction of progressive decline in kidney function within 5 years over KDIGO and standard clinical models in patients with early DKD [diabetes-related CKD].” Given the retrospective study design and marginal statistical significance, further investigation in the setting of a large, ideally randomized, trial is warranted to establish whether use of KidneyIntelX materially improves net health outcomes compared to established alternatives, such as the KDIGO guideline’s specified sociodemographic risk factors, pertinent health history and clinically relevant biomarkers.

In 2022, Lam and colleagues published results from a retrospective study of samples collected during conduction of a prospective randomized controlled trial, CANagliflozin cardioVascular Assessment Study (CANVAS). In total, 1325 CANVAS participants with diabetic kidney disease and baseline plasma samples were enrolled into the study. KidneyIntelX risk scores were generated from the available samples at baseline and years 1, 3, and 6 of study follow-up. The study’s primary aim was to assess the association of changes from baseline in KidneyIntelX scores with progression of diabetic kidney disease; composite outcomes included, (1) rapid kidney function decline, (2) a sustained 40% decline in eGFR, or (3) kidney failure. During the mean follow-up of 5.6 years, 131 study participants (9.9%) experienced a composite kidney outcome. Using risk cutoffs established from previous validation studies, KidneyIntelX stratified participants into low- (42%), intermediate- (44%), and high-risk (15%) groups with cumulative incidence for the outcomes of 3%, 11%, and 26%, respectively (risk ratio=8.4; 95% confidence interval [CI], 5.0-14.2) for the high-risk versus low-risk groups. Changes in KidneyIntelX score within the first year were significantly associated with future risk of a composite outcome (odds ratio [OR; per 10 unit decrease]=0.80; 95% CI, 0.77, 0.83; p<0.001). Study authors conclude that “KidneyIntelX risk-stratified a large multinational external cohort for progression of DKD [diabetic kidney disease]…”. Given the retrospective design and unclear clinical significance of these findings, further study is warranted to determine the impact of KidenyIntelX on net health outcomes.

In 2022, Tokita and colleagues evaluated the clinical performance of KidneyIntelX in a large hospital system over a 6-month follow-up period. Study outcomes included visit frequency, medication management changes, referral to a specialist and selected lab values. A total of 1686 individuals were enrolled and tested using KidneyIntelX scoring and the hospital’s pathway management for individuals with stages 1 to 3 diabetic kidney disease. Following testing with KidneyIntelX, a clinical encounter occurred in the first month in 13%, 43%, and 53% of low-risk, intermediate-risk, and high-risk individuals, respectively and in 46%, 61%, and 71% of the study participants at least 1 action was implemented within the first 6 months. Participants classified as high-risk were more likely to be placed on SGLT2 inhibitors (OR=4.56; 95% CI, 3.00-6.91 vs low-risk), and more likely to be referred to a specialist such as a nephrologist, endocrinologist, or dietician (OR=2.49; 95% CI, 1.53-4.01) compared to participants classified as low-risk. Systolic blood pressure (49% of participants were hypertensive at baseline) and eGFR remain unchanged across all 3 risk stratification levels throughout the study. The addition of the KidneyIntelX to the management of individuals with early diabetic kidney disease did not demonstrate a clinically meaningful benefit in this prospective trial.

In 2024, Tokita and colleagues conducted a post-hoc subset analysis of pre- and post-test measurements in 5348 individuals diagnosed with type 2 diabetes and diabetic kidney disease followed for at least 12 months (Original Observational Trial: NCT04802395). The baseline risk level predicted by the KidneyIntelX assay was low in 49% of study participants, intermediate in 40%, and high in 11%. New prescriptions or referrals to specialists were made for 19%, 33%, and 43% of these risk groups, respectively. The median A1C decreased from 8.2% pre-test to 7.5% post-test in the high-risk group (p<0.001). The median eGFR slope improved in high-risk (p=0.0003), intermediate-risk (p<0.001), and in low-risk (p<0.001) study participants. The use of KidneyIntelX in this study was associated with increased actions to enhance cardio-kidney-metabolic health, including medications and specialist referrals. While this investigation demonstrated divergence in clinical decision making potentially associated with KidneyIntelX stratification, clinically relevant outcome measures did not differ significantly between the 3 risk categories at 12 months.

In 2023, Sanmarchi and colleagues published results from a systematic review which aimed to assess the application of artificial intelligence (AI) and machine learning in predicting, diagnosing, and treating CKD. From an initial 648 studies, 68 met the inclusion criteria. The most common use of AI was prognostic prediction (e.g., KidneyIntelX), followed by diagnosis of CKD. The metrics reported varied across articles which limited direct comparisons of performance and generalizability of the algorithm via testing in diverse populations was rarely taken into account. Only a small portion (6 out of the 68 studies) had clinical evaluation and validation of the models or algorithms. The authors conclude,

Machine learning is a promising tool for the prediction of risk, diagnosis, and therapy management for CKD patients. Nonetheless, future work is needed to address the interpretability, generalizability, and fairness of the models to ensure the safe application of such technologies in routine clinical practice.

The most recent KDIGO Clinical Practice Guideline for the Evaluation and Management of CKD states the following:

…3 recent publications present models for people with or without diabetes, using both regression and machine learning–based methods, with or without biomarkers. Given the potential utility of these new models to identify high-risk people for early intervention, they should be used to predict disease progression in people with CKD G1–G2 and may supplement established risk equations among people with CKD G3. People with CKD identified as intermediate risk (e.g., >1% per year) with these tools may benefit from the earlier initiation of therapy and closer follow-up, and those identified as high risk (e.g., >5% per year) may have the largest benefit from multidrug therapy to slow progression (KDIGO, 2024).

Other Relevant Information

KidneyIntelX was granted FDA authorization to aid in the assessment of risk for progressive kidney function decline in adults diagnosed with Type 2 diabetes and existing chronic kidney disease. The FDA notes that KidneyIntelX is not intended for screening or as a stand-alone diagnostic test. The Centers for Medicare & Medicaid Services (CMS) Local Coverage Determination (LCD) L39726 entitled, KidneyIntelX and KidneyIntelX.dkd Testing, provides a framework for coverage consistent with demonstration of clinical utility. No CMS NCD addressing the use of KidneyIntelX was identified.

Summary

Although KidneyIntelX is listed as a model with ‘potential utility’ in the KDIGO consensus guideline statement, the only cited clinical study is a retrospective cohort; limitations previously described (Chan, 2020). The published peer-reviewed medical literature has not established KidneyIntelX, or any technology like it, as a proven method that materially improves net health outcomes nor has any benefit been established beyond currently available alternatives.

In 2023, approximately 35.5 million Americans reportedly had chronic kidney disease (CKD), with over 131, 000 requiring initiation of treatment for kidney failure, also known as end stage renal disease (ESRD). There was a steady rise in the rate of ESRD from 1980 to 2011, since then, the incidence rate of ESRD has started to decline. The most prevalent causes of kidney disease are diabetes and hypertension, which account for approximately 38% and 27% of ESRD cases, respectively (CDC, 2023). On average, 50,000 individuals with diabetic kidney disease progress to kidney failure annually in the United States (Chan, 2020).

Predicting which newly diagnosed diabetic kidney disease cases may progress to ESRD has proved challenging for clinicians, potentially resulting in delayed diagnosis of individuals and the subsequent need for life-saving dialysis or kidney transplants. Typically, prognosis is achieved through integration of established sociodemographic risk factors (i.e., smoking, obesity, and race/ethnicity) along with clinically relevant biomarkers, such as glycemic levels, eGFR, and lipid levels. KidneyIntelX is described by the manufacturer (RenalytixAI) as a validated machine-learned, prognostic risk score which combines data from EHRs and circulating biomarkers to predict diabetic kidney disease progression. More specifically, KidneyIntelX combines three blood-based biomarkers (tumor necrosis factor receptor [TNFR]1, TNFR2 and kidney injury molecule 1) with seven clinical indicators (eGFR, urine albumin-creatinine ratio, serum calcium, hemoglobin A1C, systolic blood pressure, platelets, and aspartate aminotransferase [AST]) to generate a risk score purported to help clinicians determine if an individual is at low, intermediate, or high risk for rapid decline in kidney function. The end goal of incorporating KidneyIntelX into practice, is to slow the progression of kidney disease and progressive kidney function decline which may result in kidney failure and ultimately long-term dialysis or kidney transplant (Chan, 2020).

Artificial Intelligence (AI): A science of computer simulated thinking processes and human behaviors, which involves computer science, psychology, philosophy and linguistics.

Chronic renal disease: The permanent loss of kidney function.

End stage renal disease: Persistent decline in renal function as documented by falling creatinine clearance in an individual diagnosed with a renal disease whose natural history is progression to renal impairment requiring renal replacement (dialysis or transplant).

Glomerular filtration rate (GFR): A test used to check how well the kidneys are functioning by estimating how much blood passes through the glomeruli each minute.

Glomeruli: A cluster of nerve endings, spores, or small blood vessels, in particular a cluster of capillaries around the end of a kidney tubule, where waste products are filtered from the blood.

The following codes for treatments and procedures applicable to this document are included below for informational purposes. Inclusion or exclusion of a procedure, diagnosis or device code(s) does not constitute or imply member coverage or provider reimbursement policy. Please refer to the member's contract benefits in effect at the time of service to determine coverage or non-coverage of these services as it applies to an individual member.

When services are Investigational and Not Medically Necessary:
For the following procedure code, or when the code describes a procedure indicated in the Position Statement section as investigational and not medically necessary.

Peer Reviewed Publications:

1. Chan L, Nadkarni G, Fleming F, et al. Derivation and validation of a machine learning risk score using biomarker and electronic patient data to predict progression of diabetic kidney disease. Diabetologia. 2021; 64(7):1504-1515.
2. Chauhan K, Nadkarni GN, Fleming F, et al. Initial validation of a machine learning-derived prognostic test (KidneyIntelX) integrating biomarkers and electronic health record data to predict longitudinal kidney outcomes. Kidney360. 2020; 1(8):731-739.
3. Lam D, Nadkarni GN, Mosoyan G, Net al. Clinical utility of KidneyIntelX in early stages of diabetic kidney disease in the CANVAS Trial. Am J Nephrol. 2022; 53(1):21-31.
4. Tokita J, Lam D, Vega A, et al. A real-world precision medicine program including the KidneyIntelX test effectively changes management decisions and outcomes for patients with early-stage diabetic kidney disease. J Prim Care Community Health. 2024; 15:1-10.
5. Tokita J, Vega A, Sinfield C, et al. Real world evidence and clinical utility of KidneyIntelX on patients with early-stage diabetic kidney disease: interim results on decision impact and outcomes. J Prim Care Community Health. 2022. Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9677284/pdf/10.1177_21501319221138196.pdf. Accessed on March 28, 2024.
6. Sanmarchi F, Fanconi C, Golinelli D, et al. Predict, diagnose, and treat chronic kidney disease with machine learning: a systematic literature review. J Nephrol. 2023; 36(4):1101-1117.

Government Agency, Medical Society, and Other Authoritative Publications:

1. Centers for Disease Control and Prevention. Chronic Kidney Disease. Chronic Kidney Disease in the United States, 2023. Available at: CKD-Factsheet-H.pdf (cdc.gov). Accessed on March 28, 2024.
2. Centers for Medicare and Medicare Services (CMS). Local Coverage Determination (LCD): KidneyIntelX and KidneyIntelX.dkd Testing (L39726). Effective 08/01/2024. Available at: https://www.cms.gov/medicare-coverage-database/view/lcd.aspx?lcdid=39726&ver=3. Accessed on September 28, 2024.
3. Kidney Disease: Improving Global Outcomes (KDIGO) CKD Work Group. KDIGO 2024 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease. Kidney Int. 2024; 105(4S):S117-S314. 
4. U.S. Food and Drug Administration (FDA) Denovo Notification Database. KidneyIntelX Summary of Safety and Effectiveness. DEN200052. Rockville, MD: FDA. June 29, 2023. Available at: https://www.accessdata.fda.gov/cdrh_docs/pdf20/DEN200052.pdf. Accessed on March 28, 2024.

1. American Diabetes Association. Type 2 diabetes. Available at: http://www.diabetes.org/diabetes-basics/type-2/?loc=db-slabnav/. Accessed on March 28, 2024.
2. American Society of Nephrology. Available at: https://www.asn-online.org/. Accessed on March 28, 2024.
3. National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). What is Kidney Failure? Updated January 2018. Available at https://www.niddk.nih.gov/health-information/kidney-disease/kidney-failure/what-is-kidney-failure. Accessed on March 28, 2024.

IntelxDKD™
KidneyIntelX

The use of specific product names is illustrative only. It is not intended to be a recommendation of one product over another, and is not intended to represent a complete listing of all products available.