Disclosures: Chauhan reports no relevant financial disclosures. Several authors report receiving financial compensation as consultants and advisory board members for RenalytixAI Inc.
July 02, 2020
2 min read

Machine-learning test may improve kidney failure prediction in patients with diabetes

Disclosures: Chauhan reports no relevant financial disclosures. Several authors report receiving financial compensation as consultants and advisory board members for RenalytixAI Inc.
You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

For patients with type 2 diabetes or the APOL1-HR genotype, a machine learning test integrating biomarkers and electronic health record data demonstrated improved prediction of kidney failure compared with commonly used clinical models.

According to Kinsuk Chauhan, MD, MPH, of Icahn School of Medicine at Mount Sinai, and colleagues, diabetic kidney disease from type 2 diabetes accounts for 44% of all patients with end-stage kidney disease, with the APOL1 high-risk genotypes also associated with increased risk for chronic kidney disease progression and eGFR decline that may ultimately result in kidney failure.

Doctor using an electronic health record
Source: Adobe Stock

“Even though these populations are on average higher risk than the general population, accurate prediction of who will have rapid kidney function decline (RKFD) and worse kidney outcomes is lacking,” the researchers wrote, noting that the current standard of using the kidney failure risk equation to predict ESKD has only been validated in patients who already have kidney disease and not in those with preserved kidney function at baseline.

“Widespread electronic health records (EHR) usage provides the potential to leverage thousands of clinical features,” the researchers added. “Standard statistical approaches are inadequate to leverage this data due to feature volume, unaligned nature of data and correlation structure.”

Collecting plasma samples from the Mount Sinai BioMe Biobank, the researchers measured plasma tumor necrosis factors 1/2 (TNFR 1/2) and kidney injury molecule-1 (KIM-1) in two separate cohorts (871 patients with diabetes; 498 patients with APOL1-HR). Data related to these biomarkers were subsequently integrated, along with EHR data, into the KidneyIntelX prognostic test, which utilizes machine-learning algorithms to predict kidney outcomes. More specifically, the test generates a risk score for a composite outcome of rapid kidney function decline (defined as eGFR decline >5 mL/min/1.73m2/year), 40% sustained eGFR decline or kidney failure, stratifying patients into risk categories.

Patients with type 2 diabetes were followed for a median of 4.6 years; patients with APOL1-HR were followed for a median of 5.1 years (23% of those with type 2 diabetes and 18% of those with APOL1-HR experienced the composite outcome during follow-up).

Researchers determined the area under the curve (AUC) of KidneyIntelX to be 0.77 in the cohort with type 2 diabetes and 0.80 in the cohort with APOL1-HR. These results showed the machine-learning test out-performed the clinical models for both groups, as these demonstrated AUCs of 0.66 and 0.72, respectively.

For patients in the high-risk strata with either type 2 diabetes or APOL1-HR (top 15%), the positive predictive values for KidneyIntelX were 62% for both cohorts compared with 46% and 39% for the clinical models.

In the low-risk strata (bottom 50%), the negative predictive values for KidneyIntelX were 92% in the type 2 diabetes cohort and 96% in the APOL1-HR cohort vs. 85% and 93% for the clinical model, respectively.

Researchers noted that for both cohorts, the features that most contributed to the improved performance of KidneyIntelX were the three plasma biomarkers (TNFR-1, TNFR-2 and KIM-1) and laboratory values or vital signs (either baseline or changes over time) that have been linked to kidney disease.

According to Chauhan and colleagues, utilizing machine learning to predict kidney outcomes has “near-term clinical implications,” and should be linked to clinical decision support and embedded care pathways within the EHR.

The researchers suggested patients with a high-risk score should be promptly referred to a nephrologist.

“In addition,” they wrote, “referral to a dietician and delivery of educational materials regarding the importance and consequences of CKD to the high-risk patients should increase awareness and facilitate motivation for changes in lifestyles and behavior. Finally, the optimization of medical therapy including renin-angiotensin aldosterone system inhibitors, statins for cardiovascular risk management and intensification of antihypertensive medication to meet guideline recommended blood pressure targets can be pursued.”