The Predictive Capacity of Machine Learning Models for Progressive Kidney Disease in Individuals With Sickle Cell Anemia

Sickle Cell Disease Clinical Trial

— PREMIER  

Official title:

Predicting Progression of Chronic Kidney Disease in Sickle Cell Anemia Using Machine Learning Models [PREMIER]

Clinical Trial Details — Status: Recruiting

Administrative data

• NCT number: NCT05214105
• Other study ID #: 2021-0746
• Secondary ID: 1R01HL159376-01
• Status: Recruiting
• Start date: July 5, 2022
• Est. completion date: January 31, 2026

Study information

• Verified date: December 2023
• Source: University of Tennessee
• Contact: Kenneth I Ataga, MD
• Phone: 901-448-2813
• Email: kataga[@]uthsc.edu
• Is FDA regulated: No
• Study type: Observational

Clinical Trial Summary

This is a multicenter prospective, longitudinal cohort study which will evaluate the predictive capacity of machine learning (ML) models for progression of CKD in eligible patients for a minimum of 12 months and potentially for up to 4 years.

Description:

Sickle cell disease (SCD) is characterized by a vasculopathy affecting multiple end organs, with complications including ischemic stroke, pulmonary hypertension, and chronic kidney disease (CKD). Albuminuria, an early measure of glomerular injury and a manifestation of CKD, is common in SCD and predicts progressive kidney disease. Kidney function decline is faster in SCD patients than in the general African American population. The prevalence of rapid decline, commonly defined as an estimated glomerular filtration rate (eGFR) decline of >3 mL/min/1.73 m2 per year, is ~ 31% in SCD, 3-fold higher than in the general population. Furthermore, high-risk Apolipoprotein 1 (APOL1) variants are associated with an increased risk of albuminuria and progression of CKD in SCD. It is well recognized that kidney disease, regardless of severity, is associated with increased mortality in SCD. The investigators have recently observed that rapid eGFR decline is also independently associated with increased mortality in SCD. Early identification of patients at risk for progression of CKD is important to address potentially modifiable risk factors, slow eGFR decline and reduce mortality. The investigators have previously reported that machine learning (ML) models can identify patients at high risk for rapid decline in kidney function. In this study, the investigators propose the conduct of a prospective, multi-center study to build a ML-based predictive model for progression of CKD in adults with SCD. A model with high predictive capacity for progression of CKD not only affords risk-stratification, but also offers opportunities to modify known risk factors in hopes of attenuating kidney function loss and decreasing mortality risk. The overall hypothesis is that ML models utilizing clinical and laboratory characteristics, additional biomarkers and genetic assessments have a higher predictive capacity for progression of CKD than persistent albuminuria alone in adults with sickle cell anemia.

Recruitment information / eligibility

• Status: Recruiting
• Enrollment: 400
• Est. completion date: January 31, 2026
• Est. primary completion date: January 31, 2026
• Accepts healthy volunteers: No
• Gender: All
• Age group: 18 Years to 65 Years

Eligibility

Inclusion Criteria:

1. HbSS or HbSß0 thalassemia, 18 - 65 years old;

2. non-crisis, "steady state" with no acute pain episodes requiring medical contact in preceding 4 weeks;

3. ability to understand the study requirements.

Exclusion Criteria:

1. pregnant at enrollment;

2. poorly controlled hypertension;

3. long-standing diabetes with suspicion for diabetic nephropathy;

4. connective tissue disease such as systemic lupus erythematosus (SLE);

5. polycystic kidney disease or glomerular disease unrelated to SCD;

6. stem cell transplantation;

7. untreated human immunodeficiency virus (HIV), hepatitis B or C infection; h) history of cancer in last 5 years; i) End-stage renal disease (ESRD) on chronic dialysis; j) prior kidney transplantation.

Related Conditions & MeSH terms

• Anemia
• Anemia, Sickle Cell
• Kidney Diseases
• Kidney Diseases, Chronic
• Renal Insufficiency, Chronic
• Sickle Cell Disease

Intervention

Other:
• Biospecimen/DNA collection and analysis
Patients will be followed longitudinally with collection of CBC and chemistries as well as research biomarkers (urine, plasma, and genomic materials).

Locations

Country	Name	City	State
United States	University of Illinois at Chicago	Chicago	Illinois
United States	The University of Tennessee Health Science Center	Memphis	Tennessee
United States	Wake Forest University	Winston-Salem	North Carolina

Sponsors (7)

Lead Sponsor	Collaborator
University of Tennessee	National Heart, Lung, and Blood Institute (NHLBI), University of Illinois at Chicago, University of Memphis, University of North Carolina, Chapel Hill, University of North Carolina, Charlotte, Wake Forest University

Country where clinical trial is conducted

United States, 

Outcome

Type	Measure	Description	Time frame	Safety issue
Primary	Develop two separate predictive models for progression of CKD (eGFR <90 mL/min/1·73 m2 and =25% drop in eGFR from baseline) and rapid eGFR decline (eGFR loss >3·0 mL/min/1·73 m2 per year) over the 12 months following the baseline clinic evaluation.	At each visit following the first 12 months, rate of eGFR change will be calculated using data from current and earlier visits.	12 months
Secondary	Alternate definitions of CKD progression as eGFR decline <90 mL/min/1·73 m2 and =50% drop in eGFR from baseline, and rapid eGFR decline as eGFR loss >5·0 mL/min/1·73 m2 per year will be evaluated.	At each visit following the first 12 months, rate of eGFR change will be calculated using data from current and earlier visits.	12 months
Secondary	Evaluate the effect of APOL1 on the predictive capacity of ML models. Genomic DNA will be extracted from whole blood collected at baseline visits using standard techniques and genotyping will be performed as previously described.	At each visit following the first 12 months, rate of eGFR change will be calculated using data from current and earlier visits	12 months