End-Stage Kidney Disease, End Stage Renal Disease (ESRD), End Stage Renal Disease on Dialysis (Diagnosis), End Stage Renal Failure on Dialysis, Peritoneal Dialysis, Peritoneal Dialysis Patients
Conditions
Keywords
Peritoneal dialysis, Artificial intelligence, Dialysis adequacy, Peritoneum transporter status, Renal replacement therapy
Brief summary
The goal of this prospective diagnostic test (correlation) study is to develop and investigate the performance of artificial intelligence in predicting peritoneum transporter status and dialysis efficiency in adult patients undergoing peritoneal dialysis (PD). The main questions it aims to answer are: Can artificial intelligence predict peritoneal transporter status based on simple clinical and biochemical measurements? Can artificial intelligence predict dialysis adequacy (Kt/V) using these features? Researchers will compare the performance of the AI model with the gold standard Peritoneal Equilibration Test (PET) and Kt/V to evaluate its accuracy and reliability. Participants will: Provide peritoneal dialysate and spot urine samples for biochemical analysis. Undergo routine dialysis adequacy and peritoneal equilibration testing (PET). Have clinical and laboratory data collected for AI model training and validation. The study will recruit approximately 350 peritoneal dialysis patients, with 280 participants in the training/validation arm and 70 participants in the test arm. The study duration is 12 months following enrollment.
Detailed description
The DETECT-PD (Dialysis Efficiency and Transporter Evaluation Computational Tool in Peritoneal Dialysis) study is a double-blind, prospective diagnostic test (correlation) study designed to evaluate the feasibility and effectiveness of artificial intelligence (AI) in predicting peritoneal transporter status and dialysis efficiency in patients undergoing peritoneal dialysis (PD). The study aims to develop a computational model that leverages clinical, biochemical, and peritoneal transport data to provide a non-invasive and efficient assessment tool, ultimately improving dialysis management and patient outcomes. Patient recruitment and data collection will be conducted during routine dialysis adequacy and peritoneal transporter status assessments. The following clinical and biochemical parameters will be collected: Demographics & Medical History Peritoneal Dialysis Data Biochemical Data The AI model will be developed using Python 3.11 and PyTorch 2.41 for deep learning and predictive analytics. The key methodological steps include: Data Preprocessing: Handling missing values, feature scaling, and one-hot encoding for categorical variables. Feature Selection: Identifying the most predictive clinical and biochemical markers. Model Training: Using deep learning regression models to predict PET and Kt/V outcomes. Performance Evaluation: Evaluating model accuracy using: Mean Absolute Error (MAE) Mean Squared Error (MSE) R² score (coefficient of determination) Bland-Altman plots and correlation coefficients for agreement with measured values.
Interventions
OTHERdata collection
An additional collection of peritoneal dialysate and spot urine samples will be collected. Participants randomized to the training/validation arm will have their data used for model development, including the training and validation phases.
OTHERdata report
An additional collection of peritoneal dialysate and spot urine samples will be collected. Participants randomized to the test arm will have their data isolated and reserved exclusively for evaluating the performance of the final AI model
Sponsors
Tuen Mun Hospital
Study design
Observational model
COHORT
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* Age 18 years or older * Diagnosis of end-stage renal failure requiring peritoneal dialysis as renal replacement therapy * Ability to give informed consent and comply with study procedures.
Exclusion criteria
* History of hernia or peritoneal leak, including pleuroperitoneal fistula (PPF), patent processus vaginalis (PPV) and retroperitoneal leak * Ongoing PD peritonitis with or without antibiotic therapy * Just finished PD peritonitis antibiotic treatment within recent 4 weeks * Pregnancy * Patient refusal
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Peritoneal Equilibration Test (PET) Parameters | Measured at baseline during study enrollment | Predictive Accuracy of AI Model for Peritoneal Equilibration Test (PET) Parameters Outcome: AI-predicted vs. actual 2-hour and 4-hour dialysate-to-plasma creatinine ratio (D/P Cr) Performance Metrics: Mean Absolute Error (MAE) Unit of Measure: Absolute error |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Dialysis Adequacy (Kt/V) parameters | Measured at baseline during study enrollment | Predictive Accuracy of AI Model for Dialysis Adequacy (Kt/V) Outcome: AI-predicted vs. actual total weekly Kt/V Performance Metrics: Mean Absolute Error (MAE) Unit of Measure: Absolute error |
| Discriminative Ability of AI Model | Measured at baseline during study enrollment | Outcome: Classification of peritoneal transporter type (low, low-average, high-average, high) based on PET Performance Metrics: Area Under the Receiver Operating Characteristic Curve (AUC-ROC) Unit of Measure: AUC-ROC value (range: 0 to 1, higher values indicate better discriminative ability) |
| Calibration Performance of AI Model | Measured at baseline during study enrollment | Outcome: Model-predicted vs. actual transporter status and dialysis adequacy (Kt/V) Performance Metrics: Calibration Slope Unit of Measure: Calibration slope (ideal value = 1) |
Countries
Hong Kong
Outcome results
None listed