Truace
Subscribe

A machine learning framework for precision dosing of daptomycin across diverse disease states

The large pharmacokinetic variability of daptomycin across disease states poses a challenge for individualized dosing. Conventional population pharmacokinetic (PopPK) models are often limited to specific conditions, lacking generalizability. This study used a machine learning (ML) framework to predict daptomycin exposure and guide dose optimization across heterogeneous populations. Nine published PopPK models were used to generate 90,000 simulated profiles across patients with infections and/or patients treated…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-1925· cite & verify
A machine learning framework for precision dosing of daptomycin across diverse disease states
United States Naval Medical Bulletin Vol. 36, Nos. 1-4, 1938 by U.S. Navy. Bureau of Medicine and Surgery. Public domain

In brief

Researchers used nine published population pharmacokinetic models to simulate 90,000 daptomycin profiles in patients with infections, kidney transplants, or renal replacement therapy. They trained seven machine learning algorithms to predict 24-hour drug exposure (AUC 0-24), and CatBoost performed best.

The approach matters because daptomycin exposure varies widely between disease states and conventional models often do not generalize across them. The validation covered only 39 patients, and the improvement in the renal replacement group was smaller than in infected patients, so clinical benefit in practice remains to be shown.

Main points

  1. CatBoost had the highest predictive accuracy among seven algorithms tested.
  2. Dose, creatinine clearance, disease state, and weight were the key predictors identified.
  3. In infected patients, RMSE was 141 versus 353 mg·h/L for PopPK models, and mean absolute prediction error was 23.5% versus 58.1%.
  4. In renal replacement therapy patients, accuracy rose from 31.6% to 52.6%, with RMSE of 197 versus 218 mg·h/L.

The gain

A machine learning model trained on simulated profiles predicted daptomycin AUC more accurately than existing population pharmacokinetic models in a small real-world validation cohort, supporting pre-treatment precision dosing.

The rundown

The training data were entirely simulated from nine published PopPK models, spanning patients with infections and/or renal transplantation and/or renal replacement therapy. Model performance was assessed with statistical metrics, clinical scenario assessments, and comparison with PopPK models.

External validation used real-world data from 20 infected patients, contributing 501 concentration measurements, and 19 patients on renal replacement therapy, contributing 513 measurements.

In infected patients the model placed 75.0% into the correct clinical AUC category, against an average overall accuracy of 30.0% for PopPK models. In renal replacement therapy patients the gains were more modest.

Sources

  1. Peer-reviewedAntimicrobial Agents and Chemotherapy2026-10-14

The debate