Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1854
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-18T18:14:46.035045Z
status: published
lens: trace
sector: health
headline: Vancomycin treatment outcomes and machine learning-based identification of heterogeneous vancomycin-intermediate Staphylococcus aureus infections
dek: Objectives Heterogeneous vancomycin-intermediate Staphylococcus aureus (hVISA) infections remain underdiagnosed with unclear clinical impact. This multicenter study evaluated vancomycin treatment outcomes and developed a machine learning (ML)-based diagnostic model. Methods We prospectively enrolled 200 patients with MRSA infections from 20 Chinese hospitals between February 2012 and July 2021. Population analysis profile-area under the curve (PAP-AUC) testing was performed on frozen isolates to confirm hVISA ph…
gain_title: A LASSO-logistic regression model built on clinical predictors identified hVISA infection in MRSA patients with a test-set AUC of 0.834, which could support earlier detection and tailored vancomycin dosing.
problem_title: In adults on vancomycin monotherapy, hVISA infection was linked to much higher treatment failure than VSSA infection, even though hVISA patients received higher doses and still reached lower PK/PD indices.
trace_subject: hVISA infections in MRSA patients treated with vancomycin
gain_reading: A LASSO-logistic regression model built on clinical predictors identified hVISA infection in MRSA patients with a test-set AUC of 0.834, which could support earlier detection and tailored vancomycin dosing.
gain_evidence: LASSO-logistic regression performed best with AUC of 0.834 | The ML model enables identification of high-risk patients
problem_reading: In adults on vancomycin monotherapy, hVISA infection was linked to much higher treatment failure than VSSA infection, even though hVISA patients received higher doses and still reached lower PK/PD indices.
problem_evidence: hVISA infection was associated with significantly higher treatment failure than VSSA infection (34.6% vs. 5.0%, P = 0.028) | Despite higher vancomycin doses, adult patients with hVISA infection achieved lower PK/PD indices
quick_read: Researchers enrolled 200 patients with MRSA infections across 20 Chinese hospitals and found that most, 117, carried heterogeneous vancomycin-intermediate Staphylococcus aureus. They tested isolates in the lab, sequenced confirmed strains, tracked vancomycin levels, and trained machine learning models to predict which patients had hVISA.

hVISA is underdiagnosed, and the results suggest it undermines standard vancomycin therapy, so a practical prediction tool could help clinicians act sooner. The model was evaluated on a held-out test set, but the evidence comes from one study and does not show that model-guided dosing improves patient outcomes.
limitation: The model's test-set confidence interval is fairly wide (0.711-0.930), and the study reports prediction performance rather than evidence that model-guided dosing improves outcomes.
tag: Dual reading
key_points: 117 of 200 MRSA patients (58.5%) had hVISA, including 56.0% of adults and 76.0% of pediatric patients. | Adults with hVISA on vancomycin monotherapy had a 34.6% treatment failure rate versus 5.0% for VSSA. | ST5, ST239 and ST764 were the predominant hVISA clones, and ST5 and ST239 had higher failure rates than ST764. | The best of 10 ML models used vancomycin MIC, tracheal intubation and prior penicillin use as key predictors.
rundown: The prospective cohort was enrolled between February 2012 and July 2021. Frozen isolates underwent PAP-AUC testing to confirm the hVISA phenotype, and confirmed strains were whole-genome sequenced. Vancomycin therapeutic drug monitoring and PK/PD analysis were also performed.

Despite receiving higher vancomycin doses, adults with hVISA reached lower PK/PD indices than those with vancomycin-susceptible S. aureus. The authors link this to the observed failure gap in monotherapy and call for early identification.

Ten machine learning models were compared, and LASSO-logistic regression had the best test-set AUC of 0.834 (95% CI 0.711-0.930). The authors say it supports phenotype- and sequence type-guided personalized dosing.
sources:
- peer_reviewed | International Journal of Antimicrobial Agents | https://doi.org/10.1016/j.ijantimicag.2026.108021 | 2026-10-05
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

1beeb03b97a21a76bd3cc1cd7bfe7f97b27cc52701435978792f5107cbe5ebbe

previous

0000000000000000000000000000000000000000000000000000000000000000