Antibiotic co-resistance phenotypes as predictors of methicillin resistance in Staphylococcus aureus: a multi-study analysis of transportability, lineage dependence and model explanations

Methicillin-resistant Staphylococcus aureus (MRSA) frequently carries resistance to non-beta-lactam antibiotics, but it is unclear how reproducibly non-beta-lactam susceptibility profiles indicate methicillin status across settings. We retrieved all laboratory-derived S. aureus susceptibility records in the BV-BRC database (45,876 records) and analysed 2,245 isolates (660 MRSA) from six studies that sampled both MRSA and methicillin-susceptible S. aureus. Using six non-beta-lactam phenotypes, we estimated study- and lineage-stratified associations and a co-resistance gradient. We compared a ciprofloxacin-only rule with logistic regression and XGBoost under leave-one-study-out validation, and compared SHAP attributions with label-based importance measures. Ciprofloxacin, erythromycin and clindamycin resistance were positively associated with MRSA in every study in which an association could be estimated (study-adjusted odds ratios 13.4, 10.4 and 8.8). However, effect sizes were highly heterogeneous (I² 93–98%) and fell by 40–62% after stratification by sequence type. Within three of the five evaluable sequence types, ciprofloxacin was uninformative. Pooled cross-validation was misleading: XGBoost reached an AUC of 0.94, and a model given only which antibiotics had been tested reached 0.69. In held-out studies, AUC ranged from 0.50 to 1.00. Wherever ciprofloxacin had been tested, the multivariable models did not consistently outperform ciprofloxacin alone, and their predicted probabilities were poorly calibrated. SHAP ranked ciprofloxacin first, in agreement with model-free measures, but lower ranks changed with the training data and partly reflected testing practice. Co-resistance phenotypes are markers of locally prevalent MRSA lineages rather than transportable predictors of methicillin resistance, and they cannot substitute for cefoxitin testing.

Authors

Publication Details

Published
2026-10-26
DOI
https://doi.org/10.1099/acmi.0.001197.v2
Primary Topic
Antimicrobial Resistance in Staphylococcus
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
preprint

Antibiotic co-resistance phenotypes as predictors of methicillin resistance in Staphylococcus aureus: a multi-study analysis of transportability, lineage dependence and model explanations

Ryan Yi Sheng Neo
Antimicrobial Resistance in Staphylococcus
preprint

Antibiotic co-resistance phenotypes as predictors of methicillin resistance in Staphylococcus aureus: a multi-study analysis of transportability, lineage dependence and model explanations

Ryan Yi Sheng Neo
preprint en

Abstract

Methicillin-resistant Staphylococcus aureus (MRSA) frequently carries resistance to non-beta-lactam antibiotics, but it is unclear how reproducibly non-beta-lactam susceptibility profiles indicate methicillin status across settings. We retrieved all laboratory-derived S. aureus susceptibility records in the BV-BRC database (45,876 records) and analysed 2,245 isolates (660 MRSA) from six studies that sampled both MRSA and methicillin-susceptible S. aureus. Using six non-beta-lactam phenotypes, we estimated study- and lineage-stratified associations and a co-resistance gradient. We compared a ciprofloxacin-only rule with logistic regression and XGBoost under leave-one-study-out validation, and compared SHAP attributions with label-based importance measures. Ciprofloxacin, erythromycin and clindamycin resistance were positively associated with MRSA in every study in which an association could be estimated (study-adjusted odds ratios 13.4, 10.4 and 8.8). However, effect sizes were highly heterogeneous (I² 93–98%) and fell by 40–62% after stratification by sequence type. Within three of the five evaluable sequence types, ciprofloxacin was uninformative. Pooled cross-validation was misleading: XGBoost reached an AUC of 0.94, and a model given only which antibiotics had been tested reached 0.69. In held-out studies, AUC ranged from 0.50 to 1.00. Wherever ciprofloxacin had been tested, the multivariable models did not consistently outperform ciprofloxacin alone, and their predicted probabilities were poorly calibrated. SHAP ranked ciprofloxacin first, in agreement with model-free measures, but lower ranks changed with the training data and partly reflected testing practice. Co-resistance phenotypes are markers of locally prevalent MRSA lineages rather than transportable predictors of methicillin resistance, and they cannot substitute for cefoxitin testing.

Antimicrobial Resistance in Staphylococcus
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon