Pretreatment lactate dehydrogenase, but not magnesium or neutrophil-to-lymphocyte ratio, is independently associated with grade ≥3 chemotherapy toxicity in breast cancer: prediction-model development and internal validation in 400 patients.

BACKGROUND

Severe chemotherapy toxicity is clinically important; existing tools were developed mainly for older adults. We evaluated three pretreatment biomarkers as an exploratory model in a younger breast-cancer cohort.

PATIENTS AND

METHODS

This single-centre cohort included 400 patients (300 retrospectively collected records and 100 prospectively collected samples); 87 (21.8%) had CTCAE grade ≥3 toxicity. Logistic regression used LDH ≥300 U/L, magnesium <1.7 mg/dL, and continuous NLR. Optimism was assessed with bootstrapping, with Firth, ridge, spline, and adjusted models. An aggregate regimen summary, not linked to individual patients, supported descriptive regimen-stratified analyses.

RESULTS

LDH ≥300 U/L was independently associated with toxicity (OR 3.64, 95% CI 2.14-6.19); magnesium <1.7 mg/dL (OR 1.55, 0.92-2.59; P = 0.098) and baseline NLR (OR 1.00, 0.76-1.33) were not. Apparent AUC was 0.678, bootstrap-corrected AUC 0.665, and corrected calibration slope 0.954. HER2 positivity was associated with toxicity in the adjusted model (OR 1.79, 1.05-3.05; P = 0.034). Crude regimen-stratified toxicity ranged from 18.2% to 30.0% (P = 0.230), with no regimen differing significantly from AC-T. Among 88 participants with complete follow-up, NLR rose from 2.4 to 5.5 during therapy and returned to 2.6 by 30 days (P < 0.001).

CONCLUSIONS

LDH ≥300 U/L was the only independent baseline predictor. HER2 positivity showed an adjusted association possibly reflecting treatment context. Baseline NLR was not independently predictive, although it changed markedly during therapy. Regimen-stratified data added context but not patient-level adjustment. Overall discrimination remained modest, and external validation is required before clinical use.