Development and validation of a predictive model of aromatase inhibitor-induced arthralgia among patients with early breast cancer
Pietro Lapidari, Maryam Lustberg, Julie Havas, Martina Pagliuca, Chayma Bousrih, Christelle Jouannaud, Marion Fournier, William Jacot, Laurence Vanlemmens, Kaderbhai Courèche, Anne Kieffer, Baptiste Sauterey, Olivier Tredan, Christelle Levy, Anne-Laure Martin, Catherine Gaudin, Gwenn Menvielle, Maria Alice Franzoi, Ines Vaz-Luis, Antonio Di MeglioAbstract
Background
Aromatase inhibitor-induced arthralgia (AIA) is common in early-stage breast cancer patients on aromatase inhibitors (AI), potentially compromising adherence and outcomes. We developed and validated prediction models for short- and longer-term AIA.
Methods
Postmenopausal patients with stage I–III breast cancer on adjuvant AI in CANTO (NCT01993498) were included. Primary outcome was AIA (any grade articular/muscular pain, Common Terminology Criteria for Adverse Events [CTCAE] v4.0) at year 1 (Y-1) and 4 (Y-4) cohort visits. Baseline clinical, behavioral, treatment-related, patient-reported (EORTC QLQ-C30, HADS) variables were modeled using multivariable logistic regression and bootstrap in development and temporal validation cohorts. Baseline inflammatory markers were examined in a sub-cohort.
Results
The Y-1 and Y-4 development cohorts included 3,065 and 2,390 patients (mean age 64.0 [SD 7.2] and 63.7 years [6.9]), respectively. AIA occurred in 61.3% (Y-1) and 66.1% (Y-4) patients. Validation cohorts included 1,313 (Y-1) and 1,009 (Y-4) patients with similar characteristics. Across models, higher BMI, greater baseline fatigue, prior articular/muscular disease, chemotherapy exposure, pre-existing pain were associated with subsequent AIA; anastrozole use was associated with AIA at Y-4. AUC was 0.62–0.65 in development cohorts; 0.61–0.67 in validation cohorts. In the biomarker sub-cohort (n = 637), higher baseline IL-8 was associated with lower Y-1 AIA odds (OR 0.74, 95% CI 0.56–0.98).
Conclusion
Using baseline clinical and patient-reported data, we generated models identifying patients at increased risk of early and persistent AIA. While performance was modest, risk stratification could help trigger stepped-care pathways to optimize AI adherence and outcomes. IL-8 association requires independent replication.