Background: Patients with HR+ MBC are treated with first-line endocrine therapy plus CDK4/6 inhibitors, but resistance often emerges through ESR1 or PIK3CA mutations (MUT). Co-mutations (co-MUT), present in ∼10-15% of patients, remain poorly understood. Here, we compare molecular features and Real-world overall survival (rwOS) among patients with co-MUT vs. those with single or no mutations (no-MUT).

Methods: We used Tempus Lens (Tempus AI, Inc., Chicago, IL) to identify patients with a primary diagnosis of MBC who had Tempus xF or xT testing from the Tempus multi-modal database (N = 8,626). Patients with tumor purity of at least 30% were assessed and all pathogenic or likely pathogenic (P/LP) somatic SNVs/indels for PIK3CA, ESR1, and TP53 were extracted. Polyclonal ESR1 and PIK3CA were defined as more than one distinct P/LP somatic SNV/indel. Tumor mutational burden (TMB) (mt/mB) was analyzed by DNA-sequencing. rwOS analyses were run for all samples collected in the year preceding any line of therapy using risk-set adjusted Cox models with the start of therapy as the index date, the later date of sequencing and treatment start as the study entry time, and a maximum follow up of 5 years.

Results: Of the total cohort, 53% of patients had no-MUTs, 32% were PIK3CA-MUT, 9% were ESR1-MUT, and 6% were co-MUT. As shown in the table, age and hormone receptor status were significantly different between the groups, but not race. The co-MUT group exhibited a higher prevalence of bone metastases, while ESR1-only patients had increased liver and lung involvement. TP53 P/LP MUT were present in 33% (N=903) of the PIK3CA-only patients and 24% (N=131) of the co-MUT patients. High TMB rate was significantly different between all groups and was highest in the co-MUT and PIK3CA patients (15% vs 14%). Polyclonal changes in both genes were also more frequent in the co-MUT group. The most common ESR1 mutation was p.D538G (6.4%), while p.His1047Arg was the predominant PIK3CA mutation (14%, n=1,187). Liquid biopsy identified 40% of co-MUT, while tissue sequencing detected 61%. Co-MUT patients had significantly worse rwOS compared to no-MUT (HR=1.75, p<0.001), though evidence of non-proportionality was observed.

Conclusions: Herein, we characterized the largest real-world cohort of co-MUT HR+ MBC patients, identifying distinct clinical and molecular features such as enriched bone metastasis, higher TMB, and increased polyclonality. The lower overall frequency of ESR1 MUT and co-MUT likely reflects that only 25% of patients had late molecular testing, suggesting these mutations often arise after prolonged endocrine therapy. In this subgroup, the ESR1 MUT rate was 24% and the co-MUT rate was 15%, as expected. Limitations include the retrospective design; future studies will address these and examine PTEN/AKT alterations.