Table 3.

Multivariate Cox regression analysis of PFS and OS by biomarker status adjusting for prognostic and clinically important baseline variablesa,b; analysis of multiplicative interaction between treatment effect and biomarker status

EGFR, HER2, and HER3 evaluated as continuous variables
T-DM1 treatment relative to CL treatment HR (95% CI)
PFScOSd
EGFR mRNA concentration ratio0.68 (0.57–0.82)0.64 (0.50–0.81)
HER2 mRNA concentration ratio0.69 (0.57–0.83)0.62 (0.49–0.78)
HER3 mRNA concentration ratio0.68 (0.57–0.82)0.61 (0.48–0.78)
EGFR, HER2, HER3, combined0.67 (0.56–0.81)0.61 (0.48–0.78)
PIK3CA mutation status evaluated as a categorical variable
T-DM1 treatment relative to CL treatment HR (95% CI)
PIK3CA mutation statusPFSeOSf
Mutated (n = 79)0.50 (0.27–0.94)0.21 (0.09–0.47)
Wild type (n = 180)0.70 (0.47–1.03)0.65 (0.38–1.10)
P value for interactiong between treatment and biomarker expression
PFSOS
EGFR mRNA concentration ratio0.830.25
HER2 mRNA concentration ratio0.070.13
HER3 mRNA concentration ratio0.520.88
PIK3CA mutation status0.220.05
  • aThe initial model also included the following variables: world region, prior chemotherapy regimen in locally advanced/metastatic setting, visceral disease, age, race, number of disease sites, prior anthracycline therapy, baseline Eastern Cooperative Oncology Group (ECOG) performance score, progesterone receptor and estrogen receptor status, baseline disease measurability, menopausal status, prior anticancer therapy, prior trastuzumab therapy, and HER2 status.

  • bA stepwise procedure was used to determine the final model.

  • cSignificant baseline risk factors were number of disease sites per independent review (<3 or ≥3 sites); baseline ECOG performance score (0 or 1); and disease measurability by independent review (yes or no).

  • dSignificant baseline risk factors were number of disease sites per independent review (<3 or ≥3 sites); baseline ECOG performance score (0 or 1); and prior trastuzumab therapy for metastatic breast cancer (yes or no).

  • eThe only significant baseline risk factor for the PIK3CA-mutated group was the number of disease sites per independent review (<3 or ≥3 sites). The only significant factor for the PIK3CA wild-type group was the presence of visceral disease (yes or no).

  • fThe only significant baseline risk for the PIK3CA-mutated group was prior anthracycline treatment (yes or no). Significant factors for the PIK3CA wild-type group were number of disease sites per independent review (<3 or ≥3 sites) and the presence of visceral disease (yes or no).

  • gThis analysis was conducted using a Cox proportional hazard model, and the P value was obtained using Wald statistics.