Table 2.

Predictive analysis based on additive and dominant polymorphism–treatment interaction models

Cetuximab and BSCBSCInteraction between biomarker and treatment
Dataset/BiomarkerAmino acidNMedian survival (95% CI)NMedian survival (95% CI)HRa (95% CI; Pb)Unadjusted PcAdjusted Pd
OS
All patients
FCGR2AH/H407.95 (5.2–10.6)414.11 (2.9–4.8)0.37 (0.2–0.6; P < 0.0001)
H/R646.60 (5.0–9.1)685.55 (4.6–7.4)0.87 (0.6–1.3; P = 0.49)0.010.005
R/R336.34 (4.5–9.1)405.82 (3.8–8.1)1.07 (0.6–1.8; P = 0.81)
R/-976.44 (5.2–8.3)1085.65 (4.7–7.1)0.93 (0.7–1.3; P = 0.66)0.0040.001
FCGR3AF/F637.72 (6.1–9.7)764.83 (4.2–5.5)0.68 (0.5–1.0; P = 0.04)
F/V606.34 (4.6–9.2)565.45 (3.9–6.5)0.83 (0.6–1.2; P = 0.36)0.460.66
V/V155.13 (2.6–10.6)183.98 (2.7–13.1)0.83 (0.4–1.8; P = 0.63)
V/-756.24 (5.2–8.3)744.99 (3.9–6.0)0.84 (0.6–1.2; P = 0.34)0.440.57
KRAS Wild type
FCGR2AH/H2510.35 (7.2–11.7)214.83 (2.0–5.5)0.36 (0.2–0.7; P = 0.003)
H/R299.49 (5.9–12.2)285.60 (4.2–8.7)0.68 (0.4–1.3; P = 0.22)0.080.04
R/R197.03 (4.5–9.8)265.34 (2.9–12.2)0.96 (0.5–1.9; P = 0.91)
R/-488.31 (6.3–9.9)545.49 (4.2–7.5)0.78 (0.5–1.2; P = 0.27)0.080.03
FCGR3AF/F369.66 (7.0–10.1)374.76 (3.6–5.9)0.58 (0.3–1.0; P = 0.04)
F/V317.95 (4.6–11.6)285.59 (4.0–7.7)0.73 (0.4–1.3; P = 0.28)0.720.62
V/V611.97 (3.6–23)115.03 (2.0–12)0.63 (0.2–2.2; P = 0.46)
V/-379.49 (5.0–11.6)395.45 (4.1–7.7)0.72 (0.4–1.2; P = 0.21)0.640.75
KRAS Mutated
FCGR2AH/H105.44 (1.0–13)133.61 (2.1–4.7)0.31 (0.1–0.8; P = 0.02)
H/R275.55 (3.4–7.6)345.55 (3.7–10.4)1.05 (0.6–1.9; P = 0.88)0.030.03
R/R105.19 (2.2–9.2)117.36 (3.0–17.5)1.47 (0.6-.4.0; P = 0.44)
R/-375.31 (3.4–6.7)455.82 (4.1–7.7)1.14 (0.7–1.9; P = 0.68)0.020.05
FCGR3AF/F225.37 (3.8–9.0)335.49 (4.6—8.3)0.93 (0.5—1.7; P = 0.83)
F/V205.42 (2.8–9.1)194.70 (1.7–7.4)0.83 (0.4–1.7; P = 0.60)0.750.24
V/V64.42 (1.0–6.4)63.25 (1.3–13.1)0.67 (0.2–2.4; P = 0.54)
V/-265.19 (3.4–6.7)253.88 (2.1–6.8)0.82 (0.5–1.5; P = 0.51)0.750.23
PFS
All
FCGR2AH/H403.63 (2.0–5.4)411.81 (1.7–2.0)0.34 (0.2–0.6; <0.0001)
H/R641.87 (1.8–3.5)681.87 (1.8–2.1)0.83 (0.6–1.2; P = 0.31)0.130.08
R/R331.84 (1.7–1.9)401.79 (1.6–1.9)0.70 (0.4–1.1; P = 0.12)
R/-971.86 (1.8–2.0)1081.84 (1.8–1.9)0.78 (0.6–1.0; P = 0.08)0.020.06
FCGR3AF/F633.48 (1.9–3.6)761.84 (1.7–2.0)0.53 (0.4–0.8; P = 0.0002)
F/V601.84 (1.8–2.2)561.84 (1.8–2.0)0.62 (0.4–0.9; P = 0.015)0.080.12
V/V151.91 (1.7–3.7)181.84 (1.7–3.6)1.02 (0.5–2.1; P = 0.96)
V/-751.86 (1.8–2.2)741.84 (1.8–1.9)0.75 (0.5–1.1; P = 0.09)0.290.48
WT KRAS
FCGR2AH/H255.49 (3.8–7.2)211.84 (1.7–2.0)0.19 (0.1–0.4; <0.0001)
H/R293.58 (1.9–5.5)282.04 (1.8–3.3)0.58 (0.3–1.0; P = 0.05)0.140.06
R/R191.91 (1.7–3.8)261.84 (1.7–1.9)0.51 (0.3–1.0; P = 0.03)
R/-483.55 (1.9–3.9)541.91 (1.8–2.0)0.53 (0.4–0.8; P = 0.003)0.070.04
FCGR3AF/F363.75 (3.1–5.5)371.87 (1.7–2.0)0.29 (0.2–0.5; <0.0001)
F/V312.96 (1.8–5.8)281.84 (1.8–2.8)0.43 (0.2–0.8; P = 0.004)0.170.18
V/V64.44 (1.9–6.8)111.91 (1.7–4.8)0.57 (0.2–1.7; P = 0.29)
V/-373.58 (1.9–5.7)391.87 (1.8–2.2)0.52 (0.3–0.9; P = 0.008)0.300.37
Mutated KRAS
FCGR2AH/H101.91 (1.0–3.6)131.68 (1.5–2.1)0.65 (0.3–1.5; P = 0.31)
H/R271.77 (1.6–1.8)341.81 (1.6—2.1)1.18 (0.7—2.0; P = 0.51)0.850.54
R/R101.76 (0.8–1.9)111.54 (0.9–1.8)0.70 (0.3–1.7; P = 0.43)
R/-371.77 (1.7–1.8)451.77 (1.6–1.8)1.07 (0.7-.7; P = 0.77)0.270.90
FCGR3AF/F221.84 (1.6–3.5)331.81 (1.6–2.1)0.96 (0.6–1.6; P = 0.87)
F/V201.77 (1.7–1.9)191.68 (1.2–1.8)0.66 (0.3–1.3; P = 0.18)0.730.95
V/V61.82 (1.0–3.7)61.68 (1.3–13.1)1.30 (0.4–4.7; P = 0.68)
V/-261.77 (1.7–1.9)251.68 (1.5–1.8)0.83 (0.4–1.5; P = 0.49)0.730.95
  • NOTE: In each section, the screening additive genetic model is presented in the first three rows. The dominant model is presented in indented format (R/-, at least one R allele; V/-, at least one V allele). As the wild-type (H/H or F/F) data remain unchanged in either model, they are only presented in the additive model.

  • Abbreviations: BSC, best supportive care; CI, confidence interval; WT, wild-type.

  • aHR of “cetuximab and best supportive care” treatment arm versus “best supportive care” arm.

  • bFrom log-rank test between “cetuximab and best supportive care” arm versus “best supportive care” arm.

  • cFrom Cox model with genotype (as ordinal categorical variable for additive screening model and as a dichotomized variable for the codominant model), treatment arm, and their interaction as covariates.

  • dFrom Cox model similar to c, but including other factors such as ECOG performance status (0–1 versus 2), gender (male vs. female), age (65 years or older vs. younger than 65 years), baseline lactate dehydrogenase level [higher than the upper limit of the normal (ULN) range vs. the ULN or lower), baseline alkaline phosphatase (higher than ULN vs. ULN or less), baseline hemoglobin [common toxicity criteria (CTC) grade 1 or higher versus CTC grade 0], number of disease sites (more than 2 vs. 2 or less), number of previous chemotherapy drug classes (more than 2 vs. 2 or less), primary tumor site (rectum only vs. colon), and presence of liver metastases (yes vs. no) as covariates. These variables were included as covariates of the multivariate models in the original clinical trial analysis.