Table 4 Sensitivity analysis using backwards selection.

From: Covariate-adjusted analysis of the Phase 3 REFLECT study of lenvatinib versus sorafenib in the treatment of unresectable hepatocellular carcinoma

 

Model 1a

Model 2b

Model 3c

HR

p-value

95% CI

HR

p-value

95% CI

HR

p-value

95% CI

Lenvatinib

0.814

0.008

0.699

0.948

0.765

0.001

0.656

0.892

0.760

<0.001

0.652

0.887

Alpha Fetoprotein >200 ng/mL

1.725

<0.001

1.478

2.014

1.685

<0.001

1.443

1.968

1.696

<0.001

1.453

1.980

Child-Pugh Score

1.676

<0.001

1.411

1.992

1.579

<0.001

1.328

1.878

1.590

<0.001

1.338

1.890

EHS

1.248

0.039

1.011

1.540

1.258

0.031

1.021

1.549

1.248

0.038

1.013

1.538

Liver Disease Involvement

2.345

<0.001

1.676

3.282

2.369

<0.001

1.692

3.317

2.371

<0.001

1.693

3.319

MPVI

1.351

0.001

1.129

1.617

1.309

0.003

1.094

1.567

1.299

0.004

1.085

1.556

Hepatitis B

1.231

0.008

1.057

1.433

1.219

0.011

1.047

1.419

1.222

0.010

1.049

1.422

Bone Disease Involvement

1.510

0.001

1.181

1.932

1.541

0.001

1.205

1.970

1.561

<0.001

1.220

1.997

Lung Disease Involvement

1.319

0.005

1.085

1.603

1.358

0.002

1.119

1.649

1.358

0.002

1.118

1.649

Other Disease Involvement

1.289

0.014

1.052

1.579

1.204

0.071

0.984

1.474

1.229

0.045

1.005

1.504

Post-treatment Therapy

0.599

<0.001

0.514

0.699

Post-treatment Medication

0.681

<0.001

0.579

0.802

Post-treatment Procedure

0.690

<0.001

0.569

0.835

AIC

8443

   

8400

   

8402

   
  1. MVPI Macroscopic Portal Vein Invasion, EHS Extra-Hepatic Spread, AIC  Aikake’s Information Criterion.
  2. aBackwards selection model starting from full multivariable model. Pr(0.05) required for covariate deletion.
  3. bBaseline covariates as Model 1 + post-treatment covariates separately.
  4. cBaseline covariates as Model 1 + post-treatment covariates as a composite variable.