Table 3. Correlations between baseline biomarker levels and clinical outcomes by ELISA test.
| Biomarker | Progression-Free Survivala |
Overall Survivala |
Overall Response Ratea |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Median (mos) | HR (95% CI) | P | Pb | Median (mos) | HR (95% CI) | P | Pb | Median (%) | OR (95% CI) | P | Pb | |
| All patients | 6.9 v 10.5 | 0.60 (0.44–0.81) | 0.001 | — | 24.5 v 26.6 | 0.80 (0.56–1.15) | 0.230 | — | 31.0 v 41.4 | 1.58 (0.84–2.96) | 0.20 | — |
| ANGPTL4 | ||||||||||||
| Low | 6.5 v 12.2 | 0.58 (0.36–0.94) | 0.023 | 0.054 (0.041) | 37.9 v 27.5 | 0.96 (0.51–1.80) | 0.900 | 0.596 (0.980) | 10.7 v 48.8 | 8.00 (2.09–30.3) | 0.002 | 0.003 |
| High | 7.3 v 9.5 | 0.73 (0.49–1.08) | 0.115 | 25.7 v 28.5 | 0.84 (0.50–1.41) | 0.520 | 41.1 v 34.1 | 0.74 (0.33–1.69) | 0.476 | |||
| HGF | ||||||||||||
| Low | 7.0 v 11.7 | 0.52 (0.35–0.76) | 0.001 | 0.020 (0.039) | 26.0 v 55.3 | 0.42 (0.20–0.89) | 0.020 | 0.010 (0.026) | 28.8 v 49.0 | 2.37 (1.04–5.38) | 0.040 | 0.134 |
| High | 6.3 v 8.5 | 0.79 (0.49–1.27) | 0.323 | 34.5 v 21.1 | 1.19 (0.74–1.90) | 0.473 | 34.4 v 31.6 | 0.88 (0.32–2.40) | 0.804 | |||
| VEGF-A121 | ||||||||||||
| Low | 6.4 v 10.8 | 0.53 (0.36–0.76) | 0.001 | 0.023 (0.028) | 28.0 v 50.0 | 0.42 (0.21–0.84) | 0.011 | 0.034 (0.032) | 28.1 v 39.3 | 1.66 (0.77–3.60) | 0.136 | 0.141 |
| High | 7.8 v 7.8 | 0.90 (0.53–1.53) | 0.70 | 29.9 v 19.6 | 1.25 (0.77–2.02) | 0.370 | 37.0 v 46.2 | 1.46 (0.49–4.37) | 0.583 | |||
aPatients were stratified according to baseline marker levels (using corresponding optimal cutoff), clinical outcomes of control group versus bevacizumab group were compared using Cox proportional hazards model (for PFS and OS), and logistic regression models (for ORR). High indicates above the corresponding cutoff and low indicates less than or equal to the corresponding cutoff. The cutoff values for ANGPTL4, HGF and VEGF121 were 1.97 ng/ml, 0.88 ng/ml and 0.59 ng/ml, respectively.
bP-value for treatment-marker interaction assessed by interaction Wald Cox proportional hazards model (for PFS and OS), and interaction logistic regression models (for ORR). In parentheses, the interaction P-value was adjusted for known clinical prognostic variables (gender, age, performance status, primary tumor site, tumor grade, prior adjuvant chemotherapy, number of metastasis site, and curative-intent metastasis resection). Abbreviations: ELISA, enzyme-linked immunosorbent assay. Statistical significance was set at 0.05 based on a two-sided test. Other footnotes as in Table 1.