It seems likely, therefore, that serum Ang-2 levels reflect the level of Ang-2 expression in the tumour excellent validation stromal compartment. Elevated serum concentrations of Ang-2 have also been reported for patients with cancers other than CRC, such as non-small-cell lung cancer and melanoma, in which high serum Ang-2 levels correlate with disease stage and poor OS (Park et al, 2007; Helfrich et al, 2009). However, the relationship between serum Ang-2 levels and clinical outcome in patients treated with VEGF-targeting drugs and chemotherapeutic agents has not been explored before, and this study is the first to investigate the impact of pre-therapeutic serum Ang-2 concentrations on the clinical outcome in patients with metastatic CRC under bevacizumab-containing therapy.

Compared with high serum Ang-2 levels, low serum Ang-2 was associated with an outstanding response rate (>80%), better disease control and excellent OS (>90% after 18 months). In accordance with previous reports (Jubb et al, 2006a), VEGF and tumour MVD were not similarly correlated to these end points. Similarly, the pericyte content of CRC was not linked to treatment outcome and did not correlate with Ang-2 serum concentrations, indicating that serum Ang-2 is probably not simply a surrogate of blood vessel morphology. On the basis of experimental models, Ang-2 has been described as an opponent of vascular normalisation that prevents blood vessels from becoming structurally and functionally stabilised (Maisonpierre et al, 1997; Scharpfenecker et al, 2005; Falcon et al, 2009; Reiss et al, 2009).

Conceivably, normalisation of tumour vessels by bevacizumab-mediated blockade of VEGF may be more difficult to achieve and chemotherapeutic drugs cannot be delivered appropriately to the tumour cells when Ang-2 serum levels are high. From a mechanistic point of view, the observation that patients with low serum Ang-2 were most likely to benefit from treatment with Entinostat respect to major clinical end points supports such a biological role of Ang-2. From a clinical perspective, our observations suggest that serum Ang-2 could hold promise as a predictive biomarker allowing bevacizumab-containing treatment to be customised in CRC patients. Having analysed a heterogeneous patient cohort of moderate size, our study is not without limitations. Nevertheless, subgroup analyses of known prognostic factors in CRC (for example, age, ECOG) showed no evidence that outcome by Ang-2 was biased by those factors.