We’ve heard the message before: until we do the trial, we don’t know the result, and negative trials, unwelcome as they are, are just as important as positive ones. In The Lancet Oncology, Rachel Kerr and colleagues present the results of the QUASAR 2 trial, 1 which now joins the NSABP C-082 and AVANT3 trials as an important negative study of bevacizumab. Kerr and coworkers assessed capecitabine alone versus capecitabine plus bevacizumab as adjuvant treatment after potentially curative surgery for histologically proven stage III or high-risk stage II colorectal cancer. 3-year disease-free survival did not differ between the groups (capecitabine and bevacizumab 75· 4% vs capacitabine alone 78· 4%, hazard ratio 1· 06, 95% CI 0· 89–1· 25, p= 0· 54). Post-hoc biomarker analyses suggested potential benefits in some subgroups of patients, but these were purely exploratory. Serious adverse events were reported in 221 patients who received capecitabine and in 350 who received capecitabine and bevacizumab. Collectively, the QUASAR 2, NSABP C-08, and AVANT trials have assessed bevacizumab for the treatment of colorectal cancer in more than 8000 patients, and together tell us several things about this drug, most of which we wish were not so. The first inescapable message, confirmed in triplicate, is that the addition of bevacizumb to adjuvant chemotherapy does not improve the outcome for patients with stage II or III colorectal cancer. The second is that we can do actual harm in terms of cancer-specific outcomes and fatal toxic effects. The third is that there might be a subgroup of patients who would benefit from treatment, but that is not entirely clear, and we