Background: It is unclear whether beta‐blocker therapy should be reduced or withdrawn in patients who develop acute decompensated heart failure (HF). We studied the relationship between changes in beta‐blocker dose and outcome in patients surviving a HF hospitalisation in COMET. Methods: Patients hospitalised for HF were subdivided on the basis of the beta‐blocker dose administered at the visit following hospitalisation, compared to that administered before. Results: In COMET, 752/3029 patients (25%, 361 carvedilol and 391 metoprolol) had a non‐fatal HF hospitalisation while on study treatment. Of these, 61 patients (8%) had beta‐blocker treatment withdrawn, 162 (22%) had a dose reduction and 529 (70%) were maintained on the same dose. One‐and two‐year cumulative mortality rates were 28.7% and 44.6% for patients withdrawn from study medication, 37.4% and 51.4% for those with a reduced dosage (n.s.) and 19.1% and 32.5% for those maintained on the same dose (HR,1.59; 95%CI, 1.28–1.98; p <0.001, compared to the others). The result remained significant in a multivariable model: (HR, 1.30; 95%CI, 1.02–1.66; p =0.0318). No interaction with the beneficial effects of carvedilol, compared to metoprolol, on outcome was observed ( p =0.8436). Conclusions: HF hospitalisations are associated with a high subsequent mortality. The risk of death is higher in patients who discontinue beta‐blocker therapy or have their dose reduced. The increase in mortality is only partially explained by the worse prognostic profile of these patients.