Objectives: To assess the potential impacts of successive lockdown-easing measures in England, at a point in the COVID-19 pandemic when community transmission levels were relatively high. Design: We developed a Bayesian model to infer incident cases and reproduction number (R) in England, from incident death data. We then used this to forecast excess cases and deaths in multiple plausible scenarios in which R increases at one or more time points. Setting: England. Participants: Publicly available national incident death data for COVID-19 were examined. Primary outcome: Excess cumulative cases and deaths forecast at 90 days, in simulated scenarios of plausible increases in R after successive easing of lockdown in England, compared with a baseline scenario where R remained constant. Results: Our model inferred an R of 0.75 on 13 May when England first started easing lockdown. In the most conservative scenario modelled where R increased to 0.80 as lockdown was eased further on 1 June and then remained constant, the model predicted an excess 257 (95% CI 108 to 492) deaths and 26 447 (95% CI 11 105 to 50 549) cumulative cases over 90 days. In the scenario with maximal increases in R (but staying ≤1), the model predicts 3174 (95% CI 1334 to 6060) excess cumulative deaths and 421 310 (95% CI 177 012 to 804 811) cases. Observed data from the forecasting period aligned most closely to the scenario in which R increased to 0.85 on 1 June, and 0.9 on 4 July. Conclusions: When levels of transmission are high, even small changes in R with easing of lockdown can have significant impacts on expected cases and deaths, even if R remains ≤1. This will have a major impact on population health, tracing systems and healthcare services in England. Following an elimination strategy rather than one of maintenance of R ≤1 would substantially mitigate the impact of the COVID-19 epidemic within England.