Date

2024-05-27

Protocol

Protocol changed

Item Changed

Target number/sample size

Previous Version

Global: 1164; Sri Lanka: 80 (40 in each arm)

Next Version

Due to a smaller final sample size the analysis has been changed as described in the endpoints/outcomes above. The final sample size is 280 participants. The original sample size calculation assumed that 1,164 participants would achieve a power of 90% to detect a risk reduction of 25% in the intervention versus the control arm (hazard ratio=0.75). This calculation was based on a survival analysis of time-to-first event using a composite of mortality, MACE and cardiac or respiratory hospitalisation and required a total of 514 participants experiencing a primary endpoint event. Due to the impact of the pandemic which prevented recruitment over an 18-month period, and during which patients with COPD were much less likely to exacerbate (and hence be eligible for PACE), we have closed recruitment at 280 participants. To capture additional information and increase power compared to a “time to first event” analysis, the primary outcome will be analysed using a Win Ratio approach37,38. Assuming a Win ratio of 1.5 (54% of wins with the beta-blocker arm) and 10% of ties, this approach will lead to 76% power. The primary statistical analysis will be by ITT i.e. according to the randomised arm and regardless of adherence to the protocol. The primary analysis will be a hierarchical win ratio analysis of the primary composite endpoint: (1) all-cause mortality, (2) cardiac and respiratory hospital admissions, (3) MACE, (4) moderate COPD exacerbations, (5) decrease in FEV1, (6) mild COPD exacerbations, (7) QOL and symptoms burden. Every participant randomised to the ?-blocker arm will be paired with every participant randomised to the placebo arm. This method compares all possible pairs, not 1 to 1 matching. If, for example, we have 100 randomised to bisoprolol and 100 randomised to placebo, we would end up with 10,000 pairs (100 x 100). If there is a slightly uneven split in the randomisation, we still create all possible pairs e.g. 70 randomised to one arm and 68 to the other, all 70 from arm 1 will be matched to all 68 from arm 2, thus creating a total of 4,760 pairs. Within each of these pairs, we then proceed to compare outcomes according to the pre-specified hierarchy until a winner is called (or a tie)". Within each pair, we will compare all-cause mortality to determine a “winner”. The winner will be the participant with the longest survival time. If neither participant died and a winner cannot be declared on mortality alone, we will then proceed to the next step and compare MACE outcomes. The approach will proceed in a stepwise fashion, moving to the next outcome in the hierarchy, until a decision has been reached for every pair. The win ratio will then be calculated as the proportion of “winners” (the participant in the beta-blocker arm had a better outcome) divided by the proportion of “losers” (the participant in the placebo arm had a better outcome). To complete the interpretation, we will also compute the win odds and net benefit.