Implications for estimating treatment effects when trials stop early, either for apparent benefit or futility