How should randomised trials including multiple pregnancies be analysed?