Adjusting for multiple prognostic factors in the analysis of randomised trials