Rejoinder to commentary by Dr Freedman of ‘assessing surrogates as trial endpoints using mixed models’