Relative risk of death from causes other than fungal infection in those who survived the fungal infection. Provided the two groups are still comparable after subtraction of fungal deaths from those randomised, the proportion of those who died from another cause than fungal infection would be expected to be the same in the two groups (fig. 1a). If, however, there was a bias in classification of cause of death in favour of the treated group, fewer deaths would be ascribed to fungal infection and more deaths to other causes in this group. This would result in a larger relative risk of death for those who did not die from fungal infection (fig. 1b). This risk is an estimate of bias overall: RR = (dexp/(Nexp-Fexp))/(dctr/(Nctr-Fctr)), where exp: experimental group; ctr: control group; Nexp, Nctr: numbers of randomised patients; Fexp, Fctr: numbers of deaths from fungal infection; dexp, dctr: numbers of deaths from causes other than fungal infection.