Recalculation of the numbers at risk - example classified as "not consistent". Kaplan-Meier plot of the trial presented in figure 1, but using fictive data. In this example it is assumed that no patient was reported as lost to follow-up in either group and minimum follow-up was 120 days. As in figure 1, we read the survival probability from the curve at 90 days. We multiplied the survival probability by the number of randomized patients in order to recalculate the number of patients at risk (number at risk: 19 in the prednisolone group vs. 13 in the control group). As the reported number at risk was smaller in the prednisolone group, four patients must have been censored before day 90. As no losses to follow up were reported, this fictive example would be classified as "not consistent".
Vervölgyi et al. BMC Medical Research Methodology 2011 11:130 doi:10.1186/1471-2288-11-130