OBJECTIVES: Increasingly, network meta-analysis (NMA) of published survival data are based on parametric survival curves as opposed to reported hazard ratios to avoid relying on the proportional hazards assumption, which may not be valid. One advantage of a Bayesian approach to NMA is that the probability of being the best treatment out of all those compared can be calculated. This directly supports decision-making. However, in the context of survival analysis multiple options are available. METHODS: Based on a case study in oncology, the probability that each treatment is best in terms of overall survival was calculated and presented based on the following underlying results: 1) the hazard over time, 2) the cumulative hazard over time, 3) the survival proportions over time, 4) the expected survival over time, 5) the expected survival at maximum follow-up, 6) expected survival when all patients have died, and 7) median survival. RESULTS: Since the NMA of survival curves results in changing hazard and survival estimates over time for the compared interventions, calculations of the probability that a certain treatment is best varies with the different alternatives. With methods 1-4 the probability that a certain treatment is best will vary as a function of follow-up, which provides relevant information. With methods 5-7 only one probability of being the best is obtained for each treatment, which is easier to understand. Method 1 does not directly relate to the survival proportion, which makes it not very intuitive.  Method 7 discards a lot of information. CONCLUSIONS: Different approaches to present the probability of being the most efficacious treatment for findings obtained with a NMA of survival curves have pros and cons. The probability that a certain treatment is best as a function of survival proportions over time, as well as expected survival over time seem the most useful and intuitive.