OBJECTIVES: The cost-effectiveness efficiency frontier is typically generated by plotting the costs and benefits for a variety of technologies, often summarised as the mean across a large number of simulations. However, the mean across simulations may mask correlations between technologies within simulations. We sought to investigate these issues using a HTA comparing 19 different interventions. METHODS: We compared 19 different breast cancer surveillance strategies for women aged less than 50 years with a BRCA1 mutation. We calculated the cost-effectiveness efficiency frontier for 5,000 simulations and also based on the mean costs and benefits for each strategy. We also investigated the probability of a strategy appearing on the frontier and correlations between whether or not different strategies appeared on the frontier. RESULTS: The efficiency frontier based on means included 6 strategies. This was the cost-effectiveness efficiency frontier in only 22 (0.44%) of 5,000 simulations. Forty four other frontiers were more likely to be generated by simulations. Two strategies not on the frontier of means had a substantial probability of being on the frontier. Some strategies were negatively correlated such that the appearance of one strategy on the frontier never or seldom co-occurred with another. These negative correlations also occurred between strategies that appeared on the frontier of means. CONCLUSIONS: The cost-effectiveness frontier generated by the average costs and benefits of each technology may mask inter-dependencies between technologies. Parameter values that render one technology efficient may render another inefficient, and vice versa. There can also be uncertainty about which interventions appear on the frontier, and the adoption of approaches to select competing interventions solely on the basis of being on the efficiency frontier may be unreasonable.