Probabilistic assessment method of the non-monotonic dose-responses-Part II: Robustness assessment

In a previous study, we presented a new method that uses a large-scale sampling system to probabilistically assess non-monotonic dose-response curves. The statistical plausibility of the characterization was governed by the probability of the dominant category, but gave no information about the specific robustness of the curve. In this paper we propose an improvement to the method by integrating a scoring system based on 4 criteria which can be used to assess the slope robustness of each of the 10,000 sampled curves. The distribution criterion which assesses the number of doses forming a slope, the intensity criterion which assesses the amplitude of the response, and the minimum and maximum confirmation criteria which increase the certainty that the response is present. The probabilistic method was tested on 294 dose-response curves taken from 2 databases and 2 other methodologies currently proposed. A total of 544 dose-response curves have been processed. The developed method offers a concrete and probabilistic characterization of the type of curve analyzed. It evaluates its statistical plausibility and its robustness according to its sampling curves. This method is applicable to all types of data (continuous and discrete) and all experimental curves starting from theoretically 3 doses at least.