maxbiostat
commented
9 years ago This is a nice point: (i) we need to clarify the origin and units of the epidemiological data; (ii) potential biases in the coalescent-based reconstructions should be better addressed. A new, smarter sensitivity analysis should do the trick.
Page 6 Demographic reconstruction: You are discussing the increase/decrease in FMDV diversity according to the reported activity of FMD and the control policies (i.e. vaccination) imposed. However, it seems that it is difficult to correlated like-with-like in your graph(s): you have doses/head of vaccine (this could be monovalent, bi-, tri-; strain(s) used), no of FMD cases (I suppose reported no of outbreak - this could be 1 individual of 1000s of animals infected) and viral diversity. One point that is completely missed in your discussion is the vaccine efficacy and this might impact in your analysis (i.e. some reports of drop in efficacy of the O campos vaccine). In addition, you comment that after 2001 an increase in vaccine doses resulted in a decrease in viral diversity: although from the FMD outbreak data is true for type A, it is not clearly valid for type O, which maintained a more stable trend (of course with some fluctuations). Is this, again, an issue due to bias in your data? A previous study (de Silva et al., 2012) describes how BSP incorrectly reconstructed a decrease in the last part of a datum epidemic when the population was still growing. This problem was related to the lack of genealogical information at later times. Would this be the case for your analysis as well?