Comment on: FIB-4 as an effective screening tool in psoriatic arthritis patients at high-risk for liver disease: a cross-sectional study using FibroScan: reply

Dear Editor, We appreciate the interest expressed in our recent report highlighting the use of fibrosis 4 (FIB-4) as an effective screening tool in PsA patients at high risk for liver disease [1], and would like to thank the readers for their constructive feedback and statistical review of our manuscript. After a thorough review of our data we can confirm that the inconsistencies noted regarding the P-values for the total:HDL cholesterol ratio, triglycerides and the reported number of patients with hypertension (n = 40 vs 38) were the result of a transposition error. We have now checked all other variables and can find no other errors of this nature. Finally, with regards to inconsistencies in CI values, this relates to an error with the log transformation of Beta/standard errors which has also been corrected in the updated Table 1. Chen et al.’s observations echo several inherent problems with statistical reporting in pilot datasets in general, many of which are discussed in the original manuscript’s limitations section. As well as human error, pilot studies are subject to a number of inherent systematic biases including type 1 and type 2 errors, rounding errors and inflation of the statistics due to small sample numbers, to name just a few. Even limited amounts of missing data can have a substantial impact, with small sample sizes leading to data not meeting the assumptions or requiring distribution to undertake the more robust methods of hypothesis testing, which often rely on the principle that the data confirm to the normal distribution, or can be transformed to appear as such via log transformations. For all these reasons, we emphasize that the primary purpose of these pilot data were hypothesis generation and informing the design and implementation of validation work. We are now planning a larger, prospective, adequately powered cohort study which addresses the limitations Chen and ourselves have identified.