Bhatt also pointed out that a prespecified analysis of "log-transform" hsCRP did not show an increase in the placebo group. He explained that log-transform analyses are conducted to limit the impact of outlier data; hsCRP is known to have a high degree of variability, and extreme outliers are often seen. "The REDUCE-IT hsCRP data fit the classic rationale for using a log transformation, with the nontransformed values showing a limited number of extreme outliers that heavily skew from normality," he said. "The log-transform analysis, which was just completed in the last couple of days, showed that the values fit a standard bell curve and show strong agreement with the line of normality. Basically, it shows when appropriately log-transformed, the hsCRP isn't increasing in the placebo arm." https://www.medscape.com/viewarticle/904737?src=rss#vp_4