Amarin Corp Plc (AMRN)

[sts66]

As ICER noted, an analysis was not presented showing results relative to hsCRP change at one year (in the same way they did LDL-c change in the NEJM paper).

It wasn't directly in the NEJM, but the hs-CRP issue was written up on the AMRN IR FAQ page and I'm 99% sure Dr. Bhatt presented this analysis at some conf. after the AHA:

https://amarincorp.gcs-web.com/static-files/97e31858-1936-43da-80ea-2f78bb7e7260

Vascepa has been shown to significantly lower hsCRP in prior studies and in REDUCE-IT. While the baseline levels of
hsCRP studied in REDUCE-IT were not considered particularly high, the results support that hsCRP was significantly
lowered and future analyses will help establish if inflammation was a contributing factor to the overall CV risk reduction
observed in REDUCE-IT.

Standard method to handle hsCRP results: Log hsCRP 1
Extreme outliers due to infections caused by temporary illness or other factors can heavily influence summary statistics
of hsCRP, even beyond what is handled by using a non-transformed data approach (e.g., a conventional mean or median
on a nominal scale). These individual outlier results can affect a mean or median population measurement in a way that
can convey a misleadingly skewed result for the population studied. For this reason, a more reliable log transformation
of hsCRP is used to incorporate outlier data appropriately within the context of the entire data set. Log transformation
of hsCRP is a standard and generally recognized method employed to put the impact of outlier data into appropriate
context within the entire data set.

hsCRP in REDUCE-IT
The REDUCE-IT statistical analysis plan (SAP) prespecified both direct hsCRP and log transformed hsCRP methods to
measure changes in hsCRP. The comment from the Global Principal Investigator regarding hsCRP not changing from
baseline in the placebo-arm of REDUCE-IT relied upon the log hsCRP method. Again, the log hsCRP method is a standard
and generally recognized method to avoid a misleadingly skewed result due to the high variability of hsCRP for the
population studied, as well as others.

Using the log hsCRP method, there was no increase in log hsCRP from baseline in the placebo arm and the between-
group change was a 22.5% reduction at Year 2. This reflects that the reduction in hsCRP was driven by Vascepa
therapy. These data are presented within the NEJM publication Supplemental Table 4 entitled “Lipid, Lipoprotein, and
Inflammatory Marker Data Over Time – ITT Population”. hsCRP data from that table is as follows:

To summarize measurements from REDUCE-IT presented in the above table, the median percent change from baseline
using the standard and generally recognized log hsCRP method, reflected the following at Year 2:
• No increase from baseline in the mineral oil placebo arm (0.0%, p=0.9203)
• Vascepa change from baseline of -21.8% (p<0.0001)
• Between-group difference in change: 22.5% reduction of hsCRP in the Vascepa versus placebo arms
(p<0.0001)
o The placebo-corrected median percent change was calculated by a Hodges-Lehmann method, whereby
each possible comparison between each individual placebo and each individual Vascepa patient change
is generated, and then a median of all possibilities is taken. This well-accepted statistical comparison
improves the predictive accuracy of determining the most likely between group difference, which at
Year 2 = -22.5%.
• Last visit results, within and between arms, are consistent with Year 2 results
A point of interest may be that the within group placebo absolute change from baseline suggested an increase of hsCRP
(+0.3 log mg/L), but the median of the within group placebo percent changes showed no change from baseline (0.0%).
Both baseline value and change from baseline value would impact individual percentage change from baseline (e.g., a
patient with high baseline hsCRP value may see an absolute increase/decrease that seems numerically impactful, but is
ultimately small on the percent scale [denominator effect]), which is why absolute values may change while observing
little or no effect in percent change from baseline in the placebo group.

Below that section is this - won't cut'n'paste the rest, but it shows how you can calculate the Log hsCRP:

Further Details on Calculation Methods

Log hsCRP data in REDUCE-IT, as published in NEJM
To obtain the log hsCRP values presented in the NEJM supplement: