Table 4.

Six-month Effect (95% CI) of 4 g/d Lovaza Treatment versus Placebo in Post MI Patients by Intention-to-treat Analysis

	LVESVI	Non-Infarct Myocardial Fibrosis	Infarct Size	LVEF
ITT Analysis (GLMM†)	−5.8% (−10.3%, −1.1%) P = 0.017, N = 358	−5.6% (−10.4%, −0.9%) P = 0.022, N = 358	−3.4% (−17.8%, 13.6%) P = 0.68, N = 358	2.4% (−0.4%, 5.2%) P = 0.094, N = 358
Per Protocol Analysis (t-test‡)	−6.6% (−11.3%, −1.8%) P = 0.0068, N = 247	−5.5% (−10.4%, −0.6%) P = 0.026, N = 171	−6.9% (−19.2%, 5.3%) P = 0.27, N = 254	2.7% (−0.3%, 5.6%) P = 0.073, N = 247
Fish Oil Absolute Change (95% CI)	−2.6 (−3.8, −1.4) [mL/m2], N = 124	−1.3 (−2.5, −0.2) [%], N=84	−1.3 (−2.6, 0.0) [%], N=130	2.2 (1.3, 3.2) [%], N=124
Placebo Absolute Change (95% CI)	−0.5, (−1.8, 0.9) [mL/m2], N=123	0.8 (−0.4, 2.1) [%], N=87	−1.6 (−2.9, −0.4) [%], N=124	0.7 (−0.5, 1.9) [%], N=123

CI was defined as confidence interval, ITT intention-to-treat, LVEF left ventricular ejection fraction, and LVESVI left ventricular end-systolic volume index.

^†The general linear mixed model (GLMM) produces unbiased estimates for responses with missing data (see statistical analysis). LVESVI and Infarct Size were natural logarithm transformed to reduce skewness and/or heteroscedasticity of residuals. Estimates are relative changes.

^‡The per protocol analysis only included patients that attended both visits. No transformations were required, instead Satterthwaite approximation was used for heteroscedasticity. Estimates are relative changes.

The paired absolute changes are calculated on raw data without any transformations.