Impact of methods used to express levels of circulating fatty acids on the degree and direction of associations with blood lipids in humans

Abstract

Numerous studies have examined relationships between disease biomarkers (such as blood lipids) and levels of circulating or cellular fatty acids. In such association studies, fatty acids have typically been expressed as the percentage of a particular fatty acid relative to the total fatty acids in a sample. Using two human cohorts, this study examined relationships between blood lipids (TAG, and LDL, HDL or total cholesterol) and circulating fatty acids expressed either as a percentage of total or as concentration in serum. The direction of the correlation between stearic acid, linoleic acid, dihomo-γ-linolenic acid, arachidonic acid and DHA and circulating TAG reversed when fatty acids were expressed as concentrations v. a percentage of total. Similar reversals were observed for these fatty acids when examining their associations with the ratio of total cholesterol:HDL-cholesterol. This reversal pattern was replicated in serum samples from both human cohorts. The correlations between blood lipids and fatty acids expressed as a percentage of total could be mathematically modelled from the concentration data. These data reveal that the different methods of expressing fatty acids lead to dissimilar correlations between blood lipids and certain fatty acids. This study raises important questions about how such reversals in association patterns impact the interpretation of numerous association studies evaluating fatty acids and their relationships with disease biomarkers or risk.

Absolute versus relative measures of plasma fatty acids and health outcomes: example of phospholipid omega-3 and omega-6 fatty acids and all-cause mortality in women

“…When FA 243 status was expressed in relative (% total), DPA, DHA, and AA were significantly higher 244 among controls but there were no differences in α-linolenic acid. Two recently published 245 cross-sectional studies compared the association between circulating FAs expressed in 246 absolute and proportion of total FAs in relation to blood lipids as disease biomarkers [12,14]. 247 Similar to our findings, both studies reported statistically significant negative associations 249 between AA [14] and linoleic acid [12] and total cholesterol when FA was measured as 250 proportion but the direction was reversed when FA was expressed as absolute concentration.…”

“…The same percentage of a FA may correspond to different absolute amounts 37 of the FA across individuals, or in reverse, the proportion of one FA can differ between 38 people although they have the same absolute amounts of that FA. Several studies have 39 previously compared the two measures of serum or plasma phospholipid FA among people 40 with and without health conditions [10][11][12][13]. Two studies showed significant differences in the 41 FA status using the two measures [10,11].…”

“…Because of this finding, Schwertner and Mosser 42 [10] recommended use of absolute concentrations of FA status as these showed the 43 differences between cases and controls that were predicted a priori whereas using relative 44 levels of FAs (wt%) were much weaker. Similarly, two cross-sectional studies that compared 45 the two FA measures in relation to blood lipids found conflicting directions of association 46 when FAs were expressed as a concentration of FA (µmol/L [12] or mg/dL [14]) or a 47 proportion (% of total FAs [12] or mol% [14]). However, these two studies reached the 48 and serum cholesterol level (mmol/L) as the serum cholesterol concentrations relate to 124 plasma FA levels [12].…”

“…Similarly, two cross-sectional studies that compared 45 the two FA measures in relation to blood lipids found conflicting directions of association 46 when FAs were expressed as a concentration of FA (µmol/L [12] or mg/dL [14]) or a 47 proportion (% of total FAs [12] or mol% [14]). However, these two studies reached the 48 and serum cholesterol level (mmol/L) as the serum cholesterol concentrations relate to 124 plasma FA levels [12]. The proportional hazards assumption and functional form of 125 continuous covariates were assessed using the ASSESS option of PHREG in SAS.…”

“…Four reported some significant differences in FA status 232 depending on the choice of FA expression used [10][11][12]14]. Schwertner and Mosser [10] 233 compared FA status among male patients with and without coronary artery disease (total 234 N=30) and they found the differences in FA composition between cases and controls were 235 much greater when they used absolute amounts (mg/L) compared with relative levels (wt%).…”