Sex Differences in Omega-3 and Omega-6 PUFA Status: A Systematic Review and Meta-analysis of Cross-Sectional Comparisons

Abstract

Sex hormones may regulate the biosynthesis of PUFAs; however, the extent and consistency of sex differences in different lipid pools remain unclear. We aimed to update and synthesize evidence on sex differences in circulating and tissue omega (ω)-3 (n–3) and ω-6 (n–6) PUFA biomarkers in healthy adults with a mean age ≤50 y. We systematically searched Medline, Embase, and Cochrane Central Register of Controlled Trials through to 21 February, 2026, for cross-sectional reports of n–3 and n–6 PUFA levels for both sexes in plasma total lipids, phospholipids, triglycerides, and cholesteryl esters (CE); erythrocyte total lipids, phosphatidylcholine, and phosphatidylethanolamine; or adipose tissue. Unadjusted mean differences (MD) between male and female n–3 and n–6 PUFA levels were pooled using random-effects meta-analysis. We identified 67 eligible publications, including 85 comparisons from observational studies and clinical trials, yielding a synthesis of 71 PUFA biomarkers. By absolute difference in percentage of total fatty acids, we observed, in males compared with females, higher n–3 docosapentaenoic acid (DPAn–3) [MD = 0.19%, 95% confidence interval (CI): 0.10%, 0.27%, P < 0.0001, I² = 74%] and lower DHA (MD = −0.32%, 95% CI: –0.46%, –0.18%, P < 0.00001, I² = 59%) in total erythrocytes and in total plasma. Males had lower plasma linoleic acid and higher CE γ-linolenic acid and dihomo-γ-linolenic acid. Heterogeneity in total erythrocyte EPA and DHA was explained by study location, with sex differences in erythrocyte DHA driven by studies in North America and Europe. The certainty of the evidence for all outcomes was graded as low or very low. Future studies should consider sex as a modifier of n–3 and n–6 PUFA status in adult males and females of childbearing age. Higher DPAn–3 and lower DHA in males point to important biological differences in this portion of the n–3 PUFA biosynthesis pathway that require further investigation.

This trial was registered at PROSPERO as CRD42023452859.

Statement of Significance.

This systematic review and meta-analysis updates the evidence on sex differences in ω-3 and ω-6 PUFA biomarkers among adults of reproductive age, demonstrating higher DPAn–3 and lower DHA levels in males compared with females in plasma and erythrocyte total lipids. For the first time, the study location explained heterogeneity for erythrocyte DHA, with sex differences driven by studies conducted in North America and Europe.

Introduction

Omega-3 (n–3) and omega-6 (n–6) PUFAs play important physiological roles as energy sources, membrane constituents [1], and precursors to a diverse group of bioactive lipid mediators that modulate immune response, inflammation, and vascular homeostasis [[2], [3], [4]]. Linoleic acid (LA; 18:2n–6) and α-linolenic acid (ALA; 18:3n–3) are nutritionally essential fatty acids from which humans can synthesize longer-chain PUFAs. EPA (20:5n–3) and DHA (22:6n–3) are the predominant long-chain n–3 PUFAs synthesized from ALA and are particularly rich in marine food sources such as oily fish. Similarly, arachidonic acid (AA; 20:4n–6) is a major n–6 PUFA highly enriched in phospholipid (PL) membranes [5] and is formed endogenously from LA while also being available from foods such as meat, eggs, and fish [6]. The biosynthesis of n–3 and n–6 PUFAs occurs through a series of competitive enzymatic reactions involving desaturases and elongases that insert double bonds and 2-carbon units to the fatty acyl chain, respectively [7,8]. Sex hormones, including estradiol and progesterone, are linked to higher expression of key synthesizing enzymes, which may promote the endogenous formation of long-chain PUFAs, including AA and DHA [[9], [10], [11], [12], [13]].

Numerous studies indicate that females have higher circulating DHA levels than males, a consistent finding across levels of dietary n–3 PUFA intake [14,15]. Sex differences in DHA have also been reported across multiple lipid fractions, including total plasma [15], plasma cholesteryl esters (CE) [12], and plasma PL [14]. PLs are tightly regulated and have important cellular functions, including membrane protein regulation, cell growth, and intracellular signaling [16]. The fatty acid composition of plasma PL and erythrocyte membranes strongly correlates with that of cardiac tissue [17]. In particular, erythrocyte EPA + DHA as a percentage of total fatty acids, or the Omega-3 Index (O3I), since its inception in 2004 [18], has been linked to cardiovascular outcomes in various studies [[19], [20], [21]]. In comparison, serum/plasma CE fatty acid composition is a short-term marker of dietary intake and compliance [22,23], and adipose tissue fatty acid composition, in contrast, reflects long-term dietary fatty acid intake over several years [5]. Furthermore, different lipid fractions have distinct patterns of fatty acid compositions [5], reflecting different metabolic pools with the potential to provide a wide range of important health insights.

Higher circulating and tissue levels of n–3 and n–6 PUFAs have been associated with lower incident total cardiovascular events [24], acute coronary syndrome [25], and all-cause mortality [26] in prospective studies. Sex-specific associations have also been reported [27], highlighting potential differences in the accumulation of PUFAs between males and females that remain to be fully understood. Investigation of sex differences in PUFA biomarkers may improve their interpretability in nutrition research, refine disease risk stratification [28,29], and inform the design of future studies linking PUFA status with health outcomes.

Individual studies are often limited in sample size and the reporting of 1 or 2 lipid fractions, which constrains the ability to draw conclusions about overall patterns of sex differences in PUFA status. An earlier meta-analysis, published in 2013 [30], provided an initial synthesis of sex differences in PUFA biomarkers, but the evidence base has expanded since then, and a focused update in healthy, premenopausal females and males of comparable age is needed. Therefore, we conducted a systematic review and meta-analysis (SRMA) of sex differences in circulating blood and tissue biomarkers of n–6 and n–3 PUFA status in young healthy adults. We aimed to: 1) quantify the effect of sex on individual PUFA levels, expressed as relative percentage of total fatty acids, across commonly reported lipid fractions; 2) explore sources of heterogeneity by quantifying effects of potential modifiers (age, BMI, n–3 PUFA status, and geographical location); and 3) evaluate the certainty of the evidence by assessing risk of bias (ROB), conducting sensitivity analyses and performing the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) assessment.

Methods

The SRMA was registered on PROSPERO (CRD42023452859) and was conducted in accordance with The Cochrane Handbook for Systematic Reviews of Interventions [31]. The review followed the PRISMA guidelines.

Search strategy

A systematic literature search was conducted to identify relevant studies reporting on n–3 and n–6 PUFA biomarkers in healthy males and females. Ovid Medline, Embase, and the Cochrane Central Register of Controlled Trials were searched from database inception to 21 February, 2026, along with a manual search through reference lists of included articles. No restrictions were placed on the language of the publication. Non-English publications were translated using Google Translate [32], and data extraction was verified by a second reviewer. The search strategy included key terms to identify commonly reported n–3 and n–6 PUFAs, lipid fractions, and both sexes (Supplementary Table 1). Our search focused on the following 16 n–3 and n–6 PUFAs: ALA, stearidonic acid (SDA; 18:4n–3), eicosatetraenoic acid (20:4n–3), EPA, n–3 docosapentaenoic acid (DPAn–3, 22:5n–3), n–3 tetracosapentaenoic acid (TPAn–3, 24:5n–3), tetracosahexaenoic acid (24:6n–3) and DHA; LA, γ-linolenic acid (GLA; 18:3n–6), dihomo-γ-linolenic acid (DGLA; 20:3n–6), AA, adrenic acid (ADA; 22:4n–6), tetracosatetraenoic acid (24:4n–6), n–6 tetracosapentaenoic acid (24:5n–6, TPAn–6), and n–6 docosapentaenoic acid (DPAn–6; 22:5n–6).

Eligibility criteria

We included studies that met the following criteria: 1) the study was conducted in healthy adults; 2) the mean age of participants was 18 to 50 y; 3) the mean BMI of participants was 18.5 to 29.9 kg/m²; 4) the majority of participants were considered generally healthy; 5) the study must have reported ≥1 of the 16 major n–3 or n–6 PUFAs in at least one of the following lipid fractions in males and females, separately: total plasma/serum, plasma/serum PL, plasma/serum triglycerides (TG), plasma/serum CE, total erythrocytes, erythrocyte phosphatidylcholine (PC), erythrocyte phosphatidylethanolamine (PE), or adipose tissue; and 6) the data were observational and cross-sectional in nature but could be derived from cohort studies, cross-sectional studies, case-control studies, or clinical trials. For experimental studies, only pre-experimental baseline data were included. For observational longitudinal studies with multiple measurements, the first measure was extracted, or if unavailable, the reported mean data were extracted. There was no restriction on the number of study participants. Reviews, meta-analyses, conference abstracts/proceedings, and animal studies were excluded.

Data extraction

An independent reviewer (DL, CS, AK, or TS) screened the titles and abstracts based on the predefined eligibility criteria. If the article appeared to meet the predefined inclusion criteria or if more information was needed, full texts were reviewed by 2 independent reviewers (DL, CS, AK or TS). Disagreements regarding the eligibility of any studies were resolved through discussion or mediation by a third reviewer.

Data for each included study were independently extracted by 2 reviewers and compared for discrepancies using Covidence [33]. The following methodological data were extracted for each study: 1) study location, 2) aim of study, 3) study design, 4) start and end dates of study, 5) study funding sources, and 6) statistical methods. The following participant data were extracted for each study: 1) health status, 2) inclusion and exclusion criteria, 3) method of recruitment, 4) baseline population characteristics (mean age, BMI, and body weight), and 5) unadjusted n–3 or n–6 PUFA levels in the lipid pool(s) of interest.

Fatty acid levels expressed as a relative proportion (%) of total fatty acids and the corresponding measure of uncertainty [SD, SE, or 95% confidence interval (CI)] were extracted for each study. Reports of fatty acid levels in mol % were converted to weight % using the molecular weight respective to the fatty acid. Absolute concentrations (μmol/L, mg/100 mL) were converted to weight % if total fatty acid data were reported. If insufficient information was provided to convert the absolute concentrations, the authors were contacted for further information. If data were unavailable for conversion to % total fatty acid by weight, those articles were excluded. Median and interquartile values were converted to mean and SD values using the methods by McGrath et al. [34]. Reports of geometric means were log-transformed and converted to arithmetic means on the raw scale before inclusion in the meta-analysis [31,35]. When unavailable, SDs of group means were obtained from SE or 95% CI using standard formulas [31].

ROB assessment

The ROB for each included study was independently evaluated by 2 reviewers (Supplementary Table 2) across 5 domains described in the Cochrane Handbook [31]. We used a series of signaling questions specific to cross-sectional comparisons presented by the Agency for Healthcare Research and Quality Methods Guide for Comparative Effectiveness Reviews [36] and adapted them to our research question. In addition, we followed the STROBE recommendations [37]. The domains of bias that were assessed were selection, performance, attrition, detection, and reporting bias. Within each domain, each study was assigned a low, high, or unclear ROB. Any disagreements in judgment were discussed between 2 reviewers, and if needed, with the third reviewer until a consensus was reached.

Certainty of evidence

All outcomes were assessed for certainty of evidence using the GRADE [38] approach and software (GRADEpro GDT) [39]. The certainty of evidence was assessed as high, moderate, low, or very low. The certainty of evidence was graded as low by default for the observational and nonrandomized design of the comparisons and downgraded or upgraded based on prespecified criteria. Criteria for downgrades were based on ROB (if the majority of studies were assigned a high ROB), inconsistency (evidence of unexplained substantial heterogeneity [I² ≥ 50% and P_Q < 0.10]), indirectness (if there were major differences between the populations being studied and the research question, related to the comparators, outcomes, population characteristics, and setting), and imprecision [if the 95% CI crossed the minimally important differences (MIDs)]. The MID was estimated by the difference in PUFA as a % of total fatty acids associated with a 10% risk difference in a health outcome. For outcomes without evidence to support a biomarker–disease relationship, MIDs were estimated based on 10% of baseline values. The estimated MIDs are presented in Supplementary Tables 6 to 13. Other considerations for downgrade were publication bias, based on a significant Egger’s or Begg’s test at P < 0.10, serious asymmetry detected from visual inspection of funnel plots, and confirmation based on imputed studies from Duval and Tweedie trim-and-fill analysis. Upgrading the evidence was considered based on assessment of the magnitude of the effect according to prespecified MIDs: large effect (≥5× MID), moderate effect (≥2× MID), small important effect (≥1× MID), and trivial/unimportant effect (<1 MID) [38].

Statistical analyses

We performed a meta-analysis to evaluate sex differences in n–3 and n–6 PUFAs using a random-effects model with the DerSimonian and Laird inverse-variance method [40] in Cochrane Review Manager (version 8.14). The data were presented as mean difference (MD) with 95% CI, where the study-level MD in PUFAs, expressed as a relative % of PUFA in total FA in each lipid pool, was determined by subtracting mean values of females from males. We prioritized unadjusted crude sex-specific values, which are most commonly reported and most comparable across studies. Publications reporting multiple subgroups (by age, geographical location, or dietary habits etc.) were included as separate comparisons. Data for plasma and serum were pooled together for each lipid fraction, considering no differences were observed between serum and plasma fatty acid compositions [41]. We used the Bonferroni correction to adjust for multiple testing of 71 biomarkers. Therefore, P values < 0.0007 were considered statistically significant. The Cochrane Q test and I² statistic were used to assess heterogeneity, where a P value of < 0.10 from the Cochrane Q test was considered significant heterogeneity and an I² value of ≥50% was considered substantial evidence of between-study heterogeneity.

We examined prespecified study-level moderators using random-effects meta-regression in Stata (meta regress) for outcomes with ≥10 comparisons to explore sources of heterogeneity. Continuous covariates were fitted as linear predictors and categorical covariates as indicator variables. We evaluated the following study-level potential modifiers: age (≤ or > mean age across studies), BMI (healthy or overweight), study location by continent, and O3I (the percentage of EPA + DHA in erythrocytes). Erythrocyte EPA + DHA has been associated with reduced risk of fatal coronary artery disease, with an estimated risk reduction of 30% comparing <4% to >8% [18,19]. Given its clinical relevance, we were interested in quantifying sex differences according to risk categories of O3I based on reported erythrocyte EPA + DHA (% total fatty acids). If unavailable, the sum of EPA and DHA (%) in erythrocytes was calculated. If erythrocytes were not assessed, the O3I was estimated by determining the sum of EPA and DHA (%) in the respective lipid pool for which sex differences were reported. Studies were grouped based on erythrocyte EPA + DHA categories, if available (low: ≤4, moderate: >4 and <8, high: ≥8) [18], or their O3I equivalent categories in total plasma (low: ≤1.7, moderate: >1.7 and <5.2, high: ≥5.2), determined using predictive equations by Schuchardt et al. [42]; plasma PL (low: ≤3.8, moderate: >3.8 and <7.6, high: ≥7.6); plasma CE (low: ≤1.3, moderate: >1.3 and <3.1, high: ≥3.1); or plasma TG (low: ≤0.6, moderate: >0.6 and <1.6, high: ≥1.6), determined by Stark et al. [43]. When referring to O3I in plasma, these are considered O3I equivalents throughout the manuscript. We assessed the credibility of prespecified effect modifiers following the Instrument for Assessing the Credibility of Effect Modification Analyses (ICEMAN) [44]. Post hoc subgroup analyses were conducted according to ROB assessments within 5 domains (selection, performance, attrition, detection, and reporting bias). Although we included studies with a mean age of ≤50 y to approximate the premenopausal status of females, certain studies included females aged >50 y; therefore, we compared studies with an age range of females >50, ≤50, and unknown. As a sensitivity analysis, to evaluate the robustness of the estimated effect, data were pooled using fixed-effect meta-analysis. To further explore sources of heterogeneity, leave-one-out sensitivity analysis was conducted by systematically removing individual comparisons and recalculating the pooled MD and evidence of heterogeneity. Studies were considered influential if they altered the statistical significance of the overall pooled MD and/or removed the evidence of heterogeneity based on P > 0.10 from the Cochrane Q test.

Results

Search results

A total of 23,109 records were retrieved from the 3 databases and the manual search (Figure 1). Duplicate studies were automatically removed by Covidence (n = 8125) and manually (n = 203). A further 13,676 records were excluded based on titles and abstracts; full text was not available for 24 abstracts, and 1081 full texts were reviewed. A total of 1014 articles were excluded after review of the full text, with 67 publications fulfilling the eligibility criteria and included in the meta-analysis, which yielded 85 comparisons. Within-study subgroups (by age, geographical location, or dietary habits) were included as separate comparisons.

FIGURE 1.

Flow of literature. Diagram outlining the number of studies identified, screened, assessed for eligibility, and included in the meta-analysis. THA, tetracosahexaenoic acid; TPAn–3, ω-3 tetracosapentaenoic acid.

Study characteristics

The characteristics of included studies are reported in Table 1 [12,14,15,[45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64], [65], [66], [67], [68], [69], [70], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87], [88], [89], [90], [91], [92], [93], [94], [95], [96], [97], [98], [99], [100], [101], [102], [103], [104], [105], [106], [107], [108]]. Studies were conducted in multiple locations globally, with 16 comparisons in Japan [42,43,60,62,68,69,82,87,94,96,99,101], 12 in Canada [14,15,47,59,63,67,73,79,85,88,93], 7 in Greece [53,58,76,77,81,83,84], 4 in Spain [55,56,102], 8 in the United Kingdom [50,57,71,91,103], 3 in India [64,78], 3 in Vietnam [66], 2 each in Fiji [97], The Netherlands [12,62], Scotland [100], Sweden [104,105], Finland [65,75], and the United States [67,89], and 1 across Italy, Belgium, and England combined [54], and 1 each in Singapore [51], Norway [80], Germany [90], Italy [49], Russia [74], Nepal [61], China [108], Nigeria [107], Tunisia [95], Palestine Territories [48], 73 Hong Kong [70], Israel [72], Taiwan [52], 89 Bulgaria [86], Belgium [92], Korea [96], Lebanon [106], Poland [98], England [75] and Mongolia [96]. Comparisons between males and females were extracted from 62 cross-sectional studies [14,43,[47], [48], [49], [50], [51],[53], [54], [55], [56],[58], [59], [60], [61], [62], [63], [64], [65], [66],[68], [69], [70], [71], [72], [73], [74],[76], [77], [78],80,82,83,86,87,[94], [95], [96], [97], [98], [99], [100], [101], [102],[104], [105], [106], [107], [108]], 1 case-control study [9], 7 nonrandomized experimental studies [12,52,79,81,85,88,93], 5 randomized controlled trials (RCTs) [15,42,57,67,103], 4 cohort studies [67,75,84,91], and 2 longitudinal studies [89,90]. Study-level MDs were calculated from crude mean values, except for studies by Giltay et al. [12] and Rosseneu et al. [92], which provided adjusted values only. From available data, study-level mean BMI in males ranged from 19.5 to 29.5 kg/m² (median 24.2 kg/m²) and in females ranged from 19.9 to 29.9 kg/m² (median 23.4 kg/m²). Study-level mean age ranged from 19.4 to 48 y (median 38.9 y) in males and 19.2 to 47.6 y (median 36.3 y) in females. The sample size of comparisons ranged from 9 to 1037 males (median 42) and 7 to 1611 females (median 57).

TABLE 1.

Baseline characteristics of the included studies

First author, publication year	Study design	Country	Males (n)	Females (n)	BMI of males (kg/m2)	BMI of females (kg/m2)	Age of males (y)	Age of females (y)	Lipid fractions reported	Fatty acids reported	Original expression of data
Abdelmagid, 2015 [47]	Cross-sectional study	Canada	327	499	23.6 ± 3.3	22.8 ± 3.4	22.8 ± 2.5	22.5 ± 2.5	Plasma total lipids	LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, EPA, DPAn–3, DHA	μmol/L, mean ± SD
Almasri, 2023 [48]	Cross-sectional study	Palestine Territories	50	99	23.8 ± 4.09	21.9 ± 3.60	19.7 ± 1.45	20.3 ± 1.46	Erythrocyte total lipids	LA, GLA, DGLA, AA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Antonini, 1970 [49]	Cross-sectional study	Italy	11	11	NR	NR	29.0 ± 6.6	27.4 ± 6.2	Adipose tissue	LA	% TFA, mean ± SD
Bakewell, 2006 [50]	Cross-sectional study	United Kingdom	13	23	23.3 ± 2.8	22.7 ± 2.3	26.0 ± 5	23.0 ± 4.0	Plasma total lipids, TG, CE, and erythrocyte PC	LA, GLA, DGLA, AA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Bi, 2019 [51]	Cross-sectional study	Singapore	62	110	23.9 ± 2.8	21.9 ± 3.7	38.2 ± 13.9	40.4 ± 13.8	Plasma total lipids	LA, GLA, DGLA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Chu, 2006 [52]	Nonrandomized experimental study	Taiwan	16	18	24.0 ± 3.3	20.3 ± 1.7	24.2 ± 2.2	23.6 ± 1.1	Plasma total lipids	LA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Detopoulou, 2018 [53]	Cross-sectional study	Greece	48	58	27.4 ± 3.8	25.7 ± 5.6	43.7 ± 13.3	42.8 ± 12.8	Erythrocyte total lipids	LA, DGLA, AA, ADA, DPAn–6, EPA, DPAn–3, DHA	% TFA, mean ± SD or median (lower-upper quartiles)
Dewailly, 2001 [14]	Cross-sectional study	Canada	179	247	NR	NR	38.7	37.8	Plasma PL	EPA, DHA	% TFA, mean ± SE
DiGiuseppe, 2009 [54]	Cross-sectional study	Italy, Belgium, and England	710	747	27.2 ± 3.0	25.8 ± 5.1	47.0 ± 7.7	44.0 ± 7.6	Plasma total lipids and erythrocyte total lipids	ALA, EPA, DPAn–3, DHA	% TFA, means ± SE for erythrocyte DPAn–3 and DHA; geometric mean (95% CI) for remaining outcomes
Fernandez-Real, 2001 [55]	Cross-sectional study	Spain	38	40	25.4 ± 6	23.4 ± 3.9	40.1 ± 13.3	38.1 ± 9.3	Serum total lipids	LA, DGLA, AA, EPA, DHA	% TFA, mean ± SD
Fernandez-Real, 2005 [56]	Cross-sectional study	Spain	76	40	24.9 ± 3.7	23.2 ± 4.3	40.3 ± 12.2	36.3 ± 10.8	Plasma total lipids	LA, GLA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Fisk, 2018 [57]	Randomized controlled crossover trial	United Kingdom	86	91	26.2 ± 0.3	24.3 ± 0.3	45.2	45.2	Plasma CE, TG	EPA, DPAn–3, DHA	% TFA (g/100 g), Median (25th, 75th percentile)
Fragopoulou, 2021 [58]	Cross-sectional study	Greece	47	55	27.4 ± 3.8	25.7 ± 5.6	43.7 ± 13.3	42.8 ± 12.8	Erythrocyte total lipids	LA, DGLA, AA, ADA, DPAn–6, EPA, DPAn–3, DHA	% TFA, mean ± SD or median (lower-upper quartile)
Garneau, 2012 [59]	Cross-sectional study	Canada	100	98	29.1 ± 6.2		34.3 ± 10.2		Plasma PL	ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Giltay, 2004 [12]	Nonrandomized experimental study	The Netherlands	72	71	NR	NR	29.6 ± 12.9	27.4 ± 10.2	Serum CE	LA, GLA, DGLA, AA, ALA, EPA, DHA	% TFA, mean (95% CI)
Hamazaki, 1986 [60]	Cross-sectional study	Japan: Farming village Fishing village	9 11	10 10	NR NR	NR NR	42 ± 6 41.0 ± 5.0	39 ± 6 38.0 ± 5.0	Plasma total lipids	AA, EPA	% TFA, mean ± SD
Hirai, 1996 [61]	Cross-sectional study	Nepal	42	38	NR	NR	29.0 ± 16	24.0 ± 10	Serum total lipids	LA, AA, ALA, EPA, DHA	mg/100mL, mean ± SD
Hirai, 2000 [62]	Cross-sectional study	Japan The Netherlands	33 20	29 19	NR	NR	University-aged students		Serum total lipids	LA, AA, ALA, EPA, DHA	mg/100mL, mean ± SD
Innis, 1988 [63]	Cross-sectional study	Canada: Innuit Vancouver	41 20	59 10	NR NR	NR NR	21–50	21–50	Erythrocyte PC, PE	LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, SDA, EPA, DPAn–3, DHA	% TFA, mean ± SE
Iwamoto, 2002 [42]	Randomized controlled trial	Japan	20	20	22.2 ± 0.5	20.7 ± 0.5	23.8 ± 0.7	23.6 ± 1.1	Serum CE	LA, AA, ALA, EPA, DHA	Mol %, mean ± SE
Jagannathan, 1969 [64]	Cross-sectional study	India	27	15	NR	NR	22 – 50	26 - 45	Adipose tissue	LA	Weight %, mean ± SE
Kaikkonen, 2014 [65]	Cross-sectional study	Finland	1020	1176	25.7 ± 4.1	24.5 ± 4.6	31.7	31.7	Serum total lipids	LA, AA	% TFA, mean ± SD
Kawabata, 2011 [43]	Cross-sectional study	Japan	20	30	20.5 ± 2.6	20.7 ± 2.0	20.2 ± 1.0	20.0 ± 1.5	Plasma PL, TG, CE	LA, DGLA, AA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Kieu, 2002 [66]	Cross-sectional study	Vietnam: Urban Suburban Rural	32 40 39	68 58 59	22.5 ± 2.6 21.2 ± 3.3 19.5 ± 2.3	22.4 ± 3.5 21.4 ± 3.7 19.9 ± 2.9	47.5 ± 5.5 46.6 (5) 46.4 (4.7)	47.5 ± 5.2 47.6 (6.4) 47.2 (5.9)	Serum total lipids	LA, GLA, DGLA, AA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Klingel, 2017 [67]	GOLDN: Intervention Study GONE FISHN’: Cohort study	United States Canada	57 37	61 57	25.3 ± 0.7 24.61 ± 3.22	26.0 ± 0.8 23.72 ± 3.47	22.3 ± 0.8 22.54 ± 2.25	22.7 ± 0.5 21.63 ± 1.51	Erythrocyte total lipids	LA, DGLA, AA, ADA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SE
Kuriki, 2003 [68]	Cross-sectional study	Japan	15	79	22.3 ± 3.6	21.5 ± 2.1	45.3 ± 10.6	47.2 ± 8.1	Plasma total lipids	LA, AA, ALA, EPA, DPAn–3, DHA	% TFA (by weight), mean ± SD
Kurotani, 2014 [69]	Cross-sectional study	Japan	291	205	23.6 ±3.1	21.0 ± 2.9	44.2 ± 10.8	40.8 ± 10.4	Serum PL, CE	LA, DGLA, AA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Lee, 2000 [70]	Cross-sectional study	Hong Kong	81	113	24.3 ± 3.6	23.7 ± 3.6	40.0 ± 6.4	40.5 ± 6.8	Serum total lipids	LA, GLA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Leeson, 2002 [71]	Cross-sectional study	United Kingdom	122	137	24.3 ± 3.3	24.1 ± 4.3	23.0	23.0	Erythrocyte total lipids	EPA, DHA	% TFA, mean ± SD
Lemaitre, 2008 [72]	Cross-sectional study	Israel	112	118	NR	NR	43.9		Erythrocyte total lipids	LA, DGLA, AA, ADA, EPA, DHA	% TFA, mean ± SD
LeMoire, 2022 [73]	Cross-sectional study	Canada	254	549	23.50 ± 3.38	22.34 ± 3.52	22.68 ± 2.35	22.50 ± 2.46	Plasma total lipids	LA, GLA, DGLA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Lyudinina, 2014 [74]	Cross-sectional study	Russia	17	17	25.4 ± 24.5–27.1	26.0 ± 24.7–30.2	40.5 (27.5–46.8)	34.0 (26.0–39.8)	Plasma total lipids	LA, AA, EPA, DHA	% TFA, median (lower-upper quartiles)
Makinen, 2025 [75] ALSPAC Cohort YFS Cohort	Cohort study	England Finland	1037 136	1611 147	NR	NR	24.3	24.2	Serum total lipids	LA, DHA	Ratio of fatty acid to total fatty acids, means ± SE
Mamalakis, 1998 [76]	Cross-sectional study	Greece	85	59	27.2 ± 3.2	23.4 ± 3.1	38.0		Adipose tissue	LA, DGLA, AA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Mamalakis, 2008 [77]	Cross-sectional study	Greece	146	178	29.5 ± 4.6	29.9 ± 5.5	44.7 ± 10.3		Serum PL, adipose tissue	LA, GLA, DGLA, AA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Manjari, 2001 [78]	Cross-sectional study	India Nonvegetarians Vegetarians	30 20	14 10	NR	NR	30.9 ± 7.67	32.8 ± 7.33	Plasma total lipids	LA, GLA, DGLA, AA, ALA, EPA, DHA	% TFA, mean ± SE
Metherel, 2009 [79]	Nonrandomized experimental study	Canada	9	7	NR	NR	22.4 ± 1.2	22.1 ± 1.8	Plasma total lipids, erythrocyte total lipids	LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Metherel, 2020 [15]	Randomized controlled trial	Canada	44	42	24.11 ± 3.49	23.10 ± 2.97	21.93 ± 2.40	21.27 ± 1.97	Plasma total lipids	LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, EPA, DPAn–3, DHA	% TFA (mol%), mean ± SE
Min, 2014 [80]	Cross-sectional study	Norway	41	40	NR	NR	20–50		Plasma TG, CE and erythrocyte PC, PE	LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Mougios, 1998 [81]	Nonrandomized experimental study	Greece	11	11	25.7 ± 0.7	27.7 ± 1.7	41.9 ± 1.6	44.5 ± 1.1	Plasma total lipids, TG, and adipose tissue	LA, GLA, AA	%TFA (mol%), mean ± SE
Nakamura, 1995 [82]	Cross-sectional study	Japan: 30 s 40 s	18 13	13 13	23.9 ± 3.3 24.9 ± 2.7	21.9 ± 2.7 23.5 ± 2.7	43.7 ± 2.60	38.1 ± 5.44	Plasma total lipids	LA, AA, ALA, EPA, DHA	% TFA (w/w), mean ± SD
Panagiotakos, 2009 [83]	Cross-sectional study	Greece	189	185	NR	NR	44.0 ± 13.0	40.0 ± 15.0	Plasma total lipids	AA, EPA, DPA, DHA	% TFA, mean ± SD
Papagiannopoulos, 2025 [84] Epirus Health Study	Prospective cohort study	Greece	615	512	26.8 (25.1, 29.4)	27.8 (22.2, 27.9)	45.0 ± 10.9	44.4 ± 10.5	Plasma total lipids	LA, DHA	% TFA, Median (25th, 75th percentile)
Patterson, 2015 [85]	Nonrandomized experimental study	Canada	11	9	NR	NR	23.3 ± 4.0		Erythrocyte total lipids	LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Petrova, 2011 [86]	Cross-sectional study	Bulgaria	50	58	29.4 ± 4.9	26.7 ± 5.8	46.0 ± 13.0	42.0 ± 11.0	Adipose tissue	LA, GLA, DGLA, AA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Poudel-Tandukar, 2012 [87]	Cross-sectional study	Japan	286	203	23.5 ± 3.2	20.8 ± 2.8	44.1 ± 10.9	40.6 ± 10.4	Serum CE	LA, GLA, DGLA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Pufahl, 2025 [88]	Single-arm trial	Canada	14	15	25.6 ± 1.9	24.1 ± 2.2	23 ± 4	22 ± 3	Plasma total lipids, erythrocyte PL	LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Reeves, 1984 [89]	Longitudinal study	United States	35	40	NR	NR	21–49	20–53	Plasma total lipids	LA, AA	mg/dl, mean ± SE
Reuter, 1984 [90]	Longitudinal study	Germany	45	23	NR	NR	<45	<45	Serum total lipids	LA	% TFA, mean ± SD
Riemersma, 2003 [91]	Cohort study	United Kingdom	21	22	23.8 ± 0.7	24.6 ± 1.1	44.0 ± 2.6	42.9 ± 2.5	Plasma total lipids	LA	% TFA, mean ± SE
Rosseneu, 1994 [92]	Case-control study	Belgium	933	907	NR	NR	22.7		Plasma CE	LA, AA	% TFA, mean ± SE
Rudkowska, 2013 [93]	Nonrandomized experimental study	Canada	13	17	29.1 ± 1.2	29.2 ± 0.8	33.5 ± 2.1	34.4 ± 2.5	Erythrocyte total lipids	EPA, DHA	% TFA, mean ± SE
Sasaki, 2000 [94]	Cross-sectional study	Japan	42	44	22.3 ± 2.5	20.8 ± 2.5	41.9 ± 8.3	43.2 ± 10.6	Serum PL	LA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Sfar, 2010 [95]	Cross-sectional study	Tunisia	54	58	NR	NR	46.11 ± 5.27	47.40 ± 6.45	Plasma total lipids	LA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Shiwaku, 2004 [96]	Cross-sectional Study	Japan Korea Mongolia	193 240 99	218 178 152	23.3 ± 3.3 24.3 ± 3.0 26.4 ± 4.2	22.5 ± 3.4 23.6 ± 3.1 25.6 ± 4.6	46.6 ± 6.9	47.5 ± 7.2	Plasma total lipids	LA, AA, ALA, EPA, DPA, DHA	%TFA (mol%), mean ± SD
Sutherland, 1995 [97]	Cross-sectional study	Fiji: urban rural	39 37	44 34	26.5 ± 3.9 25.1 ± 2.8	27.2 ± 5.1 26.4 ± 46.0	39.0 ± 16.0	36.0 ± 15.0	Erythrocyte total lipids	LA, GLA, DGLA, AA, ADA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Szewczyk, 2025 [98]	Cross-sectional study	Poland	26	34	NR	NR	46 ± 5		Plasma total lipids	LA, GLA, DLGA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Takita, 1996 [99]	Cross-sectional study	Japan (by age): 20–29 30–39 40–49	28 87 81	15 58 48	NR	NR	20–29 30–39 40–49	20–29 30–39 40–49	Plasma total lipids	LA, AA, EPA, DHA	% TFA, mean ± SD
Tavendale, 1992 [100]	Cross-sectional study	Scotland	529 508	518 469	NR	NR	40–44 45–49	40–44 45–49	Adipose tissue	LA, GLA, DGLA, AA, DHA	% TFA, mean ± SD
Umemura, 2005 [101]	Cross-sectional study	Japan	175	246	21.9 ± 2.9	21.1 ± 3.0	19.4 ± 1.0	19.2 ± 0.5	Serum total lipids	LA, GLA, DGLA, AA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD
Valles, 1988 [102]	Cross-sectional study	Spain	Younger group: 17 Older group: 16	Younger group: 17 Older group: 16	NR	NR	Younger group: 29 ± 6 Older group: 48 ± 5	Younger group: 28 ± 7 Older group: 47 ± 5	Plasma PL, TG, CE	LA, EPA	% TFA, mean ± SD
Walker, 2014 [103]	Randomized controlled trial	United Kingdom	99	104	NR	NR	50.0 ± 15.4		Plasma TG, CE, erythrocyte total lipids, adipose tissue	EPA, DHA	% TFA, mean ± SD
Warensjo, 2006 [104]	Cross-sectional study	Sweden	554	295	24.7 ± 3.2	23.7 ± 3.8	40.6 ± 9.1	40.6 ± 9.9	Serum CE	LA, GLA, DGLA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Wennberg, 2009 [105]	Cross-sectional study	Sweden	88	92	NR	NR	45 ± 11.3		Erythrocyte total lipids	LA, ALA, EPA, DHA	% TFA, Geometric means (95% CI)
Yammine, 2018 [106]	Cross-sectional study	Lebanon	129	266	27.4 ± 5.0	29.6 ± 5.9	38.8 ± 16.3	47.3 ± 14.0	Serum PL	LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, EPA, DPAn–3, DHA	% TFA, mean (95% CI)
Yeh, 1996 [107]	Cross-sectional study	Nigeria	236	161	22.14 ± 3.37	24.27 ± 5.06	40.96 ± 9.75	37.82 ± 9.17	Serum total lipids	LA, GLA, DGLA, AA, ALA, EPA, DHA	% TFA, mean ± SD
Zhang, 2023 [108]	Cross-sectional study	China	24	30	22.9 ± 2.3	21.0 ± 2.4	26.0 ± 4.0	25.0 ± 5.0	Serum total lipids	LA, GLA, DGLA, AA, ADA, ALA, EPA, DPAn–3, DHA	% TFA, mean ± SD

Abbreviations: AA, arachidonic acid; ADA, adrenic acid; ALA, α-linolenic acid; ALSPAC, Avon Longitudinal Study of Parents and Children; CE, cholesteryl esters; CI, confidence interval; DGLA, dihomo-γ-linolenic acid; DPA, docosapentaenoic acid; DPAn–3, n–3 docosapentaenoic acid; DPAn–6, n–6 docosapentaenoic acid; GLA, γ-linolenic acid;GOLDN, The Genetics of Lipid Lowering Drugs and Diet Network; LA, linoleic acid; NR, not reported; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PL, phospholipid; SDA, stearidonic acid; TFA, total fatty acids; TG, triglycerides; YFS, Young Finns Study.

Twenty-three studies reported fatty acid compositions by sex in total plasma [15,47,[50], [51], [52],54,56,60,68,73,74,78,79,[81], [82], [83], [84],88,89,95,96,98,99], 13 in total erythrocytes [48,53,54,58,67,71,72,79,85,93,97,103,105], 1 in erythrocyte PL [88], 3 in erythrocyte PC [50,63,80], 2 in erythrocyte PE [63,80], 11 in total serum [55,61,62,65,66,70,75,90,101,107,108], 8 in adipose tissue [49,64,76,77,81,86,100,103], 7 in plasma CE [43,50,57,80,92,102,103], 5 in serum CE [12,42,69,87,104], 4 in plasma PL [14,43,59,102], 3 in serum PL [69,77,94], and 7 in plasma TG [43,50,57,80,81,102,103]. For the purposes of the meta-analysis, data from serum and plasma were grouped together. Total plasma/serum, plasma/serum CE, and plasma/serum PL will be referred to as total plasma, plasma CE, and plasma PL for the remainder of the paper. Some studies were further stratified by village [60], location [62,63,66,96,97], cohort [67,75], dietary preference [78], and age [57,82,97,100,102].

ROB

In all 5 ROB domains (Supplementary Figure 1), the majority of studies had a low or unclear ROB, with <20% of studies evaluated as high ROB in each domain. Risk of selection bias (67% low, 28% unclear, and 5% high) indicates that most publications uniformly applied eligibility criteria to the comparison groups. Risk of attrition bias (35% low and 65% unclear) reveals that most studies did not clearly address sampling strategy or missing data. Risk of detection bias (31% low, 55% unclear, and 14% high) indicated that, in most studies, the methodology used for identification of fatty acids and total fatty acids was unclear, whereas 9 studies determined total fatty acids using fewer than the 13 fatty acids (at >0.55% abundance) that represent >92.9% of all fatty acids [109]. For risk of performance bias (1% low, 85% unclear, and 14% high), it was mostly unclear whether concurrent exposures or interventions influenced the results. Finally, the risk of reporting bias (26% low, 73% unclear, and 1% high) was driven by most studies not providing sufficient information regarding the primary/secondary outcomes of the study.

Meta-analysis of sex differences in PUFA biomarkers

The pooled analyses for n–3 and n–6 PUFAs with ≥2 comparisons are reported and presented for total plasma, plasma PL, plasma TG, and plasma CE in Figure 2, and total erythrocytes, erythrocyte PC, erythrocyte PE, and adipose tissue in Figure 3. As a result, pooled analyses in LA, GLA, DGLA, AA, ADA, DPAn–6, ALA, SDA, EPA, DPAn–3, and DHA across 8 lipid fractions are reported. Individual forest plots are available in Supplementary Figures 2 to 72.

FIGURE 2.

Sex differences in omega (ω)-3 and ω–6 PUFAs in plasma total lipids, plasma phospholipids, plasma triglycerides, and plasma cholesteryl esters. Black diamonds represent pooled mean differences for each respective outcome, with the lines representing 95% confidence intervals (CIs). N comparisons describe the number of comparisons pooled for each outcome, and N describes the number of participants included in each outcome. P_MD represents the P value derived from the test of the overall mean difference (MD) in the inverse-variance random-effects meta-analysis. I² describes the percentage of variability in effect estimates due to heterogeneity rather than sampling error, and P_Q is the P value derived from the chi-squared test of heterogeneity (Cochrane’s Q statistic).

FIGURE 3.

Sex differences in omega (ω)-3 and ω-6 PUFAs in erythrocyte total lipids, erythrocyte phosphatidylcholine, erythrocyte phosphatidylethanolamine, and adipose tissue. Black diamonds represent pooled mean differences for each respective outcome, with the lines representing 95% confidence intervals (CIs). N comparisons describe the number of comparisons pooled for each outcome, and N describes the number of participants included in each outcome. P_MD represents the P value derived from the test of the overall mean difference (MD) in the inverse-variance random-effects meta-analysis. I² describes the percentage of variability in effect estimates due to heterogeneity rather than sampling error, and P_Q is the P value derived from the chi-squared test of heterogeneity (Cochrane’s Q statistic).

Sex differences in n–6 PUFA biomarkers

In overall pooled analysis, males had lower total plasma LA (MD = –0.96%, 95% CI: –1.32%, –0.60%; P < 0.00001, I² = 85%) (Figure 2), and higher GLA and DGLA in plasma CE compared with females (CE GLA: MD = 0.09%, 95% CI: 0.06%, 0.12%, P < 0.00001; I² = 26%; CE DGLA: MD = 0.08%, 95% CI: 0.05%, 0.12%, P < 0.00001; I² = 84%) (Figure 2). Sex differences in LA, GLA, and DGLA were not observed in other lipid fractions, and no differences were observed for AA, ADA, and DPAn–6 in any lipid fractions (FIGURE 2, FIGURE 3). Evaluation of potential sources of heterogeneity contributing to sex differences in n–6 PUFAs is summarized in Supplementary Figures 73 to 103.

Modification of sex differences in n–6 PUFA biomarkers

Modifiers of LA

O3I was inversely associated with sex differences in total plasma LA (test of group differences: P < 0.001; residual I² = 66%), with the largest difference in studies with O3I ≥5.2% (MD = −2.38%, 95% CI: −3.52%, −1.24%), followed by O3I >1.7% and < 5.2% (MD = −0.72%, 95% CI: −1.20%, −0.24%) and not different for O3I ≤1.7% (MD = 1.83%, 95% CI: −0.25%, 3.90%) (Supplementary Figure 73). Continuously, every % of O3I was inversely associated with sex differences in plasma LA (β = −0.45%, 95% CI: −0.61%, −0.28%, P < 0.001, residual I² = 55%) (Supplementary Figure 74). In addition, studies with mean age >35.8 y were associated with larger sex differences (MD = −1.40%, 95% CI: −1.92%, −0.88%) compared with ≤35.8 (MD = −0.15%, 95% CI: −0.68%, 0.37%; test of group differences: P < 0.001, residual I² = 83%) (Supplementary Figure 73), and every year of age was inversely associated with sex differences (β = −0.07%, 95% CI: −0.10%, −0.04%, P < 0.001, residual I² = 81%) (Supplementary Figure 74). Larger sex differences were observed in studies conducted in Asia (MD = −1.41%, 95% CI: −2.13%, −0.68%) compared with Europe (MD = −0.42%, 95% CI: −0.97%, 0.14%) (test of group differences: P = 0.04, residual I² = 85%) (Supplementary Figure 73), with a trend also detected in CE LA (test of group differences: P < 0.001, residual I² = 58%) (Supplementary Figure 77).

Modifiers of γ-linolenic acid

For total plasma GLA, Asian studies yielded lower levels (MD = −0.06%, 95% CI: −0.11%, −0.00%), and North American studies higher levels in males (MD = 0.06, 95% CI: 0.05%, 0.08%) (P = 0.02, residual I² = 74%) (Supplementary Figure 84).

Modifiers of dihomo-γ-linolenic acid

A priori categorical and continuous analyses did not show any subgroup effects for total plasma DGLA (Supplementary Figures 88 and 89).

Modifiers of AA

Total plasma AA sex differences varied by O3I (P = 0.002; residual I² = 55%) and location (P = 0.01; residual I² = 59%) (Supplementary Figure 96). Studies with high O3I were associated with significant sex differences (MD = −0.49%, 95% CI: −0.76%, −0.23%), whereas studies with low (MD = 0.11%, 95% CI: −0.31%, 0.52%) and moderate O3I (MD = 0.03%, 95% CI: −0.12%, 0.17%) were not. Continuously, study mean O3I and age were inversely associated with sex differences in plasma AA (β per 1% O3I = −0.10%, 95% CI: −0.15%, −0.05%, P < 0.001; residual I² = 55%; β per 1-y age: −0.02%, 95% CI: −0.03%, −0.01%, P < 0.001, residual I² = 53%) (Supplementary Figure 97). The inverse association with age was consistently seen in erythrocyte total lipids (Supplementary Figures 100 and 101).

Sex differences in n–3 PUFA biomarkers

Males had higher total plasma DPAn–3 (MD = 0.05%, 95% CI: 0.03%, 0.07%; P < 0.00001; I² = 76%) and lower DHA (MD = −0.25%, 95% CI: −0.30%, −0.19%; P < 0.00001; I² = 81%) compared with females (Figure 2), and in total erythrocytes, a similar pattern was identified of higher DPAn–3 (MD = 0.19%, 95% CI: 0.10%, 0.27%, P < 0.0001; I² = 74%) and lower DHA in males compared with females (MD = −0.32%, 95% CI: −0.46%, −0.18%, P < 0.00001, I² = 59%) (Figure 3). However, substantial heterogeneity was detected for total plasma and erythrocyte DPAn–3 and DHA. Evaluation of potential sources of heterogeneity contributing to sex differences in n–3 PUFAs is summarized in Supplementary Figures 104 to 147.

Modification of sex differences in n–3 PUFAs

Modifiers of EPA

North American (MD = 0.07%, 95 CI%: 0.02%, 0.11%) and European studies (MD = 0.05%, 95% CI: 0.02%, 0.08%) resulted in higher erythrocyte EPA levels in males compared with females that was not observed in Asian studies (MD = 0.01%, 95% CI: −0.02%, 0.03%) (Supplementary Figure 112) and that explained the heterogeneity (test of group differences: P = 0.03, residual I² = 21%).

Every year in age was positively associated with sex differences in plasma PL EPA (0.02%, 95% CI: 0.003%, 0.03%, residual I² = 80%) (Supplementary Figure 117).

Modifiers of DHA

Similar to EPA, a subgroup effect by study location was detected for erythrocyte DHA, which explained most of the heterogeneity (test of group differences: P = 0.001, residual I² = 20%) (Supplementary Figure 140). Lower erythrocyte DHA in males was driven by North American (−0.47%, 95% CI: −0.64%, −0.29%) and European studies (−0.36%, 95% CI: −0.51%, −0.21%), with no differences in Asian studies (0.02%, 95% CI: −0.15%, 0.18%). Similarly, lower CE DHA in males was seen in European studies (−0.07, 95% CI: −0.12, −0.02) and not in Asian studies (0.06%, 95% CI: −0.05%, 0.16%) (Supplementary Figure 144).

Credibility of subgroup findings

Applying the ICEMAN credibility assessments to a priori subgroup analyses, the credibility of effect modification of LA and AA outcomes was mostly rated as “moderate.” Effect modification of EPA and DHA outcomes was rated “low” to “moderate.” Complete assessments are available in Supplementary Table 3.

Post hoc subgroup analyses

Female age range

As an indirect measure of premenopausal status (≤50 y age range) compared with mixed age of females (>50 y age range), we found sex differences in CE LA and plasma total DGLA only in studies with mixed-age females (Supplementary Figures 78 and 90). In contrast, sex differences in plasma total GLA and CE DHA were only observed in studies limited to ≤50 y in age range (Supplementary Figures 86 and 146).

ROB

Assessing potential influence of methodological ROB as sources of heterogeneity, we found that sex differences in plasma LA were driven by studies with unclear reporting ROB (Supplementary Figure 76), and in plasma GLA, females had higher levels in studies with unclear attrition ROB but lower levels in studies with low attrition ROB (Supplementary Figure 87). Finally, unclear selection ROB was associated with a smaller difference in plasma DHA compared with low selection ROB (Supplementary Figure 139).

Sensitivity analyses

PUFA biomarkers with altered significance of pooled sex difference in a fixed-effect meta-analysis are presented in Supplementary Table 4. Importantly, all of the significant differences in the random-effects model persisted in the fixed-effect model without substantial changes in the magnitude or direction of effect (data not shown), suggesting no evidence of small-study effects on significant findings [31]. However, a gain of significance was observed in the fixed-effect model for plasma PL and TG DHA; plasma CE LA and ALA; and erythrocyte ALA and EPA, suggesting small-study effects could be contributing to these null findings.

Influential comparisons resulting from the leave-one-out analysis are presented in Supplementary Table 5. Removal of the individual comparisons did not result in loss of significance for any n–3 or n–6 PUFA biomarkers, except for erythrocyte DPAn–3 (P = 0.0008), suggesting robustness of the pooled estimates. Removal of individual comparisons explained heterogeneity for several outcomes in plasma TG, CE, adipose tissue, erythrocyte total lipids, and erythrocyte PC, indicating that heterogeneity was mainly driven by 1 comparison.

Publication bias

Egger’s test revealed potential publication bias for EPA in total plasma (Begg’s test P = 0.46; Egger’s test P = 0.08), plasma CE (Begg’s test P = 0.14; Egger’s test P = 0.02), and plasma PL (Begg’s test P = 0.27; Egger’s test P = 0.03), and for DGLA in total plasma (Begg’s test P = 0.30; Egger’s test P = 0.08) (Supplementary Figures 148–151). Visual inspection of the funnel plots indicated mild asymmetry. However, trim-and-fill analysis did not impute any missing studies or adjust the pooled effect size (Supplementary Figures 152–155). The detected publication bias could be reflecting the presence of substantial heterogeneity (>50%) in these outcomes. Publication bias was not detected in any of the remaining outcomes with ≥10 comparisons (data not shown).

GRADE

All outcomes were assessed as low or very low certainty of evidence (see Supplementary Tables 6–13 for detailed justifications). A common reason for downgrading was serious imprecision due to substantial heterogeneity that was not explained by subgroup analysis or sensitivity analysis. Certain outcomes were downgraded due to wide 95% CIs that crossed the MID and altered the significance of the pooled effect in sensitivity analyses. Across all outcomes, there was no major concern for ROB, publication bias, or indirectness that warranted downgrades. There were no upgrades to the certainty of evidence because we did not observe large magnitudes of effects, and dose–response analysis was not applicable.

Discussion

This SRMA of 67 publications and 85 cross-sectional comparisons evaluated sex differences in circulating and adipose levels of n–3 and n–6 PUFAs in healthy participants. We updated an SRMA from 2013 and restricted the inclusion of studies with a mean age ≤50 y to better target sex differences between premenopausal females and males of similar age. This is particularly relevant given the potential hormone-related differences in PUFA levels based on menopause status [12,13,110] and the lower incidence of cardiovascular disease (CVD) in premenopausal females compared with age-matched males [111].

Compared with females, males were associated with having higher DPAn–3 and lower DHA status in total plasma and total erythrocytes. Although reporting only levels limits interpretations on metabolism, this is consistent with previous research indicating potential for higher DHA synthesis from DPAn–3 in females that is likely mediated by sex hormones. Isotope-labeled ALA feeding studies indicate a higher percent conversion of dietary ALA to DHA in females compared with males, with DHA levels elevated in females taking estradiol oral contraceptives [110,112]. Our reported lower DPAn–3 in females aligns with compartmental modeling data indicating 3-fold greater utilization of DPAn–3 for DHA synthesis in females compared with males [113]. In the n–3 PUFA biosynthetic pathway, elongation of DPAn–3 to TPAn–3 is catalyzed by ELOVL2, and in vitro studies point to estradiol as a potent stimulator of ELOVL2 expression [10]. Although other potential mechanisms exist, circulating estradiol may be an important underlying mechanism driving our reported sex differences in DPAn–3 and DHA status.

Males displayed lower total plasma LA and higher plasma CE GLA and DGLA, congruent with reported sex differences in the circulating lipidome profile, including higher levels of most CE species (including 18:3 and 20:3 fatty acids) in middle-aged males compared with females (45–50 y) [114]. However, because we are reporting differences in levels, and plasma CE is influenced by recent diet [5] and largely reflects lecithin-cholesterol acyl transferase activity [5,115], sex-specific metabolism of n–6 PUFAs warrants further investigation while considering these variables. Notably, despite matched dietary LA intake, animal studies have shown greater accumulation of exogenous deuterium-labeled LA in females than males across varying ALA levels [116], suggesting potential sex differences in LA levels independent of dietary LA that remain to be elucidated in humans.

Numerous modifiable and nonmodifiable factors have been shown to affect PUFA status [117]; therefore, we also investigated how study-level BMI, O3I, and age influenced sex differences in n–3 and n–6 PUFA biomarkers. Higher study-level O3I status, estimated using plasma EPA + DHA, was associated with larger sex differences, with higher plasma LA and AA levels in females. Higher circulating EPA + DHA reflects higher dietary EPA + DHA [118,119], suggesting that females may have a greater capacity than males to maintain LA/AA status as dietary EPA + DHA increases. AA is essential for early-life development [120] and is the predominant long-chain PUFA in human milk [121]. Previous studies show relatively consistent AA levels in human milk between mothers consuming heterogeneous diets [121,122], which are likely maintained by maternal stores [123].

With increasing study mean age, sex differences in plasma n–6 PUFAs (LA, AA) became larger with higher levels in females. In support, NHANES data showed higher plasma LA and AA concentrations in older adults (40–59 y) compared with younger adults (20–39 y), although sex-specific differences in age-related change were not reported [124]. In a metabolomics study with >2000 individuals aged 48 to 94 y, a strong positive correlation was observed between age and plasma LA concentrations in females only and not in males [125]. Extending the observation of sex-specific relationships between age and LA, we quantified the change in female-higher differences with every year of age in plasma LA (0.07%/y), whereas previous cohort analyses generally compared age categories [125]. Conversely, age was associated with sex differences in plasma PL EPA in the opposite direction, where increasing age corresponded with higher levels in males relative to females. This may be driven by the overall increase in EPA with age, together with higher EPA levels in males than females across age groups [126]. However, it is unclear whether age-related widening of the sex differences in our analysis reflects decreasing levels in males, increasing levels in females, or a combination of both. The subgroup findings for plasma LA and AA were generally consistent in categorical and continuous analyses, but should be interpreted with caution as we rated the credibility as “moderate.”

Study locations may reflect genetic and lifestyle differences, including dietary preferences, cultural habits, availability of local foods, and the nutrient composition of those foods. As such, we discovered that data from Asia differed from those from North America and Europe across multiple PUFA biomarkers. The sex difference in CE LA, higher in females, was larger in studies conducted in Asia than in Europe. Plasma GLA and AA were higher in females from studies in Asia, but lower in studies from North America. GLA is synthesized by Δ-6 desaturase, and AA is synthesized by Δ-5 desaturase (encoded by FADS2 and FADS1, respectively). The AA:LA ratio, a proxy measure of desaturase activity, has been reported to be higher in Caucasians than in Asians and to be strongly associated with a single-nucleotide polymorphism in FADS1 (rs174547) [127]. Specifically, the C allele was associated with reduced desaturase activity in both ethnicities; however, it is the minor allele in Caucasians and the major allele in Asians, highlighting ethnic-specific genetic regulation of desaturase activity that may contribute to PUFA status. Further investigation into sex by genotype interactions in plasma n–3 and n–6 PUFAs may provide additional insight into the regulation of PUFA synthesis across different ethnic groups. Moreover, higher erythrocyte EPA and lower erythrocyte DHA in males were driven by North American and European studies, with no detectable sex differences in Asian studies. This underscores the impact of study location on n–3 PUFA level comparisons and that overall sex-specific patterns in erythrocyte EPA and DHA may only be generalizable to North America and Europe. Notably, location explained substantial heterogeneity in both erythrocyte EPA and DHA. Although credibility was rated as low and moderate, respectively, sex differences in these biomarkers could be considered separately by region.

An important limitation of our study is the cross-sectional nature of the data, which precludes determining causal relationships. To assess potential bias due to confounding, we evaluated whether major exposures affecting PUFA status [117] were adjusted/accounted for, including n–3 and n–6 PUFA dietary intake, n–3 PUFA supplements, age, BMI, medication use, smoking, and alcohol consumption. In most studies, this was unclear, which suggests that findings are possibly confounded by these factors in addition to unknown factors. For example, physical activity has been identified to potentially impact n–3 PUFA status [117], with exercise shown to affect the composition of PL DHA in skeletal muscle in a 4-wk endurance training study [128]. An ideal analysis may have included only estimates that are adjusted for a minimum set of confounding factors, but this would have substantially reduced the number of eligible comparisons and limited the comprehensiveness of our review. Moreover, it would be important to evaluate sex differences in response to differing dietary patterns. However, an insufficient number of studies reported dietary intake to investigate this relationship, identifying a critical gap in the n–3 PUFA literature as it pertains to observed sex differences in PUFA status.

Overall, the certainty of the evidence was graded as low or very low for all outcomes. Evidence of substantial heterogeneity was not explained by subgroup or sensitivity analyses for most biomarkers. Heterogeneity could be influenced by differences in how studies derived “% total fatty acids” (e.g., % of total chromatographic peaks compared with % of total identified fatty acids) and differences between-study designs (RCTs compared with observational studies) [129]. Most of the studies had unclear ROB in performance, attrition, detection, and reporting bias, emphasizing the need for improved reporting of study methodologies, including complete descriptions of sampling methods, missing data, primary/secondary objectives, and fatty acids measured. Moreover, the observed summary estimates were considered trivial magnitudes of effects. For example, we estimated that 0.73% and 2.2% in erythrocyte DPAn–3 and DHA, respectively, were associated with a 10% lower risk in all-cause mortality based on a published analysis of 17 prospective cohort studies [26]. By comparison, we found sex differences in erythrocyte DPAn–3 (0.2%) and DHA (0.3%) that may be trivial; however, n–3 PUFA levels may not reflect metabolic consumption, and the biological significance remains unclear.

Our SRMA has some notable strengths. We quantified sex differences across a comprehensive set of PUFA biomarkers spanning 8 commonly measured lipid pools. We applied a conservative correction for multiple testing across main outcomes, improving the confidence in our findings. Importantly, blood and tissue PUFA levels are objective measures of exposures that are not affected by recall or social desirability biases. In addition, comparing the relative proportion of PUFAs minimizes confounding by body fat, which is typically higher in females than in males [130]. We also excluded studies that reported metabolic disorders or abnormal lipid levels, considering fatty acid composition can be influenced by levels of lipoproteins [5]. Together, with the pooled data being observational in nature, the results are representative of healthy, free-living adults.

We found clear evidence for an association between sex and PUFA status, particularly influencing plasma LA, plasma/erythrocyte DPAn–3 and DHA, and plasma CE GLA and DGLA. As direct precursors to various proinflammatory and anti-inflammatory lipid mediators, these findings point to potential sex differences in a wide variety of downstream products that require further investigation in RCTs with time-course data. For instance, DPAn–3 and DHA give rise to specialized proresolving mediators (SPMs), namely resolvins, protections, and maresins that are actively synthesized in the resolution of inflammation [2,8]. Compared with males, females demonstrate accelerated resolution of systemic inflammation and elevated levels of proresolving lipid mediators, particularly D-resolvins [131]. As new roles of n–3 and n–6 PUFA-derived lipid mediators become uncovered, a greater understanding of the factors influencing sexual dimorphism in circulating n–3 and n–6 PUFA and their related SPMs will be of increasing importance. This may lead to further insights into the sex-specificity of inflammatory diseases and their treatment by dietary intervention [[132], [133], [134]].

Critically, our SRMA highlights the need for future studies on PUFA–disease relationships to account for sex-specific variation. Reporting sex-stratified PUFA levels at baseline and after EPA and DHA supplementation is needed, given prior findings of female-specific reductions in CVD risk with n–3 PUFA supplementation [135] and in myocardial infarction risk with fish intake [136]. Lower DPAn–3 and higher DHA status in females compared with males may inform future research to investigate the sex-specific effects of varied doses and ratios of EPA and DHA supplementation on cardiovascular protection. Beyond n–3 PUFAs, LA consumption is increasing worldwide and represents the most highly consumed PUFA in Western diets [137]. As a result, LA-derived oxylipins are a major class of bioactive oxylipins in blood and tissue but are comparatively much less understood [8]. Evidence suggests that LA-derived oxylipins play roles in pain regulation, inflammation, ischemic brain injury response, and neurotransmission [8,138,139], with underlying mechanisms and the potential impact of sex variation yet to be fully understood. Overall, our study draws attention to the influence of biological sex on highly prominent n–3 and n–6 PUFAs that are important to consider in the numerous research studies investigating the impact of PUFA status on health and disease outcomes.

Author contributions

The authors’ contributions were as follows – DL, AHM: designed research and primarily responsible for final content; DL, CS, AK, TS, MRG, SBM: conducted the research; AHM, LC, JLS: provided essential materials necessary for the research; DL, CS, TAK: analyzed data or performed statistical analysis; DL, CS: wrote paper; and all authors: read and approved the final manuscript.

Data availability

Data described in the manuscript, code book, and analytic code will be made available on request.

Declaration of generative AI and AI-Assisted technologies in the writing process

The authors declare that no generative AI or AI-assisted technologies were used in the writing of this manuscript.

Funding

There was no direct funding for this work. DL was supported by the Heart & Stroke Postdoctoral Award for Women’s Heart and Brain Health and the Canadian Institutes of Health Research (CIHR). CS and MRG were supported by the Undergraduate Research Opportunity Program at the University of Toronto. AHM was supported by CIHR (PJT-183696).

Conflict of interest

The authors report no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: