A High Omega-3, Low Omega-6 Diet Reduces Headache Frequency and Intensity in Persistent Post-Traumatic Headache: A Randomized Trial

Abstract

Targeted manipulation of dietary omega-3 and omega-6 fatty acids has previously been shown to decrease non-traumatic headaches in controlled trials. This study assessed the effects of a diet high in omega-3 fatty acids and low in omega-6 linoleic acid (H3L6 diet) on headache frequency and severity, headache impact, and plasma nociceptive mediators in a persistent post-traumatic headache (pPTH) population. One hundred twenty-two participants with pPTH were randomized 1:1 to 12 weeks of either the H3L6 (n=62) or a control (n=60) diet. A priori primary endpoints were the plasma levels of the antinociceptive docosahexaenoic acid (DHA) derivative 17-hydroxy-DHA and the Headache Impact Test (HIT-6) score. Secondary endpoints included headache days/month and average daily headache pain intensity (0–10 scale). Statistical analyses followed intention-to-treat principles and were adjusted for baseline values. Relative to the control group, the H3L6 group significantly reduced headache days/month (−2.1, 95% confidence interval: −3.5 to −0.8, p=0.002) and average headache intensity (−0.9, 95% confidence interval: −1.2 to −0.5, p<0.001) and increased circulating 17-hydroxy-DHA (ng/mL) (difference 0.07, 95% confidence interval: 0.02 to 0.11, p=0.003), although it did not significantly improve HIT-6 scores (−1.6, 95% confidence interval: −4.0 to 0.8, p=0.18). In conclusion, the H3L6 diet reduced headache pain and increased antinociceptive mediators, supporting its potential as an adjunct non-pharmacologic pPTH therapy.

Introduction

Persistent post-traumatic headache (pPTH) is a frequent sequela of traumatic brain injury (TBI), that contributes significantly to long-term disability and reduced quality of life in affected individuals.^1,2 Epidemiological data indicate that pPTH develops more often after mild head trauma, with headache symptoms reported to persist beyond a year in over 40% of TBI patients.^3,4 Despite its prevalence, pPTH remains challenging to treat. Existing therapeutic strategies—chiefly pharmacological interventions aimed at symptomatic relief—often provide suboptimal or only transient benefit,⁵ underscoring the pressing need for new therapeutic options.

While the pathophysiology of pPTH is not well understood, it includes complex interactions between neuroinflammatory responses and dysregulated pain processing pathways.^6,7 Recent evidence points to the role of omega-3 (n-3) and omega-6 (n-6) polyunsaturated fatty acids (PUFAs) as endogenous modulators of inflammation and pain.^8–13 Specifically, n-6 linoleic acid (LA) is metabolized into pro-nociceptive mediators, whereas n-3 fatty acids such as eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) are metabolized into anti-nociceptive and anti-inflammatory lipid mediators including the hydroxy-DHA (HDHA) and epoxy-DHA families of oxylipins that have favorable effects on neuroinflammation and neuronal sensitization in preclinical models.^8–13 A potentially important molecule in this pathway is 17-hydroxy-DHA (17-HDHA), the 15-lipoxygenase derivative of DHA that is the precursor of several oxylipins reported to have antinociceptive effects in experimental models^14–24 and has been linked to lower pain sensitivity in humans.²⁵

We have previously shown that increasing dietary n-3 fatty acids (EPA+DHA) while reducing n-6 LA, (the H3L6 diet) can decrease non-traumatic headaches in randomized controlled trials of patients with chronic daily headache and episodic/chronic migraine.^8–12 These studies also showed that the H3L6 diet produced biochemical changes consistent with an anti-nociceptive profile, including altered n-3 and n-6 fatty acids and several of their nociceptive oxylipin derivatives in plasma, serum, erythrocytes and immune cells. However, it was unknown whether these findings would extend to pPTH.

In this 12-week randomized controlled trial (RCT), we tested the clinical and biochemical effects of the H3L6 diet versus a control diet in individuals with pPTH. The primary clinical endpoint was the change in headache-related quality of life measured by the Headache Impact Test (HIT-6). The primary biochemical endpoint was the change in plasma levels of 17-HDHA. Secondary endpoints included changes in headache days/month, headache intensity, and other DHA-derived oxylipins in plasma. We hypothesized that the H3L6 diet would improve headache outcomes and increase levels of anti-nociceptive oxylipins.

Methods

Participants were adult military healthcare beneficiaries, including active-duty personnel, veterans, and dependents under the care of a clinician for headache management. Inclusion criteria required participants to have both a TBI of any severity (diagnosed by Veterans Affairs/Department of Defense criteria^26,27 at least six months earlier), an International Classification of Headache Disorders (3rd edition) diagnosis of pPTH that included migraine features, and eight or more headaches per month for at least six months. Other inclusion criteria included the ability to communicate in English, comply with study procedures, follow assigned study diet, and maintain a headache diary. Exclusion criteria included severe food allergies, aversions to fish, recent consumption of fish oil supplement, recent injectable headache prophylactic (e.g. botulinum toxin), anticipated deployment, change of duty station during the intervention phase, or current/anticipated pregnancy. Complete inclusion and exclusion criteria are listed in the supplement (Table S1).

The trial was conducted at three military medical treatment facilities—Walter Reed National Military Medical Center (WRNMMC) in Maryland, Alexander T. Augusta Military Medical Center in Virginia, and Womack Army Medical Center in North Carolina—from August 2017 to March 2022. Written informed consent was secured from all participants prior to enrollment. All participants recorded headaches and medication use in a daily headache diary for four weeks (preintervention run-in phase). Participants who completed at least 80% of the diary entries during this time were randomly assigned 1:1 to one of two dietary interventions to be maintained for 12 weeks. Participants were advised to continue usual care throughout the trial.

Trial procedures were approved by the Institutional Review Board (IRB) of record at WRNMMC (IRB #416047–001) for each of the three enrollment sites, as well as at the data coordinating center at the University of North Carolina at Chapel Hill (UNC) (IRB #153220). The study protocol, including modifications made due to the COVID-19 pandemic (e.g., transition from in-person to virtual visits, changes in blood collection procedures) were previously published.²⁸ All protocol changes were cleared by the supervising IRB at WRNMMC.

Randomization & Blinding

Participants were randomized using a centrally administered, computer-generated permuted block design with block sizes of 2–6, stratified by clinical site to ensure balance. Allocation concealment was maintained through a secure online system managed by a data manager at UNC. Study dietitians entered participant identifications into the system, which provided the diet assignment and recorded it in a permanent, date-stamped log. The study employed a modified double-blind design, in which a dietitian and, as necessary, a research assistant at each site were aware of the diet assignments, but all other investigators, analysts, and laboratory technicians were masked to group assignment for the duration of the trial. Participants received counseling, food provision, and website access in accordance with their assigned intervention and were masked to the nature of the study diets.

Interventions

The two diets were designed to be eucaloric and equally credible and acceptable, with equivalent intensity and amounts of dietitian counseling, intervention materials, study foods and macronutrient intake. The H3L6 diet was designed to increase intake of n-3 EPA and DHA to 1.5 g/day while reducing n-6 linoleic acid intake to ≤1.8% of total energy. The control diet was designed to maintain a typical United States (U.S.) fatty acid intake, characterized by relatively low EPA+DHA intake (below 150 mg/day) and relatively high linoleic acid intake (approximately 7% of total energy). To achieve dietary intake of the targeted fatty acids, participants in the H3L6 group were instructed to consume study-provided fish daily, specifically selected for high n-3 content, while concurrently using the low n-6 study provided oil blend in place of all other oils used in the home. The control group was instructed to consume study-provided poultry and low EPA+DHA fish regularly and to use the high LA study-provided oil blend in place of all other oils used in the home. Participants in both the H3L6 and control diet groups were given study foods that provided about 75% of their daily caloric needs throughout the 12-week intervention. Intensive dietitian counseling and study food supplies were provided at randomization and at 2-week intervals throughout the 12-week intervention to all participants regardless of dietary assignment. More information on the interventions is provided in the supplement (pages 2–6; Tables S2–S4).

Headache-Related Endpoints

The a priori primary clinical endpoint was the headache impact test (HIT-6), a self-reported questionnaire measuring the “impact that headaches have on the ability to function on the job, at school, at home, and in social situations”.^29,30 A higher score indicates a greater impact of headaches on quality of life. The between-group minimally important difference in HIT-6 score was estimated at 1.5 points in a primary care population of people with migraine.³¹ This corresponds to a “somewhat better” perceived clinical improvement after three months compared with “about the same”.

Participants were asked to keep a daily electronic headache diary (paper if away from internet). They were given a template on which to enter the number of headache hours in the previous day, along with a rating of average pain intensity on a 0–10 visual analog scale (VAS) (0 = no pain and 10 = the most severe pain) and the worst daily headache intensity they experienced, also on a 0–10 VAS. A sample template is shown in the supplement on page 7. If a participant did not complete the diary for a given day, a text or email reminder was sent the following day. The number of headache days per month (number of days with at least one hour of headache) was derived from these data.

Specimen Collection, Processing, and Biochemical Analyses

Fasting blood was collected and immediately processed to aliquots of plasma and packed erythrocytes at randomization and after 6 and 12 weeks of diet exposure, as previously described.²⁸ The aliquots were prepared and stored on ice for transport to −80°C freezers. After freezing, they were transported on dry ice. Erythrocyte levels of n-3 and n-6 fatty acids and plasma free oxylipins were measured as biochemical markers for intervention effects on lipid-mediated pain pathways. Fatty acids were analyzed by gas chromatography with flame ionization detector³² at the University of California Davis West Coast Metabolomics Center. Oxylipins were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) at the intramural program of the National Institute on Aging.^33–36 We focused on the HDHA and epoxy-DHA families of oxylipins due to their potential role in reducing neuroinflammation and neuronal sensitization.

Adverse Events

Adverse events were assessed by the dietitians at each visit and investigated by the study staff, including a study clinician, as needed (for symptoms more than transient). Participants were followed until resolution. All potentially serious adverse events were reported to the principal investigator who made the determination of severity and relatedness. Adverse events were reported biannually to the UNC Data and Safety Monitoring Board using the Common Terminology Criteria for Adverse Events, Version 5 (CTCAE).³⁷

Statistical Analysis

Analyses were conducted in Stata version 18.5. Sample size calculations for this study were based on our previous trial in a chronic daily headache population.⁸ Using α = 0.05 and 1 – β = 0.90 to detect a 5-point difference in HIT-6 score between the H3L6 and control diet groups, the study aimed to randomize 120 participants. With an expected dropout rate of 15%, we expected to have at least 96 participants completing the trial (48 per group).

The prespecified primary analysis for HIT-6 and 17-HDHA was an analysis of covariance, with the week-12 values as dependent variables and diet group as an independent variable, adjusted for age, sex, botulinum toxin use, clinical site, and baseline value of the respective outcome. Longitudinal analyses using generalized estimating equation models were prespecified for assessment of between-group differences in variables from the headache diary. These models included continuous time-by-group interactions and were also adjusted for age, sex, botulinum toxin use, clinical site and baseline value of the respective outcome. Headache days/month was analyzed with population averaged (generalized estimating equation) Poisson regression with autoregressive structure for within person correlation. Headache hours/day, daily average headache intensity, and daily worst headache intensity were analyzed with population averaged negative binomial regression models to account for overdispersion in the values of the daily data.³⁸ Analyses of between group differences in HIT-6, 17-HDHA, and diary endpoints included all randomized participants unless otherwise specified. Missing data were imputed by using multiple imputation procedures. The imputation model included baseline demographic and clinical variables and used within group chained equations with a predictive mean matching algorithm generating 30 imputations per missing value. We ran analysis models for each of the 30 imputed datasets, and the results were combined using Rubin’s rules.

Statistical analyses of blood fatty acids and oxylipins (other than the primary biochemical outcome, 17-HDHA) included only participants with complete data at baseline and week 12. For participants who dropped out before week 12, data from the last follow-up visit completed (week 6) were substituted whenever available. Analysis of covariance (with variable transformations, if necessary) was used to assess between group differences at the final visit.

Results

Participant Characteristics

A total of 122 participants were randomized: 62 to the H3L6 diet and 60 to the control diet (Figure 1). Two participants (one in each group) were found to be ineligible after randomization and did not receive the intervention (undisclosed shrimp allergy and recent recurrent TBI). Ninety-nine of 122 completed the 12-week intervention: 87% of the H3L6 group, 75% of the control group. Three were unable to complete the study due to temporary study site closures during the COVID-19 pandemic and four were deployed.

Figure 1.

CONSORT Diagram.

At baseline, participants had an average of 22 headache days/month and an average daily headache intensity of 3.5. Mean HIT-6 score was 64, indicating severe impact of headaches on quality of life. Mean age was 40 years, 79% were male, and mean body mass index (BMI) was 28 (Table 1). The time since injury was variable ranging from 0.6 to 36.1 years (mean 9 years, SD 6.35). TBI classifications included 90 (74%) with mild TBI, 27 (22%) with moderate TBI, one severe TBI and four with severity indeterminant (possibly severe based on memory deficits).

Table 1.

Demographic and Clinical Characteristics (n=122)

	Control (n=60)	H3L6 (n=62)	Total
Age, y, mean (SD)	39.6 (8.0)	40.6 (9.1)	40.1 (8.6)
Body mass index, mean (SD)	28.5 (3.7)	28.3 (3.4)	28.4 (3.6)
Male, n (%)	49 (82%)	47 (76%)	96 (79%)
Race, n (%)
Black	10 (17%)	10 (16%)	20 (16%)
White	33 (55%)	40 (65%)	73 (60%)
Othera	8 (13%)	8 (13%)	16 (13%)
Unknown/Not reported	9 (15%)	4 (6%)	13 (11%)
Ethnicity, n (%)
Hispanic	12 (20%)	8 (13%)	20 (16%)
Non-Hispanic	39 (65%)	46 (74%)	85 (70%)
Unspecified	9 (15%)	8 (13%)	17 (14%)
Highest level of education completed, n (%)
≤ Some collegeb	18 (30%)	14 (23%)	32 (26%)
Associate’s/Technical	8 (13%)	15 (25%)	23 (19%)
Bachelor’s	18 (30%)	14 (23%)	32 (26%)
Post graduate	16 (27%)	18 (30%)	34 (28%)
Annual household income
< $50,000	6 (10%)	4 (6%)	10 (8%)
$50,000-$74,999	9 (15%)	9 (15%)	18 (15%)
$75,000-$99,999	10 (17%)	16 (26%)	26 (21%)
≥ $100,000	30 (50%)	26 (42%)	56 (46%)
Refused/Unknown	5 (8%)	7 (11%)	12 (10%)
HIT-6 score, mean (SD)	65.2 (6.0)	62.9 (4.6)	64.0 (5.4)
Headache hours/day, mean (SD)	6.2 (5.1)	6.0 (5.0)	6.1 (5.0)
Average headache pain (0–10), mean (SD)c	3.6 (1.7)	3.3 (1.6)	3.5 (1.6)
Worst headache pain (0–10), mean (SD)c	4.5 (1.8)	4.2 (1.7)	4.4 (1.7)
Headache days/month, mean (SD)d	22.6 (6.2)	22.1 (6.0)	22.3 (6.1)
Clinical site, n (%)
Walter Reed	14 (23%)	14 (23%)	28 (23%)
Ft. Belvoir	22 (37%)	23 (37%)	45 (37%)
Ft. Bragg/Womack	24 (40%)	25 (40%)	49 (40%)

^aMore than one race (n=8), American Indian (n=5), Asian (n=3).

^bFour subjects had a high school diploma/GED or less.

^cScale is from no distress (0) to highest possible distress (10).

^dA headache day is defined as having at least one hour of headache during the day.

Primary Endpoints

In intention-to-treat analyses (n=122, Table 2), the improvement in quality of life (HIT-6 scores) in the H3L6 group versus the control group was clinically relevant³¹ but not statistically significant (baseline-adjusted mean difference −1.6, 95% confidence interval (CI): −4.0 to 0.8, p=0.18; Cohen’s d −0.25). However, the H3L6 diet increased plasma 17-HDHA (ng/mL) compared with the control group. The baseline-adjusted mean difference was 0.07 (95% CI: 0.02 to 0.11, p=0.003).

Table 2.

Primary and secondary endpoints at week 12 (intention-to-treat analysis) ^a

		Control (n=60)	H3L6 (n=62)
Primary Outcomes
HIT-6 scoreb	Estimate (95% CI)	59.5 (57.8 to 61.2)	57.9 (56.3 to 59.4)
			Versus Control
	Difference (95% CI)		−1.6 (−4.0 to 0.8)
	p-value		0.18
	Cohen’s d		−0.25
17-HDHA (ng/mL)	Estimate (95% CI)	0.12 (0.09 to 0.15)	0.19 (0.16 to 0.22)
			Versus Control
	Difference (95% CI)		0.07 (0.02 to 0.11)
	p-value		0.003
	Cohen’s d		0.57
Secondary Clinical Outcomes
Headache hours/dayc d	Estimate (95% CI)	5.1 (4.4 to 5.9)	4.8 (4.1 to 5.4)
			Versus Control
	Difference (95% CI)		−0.4 (−1.3 to 0.6)
	p-value		0.45
	Cohen’s d		−0.13
Average headache pain (0–10)c d e	Estimate (95% CI)	3.0 (2.7 to 3.3)	2.1 (1.9 to 2.4)
			Versus Control
	Difference (95% CI)		−0.9 (−1.2 to −0.5)
	p-value		<0.001
	Cohen’s d		−0.80
Worst headache pain (0–10)c d e	Estimate (95% CI)	3.6 (3.3 to 4.0)	2.7 (2.4 to 3.0)
			Versus Control
	Difference (95% CI)		−0.9 (−1.3 to −0.5)
	p-value		<0.001
	Cohen’s d		−0.75
Headache days/monthc f g	Estimate (95% CI)	18.9 (17.8 to 19.9)	16.7 (15.7 to 17.7)
			Versus Control
	Difference (95% CI)		−2.1 (−3.5 to −0.8)
	p-value		0.002
	Cohen’s d		−0.52

^aMissing data were imputed using multiple imputation procedures.

^bRegression models were adjusted for recruitment site, sex, age, botox use, group, and baseline outcome.

^cModels were adjusted for recruitment site, sex, age, botox use, time, and group-by-time interaction.

^dPopulation-averaged model, negative binomial distribution. Within-panel correlation structure is autoregressive of order 1.

^eScale is from no distress (0) to highest possible distress (10).

^fPopulation-averaged model, Poisson distribution. Within-panel correlation structure is autoregressive of order 1.

^gA headache day is defined as having at least one hour of headache during the day.

Headache Frequency, Duration, and Intensity

In intention-to-treat analyses (Figure 2 and Table 2), the H3L6 group significantly reduced the number of headache days/month (−2.1, 95% CI: −3.5 to −0.8; Cohen’s d −0.52), the daily average headache pain intensity (−0.9, −1.2 to −0.5; Cohen’s d −0.80), and the worst daily headache pain intensity (−0.9, −1.3 to −0.5; Cohen’s d −0.75).

Figure 2.

Longitudinal Models of Headache Diary Endpoints.

^a Models were adjusted for recruitment site, sex, age, botox use, time, and group-by-time interaction. P-values are for the group-by-time interaction. Missing data were imputed using multiple imputation procedures.

^b A headache day is defined as having at least one hour of headache during the day.

^c Population-averaged model, Poisson distribution. Within-panel correlation structure is autoregressive of order 1.

^d Population-averaged model, negative binomial distribution. Within-panel correlation structure is autoregressive of order 1.

^e Scale is from no distress (0) to highest possible distress (10).

Biochemical Endpoints

Pre- and post-intervention blood samples were available for 95 participants due to pandemic-related modifications. The H3L6 intervention increased n-3 DHA levels and decreased n-6 arachidonic acid in red blood cells relative to the control group, but it did not alter levels of n-3 EPA or n-6 LA (Table 3). In addition, the H3L6 diet increased circulating levels of several DHA-derived oxylipins measured in plasma: 4-HDHA, 17-HDHA, 16,17-EpDPA, and 19,20-EpDPA (Table 4). There were no differences in oxylipin derivatives of n-6 LA or arachidonic acid, except 9,10-DiHOME (p=0.022). Note that oxylipin measures were log transformed to improve model fit.

Diet-induced changes in percent of total fatty acids (red blood cells)

		Values at Baseline and Week 12		Differences Between Groups at Week 12a
		Control (n=45)	H3L6 (n=50)
		Median (IQR)	Median (IQR)	H3L6 vs Control
Omega-3 fatty acids
Eicosapentaenoic acid	Baseline	1.8 (1.3 to 2.1)	1.6 (1.1 to 2.1)
	Week 12	1.7 (1.2 to 2.1)	1.8 (1.2 to 2.7)	Difference	0.35
	% Change	−2%	+8%	p-value	0.52
Docosapentaenoic acid	Baseline	2.2 (1.5 to 2.5)	2.2 (1.7 to 2.5)
	Week 12	2.1 (1.7 to 2.5)	2.2 (1.4 to 2.5)	Difference	−0.12
	% Change	−3%	−1%	p-value	0.43
Docosahexaenoic acid	Baseline	4.3 (3.5 to 5.5)	4.1 (3.1 to 5.1)
	Week 12	4.7 (3.4 to 5.6)	5.3 (4.4 to 7.2)	Difference	1.1
	% Change	+10%	+29%	p-value	0.005
Alpha-linolenic acid	Baseline	0.02 (0.014 to 0.039)	0.02 (0.012 to 0.031)
	Week 12	0.022 (0.014 to 0.043)	0.021 (0.015 to 0.066)	Difference	0.015
	% Change	+11%	+7%	p-value	0.16
Omega-6 fatty acids
Linoleic acid	Baseline	15 (14 to 17)	15 (13 to 17)
	Week 12	15 (13 to 17)	14 (12 to 17)	Difference	−0.38
	% Change	0%	−5%	p-value	0.61
Arachidonic acid	Baseline	18 (15 to 20)	17 (16 to 19)
	Week 12	18 (16 to 19)	16 (14 to 18)	Difference	−1.9
	% Change	+2%	−11%	p-value	0.014
Monounsaturated fatty acids
Palmitoleic acid	Baseline	1.6 (1.3 to 1.9)	1.5 (1.3 to 1.9)
	Week 12	1.5 (1.2 to 1.8)	1.5 (1.2 to 2.0)	Difference	−0.011
	% Change	−7%	−4%	p-value	0.91
Oleic acid	Baseline	15 (14 to 16)	15 (3.6 to 16)
	Week 12	15 (13 to 17)	15 (2.8 to 16)	Difference	−0.79
	% Change	−1%	+2%	p-value	0.47
Saturated fatty acids
Palmitic acid	Baseline	26 (25 to 29)	26 (25 to 29)
	Week 12	26 (25 to 27)	27 (25 to 29)	Difference	1.0
	% Change	−1%	+2%	p-value	0.10
Stearic acid	Baseline	18 (16 to 19)	18 (16 to 19)
	Week 12	18 (16 to 19)	18 (15 to 19)	Difference	1.2
	% Change	−1%	0%	p-value	0.27

^aBaseline-adjusted values are based on ANCOVA controlling for the recruitment site.

Diet-induced changes in oxylipins (ng/mL) ^a

		Values at Baseline and Week 12		Differences Between Groups at Week 12b
		Control (n=45)	H3L6 (n=45)
		Median (IQR)	Median (IQR)	H3L6 vs Control
DHA derivatives
4-HDHA	Baseline	0.12 (0.11 to 0.14)	0.13 (0.11 to 0.14)
	Week 12	0.12 (0.11 to 0.14)	0.16 (0.14 to 0.21)	Difference	0.31
	% Change	−2%	+27%	p-value	<0.001
14-HDHA	Baseline	0.039 (0.039 to 0.12)	0.039 (0.039 to 0.12)
	Week 12	0.039 (0.039 to 0.11)	0.12 (0.039 to 0.16)	Difference	0.29
	% Change	0%	+196%	p-value	0.054
17-HDHA	Baseline	0.066 (0.066 to 0.18)	0.066 (0.066 to 0.17)
	Week 12	0.066 (0.066 to 0.17)	0.19 (0.066 to 0.27)	Difference	0.56
	% Change	0%	+192%	p-value	<0.001
16,17-EpDPA	Baseline	0.024 (0.024 to 0.05)	0.024 (0.024 to 0.051)
	Week 12	0.024 (0.024 to 0.024)	0.058 (0.024 to 0.069)	Difference	0.55
	% Change	0%	+139%	p-value	<0.001
19,20-EpDPA	Baseline	0.11 (0.088 to 0.14)	0.11 (0.092 to 0.13)
	Week 12	0.1 (0.089 to 0.13)	0.15 (0.12 to 0.21)	Difference	0.39
	% Change	−7%	+38%	p-value	<0.001
AA derivatives
PGE2c	Baseline	0.025 (0.025 to 0.025)	0.025 (0.025 to 0.025)
	Week 12	0.025 (0.025 to 0.025)	0.025 (0.025 to 0.025)	Difference	−0.11
	% Change	0%	0%	p-value	0.10
TxB2	Baseline	0.025 (0.025 to 0.047)	0.025 (0.025 to 0.052)
	Week 12	0.027 (0.025 to 0.067)	0.025 (0.025 to 0.054)	Difference	−0.17
	% Change	+8%	0%	p-value	0.23
5-HETE	Baseline	0.14 (0.1 to 0.18)	0.13 (0.11 to 0.17)
	Week 12	0.13 (0.11 to 0.18)	0.14 (0.1 to 0.17)	Difference	−0.094
	% Change	−6%	+2%	p-value	0.29
12-HETE	Baseline	0.2 (0.17 to 0.25)	0.2 (0.18 to 0.25)
	Week 12	0.21 (0.18 to 0.24)	0.2 (0.17 to 0.26)	Difference	−0.096
	% Change	+3%	+1%	p-value	0.30
15-HETE	Baseline	0.26 (0.22 to 0.34)	0.27 (0.23 to 0.29)
	Week 12	0.26 (0.22 to 0.31)	0.25 (0.22 to 0.3)	Difference	−0.05
	% Change	+1%	−5%	p-value	0.37
LA derivatives
9-HODE	Baseline	1.5 (0.79 to 2.0)	1.4 (0.95 to 1.9)
	Week 12	1.4 (0.9 to 2.5)	1.2 (0.85 to 2.2)	Difference	−0.16
	% Change	−7%	−12%	p-value	0.24
13-HODE	Baseline	4.6 (3.4 to 6.5)	4.4 (3.4 to 6.4)
	Week 12	4.8 (3.7 to 7.9)	4.9 (3.3 to 6.7)	Difference	−0.12
	% Change	+4%	+11%	p-value	0.31
9,10-EpOMEc	Baseline	0.5 (0.5 to 0.5)	0.5 (0.5 to 0.5)
	Week 12	0.5 (0.5 to 0.5)	0.5 (0.5 to 0.5)	Difference	−0.021
	% Change	0%	0%	p-value	0.50
12,13-EpOME	Baseline	0.25 (0.25 to 0.76)	0.25 (0.25 to 0.78)
	Week 12	0.25 (0.25 to 0.8)	0.25 (0.25 to 0.61)	Difference	−0.063
	% Change	0%	0%	p-value	0.68
9,10-DiHOME	Baseline	0.9 (0.63 to 1.2)	0.85 (0.59 to 1.1)
	Week 12	0.99 (0.6 to 1.7)	0.83 (0.52 to 1.1)	Difference	−0.35
	% Change	+10%	−3%	p-value	0.022
12,13-DiHOME	Baseline	1.2 (0.87 to 2.0)	1.4 (1.0 to 1.9)
	Week 12	1.4 (0.92 to 1.9)	1.4 (1.0 to 1.9)	Difference	−0.11
	% Change	+12%	−1%	p-value	0.36
9H-12,13E-LA	Baseline	0.025 (0.025 to 0.067)	0.025 (0.025 to 0.071)
	Week 12	0.057 (0.025 to 0.083)	0.025 (0.025 to 0.082)	Difference	−0.17
	% Change	+126%	0%	p-value	0.24
13H-9,10E-LA	Baseline	0.025 (0.025 to 0.079)	0.025 (0.025 to 0.11)
	Week 12	0.054 (0.025 to 0.088)	0.052 (0.025 to 0.079)	Difference	−0.17
	% Change	+115%	+106%	p-value	0.25
9,10,13-TriHOMEc	Baseline	0.25 (0.25 to 0.25)	0.25 (0.25 to 0.25)
	Week 12	0.25 (0.25 to 0.25)	0.25 (0.25 to 0.25)	Difference	0.02
	% Change	0%	0%	p-value	0.31
9,12,13-TriHOMEc	Baseline	0.5 (0.5 to 0.5)	0.5 (0.5 to 0.5)
	Week 12	0.5 (0.5 to 0.5)	0.5 (0.5 to 0.5)	Difference	−0.014
	% Change	0%	0%	p-value	0.70

^aOxylipin measures that were listed as “BLQ”, “No Peak”, or “< 0” were imputed with 1/2 of the minimum value in the sample at the respective time point.

^bBaseline-adjusted values are based on ANCOVA controlling for the recruitment site. Presented in natural log units.

^cMore than half of the available subjects had no data for this oxylipin at both time points.

Adverse Events

Adverse event categories as classified in the CTCAE showed no clear pattern. The most reported adverse events classified as moderate or higher (requiring medical attention) were gastrointestinal (n=5), infections (n=5), and musculoskeletal (n=4) with no significant differences between groups (Table S5). In addition, there were no statistically significant differences in the total number of participants experiencing at least one adverse event (including mild, transient symptoms), with 22 (35.5%) participants in the intervention group and 16 (26.2%) in the control group (p=0.33) (Table S6). No life-threatening events occurred in either group. CTCAE-defined mild, moderate, and severe events occurred in 13 (21.0%), 13 (21.0%), and 1 (1.6%) participant(s) in the intervention group versus 10 (16.4%), 8 (13.1%), and 1 (1.6%) participant(s) in the control group, respectively, with all p-values ≥0.34. The two serious adverse events (one in each group) were unrelated to the interventions (i.e., a motor vehicle accident and an injury on a water slide).

Discussion

In this RCT, we investigated the clinical and biochemical effects of a 12-week H3L6 diet (versus an equally intensive control diet with a fatty acid composition that mimics the typical American diet), on headache-related outcomes in individuals with pPTH. Compared to the control diet, participants assigned to the H3L6 diet experienced a statistically significant reduction in the number of headache days/month (about two days less), along with approximately 30% reduction in the average headache pain intensity. Although the H3L6 diet decreased headache impact on quality of life (HIT-6; primary endpoint) by a clinically relevant³¹ amount (1.6 points) relative to the control diet, this difference in improvement was not statistically significant. Biochemically, the H3L6 diet increased plasma levels of 17-HDHA and other monohydroxy- and epoxy-derivatives of DHA that have anti-nociceptive properties in preclinical models,²⁴ supporting the hypothesis that modulation of lipid mediators contributes to the H3L6 diet’s antinociceptive effects.

Findings from this trial are consistent with two prior RCTs from our group that have shown benefits of increasing dietary n-3 and decreasing dietary n-6 fatty acid intake in other (non-traumatic) headache populations.^8,11 In Ramsden et al.,⁸ the H3L6 diet led to significant reductions in headache frequency and severity and pain medication use among participants with chronic daily headache compared with a diet that reduced n-6 alone (n=67). Additionally, a more recent RCT by our group (n=182) found that similar dietary alterations significantly reduced both headache frequency and severity in adults with chronic and episodic migraine, although like this study, improvements in headache-specific quality of life, as measured by the HIT-6, were not significant.¹¹ The consistent finding of significant, clinically relevant pain reduction in these headache populations highlights the potential for targeted dietary changes as an adjunct, non-pharmacologic treatment for chronic headaches. This is notable in part because the present study recruited mostly male participants rather than the primarily female cohorts in our prior RCTs. It is also notable that, in this trial, a greater effect was seen on headache pain intensity (Cohen’s d from −0.75 to −0.80) rather than headache days per month (Cohen’s d −0.52). This may be due to the tendency for pPTH to be constant as compared with migraine (e.g.,³⁹).

Diet-induced increases in DHA and its HDHA and epoxy-DHA derivatives as well as decreases in arachidonic acid may have contributed to the observed pain relief in the H3L6 group. Circulating 4-HDHA has been previously linked to the resolution of severe headache pain following mild TBI.³² 4-HDHA has also previously been shown to decrease angiogenesis via activation of PPARy,⁴⁰ and both angiogenesis and decreases in PPARy have been implicated in headache and migraine.^41,42 DHA-derived monoepoxides have been shown to selectively modulate nociceptive signaling in preclinical pain models.²⁴ 17-HDHA and several of its downstream derivatives also exhibit antinociceptive effects in preclinical pain models (reviewed in^11,43). Arachidonic acid is the precursor to prostaglandins and other oxylipins that are linked to inflammation, vasodilation, and pain in migraine and other pain conditions.^44,45 Future mechanistic studies are needed to determine whether diet-induced alterations in circulating pools of DHA, DHA-derived oxylipins, and arachidonic acid are accompanied by changes in tissues that are more directly involved in headache pathogenesis.

Strengths and Limitations

Strengths of our study include its randomized controlled design, use of a well-defined pPTH population, detailed biochemical assessments, and relatively high retention rates despite challenges such as the COVID-19 pandemic and military deployments. However, several limitations warrant consideration. First, the choice of HIT-6 as the primary outcome rather than a specific pain measure (e.g. headache days/month) is a limitation of this study as quality of life was only expected to improve secondary to pain reduction, but not as a direct effect of the H3L6 intervention. Notably, the recently published International Headache Society guidelines for pPTH RCTs recommend the use of headache days/month as the primary endpoint when assessing drug efficacy.⁴⁶ Second, while the 12-week intervention period aligns with the standard duration in many pharmacological and dietary trials for headache disorders, it may not have been long enough to capture the full extent of dietary influences on headache patterns or their impact on quality of life. Future trials of longer duration could potentially yield greater or more sustained effects. Moreover, while the study team implemented major efforts at each site to overcome challenges related to the COVID-19 pandemic to complete the trial, the necessary modifications reduced the intensity of the intervention (e.g., less in-person contact), potentially reducing the adherence to the diets. In addition, changes to blood collection procedures necessitated by the pandemic resulted in greater variability in blood processing and fewer biochemical data points, thereby potentially limiting the power and interpretation of the biochemical aspects of the trial. However, blood assessments suggest that most participants were able to follow the diet, at least to a degree sufficient to produce measurable biochemical changes (Tables 3–4). Protocol changes due to the pandemic are described in our protocol paper.²⁸

Clinical Implications and Future Directions

Our mechanism-based approach addresses a modifiable risk factor for pain—high n-6 and low n-3 fatty acid dietary intake. The findings from this trial, alongside our randomized trial evaluating the H3L6 diet in migraine, support recommendations for targeted dietary modification of n-3 and n-6 fatty acids as an adjunct clinical strategy for the management of chronic headache disorders, particularly since a dietary intervention is not only low risk, but it may also provide additional health benefits that were not assessed in this study. Four areas of research are needed to ensure that this approach is both effective and practical. First, further mechanistic studies should explore how specific dietary modifications influence headache pathophysiology. This could involve using animal models to test mechanistic hypotheses in affected tissues, as well as human studies leveraging more detailed biochemical analyses and advanced imaging or physiological techniques to directly assess changes in brain inflammation and neuronal activity in response to diet.⁶ In addition, it will be important to investigate the heterogeneity of responses to the intervention, potentially related to genetic polymorphisms or differences in effects on the gut microbiome. Second, dietary supplement studies are needed. Although a dietary intervention component is essential to lower n-6 intakes, replacing some or all the fish intake with high-quality n-3 supplements may make the diet easier to follow and less costly. However, it will be important to assess if a dietary supplement product has the same effect on anti-nociceptive oxylipins, as was documented in our trials based on n-3 food intakes, before this approach can be recommended. In addition, the safest and most effective product formulation for n-3 EPA and DHA dosage is unclear. Prior studies of n-3 supplementation on pain outcomes have used a variety of doses and have shown inconsistent results.^47–49 Third, pragmatic clinical trials are needed to increase the accessibility of the H3L6 intervention. This includes developing feasible and cost-effective strategies for patient education and assessing long-term adherence in diverse patient populations. Such research will also help clarify whether modifications or personalized adaptations to the H3L6 regimen are necessary to optimize outcomes for specific subgroups. Finally, future implementation research could involve assessing organizational readiness or capacity building for future implementation of system interventions for dietary changes and establishing practical support systems within healthcare settings. Collectively, these avenues of inquiry hold promise for refining dietary recommendations, elucidating the pathophysiological underpinnings of pPTH, and expanding the repertoire of effective, non-pharmacological interventions available to pPTH patients

Conclusion

The results of this RCT in a pPTH population add to growing evidence that the H3L6 diet can favorably alter lipid mediators of pain and reduce headache frequency and pain intensity in individuals with chronic headaches. While HIT-6 improvements did not reach statistical significance, the clinical and biochemical changes observed highlight the potential of this non-pharmacological, mechanism-based intervention to help in the treatment of pPTH. These findings represent an important step forward in promoting relief in individuals with pPTH, a complex condition for which few treatments have proven successful. Further research is warranted to optimize the dietary interventions for a broader population and to better understand the mechanistic pathways through which the H3L6 diet exerts its analgesic effects.

Transparency, Rigor, and Reproducibility

The study design and analysis plan were preregistered on August 21, 2017, at ClinicalTrials.gov (NCT03272399). The prespecified sample size was 120, (60 per group) with 96–99 subjects who complete the trial, yielding statistical power of (80%) for detection of a 3.6-point difference (SD 7.06) between groups in the primary clinical outcome measure, the Headache Impact Test and greater than 85% for the primary biochemical outcome measure, 17-hydoxy docosahexaenoic acid. All subjects were assigned to the H3L6 or control diet using a random number generator with randomized blocks and stratification by site, yielding groups that did not differ meaningfully in baseline characteristics. In the trial, 122 subjects were randomized and primary outcomes were assessed in 122 subjects after no deaths and 21 incomplete assessments (multiple imputations employed) for clinical outcomes. For the biochemical outcomes, 122 subjects were randomized and 90 outcomes were assessed for subjects with complete data. All primary outcomes were assessed by investigators blinded to group assignment and could guess the group assignment with accuracy no greater than chance. The fidelity of dietary counseling was confirmed using reviews and discussion of counseling recordings. The primary clinical outcome measure is a standard in the field, has a test-retest reliability of 77%,³⁰ and is distributed normally. The findings have not yet been replicated or externally validated. Data and analytic code will be available upon request if approved by the human research ethics agency, and the manuscript is open access.