Objectively and Subjectively Measured Effects of a Palmitoleate Dietary Supplement: A Placebo-Controlled Double-Blind Randomized Crossover Clinical Trial

Background: Supplementation with dietary palmitoleate was reported to lower serum high-sensitivity C-reactive protein (hsCRP) for those with elevated hsCRP. The objective of the study was to evaluate subjectively and objectively measured outcomes of an omega-7 fatty acid (n7FA) dietary supplement on people with non-specic chronic musculoskeletal discomfort. Methods: A double-blind, placebo-controlled, 1:1 randomized single crossover trial was composed of Group “A” which was given placebo rst and verum second while group “B” was given verum rst and placebo second. Volunteers with uncomplicated chronic musculoskeletal discomfort or a baseline serum hsCRP >1.0 mg/L were recruited from 2017 – 2019. Verum capsules contained palmitoleate and the placebo contained medium chain fatty acids (MCT). Participants took two capsules per day from the rst bottle for three weeks followed by two capsules per day from the second bottle without washout for an additional three weeks. Subjects were instructed to maintain their current diet and lifestyle. Results: From 211 volunteers consented and screened, 56 subjects were randomized. Six subjects dropped out, and fty subjects contributed to the statistical analyses. At baseline, none of the investigated biomarkers were signicantly correlated to subjectively assessed musculoskeletal discomfort levels, and Group B showed higher HDL and lower triglyceride than Group A. Main Outcomes of crossover analysis found no statistical signicance in all measurements. The mean difference of subjectively measured PROMIS® scores between the rst and second visits showed improvement for both placebo and verum groups. Conclusions: Supplementation for three weeks with 688mg per day of a palmiteolate dietary supplement did not reduce serum inammatory biomarkers nor did it improve subjectively measured quality of life (QoL) for those with chronic non-specic musculoskeletal discomfort. Future studies should explore time, dosing and verication of cis-palmiteolate versus trans-palmiteolate. Future studies should also consider selection of patients with chronic inammation, for example individuals with higher body-mass index (BMI).


Introduction
In the U.S., cytotoxic or noble agents must be tested for therapeutic use by well-designed and phased trials as speci ed by the FDA, which oversees safety, tolerability and e cacy requirements 1 . On the other hand, dietary supplements and other well-known essential nutrients are not required by law to follow the FDA phased trial guidelines and are thus not allowed to advertise any intention to diagnose, treat, cure, prevent or mitigate a clinical condition. Dietary supplements must be advertised the above limitation so that they do not violate FTC regulations. Because of this legal ability to avoid the normal process of clinical evaluation and validation, establishing a solid and compelling evidence-informed or evidence-based medicine (EBM) for dietary supplements and other lifestyle/dietary modalities has been challenging.
In this article, in order to clarify what is referred to as complementary and integrative health (CIH) versus complementary and alternative medicine (CAM), we propose that CAM focuses on modalities, whereas CIH refers to a setting or environment that may provide a combination of modalities but also embodies diverse alternative health philosophies and principles. In many cases, CAM trials are conducted with the hope of validating traditional and anecdotal health uses or to validate the health claims of a product. However, scienti c rigor in many of these studies has often been lacking for a variety of reasons, including: 1) lack of nancial support 2 , 2) lack of both research infrastructure and competent researchers at CIH institutions 3 , and 3) in uence exerted by interest groups 4 . Current research best practices often includes outcome measurements that are subjective (e.g., Patient Reported Outcomes) in order to re ect the values and preferences of patients 5 , and the inclusion of both subjective and objective measurements may enhance scienti c rigor when moving toward a fully EBM model of research on CAM.
A double-blind, placebo-controlled, 1:1 randomized single crossover (3 weeks x 2) trial of an omega-7 fatty acid (n7FA) supplement containing palmitoleate was conducted from 2017-2019. The investigational supplement was provided by Barlean's Organic Oils, LLC (Ferndale, WA), which was also a sponsor of this trial, together with Bastyr University Research Institute. The purpose of the trial was to test the hypothesis that supplementation with a n7FA-containing oil alleviates musculoskeletal discomfort and reduces serum in ammatory markers. The purported active constituent of the verum was palmitoleate, which is a 16-carbon monounsaturated (16:1) fatty acid with a double bond occurring at the 7th carbon atom away from the methyl group. There is no Recommended Daily/Dietary Allowance (RDA) for n7FA, which is not an essential fatty acid, meaning the body can produce it de novo 6 . The aim of this study was to enroll su cient numbers of eligible volunteers to reach 50 participants who would provide data on three time points. This clinical trial was registered on ClinicalTrials.gov (NCT03669575) and conducted in accordance with the International Council of Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) and Good Clinical Practice (GCP). The protocol was approved by the Bastyr University Institutional Review Board (IRB#16-1575).

Administrative considerations
Because the sponsor company is a longtime supporter of the university, the O ce of Research Integrity (ORI) required that the research team to maintain a strict and conscientious boundary between the sponsor organization and the research team. Our ORI is the administrative o ce for the IRB and oversees the scienti c and ethical integrity of research, including managing any potential con ict of interest. The a priori protocol and double blinding were strictly maintained, and all acquired data per protocol were used for the analysis. All analyses, including tallying adverse events, dropouts and crossover analyses were performed while a statistician was given the A or B designations only. Outliers were detected by the Grubb's Extreme Studentized Deviate (ESD) test. The code was broken in a group meeting after all analyses were completed. The trial took place at the Clinical Research Center at Bastyr University in Kenmore, Washington.

Page 4/19
The investigational products were transported to the trial site with dosing and storage instructions. Each bottle was stamped with a unique lot number, which distinguished a placebo or verum bottle identi able by a single designated researcher. All staff including the PI were blinded to the bottle identity. Fifty-six bottles of the verum and 56 bottles of placebo (112 bottles of study product) were prepared. Fourteen blocks of four (2 verum and 2 placebo) were randomized as A or B, assuring an equal number of participants assigned to both groups for every four subjects enrolled 7 . Each participant received one of the verum and one of the placebo bottles based on the sequence de ned by A or B designation.

Verum and placebo
Per the certi cate of analysis, the verum capsule contained 413 mg omega-7 Fish Oil Blend (Barlean's) derived from Alaskan Pollock (Gadus Chalcogrammus) consisting of 227 mg palmitoleate per capsule. The rancidity of the oil was measured by the level of peroxide and p-Anisidine, and indicated as 1.16 mEq/kg and 3.91 mEq/kg, respectively. Peroxide values (PVs) under 10-20 mEq/kg of sh oil are considered very fresh and unlikely to exhibit rancid smells 8 . Hung and Slinger (1980) have indicated that for salmon oil, PVs around 26 mEq/kg is 'moderately oxidized,' 120 mEq/kg is 'highly oxidized,' and 314 mEq/kg is 'extremely oxidized' 9 . Placebo capsules contained the equivalent quantity of fatty acid in the form of medium chain triglycerides (MCT) consisting of 65% caprylic acid (8:0), 45% capric acid (10:0), and < 2% lauric acids (12:0) per the company's information. The investigational products were stored at room temperature in a locked closet.

Outcome measurements
To assure su cient statistical power for the primary serum biomarkers of in ammation, we modeled the previously studied biomarker used for n7FA, which was high-sensitivity C-reactive protein (hsCRP). Bernstein et al. (2014) used a pill containing "puri ed" 52.5% cis-palmitoleate, 220.5 mg from "double metabolically distilling anchovy oil after the removal of omega-3s" and used a dose of one pill per day for 30 days 10 .
According to the reported intervention effect estimate of -1.9 mg/L and a conservative estimate of a standard deviation of 0.2 mg/L between verum and placebo, only four subjects would be needed to reach 95% power at the alpha level of 0.05 11 . A recruitment size of 56 was chosen for secondary biomarker outcomes, which included Tumor Necrosis Factor alpha (TNFα) and interleukin-6 (IL-6) from the published literature 12− 14 . Using a crossover design, 28 participants received the verum rst and another group of 28 participants received the placebo rst, and this data would then be analyzed by a mixed pre-post t-test. Blood was collected from the antecubital vein. Several factors that deviated from Bernstein's study were: 1) our study did not recruit subjects with a serum hsCRP range of 2-5 mg/L, 2) the dosage was tripled, 3) the length of study was 21 days, and 4) our study did not con rm the cis-isoform of n7FA. In addition to the objective outcomes, three subjective measurements were selected from the questionnaire-bank of the REDCap database system 15 , which were: 1) fatigue, 2) pain interference, and 3) physical function domains from the Patient-Reported Outcomes Measurement Information System (PROMIS®) 16 . These instruments were lled out at the baseline, the rst follow-up and the last visit.
Trial design, procedures, and sequence of events The trial was characterized by the following: 1) receiving voluntary contact, 2) consenting, 3) the baseline assessment, 4) no run-in re-assessment, 5) taking two gel capsules a day from the rst assigned bottle for three weeks, 6) returning for the rst follow up assessment, 7) no washout period, 8) taking two gel capsules a day from the second assigned bottle for three weeks, 9) returning for the last assessment.

Crossover design and statistics
The crossover design aims to eliminate between-subject variability of pre-post as well as verum-placebo effects; however, the level of sensitivity and the power of statistical calculation are determined by three different assumptions between the verum and placebo responses: 1) having a positive correlation, 2) having no correlation, or 3) having a negative correlation 17 . A positive correlation between the verum and the placebo means that if you have a large placebo effect, your verum effect will also be greater. A negative correlation between the verum and the placebo means that a high placebo responder will have a smaller effect for the verum. In this study, no correlation was assumed.

Results
Of the 211 subjects consented and screened, 56 subjects were randomized. Five subjects dropped out during the verum dosing and one during the placebo dosing, which resulted in 50 subjects as shown in Fig. 1. At the baseline, twelve subjects (24%) were using some sort of dietary supplement such as a multivitamin, sh oil, turmeric; alone or in combination with other pharmaceuticals. Non-steroidal anti-in ammatory drug use was 16%; prescriptions of muscle relaxants or non-opioid and non-barbiturate pain medications was 12%; and 8% of enrollees were using other pharmaceuticals such as anti-hypertensives or antidepressants. Overall, the participants were generally healthy adults with chronic musculoskeletal discomfort.
At the baseline, none of the plasma biomarkers, including hsCRP, was statistically signi cantly correlated with the musculoskeletal discomfort assessment. Hsu et al. (2019) also reported that the level of serum hsCRP was not statistically different for chronic kidney disease patients with or without musculoskeletal pain 18 . Demographic information is summarized in Table 1.

Crossover analysis
For the placebo effect, the baseline value was subtracted from the followup_1 value for Group A (placeboverum sequence) and the followup_1 value was subtracted from the followup_2 value for Group B (verum -placebo sequence). For the verum effect, the followup_1 value was subtracted from followup_2 value for Group A and the baseline value was subtracted from followup_1 value for Group B. The aggregate average of the difference within each subject (n = 50) were then compared between the placebo and verum by multiple t-tests. One outlier was identi ed by the Grubb's ESD test and excluded in series from the aggregate calculations. Using the Holm-Sidak multiple-comparison adjustment, none of the factors was statistically signi cantly different between the verum and placebo. (Fig. 2) The hsCRP was analyzed by LabCorp, Inc., and cytokines were analyzed by Eve Technologies Corporation, Ontario, Canada.

Fatty acid analysis
The placebo capsule, verum capsule and plasma were analyzed for 13 free fatty acids (Lipid Analytical Laboratories, Ontario, Canada). The frozen plasma specimens of seven subjects who were higher than 90% compliant by pill count were selected for fatty acid analysis to investigate whether or not an elevated quantity of n7FA was detectable when verum was dosed compared to placebo. Bradbury et al. (2011) suggested that the 16:1 fatty acid detected would be 16:1n7 (with no 16:1n9); and for human serum/plasma lipids, the 16:1n9 would be undetectable or only trace amounts detected 19 .
However, the desaturation ratio or other free fatty acids concentrations alone did not distinguish plasma from baseline, verum or placebo. For instance, we expected an elevation of medium chain fatty acids in plasma of placebo-dosed participants; however, uctuation in the values was negligible. We were unable to nd a particular pattern in the fatty acids pro le by dosing. The time of the last dosage was not documented. Because the visit appointment was usually in the morning, the last dosage was likely the night before or 10-12 hours prior to the blood draw. Two additional ndings were: 1) contrary to the certi cate of analysis provided (227 mg/capsule 16:1n7FA), our third-party independent analysis found the quantity of 16:1n7FA to be 344 mg per capsule, and 2) the verum also contained 199 mg per capsule of palmitate.
Therefore, the dosage used in our verum would be 688 mg of 16:1n7FA and 398 mg of palmitate per day.

Post-hoc analyses
For the rst post-hoc analysis, Fig. 3 indicates the baseline, followup_1, and followup_2 measurements of Group A (n = 26) and B (n = 24) for three objective and three subjective measurements. Three biomarkers demonstrated no pre-post signi cance for verum and placebo, whereas subjective measurement showed statistically signi cant reductions in Fatigue and Pain Interference measurements between the baseline and followup_1 for the placebo only; however, this reduction did not continue to the third measurement (followup_2).

Effect size considerations for future cross-over studies
The second analysis focused on the estimation of effect size or power analysis for available biological measurements purportedly modulated by n7FA dietary supplementation. The p-values in Table 2 show the results of unadjusted one sample t-tests using zero as the reference value, and thus indicating an upward or downward trend for the placebo or verum supplementation.
Criteria for considering viable biomarkers are: 1) the coe cient of variation (CV) of the placebo and verum is approximately equal (< 20% difference), 2) CV is also small (< 1000%), and 3) past literature articles linking the biomarker with omega-7 fatty acid levels. Because this analysis was post-hoc, no threshold value for the CV was set; however, in Table 2, three biomarkers were identi ed as a possibility: HDL, MG-CSF, and IL-8. The difference variability between placebo and verum for total cholesterol is also small; however, the CVs are large and the directional change for the mean of placebo and verum are the same. Mozaffarian, et al. (2010) reported that higher serum palmitoleate concentration was independently associated with higher HDL in 3,630 US men and women in the Cardiovascular Health Study. 21  in ammatory fatty acids such as oleic and palmitic acids 27 , or proposing the modulation of SCD1 28 . Our study did not show a decreased trend for cytokines TNFα or IL-6.
As for the limitations of this study, we did not exactly replicate the Bernstein study (2014). A different n7FA product and a different baseline hsCRP range may account for the observed differences between the two studies; however, we also noted that in the Bernstein study, the effect estimate of -1.9 observed for n7FA versus placebo came to an effect size of 9.5 11 . This value is actually considerably larger than many pharmaceutical products, which seems extraordinary for a natural product. It is a well-described phenomenon that most natural products clinical trials have failed to show signi cant effect sizes beyond placebo 29 . Given the fact that we were unable to see any difference between verum and placebo with n7FA supplementation, it would be of value for other research groups to repeat the Bernstein study as precisely as possible, given that these types of studies are used as non-FDA approved evidence of safety, tolerability and e cacy.
This study strongly suggests that researchers should pay attention to fatty acids that are absorbed into the body quickly. Neither the orally taken MCT fatty acids nor the n7FA appeared in the plasma at detectable levels after 12 hours. Bioavailability data on the essential fatty acid (n3FA) showed a maximum absorption peak at 7-10 hours after oral ingestion 30 . In the case of non-essential FAs, biofeedback mechanisms may play a role in controlling the body's FA concentration even within three weeks. A comparison study of a longterm low dose versus a short-term high dose non-essential FAs may provide additional information. For bioavailability detection, the analysis of red blood cell membranes would be advised for future studies.
This study also demonstrated a signi cant placebo effect in the subjective assessments. CAM clinical trials, including vitamins, minerals and other dietary supplements, are likely to have low effect sizes because the underlying purpose of using CAM is most often to assist the inherent healing process of the individual and this may be easily confounded by the placebo effect. From the results of this study, we suggest that scienti cally validated biological surrogate measurements should be included in all CAM clinical trials.
Placebo effects are often called nonspeci c biological responses (NBR) because most scientists acknowledge the placebo effect is real but inconsistent and uncontrollable. Using a particular CAM modality may make the NBR more speci c and predictable. We believe that CAM clinical trials can and should be designed to collect a large dataset that investigates the threshold of these speci c and predictable placebo effects associated with speci c modalities as a part of EBM in CAM.
When designing our study, pure 16:1n7FA was both cost-prohibitive and not safe for use as a control because the LD 50 has not been established 31 . Dietary supplement companies regularly refer to their product as 'cis-palmitoleate,' perhaps because partially hydrogenated fatty acids or processed trans-isoforms have been shown to be harmful 32 . Traditionally, researchers have assumed that the endogenously produced or naturally occurring palmitoleate is the cis-isoform; however, trans-palmitoleate can be naturally produced from enzymatic shortening of dietary vaccenic acid in vivo 33 as well as in dairy products, and these forms of trans-palmiteolate act like lipokine hormones 34,35 . A healthy French-Canadian population study found that high fat dairy intake was positively correlated with plasma trans-palmitoleate levels (r = 0.15; p = 0.03), and total dairy consumption was positively correlated with plasma CRP (r = 0.15; p = 0.03) 36 . Other large crosssectional dietary studies report the association of plasma trans-palmitoleate with: 1) high intake of whole-fat dairy products 37 ; 2) in ammatory biomarkers in an elderly Swedish population 13 ; and 3) other in ammatory markers such as hsCRP, RANTES, IL-1Ra, Interferon gamma, IL-10, PDGF-bb 38 . Therefore, veri cation of the cisversus the transform of palmitoleate used in any future clinical trials on n7FA may prove essential.
From the viewpoint of clinicians, this study may be criticized for a lack of clinical insight because subjects were selected by voluntary participation rather than active selection by clinicians who might identify patients who would bene t from supplementing with n7FA in order to alleviate musculoskeletal discomfort. The authors acknowledge this limitation; however, volunteerism and transparency are the nature of human subjects research. As n7FA supplements are readily available and circulating without prescription or FDA approval, typical consumers are not clinically trained but self-dosing. Thus, this study focused on the typical consumer population rather than a clinically handpicked population.

Conclusions
Data presented in this article does not support one of the benchmark clinical trials published on n7FA dietary supplementation to lower serum hsCRP. Many questions remain about why we did not observe a lowering of the primary outcome measure for this study, hsCRP, and why we did not observe lowering of other in ammatory markers such as TNFα or IL-6. It may indeed be the case that in individuals with higher hsCRP levels (hsCRP > 3 mg/L) that is due to higher BMI 39,40 , supplementation with cis-palmiteolate would lower their hsCRP levels, along with lowering subjectively assessed musculoskeletal pain; however, in our subanalysis, subjects with higher hsCRP levels showed no improvement with 3 weeks of supplementation with 688 mg of palmiteolate, nor did they show an improvement in subjectively assessed musculoskeletal discomfort beyond placebo. Future studies should consider identifying individuals with higher hsCRP due to high BMI, the length of dosing with n7FA, and prior determination of what percent of the palmiteolate fatty acid dietary supplement is in the cis-isoform. Our study suggests that for the general population with a range of hsCRP levels and subjectively assessed musculoskeletal discomfort, there is no bene t to n7FA supplementation. PSA and MS contributed to data analysis, drafting or revising the article, gave nal approval of the version to be published. MJB and MS participated in o ce visits and data collection. All authors agree to be accountable for all aspects of the work. Table 3: Objective biomarker measurements in this study are listed in the left column. N = number of data points, p-value = unadjusted one sample t-test, ES = effect size for one sample t-test, CV = coe cient of variation, P -V= the difference in ES between placebo and verum. ES calculation for viable factors was determined by a: 1) smaller CV, 2) smaller P -V, and 3) supportive of a link between palmitoleate in published literature. Calculated ES for the palmitoleate dosed in this study for HDL was 0.21, requiring a sample size of 147 in order to achieve a statistical power of 0.80 at alpha level 0.05. Likewise, 184 and 614 sample sizes are needed for GM-CSF and IL-8. Figure 1 With the single crossover design, subjects who participated in both verum and placebo measurements were included in the analysis per protocol. * One hsCRP data point was detected as an outlier by the Grubb's HSD at the alpha=0.05. This subject reported a u episode on the date when the outlier was detected. This subject was not excluded; however, for aggregate statistics, the data points (follow-up 2) of this subject were excluded in series.

Figure 2
Fifty data points were provided from 26 subjects who were assigned to taking the placebo-verum sequence and 24 subjects who were assigned to the verum-placebo sequence. A mixed t-test or split-plot analysis showed no signi cant difference between the verum and placebo measurements in both biomarkers and subjective assessments. One data point in placebo measurement was found an outlier by Grubb's ESD test at alpha level of 0.05. The multiple t-tests were adjusted by the Holm-Sidak adjustment. PainInt = pain interference. PhyFunc = physical functioning. P = placebo. V = verum. Subjective measures demonstrated a reduction in three PROMIS® measures at the time of entry into the study to the rst follow up visit; however, the reduction did not continue for either the placebo or the verum