Dose–Response Effects of Cottonseed Oil on Blood Lipid Responses in Adults at Risk of Cardiovascular Disease: A Randomized Controlled Trial

Abstract

Background

Cottonseed oil (CSO) has displayed beneficial effects regarding blood lipids, albeit in relatively high doses.

Objectives

The aim was to examine the effects of 3 doses of CSO compared with a control (CON) diet on blood lipids in adults at risk of cardiovascular disease (CVD).

Methods

Ninety-one adults with obesity and/or hypercholesterolemia completed this randomized, parallel clinical trial with a 4-wk partial outpatient feeding intervention. Participants were randomly assigned into 1 of 4 groups: intervention group received 10% estimated energy needs (EEN) from cottonseed oil (LOW) (n = 23), intervention group received 20% EEN from cottonseed oil (MID) (n = 23), or intervention group received 30% EEN from cottonseed oil (HIGH) (n = 22) providing 10%, 20%, or 30% of EEN from CSO, respectively, or control (CON) (n = 23), receiving 10% of EEN from a combination of oils. Fasting blood lipids were obtained at preintervention and postintervention visits, followed by a high-fat meal challenge (35% EEN) with 5-h postprandial blood draws to evaluate lipid metabolism.

Results

Compared with the control diet, total cholesterol decreased in LOW, MID, and HIGH, respectively (–10.2 mg/dL; –12.5 mg/dL; –13.9 mg/dL, respectively; P < 0.05 for all). Similarly, non–high-density lipoprotein cholesterol decreased in LOW, MID, and HIGH (–10.2 mg/dL; –15.8 mg/dL; –16.5, respectively; P < 0.05 for all) compared with CON. Low-density lipoprotein cholesterol reduced in MID and HIGH (–11.7 mg/dL and –12.9 mg/dL, respectively; P < 0.05 for both) with a trend for reduction in LOW (–6.8 mg/dL; P = 0.08) compared with no change in CON. Postprandial triglycerides were reduced in HIGH (–14.7 mg/dL, P = 0.02) compared with LOW, MID, and CON.

Conclusions

Enriching diets with 10%, 20%, and 30% CSO improved fasting blood lipids in CVD at-risk adults, with the best results in the 20% and 30% doses.

This trial was registered at clinicaltrials.gov as NCT05686954.

Introduction

Cardiovascular disease (CVD) is the leading cause of death globally, accounting for nearly 18 million deaths annually [1]. A primary inducing condition in the development of CVD is atherosclerosis. Two commonly identified risk factors in the development of atherosclerosis are elevated blood lipids [2] and obesity [3]. Common therapeutic approaches for reducing CVD risk involve targeted modulation of blood lipids through manipulations to dietary nutrient intake and/or overall dietary patterns. Dietary fatty acid composition has been shown to influence fasting and postprandial blood lipids [[4], [5], [6]]. For example, an effective strategy in mitigating hypercholesterolemia is the replacement of SFA with unsaturated fatty acids [6,7] because high intake of some SFAs has been attributed to increased CVD risk [8]. Nevertheless, the intake of unsaturated fats, specifically PUFAs, is known to demonstrate favorable responses concerning blood lipids and overall CVD risk [6,9]. Furthermore, the omega-6 (n-6) PUFA, linoleic acid (LA; 18:2n–6), uniquely influences gene expression in hepatic lipid and cholesterol mechanisms to lower plasma total cholesterol (TC) and LDL cholesterol concentrations [10].

Cottonseed oil (CSO) is a regularly used vegetable oil in the United States, with 160,000 tons consumed per year [11]. Common food sources include mixed oil blends, salad dressings and other condiments, snack foods, and its use as a frying oil. Despite its prevalence in the food system, the body of literature on its effects on human health remains slight. CSO is rich in PUFAs, especially LA, and has exemplified therapeutic effects regarding improvements in blood lipids in both healthy adults [12,13] and adults with hypercholesterolemia [14]. CSO is also unique from other high LA oils as it contains a bioactive lipid, dihydrosterculic acid (DHSA), a known inhibitor of the lipogenic enzyme stearoyl-CoA desaturase-1 (SCD1) [15]. Therefore, CSO has a unique cardioprotective makeup by preventing lipid accumulation via its high PUFA/LA content’s ability to suppress lipogenic transcription factors [10], and the presence of DHSA that channels SFA utilization into free fatty acid (FFA) oxidation to prevent storage/synthesis of fatty acids [15]. There may also be other bioactive components of CSO that could contribute to the cholesterol-lowering effect, including tocopherols and β-sitosterol [16]. Importantly, though, the previously observed benefits were achieved with relatively high doses of CSO, which resulted in a high dietary fat intake [>35% estimated energy needs (EEN)]. Specifically, a high-fat (HF) diet (50% EEN) enriched with 44% of fat from CSO resulted in reductions in fasting TC, LDL cholesterol, and fasting and postprandial triglycerides (TG) in healthy young males [12]. These results were complemented by an 8-wk trial in hypercholesterolemic adults in which participants had improvements in fasting TC, LDL cholesterol, HDL cholesterol, TC:HDL cholesterol ratio, and non-HDL cholesterol after a partial outpatient feeding trial providing an HF diet (49% EEN) that included a high dose of CSO (30% EEN) [14].

Despite evidence demonstrating beneficial effects of n-6 PUFA intake on cardiometabolic outcomes, there is a lack of research on the dose–response effects of differing amounts of PUFAs in the diet. The purpose of this study was to determine the impact of different doses of CSO compared with a control (CON) diet over a 4-wk period on fasting blood lipids and postprandial lipid metabolism after a high SFA meal challenge in adults at risk of CVD. This was completed by enriching diets with 10%, 20%, or 30% of EEN with CSO; the CON group received 10% of their EEN from a mixture of oils that matched the fatty acid composition of the typical American’s diet. We hypothesized that, in a dose-dependent manner, the CSO-enriched diets would improve fasting and postprandial blood lipids (decreases in blood lipid measures except for increases in HDL cholesterol), and all would be different than CON.

Methods

Study design

This was a single-blind, parallel randomized control trial (clinicaltrials.gov: NCT05686954). We recruited adults with overweight/obesity and/or hypercholesterolemia. There were 4 groups in which participants were randomly assigned to: the intervention group received 10% EEN from CSO (LOW) group received 10%, the intervention group received 20% EEN from CSO (MID) group received 20%, and the intervention group received 30% EEN from CSO (HIGH) group received 30% of their EEN from CSO. The fourth group served as a CON group in which participants received 10% of their EEN from a mixture of oils designed to match the fatty acid profile of the typical American diet [17]. An individual not involved in the data collection or analysis created the randomization. Participants were randomly assigned into 1 of the 4 groups, with an allocation ratio of 1:1:1:1, using a balanced block randomization list stratified based on the combination of both LDL cholesterol [LDL was categorized into 4 strata: optimal (<100 mg/dL); above optimal (100–129 mg/dL); high (130–189 mg/dL); very high (≥190 mg/dL)] and BMI [BMI was categorized into 5 strata: healthy (18.5–24.9 kg/m²); overweight (25–29.9); obese class I (30–34.9); obese class II (35–39.9); obese class III (≥40)]. Participants were blinded to their group assignment. Data collection occurred from March 2023 to May 2024, when the goal of ≥20 participants per group was achieved. This study consisted of a 28-d intervention with 6 visits: 1 screening visit (v0), a preintervention visit (v1), short weekly visits during the intervention (end of weeks 1, 2, and 3; v2–4), and a postintervention visit (end of week 4; v5) (Supplemental Figure 1). The primary outcomes included fasting blood lipids (TC, HDL cholesterol, TG, and LDL cholesterol). Other primary outcomes of interest included postprandial measures of TG and nonesterified fatty acids (NEFAs). These were included with the fasting cholesterol profiles to provide a complete picture of lipid metabolism. All primary outcomes were measured at v1 and v5. Other exploratory outcomes included compliance, perceived stress, total metabolic equivalent of task (MET) minutes, and self-reported dietary intake. This study was approved by the University of Georgia Institutional Review Board, and written informed consent was obtained from all participants before any study procedures began.

Participants

Ninety-one (n = 23 LOW; n = 23 MID; n = 22 HIGH; n = 23 CON) sedentary adults between the ages of 25 and 75 y at increased risk of CVD were included in the per-protocol analysis (Figure 1). Increased risk for CVD was defined by either elevated cholesterol profiles or overweight/obesity. Elevated cholesterol profiles were defined as “At risk” in 2 or more of the variables (TC: 180–239 mg/dL, LDL cholesterol (110–159 mg/dL), TGs (130–199 mg/dL) or “High” in TC (≥240 mg/dL), LDL cholesterol (≥160 mg/dL) or TGs (between 200 and 350 mg/dL). Overweight/obesity was defined by BMI (BMI ≥ 25).

FIGURE 1.

CONSORT flow diagram selection of participants. HIGH, intervention group received 30% estimated energy needs from cottonseed oil; LOW, intervention group received 10% estimated energy needs from cottonseed oil; MID, intervention group received 20% estimated energy needs from cottonseed oil.

To rule out participants with probable familial hypercholesterolemia [18], individuals with TC >290 mg/dL or LDL cholesterol > 190 mg/dL plus a family history of myocardial infarction (MI) before 50 y in a 2nd-degree relative, or below age 60 in a 1st-degree relative were not eligible to participate in this study. Other exclusion criteria included females on hormone replacement therapy for <2 y, females who are pregnant or nursing, individuals who engaged in structured exercise >3 h/wk, weight gain or loss >5% of body weight in the past 3 mo, plans to begin a weight loss/exercise regimen during the trial, history of medical or surgical events that could affect digestion or swallowing, gastrointestinal surgeries, conditions or disorders, any chronic diseases (including moderate to severe asthma, chronic lung disease, and kidney disease), metabolic diseases, atherosclerosis, previous MI or stroke, cancer, fasting blood glucose levels >126 mg/dL or blood pressure >180/120 mmHg were also excluded. Individuals with medication use affecting digestion and absorption, metabolism (e.g. thyroid meds), lipid-lowering medications, medications for diabetes, steroid/hormone therapies, or current antibiotic cycles were excluded. Individuals who were on a medically prescribed or special diet or had food allergies (specific to the foods in the study) were not eligible. Additionally, those taking fish oil supplements, excessive alcohol use (>3 drinks/d for males; >2 drinks/d for females), tobacco or nicotine use, or underweight BMI (<18.5) were excluded.

Screening visit (v0)

Participants completed an online screening form to assess preliminary qualification. After preliminary qualification, participants underwent an initial phone screening, then participants reported to the Human Nutrition Lab (HNL) after an 8–12 h overnight fast and abstaining from exercise and alcohol for ≥24 h. Height, weight, and blood pressure were recorded, and BMI inclusion criteria were confirmed. A fasting blood sample was obtained to assess for standard blood lipids and blood glucose to determine eligibility as described above in exclusion criteria (Piedmont Athens Regional Laboratory). An Alcohol Use Disorders Identification Test was given to participants to determine alcohol consumption habits [19]. Next, resting metabolic rate (RMR; kilocal/d) was measured utilizing indirect calorimetry (TrueOne 2400, Parvo Medics) following standard procedures [20]. Total daily EEN were calculated based on a participant’s RMR multiplied by an activity factor of 1.55, which is slightly lower than the average United States adult physical activity level at 1.63 [21] but was appropriate for our study population. This calculation was used for the 28-d diet and meal challenges at the preintervention and postintervention visits. Individuals who qualified for the study were then randomly assigned into one of the 4 groups with a random number generator. Participants were instructed by a member of the research team, which included a registered dietitian and graduate students in a dietetics program, to record a 3-d food record (2 weekdays, 1 weekend day). Participants were asked to record everything they consumed, including food, beverages, cooking oils/butter, condiments, and brands using serving sizes, measurement (weight and/or volume measurements), and utilizing hand size approximations when direct measurements were not available (i.e., meals consumed at restaurants).

Preintervention visit (v1)

Participants arrived at the HNL in a fasted state (8–12 h fast) and 24 h without alcohol or exercise. Upon arrival, participants turned in a 3-d food record [22]. All participants consumed a standardized meal and snack for dinner the night before v1 comprised ∼35% of total energy from fat, 15% protein, and 50% carbohydrate. In the laboratory, anthropometrics were measured as described in the screening visit (height and weight). Waist and hip circumference were measured in triplicate to calculate waist-to-hip ratio. Body composition was measured using a Bod Pod (Cosmed USA, Inc.). Then, an intravenous (IV) catheter was placed, and a fasting blood sample was collected 5 min after IV insertion. The line was kept patent with a saline solution. Next, participants consumed a high-fat meal challenge rich in SFAs within a 10-min period. The liquid meal contained a milk-chocolate, ready-to-drink shake (Ensure, Abbott Laboratories, Inc.), unsalted butter, red palm oil, coconut oil, soy lecithin granules, and powdered chocolate drink mix. The meal provided 35% of the participant's total EEN, and the composition of the meal can be found in Prater et al. [14]. A validated sensory questionnaire, using a modified 9-point hedonic scale with “1” indicating “dislike extremely” and “9” indicating “like extremely,” was given to assess perception of the study foods [23]. Blood draws then occurred at 30, 60, 90, 120, 150, 180, 240, and 300 min (5 h) postprandially. To collect data on potential confounding factors such as stress, anxiety, and physical activity, participants completed questionnaires including the Perceived Stress Scale (PSS) [24], the State Trait Anxiety Inventory (STAI) [25], and the International Physical Activity Questionnaire (IPAQ) [26].

28-d dietary intervention

Following v1, participants began the 28-d diet intervention. Before leaving the HNL after v1, participants were given a 1-wk supply of daily shakes and snacks that corresponded to their assigned diet group. Participants in the MID, LOW, and CON group received all their oil in the form of a breakfast shake with an additional snack item that did not contain oil. The HIGH group received oil through both their breakfast shake and the daily snack (Table 1). The ingredients for the breakfast shakes and snacks were identical between groups, with the only differences being the amount of the assigned oil incorporated into the foods. Participants were provided with the amount of calories they would be receiving from their provided intervention foods. Participants were counseled to consume the shake as a meal replacement and replace the snack with a food item of a similar caloric density from their diet. To implement this, research personnel worked with participants to understand or identify foods in their usual intake (from their food records) that could be suitable replacements for the intervention foods. To provide doses to the groups at 10%, 20%, and 30% of energy needs from oils, by design, participants received different amounts of total energy from the intervention foods. This was not a complete inpatient or outpatient feeding study (where research personnel could adjust the energy content of other meals to ensure energy balance for each individual), so the self-selection of energy content and amount for remaining meals was left to the participant in this free-living scenario.

TABLE 1.

Description of diet interventions

Intervention group	Percentage of energy provided from oil per day (%)	Percentage of energy from provided foods per day (%)	Provided food items that contained oil	Provided food items that do not contain oil
CSO HIGH	30	47	Breakfast shake and 1 snack1,2	—
CSO MID	20	37	Breakfast shake2	1 snack1
CSO LOW	10	27	Breakfast shake2	1 snack1
CONTROL	10	27	Breakfast shake3	1 snack1

Abbreviation: CSO, cottonseed oil.

¹The snack was either a yogurt or pudding.

²CSO was utilized in food. The CSO contained 54%, 19%, 27%, and 54% of PUFA, MUFA, SFA, and linoleic acid, respectively.

³Control oil mixture containing canola, coconut, corn, safflower, and soybean oil was utilized in food. The control oil contained 25%, 39%, 36%, and 21% of PUFA, MUFA, SFA, and linoleic acid, respectively.

During weeks 2 and 4 of the intervention, participants completed 3-d food records (2 weekdays and 1 weekend day). Daily nutrient intakes from the food diaries were assessed using The Food Processor SQL software (version 11.11.32). Participants were also asked to record the foods consumed daily during the intervention using the compliance checklist. Participants who consumed <75% of the provided foods were deemed noncompliant and removed from the final data analysis. Additionally, participants were asked to return empty food containers at their weekly visits and the postintervention visit to aid in compliance checks (v2-5).

Study foods

The differences in fatty acid composition between CSO and the control are presented in Table 1. A fatty acid analysis for the CSO and control treatment is provided in this study is provided in Supplemental Table 1. The control oil contained 25% safflower oil (7% SFA, 79% MUFA, 14% PUFA) 13% canola oil (7% SFA, 64% MUFA, 29% PUFA), 33% coconut oil (93% SFA, 7% MUFA, 0% PUFA), 13% soybean oil (14% SFA, 28% MUFA, 57% PUFA), and 17% corn oil (14% SFA, 28% MUFA, 57% PUFA). All gram amounts for the study were based on a 2000 kcal diet that was scaled based on the participant’s EEN. For the provided study foods, participants were given a bag of Instant Carnation Breakfast Shake mix (27 g carbohydrate, 5 g protein, and 1 g fat per 36 g serving) and The Vitamin Shop soy lecithin (0 g carbohydrate, 0 g protein, and 4 g fat per 7.5 g serving) as an emulsifier to mix with a bottle of their preweighed intervention oil. The snack foods provided in the study were Dannon Greek Light and Fit Yogurt (9 g carbohydrate, 12 g protein, and 0 g fat per 150 g serving) and Snack Pack pudding (17 g carbohydrate, 0 g protein, and 3 g fat per 92 g serving).

Visits 2, 3, and 4 (v2-4)

During the 28-d intervention, participants returned to the HNL each week for short study visits (v2-v4). At these visits, a fasting blood draw and body weight were collected. Additionally, at v3, the IPAQ was completed. Participants consumed their breakfast shake in the laboratory, turned in study documents, and collected their food for the next 7 d. Those documents included: 1) compliance checklist, 2) saved food containers, 3) 3-d food records (v3 only), and 4) the sensory evaluation logs (as completed). At v4, participants were additionally provided the lead-in meal to be consumed the night before v5.

Postintervention visit (v5)

The day after the 28-d intervention, participants reported to the HNL for v5 following the same 8–12 h fast and 24 h abstinence from alcohol and exercise. The evening before v5, participants consumed the dinner meal and snack provided at v4, which was the exact same meal as that consumed before v1. All procedures performed at the preintervention visit (v1) were repeated including consumption of the same high SFA meal, anthropometric measurements, questionnaires, and blood draws.

Sample analysis

All blood samples were drawn into K2 EDTA-coated vacutainers (Becton, Dickinson, and Company), and immediately placed on ice. For the lipid panel at preintervention and postintervention visits (v1 and v5), a portion of the fasting blood sample was drawn into a serum separator clot activator evacuated tube and kept at room temperature for 30 min before centrifugation. Blood samples were centrifuged for 15 min at 3000 ×g at 4°C. The serum was held at 4°C until the advanced blood lipid panel was performed (Quest Nichols Institute). This lipid panel, which was the primary outcome, included TC, HDL cholesterol, TG, LDL cholesterol, TC:HDL cholesterol ratio, and non-HDL cholesterol through spectrophotometry and ion mobility lipoprotein fractionation. The remaining fasting sample and all postprandial samples were centrifuged under the same conditions, then aliquoted, and stored at –80°C until analysis. Primary postprandial outcomes of plasma TG [interassay coefficient of variation (CV): 7.3%; intra-assay CV: 3.1%] and NEFAs (interassay CV: 7.5%; intra-assay CV: 2.0%) were measured by enzyme-based calorimetric assays (Wako Chemicals USA, Inc.). All samples for a participant were analyzed within the same assay kit.

Statistical analysis

The SAS version 9.2 statistical package (SAS Institute INC.) was used for all data analysis, and statistical significance was set at P ≤ 0.05. All results were reported as mean ± SEM unless otherwise noted. A sample size of 72 (18/group) was estimated to detect a difference in LDL cholesterol (Cohen’s F of 0.335) based on the results of Prater et al. [14] using G∗power 3.19.7 and assuming ≥80% power with an α of 0.05. The decision to use per-protocol analysis was made a priori. The normality assumption was tested using the Shapiro–Wilk test and normal probability (Q–Q) plots. All models were adjusted for baseline age (y), sex (male or female), weight (kg), and BMI. Data were assumed to be missing completely at random. This was validated by conducting a logistic regression analysis to assess whether missingness was associated with group, age, or sex, which revealed no significant differences between the missingness indicator and outcomes with missing data. Missing data were handled within our linear mixed model utilizing the restricted maximum likelihood estimation without the need for advanced imputation.

To determine between-group differences from preintervention to postintervention for primary outcomes of blood lipids, repeated measures (RM) linear mixed models including factors of group and visit were utilized. For primary outcomes for meal response data (NEFA, TG), RM linear mixed models including the fixed effect of group and repeated effects of visit and postprandial timepoint were utilized.

To account for multiple outcomes and type I error, P values were adjusted with the false discovery rate (FDR) method. The Benjamini–Hochberg procedure was applied, which ranks raw P values and calculates adjusted P values (q values) based on their ranks. FDR-adjusted P values of ≤0.05 were considered statistically significant. Primary interpretations of the data were based on unadjusted P values for hypothesis testing, whereas FDR-adjusted results were used to account for multiple outcomes and identify strong associations.

For all other outcomes, such as compliance, perceived stress, total MET minutes, and self-reported dietary intake, RM linear mixed models for group and visit were used to assess differences between and within groups from preintervention to postintervention. An “intervention average” was calculated for total MET minutes [IPAQ from mid-intervention (week 2) and end of intervention (week 4)] and dietary intake (2 weekdays and 1 weekend day food records from mid-intervention), which was compared with baseline averages. To determine between-group differences for sensory evaluations of the intervention foods, RM linear mixed models for group were utilized. If a significant group × visit interaction was found, a Tukey–Kramer post hoc test was performed to adjust for multiple comparisons. For all linear mixed models, participants were modeled as a random effect.

Results

Participants

Baseline characteristics and impact of dietary intervention

Ninety-five participants were randomly allocated into intervention groups (n = 23 LOW; n = 25 MID; n = 24 HIGH; n = 23 CON); however, 4 participants were not included in the final analysis (Figure 1). Therefore, 91 participants completed the intervention (20 females and 3 males in the LOW group, 20 females and 3 males in the MID group, 20 females and 2 males in the HIGH group, and 21 females and 2 males in the CON group) and were included in this per-protocol analysis. Baseline and week 4 participant demographics are presented in Table 2. There were no differences between groups at baseline. For body weight, there was a group × visit interaction (P = 0.01). Post hoc analyses revealed that this interaction was driven by increases in LOW, MID, and HIGH (P < 0.01, P = 0.01, P < 0.001, respectively) compared with no change in CON (P = 0.17). For BMI, there was a trend for a group × visit interaction (P = 0.097). Post hoc analyses showed the trend in the interaction was driven by increases in LOW, MID, and HIGH (P = 0.03, P < 0.01, P < 0.001, respectively) compared with CON (P = 0.20). There were no other interaction effects for body composition, stress (PSS), anxiety (STAI), or physical activity (IPAQ).

TABLE 2.

Characteristics of adults at-risk of cardiovascular disease at baseline and week 4 in CSO or control groups¹

Characteristic	CSO LOW (n = 23)		CSO MID (n = 23)		CSO HIGH (n = 22)		Control (n = 23)		P values
	Baseline	Week 4	Baseline	Week 4	Baseline	Week 4	Baseline	Week 4	Interaction
Female (%)	86	—	86	—	91	—	91	—	—
Age (y)	47 ± 3	—	46 ± 3	—	45 ± 3	—	49 ± 2	—	—
Height (cm)	167.0 ± 1.6	167.0 ± 1.7	165.5 ± 2.2	165.3 ± 2.2	165.4 ± 1.4	165.4 ± 1.4	163.8 ± 1.3	163.8 ± 1.3	0.49
Weight (kg)	91.5 ± 3.9	92.3 ± 4.02	85.1 ± 3.7	86.0 ± 3.82	90.2 ± 4.3	91.8 ± 4.32	87.8 ± 4.7	88.4 ± 4.7	0.01
BMI (kg/m2)	32.8 ± 1.8	33.0 ± 1.32	31.1 ± 1.3	31.5 ± 1.32	33.0 ± 1.5	33.5 ± 1.52	32.6 ± 1.6	32.9 ± 1.6	0.10
WC (cm)	98.2 ± 2.9	98.4 ± 3.1	95.0 ± 2.6	96.5 ± 2.6	99.7 ± 2.8	102.0 ± 3.3	96.8 ± 3.6	95.7 ± 3.7	0.15
HC (cm)	117.6 ± 2.1	116.5 ± 2.2	114.1 ± 2.3	114.9 ± 2.4	117.5 ± 3.1	118.8 ± 3.3	116.7 ± 3.4	116.7 ± 3.3	0.21
WHR	0.8 ± 0.0	0.8 ± 0.0	0.8 ± 0.0	0.8 ± 0.0	0.8 ± 0.0	0.9 ± 0.0	0.8 ± 0.0	0.8 ± 0.0	0.61
SBP (mmHg)	135.4 ± 3.6	132.3 ± 3.2	128.8 ± 4.3	137.9 ± 3.5	133.2 ± 3.2	136.5 ± 3.5	135.2 ± 3.6	132.5 ± 3.5	0.78
DBP (mmHg)	84.5 ± 2.9	81.8 ± 2.4	78.7 ± 3.2	78.0 ± 2.9	79.9 ± 1.9	81.0 ± 1.6	82.0 ± 2.6	83.4 ± 2.1	0.31
Body fat (%)	40.5 ± 1.6	40.7 ± 1.6	38.5 ± 1.7	38.6 ± 1.8	41.3 ± 2.0	42.0 ± 2.2	40.8 ± 2.0	40.5 ± 2.0	0.36
Total MET1 (min/wk)	2007 ± 384	2561 ± 606	2204 ± 463	1918 ± 274	1949 ± 341	2131 ± 401	1494 ± 237	1581 ± 242	0.53
STAI	50.1 ± 2.7	52.3 ± 3.5	53.6 ± 3.1	51.1 ± 2.8	58.8 ± 5.6	59.7 ± 6.9	50.3 ± 3.1	48.0 ± 2.0	0.23
PSS	12 ± 1	11 ± 1	11 ± 1	12 ± 1	12 ± 2	12 ± 2	12 ± 1	10 ±1	0.11

Abbreviations: CSO, cottonseed oil; DBP, diastolic blood pressure; HC, hip circumference, HIGH, intervention group received 30% estimated energy needs from cottonseed oil; LOW, intervention group received 10% estimated energy needs from cottonseed oil; MID, intervention group received 20% estimated energy needs from cottonseed oil; MET, metabolic equivalent (min/wk); PSS, Perceived Stress Scale; SBP, systolic blood pressure; STAI, state strait anxiety inventory; TFEQ, three factor eating questionnaire; WC, waist circumference; WHR waist-to-hip ratio.

¹All values are means ± SEMs. Anthropometrics were analyzed with a 2-way (group × visit) linear mixed model. Week 4 values represent an average of total MET min recorded during the 4-wk intervention.

²Indicates a significant difference from baseline to week 4 compared with control.

Nutrient intake and impact of dietary intervention

On average, participants in the LOW, MID, HIGH, and CON group reported consuming 94 ± 1%, 94 ± 1%, 94 ± 2%, 93 ± 1%, of the provided intervention foods, respectively, and compliance was not different between groups. However, 1 participant from the HIGH group was excluded from the final analysis due to poor compliance (defined as <75% study food consumption). Self-reported dietary intake is presented in Table 3. There were no differences between groups at baseline for the nutrients analyzed. With respect to total fat intake, there was a group × visit interaction (P < 0.01) driven by increases LOW (P < 0.01) and HIGH (P < 0.01) compared with no change in MID (P < 0.82) and CON (P = 0.92). Furthermore, there was a group × visit interaction (P < 0.01) for n-3 fatty acid intake driven by decreases in MID (P < 0.01) and HIGH (P < 0.01) compared with no change in LOW (P = 0.47) and CON (P =0.99). Regarding total PUFA as well as n-6 fatty acid intake, there were group × visit interactions (P < 0.01 for both), driven by increases in LOW, MID, and HIGH (P < 0.01 for all) compared with no change in CON (P = 0.97 for both). There were no interaction effects for other self-reported dietary intake values.

TABLE 3.

Self-reported nutrient intake at baseline and intervention period in CSO or control groups¹

Measure	CSO LOW (n = 23)		CSO MID (n = 23)		CSO HIGH (n = 22)		Control (n = 23)		P values
	Baseline	Intervention	Baseline	Intervention	Baseline	Intervention	Baseline	Intervention	Interaction
Energy (kcal/d)	2010 ± 135	2595 ± 142	2369 ± 249	2670 ± 268	2230 ± 223	2782 ± 172	2151 ± 142	2362 ± 184	0.27
Protein (g/d)	74.9 ± 5.8	88.1 ± 4.7	86.9 ± 8.5	121. 4 ± 34.7	89.0 ± 10.5	89.0 ± 6.2	76.7 ± 5.2	87.3 ± 5.5	0.47
Carbohydrate (g/d)	234.7 ± 18.5	258.4 ± 19.0	249.3 ± 21.8	255.5 ± 28.0	234.7 ± 26.5	236.3 ± 17.3	226.1 ± 19.3	243.5 ± 23.0	0.86
Fiber (g/d)	14.7 ± 1.5	14.3 ± 1.4	19.3 ± 2.6	14.5 ± 1.7	16.8 ± 2.0	16.2 ± 2.2	17.8 ± 1.8	13.4 ± 1.7	0.17
Sugar (g/d)	94.8 ± 10.3	121.1 ± 8.6	90.0 ± 9.1	125.7 ± 19.7	74.5 ± 9.4	111.3 ± 10.9	89.6 ± 11.0	116.0 ± 11.3	0.90
Fat (g/d)	78.0 ± 5.6	124.0 ± 6.52	103.5 ± 15.5	118.7 ± 6.0	93.2 ± 9.5	151.6 ± 10.12	92.9 ± 7.2	105.5 ± 9.0	<0.01
Saturated fat (g/d)	26.0 ± 1.8	38.2 ± 2.2	35.3 ± 4.9	36.4 ± 1.8	30.1 ± 3.6	45.4 ± 1.2	33.2 ± 3.6	38.9 ± 4.5	0.05
Trans fat (g/d)	0.4 ± 0.1	0.4 ± 0.1	0.5 ± 0.1	0.3 ± 0.1	0.8 ± 0.3	0.4 ± 0.1	0.4 ± 0.1	0.5 ± 0.1	0.07
MUFA (g/d)	31.8 ± 2.8	34.5 ± 1.8	39.0 ± 5.7	33.4 ± 2.2	38.0 ± 4.1	41.4 ± 3.7	38.3 ± 2.8	41.4 ± 3.4	0.12
PUFA (g/d)	19.8 ± 1.6	51.9 ± 3.02	28.7 ± 5.6	48.6 ± 2.52	24.3 ± 2.0	64.7 ± 3.12	21.0 ± 1.8	24.6 ± 2.3	<0.01
n-3 fatty acid (g/d)	2.2 ± 0.4	1.4 ± 0.1	2.6 ± 0.5	1.1 ± 0.12	3.0 ± 0.6	1.4 ± 0.22	1.8 ± 0.2	2.1 ± 0.2	<0.01
n-6 fatty acid (g/d)	17.7 ± 1.5	50.5 ± 3.02	26.1 ± 5.1	47.5 ± 2.52	21.3 ± 1.8	63.3 ± 3.02	19.1 ± 1.7	22.5 ± 2.1	<0.01
Alcohol (g/d)	2.1 ± 1.2	1.8 ± 0.5	4.0 ± 1.4	3.1 ± 2.8	5.0 ± 2.1	5.5 ± 1.8	5.1 ± 1.3	3.6 ± 1.0	0.87

Abbreviations: CSO, cottonseed oil; HIGH, intervention group received 30% estimated energy needs from cottonseed oil; LOW, intervention group received 10% estimated energy needs from cottonseed oil; MID, intervention group received 20% estimated energy needs from cottonseed oil; n-3, omega 3; n-6 omega 6.

¹All values are means ± SEMs. Baseline values are an average of 3 food diaries before the intervention. Intervention values are an average of all 6 food diaries kept throughout the intervention. All values were analyzed with a 2-way (group × visit) linear mixed model. To convert kilocalories to kilojoules multiply by 4.184. The average contribution of study foods to total energy intake was 662, 880, 1063, and 643 kcal in the LOW, MID, HIGH, and CON groups, respectively. The average contribution of fat intake from those study foods to total energy intake was 550, 775, 1038, and 534 kcal in the LOW, MID, HIGH, and CON group, respectively.

²Indiates a significant between-group difference from baseline to intervention compared with control.

Fasting blood lipid markers

Fasting biochemical markers are presented in Figure 2 and Table 4. For TC, there was a group × visit interaction (P = 0.02), driven by reductions in the LOW, MID, and HIGH CSO groups (P < 0.01, P < 0.001, P < 0.001, respectively) compared with no change in CON (P = 0.95). For LDL cholesterol, there was a group × visit interaction (P = 0.03) driven by reductions in the MID and HIGH CSO groups (P < 0.001 for both), and a trend in the LOW group (P = 0.08), compared with no change in CON (P = 0.75). Furthermore, there was an interaction effect for non-HDL cholesterol (P < 0.01), driven by reductions in the LOW, MID, and HIGH CSO groups (P < 0.01, P < 0.001, P < 0.001, respectively) compared with no change in CON (P = 0.75). Similarly, there was a group × visit interaction for TC:HDL cholesterol ratio (P < 0.01), driven by reductions in the MID and HIGH (P < 0.001 for both) and a trend for reductions in LOW (P = 0.09) with no change in CON (P = 0.74).

FIGURE 2.

Fasting serum levels of: (A) total cholesterol (TC), (B) LDL cholesterol (c), (C) HDL-c, (D) serum triglycerides, (E) TC:HDL-c ratio, and (F) non-HDL-c from prediet to postdiet intervention in adults at risk of cardiovascular disease [n = 23 CSO LOW; n = 23 CSO MID; n = 22 CSO HIGH; n = 23 control (CON)]. ∗Indicates a difference compared with control (P ≤ 0.05); # indicates a trend for differences compared with control (CON) (P = 0.06 to 0.10). All data are presented as mean ± SEM. All values were analyzed with a 2-way (group × visit) linear mixed model. CSO, cottonseed oil; CSO LOW, received 10% energy needs from cottonseed oil; CSO MID, received 20% energy needs from cottonseed oil; CSO HIGH, received 30% energy needs from a cottonseed oil; CON, received 10% energy needs from a mixture of oils representative of the fatty acid composition of the United States diet; pre, preintervention; post, postintervention; TC, total cholesterol.

TABLE 4.

Fasting biochemical markers of adults at risk of cardiovascular disease at baseline and week 4 in CSO or control groups¹

Measure	CSO LOW (n = 23)		CSO MID (n = 23)		CSO HIGH (n = 22)		Control (n = 23)		P values
	Baseline	Week 4	Baseline	Week 4	Baseline	Week 4	Baseline	Week 4	Interaction
TC (mg/dL)	200.3 ± 9.0	190.3 ± 8.92	193.2 ± 6.8	180.6 ± 6.12	193.9 ± 6.8	180.0 ± 5.92	186.7 ± 7.2	183.7 ± 7.4	0.02
LDL-c (mg/dL)	115.5 ± 7.3	109.0 ± 6.43	113.1 ± 5.1	101.3 ± 5.12	114.4 ± 5.5	101.5 ± 5.02	106.9 ± 6.5	103.0 ± 6.6	0.03
HDL-c (mg/dL)	59.3 ± 3.3	59.3 ± 3.4	56.3 ± 3.2	59.5 ± 3.5	56.6 ± 3.1	59.2 ± 3.5	55.8 ± 2.2	57.0 ± 2.4	0.20
TG (mg/dL)	115.4 ± 11.4	107.4 ± 12.3	108.5 ± 10.5	102.8 ± 10.4	109.5 ± 9.8	97.2 ± 8.0	129.4 ± 14.2	136.9 ± 15.0	0.10
Non-HDL-c (mg/dL)	141.0 ± 8.2	131.0 ± 7.62	136.9 ± 6.2	121.1 ± 5.92	137.3 ± 6.3	120.8 ± 5.92	130.8 ± 7.1	126.7 ± 7.0	<0.01
TC:HDL-c ratio	3.6 ± 0.2	3.3 ± 0.22	3.6 ± 0.2	3.2 ± 0.22	3.6 ± 0.2	3.2 ± 0.22	3.4 ± 0.2	3.3 ± 0.2	<0.01
NEFA (mEq/L)	0.49 ± 0.04	0.41 ± 0.03	0.43 ± 0.04	0.44 ± 0.03	0.49 ± 0.05	0.45 ± 0.05	0.51 ± 0.04	0.47 ± 0.04	0.33

Abbreviations: c, cholesterol; CSO, cottonseed oil; HIGH, intervention group received 30% estimated energy needs from cottonseed oil; LOW, intervention group received 10% estimated energy needs from cottonseed oil; MID, intervention group received 20% estimated energy needs from cottonseed oil; NEFA, nonesterified free fatty acid; TC, total cholesterol; TG, triglyceride.

¹All values are means ± SEMs. All values were analyzed with a 2-way (group × visit) linear mixed model. All outcomes were measured in serum, except for NEFA, which were measured in plasma. To convert LDL-c and HDL-c to mmol/L divide by 38.67. To convert TG to mmol/L divide by 88.57.

²Indicates a significant between-group difference from baseline to week 4 compared with control.

³Indicates a trend for a significant between-group difference from baseline to week 4 compared with control.

For TG, there was a trend for a group × visit interaction (P = 0.10); however, post hoc analyses revealed this interaction to be driven by nonsignificant decreases in the LOW, MID, HIGH groups, and increases in CON (P = 0.65, P = 0.93, P = 0.49, P = 0.94, respectively). There were no other group-by-visit interactions for any blood lipid variables. There were also no interaction effects in fasting HDL cholesterol or NEFAs.

FDR correction using the Benjamini–Hochberg procedure

After controlling for multiple outcomes using FDR correction, significant interactions remained for TC (P = 0.04), LDL cholesterol (P = 0.05), non-HDL cholesterol (P = 0.01), and TC:HDL cholesterol ratio (P = 0.02). There were no other significant interactions after adjusting for multiple comparisons for HDL cholesterol, TG, or NEFA.

Postprandial biochemical markers

The meal responses for TGs are presented in Figure 3. Postprandial TGs were suppressed in HIGH (P = 0.02) compared with no change in LOW (P = 0.4), MID (P = 0.4), or CON (P = 0.4). There were no interaction effects for NEFAs.

FIGURE 3.

(A–D) Time course for plasma triglycerides and nonesterified fatty acids (NEFA; E–H) for each group at preintervention and postintervention in adults at risk of cardiovascular disease [n = 23 LOW; n = 23 MID; n = 22 HIGH; n = 23 control (CON)]. Participants consumed a high-saturated-fat meal after time 0. ∗Indicates a significant difference between preintervention and postintervention meal responses compared with control (CON) (P ≤ 0.05). All data are presented as mean ± SEM. All values were analyzed with a linear mixed model with a fixed effect of group and repeated effects of visit and postprandial timepoint. CON, received 10% energy needs from a mixture of oils representative of the fatty acid composition of the United States diet; HIGH, received 30% energy needs from cottonseed oil; LOW, consumed 10% energy needs from cottonseed oil; MID, received 20% energy needs from cottonseed oil; pre, preintervention; post, postintervention.

Sensory evaluations

Sensory evaluations for the first time a study food was consumed during the intervention are presented in Table 5. There were no differences between groups for any characteristic (appearance, taste/flavor, texture/consistency, aroma/smell, and overall acceptability) of the CSO- or control oil-enriched study foods.

TABLE 5.

Sensory evaluation of intervention foods in the CSO or control groups

	CSO LOW (n = 23)		CSO MID (n = 23)		CSO HIGH (n = 22)		Control (n = 23)		P values
Appearance	7.2 ± 0.2	—	7.2 ± 0.2	—	6.6 ± 0.3	—	7.1 ± 0.2	—	0.18
Taste/flavor	7.1 ± 0.2	—	7.2 ± 0.2	—	7.0 ± 0.2	—	7.3 ± 0.2	—	0.67
Texture/consistency	7.1 ± 0.2	—	7.2 ± 0.2	—	6.7 ± 0.3	—	7.2 ± 0.2	—	0.27
Aroma/smell	7.2 ± 0.2	—	7.3 ± 0.2	—	6.8 ± 0.3	—	7.2 ± 0.2	—	0.32
Overall acceptability	7.2 ± 0.2	—	7.3 ± 0.3	—	7.0 ± 0.3	—	7.4 ± 0.1	—	0.46

Abbreviations: CSO, cottonseed oil; HIGH, intervention group received 30% estimated energy needs from cottonseed oil; LOW, intervention group received 10% estimated energy needs from cottonseed oil; MID, intervention group received 20% estimated energy needs from cottonseed oil; pre, preintervention; post, postintervention.

All values are means ± SEMs. Ratings are based on a 9-point hedonic liking scale with 1 indicating “extremely dislike,” 5 indicating “neither like or dislike,” and 9 indicating “extremely like.” All sensory evaluations of intervention foods were analyzed with a 1-way (group) linear mixed model. Evaluations of intervention foods were done on the first day of the intervention only.

Discussion

For the first time, in a population at risk for CVD, we have shown that CSO resulted in significant improvements in fasting blood lipids at doses ≤30% EEN. Specifically, we observed decreases in fasting TC, non-HDL cholesterol, and TC:HDL cholesterol ratio in diets enriched with 10%, 20%, and 30% CSO compared with a CON diet. Furthermore, enriching diets with 20% and 30% CSO led to significant reductions in LDL cholesterol, whereas the 10% CSO dose group displayed only a trend for reductions in LDL cholesterol. Additionally, the 30% CSO dose suppressed postprandial TGs after an HF meal challenge. Altogether, this provides further evidence of CSO’s lipid-lowering effects, even at lower doses of consumption (≤30% EEN).

CSO has a unique fatty acid composition because of its high PUFA content, but also high SFA content (27% SFA, 19% MUFA, 54% PUFA; Supplemental Table 1). Given its high SFA content, CSO could be regarded as a “less healthy” oil. Yet, CSO is very effective at reducing blood lipid levels [[14], [15], [16]], even when compared with other high n–6 PUFA sources in preclinical models [15,[27], [28], [29], [30], [31]]. It was not surprising that we observed improvements in blood lipids in the HIGH group (30% EEN from CSO) based on past literature utilizing this dose to yield similar effects [12,14]. Although intake of ∼15%–20% of energy from canola and corn oil has been identified for optimal changes to blood lipids [32,33], before this study, there had been no dose–response clinical trials for CSO. Therefore, this is the first study to demonstrate that doses lower than 30% EEN of CSO improve blood lipids, even when slight, but statistically significant, weight gain occurs.

The drawback with utilizing high doses of an intervention oil is that total fat intake typically exceeds the acceptable macronutrient distribution range (AMDR) for total fat as determined by the Dietary Guidelines (current AMDR is 20%–35%) [34]. Previous studies demonstrating lipid-lowering effects with CSO have corresponded to participants’ total fat intake ranging from 40% to 50% of total energy [[12], [13], [14]]. Therefore, the idea behind this dose–response study was to determine whether lower doses of CSO would be effective at lowering blood lipids, while hopefully keeping total daily fat intake within the AMDR to encourage healthier dietary patterns. Unfortunately, the former was more successful than the latter. The actual percentage of energy from fat was 45.1%, 44.2%, and 50.6% in the LOW, MID, and HIGH CSO groups, respectively, and 41.5% in the CON. Therefore, even with our LOW dose (and CON) where 10% of energy from fat was provided in the intervention, the fat intake for the remaining self-selected meals/snacks was not low enough for participants to fall within the AMDR. Importantly, in our CVD at-risk population, self-reported intake revealed that participants were already following a HF diet before the start of the intervention (LOW: 36.1%; MID: 39.0%; HIGH: 39.3%; CON: 40.0%), which is slightly higher than the United States averages (∼36% of energy) [35]. Thus, even with isocaloric integration of the study foods, participants would likely still have been on an HF diet. Despite an increase in reported energy intake during the intervention, the calculated proportion of CSO consumed relative to actual self-reported intake was 10%, 20%, and 28% in the LOW, MID, and HIGH groups, respectively. The average contribution of study foods to total energy intake was 662, 880, 1063, and 643 kcal in the LOW, MID, HIGH, and CON groups, respectively. However, the actual increase in energy intake in the same groups was 585, 301, 552, and 211 kcal, indicating that participants at least partially substituted the study foods for other items in their diet. To illustrate this further, we used the National Institute of Health Body Weight Planner [36] to estimate the weight gain expected if participants had consumed the study foods as in addition to their usual diet with no dietary substitution or compensation. Predicted weight gains were 1.0, 3.3, 3.2, and 1.3 kg in the LOW, MID, HIGH, and CON groups, respectively, compared with the observed gains of only 0.8, 0.9, 1.6, and 0.6 kg.

This HF diet consumption, particularly in the HIGH group, was a possible cause of the weight/BMI gain that occurred during the intervention in the groups because dietary fat has such a high energy content. Although dietary substitution instructions were provided, and the study foods were well-liked and easy for participants to incorporate into their daily lives based on sensory data, a complete energy balance was still not achieved. Although the body adapts to HF diets by increasing fat oxidation to match the higher fat intake [37], a positive energy balance can still occur, which can negatively impact health over time [38]. This highlights another potential benefit of utilizing lower doses (≤30% EEN) of CSO for blood lipid reductions so as not to negatively influence other aspects of health long term. These findings underscore the need for dietetic counseling in future nutrition interventions utilizing lower doses of CSO to promote energy balance and adherence to the AMDR. This would aid in the incorporation of CSO into one’s diet to support cholesterol reduction within an overall healthy dietary pattern.

In human studies, it is important to consider not just statistical significance but the clinical relevance of the findings as well. We found that the reductions in TC for the LOW, MID, HIGH CSO groups were 4.9%, 6.3%, and 6.7%, respectively. For LDL cholesterol, the MID and HIGH CSO groups elicited a comparable reduction of 10.4% and 10.8%, respectively, compared with 4.1% reduction in the CON group. These results are clinically meaningful because every 1% of LDL cholesterol reduction equates to a 1.2%–2.0% reduction of coronary artery disease (CAD) risk [39,40], resulting in an estimated ∼12.3% to 21% CAD risk reduction with MID and HIGH doses of CSO. Moreover, LDL cholesterol reduction serves as a primary therapeutic target in mitigating and preventing CVD [[41], [42], [43]]. Therefore, the magnitude of LDL cholesterol reduction in our 2 highest CSO doses was comparable with that of commonly used lipid-lowering pharmacological therapies including bile acid sequestrants, fibrates, and nicotinic acid [44]. There were no reported adverse side effects from participants in this study. LDL cholesterol concentration reflects the cholesterol content in LDL cholesterol particles, but it does not account for other atherogenic lipoproteins encompassed by non-HDL cholesterol. Therefore, non-HDL cholesterol is recognized as a strong predictive indicator of CVD risk [45]. In this study, non-HDL cholesterol was reduced by 6.7%, 11.5%, and 11.7%, in the LOW, MID, and HIGH group, respectively, which correlates to reduced CVD risk. Importantly, this outcome remained statistically significant after adjusting for multiple outcomes, highlighting the robustness and clinical relevance of this finding.

Although we hypothesized that the LOW, MID, and HIGH CSO groups would elicit improvements in blood lipids in a dose-dependent manner, it is notable that the intake of 20% and 30% EEN from CSO produced a similar magnitude of effect whereas the LOW dose appeared to have a slightly less magnitude of response. Still, the significant reductions observed within the LOW group for TC, non-HDL cholesterol, and TC–HDL cholesterol ratio, and a trend for improvements in LDL cholesterol, present a novel finding with such a low intake of CSO. Given that the LOW group was consuming ≥5% EEN from LA (12.0–16.7 g LA/d, dependent on EEN), their reductions align with existing evidence that consuming 5%–10% of energy from LA results in improvements in LDL cholesterol and TC [46]. We can theorize that a longer intervention period may lead to greater reduction in LDL cholesterol and reveal statistically significant reductions with the LOW dose. This theory is corroborated by the continued reductions observed with a previous CSO intervention from 4 to 8 wk in a longer-term trial [14]. Nevertheless, evaluating doses <20% in longer-term studies is warranted to understand the relationship between the length of the intervention and the magnitude of response compared with a control diet.

There are specific components of CSO thought to be responsible for its beneficial effects on lipid metabolism. CSO has the ability to interact with transcription factors through its high PUFA content to regulate lipid and cholesterol metabolism [47]. Beyond just the high LA content, though, CSO is unique in that it contains a bioactive lipid, DHSA, a known inhibitor of the lipogenic enzyme SCD1 [15]. Specifically, SCD1 is a lipogenic enzyme that catalyzes the desaturation of SFA to MUFAs, promoting lipid storage rather than FFA oxidation [48]. The concentration of DHSA in CSO has been confirmed at 0.3% in other human feeding trials [49]. This coupled effect of high PUFA content and DHSA may be necessary to elicit the strong effects for blood lipids observed in previous CSO trials. CSO also contains tocopherols, which protect LDL from oxidation, and beta-sitosterol, which reduces cholesterol absorption through competitive inhibition [16]. Nonetheless, it is important to consider that CSO as whole food may be required to elicit the marketed changes in blood lipid metabolism rather than an isolated component.

United States adults spend nearly two-thirds of their day in the postprandial state, a period known to exacerbate disease development [50,51]. Our study design allowed us to assess the effects of chronic CSO consumption on postprandial metabolism following an SFA-rich, HF meal typical of Western dietary patterns. Although we did not observe significant suppression of NEFA with CSO consumption, as reported in a previous 8-wk CSO trial following the same meal challenge design [14], we did observe significant suppression of postprandial TGs with our 30% CSO dose group. This aligns with findings from Polley et al. [12] who reported a similar suppression of postprandial TGs with acute CSO consumption. Importantly, the suppression of postprandial TGs in this study may be partially attributed to the nonsignificant decrease in fasting TGs in the HIGH group. Even so, this finding suggests that CSO may offer protective effects by attenuating postprandial TGs following high SFA meal ingestion, but potentially only at high daily doses of CSO. Further investigation is warranted to understand the relationship between length and dose of CSO consumption and postprandial responses.

There were some limitations with this study. Our control group received a mixture of oils matching the FA composition of the United States diet. These oils were mixed into the same foods that the CSO groups received intentionally to isolate the effect of CSO; however, we did not have a control group that received no food or oils. To keep the groups at 10%, 20%, and 30% of energy from oils, by design, participants received different amounts of total energy from intervention foods (Table 1). Our study foods were palatable (Table 5) and were partially provided as liquid kilocalories, which could have contributed to the lack of complete isocaloric substitution by the participants [52]. Similarly, the amount of energy provided in each group was moderate to large, so participants may not have been sufficiently experienced with listening to internal appetite cues to fully compensate for this amount of added energy. The slight, nonsignificant improvements in some outcome measures in the control group may have prevented the observation of significance in those measures with our LOW group, such as with LDL cholesterol. Also, participants were free-living, making it possible to consume other foods and beverages. Again, this was done by design as our goal was to determine whether incorporation of CSO into one’s diet could improve CVD risk measures. Measures of dietary intake, physical activity, stress, and compliance were all self-reported, which may contain some degree of under-reporting or over-reporting, and we did not utilize direct measures or biomarkers to assess compliance such as LA. We did try to control for error in reporting compliance by counting empty food containers brought back by participants on their weekly visits and comparing that to their compliancy log. This study also utilized a per-protocol analysis, which introduces some degree of bias and should be considered when interpreting these results. We did complete an analysis of baseline characteristics of participants lost to follow-up and included in the analysis. This revealed no significant differences. Additionally, although data appeared to be missing completely at random based on analyses of available covariates, this assumption cannot be verified and should be interpreted with caution. Although we recruited both males and females, our study had predominantly female participants, so future work should also emphasize male participants. Finally, this design was single blinded.

In conclusion, we have shown that diets enriched with 10%, 20%, and 30% from CSO reduced TC, non-HDL cholesterol, and TC:HDL cholesterol ratio compared with a control diet in adults at risk for CVD. Diets enriched with 20% and 30% from CSO also had significant reductions in LDL cholesterol whereas the 10% CSO dose only exhibited a trend in LDL cholesterol improvements. These reductions in blood lipids at lower doses (≤30% EEN) of CSO may enable the adoption of healthier dietary patterns (i.e., lower fat intake) while still reaping the benefits of CSO consumption; however, it must be incorporated into an energy-balanced diet, which may prove challenging. Sensory data indicated that participants liked the CSO-enriched foods, supporting the feasibility and acceptability of incorporating CSO in the diet. Future studies should investigate CSO diet enrichment in different populations and at lower doses, especially the 10% dose, for longer durations. Also, future studies should evaluate other high-PUFA oils compared with CSO to decipher if CSO has superior lipid-lowering effects. Lastly, to provide a more robust understanding of CSO’s effects on CVD risk, other studies should assess markers of other chronic disease risk factors.

Author contributions

The authors’ responsibilities were as follows – JAC, CMP: project conceptualization, funding acquisition, study oversight, methodology, and provided essential reagents and materials; JAC, MCP: developed the overall research plan and methodology; MCP, CDH, CAW: conducted the research; CDH, CAW: analyzed the data; JAC, CDH: wrote the manuscript; and all authors: read and approved the final manuscript.

Data availability

Data generated and analyzed during this study can be available from the corresponding author on reasonable request.

Funding

This study was funded by Cotton Incorporated (AWD00015914). The funders had no role in the design, implementation, analysis and interpretation of the data.

Conflict of interest

JAC reports financial support was provided by Cotton Incorporated. All other authors report no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: