Are long‐term FAD diets restricting micronutrient intake? A randomized controlled trial

Abstract The micronutrient adequacy of common fad diets is rarely assessed. We compared a high‐fiber diet [HF] with a low‐carbohydrate diet [LC] to assess their effects on body weight and micronutrient adequacy. One hundred and seventy‐three adult males and females with or without diabetes with a mean body mass index of 36 kg/m2 were randomized to either a HF or LC diet. Differences in anthropometrics, blood lipids, glucose, blood pressure, and micronutrient consumption between groups were assessed after 52 weeks. Differences between groups a priori were assessed using independent t tests and chi‐squared tests. Post hoc differences in nutrient consumption between groups while controlling for gender were assessed using factorial analysis of variance. After 52 weeks, LC dieters (n = 24) retained weight loss better than their HF counterparts (n = 30) (p = .06). LC dieters consumed more vitamin K (mcg) [HF = 124.0 ± 15.0; LC = 220.0 ± 39.1; p = .025] and vitamin B12 (mcg) [HF = 3.1 ± 0.3; LC = 4.1 ± 0.4; p = .026]. The HF group consumed more folate (mcg) [HF = 479.9 ± 34.0); LC = 333.8 ± 22.1; p < .001], magnesium (mg) [HF = 353.1 ± 17.4; LC = 281.1 ± 18.0; p < .001], and iron (mg) [HF = 14.6 ± 0.8; LC = 10.7 ± 0.6; p < .001. Both groups consumed less than the respective EAR for vitamins D and E and less than the AI for potassium. While a LC diet may be more effective for long‐term weight loss, both diets were deficient in micronutrients.

fad diets may reduce total daily calorie consumption. However, an unintended consequence may be inadequate micronutrient consumption. Micronutrient deficiencies have been found in those with obesity and may increase the risk of developing comorbidities such as diabetes, heart disease, cancer, and osteoporosis (Calton, 2010).
The popularity of low-carbohydrate, high-protein diets for body weight management remains high, and meta-analyses seem to support their effectiveness for weight loss (Brehm, Seeley, Daniels, Seeley, Daniels, & D'Alessio, 2003;Hu et al., 2012;Samaha et al., 2003). This may be due to increased satiety, lipid oxidation, and energy expenditure, all of which may contribute to negative energy balance (Brehm et al., 2003). Long-term effects of low-glycemic, high-fiber diets have also shown favorable effects on body weight and reduced risk of cardiovascular disease, type 2 diabetes, and colon cancer (Brouns, 2018;Reynolds et al., 2019). It has been proposed that high-fiber diets promote weight loss through various mechanisms, including delayed gastric emptying and insulin release as well as triggering the secretion of satiety hormones such as ghrelin, glucagon-like peptide, and polypeptide YY (Slavin, 2013). However, their respective micronutrient densities have not been adequately examined. We previously reported that a low-carbohydrate diet was superior to a high-fiber bean-rich diet for body weight management over the long term, whereas those adhering to the high-fiber bean-rich diet experienced more favorable effects on atherogenic blood lipids (Tonstad, Malik, & Haddad, 2014). Bean consumption has been associated with a reduced risk of diabetes, heart disease, and colon cancer (Mitchell, Lawrence, Hartman, & Curran, 2009). It has been proposed that adherence to a high-fiber plant-based diet promotes regular consumption of antioxidants which may result in reduced systemic inflammation (Castro-Quezada, Román-Viñas, & Serra-Majem, 2014). However, analyses of the micronutrient composition of low-carbohydrate and high-fiber bean-rich diets have yet to be explored. The present study served as a secondary analysis and compared a high-fiber bean-rich diet with a low-carbohydrate diet with respect to micronutrient adequacy.

| Study design
As previously reported by Tonstad et al. (2014), a prospective, randomized controlled trial was conducted as a proof-of-concept study.
Participants were randomly assigned to one of two diets: (a) a highfiber, bean-rich diet [HF] or (b) a low-carbohydrate diet [LC]. Women in the HF diet were encouraged to consume ≥ 40 g of fiber per day, whereas males aimed to consume ≥ 50 g of fiber per day. LC dieters were instructed to consume < 120 g of carbohydrates per day.
Participants were followed for 52 weeks. Follow-up visits were scheduled at regular intervals during the first 16 weeks and every 3 months thereafter. Participants met with a registered dietitian (RD) who conducted a 24-hr dietary recall to monitor compliance. Three additional 24-hr dietary recalls were conducted via phone at random between weeks 16 and 52. Changes in body weight, body mass index (BMI), waist and hip circumferences, and blood pressure were measured. Blood glucose levels, hemoglobin A 1c (HbA 1c ), and lipids were measured at 16 weeks and 52 weeks.
Based on Cohen's estimation of a medium effect size (d = 0.50), power = 80%, and α = .05, a total of 128 participants were needed (64 in each group) (Cohen, 1992). Accounting for a 25% dropout rate, 170 total participants were needed.

| Participants
Individuals were recruited using flyers and informational meetings at the Diabetes Treatment Center in Loma Linda, CA. The Loma Linda University Institutional Review Board (IRB) approved the study protocol and all participants were provided written informed consent.
Participants were deemed eligible if they were ≥ 18 years of age with a BMI between 30.0 and 44.0 kg/m 2 . Those with type 2 diabetes were eligible if they were clinically stable within the past 3 months.
Exclusion criteria included type 1 diabetes, pregnancy, lactation, active cancer, participation in another trial, dieting within the last 3 months with weight loss exceeding 4 lbs. (1.81 kg), adherence to a vegan diet, active eating disorder, current acute or chronic infection, use of weight loss medications, or gastrointestinal problems made worse by a high-fiber diet.

| Procedures
A total of 201 individuals were screened, in which 23 of them were ineligible. Ultimately, 173 subjects were randomized (Figure 1). A total of 91 participants were assigned to the HF group and 82 to the LC group. Married couples were randomized to the same group to improve compliance. Participants were stratified according to their diabetes status to ensure that those with diabetes were equally represented in both groups. An RD provided instructional materials (see Appendices A and B) and provided dietary education. This was followed by a 3-week preparation phase.
During this phase, those in the HF group began eating one serving (125 ml or ½ cup) of cooked beans with one meal each day.
Participants were encouraged to gradually increase their dietary fiber intake by consuming one serving of cooked beans with each of three meals per day. LC dieters were encouraged to consume fewer carbohydrates while increasing their intake of protein-rich foods such as lean meats, fish, and eggs. A 1-week dietary induction phase followed at week 4 after baseline. Each participant was provided seven meals over the course of 3 days (two breakfasts, three lunches, and two dinners), which were distributed from the Loma Linda University Research Kitchen. Meals were consumed and were designed to educate participants with regard to portion sizes and menu options for their respective diets. Meals for the high-fiber group provided half a cup of beans or lentils with breakfast, lunch, and dinner and emphasized whole grains, vegetables, and fruit. Meals for the low-carbohydrate group consisted of a large meat, egg or vegetable protein portion, one or two servings of nonstarchy vegetables, and a small fruit serving.
Participants were instructed to return for maintenance visits with an RD during weeks 6, 8, 12, and 14. Additional follow-up visits occurred at months 6, 9, and 12. During each visit, changes in body weight, BMI, waist and hip circumferences, dietary compliance, and physical activity patterns were assessed. Participants' physical activity levels were assessed using a brief questionnaire asking participants to indicate whether their physical activity levels had increased, decreased, increased vigorously, or remain unchanged since the previous visit.

| Anthropometric measurements
Participants' body weight and height were measured without shoes.

| Laboratory analyses
Participants were required to fast for 8 hr prior to all blood draws.
Samples were analyzed at the Loma Linda University Medical Center Central Laboratory. Total cholesterol levels were assessed using the Roche Cholesterol Assay (Roche). Direct assay was used to measure LDL cholesterol, whereas HDL cholesterol used polyethylene glycolmodified enzymes and dextran sulfate. Triglyceride levels were assessed using enzymatic calorimetric tests. The hexokinase method determined fasting blood glucose levels, and HbA 1c was assessed using the turbidimetric inhibition immunoassay for hemolyzed whole blood.

| Dietary compliance
During in-person maintenance follow-up visits, trained research RDs conducted 24-hr dietary recalls with each participant. Each participant was also provided with a blank food journal and encouraged to F I G U R E 1 Study timeline log their food and beverage intake in addition to dates, times of day, and portion sizes. These logs were also reviewed by the RDs during maintenance follow-up visits but were not included in our data analyses. Rather, the logs were used to encourage dietary compliance. Between weeks 6-16 and months 6-12, subjects were also telephoned unannounced at random and asked to participate in a 24-hr dietary recall to assess dietary compliance. Nutrition Data System for Research, version 2008 (2008) (Nutrition Coordinating Center, University of Minnesota, Minneapolis, MN, USA), was used to collect 24-hr dietary recall data via telephone. Data were collected using standardized interview prompts. These data were used in our statistical analyses.
In addition to presenting data on changes in body weight, micronutrient intakes were compared between diet group and gender. Diet adequacy was determined using the established estimated average requirement (EAR) values. The following nutrients were included when determining the adequacy of the respective diets: vitamin A, thiamin, riboflavin, niacin, vitamin B6, folate, vitamin B12, vitamin C, vitamin E, magnesium, phosphorus, selenium, and zinc. It has been reported that dietary iron intake is often not normally distributed, which may lead to difficulties interpreting iron intakes (Barr, Murphy, & Poos, 2002). Therefore, iron adequacy was not assessed in our analyses. Rather, iron intakes between groups were included for comparison purposes. It should also be noted that EARs have not been established for vitamin K and pantothenic acid (IOM, 2011), and therefore, only comparisons between groups were conducted for these nutrients. EAR has not been established for potassium; however, an adequate intake (AI) cutoff has been determined (National Academies of Sciences, 2019) which was used in its place.

| Statistical analyses
Data were entered into SPSS, version 25 (IBM, Inc.). After randomization, baseline differences between groups and changes over time were assessed using independent t tests and chi-squared tests. Statistical analyses were performed regardless of whether all test assumptions are met. When test assumptions were not met, nonparametric tests were conducted and compared to the initial analyses. If p-values significantly differed, the variables were transformed and re-analyzed. Nutrient intake comparisons between groups and gender were assessed using factorial analysis of variance. Comparisons of vitamin and mineral consumption between groups controlling for gender were conducted (Table 2). Using the last observation carried forward, results for the completers are shown where p < .05 was considered statistically significant.

| RE SULTS
Of the 173 participants that were randomized, 54 completed the 52-week intervention and data collection procedures (n = 24 in the LC group; n = 30 in the HF group). Baseline characteristics of the study sample are presented in Table 1. A higher proportion of women were enrolled in the study (74%) when compared with their male counterparts (26%). Twenty percent of those randomized were diabetic. The overall mean (±SD) age was 49 ± 1 years, the majority of whom identified as White (51%), followed by Hispanic/ Latino (25%), Black/African American (18%), and "other" (6%). At randomization, mean (±SD) body weight was 99 ± 2 kg. Glucose levels differed initially between groups. This was likely due to a higher proportion of diabetics being randomly assigned to the LC group.
As previously reported by Tonstad et al. (2014), while not statistically significant, the LC group tended to retain weight loss better than the HF group (p = .06), even though total energy intake and physical activity levels did not differ (data not shown).
The mean (SD) intakes by diet group and gender of select fatand water-soluble vitamins are presented in Table 2. Results of the statistical analyses revealed significant differences between the two groups. Expectedly, the LC group consumed higher quantities of vitamins commonly found in animal-based products such as vitamin K and vitamin B12 (Table 2). Unsurprisingly, those assigned to the HF group consumed more folate. Mean (SD) differences in mineral consumption between groups and gender intakes are seen in Table 3.
Somewhat expectedly, HF dieters consumed more magnesium and less calcium when compared to those in the LC group. However, quite surprisingly, these individuals consumed more iron. Of note, males and females in both groups consumed less than the EAR for vitamin D and vitamin E and less than the AI for potassium. EARs were used as the cutoff values. Recommended dietary allowances (RDAs) represent intakes that exceed the requirements of 98% of all individuals; therefore, this may overestimate the prevalence of nutrient inadequacies (Barr, Murphy, & Poos, 2002). Rather, EARs are estimates of median intakes and therefore exceed the needs of half of the sample and fall short of the needs for the other half. Because of this, EARs are considered more appropriate when assessing nutrient adequacy among groups.

| D ISCUSS I ON
We previously reported that LC dieters had significantly higher mean intakes of total, saturated, monounsaturated, polyunsaturated, and trans fats and cholesterol (Tonstad et al., 2014). These analyses indicate that micronutrient deficiencies may be present in those following fad diets.  Long-term consumption of vitamin D below the EAR may increase chronic disease risk. Osteomalacia has been well established as a result of long-term vitamin D deficiency in both children and adults (IOM, 2010a). A positive correlation between obesity and vitamin D deficiency has also been reported (Bell et al., 1985).
Vitamin D is stored in body fat cells; however, it may not be bioavailable which may account for this correlation (Holick, 2004

| Vitamin E
While vitamin E is found in a variety of forms, alpha-tocopherol is considered the most significant in humans (IOM, 2000). However, the majority of dietary vitamin E is consumed as gamma-tocopherol (Dietrich et al., 2006 Vitamin E is involved in immune function, cell signaling, and regulation of gene expression (Traber, 2006). It has been proposed that vitamin E protects cells against damage caused by free radicals by ceasing production of reactive oxygen species during fat metabolism. Vitamin E may also aid in the dilation of blood vessels and platelet aggregation promoting the release of prostacyclin from the endothelium (IOM, 2000).

| Calcium
It has been reported that across the lifespan, females consume less dietary calcium from food (Ervin, Wang, Wright, & Kennedy-Stephenson, 2004). The majority of study participants identified as female (74%) ( Table 1). Mean intakes among females in the HF group were lower than their LC counterparts (Table 3). U.S. NHANES 2003-2006 data revealed that mean calcium intake from foods and supplements in women ranges from 918 to 1,296 mg/day (Bailey et al., 2010). However, it has been estimated that only 30% of calcium is absorbed from food (IOM, 2010b). Beginning in childhood, the absorption rate of calcium in the gut begins to decrease such that by adulthood, only 15%-20% of dietary calcium is absorbed (Office of Dietary Supplements, 2020). The absorption rate continues to decrease throughout adulthood, which partially explains the increased EAR for calcium among females older than 50 years and among males and females older than 70 years (1,000 mg/day) (Heaney, Recker, Stegman, Recker, Stegman, & Moy, 1989;NIH, 1994). However, consuming calcium in the presence of vitamin D promotes the absorption of calcium in the gut (USDA, 2020).
Good food sources of calcium (≥20% DV) include yogurt, milk, cheddar cheese, and fortified juices and beverages (such as soy milk) (USDA, 2020). Calcium deficiency may not be immediately evident Dairy products, which are considered good sources of calcium (≥20% DV), may also reduce blood pressure (Appel et al., 1997). However, low-fat dairy products may be most beneficial for reducing blood pressure due to the potential atherogenic effects of full-fat dairy.

| Magnesium
Magnesium supports many physiological functions, including DNA synthesis, energy production, muscle contraction, and regulation of blood glucose and blood pressure (IOM, 1997;Rude, 2010Rude, , 2012. Magnesium homeostasis is closely regulated by the kidneys through modifications in urinary excretion (Office of Dietary Supplements, 2019a). Dietary magnesium can be obtained by consuming common foods, particularly those that contain dietary fiber such as spinach, nuts, seeds, and legumes (USDA, 2020). Males and females in the LC group consumed less than the respective established EAR for gender.
While excess magnesium consumption from supplements may lead to a diuretic effect, this has not been demonstrated with foods.
Whole, unprocessed foods are typically higher in magnesium as the refining process may affect magnesium content (IOM, 1997). U.S. NHANES 2005NHANES -2006 data estimate that those from all age groups consumed 284 mg/day, which is less than EAR (Moshfegh, Goldman, Ahuja, Goldman, Ahuja, Rhodes, & LaComb, 2009). It has been reported that dietary magnesium intakes greater than 250 mg/day may reduce the risk for ischemic heart disease and hemorrhagic stroke (Del Gobbo et al., 2013;Larsson, Orsini, & Wolk, 2012). We found that men and women in the LC group consumed more than this cutoff on average (Table 3). Of particular interest given the study sample, meta-analyses (Larsson et al., 2012;Schulze et al., 2007) have revealed an inverse relationship between magnesium intake and risk of type 2 diabetes, particularly among those with a BMI ≥ 25.0 kg/m 2 (Dong, Xun, He, Xun, He, & Qin, 2011). Increased consumption of magnesium-rich foods may be recommended for those with overweight or obesity as a potential preventive measure against metabolic disease with minimal side effects. The most commonly consumed magnesium-containing foods among those in the HF group were almonds and legumes (beans), whereas those in the LC group likely consumed the majority of their magnesium in the form of peanut butter (data not shown). Almonds, cooked spinach, and cashews are considered good sources of magnesium (≥20%DV) (USDA, 2020). HF dieters consumed more almonds, which may partly explain the differences in consumption across groups.

| Potassium
It has been well established that those with obesity are at an increased risk for developing hypertension (Koliaki & Katsilambros, 2013;Kotsis, Stabouli, Papakatsika, Stabouli, Papakatsika, Rizos, & Parati, 2010). The physiological mechanisms by which this occurs are complex and involve multiorgan systems. Those with obesity may experience low-grade chronic inflammation which in turn may increase insulin resistance, endothelial resistance, and sodium retention (Kotsis et al., 2010;Ramos et al., 2003). The IOM (2005) has established an AI of 1,200-1,500 mg/day and a tolerable upper limit of 2,300 mg/day. A meta-analysis conducted by the World Health Organization (WHO) (2012) revealed that normotensive individuals and those with hypertension that consume > 2,000 mg/day may increase their risk of developing hypertension. However, increased potassium consumption may offset this effect (Geleijnse, Kok, & Grobbee, 2003) by decreasing blood volume through vasodilation while increasing the rate of sodium excretion (IOM, 2005).
Reductions in blood pressure have been seen in those that regularly consume potassium-rich foods such as fruits and vegetables (Champagne, 2006) which is preferred to supplementation (Appel et al., 2006). However, results from national surveys estimated that among those 20 years of age or older, mean daily potassium intakes from foods were 3,016 mg for men and 2,320 mg for women (Oria, Harrison, & Stallings, 2019), which fall below the respective AI (3,400 mg/day for males; 2,600 mg/day for females).
In our study, those assigned to the HF group consumed more potassium when compared with those in the LC group. However, participants in both groups consumed less than the AI. The most commonly consumed food sources of potassium among HF dieters were beans, lentils, tomatoes, and bananas. In contrast, LC diet- ers regularly consumed chicken, cheese, and Greek yogurt (data not shown). Apricots are considered good sources of potassium (≥ 20%DV), whereas lentils, kidney beans, and bananas are considered moderate sources (9%-16%DV) (USDA, 2020). HF participants were encouraged to increase their intake of dietary fiber by consuming beans and lentils, which specifically may have led to the observed differences. After 1 year, participants in the HF group did not experience significant reductions in either diastolic or systolic blood pressure, despite their potassium consumption (Table 1).

| Study strengths and limitations
The study design, including randomization of participants and its duration, were key strengths. Regular follow-up visits with an RD and the collection of multiple 24-hr dietary recall data clearly showed diet compliance among those in both groups. This was true despite the fact that, historically, adherence to prescribed diets among study participants is quite low (Moyad, 2005

ACK N OWLED G M ENTS
We also thank Cathy Nehl, Truitt Brothers, Haricot Farms, and Central Bean Co. for their donation of beans for the study. We would also like to thank Stephanie Gomez for her assistance. We have received permission from those named to be acknowledged.

CO N FLI C T O F I NTE R E S T
The authors declare that there are no conflicts of interest.

TR A N S PA R EN C Y D ECL A R ATI O N
The lead author affirms that this manuscript is an honest, accurate, and transparent account of the study being reported.