Effect of a Dietary Portfolio of Cholesterol-Lowering Foods Given at 2 Levels of Intensity of Dietary Advice on Serum Lipids in Hyperlipidemia:  A Randomized Controlled Trial FREE

Agencies concerned with cardiovascular health have uniformly stressed the importance of diet and lifestyle as the primary means of lowering serum lipids and coronary heart disease (CHD) risk.^1- 3 Nevertheless, the introduction of statins in the late 1980s highlighted the relative ineffectiveness of conventional dietary advice.⁴ As a result, alternative dietary paradigms have been proposed^5- 7 and efforts have been made to enhance the ability of conventional dietary therapy to reduce serum cholesterol through the inclusion of specific foods or food components with known cholesterol-lowering properties,^1,8 singly or in combination (dietary portfolio).⁹ Use of these cholesterol-lowering dietary components in combination in short-term studies with all food provided has been shown to reduce serum low-density lipoprotein cholesterol (LDL-C) to a similar extent as first-generation statins.^10- 11 The longer-term effect of such diets compared with conventional dietary advice has not been assessed.

We therefore undertook a multicenter trial to determine whether advice to eat a dietary portfolio consisting of foods recognized by the US Food and Drug Administration (FDA) as associated with lowering serum cholesterol achieved significantly greater percentage decreases in LDL-C compared with a control diet at 6-month follow-up. The control diet emphasized high fiber and whole grains but lacked portfolio components. To increase the relevance of the study for routine clinical application, the advice was given at 2 levels of intensity, either as a routine dietary portfolio (2 clinic visits of 40- to 60-minute sessions) or an intensive dietary portfolio (7 clinic visits of 40- to 60-minute sessions).

Three hundred fifty-one participants with hyperlipidemia (137 men and 214 postmenopausal women) were randomized after recruitment through advertisement by 4 participating centers across Canada (Quebec City, Toronto, Winnipeg, and Vancouver). Six randomized participants (3 men and 3 women) were either withdrawn for medical reasons before starting the study (n = 3) or failed to attend for baseline measurements (n = 3), resulting in 345 participants with data available for the final (modified intention-to-treat) analysis (Figure 1). Inclusion criteria included men and postmenopausal women in the low (0%-10%) and intermediate (10%-19%) Framingham 10-year risk categories who had LDL-C values ranging from 135 to 205 mg/dL and 116 to 178 mg/dL, respectively (to convert to millimoles per liter, multiply by 0.0259). Exclusion criteria included a history of cardiovascular disease, cancer or a strong family history of cancer, untreated hypertension (blood pressure >140/90 mm Hg), diabetes, renal or liver disease, and currently taking lipid-lowering medications.

Recruitment was achieved by advertising in newspapers, subway cars, lipid clinics, and family practice offices, and randomization took place between June 25, 2007, and February 19, 2009. Participants were stratified by center, sex, and pretreatment LDL-C levels (≥158 vs <158 mg/dL), and randomized in blocks of 75 participants. The blocks were created by a statistician (E.V.) and the allocations were placed in numbered envelopes to be opened by the dietitian in the presence of the participants. Neither the dietitian nor the participants could be blinded to treatment; however, the statistician, investigators, and laboratory staff who analyzed the samples were unaware of participant allocation. Treatments were coded as 1, 2, and 3. Participants were randomized to a parallel-design study of dietary advice to take either a therapeutic low-fat diet (control) or dietary portfolio of cholesterol-lowering foods with either 2 visits (routine) or 7 visits (intensive) during a 6-month period.

Study visits occurred at baseline (week 0) and at 3 and 6 months for the control and routine dietary portfolio interventions, and baseline, 2 weeks, and subsequently at monthly follow-up for the intensive dietary portfolio intervention (eFigure). At each study visit, the preceding 7-day diet histories were assessed by the dietitian and discussed with the participant. Body weight was measured and a fasting blood sample was obtained (eFigure). At each visit, blood pressure was measured 3 times in the nondominant arm using a digital blood pressure monitor (Omron HEM-907XL, Omron Healthcare Inc, Vernon Hills, Illinois). Participants completed a race/ethnicity questionnaire at week 0 with predefined options to determine the applicability of the results to the general population.

The study was approved by the ethics committees of St Michael's Hospital and the universities of Toronto, Manitoba, British Columbia, and Laval, and the National Health Products Directorate, Health Canada. Informed consent was obtained in writing from the participants.

Table 1 shows diets for participants in each study group before enrollment. During the 6-month study period, dietitians counseled participants to follow weight-maintaining vegetarian diets from foods available in supermarkets and health food stores. Counseling periods were 1-hour duration for the first visit and 30 to 40 minutes at subsequent visits. For participants in the dietary portfolio interventions, dietitians focused on incorporating study foods (eTable 1) into the participants' regular diets using their 7-day food diaries as templates. Participants were provided with a 7-day study food checklist and an illustrated study booklet. The goal of the dietary portfolio was to provide 0.94 g of plant sterols per 1000 kcal of diet in a plant sterol ester–enriched margarine; 9.8 g of viscous fibers per 1000 kcal of diet from oats, barley, and psyllium; 22.5 g of soy protein per 1000 kcal as soy milk, tofu, and soy meat analogues; and 22.5 g of nuts (including tree nuts and peanuts) per 1000 kcal of diet. Consumption of peas, beans, and lentils was also encouraged. This dietary portfolio has been described in detail previously.¹⁰

The control dietary advice focused on low-fat dairy and whole grain cereals together with fruit and vegetables and avoidance of the specific portfolio components. Participants were provided with measuring cups and measuring spoons to assist in portion control.

All samples from a given individual were labeled by code and analyzed in the routine laboratory of the hospital for the Toronto and Vancouver centers (Beckman SYNCHRON LX Systems, Mississauga, Ontario, Canada, and Advia 1650, Siemens, Deerfield, Illinois, respectively) and the research laboratories of the Quebec and Winnipeg centers (Roche Hitachi 917 Chemistry Analyzer, Roche Diagnostics GmbH, Mannheim, Germany, and Vitros 350, Orthagonal Diagnostics, Johnson & Johnson, Markham, Ontario, Canada). Serum lipid standards (Solomon Park Research Laboratories, Kirkland, Washington) were used to quality control the lipid analyses at the 4 collaborating sites. The mean interbatch coefficients of variation at each site were all less than 3.9% for total cholesterol (TC), LDL-C, high-density lipoprotein cholesterol (HDL-C), and triglyceride levels. LDL-C level was calculated using the Friedewald equation¹² unless triglyceride levels were higher than 398 mg/dL, in which case the prebaseline LDL-C value was used for baseline but no LDL-C value was ascribed for the 2 participants with increased triglyceride levels at week 24. Serum samples, stored at −70°C, were analyzed in the J. Alick Little Lipid Laboratory at St Michael's Hospital, Toronto, Ontario, Canada, for apolipoprotein A-l and B by nephelometry (intra-assay coefficients of variation, 2.2% and 1.9%, respectively)¹³ and C-reactive protein (CRP) by end point nephelometry (coefficient of variation, 3.5%) (N high sensitivity CRP reagent; Dade-Behring, Mississauga, Ontario, Canada). Plant sterols in serum were measured in the sterol laboratory of the Richardson Center by gas chromatography with flame ionization detectors (6890 GC, Agilent Technologies, Palo Alto, California).¹⁴

Diets were analyzed using a program based on US Department of Agriculture data (ESHA Food Processor SQL version 10.1.1; ESHA, Salem, Oregon) with addition of data on foods relevant to ongoing studies.

Adherence with the 4 portfolio components was estimated from the 7-day food records by expressing the recorded intake for each of the 4 main components as 25%. The sum of the 4 components if consumed as prescribed would equal 100% adherence.

A modified intention-to-treat analysis was undertaken on the 345 participants to whom randomization had been revealed. Missing data were multiple imputed using PROC MIXED and MIANALYZE in SAS version 9.2 (SAS Institute Inc, Cary, North Carolina). Multiple imputation assumes the values imputed are randomly sampled from the distribution of true (unobserved) missing values. This process results in valid statistical inferences that properly reflect the uncertainty due to missing values, producing valid standard errors (and 95% confidence interval [CI]) for parameters. For example, using LDL-C at 24 weeks, the strongest predictors of missingness at 24 weeks were age (β = 0.0454, P = .01) and sex (women vs men) (β = −0.2807, P = .11), and these, along with the observed values, were used to inform the imputations. Site and diet assignment did not predict missingness (P ≥ .16). PROC MIXED was used to determine the significance of the change in outcome from week 0 to week 24 as the response variable with treatment, sex, and treatment × sex as main effects and baseline as a covariate. A Tukey adjustment was used for multiple comparisons.¹⁵ Percentage change in LDL-C adjusted for sex was the primary outcome. In addition, the 10-year Framingham CHD risk score was calculated¹⁶ for this predominantly white cohort. Pearson correlation was used to calculate the correlation coefficient between dietary adherence and percentage change in LDL-C. χ² Test was used to determine the differences in participants' categorical baseline characteristics.

One hundred ten participants were required per treatment group. Assuming a maximum effect size of 10% (17 mg/dL) for percentage LDL-C change, a 10% SD of effect with α = .05 and 1−β = .80, and a 20% attrition rate, we had sufficient power to detect a 5% change (9 mg/dL) in LDL-C between treatments.

The study was conducted between 2007 and 2009, and randomization took place between June 25, 2007, and February 19, 2009. The baseline characteristics of the participants were similar for all 3 treatments, with the exception of the ratio of men to women, which was higher on the intensive portfolio than on the other 2 treatments (Table 2). No participants were taking medications known to influence serum lipids, except 31 women and 4 men who were all receiving stable doses of thyroxine and 12 women who were receiving estrogen therapy. Fifty-one participants (14%) had been taking statins before the study commenced and had discontinued taking the medications at least 2 weeks before the study (Table 2).

The mean overall adherence for all participants to the intensive dietary portfolio was 46.4% (95% CI, 40.4%-52.4%) and to the routine dietary portfolio was 40.6% (95% CI, 34.6%-46.6%). Participants lost a similar amount of weight while taking part in all 3 treatments (intensive dietary portfolio, −1.2 kg; 95% CI, −1.9 to −0.6 kg; P < .001; routine dietary portfolio, −1.7 kg; 95% CI, −2.4 to −1.0 kg; P < .001; and control, −1.5 kg; 95% CI, −2.1 to −0.8 kg; P < .001) (Table 3).

No differences were observed between the 3 treatment groups in baseline blood measurements. In the modified intention-to-treat analysis, the percentage changes from baseline to week 24 in the control diet were −3.0% (95% CI, −6.1% to 0.1%; P = .06) or −8 mg/dL (95% CI, −13 to −3 mg/dL; P = .002) for LDL-C and −1.4% (95% CI, −4.3% to 1.6%; P = .37) or −0.09 (95% CI, −0.24 to 0.05; P = .18) for TC:HDL-C ratio. In the routine and intensive dietary portfolio interventions, the respective percentage changes were −13.1% (95% CI, −16.7% to −9.5%; P < .001) or −24 mg/dL (95% CI, −30 to −19 mg/dL; P < .001) and −13.8% (95% CI, −17.2% to −10.3%; P < .001) or −26 mg/dL (95% CI, −31 to −21 mg/dL; P < .001) for LDL-C; and −8.2% (95% CI, −11.2% to −5.3%; P < .001) or −0.37 (95% CI, −0.52 to −0.23; P < .001) and −6.6% (95% CI, −9.8% to −3.4%; P < .001) or −0.39 (95% CI, −0.56 to −0.24; P < .001) for TC:HDL-C ratio, respectively (Figure 2). The percentage change and absolute treatment differences between the control and both the dietary portfolio interventions were significant for LDL-C (P < .001) and in absolute units for the TC:HDL-C ratio (P = .004 for intensive dietary portfolio and P = .006 for routine dietary portfolio), with no significant differences between the dietary portfolios (P = .66) (Table 3). The apolipoproteins reflected the lipid and lipoprotein changes (Table 3). No significant differences were observed between treatments in CRP. No treatment differences in response were observed between men and women.

The intensive dietary portfolio led to a nonsignificant reduction in systolic blood pressure of 2.6 mm Hg (95% CI, −5.4 to −0.2 mm Hg; P = .07) and a significant reduction in diastolic blood pressure of 2.1 mm Hg (95% CI, −3.7 to −0.4 mm Hg; P = .01) compared with the control diet (Table 3).

The routine dietary portfolio reduced the calculated 10-year CHD risk by 10.8% (95% CI, −16.8% to −5.0%; P < .001; absolute risk change, −0.9%; 95% CI, −1.4% to −0.5%), with a comparable reduction in the intensive dietary portfolio of −11.3% (95% CI, −17.1% to −5.4%; P < .001; absolute risk change, −1.2%; 95% CI, −1.5% to −0.8%). These percentage reductions were significantly different (P = .02 and P = .01, respectively) from the nonsignificant decrease in the control diet (−0.5%; 95% CI, −6.0% to 5.0%; P = .87; absolute risk change, −0.3%; 95% CI, −0.7% to 0.1%; P = .12) (Table 3).

The mean percentage reductions in LDL-C were significant at −15.0% (95% CI, −18.6% to −11.4%; P < .001) or −27 mg/dL (95% CI, −32 to −22 mg/dL) for routine dietary portfolio and −15.5% (95% CI, −19.0% to −12.0%; P < .001) or −28 mg/dL (95% CI, −32 to −22 mg/dL) for intensive dietary portfolio, but not for the control diet (−2.5%; 95% CI, −6.0% to 1.0%; P = .16; or −9 mg/dL; 95% CI, −14 to −4 mg/dL). Both portfolio treatments were different from the control (P < .001) but were similar to each other.

Overall adherence with the 4 principal portfolio components (nuts, soy, viscous fiber, and plant sterol) was significantly associated with the percentage reduction in LDL-C in participants who completed the study (r = −0.34, n = 157, P < .001).

The mean percentage LDL-C response of the 2 portfolio treatments for Vancouver (−19.9%; 95% CI, −25.0% to −14.8%; or −33 mg/dL; 95% CI, −41 to −25 mg/dL) differed significantly from both Toronto (−11.6%; 95% CI, −15.6% to −7.6%; or −23 mg/dL; 95% CI, −30 to −17 mg/dL) and Winnipeg (−11.1%; 95% CI, −16.4% to −5.8%; or −22 mg/dL; 95% CI, −30 to −14 mg/dL) for the difference in percentage reduction between centers (P = .01 and P = .02, respectively), and the difference between Vancouver and Quebec (−12.7%; 95% CI, −18.2% to −7.2%; or −26 mg/dL; 95% CI, −35 to −17 mg/dL) was of borderline significance for the percentage reduction in LDL-C (P = .06). Vancouver (48.3%) differed from Winnipeg (34.0%) in terms of adherence (P = .001).

There were no serious adverse events or events that required hospitalization (eTable 2). However, 36 minor events were reported during the study (15 in the intensive dietary portfolio, 12 in the routine dietary portfolio, and 9 in the control diet), with no treatment difference in event number (P = .20). None of the events were directly linked to the study intervention, except for 1 man who had recurrent facial flushing and itching at the back of the neck and was found to have positive skin test for soy in the routine dietary portfolio, and 1 woman with a rash and positive skin test for soy and almonds in the intensive dietary portfolio.

Our data demonstrate the cholesterol-lowering potential of a dietary portfolio intervention that counsels participants to increase consumption of cholesterol-lowering foods denoted by the US FDA to have a heart health benefit and that have also been recommended in national guidelines to enhance the effectiveness of cholesterol-lowering therapeutic diets.^1,10 Our study also represents the first randomized trial to our knowledge to assess the ability of an intervention that counsels for consumption of these cholesterol-lowering foods to reduce LDL-C at 6-month follow-up in real-world conditions. The reductions in LDL-C in the dietary portfolio intervention were approximately half those observed with early statin trials that were associated with 20% reductions in CHD mortality.¹⁷

Although there are no data on plant sterol consumption and reduction in CHD risk, cohort studies have consistently shown that consumption of 5 servings of nuts a week, similar to the 26 to 31 g/d consumed in our study, is associated with a decrease in CHD events by 40% to 60%.^18- 19 Fiber intake has been negatively associated with CHD risk.^20- 21 Increased consumption of vegetable sources of protein and fat is associated with reductions in CHD risk.²² Similar associations were identified for soy protein consumption in the Shanghai cohort.²³ Further study is needed to determine whether cholesterol reduction using these portfolio components is associated with lower rates of CHD events.

The specific food components used in the portfolio have well-established cholesterol-lowering properties and are recognized by the US FDA as justifying a heart health claim.¹⁰ Review articles and meta-analyses have confirmed LDL-C benefits for viscous fibers,^24- 25 plant sterols,^26- 27 soy protein,^28- 30 and nuts.³¹

On the basis of the reported intake of portfolio components, one might expect a 4% LDL-C reduction from viscous fiber, 2% each from nuts and soy, and 5% from plant sterols, resulting in 13% LDL-C reduction. A reduction of a similar magnitude (13%-14%) was observed in our study.

This is the first study to our knowledge to assess the effect of frequency of visit on dropout rate, adherence, and outcome. Although a small reduction was observed in the dropout rate by increasing the frequency of visits from 2 to 7 during the 6-month period, no advantage was observed in terms of dietary adherence or the percentage LDL-C reduction at 24 weeks. More frequent clinic attendance therefore appears to be unnecessary in achieving a significant percentage reduction in LDL-C. The near maximal effectiveness of only 2 clinic visits enhances the suitability of this dietary approach for clinical application.

The study had limitations. First, the intervention was complex. Second, colinearity between the different dietary components did not permit attribution of the lipid-lowering effect to specific components of the intervention. Third, the study was not metabolically controlled in terms of providing all food to the participants. However, our goal was to assess the effect of dietary advice in real-world conditions. Studies of longer duration in which provision of specific diets was possible have relied on workplace or institutional environments.^7,32- 34 Fourth, the study had a high overall dropout rate of 22.6%. This attrition rate is common to dietary studies provided at these levels of intensity.^35- 37 In addition, participants were predominantly white with low to intermediate risk of cardiovascular disease and relatively low mean body mass index levels. The generalizability of this clinical trial to higher-risk, more overweight, or obese patient populations is unknown.

The study advantages include its multicenter nature with centers from across the continent. Participants who joined the study were already consuming an acceptable background diet low in saturated fat and cholesterol to provide a fairer illustration of the type of patients for whom standard dietary advice has failed to achieve therapeutic targets. This approach may underestimate the effectiveness of the diet when applied to those individuals who are not already following therapeutic diets. However, it is also possible that participants in this study are better able to adhere to healthy diets than those who chose not to participate. Many of the foods have other attributes, including lowering the glycemic index, which may aid in reducing disease risk for CHD, diabetes, and obesity.³⁸ In addition, the dietary portfolio treatments lowered LDL-C without also lowering HDL-C.

In conclusion, this study indicated the potential value of using recognized cholesterol-lowering foods in combination. We believe this approach has clinical application. A meaningful 13% LDL-C reduction can be obtained after only 2 clinic visits of approximately 60- and 40-minute sessions. The limited 3% LDL-C reduction observed in the conventional diet is likely to reflect the adequacy of the baseline diet and therefore suggests that larger absolute reductions in LDL-C may be observed when the dietary portfolio is prescribed to patients with diets more reflective of the general population.