Two Years of Calorie Restriction and Cardiometabolic Risk Factors

Abstract

BACKGROUND

For several cardiometabolic risk factors, values considered within normal range are associated with increased risks of cardiovascular morbidity and mortality. However, very little is known about the short- and long-term effects of caloric restriction (CR) with adequate nutrition on these risk factors in healthy lean or slightly overweight young and middle-aged individuals.

METHODS

Cardiometabolic risk factor responses to a prescribed 25% CR diet for 2-years were evaluated in a multicenter randomized controlled trial in 218 young and middle-aged (21 to 50 y), healthy non-obese (body mass index 22 to 27.9 kg/m²) men and women.

RESULTS

Over two years, participants in the CR group achieved 11.9% CR and a sustained 10% weight loss, of which 71% was fat mass loss. CR caused a significant and persistent reduction of all measured conventional cardiometabolic risk factors, including LDL-cholesterol, total cholesterol to HDL-cholesterol ratio, systolic and diastolic blood pressure. In addition, CR resulted in a significant improvement in C-reactive protein, glucose tolerance, insulin sensitivity index, and metabolic syndrome score relative to control. A secondary analysis revealed the responses to be robust after controlling for relative weight loss changes.

CONCLUSIONS

Two years of moderate caloric restriction significantly reduced multiple cardiometabolic risk factors in young, non-obese men and women. These findings suggest the potential for significant cardiovascular advantage of practicing moderate CR in young and middle-aged healthy individuals, and they offer promise for significant long-term population health benefits.

INTRODUCTION

Cardiovascular disease (CVD) is the leading cause of world-wide morbidity, disability and death. Approximately 2,200 men and women die daily in the US due to CVD — an average of one death every 40 seconds.¹ In Rhesus monkeys, young-onset calorie restriction (CR) reduces the risk of developing and dying of CVD by at least 50%.^2,3,4 In observational human studies, severe CR provides a powerful protective effect against multiple atherosclerotic risk factors.⁵ These effects include less carotid artery intima-media thickening; improved left ventricular diastolic function; and increased beneficial heart rate variability than matched controls not practicing CR.^5–7 However, little is known about the extent of cardiometabolic adaptations in response to CR of greater than one year in healthy non-elderly normal weight or slightly overweight human individuals. Critically important data are needed on the effects of CR in early life, corresponding to the studies in animal models, on cardiometabolic health. This represents an important novel aspect of this study.

From risk factor values well below the conventional clinical disease thresholds, cardiovascular incidence and mortality risk increase continuously for all conventional cardiometabolic risk factors;^8–13, and individuals with optimal cardiometabolic health experience a dramatically lower lifetime risk of developing cardiovascular disease — even in the face of a much longer average lifespan.¹⁴

The two-year multicenter randomized controlled Comprehensive Assessment of Long-term Effects of Reducing Intake of Energy (CALERIE) trial was designed to evaluate the potential for CR to promote anti-aging adaptations in resting metabolic rate and body core temperature; findings for these outcomes have been previously reported.¹⁵ Here, our objective was to evaluate — in all persons completing the 2-year CALERIE protocol — the short- and long-term effects of randomization to a 25% CR regimen on multiple cardiometabolic risk factors implicated in the development and progression of atherosclerotic CVD in a healthy, young and middle-aged (21 to 50 years old), non-obese (BMI, 22–27.9 kg/m²) population of men and women.

METHODS

Study population

CALERIE was a large multicenter randomized, controlled trial aimed at evaluating the time-course effects of 25% CR below the individual’s baseline level over a two-year period in healthy normal weight and slightly overweight (BMI, 22 to 27.9 kg/m²) young men (21 to 50 y) and premenopausal women (21 to 47 y).¹⁶ The study protocol (No. NCT00427193) was approved by Institutional Review Boards at three clinical centers — Washington University Medical School (St. Louis, MO), Pennington Biomedical Research Center (Baton Rouge, LA), Tufts University (Boston, MA) — and coordinating center at Duke University (Durham, NC). Study volunteers provided written informed consent. Study oversight was provided by a Data and Safety Monitoring Board. The CONSORT diagram for enrolled participants is available in Figure S1 in the Supplementary Appendix.

Treatment assignment and intervention

After baseline testing, participants were randomized in a ratio of 2:1 to a CR behavioral intervention designed to achieve a predicted longitudinal weight loss trajectory estimated from previous data of 25% CR in a phase I pilot study, or to an ad libitum (AL) control group; randomization was stratified by site, sex, and BMI.¹⁵ Extensive details of the intervention are provided in several publications.^15–17 Participants in CR were prescribed a 25% restriction in calorie intake based on energy requirements estimated from doubly labeled water measurements over a four-week period at baseline. Participants were fed in house at the three clinical centers for one month, during which they were instructed on the essentials of CR. They chose from one or more of six eating plans modified from different ethnic sources. Participants continued one in house meal provided with intensive group and individual behavioral counseling sessions. There were 24 group and individual counseling sessions over the first 24 weeks of the intervention (Table 1 of that manuscript); there was a detailed algorithmic strategy to monitor and respond to changes in adherence (Clinical Tracking System (CTS) conducted by behavioral interventionist. Adherence to the CR intervention was determined in real-time by the degree to which individual weight change adhered to an individualized weight loss trajectory (15.5% weight loss at one yea followed by weight loss maintenance). Additionally, the precise level of calorie restriction was retrospectively quantified by calculating the total daily energy expenditure (TDEE) by doubly-labeled water and adjusting TDEE for changes in body composition.¹⁸

Table 1.

Baseline study subject characteristics (Mean (SD)^*

	Ad Libitum (n=75; 22M)	Calorie Restriction (n=143; 44M)
Race
White	57 (76.0%)	111 (77.6%)
African American	11 (14.7%)	15 (10.5%)
Other	7 (9.3%)	17 (11.9%)
Age, y	37.9 (6.94)	38.0 (7.34)
Height, m	168.4 (8.31)	168.9 (8.60)
Baseline Weight, kg	71.5 (8.65)	72.0 (9.49)
Baseline BMI, kg/m2	25.1 (1.64)	25.2 (1.78)
Body Fat, %	33.6 (6.57)	32.9 (6.07)
Fat free mass, kg	47.6 (8.61)	48.5 (9.21)
Energy and macronutrient intake
Energy intake, kcal/d	2390 (384.8)	2467 (405.6)
Protein, g/kg/d	1.2 (0.04)	1.2 (0.02)
Protein, % of energy	17.2 (3.48)	16.6 (3.04)
Fat, % of energy	34.7 (5.12)	33.5 (4.93)
Carbohydrates, % of energy	45.1 (6.33)	46.8 (6.48)

^*There were no significant between-group differences in baseline characteristics

Body weight and body composition

With the subject wearing a hospital gown and no shoes, body weight was measured in duplicate in the morning following a twelve-hour fast. Height was measured without shoes to the nearest 0.005 m (0.5 cm) and BMI calculated. Fat mass (FM) and fat-free mass (FFM) was measured by dual-energy x-ray absorptiometry (DXA; Hologic 4500A, Delphi W or Discovery A scanners) with all scans analyzed at University of California San Francisco. Scanner performance was monitored with baseline and longitudinal phantom cross calibrations.

Dietary intake

Dietary intakes were determined using six-day food diaries analyzed with Nutrition Data System for Research (Minneapolis, MN) by a central reading center at the University of Cincinnati.

Blood chemistries

Venous blood was sampled for lipid and hormone concentrations after an overnight fast. Samples were collected in EDTA plasma tubes, immediately centrifuged to separate the plasma, aliquoted, and stored in −80°C freezer until use. Serum lipid and lipoprotein-cholesterol concentrations were determined in the Laboratory for Clinical Biochemistry at the University of Vermont. Cholesterol and glycerol-blanked triglycerides were measured with automated enzymatic commercial kits (Miles-Technicon, Tarrytown, NY). High-density lipoprotein cholesterol (HDL) was measured in plasma after precipitation of apolipoprotein B-containing lipoproteins by dextran sulfate (50000 MW) and magnesium. Low-density lipoprotein cholesterol (LDL) was calculated using the Friedewald Equation.¹⁹ High sensitivity C-reactive protein (hsCRP) was measured using a high sensitivity ELISA kit (ALPCO Diagnostics Ltd, Windham, NH).

Blood pressure

Blood pressure was measured with an oscillometric blood pressure monitor (Dinamap Procare 200, GE Healthcare, Waukesha, Wisconsin, USA) in the morning after 12 hours fast in a seated position. Blood pressure was measured according to a common procedure; the measurement process was regularly monitored to assure protocol adherence. Three sequential measurements were obtained at each sitting. If one differed by the other two by ≥15 mmHg, it was discarded. Otherwise the three measures were averaged.

Oral Glucose Tolerance Testing (OGTT)

Two-hour, 0.075-kg oral glucose tolerance tests were performed at baseline and at 12 and 24 months with blood samples collected at baseline, and 30, 60, 90 and 120 minutes after the glucose consumption. Adequate carbohydrate intake over the prior three days (>150 g/d) was ensured by prescription and subsequent interview with a study dietician. Plasma glucose was measured by the glucose oxidase method (YSI Instruments, Fullerton, CA); insulin was measured by chemiluminescent immunoassay (Elecsys, Roche Diagnostics, Indianapolis, IN).

Insulin resistance was calculated by using homeostasis model assessment (HOMA) (HOMA-IR = [fasting glucose {mmol/l} × fasting insulin {mU/L}]/22.5).²⁰ Beta-cell function was calculated using HOMA-beta (%) = [360 * fasting insulin (mU/L)]/[fasting glucose (mg/dL)-63].²⁰ Area under the curve insulin (AUC-insulin) and area under the curve glucose (AUC-glucose) values from the OGTT were determined using the trapezoidal method.²¹

Other derived glucose homeostasis measures

Insulin response was calculated at the ratio of change in plasma insulin from baseline to 30 minutes to the change in plasma glucose over the same period (= ΔI0min-30min/ΔG0min-30min). Insulin sensitivity index was calculated as 1/fasting insulin.^22,23 The Oral Disposition Index (DIo) was calculated as the product of insulin response and insulin sensitivity.^24,25

Metabolic Syndrome Score

Sex-specific metabolic syndrome scores were developed for each subject at baseline, at 12 and 24 months using the method outlined below referenced to the standard deviation (SD_value) of the entire CALERIE study population at baseline.^26,27 Metabolic syndrome score (MSS) was developed using mean blood pressure (MBP; = [2 × diastolic blood pressure + systolic blood pressure]/3), high density lipoprotein cholesterol (HDL), triglycerides (TG), waist circumference (WC) and fasting blood glucose (FBG):

Women: MSS = [45-HDL]/SD_HDLW+ [TG-150]/SD_TG + [WC-88]/SD_WCW + [FBG-100]/SD_FBG + [mean BP-100]/SD_MBP

Men: MSS = [40-HDL]/SD_HDLW+ [TG-150]/SD_TG + [WC-102]/SD_WCW + [FBG-100]/SD_FBG + [mean BP-100]/SD_MBP

Statistical analysis

The same statistical methodologies used in the parent CALERIE study were applied.^18,28 Intention-to-treat analyses were performed by including all available observations in the analysis. Wilcoxon and Fisher exact tests were used to evaluate between-group differences with respect to baseline characteristics. Repeated measures analysis of covariance, as implemented under mixed models,²⁹ was applied with change from baseline as the dependent variable, and treatment, time, and the treatment × time interaction as independent variables. The approximate normality of each outcome and of the change score of the outcome was confirmed by examination. Site, sex, BMI stratum (normal weight, i.e., 22.0 to 24.9 kg/m², and overweight, i.e., 25.0 to 27.9 kg/m²), and the baseline value were included as covariates to insure statistical balance not captured by randomization; and more important, to reduce error variance.³⁰ To avoid arbitrary modeling assumptions with respect to linearity, time was treated as a categorical variable; similarly, an unstructured model was applied for the covariance matrix among the repeated observations. Hypotheses of specific interest — e.g., between-group differences at the individual time points and within-group changes over time — were tested by defining contrasts among the regression parameters; predicted mean change ± SE are the adjusted values from this model. For any outcome, type-I error was controlled using a hierarchical gatekeeping strategy.³⁰ The treatment-by-visit interaction term was tested first. If significant, between-group differences at each time point were tested at α=0.05. If not significant, the treatment main effect was tested next. If significant, then between-group differences at each time point were tested at α=0.05. Otherwise, Bonferroni correction was applied at each time point, with p-values adjusted by multiplying the nominal p-value by the number of tests (truncated at 1.0).¹⁸

To address the robustness of the Intention-to-treat analysis — specifically with respect to the issue of non-random of missing data due to differential drop out and adherence to intervention arms — we performed Marginal Structural Modeling as previously described.³² This approach allowed us to test the effects of our findings using intent-to-treat modeling versus the effects considering full compliance and effect with 25% CR through the entire study.

Role of the Funding Source

The CALERIE Study was conducted under the National Institutes of Health (NIH) U-grant mechanism, which implies that the study design and conduct was collaborative effort between internal NIH and external study investigators.

RESULTS

Of the more than 10,000 men and women assessed for eligibility, the screening procedures excluded 45% for age or BMI; 14% for health or medication reasons; and 30% refused to participate due to concerns about their ability to adhere to the protocol, personal, or other study-related issues.¹⁵ Of the 238 participants who began baseline assessments, 220 were randomized; 218 started the assigned intervention; 82% of CR and 95% of AL, respectively, completed the study (Figure S1 in the Supplementary Appendix). Table 1 presents the baseline characteristics of the 218 participants with means (±SD). As expected there were no statistically significant differences between groups at baseline for all presented variables. Effects of CR on resting metabolic rate, core body temperature, and hormones were reported previously¹⁵.

Intervention adherence, body weight and composition

Detailed information regarding the observed adherence to the intervention has been published previously.¹⁵ Also shown in Table 1, there was a very small mean difference of 77 kcal/d in average energy intake at baseline (p=0.15) — assessed as TDEE during a period of weight stability before the intervention — between the CR (2467±34 kcal/d; mean ± SE) and AL (2390±45 kcal/d) groups. Of note, although participants were recruited to be within a BMI range of 22 to 28 kg/m² for this study, men had particularly high percent body fats for men (26.0 ± 0.6) and moderately high percent body fats for women (36.3 ± 0.5). In the CR group, energy intake was reduced by 19.5±0.8% (480 kcal/day) during the first six months; decreased over time as might be expected to 9.1% (±0.7%) after 6 months, but averaged 11.7% (±0.7%) over the entire, two-year intervention. In the control group, average daily energy intake was unchanged over the period (data not shown).

Changes in body composition by intervention group and by sex are detailed in a previous report.³¹ In sum, weight loss from baseline averaged 8.4±0.3 kg (11.5%) at one year and 7.5±0.3 kg (10.4%) at two years in the CR group (P<0.001); it did not change significantly in the AL group (Table 2). Body fat decreased from baseline by 6.1±0.2 kg at one year and 5.3±0.3 kg at two years in the CR group (P<0.001); it did not change in the AL group. Percentage weight loss did not differ by sex. Percent responses in waist circumference, body mass index, total body fat and appendicular fat to caloric restriction did not differ by sex.³¹ Fat loss at two years accounted for approximately 71% of the weight loss (Table 2).

Table 2.

Body composition changes throughout the CALERIE study

	AL (n=75)	CR (n=143)
	mean (SE)	mean (SE)	Between group P value
Body Weight (kg)
Baseline	71.5 (1.0)	72.0 (0.8)	0.98
Δ Month 12	−0.7 (0.4)*	−8.4 (0.3)***	<0.001
Δ Month 24	0.1 (0.5)	−7.5 (0.4)***	<0.001
Body Mass Index (kg/m2)
Baseline	25.1 (0.2)	25.2 (0.2)	0.94
Δ Month 12	−0.2 (0.1)	−2.9 (0.1)***	<0.001
Δ Month 24	0.1 (0.2)	−2.6 (0.1)***	<0.001
% Body Fat
Baseline	33.6 (0.8)	32.9 (0.5)	0.34
Δ Month 12	−0.47 (0.3)	−5.5 (0.2)***	<0.001
Δ Month 24	0.13 (0.3)	−4.6 (0.3)***	<0.001
Fat Mass (kg)
Baseline	23.8 (0.6)	23.5 (0.4)	0.61
Δ Month 12	−0.34 (0.3)	−6.1 (0.2)***	<0.001
Δ Month 24	0.38 (0.4)	−5.3 (0.3)***	<0.001
Fat Free Mass (kg)
Baseline	47.6 (1.0)	48.5 (0.8)	0.48
Δ Month 12	−0.3 (0.2)	−2.2 (0.1)***	<0.001
Δ Month 24	−0.2 (0.2)	−2.2 (0.2)***	<0.001

^aBaseline values are the observed mean (SE); change scores are the least-squares adjusted means (SE) from the ITT repeated measures analysis.

^bBetween-group p-value tests for a significant between-group difference in the change score at the time point. All p-values reflect Bonferroni corrections, truncated at 1.0, as appropriate (see text).

*, **, and *** indicated within group differences at the p < 0.05, 0.01, and 0.001 values, respectively

Self-reported energy and nutrient intake

By seven-day food records, the CR group significantly restricted their energy intake (−279±29 kcal/d at one year and −216±33 kcal/d at two years); the AL group maintained its intake at one year (−83±38 kcal/d) and reported −121±43 kcal/d at two years (p=0.005)).

Lipids and lipoprotein profile

Total cholesterol and LDL-cholesterol decreased significantly both at 1 and 2 years in the CR group; it did not in the AL group (Table 3). HDL-cholesterol was increased by CR at one and two years, but the change was significantly different from AL only at year two. Accordingly, the ratio of total cholesterol to HDL-cholesterol decreased significantly and persistently in the CR group (Table 3). CR — but not AL — caused a major drop in serum triglycerides (Table 3).

Table 3.

Lipid and lipoproteins before and after intervention

	AL (n=75)	CR (n=143)
	Mean (SE)	Mean (SE)	Between group p-value
Total Cholesterol (mmol/L)
Baseline	4.52 (0.10)	4.30 (0.06)	0.10
Δ Month 12	−0.04 (0.06)	−0.32 (0.04)***	0.0001
Δ Month 24	0.03 (0.07)	−0.25 (0.05)***	0.0010
LDL-Cholesterol (mmol/L)
Baseline	2.71 (0.09)	2.51 (0.06)	0.0670
Δ Month 12	−0.05 (0.05)	−0.25 (0.04)***	0.0015
Δ Month 24	0.03 (0.05)	−0.23 (0.04)***	0.0001
HDL-Cholesterol (mmol/L)
Baseline	1.26 (0.03)	1.26 (0.03)	0.723
Δ Month 12	0.03 (0.02)	0.07 (0.02)***	0.107
Δ Month 24	0.03 (0.03)	0.11 (0.02)***	0.007
Triglycerides (mmol/L)
Baseline	1.19 (0.08)	1.15 (0.05)	0.97
Δ Month 12	−0.03 (0.05)	−0.29 (0.04)***	<0.0001
Δ Month 24	−0.03 (0.05)	−0.27 (0.04)***	0.0002
Total Cholesterol:HDL-c ratio
Baseline	3.76 (0.128)	3.65 (0.097)	0.47
Δ Month 12	−0.124 (0.061)	−0.505 (0.046)***	<0.0001
Δ Month 24	−0.047 (0.065)	−0.532 (0.050)***	<0.0001

^aBaseline values are the observed mean (SE); change scores are the least-squares adjusted means (SE) from the ITT repeated measures analysis.

^bBetween-group p-value tests for a significant between-group difference in the change score at the time point. All comparisons are controlled for baseline values. All p-values reflect Bonferroni corrections, truncated at 1.0, as appropriate (see text).

*, **, and *** indicated within group differences at the P<0.05, 0.01, and 0.001 values, respectively

Blood pressure

Baseline blood pressure values were low normal in the young non-obese individuals enrolled in the study; nevertheless, CR but not AL resulted in a significant reduction in systolic, diastolic and mean blood pressure (Table 4). The lowering of blood pressure was evident as early as six months; however, it reached statistical significance only at one year and persisted for the entire duration of the study (Table 4).

Table 4.

Blood pressure before and after intervention

	AL (n=75)	CR (n=143)	p-value†
Systolic Pressure (mmHg)
Baseline*	111.2	112.1	0.48
Δ Month 6	−1.21 (0.86)	−2.97 (0.65)***	0.092
Δ Month 12	2.26 (1.03)*	−1.87 (0.77)*	0.001
Δ Month 18	0.60 (0.95)	−2.89 (0.74)***	0.003
Δ Month 24	2.15 (1.06)*	−2.20 (0.82)**	0.001
Diastolic Pressure (mmHg)
Baseline*	71.2	72.1	0.50
Δ Month 6	−2.70 (0.76)***	−3.84 (0.57)***	0.22
Δ Month 12	1.30 (0.74)	−3.38 (0.55)***	<0.0001
Δ Month 18	−0.58 (0.79)	−3.31 (0.61)***	0.005
Δ Month 24	1.55 (0.80)*	−3.40 (0.62)***	<0.0001
Mean BP (mmHg)
Baseline*	84.5	85.4	0.48
Δ Month 6	−2.20 (0.74)**	−3.57 (0.55)***	0.12
Δ Month 12	1.62 (0.76)*	−2.90 (0.57)***	<0.0001
Δ Month 18	−0.19 (0.77)	−3.20 (0.60)***	0.002
Δ Month 24	1.75 (0.81)*	−3.02 (0.63)***	<0.0001
Pulse Pressure (mmHg)
Baseline*	40.0	40.1	0.63
Δ Month 6	1.62 (0.63)*	0.99 (0.47)*	0.41
Δ Month 12	1.15 (0.75)	1.63 (0.55)**	0.60
Δ Month 18	1.35 (0.71)	0.56 (0.55)	0.36
Δ Month 24	0.79 (0.77)	1.32 (0.59)*	0.58

^aBaseline values are the observed mean (SE); change scores are the least-squares adjusted means (SE) from the ITT repeated measures analysis.

^bBetween-group p-value tests for a significant between-group difference in the change score at the time point. All comparisons are controlled for baseline values. All p-values reflect Bonferroni corrections, truncated at 1.0, as appropriate (see text).

*, **, and *** indicated within group differences at the P<0.05, 0.01, and 0.001 values, respectively

Glucose tolerance, insulin action and inflammation

Fasting and AUC-insulin were both significantly reduced in the CR group as compared with the AL group at one and two years (Table 5). Fasting glucose was significantly reduced by CR at year one, but not at year two. In contrast, no significant reduction in the AUC-glucose was observed in this study (Table 5). Nonetheless, CR induced improvements in insulin sensitivity, as reflected in significantly lower HOMA-IR, insulin response (at two years only) and increased insulin sensitivity index (Table 5). Oral disposition increased in the CR group more than in the AL group though the difference between groups did not reach statistical significance. Plasma high-sensitivity C-reactive protein concentrations were significantly reduced in the CR, but not in the control group, at two years (Table 5). CR also resulted in a major and persistent reduction in the metabolic syndrome score (Table 5). CR-induced responses in clinical cardiometabolic risk parameters when controlling for simultaneous changes in body mass are shown in Supplementary Material, Table 1. In sum, there is substantial residual and significant dose-response effects of caloric restriction on cardiometabolic risk factors when controlling for weight loss.

Table 5.

Glucose tolerance, insulin action, and inflammation before and after intervention

	AL (n=75)	CR (n=143)	p-value
Fasting Insulin (μIU/mL)
Baseline	5.79	5.38	0.27
Δ Month 12	−0.14 (0.24)	−1.59 (0.18)***	<0.0001
Δ Month 24	0.14 (0.21)	−1.71 (0.16)***	<0.0001
Fasting Glucose (mg/dL)
Baseline	4.64	4.55	0.12
Δ Month 12	0.02 (0.03)	−0.07 (0.02***	0.01
Δ Month 24	−0.01 (0.03)	−0.05 (0.02)*	0.26
HOMA-IR
Baseline	1.20	1.11	0.15
Δ Month 12	−0.031 (0.050)	−0.347 (0.038)***	<0.0001
Δ Month 24	0.027 (0.046)	−0.364 (0.035)***	<0.0001
HOMA-β (%)
Baseline	107.4	128.1	0.92
Δ Month 12	−17.30 (9.09)	−33.01 (6.78)***	0.16
Δ Month 24	−16.71 (6.25)	−43.92 (4.85)***	0.0004
AUC Insulin (μIU-h/mL)
Baseline	98.8	96.2	0.68
Δ Month 12	7.33 (6.25)	−23.61 (4.86)***	<0.0001
Δ Month 24	6.25 (4.98)	−19.34 (4.08)***	<0.0001
AUC Glucose (mg-h/mL)
Baseline	260.8	260.1	0.97
Δ Month 12	4.52 (4.70)	−4.70 (3.72)	0.23
Δ Month 24	2.89 (4.82)	0.10 (3.88)	1.0
Insulin Response
Baseline	0.93	0.93	0.83
Δ Month 12	0.037 (0.149)	−0.055 (0.111)	1.0
Δ Month 24	0.096 (0.055)	−0.143 (0.045)**	0.0014
Insulin Sensitivity
Baseline	0.22	0.24	0.27
Δ Month 12	0.018 (0.015)	0.078 (0.012)***	0.001
Δ Month 24	−0.013 (0.020)	0.099 (0.015)***	<0.0001
Oral Disposition
Baseline	0.19	0.24	0.44
Δ Month 12	−0.036 (0.052)	0.042 (0.039)	0.46
Δ Month 24	−0.027 (0.030)	0.035 (0.024)	0.18
hsCRP (nmol/L)
Baseline	0.114	0.155	0.91
Δ Month 12	0.030 (0.037)	−0.045 (0.028)	0.105
Δ Month 24	0.002 (0.023)	−0.068 (0.018)**	0.012
Metabolic Syndrome Score
Baseline	−7.9	−8.3	0.35
Δ Month 12	−0.156 (0.212)	−2.646 (0.160)***	<0.0001
Δ Month 24	−0.064 (0.252)	−2.669 (0.193)***	<0.0001

^aBaseline values are the observed mean (SE); change scores are the least-squares adjusted means (SE) from the ITT repeated measures analysis.

^bBetween-group p-value tests for a significant between-group difference in the change score at the time point. All comparisons are controlled for baseline values. All p-values reflect Bonferroni corrections, truncated at 1.0, as appropriate (see text).

*, **, and *** indicated within group differences at the P<0.05, 0.01, and 0.001 values, respectively

The findings of the Marginal Structural Modeling are presented in the Supplementary Material, Figure 2 and Tables 2–4. This analysis allows us to account for biases created by unbalanced drop out and non-adherence to the intervention by inferring a 25% caloric restriction to all CR participants using trial data and weighted by data obtained from individuals that achieved the goal caloric restriction.³² We found that assuming full compliance and effect in the calorie restriction arm had marginal effects on the p-values and essentially no effect on the conclusions of the study on the cardiometabolic factors of interest, with the exception of hsCRP, for which the between group difference was not statistically significant at 2 years in the MSM analysis.

DISCUSSION

Little is known about the effects of prolonged caloric restriction on cardiometabolic health in normal weight, young individuals. Since the studies of caloric restriction on lifespan in animal models starts early in lifespan in normal weight individuals, to understand similar effects in humans, it is critical to conduct caloric restriction experiments in similarly young, healthy individuals. In this two-year multicenter randomized clinical trial, we tested — for the first time — the time-course of cardiometabolic adaptations to two years of moderate CR in healthy non-obese young and middle-aged individuals, having clinically normal risk factor at baseline (percent changes in six parameters of cardiometabolic list are shown in Figure 1). Results of this trial provide evidence that CR with adequate nutrition leads to improvements in multiple cardiometabolic risk factors — even when implemented in healthy young and middle-aged men and women with normal baseline values; further, we found that CR in this population improved already normal risk factors, implying improvement in long term cardiovascular risk. Moreover, these data indicate that sustained CR over two years exerts beneficial effects on cardiometabolic health over and above those conferred by the concordant weight loss.

Figure 1. Percent Changes in Cardiometabolic Parameters.

The within group percent change values (means±SE) by intervention group are shown for cardiometabolic parameters: (A) Body Mass Index (BMI), (B) Metabolic Syndrome Score, (C) Blood Pressure (BP), D) Low Density Lipoprotein-Cholesterol (LDL-C), (E) High Density Lipoprotein-Cholesterol (HDL-C), (F) Area Under the Curve (AUC) Insulin. Significance values are indicated in Table 2–5.

The development of atherosclerotic disease is a multistage process modulated by several cardiometabolic risk factors, of which elevated LDL-cholesterol (LDL-C) and low HDL-C are among the most important triggers and predictors of future CVD events.^12,33–35 According to the guidelines of the National Cholesterol Education Program ATP-III, plasma LDL-C levels should be below 2.56 mmol/L.³⁶ However, data from randomized placebo-controlled trials using statins for primary prevention of CVD events indicate that optimal LDL-C levels should be much lower than 2.56 mmol/L; and starting as low as 1.46 mmol/L, there is a linear relationship between serum LDL-C concentration and the risk of developing CVD.¹² In our study, moderate CR but not AL, reduced serum LDL-C concentrations by 7% from 2.51 to 2.33 mmol/L. Importantly, the reduction in LDL-C was accompanied by a significant increase in serum HDL-cholesterol concentration (from 1.26 to 1.36 mmol/L) and a substantial reduction in serum TG concentration (from 1.15 to 0.90 mmol/L) — two factors modulating cardiovascular risk independently of LDL-C ^8,37,38

Elevated blood pressure is another key risk factor for the development of myocardial infarction, stroke, heart and renal failure. According to the new AHA/ACC blood pressure guidelines, optimal blood pressure is lower than 120/80 mmHg in persons 30 through 59 years of age, and blood pressure values between this level and 130/85 are considered “elevated”. [Whelton, 2018 #2163] In fact, data from observational studies involving more than 1 million men and women clearly indicate that in all age groups, the risk of dying from both coronary heart disease and stroke rises drastically starting from a blood pressure as low as 115/75 mm Hg.¹⁰ The death risk from CHD and stroke doubles for every 20 mm Hg systolic or 10 mm Hg diastolic increase in BP. ¹⁰ In our study, individuals randomized to the CR diet experienced a rapid and significant reduction in systolic, diastolic and mean blood pressure, whereas blood pressure increased in the control group eating an ad-libitum diet. The mechanism mediating this rapid and persistent drop in blood pressure induced by CR are not completely known; it may be due to a CR-mediated reduction in oxidative stress, inflammation and preservation of endothelial nitric oxide bioavailability and function.³⁹

Insulin resistance is a powerful risk factor for the development of type 2 diabetes, coronary heart disease, and some forms of cancer.^9,40 Insulin resistance predicts the risk of developing type 2 diabetes as early as thirteen years before diagnosis. Further and relevant, a rise in plasma glucose concentrations also predict the development of diabetes as early as twelve years before diagnosis; they are normally still within the normal, non-diagnostic range until two to five years before diagnosis, when a rapid deterioration of insulin secretion (β-cell function) and a parallel elevation of glycemia occurs.⁴¹ Moreover, insulin resistance and the compensatory hyperinsulinemia play a role in the pathogenesis of hypertension; in the inhibition of fibrinolysis; and in the stimulation of vascular smooth muscle proliferation and migration, all leading to atherosclerosis.⁴⁰ In our study, CR did not significantly change glucose tolerance at a point in the development of diabetes where it may still be in the normal range (see above); however, it profoundly improved insulin sensitivity; and reduced plasma fasting glucose and glucose-stimulated insulin concentration. This reduction in circulating levels of insulin and the improvement of the metabolic syndrome score may exert beneficial effects not only in lowering the risk of atherosclerosis, but also in the primary and secondary prevention of certain common cancers, such breast cancer.^42,43

Low-grade chronic inflammation is also deeply implicated in the pathogenesis of coronary heart disease, but also of cancer, cognitive impairment and in the biology of aging itself.^44,45 According to the guidelines of the American Heart Association, individuals with circulating high sensitivity C-reactive protein (hs-CRP) concentrations less than 9.5 mmol/L have a 50% less risk of CHD than those with hs-CRP between 9.5 and 28.6 mmol/L.⁴⁶ In our study, individuals randomized to CR, but not the control group, experienced 14.1 to 8.38 mmol/L over the two-year study. Interestingly, changes in hs-CRP were only significant after two years of CR even though there was some weight regain between one and two years of CR; this emphasizes the benefits of sustained CR over short-term weight loss for CVD risk reduction. Of further interest, in a pharmacological interventional clinical trial with atorvastatin of 502 patients with angiographically-documented coronary disease, a similar reduction of hs-CRP from 27.6 to 21.9 mmol/L resulted in a significant regression of atherosclerosis as determined by intravascular ultrasonography.¹¹

We believe these results have a profound public health relevance for the lifetime reduction of atherosclerotic risk for at least two reasons. First, in the Framingham Heart Study, individuals with optimal cardiometabolic risk factors at age 50 years have a 13-times lesser risk of developing CVD during their remaining lifetime than do men and women with two or more abnormal risk factors, despite a much longer lifespan.¹⁷ Second, early reduction in cardiometabolic risk is much more effective than late reductions in preventing coronary events. A 30% reduction of serum LDL-C with statins results in a 30% lesser risk of coronary events; indeed, a similar 30% reduction of LDL-C in individuals born with the rare PCSK9 genetic variant have lesser coronary artery disease risk by approximately 90%.¹¹

To our knowledge, this is the first and adequately powered two-year dietary randomized clinical trial to demonstrate such a profound effect on lowering all cardiometabolic risk factors beyond normal levels, even in rather young lean individuals. There are no pharmacologic agents with such profound effect on such a broad range of cardiometabolic risk factors. These should provide a new tool for clinicians in fighting the ravages of the twenty-first century American lifestyle.

Major strengths of this study include the relatively large sample size for an intensive physiologic intervention study of this type; the intention-to-treat randomized controlled trial design minimizing the potential for selection bias; and the long duration of the study. Moreover, in all the participants we carefully measured their energy intake and energy expenditure by using both nutrition assessment software and direct analysis through double-labeled water. Our study had a high retention rate of enrolled participants, and good adherence to the study interventions as shown by the successful weight reduction over two years. A major limitation of this study is the lack of a clinical measurement of atherosclerotic plaque modifications.

In conclusion, the results of this large randomized controlled trial provide evidence that a moderate CR-induced negative energy balance improves multiple cardiometabolic risk factors — waist circumference, blood pressure, HDL-cholesterol, LDL-cholesterol, triglycerides, insulin resistance and glucose control, metabolic syndrome, and chronic inflammatory tone — well below the conventional risk thresholds used in the clinical practice. These findings are of significant public health import even when started in healthy young and middle-aged non-obese men and women. These data combined with previously published safety data for CR,⁴⁷ indicate that inexpensive and safe dietary interventions, such as moderate CR, can be implemented early in life to optimize cardiometabolic health and reduce the lifetime risk of developing some of the most common, disabling and expensive chronic diseases, such as hypertensive and atherosclerotic cardiovascular disease. It will be important to conduct additional research in this area to understand the biological factors — physiologic and molecular — whereby the described adaptations are achieved. Such understanding may lead to pharmacologic therapies of use for improving human health.

RESEARCH IN CONTEXT.

Evidence before this study

In Rhesus monkeys, young-onset calorie restriction (CR) reduces the risk of developing and dying of cardiovascular disease by at least 50%. In observational cross-sectional human studies, severe CR provides a powerful protective effect against multiple atherosclerotic risk factors: older adults practicing CR have healthier physiologic cardiovascular parameters. However, little is known about the extent of cardiometabolic adaptations in prospective human studies in response to CR of greater than one year in healthy non-elderly normal weight or slightly overweight human individuals.

Added value of this study

This is the first study of medium term (two-years) of true calorie restriction in humans—and its effect on cardiometabolic intermediary outcomes.

Implications of all the available evidence

The effects of two years of 13% calorie restriction on a myriad of cardiometabolic risk factors — in particular those composing the five components of metabolic syndrome, as well as glucose tolerance using an oral glucose tolerance test and derivatives thereof — raise the possibility of being able to modify the ravages of cardiovascular disease, diabetes and obesity in Western countries with relatively modest lifestyle interventions.