Evidence-based formulas relating long-term diet, exercise, tobacco use, body mass index, and gender to total cholesterol / high density lipoprotein cholesterol ratio

 

David K. Cundiff, MD

 

Los Angeles County + USC Medical Center (Retired)

 

Correspondence to dkcundiff@whistleblowerdoctor.org

 

 

Abstract

Background The precise quantification of how diet, exercise, tobacco use, body mass index (BMI) and gender interact with each other and interactively influence the total cholesterol/high-density lipoprotein cholesterol ratio (TC/HDL-C) could facilitate cardiovascular disease risk analysis and inform disease prevention strategies.

Objective Develop an evidence-based methodology to predict TC/HDL-C

Design Retrospective univariate and multivariate statistical analysis of TC/HDL-C related to cardiovascular disease risk factors (macronutrient profiles, exercise, tobacco use, BMI, and gender) from the Diabetes Control and Complications Trial (DCCT) and Food and Agriculture Organization/World Health Organization (FAO/WHO)

Setting DCCT: 27 diabetes clinics in North America, FAO/WHO: female and male cohorts from 167 countries worldwide

Participants DCCT database: 681 female and 760 male participants ≥ 12 years old from North America with type 1 diabetes; and FAO/WHO databases: population-based samples of female and male cohorts (age ≥ 15 years) from 167 countries

Main Outcomes and Measures Formulas will be derived from multiple regression analysis relating TC/HDL-C (dependent variable) to macronutrient profiles, exercise, tobacco use, BMI, and gender (independent variables).

 Results DCCT database: formula relating TC/HDL-C to macronutrient profiles, exercise, tobacco use, BMI, and gender =((23.44 protein (g) +25.856 carbohydrates–44.72 fiber+56.61 total fats–23.535 polyunsaturated fatty acids–45.248 alcohol) / kilocalaries–0.11820 exercise (DCCT 1-4 scale)+0.36101 tobacco use (yes=1, no=0) * 1.00444+0.04782 BMI)*3.292444–0.45111 gender (female=1, male=0)) * 0.967857–17.8187; (n=1441, R²=0.12, P<0.0001).

FAO/WHO database: Simulated TC/HDL-C (countries worldwide)=((51.744 protein (g)+75.36 carbohydrates –85.30 fiber+250.794 total fat–341.406 polyunsaturated fatty acids)/kilocalories+0.01519 alcohol–0.09941 exercise (1-4 DCCT scale)+1.43433 tobacco use (1=yes, 0= no)) * 1.19127–16.99615) * 0.80795+0.02070 BMI) * 1.1896–0.48011) * 0.82219–0.17610 gender (female=1, male=0) * 1.18239+ 0.20453; (n=334, R²=0.72)

 Conclusions Multiple regression derived formulas from two very different databases similarly predicted TC/HDL-C. This methodology should be explored in prospective studies and used to help guide diet and lifestyle.

 

 

 

Introduction

High serum cholesterol/high density lipoprotein cholesterol ratio (TC/HDL-C) is associated with higher risks of cardiovascular disease,[1] at least in developed countries. Diet, physical activity, tobacco use, BMI, and gender interact with each other and with environmental and genetic predispositions in complex and poorly understood ways to influence the serum lipid profile.

According to an investigation by The BMJ, the US Dietary Guidelines for Americans recommendation for low intakes of saturated fatty acids to avoid cardiovascular disease (CVD) events relied on unscientific evidence.[2] Likewise, The BMJ analysis faulted the Dietary Guidelines for Americans for failure to comprehensively study the potential for low carbohydrate diets in improving risk factors for CVD. The BMJ report indicated that new strategies of assessing the relationship between diet and CVD risk should be considered.

Dietary saturated fat, trans fat, and cholesterol have been associated with dyslipidemias,[3, 4] high serum cholesterol,[3, 5, 6] high serum cholesterol/high density lipoprotein cholesterol ratio (TC/HDL-C),[7, 8] and tobacco use.[3, 5, 6, 9] Inverse correlations have been shown with HDL-C and CVD risk.[5, 6] Physical inactivity and tobacco use have been associated with low levels of HDL-C.[10] Total cholesterol, LDL cholesterol (LDL-C), and TC/HDL-C have been reported to be reduced by exercise.[10, 11] The positive associations between body mass index (BMI) and dyslipidemias are strongest in people 20 – 39 years old but persist at older ages.[12]

Diet may interact with other CVD risk factors. For example, using the Diabetes Control and Complications Trial (DCCT) database, I previously reported that tobacco users consumed significantly more total fat, saturated fat, cholesterol, and alcohol than never smokers. HDL-Cs were significantly lower in both female and male smokers than non smokers. Macronutrient intakes of former tobacco users resembled that of never users rather than current users.[13]

The current study introduces a new evidence-based methodology to study the relationship of diet and other risk factors with TC-HDL-C. It utilizes DCCT and Food and Agriculture Organization and World Health Organization (FAO/WHO) data to relate TC/HDL-C (dependent variable) to diet, tobacco use, body mass index (BMI), physical activity, and gender (independent variables). Multiple regression derived formulas from the DCCT and the FAO/WHO databases are derived and compared.

The hypotheses being tested were (1) clinically plausible multiple regression formulas can be derived from risk factor data from each of two databases and (2) the formulas will correlate with each other. For both databases, I predicted that TC/HDL-C would be decreased with exercise and that TC/HDL-C would be increased by a high saturated fat, low fiber diet, tobacco use, high BMI, and male gender.

Utilising the methodology here presented, formula predictions of TC/HDL-C based on diet, exercise, tobacco use, BMI, and gender are offered in an online computer application to for physicians and individuals.

Methods

DCCT analysis

The DCCT was conducted by the DCCT Research Group and supported by National Institute of Health grants and contracts and by the General Clinical Research Center Program from the National Center for Research Resources. The data from the DCCT was supplied by the National Institute of Diabetes and Digestive and Kidney Diseases Central Repositories. This manuscript was not prepared under the auspices of the DCCT and does not represent analyses or conclusions of the DCCT study group, the National Institute of Diabetes and Digestive and Kidney Diseases Central Repositories, or the National Institutes of Health.

The DCCT eligibility criteria and screening methods and the baseline characteristics of the study subjects have been reported in detail.[14] At the start of the study, 1441 participants (761 males and 680 females; 96.5% white, 2.0% black, and 1.5% other races) from 27 diabetes specialty clinics in the U.S. and Canada were between 13 and 39 years of age, C peptide-deficient, and in good general health. Institutional review boards in all the diabetes specialty centers approved the trial, so further human subjects’ research committee approval was not required. At the end of the DCCT trial in 1993, the mean time on-study was 6.5 years (range, 3.8-9.7 years).[15]

Trained dietitians interviewed participants and conducted a modified Burke type diet history, collecting quantitative and qualitative information representing a usual week of dietary intake over the previous year.[16] Subsequently, the dietitians obtained follow-up diet histories at years two and five and at the end of the study.[17] By analyzing these diet histories with nutritional component software, the dietitians and statisticians generated a dataset consisting of each subject’s consumption of macro and micronutrients. This diet analysis instrument had a high reproducibility on repeated administrations of the diet history.[18]

Macronutrients included for univariate and multivariate analysis with mean TC/HDL-C were the following:

• protein (g and % of kilocalories (kcals)),
• carbohydrates (g and % of kcals),
• dietary fiber (g and g/1000 kcals),
• polyunsaturated fatty acids (PUFA: g and % of kcals),
• monounsaturated fatty acids (MUFA: g and % of kcals),
• saturated fatty acids (SFA: g and % of kcals),
• total fats (g and % of kcals), and alcohol (g and % of kcals),
• daily average kilocalories (kcals) consumed.

Compared with females, males averaged 51% more calories consumed. Average caloric intake decreased with increasing age (females: – 7.2 kcals/day per year; males: – 29.6 kcals/day per year). To eliminate confounding by the major influences of age and gender on food intake in the multivariate analyses, macronutrient consumption was measured in percent of kcals rather than grams (g) except dietary fiber where it was g/1000 kcals instead of g.

The patterns of tobacco use and exercise among the 1441 DCCT participants were assessed as follows:

• tobacco use during any part of the trial (yes = 1; no = 0, mean [90% CI] = 0.209 [0.0 -1.0])
• mean exercise level while on study (on a 4 point scale: 1 = sedentary, 2 = mild activity, 3 = moderate activity, and 4 = strenuous activity: 27.7% of subjects averaging ≤ 1.5, 65.0% of subjects averaging > 1.5 – 2.5, 7.3% of subjects averaging > 2.5 – 3.5 and 1.0% averaging > 3.5 – 4: mean [90% CI] = 1.82 [1.00-2.67]).

Data on exercise, tobacco use, serum cholesterol, HDL-C, and BMI were collected yearly.

 

FAO/WHO analysis

Of 200 countries in the Global Health Observatory Data Repository of the WHO and the FAO databases, complete data on food commodity availability in kcals per capita per day by the FAO[19] and mean serum cholesterol data for female and male cohorts by the WHO[20] were available for 167 countries. The food commodity availability data were broken down to the following:

• cereals (e.g., rice, maize, and corn),
• vegetable oils (e.g., soy, rapeseed, mustard seed, and palm),
• sugar and sweeteners (e.g., sucrose and fructose from sugar cane, corn, beets, and honey),
• meat (e.g., cow, pig, sheep, goat, offals),
• animal fats,
• roots and tubers (e.g., potatoes and cassavas),
• fruit (including juices),
• pulses (e.g., beans and lentils), and
• milk, cheese, and eggs.

The percent of total available kcals percapita per day for each food group in a given country comprised that country’s food group profile. Data for percapita alcohol “consumption”, in contrast with “availability,” came as the variable “g/day consumed” from the WHO[21] rather than as percent of total available kcals.

The WHO evaluated physical activity of female and male cohorts in countries worldwide with the variable “insufficient physical activity” (0%-100% scale). According to the WHO, “adults aged 18–64 should do at least 150 minutes of moderate-intensity aerobic physical activity throughout the week or do at least 75 minutes of vigorous-intensity aerobic physical activity throughout the week or an equivalent combination of moderate and vigorous-intensity activity.”[22] The WHO defined “insufficient physical activity” as less than this recommended level of physical activity.

While 112 out of the 167 countries with food and serum cholesterol data had complete data on physical activity,[22] 55 countries lacked physical activity data. Imputed estimates of the WHO variable “insufficient physical activity” were obtained by using a multiple regression analysis generated formula. As the dependent variable for this formula, I used insufficient physical activity of the 224 female and male cohorts from the 112 countries with data. The independent variables consisted of food groups (% kcals), gender, BMI in 2008, and country percapita gross domestic product (GDP).

The WHO provided tobacco use data (percent of female and male cohorts that used tobacco) for 307 cohorts from the 167 countries.[23]  As with the missing physical activity data, imputed tobacco use estimates from a multiple regression formula were used for the 27 cohorts (14 female and 13 male cohorts) for which WHO provided no data.

To facilitate comparisons, the same macronutrients were included in the FAO/WHO analysis as in the examination of DCCT data. To generate the FAO/WHO macronutrient profile for each country from the food group profile, the FAO food commodity availability data were broken down to macronutrient availability by analyzing samples of each commodity using the United States Department of Agriculture (USDA) Nutrient Database for Standard Reference, Release 24,[24] as described previously.[25]

The FAO/WHO database had a serum cholesterol variable but no HDL-C. An evidence-based simulation of TC/HDL-C would greatly aid in the comparisons of the risk factors between the DCCT and FAO/WHO databases. This simulation assumes that increases in serum cholesterol correspond to nonlinear increases in TC/HDL-C in a logical way modeled from the means and standard deviations (SDs) of the TC/HDL-C in the DCCT database. To simulate a TC/HDL-C variable for this WHO database to compare with TC/HDL-C from the DCCT for univariate and multivariate correlations, the following steps were undertaken:

1. The FAO/WHO database reported mean serum cholesterol in mmol/l. Cholesterol (mmol/l) was converted to cholesterol (mg/dl) in the FAO/WHO database (i.e., multiplying cholesterol mmol/l by 38.61).
2. The DCCT’s cholesterols (means and standard deviations (SDs)), HDL-C levels (means, SDs), and TC/HDL-Cs (means, SDs) were determined by gender. These means and SDs from the DCCT database were used as the reference from which to match the means and SDs of the same variables from the FAO/WHO database. This allowed for the simulation of TC/HDL-C from the serum cholesterol in the FAO/WHO database.
3. The SDs of the distributions of serum cholesterols of the FAO/WHO cohorts by gender were adjusted to match the SDs of the DCCT subjects by gender by multiplying by the appropriate factors (e.g., FAO/WHO simulated serum cholesterol of country A’s female cohort = mean cholesterol of country A’s female cohort + ((mean cholesterol of country A’s female cohort – 80234 (mean cholesterol of all female cohorts)) * SD adjustment multiplying factor (i.e., the SD adjustment factor that will harmonize the FAO/WHO female cohort cholesterol SD with the DCCT female cholesterol SD)).
4. A simulated FAO/WHO HDL-C variable was created that inversely correlated with FAO/WHO serum cholesterol and had the same SD by gender as in the DCCT database. For example, simulated HDL-C for country A’s female cohort = mean cholesterol of country A’s female cohort * 3039 (0.3039=mean HDL-C /cholesterol for females in DCCT database) – (mean cholesterol of country A’s female cohort – 183.80 (183.80 = mean FAO/WHO female cholesterol) * 0.517 (0.517 = multiplying factor calculated to make the SD of the FAO/WHO female cohorts’ simulated HDL-Cs match the actual SD of the HDL-Cs of the DCCT females.
5. Finally, the SD by gender of the FAO/WHO variable “simulated TC/HDL-C” was adjusted by a multiplying factor calculated to match the SD of the FAO/WHO database TC/HDL-C to the SD of the DCCT TC/HDL-C variable.

 

Statistical analysis

Pearson correlation coefficient analysis determined the positive or negative associations of TC/HDL-C (DCCT) and simulated TC/HDL-C (FAO/WHO) with the diet (macronutrient profiles), exercise, tobacco use, BMI, and gender in the respective databases.

To assess the interaction of macronutrients, exercise, tobacco use, BMI, and gender with TC/HDL-C (dependent variable), multiple regression analysis was utilized with the non-experimental regression method. To assure the inclusion of the highest proportion of the components of the macronutrient profile in the formulas, the multiple regression analysis was done empirically in stages to maximize the overall percent of kcals represented by the macronutrients included in the formula. For example, if including alcohol (≈ 1% of macronutrients) excluded carbohydrates (≈ 46% of macronutrients) from the DCCT multiple regression formula, alcohol was moved from stage 1 to stage 2 of the multiple regression analysis. And if the DCCT MUFA variable (≈13% of macronutrients) excluded total fat (≈ 35% of macronutrients), the variables MUFA and SFA would be eliminated leaving total fat and PUFA to better represent the overall influence of subcategories of fat in the multiple regression formula.

By multiplying each significant variable in the multiple regression by its parameter estimate, these multiple regression generated formulas quantified the relationship of TC/HDL-C with macronutrient intake (percentages of macronutrients), exercise, tobacco use, BMI, and gender. The macronutrient variables in the formulas were then transformed into grams, making identical formulas that could be used for prospective predictions of TC/HDL-C. To transform the DCCT and FAO/WHO TC/HDL-C formula outputs into percentiles as a prediction tool for individuals, the formulas were standardized (adjust the standard deviation (SD) to 1.00 and then the mean to 0.00) and then set equal to z.

Then the formula for percentiles was employed:

Estimated percentile of cardiovascular risk=50 + (z * (43.73 + 17.68 * z²) / (1 + Abs(z) * (0.4331 + 0.3666 * z² ) ));

I performed the data analysis with SAS statistical software 9.1 (SAS Institute, Cary, NC).

 

Results

DCCT analysis

With DCCT participants, Table 1 shows the correlations of macronutrients consumed (g and percent of kcals), exercise, tobacco use, BMI, gender, and age with the TC/HDL-C. Dietary fiber (g/1000 kcals) was most significantly negatively correlated with TC/HDL-C (r= –0.10, P<0.0001), while energy in kcals and MUFA were most significantly positively correlated (r= 0.13, P<0.0001). TC/HDL-C positively correlated with smoking (r= 0.16, P<0.0001) and BMI (r= 0.19, P<0.0001) and negatively with exercise (r= –0.07, P=0.0069) and female gender (r= –0.25, P<0.0001).

Table 1. Correlations of macronutrient consumption and other variables of DCCT participants with simulated total cholesterol / HDL ratio (n=1441)

Macronutrients and Other Variables	On Trial  Results Mean	90% CI	Correlation with Total Cholesterol/ HDL r (P)
Energy (kcals)	2310	1362-3612	0.13 (<0.0001)
Protein g	102	62.5-159	0.12 (<0.0001)
Protein % of kcals	17.9	14.5-21.6	–0.03 (0.30)
Total carbohydrate g	264	161-410	0.12 (<0.0001)
Total carbohydrate % of kcals	46.3	36.9-55.6	–0.04 (0.11)
Dietary fiber g	24.2	13.7-39.1	0.04 (0.13)
Dietary fiber g /1000 kcals	10.8	6.72-15.9	–0.10 (<0.0001)
PUFA g	19.3	9.33-33.9	0.08 (0.0025)
PUFA % of kcals	7.5	4.97-10.4	–0.02 (0.53)
MUFA g	35.9	16.6-63.7	0.13 (<0.0001)
MUFA % of kcals	13.7	9.65-17.6	0.10 (0.0002)
SFA g	32.8	15.1-59.7	0.11 (<0.0001)
SFA % of kcals	12.6	8.52-16.5	0.06 (0.0178)
Total fat g	94.9	46.5-165	0.12 (<0.0001)
Total fat % of kcals	36.4	26.9-45.3	0.07 (0.0110)
Total fat – PUFA g	75.6	36.3-133	0.13 (<0.0001)
Total fat – PUFA %	29.0	21.0-36.4	0.09 (0.0012)
Alcohol g	3.31	0.00-14.9	–0.06 (0.0376)
Alcohol % of kcals	0.93	0.00-3.98	–0.09 (0.0010)
Age (mean on study years)	30.1	17.7-40.8	0.06 (0.0255)
Female (yes=1, no=0)	0.47	–	–0.25 (<0.0001)
Exercise (1-4  scale)	1.82	1.00-2.67	–0.07 (0.0069)
Tobacco use (yes=1, no=0)	0.29	1.00-2.00	0.16 (<0.0001)
BMI at close of the trial	25.8	20.8-32.6	0.19 (<0.0001)

The DCCT multiple regression formula for predicting TC/HDL-C (dependent variable) based on macronutrient profiles, exercise, tobacco use, BMI, and gender (independent variables) is as follows:

TC/HDL-C = ((23.44 protein (g) + 25.856 carbohydrates (g) – 44.72 fiber (g) + 56.61 total fats (g) – 23.535 PUFA (g) – 45.248 alcohol (g)) / kcals – 0.11820 exercise (DCCT 1-4 scale) + 0.36101 tobacco use (yes = 1, no = 0) * 1.00444 +0.04782 BMI) *  3.292444 – 0.45111 gender (female = 1, male = 0)) * 0.967857 – 17.8187; (n= 1441, R²=0.12, P < 0.0001)

 

FAO/WHO database analysis

For the 167 countries with complete data on dietary intake[19] and cholesterol levels,[20] Appendix 1 (online) shows the mean levels of physical inactivity (0% – 100%),[26] tobacco use (% of cohort),[23] and alcohol consumption (g/day per capita).[21] Imputed estimates of “insufficient physical activity” and tobacco use (%) are indicated by “*”.

Appendix 2 (online) shows FAO data on availability of selected food groups (% of total available kcals) from worldwide countries.[19] Caloric intake of males was considered to be equated to 1.2 times the mean population caloric intake and for females 0.8 times the mean, in accordance with USA data from the National Health and Nutrition Examination Survey.[27]

From the FAO/WHO databases, Table 2 shows the univariate correlations between simulated TC/HDL-C and the major food groups. Except for sugar, vegetable oils, and fruit; plant-based foods negatively correlated with TC/HDL-C while animal-based foods directly correlated.

 

Table 2. Availability of foods correlated with simulated cholesterol/HDL-C (FAO data from 167 countries[19])

Food Groups and Kcals	Mean % of Kcals	% of kcals 90% CI	Simulated Cholesterol/ HDL-C mg/dl r (P)
Cereals	41.9	21.5-67.4	–0.51 (<0.0001)
Wheat	18.3	1.71-43.5	0.34 (<0.0001)
Rice	11.4	0.59-42.8	–0.29 (<0.0001)
Maize	7.58	0.00-31.2	–0.39 (<0.0001)
Vegetable oils	9.33	2.65-16.6	0.21 (0.0002)
Soy	2.48	0.03-8.69	0.22 (0.0007)
Rape and mustard seed	0.63	0.00-4.31	0.42 (<0.0001)
Palm	1.61	0.00-6.23	–0.30 (<0.0001)
Sugar and sweets	9.87	1.99-17.2	0.50 (<0.0001)
Raw sugar	9.02	1.90-16.0	0.44 (<0.0001)
Meat and offals	7.37	1.43-16.3	0.68 (<0.0001)
Poultry	1.94	0.11-5.43	0.43 (<0.0001)
Cow	1.77	0.31-4.16	0.29 (<0.0001)
Pig	2.45	0.00-8.21	0.60 (<0.0001)
Sheep/Goat	0.65	0.00-2.50	0.13 (<0.0127)
Offals	0.3	0.01-0.64	0.24 (<0.0001)
Roots and tubers	7.43	1.08-32.3	–0.41 (<0.0001)
Potatoes	2.23	0.02-7.03	0.41 (<0.0001)
Cassava	2.91	0.00-15.3	–0.42 (<0.0001)
Milk, eggs, and fish	7.48	1.47-14.5	0.66 (<0.0001)
Milk	3.91	0.32-9.64	0.25 (<0.0001)
Cheese	1.18	0.00-4.80	0.71 (<0.0161)
Eggs	0.76	0.08-1.72	0.69 (<0.0001)
Fruit	5.7	1.25-10.8	0.11 (0.0383)
Animal fats	2.34	0.17-7.35	0.66 (<0.0001)
Pulses	2.33	0.11-6.70	–0.46 (<0.0001)
Other food groups	6.26	1.75-11.4	0.25 (<0.0001)

My template for using data from the USDA Nutrient Database for Standard Reference, Release 24[24] to convert FAO food group data to macronutrient profiles for each country has been previously reported.[25] As an example of converting the FAO food group data to the macronutrient profile of a country, Table 3 gives the breakdown of the macronutrient profiles of the 10 major food group components of the USA diet.

Table 3 Converting FAO Food Group Availability Data in % Kcals to Macronutrients (g) for the USA

FAO Food Groups	Kcals	Protein	Carbs	Dietary Fiber	PUFA	MUFA	SFA	Total Fat	Alcohol
Cereals	821	22.0	164	13.2	5.5	4.6	2.4	12.2	0.0
Vegetable oil	667	0.0	0.0	0.0	19.5	25.9	20.0	74.3	0.0
Sugar and sweeteners	638	0.0	161	0.0	0.0	0.0	0.0	0.0	0.0
Meats	455	28.5	1.5	0.0	3.5	19.8	13.0	36.2	0.0
Roots and tubers	96	1.5	22.8	2.8	0.1	0.0	0.0	0.1	0.0
Milk, eggs, fish	466	45.4	15.0	0.0	4.6	10.4	8.7	23.6	0.0
Fruits	193	2.4	46.2	5.1	0.4	1.8	0.4	2.5	0.0
Animal fat	109	0.0	0.0	0.0	1.5	6.7	4.6	11.9	0.0
Pulses	41	2.4	5.6	1.8	0.5	0.7	0.2	1.2	0.0
Alcohol	124	0.0	5.6	0.0	0.00	0.0	0.0	0.0	14.2
Totals	3659	102	422	22.8	35.6	69.9	49.3	162	14.2

Appendix 3 shows macronutrient profiles of female and male adult cohorts from 167 worldwide countries derived from FAO food availability data using the USDA database as previously described.[25] Using data from Appendix 3, Table 4 presents the correlations of simulated TC/HDL-C with energy (kcals), macronutrients, exercise (converted from the WHO “insufficient physical activity” variable), tobacco use, BMI, and gender.

Table 4. Selected correlations of macronutrients and other variables with cholesterol/HDL-C from 167 countries (FAO/WHO data[19], n=334 cohorts)

 

Macronutrients and other variables	Mean	90% CI	Simulated TC/HDL-C mg/dl r (P)
Energy (kcals)	2613	1855-3353	0.68 (<0.0001)
Protein g	70.2	38.5-111	0.67 (<0.0001)
Protein % of kcals	10.7	8.60-13.3	0.19 (0.0004)
Carbohydrate g	382	248-536	0.36 (<0.0001)
Carbohydrate % of kcals	59.3	45.8-72.0	–0.69 (<0.0001)
Dietary fiber g	30.4	18.2-46.2	–0.02 (<0.72)
Dietary fiber g per 1000 kcals	12.0	7.60-18.1	–0.69 (<0.0001)
PUFA g	20.4	10.5-33.9	0.65 (0.0184)
PUFA % of kcals	7.0	5.39-8.84	0.26 (<0.0001)
MUFA g	36.8	14.7-72.9	0.83 (<0.0001)
MUFA % of kcals	12.3	7.28-18.0	0.73 (<0.0001)
SFA g	24.2	8.62-49.4	0.82 (<0.0001)
SFA % of kcals	8.1	4.42-12.3	0.71 (<0.0001)
Total fat g	84.5	34.0-164	0.81 (<0.0001)
Total fat % of kcals	28.3	17.3-40.4	0.69 (<0.0001)
Total fat g – PUFA g	64.0	44.5-130	0.83 (<0.0001)
Total fat % of kcals – PUFA % of kcals	21.3	10.3-31.6	0.73 (<0.0001)
Alcohol g	6.9	0.27-16.7	0.50 (<0.0001)
Exercise (1-4  scale)	1.97	1.14-2.71	–0.28 (<0.0001)
Tobacco use (portion of population: 0.00-1.00)	0.22	0.14-0.51	0.42 (<0.0001)
BMI (kg/m2)	25.2	21.2-28.9	0.52 (<0.0001)
Female (yes=1, no=0)	0.50	–	–0.25 (<0.0001)

Simulated TC/HDL-C ratio (FAO/WHO data set) = ((51.744 protein (g) + 75.36 carbohydrates (g) – 85.30 fiber (g) + 250.794 total fat (g) – 341.406 PUFA (g)) / kcals + 0.01519 alcohol (g) – 0.09941 exercise (1-4 DCCT scale) + 1.43433 tobacco use (1=yes, 0= no)) * 1.19127 – 16.99615) * 0.80795 + 0.02070 BMI) * 1.1896 – 0.48011) * 0.82219 – 0.17610 gender (female = 1, male = 0) * 1.18239 + 0.20453; (n=334, R²=0.72)

The univariate correlations of macronutrients with TC/HDL-C from the FAO/WHO databases were similar to those from the DCCT (Table 1) but much stronger. As the exception, alcohol was associated with lower TC/HDL-C in the DCCT database but higher in the FAO/WHO database. The correlations of TC/HDL-C with gender, exercise, tobacco use, BMI, and gender were in the same directions as those from the DCCT.

The FAO/WHO multiple regression derived formula relating macronutrient availability, exercise, tobacco use, BMI, and gender to simulated TC/HDL-C has a striking similarity to the corresponding DCCT formula (above):

Table 5 shows that the formula outputs for TC/HDL-C. The correlations are strong in both the DCCT database and FAO/WHO database.

 

Table 5. DCCT and FAO/WHO simulated TC/HDL-C formulas compared using the DCCT database (n=1441) and FAO/WHO database (n=334)

Database/ formula	DCCT database: DCCT TC/ HDL-C formula r (P)	FAO/WHO database: DCCT TC/HDL-C  formula r (P)	DCCT database: FAO/WHO TC/ HDL-C formula r (P)	FAO/WHO database: FAO/WHO TC/HDL-C  formula r (P)
DCCT TC/HDL-C	0.35 (<0.0001)	–	0.24 (<0.0001)	–
FAO/WHO TC/HDL-C	–	0.76 (<0.0001)	–	0.85 (<0.0001)
DCCT TC/HDL-C formula	1.0	–	0.73 (<0.0001)	0.88 (<0.0001)

 

The online interactive health enhancement app based on these formulas (http://grandbargainsbook.com/healthTool/public/) allows individuals, healthcare providers, and cardiovascular risk researchers to enter data on foods and beverages consumed, exercise, tobacco use (yes or no), BMI, and gender into a form on the website in order to predict their TC/HDL-C based on averages of multiple data inputs over the long term (i.e., years). It also predicts their ranking in percentiles of TC/HDL-C levels from the FAO/WHO worldwide databases. For these predictions, the outputs of the FAO/WHO and DCCT formulas are averaged and the predictions reported.

 

Discussion

In the DCCT participants, the potential beneficial effects of exercise on TC/HDL-C may be underestimated because of the low physical activity levels (mean exercise level = 1.82 on the DCCT 1-4 scale, i.e., < mild activity).  The mean of the activity levels in the 334 cohorts in the FAO/WHO database was only slightly higher (mean exercise level = 1.97 on the DCCT 1-4 exercise scale).

Alcohol consumption (% of kcals) inversely correlated with TC/HDL-C in the DCCT database (r= –0.09, P=0.0010, Table 1), in accordance with previous studies,²⁴ and decreased TC/HDL-C in the DCCT multiple regression formula. However, with FAO/WHO data, alcohol directly correlated with TC/HDL-C and increased outputs of the multiple regression formula. The moderately strong correlation of TC/HDL-C and alcohol in the FAO/WHO database (r=0.50, P < 0.0001, Table 4) may have been due, in part, to the inverse relationship of alcohol (g) with fiber (g / 1000 kcals, r= –0.35, P < 0.0001) and direct correlations with saturated fat (% kcals, r= 0.40, P < 0.0001) and tobacco use (yes=1, no=0, r= 0.36, P < 0.0001).

 

Strengths of the study

• Two large databases were analysed using the same dependent and independent variables.
• Multiple regression derived formulas from the 2 very different databases could be compared.
• The FAO/WHO database of cohorts around the world provided a very wide range of values for both the dependent and independent variables
• The multiple regression formulas resulting from the analysis can be utilised by physicians and lay people with an interactive online app to predict TC/HDL-C based on diet, exercise, tobacco use, and gender.

Limitations of the DCCT analysis

• retrospective analyses,
• subjects with a single disease,
• infrequent assessments of dietary input, exercise levels, and tobacco use, and
• subjects living only in the USA and Canada.

Limitations of the FAO/WHO analysis

• tracking percapita food availability rather than food consumption,
• TC/HDL-C was simulated from FAO/WHO serum cholesterol, harmonizing with means and SDs from the DCCT,
• changes in diet, exercise, and tobacco use over generations were not assessed,
• there were no data on food commodities consumed (% of kcals available), so food commodities available was used,
• kcals available by gender and consequently food group availability by gender were estimated based on a previous study,
• exercise data for 110 cohorts were absent and imputed from multiple regression analysis,
• tobacco use data for 27 cohorts were imputed, and
• data on vegetables (other than vegetable oils), nuts, and fish were not available.

These data confirm previous findings that TC/HDL-C is influenced by diet, physical inactivity, tobacco use, BMI, and gender. However, these formulas better quantify the interrelationship of these variables. Animal products tend to increase TC/HDL-C and plant products (other than sugar, vegetable oils, and fruit juices) are apt to decrease or maintain low TC/HDL-Cs. While the strength of the correlations between these two formulas from disparate databases serves to validate them both to a great degree, these multiple-regression equations relating TC/HDL-Cs to diet and lifestyle should be further validated with similar analyses from other databases and in prospective studies. The web based interactive TC/HDL prediction tool (http://grandbargainsbook.com/healthTool/public/) may be useful for patients for monitoring risk reduction strategies. Researchers may use this interactive tool for conducting observational and interventional studies.

 

 

 

Acknowledgements

This analysis was not supported by public or private funding. The creation of the interactive website containing the formulas predicting TC/HDL-C was supported by a grant from the Pollination Project (http://thepollinationproject.org/).  Pollination Project personnel had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.

John Pezzullo, PhD, Nikunj Raghuvanshi, PhD, Brent Hill, and Nikifor Porlanco used Java Script coding to create the online interactive TC/HDL-C predictor based on formula estimates of TC/HDL-C. They were not involved with the data analysis of the study.

 

 Transparency declaration

The author affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

 

Ethics approval

NA

 

Data sharing statement

The author will make the data used to generate this study available to other researchers on request.

 

References

1. Estimate of 10-year risk for coronary heart disease: Framingham point scores. National Heart, Lung, and Blood Institute of the US Department of Health and Human Services http://www.nhlbi.nih.gov/health-pro/guidelines/current/cholesterol-guidelines/quick-desk-reference-html/10-year-risk-framingham-table
2. Teicholz N. The scientific report guiding the US dietary guidelines: is it scientific? BMJ 2015;351 doi: 10.1136/bmj.h4962published. http://www.bmj.com/content/351/bmj.h4962
3. Ecological analysis of the association between mortality and major risk factors of cardiovascular disease. The World Health Organization MONICA Project. Int J Epidemiol 1994;23(3):505-16. http://www.ncbi.nlm.nih.gov/pubmed/7960374
4. Dietary Reference Intakes for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol, Protein, and Amino Acids (Macronutrients): The National Academies Press, 2005. http://www.nap.edu/openbook.php?record_id=10490
5. D’Agostino RB, S.Vasan R, Pencina MJ, et al. General Cardiovascular Risk Profile for Use in Primary Care. The Framingham Heart Study. Circulation 2008;117(6):743-53. http://circ.ahajournals.org/cgi/content/abstract/CIRCULATIONAHA.107.699579v1
6. Anderson KM, M.Odell P, Wilson PW, et al. Cardiovascular disease risk profiles. American Heart Journal 1991;121(1 Pt 2):293-8. http://www.ncbi.nlm.nih.gov/pubmed/1985385
7. Arntzenius AC, Kromhout D, Barth JD, et al. Diet, lipoproteins, and the progression of coronary atherosclerosis. The Leiden Intervention Trial. New England Journal of Medicine 1985;312(13):805-11. http://www.nejm.org/doi/full/10.1056/NEJM198503283121301
8. Lichtenstein AH, Ausman LM, Jalbert SM, et al. Effects of different forms of dietary hydrogenated fats on serum lipoprotein cholesterol levels. N Engl J Med 1999;340:1933–40. http://www.nejm.org/doi/full/10.1056/NEJM199906243402501
9. Ockene IS, Miller NH. Cigarette smoking, cardiovascular disease, and stroke: a statement for healthcare professionals from the American Heart Association, American Heart Association Task Force on Risk Reduction. Circulation 1997;96:3243–47. http://circ.ahajournals.org/content/114/1/82.full#ref-26
10. Kraus WE, Houmard JA, Duscha BD, et al. Effects of the Amount and Intensity of Exercise on Plasma Lipoproteins. N Engl J Med 2002;347(19):1483-92. http://content.nejm.org/cgi/content/abstract/347/19/1483
11. Silberman A, Banthia R, Estay IS, et al. The effectiveness and efficacy of an intensive cardiac rehabilitation program in 24 sites. Am J Health Promot 2010;24(4):260-6. http://ajhpcontents.org/doi/10.4278/ajhp.24.4.arb?url_ver=Z39.88-2003&rfr_id=ori%3Arid%3Acrossref.org&rfr_dat=cr_pub%3Dpubmed&
12. Brown CD, Higgins M, Donato KA, Rohde FC, Garrison R, Obarzanek E, Ernst ND, Horan M. Body mass index and the prevalence of hypertension and dyslipidemia. Obes Res 2000;8(9):605-19. http://www.ncbi.nlm.nih.gov/pubmed/11225709
13. Cundiff DK. Diet and tobacco use: analysis of data from the Diabetes Control and Complications Trial, a randomized study. MedGenMed 2002;4(1):1-8. http://www.medscape.com/viewarticle/414943
14. The Diabetes Control and Complications Trial Research Group. DCCT Protocol. Springfield, Va: National Technical Information Service. US Dept of Commerce publication, 1988.
15. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993;329(14):977-86. http://www.ncbi.nlm.nih.gov/pubmed/8366922
16. Burke BS. Diet History as a tool in research. JAMA 1947;23:1041-46.
17. The Diabetes Control and Complications Trial (DCCT): design and methodologic considerations for the feasibility phase. Diabetes 1986;35:530-45. http://diabetes.diabetesjournals.org/content/35/5/530.long
18. Schmidt LE, Cox MS, Buzzard IM, et al. Reproducibility of a comprehensive diet history in the Diabetes Control and Complications Trial. J Am Diet Assoc 1994;94(12):1392-7. http://www.ncbi.nlm.nih.gov/pubmed?term=(Reproducibility%20of%20a%20comprehensive%20diet%20history%20in%20the%20Diabetes%20Control%20and%20Complications%20Trial)
19. FAOSTAT. Food and Agriculture Organization (FAO) Statistical Yearbook 2010, Volume D9 – Consumption of 10 major vegetal foods (2005-2007) and D10 – Consumption of 10 major animal foods (2005-2007). Food and Agriculture Organization 2014 2014 http://www.fao.org/economic/ess/ess-publications/ess-yearbook/ess-yearbook2010/yearbook2010-consumption/en/
20. Global Health Observatory Data Repository: Risk factors: Cholesterol. World Health Organization: Global Burden of Disease 2013 http://apps.who.int/gho/data/node.main.A883?lang=en
21. Quantity, Consumption of pure alcohol by type of beverage. WHO Global Health Observatory Data Repository 2013 http://www.who.int/gho/alcohol/en/index.html
22. Hallal PC, Andersen LB, Bull FC, et al. Global physical activity levels: surveillance progress, pitfalls, and prospects. Lancet 2012;380(9838):247-57. http://www.ncbi.nlm.nih.gov/pubmed/22818937
23. Global Health Observatory Data Repository: Risk factors: Tobacco Data by country World Health Organization: Global Burden of Disease 2013 http://apps.who.int/gho/data/node.main.65?lang=en
24. USDA Nutrient Database for Standard Reference, Release 24. U.S. Department of Agriculture. Agricultural Research Service 2011 http://www.ars.usda.gov/research/publications/publications.htm?seq_no_115=272030
25. Cundiff DK, Raghuvanshi N. Future body mass index modelling based on macronutrient profiles and physical activity. Theoretical Biology and Medical Modelling 2012;9(43):epub. http://www.tbiomed.com/content/9/1/43
26. Global Health Observatory Data Repository: Risk factors: Physical inactivity. World Health Organization: Global Burden of Disease 2013 2013 http://apps.who.int/gho/data/node.main.A892?lang=en
27. National Health and Nutrition Examination Survey. NHANES 2001-2002 Data Files – Data, Docs, Codebooks, SAS Code. US Department of Health and Human Services. Centers for Disease Control and Prevention. National Center for Health Statistics 2011 2011 http://www.ars.usda.gov/SP2UserFiles/Place/80400530/pdf/0102/usualintaketables2001-02.pdf
28. Sugars intake for adults and children Guideline. World Health Organization 2015 http://www.who.int/nutrition/publications/guidelines/sugars_intake/en/