Associations of the MCM6-rs3754686 proxy for milk intake in Mediterranean and American populations with cardiovascular biomarkers, disease and mortality: Mendelian randomization

Controversy persists on the association between dairy products, especially milk, and cardiovascular diseases (CVD). Genetic proxies may improve dairy intake estimations, and clarify diet-disease relationships through Mendelian randomization. We meta-analytically (n ≤ 20,089) evaluated associations between a lactase persistence (LP) SNP, the minichromosome maintenance complex component 6 (MCM6)-rs3754686C>T (nonpersistence>persistence), dairy intake, and CVD biomarkers in American (Hispanics, African-American and Whites) and Mediterranean populations. Moreover, we analyzed longitudinal associations with milk, CVD and mortality in PREDIMED), a randomized Mediterranean diet (MedDiet) intervention trial (n = 7185). The MCM6-rs3754686/MCM6-rs309180 (as proxy), LP-allele (T) was strongly associated with higher milk intake, but inconsistently associated with glucose and lipids, and not associated with CVD or total mortality in the whole population. Heterogeneity analyses suggested some sex-specific associations. The T-allele was associated with higher CVD and mortality risk in women but not in men (P-sex interaction:0.005 and 0.032, respectively), mainly in the MedDiet group. However, milk intake was not associated with CVD biomarkers, CVD or mortality either generally or in sub-groups. Although MCM6-rs3754686 is a good milk intake proxy in these populations, attributing its associations with CVD and mortality in Mediterranean women to milk is unwarranted, as other factors limiting the assumption of causality in Mendelian randomization may exist.

: P-values for differences in sex and differences in race. Chi-squared tests were used to test differences in percentages. We used ANOVA test to compare means of continuous variables.

African American Hispanic American
*: Values are expressed as mean ± standard deviation for continuous variables or as % for categorical variables. The rs3754686 SNP was determined in PREDIMED and imputed in BPRHS. The proxy rs309180 was genotyped in GOLDN and WHI studies.
*: Values are means ± Standard Error of Mean. The rs3754686 SNP was determined in PREDIMED and imputed in BPRHS. The proxy rs309180 was genotyped in GOLDN and WHI studies.
**: General Linear Regression models with multivariable adjustment for the indicated covariates were fitted for each population. 1: P adjusted by sex, age, field center or race.

Supplemental table S6. Associations of MCM6-rs3754686 proxy for milk intake with fasting glucose and lipids in women in the studied populations WHI African Americans WHI Hispanic Americans
*: Values are means ± Standard Error of Mean. The rs3754686 SNP was determined in PREDIMED and imputed in BPRHS. The proxy rs309180 was genotyped in GOLDN and WHI studies.
1: P adjusted by sex, age, field center or race. 2: P adjusted for sex, age, field center or ancestry (BPRHS, WHI), family (GOLDN), BMI, smoking, drinking, physical activity, diabetes, medication and total energy intake. In PREDIMED, some variables (glucose, LDL-C, HDL-C and triglycerides) included missing data point. Biochemical data were available for fasting glucose (n = 3886 participants) total cholesterol (n = 3964 participants), HDL cholesterol (n = 3912 participants), LDL cholesterol (n = 3883 participants), and triglycerides (n = 3935 participants). **: General Linear Regression models with multivariable adjustment for the indicated covariates were fitted for each population. : P adjusted by sex, age, field center or ancestry (BPRHS, W HI), family (GOLDN), BMI, smoking, drinking, physical activity, diabetes, medication and total energy intake. Beta indicates the increase/decrease of the CVD risk factor in mg/dL per 100 g/d of milk consumed. SE is expressed in the same units too.

Women n = 2,726
Per variant allele (T)**** Model 4: Model 3 additionally adjusted for total fat and carbohydrates at baseline.

Supplemental table S10. Incidence and hazard ratios (HR) for CVD depending on the MCM6-rs3754686 polymorphism after 4.8 years of median follow-up for the Mediterranean Diet intervention group and stratified by sex
Model 3: Model 2 adjusted for variables in model 2 plus total milk intake.
*: Crude incidence rates were expressed per 1000 person-years of follow-up. **: Codominant model. ***: Recesive model.****: Additive model. We used multivariable Cox regression models with length of follow-up as the primary time variable. Separate models were fitted for CVD and total mortality to estimate the corresponding HRs depending on the model.

Supplemental table S11. Incidence and hazard ratios (HR) for CVD depending on the MCM6-rs3754686 polymorphism after 4.8 years of median follow-up for the Control group and stratified by sex Total (men + women) n = 2,340
Model 3: Model 2 adjusted for variables in model 2 plus total milk intake.
*: Crude incidence rates were expressed per 1000 person-years of follow-up. **: Codominant model. ***: Recesive model.****: Additive model. We used multivariable Cox regression models with length of follow-up as the primary time variable. Separate models were fitted for CVD and total mortality to estimate the corresponding HRs depending on the model.

Women n = 2,726
Model 4: Model 3 additionally adjusted for total fat and carbohydrates at baseline.

Supplemental table S12. Incidence and hazard ratios (HR) for total mortality depending on the MCM6-rs3754686 polymorphism after 4.8 years of median follow-up for the Mediterranean Diet intervention group and stratified by sex
Model 3: Model 2 adjusted for variables in model 2 plus total milk intake.
*: Crude incidence rates were expressed per 1000 person-years of follow-up. **: Codominant model. ***: Recesive model.****: Additive model. We used multivariable Cox regression models with length of follow-up as the primary time variable. Separate models were fitted for CVD and total mortality to estimate the corresponding HRs depending on the model. Model 1: Adjusted for sex, age, field center and dietary intervention group. Model 2: Model 1 adjusted for variables in model 1 plus BMI, diabetes, drinking, smoking, physical activity, medication (hypertension, dyslipemia and glucose) and total energy intake at baseline.

Women n = 1,394
Model 4: Model 3 additionally adjusted for total fat and carbohydrates at baseline.

Supplemental table S13. Incidence and hazard ratios (HR) for total mortality depending on the MCM6-rs3754686 polymorphism after 4.8 years of median follow-up for the Control group and stratified by sex
Model 3: Model 2 adjusted for variables in model 2 plus total milk intake.
*: Crude incidence rates were expressed per 1000 person-years of follow-up. **: Codominant model. ***: Recesive model.****: Additive model. We used multivariable Cox regression models with length of follow-up as the primary time variable. Separate models were fitted for CVD and total mortality to estimate the corresponding HRs depending on the model.

Study: Boston Puerto Rican Health Study BPRHS
Biochemical measurements were measured in a state and federal licensed laboratory according to standard operating procedures. Control of pre-analytical variation was maximized through adherence to a manual of operations. Glucose was measured with intra-and inter-assay C.V.s of 2.0% and 3.2% respectively. Total cholesterol was measured by an enzymatic procedure with intra-and inter-assay CVs of 2.0% and 2.8% respectively. Triglycerides were measured by a series of coupled enzymatic reactions with intra-and interassay CVs of 2.0% and 3.4% respectively. HDL was measured by a two-phase reaction with colorimetric endpoint detection, with intra-and inter-assay CVs of 3.0% and 5.0% respectively.

Health Data Collection Genotyping
Weight was taken with minimal clothing on a balance. Results were recorded to the nearest pound, rounding down. Height was measured while standing as straight as possible without shoes with feet flat on the floor. Height was recorded to the nearest centimeter, rounding down to the nearest centimeter. Clinical and lifestyle questionnaires were administered by a trained interviewer.
F o r t h e G O L D N s t u d y , g e n o t y p e s w e r e obtained using the genome-wide Human SNP Array 6.0 (Affymetrix, Santa Clara, CA, USA, www.affymetrix.com). In GOLDN, the proxy SNP MCM6-rs309180, with a high LD with the MCM6-rs3754686 (D'=1 and r 2 =0.95), was genotyped.

Blood sample collection and handling
Blood was collected after a 12 hour overnight fast. Specimens were required to be non-hemolyzed. Blood from visits 1-4 were stored at -70 degrees C at field centers until all visits were complete, and then analysis was performed. For the current study, analyses were performed with blood obtained at visit 2. Blood for glucose and lipids was stored in yellow cap tubes.

Laboratory Quality Control
Quality control procedures were performed quarterly by the Collaborative Studies Clinical Laboratory (CSCL) Minneapolis, Minnesota in conjunction with the CDC. Lipid proficiency specimens (n=36) were distributed by the CDC and assayed by the CSCL in batches at weekly intervals. Acceptable accuracy limits were as follows: total cholesterol: CDC mean ± 3%, triglycerides: <200 = CDC mean ± 8 mg/dL; triglycerides >200= CDC mean ± 5%, HDL-C: CDC mean ± 5% The PREDIMED trial (Prevención con Dieta Mediterránea) was a parallel-group, multicenter, randomized trial. The trial was designed and conducted by the authors, and the protocol was approved by the institutional review boards at all study locations.
Beginning on October 1, 2003, participants were randomly assigned, in a 1:1:1 ratio, to one of three dietary intervention groups: a Mediterranean diet supplemented with extra-virgin olive oil, a Mediterranean diet supplemented with nuts, or a control diet. Randomization was performed centrally by means of a computer-generated random-number sequence. From October 2003 through June 2009, a total of 8713 candidates were screened for eligibility, and 7447 were randomly assigned to one of the three study groups.
Eligible participants were men (55 to 80 years of age) and women (60 to 80 years of age).

Health Data Collection Genotyping
Weight, height, and waist circumference were directly measured. A general medical questionnaire, a 137-item validated food-frequency questionnaire, and the Minnesota Leisure-Time Physical Activity Questionnaire were administered on a yearly basis. Information from the foodfrequency questionnaire was used to calculate intake of energy and nutrients.
The primary end point was a composite of myocardial infarction, stroke, and death from cardiovascular causes. Secondary end points were stroke, myocardial infarction, death from cardiovascular causes, and death from any cause. We used four sources of information to identify end points: repeated contacts with participants, contacts with family physicians, a yearly review of medical records, and consultation of the National Death Index. All medical records related to end points were examined by the end-point adjudication committee, whose members were unaware of the study-group assignments. Only end points that were confirmed by the adjudication committee and that occurred between October 1, 2003, and December 1, 2010, were included in the analyses.
Genomic DNA was extracted from buffy coat. We genotyped the MCM6-rs3754686 polymorphism in all PREDIMED participants with DNA available on a 7900HT Sequence Detection System (Applied Biosystems) by using a fluorescent allelic discrimination TaqMan assay.

Blood sample collection and handling
Blood samples were obtained after an overnight fast and were frozen at −80°C. Fas ng glucose, total cholesterol, triglycerides, HDL cholesterol, and LDL cholesterol were measured by using standard enzymatic methods. In participants whose triglyceride levels were <400 mg/dL, LDL cholesterol concentrations were estimated by using the Friedewald formula. Biochemical measures were available for nearly 7000 participants at baseline.

Laboratory Quality Control
Biochemical analysis was carried out in regional and national licensed laboratories according to standard operating procedures. Control of pre-analytical variation was maximized through adherence to a manual of operations. Glucose was measured with intra-and inter-assay C.V.s of 2.0% and 3.2% respectively. Total cholesterol was measured by an enzymatic procedure with intra-and inter-assay C.V.s of 2.0% and 2.8% respectively. Triglycerides were measured by a series of coupled enzymatic reactions with intra-and interassay C.V.s of 2.0% and 3.4% respectively. HDL was measured by a two-phase reaction with colorimetric endpoint detection, with intra-and inter-assay C.V.s of 3.0% and 5.0% respectively.

Study History and Recruitment
Age/sex WHI began in 1993 and is ongoing. Participants were recruited and enrolled at 40 clinical centers throughout the US. Recruitment methods included: mass mailings (primary method), community presentations, newspapers and articles, TV and radio, and health fairs. A total of 161,808 postmenopausal women aged 50-79 years old were recruited. Data for the current study were collected at baseline from the Observational Study, and were limited to African American and Hispanic women.

Health Data Collection Genotyping
Health information relevant to the current study (such as weight and height) was collected at a clinic visit by trained clinical stuff. Other information, such as age and ethnicity were collected at baseline by self-report. At baseline and the first follow-up clinic visit, which occurred 3 years after baseline, Observational Study participants completed questionnaires on medical, lifestyle and psychosocial characteristics.

F o r t h e W H I S N P H e a l t h A s s o c i a t i o n
Resource (SHARe) study, genotypes were obtained using the genome-wide Human SNP Array 6.0 (Affymetrix, Santa Clara, CA, USA, www.affymetrix.com). In WHI, the proxy SNP MCM6-rs309180, with a high LD with the MCM6-rs3754686 (D'=1 and r 2 >0.96), was genotyped.

Blood sample collection and handling
Blood was collected after a 12 hour fast and was maintained at 4 degrees C for up to 1 hour until plasma or serum was separated from cells. Centrifuged aliquots were stored in freezers (at -70 degrees C) within 2 h of collection and sent on dry ice to the central repository, where storage at -70 degrees C was maintained.

Laboratory Quality Control
The accuracy and precision of the lipid assays were regularly monitored with the CDC/NHLBI Lipid Standardization Program to control for any potential drift over time. The CDC/NHLBI Laboratory Quality Assurance and Standardization Program provides the clinical laboratory community with performance guidelines and equipment recommendations that meet the following standards: 1) cholesterol tests with bias from the reference method ≤ 3.0% and a coefficient of variation (CV) ≤ 3.0% 2) HDL cholesterol should be measured with a bias from the reference method ≤ 5% and methods perform with a CV ≤ 4% at ≥ 42 mg/dL (1.09mmol/L) and a standard deviation of ≤ 1.7 mg/dL (0.044 mmol/L) at < 42 mg/dL (1.09 mmol/L) and 3) LDL cholesterol with a bias from the reference method ≤ 4% and perform with a CV ≤ 4%.