The ingested choline and betaine was quantified, during the baseline visit (1987–1989), with a 66-item version of the Willett semi-quantitative food frequency questionnaire, FFQ [15]. The participants were asked how often, on average, they had consumed listed food items during the previous year. Nine frequency responses were listed ranging from more than six per day to almost never. We calculated daily servings by converting the consumption frequency to servings per day. Dietary choline and betaine were estimated as the sum of daily intakes, using a choline and betaine database composed with the U.S. Department of Agriculture (USDA) choline and betaine content in common foods database, database that contains 207 food items [10].

Incident non-fatal MI and fatal CHD were ascertained, validated and classified following the standardized ARIC cohort and community surveillance protocol [16]. We considered the intakes of choline, and choline plus betaine, in multivariable models. We adjusted for total energy intake as a continuous variable after evaluating that adjustment using the residual method [17] produced similar results. The following confounding variables were also included in the models: age, sex, education, dietary folate, methionine, B₆ vitamin and cholesterol, race, diabetes, ARIC center, menopausal status (reported cessation of menses), and a series of CHD risk factors such as smoking, hypertension, body mass index (BMI) and family history of CHD. Using a likelihood ratio test we assessed the effect measure modification of dietary folate intake, sex, menopausal status, race, education, ARIC center and alcohol intake. Due to women's capacity to form the choline moiety de novo the amount of choline necessary in the daily diet is influenced by sex and menopausal status. Alcohol intake, a known folate antagonist, may plausibly increase the requirement for folate intake. Folate and alcohol were categorized with a cutoff point at the lowest and the highest quartile, respectively. We calculated hazard ratios (HR), per quartiles of dietary intake, using Cox proportional hazard regression. Verification of the proportional hazard assumptions was assessed using plots of the log(-log) survival curves. We contrasted the results with those obtained using the procedure described below to correct for measurement error. Statistical analyses were conducted using SAS software [18]. All p-values were two-tailed.

Over an average of 14 years of follow-up (1987–2002), 1072 validated incident CHD events occurred. Neither higher intakes of dietary choline nor betaine were significantly associated with incident CHD. Compared with the lowest quartile of intake, incident CHD risk was 22% higher [HR = 1.22 (0.91, 1.64)] and 14% higher [HR = 1.14 (0.85, 1.53)] in the highest quartile of choline and choline plus betaine respectively, controlling for age, sex, education, total energy intake, and dietary intakes of folate, methionine and vitamin B₆ (Table ​(Table3).3). Further adjustment for race, diabetes status, ARIC field center and dietary cholesterol, vitamins B₁₂ and B₂, as well as for other CHD risk factors, such as obesity (defined as a body mass index, BMI, higher than 30), hypertension, smoking status and estimated family history of CHD, produced similar risk estimates.

Because eggs, milk and green vegetables are important contributors to choline and betaine intake, we performed sensitivity-type analyses to further adjust for intake of saturated fatty acids, animal fat, dietary fiber and animal protein. The hazard ratios across the quartiles of choline and total choline intakes (considering the same lowest quartile as referent, and adjusting for all confounders considered in models #2, Table ​Table3,3, namely age, sex, education, total energy intake, dietary intakes of folate, methionine, vitamin B6, race, diabetes status, ARIC field center, menopausal status and dietary cholesterol) remained practically unchanged, suggesting that the higher risk associated with the highest intake is not due to confounding from uncontrolled dietary factors (Table ​(Table33).

No effect measure modification was detected by sex, menopausal status, race, ARIC center, education or folate intake analyzed as a continuous variable. Correction for measurement error in the dietary intake of choline and related nutrients, using models containing continuous nutrient variables, provided similar results. For choline, the hazard ratio for those in the highest quartile, compared with those in the lowest, was 1.34 (0.98, 1.85). When using the bootstrapping procedure and obtained the 95% confidence interval for the corrected choline regression coefficient, the risk estimate remained non-significant.

To examine the effect of folate fortification that was mandated after January 1998 we conducted a supplemental analysis with exclusion of events that occurred prior to that date. For choline and for choline plus betaine the risk estimates were weakened and still not significant. Specifically, compared with the lowest quartile, the hazard ratios for choline were 0.80 (0.62, 1.04), 1.01 (0.76, 1.36) and 0.97 (0.64, 1.47) for the second, the third and the fourth quartile of the dietary intake, respectively. For choline plus betaine, the risks of incident CHD by quartiles were 0.86 (0.67, 1.11), 0.98 (0.73, 1.30) and 0.90 (0.61, 1.35), respectively.

Choline, an essential nutrient for humans [23], is found in several compounds that are methyl-donors. Supplementation in the dietary intake range of betaine, a methyl-donor continuously produced from choline [24], leads to immediate and long term lowering of plasma homocysteine [25], a putative CHD risk factor [26]. Homocysteine, a sulfur aminoacid whose metabolism has a direct cytotoxicity effect on vascular endothelium [27], stands at the intersection of two pathways [28]. One catalyzes the synthesis of the amino acid cysteine and the other remethylation to form methionine, a process that requires folate and vitamin B₁₂. In an alternative reaction, betaine, the oxidative by-product of choline, serves as a donor of methyl groups to homocysteine to form methionine [29]. Thus, the two metabolic pathways provide alternate mechanisms for removal of homocysteine as shown in Figure ​Figure1.1. The increase in blood homocysteine after a methionine load [3] and consequent vascular cytotoxicity, or the aberrant methylation produced by a low plasma choline and plasma betaine with possible increased atherogenesis [5,6], provide the putative mechanisms that could explain an increase in CHD risk. Alternatively, the intrinsec antioxidant properties of choline [30] could decrease the endothelial dysfunction and the risk for occlusive vascular events in those with a relatively higher intake of this nutrient.

Whereas there is research suggestive of an inverse association between dietary folate and incident CHD [31], the extant literature on dietary choline is small [11-13]; the current study is the first to explore choline as well as betaine in relationship with cardiovascular outcomes in a biracial cohort of both men and women. No relationship was also found recently between dietary intake of choline and betaine and cardiovascular events in a cohort of postmenopausal women enrolled in the Dutch EPIC cohort [13]. Until recently it was not possible to calculate dietary choline intake in humans and there are no nationally representative estimates of this intake from food [23] because the choline content of foods had not been included in major nutrient databases. Food choline data were unreliable due to older, imprecise assay procedures. As a consequence an estimated Average Requirement (EAR) for choline remains to be established. The proposed adequate intake (AI) for choline was set at 550 mg/day for men and 425 mg/day for women [32]. It is unknown whether intakes of choline in the U.S. meet the AI. In the ARIC cohort, the median and 75^th percentiles of choline intake were 284 and 367 mg/day, respectively (unpublished results). Only 6% of men and 11% of women had an intake of choline above that proposed as the AI for this population. Choline intake was associated with sex, race and menopausal status.

Repeat measurement mixed modeling permits the estimation of measurement reliability, which in turn permits adjustment for measurement error such as through a regression calibration [19-21]. One approach is to replace the observed values of the variables measured with error (in our case, the nutrients of interest) with multivariate Stein estimates of the true values, conditional on the values of the variables measured without error and the observed values of the variables measured with error [33]. Because the risk estimates relating nutrient intake to CHD were very close to the null, little change in the hazard ratio was observed when measurement error correction was applied.

An explanation for the slightly increased CHD risk toward the higher end of choline intake distribution, opposite to what we initially expected, could be that choline is required for normal secretion of very low density lipoprotein from liver; perhaps provision of choline mobilizes cholesterol from hepatic stores into the vascular pool permitting deposition in atheromas [33]. A higher intake of choline and betaine, which increases the methylation potential of methionine, may result in a change of the cell phenotype that also promotes the development of atherosclerotic plaque [35].

There are several limitations of the present study. First is that the food frequency questionnaire used in ARIC tends to underestimate the absolute dietary intake for a particular nutrient, as is commonly the case for semi-quantitative dietary assessment tools [15]. Nevertheless, the FFQ used in this study was designed to rank participants and it is likely that we properly discriminated individuals in the highest and lowest categories of intake which was the focus of our analyses. The foods with the highest content in choline – eggs, milk, liver, red meat, poultry and fish, as well as the foods with the highest content in betaine – spinach, white bread and breakfast cereals, were items included in the ARIC FFQ. One can overcome the limitations of a dietary tool through the use of a biomarker. However, there is no reliable blood biomarker for dietary intake of choline. As it is known from clinical settings, plasma concentrations of choline and betaine decrease when subjects are fed a low choline diet, but the amount of decrease is not highly correlated with susceptibility to develop organ dysfunction while on this diet [36]. Another limitation of our investigation is the absence of supplemental B vitamin information including folate, which was queried only during subsequent examinations of the ARIC cohort.