Study design and population
The LifeLines Cohort Study is a large ongoing observational population-based cohort study that investigates health and health-related behaviors of more than 167,000 individuals. A detailed description of the Lifelines Cohort Study can be found elsewhere [8, 9]. Participants were recruited from the three Northern provinces of the Netherlands between 2006 and 2013. In short, the first group of participants was recruited via local general practices. Participants could indicate whether family members were interested as well. In addition, individuals who were interested in the study had the possibility to register via an online self-registration. Individuals with insufficient knowledge of the Dutch language, with severe psychiatric or physical illness, and those with limited life expectancy (< 5 years) were excluded from the study. Participants completed several questionnaires, including topics such as the occurrence of diseases, general health, medication use, diet, physical activity, and personality. Participants were invited to the Lifelines Research sites for a comprehensive health assessment and to allow storage of biological samples, including plasma, serum, and 24-h urine samples in the biobank underlying the LifeLines cohort. All participants provided written consent. The Lifelines Cohort Study was conducted according to the principles of the Declaration of Helsinki and approved by the Medical ethical committee of the University Medical Center Groningen, the Netherlands (METc approval number 2007/152).
The present study included individuals of the LifeLines-MINUTHE (MIcroNUTrients and Health disparities in Elderly) subcohort of the LifeLines Cohort Study. This subcohort consists of 1605 individuals aged between 60 and 75 years, with available plasma, serum, and 24-h urine samples from the biobank of the LifeLines cohort. The 1605 individuals comprised 400 men and 403 women with low socioeconomic status (SES) and 402 men and 400 women with high SES. Since education is more differentiating than income in the Dutch population, the classification of SES was based on educational status. Low SES was defined as never been to school or elementary school only, or completed lower vocational or secondary schooling; high SES was defined as completed higher vocational schooling or education. In the present study, we included 1533 individuals with available MMA, vitamin B12, and eGFR measurements.
Data collection and measurements
Data regarding demographics, education, smoking status, and general health were collected from self-administered questionnaires. Anthropometric measurements and blood pressure were measured by well-trained staff. BMI was calculated as weight (kg) divided by height squared (m2). Systolic and diastolic blood pressures were measured 10 times during a period of 10 min using an automated Dinamap Monitor (GE Healthcare, Freiburg, Germany). The average of the final three readings was used for each blood pressure parameter.
Blood samples were collected in a fasting state between 8.00 and 10.00 a.m. and subsequently transported to the Central Lifelines Laboratory in the University Medical Center Groningen. MMA was measured using LC-MS/MS. Vitamin B12 was measured using an electrochemiluminiscence immunoassay on a Roche Cobas chemistry analyzer (Roche, Mannheim, Germany). Serum creatinine (SCr) was measured via an enzymatic assay with colorimetric detection on a Roche Cobas chemistry analyzer (Roche, Mannheim, Germany). The creatinine-based CKD-EPI formula was used to obtain the estimated glomerular filtration rate (eGFR) [10]. Other laboratory measurements, including plasma total homocysteine, were assessed by commercially available assays on a Roche Cobas chemistry analyzer (Roche, Mannheim, Germany).
Clinical endpoints
In the present study, we investigated the association of MMA with all-cause mortality. Data on mortality were obtained from the municipal register.
Statistical analyses
Statistical analyses were performed using SPSS version 25 for Windows (IBM Corporation, Chicago, IL), STATA version 13.1 (StataCorp LP, TX, USA), and R version 3.5.2 (R Core Team (2017); R: A language and environment for statistical computing; R Foundation for Statistical Computing, Vienna, Austria; URL https://R-project.org/). Results were expressed as mean ± standard deviation (SD) or median (interquartile range) for normally and non-normally distributed data, respectively. Nominal data were presented as the total number of patients (percentage). A two-sided P < 0.05 was considered to indicate statistical significance.
Baseline characteristics are presented for the total study population and for tertiles of baseline MMA concentrations. P values for differences between tertiles were assessed using ANOVA for normally distributed data, Kruskal-Wallis test for skewed data, and the χ2 test for nominal data.
We used linear regression analyses to investigate the cross-sectional associations of MMA with vitamin B12, eGFR, and other parameters including SES. Logarithmic transformation of variables was used to fulfill criteria for linear regression analyses if necessary. First, univariable linear regression analyses were conducted. In addition, we tested for interactions between variables using multivariable linear regression analyses. Finally, multivariable linear regression models were developed using stepwise backward selection, without and with inclusion of the interaction term for vitamin B12 and eGFR (model 1 and model 2, resp.). Variable exclusion in the backward stepwise selection procedure was set at a P value of 0.2; the P value for subsequent variable inclusion was set to 0.05. Results for variables with a P value of > 0.2 in univariable and multivariable linear regression analyses were not shown. R2 and adjusted R2 values were obtained to assess the proportion of variability in the data accounted for by single variables and the multivariable models. The R package plot3D was used to depict the cross-sectional interaction between vitamin B12 and eGFR with MMA levels.
We used Cox regression analyses to investigate the prospective association of MMA with all-cause mortality. We applied a log2 transformation of MMA values so the hazard ratios were expressed as an increase in risk per doubling of baseline MMA values. Cox regression analyses were also used to test for interaction between MMA and eGFR. Various Cox regression models were built to adjust for possible confounders. The first model depicts the interaction between log2 MMA and eGFR for the risk of mortality; model 2 was adjusted for age and sex; and model 3 was additionally adjusted for SES, smoking, alcohol intake, BMI, SBP, vitamin B12, and use of vitamin. In secondary analyses, we investigated whether prospective associations for MMA were paralleled by prospective associations for plasma total homocysteine. Model 1A and model 1B depict the interaction between log2 total plasma homocysteine and eGFR with respect to mortality and the interaction between log2 MMA and eGFR with respect to mortality; model 2A and model 2B were fully adjusted for potential confounders, with additional adjustment for log2 MMA in case of the analysis for total plasma homocysteine (model 2A) and additional adjustment for log2 total plasma homocysteine in case of the analysis for MMA (model 2B); and model 3 was fully adjusted for potential confounders and included both the interaction between log2 total plasma homocysteine and eGFR with respect to mortality and the interaction between log2 MMA and eGFR with respect to mortality in one model. The assumption of proportional hazards was investigated by inspecting the Schoenfeld residuals. The R package plot3D was used to depict the interaction of MMA and eGFR in their association with mortality. As sensitivity analyses, we stratified Cox regression analyses for SES. In addition, we repeated Cox regression analyses after exclusion of subjects that used multivitamin or vitamin B supplements.