*2.1. Variable Definitions*

The definition of hyperhomocysteinemia in healthy children and adolescents is not well defined in WKHOLWHUDWXUH&XWRIIVKDYHYDULHGIURPWRȝPRO/DQGGLIIHUE\DJHDQGHWKQLFLW\>–22]. In adults, hyperhomocysteinemia is commonly defined as FRQFHQWUDWLRQVJUHDWHUWKDQȝPRO/>@ RUȝPRO/>@%HFDXVHWKHUHLVQRVSHFLILFFXW-off for defining high plasma homocysteine level in children, population-specific 85th percentile values may guide the definition of hyperhomocysteinemia for FKLOGUHQ >@ ,Q WKLV DQDO\VLV WKH WK SHUFHQWLOH RI +F\ ZDV ȝPRO/ FORVH WR WKH ZLGHO\ accepted cutoff for hyperhomocysteinemia in adults, and used in previously published reports [22]. 7KHUHIRUHZHFKRVHDFXWRIIRIȝPRO/WRLQGLFDWHhyperhomocysteinemia, despite lower cutoffs sometimes used in children, because it better reflected the risk of higher homocysteine observed in Asian populations [22] and because it is conventionally used. We have established and reported previously the cutoffs used for other components of MetS and the definition of MetS [9,25–28], which are summarized in Table 1.


**Table 1.** Variable definitions.

(1) As described by the NCEP criteria set for adults, because there is no separate recommendation for children [26]; (2) As described by the NCEP criteria for cholesterol in children and adolescents [27]; BMI, basal metabolic rate; BP, blood pressure; TG, triglyceride; HDL, high density lipoprotein; PG, plasma glucose; FPI, fasting plasma insulin; MetS, metabolic syndrome.

#### *2.2. Statistical Analysis*

Anthropometric measures, BP and biomarkers were examined by Hcy ( ȝPRO/ *vs*. !ȝPRO/ using independent sample t-tests. Values were expressed as mean ± SD.

The association of hyperhomocysteinemia (>12.0 ȝPRO/ZLWK0HW6DQGGLFKRWRPL]HGLQGLYLGXDO components of MetS was expressed as odds ratios (OR) determined by separate logistic regressions for each outcome variable, with adjustment for age and gender. We also estimated OR for the combined risk of having multiple MetS components (such as low HDL and high BMI combined) against hyperhomocysteinemia through multiple logistic regression. Since we had previously shown an effect of maternal antenatal micronutrient supplementation, particularly with folic acid, on MetS in these children [9], we also initially adjusted our regression models for maternal intervention groups and birth weight, but this adjustment is not reported as these variables were not statistically significant. Likewise we examined statistical models adjusted for various aspects of socioeconomic status (SES), including ownership of televisions, radios, bicycles, livestock and use of electricity, as well as for seasonal effects. However, adjustment for these variables is not reported as their influence was not statistically significant. Finally, we explored models adjusted for fasting status and observed no effect on the OR; thus, that adjustment was not included in our results. Risks were expressed as OR and associated 95% confidence intervals (CI). All statistical analyses were done with IBM SPSS® (Statistical Package for Social Sciences, IBM Corp., Armonk, NY, USA) software version 21 for Windows®.
