SINCE THE PUBLICATION OF SEVERAL RANDOMIZED TRIALS and a meta-analysis indicating that lowering homocysteine levels with B vitamins (to reduce the effects of homocysteine on the vascular endothelium) did not result in cardiovascular benefit, the use of vitamin therapy to lower homocysteine levels is widely regarded as ineffective. However, there has been renewed interest in the issue because an analysis of studies from the National Health and Nutrition Examination Survey and the Multi-Ethnic Study of Atherosclerosis showed that adding total homocysteine level to a Framingham risk score was associated with an approximately 20% net reclassification of intermediate-risk patients. A recent meta-analysis raised the question of whether population folate intake and polymorphisms of methylenetetrahydrofolate reductase may alter interpretation of these clinical trials; mendelian randomization analysis suggested that the polymorphisms might be important only among individuals with low folate status. A commentary about that article noted the hazards of the “spinal reflex” that leads to “automatic rejection of observational data when they appear to be discrepant from trials.” A similar spinal reflex, automatic rejection of subgroup analyses, is another hazard of interpretation of clinical trials. Subgroup analyses founded in biology may have the potential to importantly inform interpretation of clinical trial results, provided sufficient caution is exercised. Thoughtfully derived subgroups, especially those formed a priori and not derived post hoc from the same data set, can stimulate further work and reinterpretation of existing data. The question of whether B vitamin therapy to lower homocysteine levels may reduce the risk of stroke is becoming more complicated, and analyses that depended on grouping all patients together may have obscured important differences among patient (and perhaps population) subgroups. For instance, the main results of the Vitamin Intervention for Stroke Prevention (VISP) trial showed no effect of B vitamins (folic acid/pyridoxine/cyanocobalamin) on the risk of recurrent stroke, death, or myocardial infarction, whereas a subgroup analysis of the VISP trial showed benefit of B vitamins in a defined subset of the population. More recently, House et al reported that B vitamin therapy was shown to increase cardiovascular risk in patients with diabetic nephropathy. Hence, subgroup analyses have shown beneficial as well as adverse effects of B vitamin therapy. The VISP subgroup analysis excluded patients with B12 deficiency (because they were all receiving injections of cyanocobalamin, regardless of the treatment to which they were randomized), and patients with significant renal impairment (because it was thought they would not respond to vitamin therapy). There was a significant reduction of stroke or myocardial infarction with B vitamins in the remaining participants, and this finding was more pronounced when patients were stratified by baseline serum B12 level. The difference between patients who entered the study with a serum B12 level above the median (ie, they could absorb B12 reasonably well) and received high-dose vitamins and those who entered the study with a serum B12 level below the median and received low-dose vitamins was a 34% reduction of stroke, death, or myocardial infarction (P=.02). At the time, it seemed that the findings were explained mainly by exclusion of patients receiving vitamin B12 injections, and this seemed supported by the finding in the Heart Outcomes Prevention Evaluation 2 (HOPE-2) trial (the only large trial to use an adequate dose of B12 for elderly patients) that B vitamins significantly reduced the risk of stroke. In 2011, Hsu et al reported that in the VISP trial, a subgroup of patients with the GG phenotype of transcobalamin 2, a transport protein for vitamin B12, were responsive to high-dose vitamin therapy. Vitamin B12 may thus have a key role in stroke prevention interventions involving vitamin therapy used to lower homocysteine levels. Homocysteine increases thrombosis and is associated with a markedly increased risk of stroke in atrial fibrillation. The prevalence of atrial fibrillation increases steeply with age, as do metabolic B12 deficiency (not necessarily reflected in serum B12 levels, discussed below) and plasma total homocysteine levels. In the Framingham cohort, 1.5% of strokes at ages 50 through 59 years were attributable to atrial fibrillation, whereas by ages 80 through