Abstract 4200: Does Homocysteine Lowering Impact Cardiovascular Outcomes? A Bayesian Meta-Analysis of NORVIT and HOPE-2 Trials
Background: In HOPE-2 trial, homocysteine lowering with B vitamins failed to reduce clinical outcomes compared to placebo except for a significant reduction in stroke (P=0.03). The stroke benefit was thought to “represent either an overestimate of the real effect or a spurious result due to play of chance, and that a meta-analysis should be able to clarify the issue”.
Objective: To perform a Bayesian meta-analysis by integrating prior information from the NORVIT trial with the empirical evidence from HOPE-2 (the only 2 large completed placebo-controlled trials of homocysteine lowering).
Methods: Posterior probabilities that the benefit (Pb) or harm (Ph) exceeds a specified difference (δ<0%, 10%) were estimated via the Bayes’ theorem: posterior α evidence x prior. Only data from comparison of combination vitamin therapy group vs placebo was utilized from NORVIT to match the HOPE-2 treatment groups for primary composite endpoint of cardiovascular death, myocardial infarction (MI) or stroke and the individual endpoints. Superiority or inferiority was inferred at a posterior probability of benefit or harm >0.950.
Results: The Ph exceeded Pb for the primary composite endpoint and each of the individual components except for stroke where the probability of any benefit (δ>0%) was 0.992 and the probability of >10% risk reduction (δ>10%) was 0.930.
Conclusions: Thus, a Bayesian meta-analysis helps clarify the issue, and suggests a low likelihood of coronary benefit and a high likelihood of stroke benefit with homocysteine-lowering therapy compared to placebo. These data are consistent with the epidemiologic observations of a stronger association of elevated homocysteine with stroke compared to CAD (adjusted odds ratio of 1.32 (1.18–1.49) vs 1.19 (1.12–1.25) for +5μmol/L increase in homocysteine level observed in prospective studies). Ongoing trials should further clarify the impact of homocysteine lowering.