To the Editor:
Lewis and colleagues report that calcium supplements had no effect on the incidence of atherosclerotic vascular disease.1 We do not agree that these results differ from previous studies or provide “compelling evidence” of the cardiovascular safety of calcium supplements. Instead, these findings are severely limited by a failure to report important data or capture all relevant events and a lack of statistical power.
Previously, we reported that calcium supplements increased the risk of myocardial infarction (MI) and stroke by 25% to 30% and 15% to 20%, respectively.2–4 Lewis and colleagues have not reported data for either of these endpoints, so their results are not comparable with ours. Instead, they chose a very broad composite endpoint of atherosclerotic vascular disease that included several conditions not necessarily related to atherosclerosis, such as congestive heart failure and atrial fibrillation. The use of composite endpoints with diverse components that occur at markedly different frequencies is contrary to current recommendations.5 In this case, the contribution of MI or stroke to the composite endpoint will be greatly outweighed by that of more frequent events that are unrelated to calcium supplementation. Thus the choice of composite endpoint likely obscured any effect calcium supplements might have on MI or stroke while producing a null effect for the composite endpoint. Importantly, Lewis and colleagues did provide trial-level data on self-reported MI and stroke from their study for our meta-analysis3 that were incorporated into the estimates of cardiovascular risk with calcium supplements described earlier.
We were surprised at the very low event rates reported by Lewis and colleagues. At 5 years, only 14% of women had a vascular death or hospitalization. In our similar 5-year study of calcium supplements, 7% of women had an MI or stroke.2 MI or stroke was the principal diagnosis in 18% of hospital admissions for heart, stroke, or vascular disease in Australia in 2001–2002.6 Based on these data, the expected proportion of women in our study with a hospital admission for heart, stroke, or vascular disease would be 30% to 35%. The event rates reported by Lewis and colleagues are probably lower because they considered only first hospitalizations and the first diagnostic code for each hospitalization. Thus it appears that if a person was admitted with atrial fibrillation and later with MI, the second admission was not considered, and if the primary diagnostic code was hip fracture but the person also sustained an MI during hospitalization, the MI was not counted. This approach seems to have considerably underestimated event rates. Consequently, the study did not have sufficient power to detect a 20% to 30% difference between groups in MI, stroke, or the chosen composite endpoint, which is reflected in the wide confidence intervals of the reported hazard ratios and renders the results far from compelling.