Dr. Cher points out some inconsistencies between two independently performed meta-analyses [1, 2] on the same research question. A comparison of the two papers shows that Bucher and colleagues identified several studies that we did not. After reviewing several of these studies, we concluded that our MEDLINE search could have been improved by including additional generic terms for dietary factors. However, it was disturbing to discover that Bucher and associates included the study by Singh and colleagues [3], a randomized trial of the effect of guava fruit intake on blood pressure in 145 patients with essential hypertension. Singh and associates reported changes of –7.5 mm Hg in systolic blood pressure and –8.5 mm Hg in diastolic blood pressure, attributable to increased potassium intake and a reduced sodium:potassium ratio. Bucher and colleagues included these blood pressure changes in their Table 2 as if they were attributable to changes in calcium intake. This goes far beyond failing to consider a confounding variable, because the main effects of other dietary factors known to influence blood pressure are ascribed to calcium. This trial represents 6% of the sample size of the meta-analysis.

Dr. Cher mentions the difference in sample size for the study by Tanji and colleagues [4]. Although 28 patients enrolled in this crossover trial, only 19 completed it; the published results are based on these 19 patients, not on the 28 randomly assigned patients. We excluded this trial from our pooled analysis because the data could not be pooled as presented; the study design was crossover, but the data were presented as if the design were parallel.

Dr. Cher notes that two of the studies were classified as having hypertensive patients in one meta-analysis and normotensive persons in the other. Bucher and colleagues reported that six studies included subgroups of hypertensive patients, although they did not define hypertension. We prospectively abstracted information on normotensive and hypertensive status as reported by the investigators for each study or study component. We classified as hypertensive seven studies that Bucher and colleagues listed in their Table 1 under the heading normotensive participants. Having examined these papers again, we stand by our original classification. We find it puzzling that Bucher and colleagues considered these studies to be trials in normotensive persons, particularly given that they abstracted in duplicate, as we did. Bucher and colleagues also included the study by Singh and associates [3] as a study of normotensive persons, even though it was a trial of 145 persons with essential hypertension.

In their letter, Bucher and colleagues question our subgroup analyses of normotensive and hypertensive persons. We did not use a 130-mm Hg systolic blood pressure cutpoint for hypertension, as they suggest. We simply reported the findings separately for normotensive and hypertensive persons from the pooled data set, as we did for parallel and crossover trials. Although we did not clearly state this as an a priori hypothesis, we had planned to do these analyses from the beginning. Results among hypertensive patients are of particular value to clinicians, whereas results among normotensive persons serve public health practitioners.

We did not do a regression analysis on the effect of calcium supplementation based on baseline blood pressure, as was done by Bucher and colleagues. Although most, but not all, of the trials that included persons receiving antihypertensive medications had a washout period, most allowed only 2 to 4 weeks. Bucher and colleagues do not mention this as a possible confounder in this analysis.

Despite the differences described above, these independently performed meta-analyses found similar estimates of the effect of dietary calcium supplementation on blood pressure. A meta-analysis of observational studies of dietary calcium intake and blood pressure [5] found effect sizes that were also similar when adjusted to the median calcium dose of the clinical trials.