EXAMPLE 9.11 Vitamin A Saves Lives of Young Children

Vitamin A is often given to young children in developing countries to prevent night blindness. It was observed that children receiving vitamin A appear to have reduced death rates. To investigate the possible relationship between vitamin A supplementation and death, a large field trial with more than 25,000 children was undertaken in Aceh Province of Indonesia. About half of the children were given large doses of vitamin A, and the other half were controls. The researchers reported a 34% reduction in mortality (deaths) for the treated children who were one to six years old compared with the controls. Several additional studies were then undertaken. Most of the results confirmed the association: treatment of young children in developing countries with vitamin A reduces the death rate, but the size of the effect varied quite a bit.

469

How can we use the results of these studies to guide policy decisions? To address this question, a meta-analysis was performed on data from eight studies.⁴ Although the designs varied, each study provided a two-way table of counts. Here is the table for the study conducted in Aceh Province. A total of children were enrolled in the study. Approximately half received vitamin A supplements. One year after the start of the study, the number of children who had died was determined.

The summary measure chosen was the relative risk: the ratio formed by dividing the proportion of children who died in the vitamin A group by the proportion of children who died in the control group. For Aceh, the proportion who died in the vitamin A group was

or 7.7 per thousand. For the control group, the proportion who died was

or 10.6 per thousand. The relative risk is, therefore,

Relative risk less than 1 means that the vitamin A group has the lower mortality rate.

The relative risks for the eight studies were

A meta-analysis combined these eight results to produce a relative risk estimate of 0.77 with a 95% confidence interval of (0.68, 0.88). That is, vitamin A supplementation reduced the mortality rate to 77% of its value in an untreated group. The confidence interval does not include 1, so we can reject the null hypothesis of no effect (a relative risk of 1). The researchers examined many variations of this meta-analysis, such as using different weights and leaving out one study at a time. These variations had little effect on the final estimate.