EXAMPLE 4 Dropouts in a medical study

Orlistat is a drug that may help reduce obesity by preventing absorption of fat from the foods we eat. As usual, the drug was compared with a placebo in a double-blind randomized trial. Here’s what happened.

The subjects were 1187 obese subjects. They were given a placebo for four weeks, and the subjects who wouldn’t take a pill regularly were dropped. This addressed the problem of nonadherers. There were 892 subjects left. These subjects were randomly assigned to Orlistat or a placebo, along with a weight-loss diet. After a year devoted to losing weight, 576 subjects were still participating. On the average, the Orlistat group lost 3.15 kilograms (about 7 pounds) more than the placebo group. The study continued for another year, now emphasizing maintaining the weight loss from the first year. At the end of the second year, 403 subjects were left. That’s only 45% of the 892 who were randomized. Orlistat again beat the placebo, reducing the weight regained by an average of 2.25 kilograms (about 5 pounds).

Can we trust the results when so many subjects dropped out? The overall dropout rates were similar in the two groups: 54% of the subjects taking Orlistat and 57% of those in the placebo group dropped out. Were dropouts related to the treatments? Placebo subjects in weight-loss experiments often drop out because they aren’t losing weight. This would bias the study against Orlistat because the subjects in the placebo group at the end may be those who could lose weight just by following a diet. The researchers looked carefully at the data available for subjects who dropped out. Dropouts from both groups had lost less weight than those who stayed, but careful statistical study suggested that there was little bias. Perhaps so, but the results aren’t as clean as our first look at experiments promised.