The triangular test to assess the efficacy of metoclopramide in gastroesophageal reflux

Background: Sequential methods are particularly interesting when recruitment is difficult because they may allow a study to be stopped early while maintaining type I and II error rates. Methods: This placebo-controlled, randomized double-blind study was aimed at assessing the efficacy of metoclopramide (0.2 mg/kg three times daily during 14 days) on gastroesophageal reflux in infancy. The main end point was the relative variation of the percentage of time at pH <4 between inclusion (day 0) and evaluation (day 14) assessed on two 24-hour esophageal pH recordings. Statistical analysis was performed with use of a sequential method, the triangular test. Results: The study was stopped after the seventh analysis (39 infants evaluated: 20 placebo and 19 metoclopramide) without showing the expected benefit. Improvement on the main end point was 30% +/- 48% (mean +/- SD). Corresponding unbiased median estimates were 22% for placebo and 39% for metoclopramide (p = 0.28, sequential analysis). On day 14, the percentage of time at pH <4 was 8.1% +/- 11.7% for placebo and 6.7% +/- 9.2% for metoclopramide (p = 0.68, t test), and the number of reflux episodes >5 minutes was 3.0 +/- 3.5 for placebo and 1.9 +/- 3.0 for metoclopramide (p = 0.33, t test). Conclusion: If a tendency for a superior improvement with metoclopramide than with placebo was observed on the main end point, it was lower than expected and the difference was not significant. Compared with the corresponding single-stage design, the triangular test allowed to stop the study with a 15% reduction in sample size.