Sirs, The recent article by Labenz et al. on the EXPO study is intriguing,1 but it should be objectively considered within the context of recently published peer review trials, such as Gillessen et al., which is appropriately powered to show equivalence between pantoprazole and esomeprazole for endoscopically confirmed healing,2, 3 and within an integrated therapeutic perspective that values the prompt resolution of a broad spectrum of gastro-oesophageal reflux disease (GERD)-related symptoms. Recognizing GERD as a heterogeneous symptom complex, rather than considering only heartburn, is of increasing relevance in reflux studies.4 This approach shows that pantoprazole is at least as effective as esomeprazole for symptom relief.5 In addition to these issues, the greatest problem with the EXPO study is its flawed statistical analysis. The EXPO study was powered to detect a clinically relevant treatment difference of 5%, assuming 8-week healing rates of 88% and 83% for esomeprazole and pantoprazole. However, data show a treatment difference of only 3.5% (survival method) which, by the authors’ own definition, is not clinically significant. Indeed, re-calculation of the EXPO data show cumulative healing rates after 8 weeks’ treatment of 91.6% for esomeprazole and 88.9% for pantoprazole, a difference of 2.7% [Blackwelder 95% confidence interval (CI) 0.6–4.8]. This suggests that esomeprazole is equivalent to pantoprazole, as even the upper limit of the CI (4.8%) is less than the 5.0% necessary to show clinically relevant superiority. Furthermore, when Labenz et al. display the breakdown of healing rates by baseline Los Angeles (LA) grade of erosive oesophagitis severity, the 95% CIs overlap, indicating a non-significant difference between the two treatment groups e.g., LA grade C healing rates: esomeprazole 40 mg 91.3% (95% CI 88.0–94.6), pantoprazole 40 mg 87.6% (95% CI 83.8–91.5). Nonetheless, Labenz et al. were able to show a significant difference in healing rates between treatment groups in patients with LA grade C using the log-rank test, which does not test the difference between healing rates directly but compares the difference of the entire survival curves from beginning to end of treatment. In addition, in the EXPO study, more patients with mild GERD (LA grade A) were included in the esomeprazole group than the pantoprazole group (33.5% vs. 30.1%). This difference of 3.4% reaches statistical significance (Fischer's exact test, P = 0.046) and is similar in magnitude to the difference in healing rates reported between the two treatment groups. Further, if the data for healing rates for the different LA grades found with esomeprazole are transferred to the distribution of LA grades reported for pantoprazole, an overall healing rate of 92.2% can be calculated for esomeprazole compared with the reported 95.5%, which is significantly different (P < 0.00001) i.e., esomeprazole is superior to esomeprazole. The same can be shown vice versa for pantoprazole, which clearly indicates that significant results smaller than the inhomogeneity at baseline can be produced without any difference in treatment. In conclusion, taken as a whole, data from peer-reviewed trials, including re-analysis of the EXPO study, show that pantoprazole and esomeprazole are equivalent for oesophageal healing and symptom relief.
Read full abstract