Letter to the Editor: “Incidence of Hypoglycemia After Gastric Bypass vs Sleeve Gastrectomy: A Randomized Trial” Free

We read with interest the article by Capristo et al. (1) that compared the incidence of hypoglycemia after either Roux-en-Y gastric bypass (RYGB) or sleeve gastrectomy (SG). This important study advances our understanding of both the incidence of hypoglycemia after these surgical procedures and the possible underlying mechanisms. The authors concluded that, compared with RYGB, reactive hypoglycemia was not less common after SG, and therefore SG is not a safer option. This conclusion is based on the lack of a statistically significant difference in the incidence of hypoglycemia during a 3-hour oral glucose tolerance test performed 1 year after surgery (the primary outcome). However, the P value for this comparison as reported in the text (P = 0.079) differs from the P value reported in Table 1 (P = 0.047), which is below the 0.05 threshold for statistical significance.

To confirm these findings, we reanalyzed the published data. The authors state that “Associations between treatments and presence/absence of hypoglycemic events were analyzed by χ² or Fisher exact tests.” When hypoglycemia incidence after RYBG and SG is compared with the χ² test, the P value is 0.047, whereas a Fisher exact test yields a P value of 0.079. It would helpful to readers if the authors were specific about what statistical method was used to test the primary outcome and whether the method was prespecified. A previously published article on the protocol stated that Fisher exact test would be used (2). Potential for discrepancy between the χ² and Fisher exact tests has been recognized, as has the tendency for the χ² test to produce smaller P values than Fisher exact test (3). The sample size and event rate are large enough that use of the χ² test would be defensible, and the P < 0.05 calls into question the conclusion that hypoglycemia was not less common after SG.

In addition to a lack of clarity about the statistical significance of the difference in the primary outcome, the study may be underpowered to detect a difference. The sample size target was calculated to detect a 30% difference between treatment groups, assuming hypoglycemia incidence of 50% in the RYBG group and 20% in the SG group. However, the observed incidences were 29% and 14%, which is only a 15% difference. Based on the observed hypoglycemia rates, the sample size needed to achieve a power of 0.9 with two-tailed P < 0.05 is 153 subjects per group, more than twice the number of patients in the study. Therefore, absent a statistically significant difference as indicated by the prespecified Fisher exact test, the study was underpowered to detect a difference in hypoglycemia incidence between groups.

To summarize, either there is a statistically significant lower incidence of hypoglycemia after SG than RYBG, or there is not a significant difference and the study was underpowered. We believe the strongly worded conclusion that hypoglycemia was not less common after SG should be moderated to take these considerations into account.

Acknowledgments

Financial Support: D.J.R. was supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health under Award Number K23DK102963. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The funders had no role in the study design, data, writing the report, or submission for publication.

Disclosure Summary: The authors have nothing to disclose.

References