Metformin has shown its effectiveness in treating obesity in adults. However, little research has been conducted in children, with a lack of attention on pubertal status. The objectives were to determine whether oral metformin treatment reduces BMI z score, cardiovascular risk, and inflammation biomarkers in children who are obese depending on pubertal stage and sex.
This was a randomized, prospective, double-blind, placebo-controlled, multicenter trial, stratified according to pubertal stage and sex, conducted at 4 Spanish clinical hospitals. Eighty prepubertal and 80 pubertal nondiabetic children who were obese aged 7 to 14 years with a BMI >95th percentiles were recruited. The intervention included 1 g/d of metformin versus placebo for 6 months. The primary outcome was a reduction in BMI z score. Secondary outcomes comprised insulin resistance, cardiovascular risk, and inflammation biomarkers.
A total of 140 children completed the study (72 boys). Metformin decreased the BMI z score versus placebo in the prepubertal group (−0.8 and −0.6, respectively; difference, 0.2; P = .04). Significant increments were observed in prepubertal children treated with metformin versus placebo recipients in the quantitative insulin sensitivity check index (0.010 and −0.007; difference, 0.017; P = .01) and the adiponectin–leptin ratio (0.96 and 0.15; difference, 0.81; P = .01) and declines in interferon-γ (−5.6 and 0; difference, 5.6; P = .02) and total plasminogen activator inhibitor-1 (−1.7 and 2.4; difference, 4.1; P = .04). No serious adverse effects were reported.
Metformin decreased the BMI z score and improved inflammatory and cardiovascular-related obesity parameters in prepubertal children but not in pubertal children. Hence, the differential response according to puberty might be related to the dose of metformin per kilogram of weight. Further investigations are necessary.
We thank Dr. David B. Allison for his kind comments about the significance of our study and we agree with him about the importance of a correct interpretation of the article.
Firstly, we would like to clarify that the “differential response according to puberty” stated in the abstract does not imply to declare a significant differential response. We mean that the effect of metformin was noticed in prepubertal children, but not in pubertal children compared to placebo, clearly demonstrated using a general lineal model for repeated measures (GLM-RM) separately. Nevertheless, to avoid this statement from leading to confusion, we agree with changing the conclusion from “Metformin decreased the BMI z score and improved inflammatory and cardiovascular-related obesity parameters in prepubertal children but not in pubertal children. Hence, the differential response according to puberty might be related to the dose of metformin per kilogram of weight. Further investigations are necessary.” to “Metformin decreased the BMI z score and improved inflammatory and cardiovascular-related obesity parameters only in prepubertal children, but a differential effect of metformin was not observed in prepubertal compared to pubertal children. Nevertheless, the doses per kilogram of weight administrated may have had an impact on the metformin effect. Further investigations are necessary.”
This study presents two independent statistical analyses. The sample size was calculated to ensure two equal and matched groups (n=80 for each group) according to the pubertal condition. Hence, the initial analysis included the prepubertal and pubertal groups separately in the GLM-RM using an interaction term for time and treatment. This assumption was based on the physiological differences between both pubertal groups, which we consider very relevant in clinical trials with young subjects. The analyses revealed that the metformin effect was significantly higher than placebo in prepubertal children (P=0.04), but not in pubertal children (P=0.19).1 The different metformin effect compared to the placebo group is not a nominal difference, as we did not use a within-group approach. We think it is licit to perform a separate analysis. By stating that “None of the previous studies used a study design to allow observing a differential response by pubertal status separately”,1 we intend to highlight that we focused the analysis by clearly separating both groups. Actually, the doses per body-weight were statistically lower in pubertal children (P<0.001), which could explain the no effect observed. Hence, we suggested that higher doses for subjects aged 10-16 years2 might improve the design in further trials. Nevertheless, given the lack of studies performed in children based on puberty status, we conducted a second a posteriori analysis including the entire cohort. For this analysis, GLM-RM was used including an interaction term for time × treatment × puberty. This analysis demonstrated no different effect of metformin according to pubertal stage (P=0.41).1 Indeed, this interaction would reflect that there are no significant differences in the effect of metformin based on pubertal stage.
Secondly, we would like to clarify that we did not consider placebo-corrected reductions in Table 2, which shows BMI z-score compared to placebo from baseline to six months after intervention. We believe that it is not appropriate to only observe the difference of the net reduction of 0.2 points between placebo and metformin in both pre- and pubertal groups when a complex model is used, since such models correct means according to the fixed effects included, taking into account several factors that influence the results.
References:
1. Pastor-Villaescusa B, Cañete MD, Caballero-Villarraso J, et al. Metformin for obesity in prepubertal and pubertal children: A randomized controlled trial. Pediatrics. 2017;140(1):e20164285
2. Brufani C, Fintini D, Nobili V, Patera PI, Cappa M, Brufani M. Use of metformin in pediatric age. Pediatr Diabetes. 2011;12(6):580–588
To the Editors,
The article by Pastor-Villaescusa may make a valuable contribution to the literature if interpreted correctly. However, in its current form it contains an important error of statistical inference.
Specifically, the abstract states: “Metformin decreased the BMI z score and improved inflammatory and cardiovascular-related obesity parameters in prepubertal children but not in pubertal children. Hence, the differential response according to puberty might be related to the dose of metformin per kilogram of weight.”
Yet, in contrast to the statement above, the body of the paper states: “No differences were found in the impact of metformin according to the pubertal stage when the interaction time × treatment × puberty was applied to the entire population (P = .41).” Furthermore, Table 2 shows that the placebo-corrected reductions in BMI z score in the prepubertal and pubertal groups are, within the precision reported in the table, identical.
In short, there is no evidence for a "differential response" by pubertal status, contradicting the authors’ conclusion statement. The authors have made a common but serious error of inference by neglecting that ‘Difference in Nominal Significance is not a Significant Difference’ as described elsewhere (1,2).
The conclusion offered by the authors is not supported by the data and should be corrected.
Sincerely,
David B. Allison, PhD
University of Alabama at Birmingham
References:
1. George, B. J., Beasley, T. M., Brown, A. W., Dawson, J., Dimova, R., Divers, J., Goldsby, T. U., Heo, M., Kaiser, K. A., Keith, S. W., Kim, M. Y., Li, P., Mehta, T., Oakes, J. M., Skinner, A., Stuart, E. and Allison, D. B. (2016), Common scientific and statistical errors in obesity research. Obesity, 24: 781–790. doi:10.1002/oby.21449.
2. Gelman, A., & Stern, H. (2006). The Difference between "Significant" and "Not Significant" Is Not Itself Statistically. The American Statistician, 60: 328-331. http://www.jstor.org/stable/27643811