Commentary: Gluteal muscle and subcutaneous tissue thicknesses in adults: a systematic review and meta-analysis

The authors of this paper endeavoured to answer two questions in relation to gluteal muscle and subcutaneous tissues in adults. Firstly, how thick are they, and, secondly, do the thicknesses differ by gender and body mass index (BMI)? The results would hopefully aid nurses and other healthcare professionals when administering intramuscular injections with ease, through the thinnest subcutaneous tissue, or needing to use longer needles. They did this by searching the available published literature and meta-analysing aggregate results from the included studies.

As a statistician, my interests lie in how these analyses were conducted. Were the statistical methods used appropriate? Were assumptions justified? And, were these analyses conducted properly?

From the six full-text articles that were included in the analysis, standardised mean differences were calculated, when necessary to do so, and were synthesised using a random-effects meta-analysis model. Random-effects models allow for differences in estimates from study to study and result in wider confidence intervals compared to fixed-effects models which assume all studies are estimating the same estimate. This is especially important when there is potential for high heterogeneity as they assume that study estimates (gluteal muscle thickness, for example) vary not only due to random sampling error but also due to fundamental differences in the way each study was designed and conducted. Obviously, this is more realistic in a practical sense than a fixed-effects model which assumes that the study design was similar throughout.

Common tests for heterogeneity were used; the Cochrane’s Q test and the I² statistic. An I² of over 50% was indicative of high heterogeneity, consistent with section 9.5.2 of the Cochrane Handbook where an I² of 50–90% indicates substantial heterogeneity. However, there is no hard-and-fast rule or definite cut-offs, and everything should be looked at within the context of the area of research.

Nevertheless, even borderline high-heterogeneity should be investigated further to understand what differences exist between studies – are they influenced by an important treatment effect modifier? In this study, the authors conducted a meta-regression model, a natural next-step after a random effects model, and adjusted for sex and BMI in this model. However, these results concluded that neither sex nor BMI was statistically significant in relation to muscle thickness, subcutaneous tissue thickness or total tissue thickness. Furthermore, the R² of these models ranged from 0.0 to 0.7%, meaning that a minute percentage of the variance of muscle or tissue thickness can be explained by sex or BMI. Clearly, there are other factors that should have been adjusted for in these models.

Moreover, the authors conducted multiple meta-regression models, investigating the effects of sex and BMI on the different outcomes. Multiplicity is an issue as testing several hypotheses within the same study simultaneously increases the type I rate – the probability of rejecting the null hypothesis of no difference. This, in turn, increases the risk in finding false-positive results. A few methods exist to combat this, such as the Bonferroni correction, which divides the alpha rate (5% in this study) by the number of models to get a new, smaller, significance level. In light of the multiplicity issue, the results of the meta-analyses should be interpreted with caution. As there were six meta-analysis models and forest plots presented, we would see the new significance level below 1%.

For example, in the abstract, the authors state ‘females had thicker subcutaneous tissue at the ventrogluteal site than males . . . p=0.0204’. Taking multiple testing into account, this result would not be statistically significant and the conclusion would change. Recommendations for further work include studies to be done on muscle and tissue thickness in other age groups of BMI categories. However, as other factors were not considered, it is possible that other important subgroups exist that need investigating further.

Biography

Mubarak Patel is a statistician currently working in health technology assessment at the University of Warwick. He is an editorial board member of the Journal of Research in Nursing as our Statistical Adviser.