Comment on 'Association between serum osteocalcin and body mass index: a systematic review and meta-analysis.'
- PDF / 219,386 Bytes
- 2 Pages / 595.276 x 790.866 pts Page_size
- 19 Downloads / 208 Views
LETTER TO THE EDITOR
Comment on 'Association between serum osteocalcin and body mass index: a systematic review and meta-analysis.' Xiaoying Liu1 Kaye E. Brock1 Tara C. Brennan-Speranza1 ●
●
Received: 20 September 2017 / Accepted: 14 October 2017 © Springer Science+Business Media, LLC 2017
Dear Editor, We read with interest, the recently published article by Kord-Varkaneh et al. [1] reporting a significant inverse association between total serum osteocalcin and body mass index (BMI) in adult healthy populations. We have noticed a number of methodological issues that limit the validity and question the generalisability of the authors’ conclusions. Our first concern is the inclusion of multiple publications from within the same study populations (n = 5 studies, n = 2 cohorts). Particularly we note that three papers from China [2–4] accessed data from the same large population base and time frame. Each of these three papers reported relatively higher correlations between BMI and osteocalcin than other studies in the authors’ meta-analysis (ranging from r = −0.292, n = 2043, wt% = 11 [2] to, r = −0.260, n = 2400, wt% = 13 [3], to r = −0.258, n = 2344, wt% = 12 [4]). This means that these three papers contributed 36% of data to the pooled correlation between BMI and osteocalcin in the authors’ meta-analysis. In addition, two smaller studies from Korea [5, 6] drew on the same population source. The usual approach in metaanalysis is to select a single publication with the largest sample size within each individual cohort. When we reanalysed the authors’ meta-analysis [1] including only one publication with the largest sample size from each cohort (n = 25 studies, n = 14,438 participants), the estimate of effect between BMI and total osteocalcin was r = −0.148; 95% confidence interval (CI) −0.182 to −0.113. This
* Xiaoying Liu [email protected] 1
Department of Physiology, University of Sydney, Sydney, NSW, Australia
correlation is substantially weaker and falls outside of the authors’ 95% CI r = −0.190, 95% CI −0.200 to −0.170 from Fig. 2 [1]. For this reason, we do not think the reported correlation is likely to represent the true pooled correlation coefficient. Furthermore, the authors’ inclusion of such duplicate studies is likely to have affected the reliability of all the subsequent conclusions throughout the paper. Our other concern was the authors’ lack of detailed definition of their sub-groups (healthy, postmenopausal and metabolic syndrome (MetS)). The sub-groups were often only a proportion of the total population. For example, of the five papers labelled as metabolic syndrome, one was a duplicate [4], the others had varying definitions of MetS and also different proportions of their population with MetS which varied from 68% [7] to 55% [8] to 39% [6] and to 9% [3] ranging from typical to very low compared to the general population of these ages [9]. Labelling these cohorts as MetS may mislead the reader and the resultant correlation coefficient of the authors’ meta-analysis is unlikely to represent the
Data Loading...
