Journal Logo

Medicine Correspondence Blog

The Medicine Correspondence Blog allows authors to post Letters to the Editors, Reviews, and other editorial writings that are not considered original research.

Friday, January 3, 2020

Letter to the Editor: Acupuncture and weight loss in Asians, a PRISMA-compliant systematic review and meta-analysis

We appreciated reading the work of Junpeng et al [1], "Acupuncture and weight loss in Asians, a PRISMA-compliant systematic review and meta-analysis", which is a meta-analysis for the association between acupuncture and weight loss. However, we found some notable data anomalies.

When estimating the weighted mean differences (e.g. Xie 2011 [2] in Figure 4 section 1.1.1), the authors assumed the correlation coefficient to be 0.4 across conditions, which is low for outcomes of body mass index (BMI), waist circumference (WC), total cholesterol (TC), and triglyceride (TG). For comparison, we used the change scores of other papers that were included in this meta-analysis and calculated the correlation coefficients to be mostly around 0.8.

We attempted to recalculate the mean differences and confidence intervals that were reported in Figures 4, 5, and 6 for the "acupuncture vs sham acupuncture" sections and encountered inaccurate/insufficient information from the publications to reproduce them. For the original paper of Hsu 2009 [3], the means of differences for the control group are -0.3 for BMI and -2.3 for WC, where in Junpeng's work [1], these were reported as 0.3 and 2.3 respectively.

In addition, the outcome of waist circumference from Yeo 2014 [4] appear to be missing from the meta-analysis, even though these statistics are reported in the original paper.

These all have affected the results of the meta-analyses to difference extents. We re-analyzed the data with the correlation coefficient 0.8 and including the previously missing WC data from Yeo 2014, and the recalculated results are presented in Table 1.

Table 1. Meta-analysis of Acupuncture vs. Sham Acupuncture
RecalculatedReported in paper
 Heterogeneity testI2Test of overall effectHeterogeneity testI2Test of overall effect
BMIChi2(3)=1.07, p=.7840.0%Z=7.74, p<.001Chi2(3)=0.34, p=.950.0%Z=7.79, p < .001
WCChi2(2)=17.22, p<.00188.4%Z=0.93, p=.351Chi2(1)=0.82, p=.370.0%Z=5.49, p < .001
TCChi2(1)=8.37, p=.00488.1%Z=0.31, p=.757Chi2(1)=7.17, p=.00786.0%Z=0.33, p = .74
TGChi2(1)=0.93, p=.3340.0%Z=0.08, p=.939Chi2(1)=0.51, p=.480.0%Z=0.26, p = .80

 

By changing the correlation coefficient from 0.4 to 0.8, the mean difference between groups for Hsu 2009 and Xie 2011 on BMI became 0.20, 95% CI [-2.12,2.52], and -0.30, 95% CI [-2.39, 1.79], respectively. The test of overall effect is still significant (p<.001)

For WC, the mean difference between groups for Hsu 2009 became 2.10, 95% CI [-0.79, 4.99]. After also including Yeo 2014 (-1.16, 95% CI [-2.09, -0.23]), the overall effect is then not significant (p=0.351 instead of p<.001; WMD: -1.28, 95% CI: [-3.96, 1.41] instead of WMD: -3.69, 95%CI: [-5.01,-2.37]). Inclusion of Yeo's paper [4] and correction of mean difference errors led to a dramatic difference on I-squared of waist circumference, which means there is heterogeneity among the three papers.

The mean difference between groups for Xie 2011 on TC became 0.16, 95% CI [0.06, 0.26]. The test of overall effect is still not significant (p=0.757).

For TG, the mean difference between groups for Xie 2011 became 0.05, 95% CI [-0.10, 0.20]. The test of overall effect is still not significant (p=0.939)

Although some of the final conclusions remain the same, others change and the analyses should be corrected.


Acknowledgments

This work is supported in part by the National Institute of Health, grant R25HL124208. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or any other organization.


Author Correspondence

Xiwei Chen, Department of Epidemiology and Biostatistics, School of Public Health, Indiana University, USA

Lilian Golzarri-Arroyo, Department of Epidemiology and Biostatistics, School of Public Health, Indiana University, USA

Stephanie L. Dickinson, Department of Epidemiology and Biostatistics, School of Public Health, Indiana University, USA

David B. Allison Department of Epidemiology and Biostatistics, School of Public Health, Indiana University, USA


References

[1] Junpeng Y, et al. Acupuncture and weight loss in Asians, A PRISMA-compliant systematic review and meta-analysis. Medicine 2019;98:33.

[2] Xie CF, Fu WS, Sun J, et al. Standardized plan for acupuncture treatment of simple obesity (Chinese). Chin J Gerontol 2011;31:4751–3.

[3] Hsu CH, Wang CJ, Hwang KC, et al. The effect of auricular acupuncture in obese women: a randomized controlled trial. J Women's Health (Larchmt) 2009;18:813–8.

[4] Yeo S, Kim KS, Lim S. Randomised clinical trial of five ear acupuncture points for the treatment of overweight people. Acupunct Med 2014;32:132–8. 

The authors of this article provided a response which can be found at this link.