Can a meta-analysis that included articles with high heterogeneity appropriately evaluate the risk of venous thromboembolism in users of combined hormonal contraceptives containing drospirenone or desogestrel?

Abstract

IR: We congratulate Mart í nez et al. 1 on their interesting systematic review. However, we found that part of the meta-analysis they performed was compromised by the high heterogeneity which affected certain studies. This issue may have led to biases in their inferences. For example, the forest plot (Figure 8 in the paper by Mart í nez et al . 1 ) depicted four studies which compared users of combined oral contraceptives (COCs) that contained either drospirenone or levonorgestrel as a progestin 2 – 5 ; two of these investigations were cohort studies 4,5 , whereas the other two were case-control studies nested in a cohort 2,3 . The meta-analysis demonstrated high heterogeneity and inconsistency among these four studies: (Chi-squared [ χ 2 ] p 0.0008 and Inconsistency [I 2 ] 82%). In the chi-squared test, a low p -value (or a large chi-squared statistic relative to the degree of freedom) provides evidence of the heterogeneity of intervention effects (that is, a variation in effect estimates beyond the variation that would be predicted by chance alone). Measures of heterogeneity are essential in metaanalyses of clinical trials. The chi-squared test has a low discriminatory power if the number of included studies is small; thus, in these situations, a chi-squared test does not necessarily exclude the presence of signifi cant heterogeneity. Recently, a metric known as inconsistency (I 2 ) has been proposed which measures the portion of the differences between studies that is due to heterogeneity. This index is often expressed as a percentage: (i) an inconsistency of 0% to 40% may be relatively unimportant; (ii) an inconsistency of 30% to 60% may indicate moderate heterogeneity; (iii) an inconsistency of 50% to 90% may imply substantial heterogeneity; and (iv) an inconsistency of 75% to 100% may denote considerable heterogeneity 6 . The European Journal of Contraception and Reproductive Health Care, 2013; 18: 137–142

DOI: 10.3109/13625187.2013.764406

Cite this paper

@article{Rosa2013CanAM, title={Can a meta-analysis that included articles with high heterogeneity appropriately evaluate the risk of venous thromboembolism in users of combined hormonal contraceptives containing drospirenone or desogestrel?}, author={Maria In{\^e}s da Rosa and Kristian Madeira and Lidia Rosi de Freitas Medeiros}, journal={The European journal of contraception & reproductive health care : the official journal of the European Society of Contraception}, year={2013}, volume={18 2}, pages={137-9} }