We read with interest the meta-analysis by Wang et al. 1 published in a recent issue of BMJ. However, we found some problems in their meta-analysis that might have influenced their results.
First, their search items were insufficient. The keywords such as “cancer”,” tumor”, “cohort study”, “longitudinal study” also should be used.
Second, we found that the authors omitted many eligible studies that met their inclusion criteria. Taking the association of fruit consumption with health outcomes for example, we identified 21 studies for total mortality, 20 studies for cardiovascular mortality, and 25 studies for cancer mortality based on their search strategy. Surprisingly, the authors only identified 7, 6 and 7, for total, cardiovascular and cancer mortality, accordingly.
Third, there was significant between-study heterogeneity for associations with majority of health outcomes. The authors used subgroup analysis to examine the source of heterogeneity. In our opinion, it was not sufficient. Meta regression analysis should be performed firstly 2 and then the subgroup analysis.
Fourth, there was significant publication bias for the associations of total mortality with fruit and vegetable combined and fruit alone (p=0.006 and 0.02, respectively). The authors used trim and fill method to address the publication bias. In our opinion, this method may partially address the publication bias in statistical aspect, but it cannot overcome the inherent publication bias. Thus, the protective effect of fruit and vegetable combined or fruit alone on risk of total mortality should be interpreted with caution.
1. Wang X, Ouyang Y, Liu J, Zhu M, Zhao G, Bao W, Hu FB. Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies. BMJ 2014;349:g4490.
2. Martin B. Meta-analysis: Dealing with heterogeneity. York: University of York; 2006.
Competing interests: No competing interests