A Serious Flaw in Nutrition Epidemiology: A Meta-Analysis Study

Document Type

Article

Publication Date

11-22-2018

Publication Title

The International Journal of Biostatistics

DOI

10.1515/ijb-2018-0079

ISSN

1557-4679

Abstract

Background

Many researchers have studied the relationship between diet and health. Specifically, there are papers showing an association between the consumption of sugar sweetened beverages and Type 2 diabetes. Many meta-analyses use individual studies that do not attempt to adjust for multiple testing or multiple modeling. Hence the claims reported in a meta-analysis paper may be unreliable as the base papers do not ensure unbiased statistics.

Objective

Determine (i) the statistical reliability of 10 papers and (ii) indirectly the reliability of the meta-analysis study.

Method

We obtained copies of each of the 10 papers used in a metaanalysis paper and counted the numbers of outcomes, predictors, and covariates. We estimate the size of the potential analysis search space available to the authors of these papers; i. e. the number of comparisons and models available. The potential analysis search space is the number of outcomes times the number of predictors times 2 c , where c is the number of covariates. This formula was applied to information found in the abstracts (Space A) as well as the text (Space T) of each base paper.

Results

The median and range of the number of comparisons possible across the base papers are 6.5 and (2 12,288), respectively for Space A, and 196,608 and (3072–117,117,952), respectively for Space T. It is noted that the median of 6.5 for Space A may be misleading as each study has 60–165 foods that could be predictors.

Conclusion

Given that testing is at the 5% level and the number of comparisons is very large, nominal statistical significance is very weak support for a claim. The claims in these papers are not statistically supported and hence are unreliable so the meta-analysis paper is also unreliable.

Link to Full Text

Share

COinS