NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Moorthy D, Chung M, Lee J, et al. Concordance Between the Findings of Epidemiological Studies and Randomized Trials in Nutrition: An Empirical Evaluation and Citation Analysis: Nutritional Research Series, Vol. 6 [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2013 May. (Technical Reviews, No. 17.6.)
Concordance Between the Findings of Epidemiological Studies and Randomized Trials in Nutrition: An Empirical Evaluation and Citation Analysis: Nutritional Research Series, Vol. 6 [Internet].
Show detailsDatabase of Nutrient-Outcome Associations
We identified a total of 34 eligible associations between nutrients and health outcomes (Table 1). Sixteen of 34 were based on meta-analyses of epidemiological studies and of RCTs that were reported in the same paper,12,19–27 11 were based on “matched” meta-analyses that were reported in separate papers,26,28–36 3 on a meta-analysis of observational studies matched with a large RCT (>1,000 participants),35,37–39 and 4 on a meta-analysis of RCTs matched with a large epidemiological study (>5,000 participants).40–45 Some of the included papers reported results from a meta-analyses of more than one nutrients and health outcome pair.
In Table 1, 20 out of 34 associations pertained to vitamins, 7 pertained to minerals (such as calcium) and trace elements (such as selenium), and 7 to either macronutrients (fiber) or fatty acids. The examined endpoints span a wide range of clinical outcomes including mortality, stroke, and other cardiovascular outcomes, various cancers, acute macular degeneration, and fractures, and surrogate outcomes such as systolic and diastolic blood pressure. In two cases nutrient intake of the mother was examined with relation to a clinical outcome in the child (prenatal multivitamin supplementation in the mother and risk of neural tube defects in the fetus, and maternal calcium intake and blood pressure in the offspring).
Concordance or Discordance Between Epidemiological and RCT Data
Table 2 shows concordance between the summary effects in the two research designs using the three definitions. Using the first definition, 6 out of 34 topics (18%) were qualitatively concordant, and the remaining 28 were classified as “unclear” (there was no topic with qualitatively discordant epidemiological and RCT data). Twelve of 34 associations (35%) had a statistically significant z-score, and using the second definition, evidence from epidemiological studies and evidence from RCTs was statistically significantly discordant. The remaining 22 associations had no evidence of significant discordance. Of the 12 association with a statistically significant z-score, in 6 the meta-analysis of the RCTs and that of the epidemiological studies pointed to different directions. Thus, six examples were quantitatively discordant using the third definition.
Citation Graphs
We were able to construct citation graphs for 28 topics (Table 3). For 6 topics we were unable to form reliable citation graphs (for technical reasons related to changes in the format of the citation information obtained from the ISI Web site). Briefly, there was variability in the total number of vertices in the 28 citation graphs: The median number of vertices (articles) was 253 (25th–75th percentiles: 95, 356), and the median number of edges (citation relationships) was 1181 (25th–75th percentiles: 255, 1620). The citation graphs are relatively sparsely connected with median density 0.018 (low density or low connectivity is the norm for citation graphs). The table shows also the mean hub and authority scores across all papers in a graph, and across the 20 papers with the highest respective scores. Graphs with higher mean hub scores and higher mean authority scores have more papers that “integrate” and “provide” information (as conveyed by citaiton relationships), respectively, compared with graphs with lower scores. When all papers in a graph were considered, the median hub score was 0.09 (25th–75th percentiles: 0.07, 0.13) and the median authority score was 0.13 (25th–75th percentiles: 0.10, 0.21). Finally, similar variability across the 28 topics was observed in the mean number of citation received, citations made, or total citations made or received. When all papers in a graph were considered the median numbers and 25th–75th percentile ranges were 4.3 (3.2, 4.7), 4.3 (3.2, 4.7) and 8.5 (6.3, 9.4), respectively. (When all papers are considered in a citation graph, the mean number of citations made equals the mean number of citations received.) The corresponding mean scores or numbers over the 20 papers with the largest respective scores or numbers are also shown (Table 3).
Associations Between Citation Graph Metrics and Qualitative and Quantitative Concordance
There was no association between the characteristics of the citation graphs and the observed qualitative or quantitative concordance in the 28 analyzable topics (Table 4). The table shows analyses in which the predictors (numbers or mean scores) were log transformed using the base 2. Thus, the odds ratios in the table represent the change in the odds of finding concordance per doubling of the value of the predictor. The p-values remain identical for any other logarithmic transformation of the predictors. Analyses using other transformations (square root, no transformation) were qualitatively similar. It appears that the examined connectivity metrics of the citations maps do not correlate with the likelihood that the summary findings from meta-analyses of epidemiological studies and of RCTs are in qualitative or quantitative agreement.
- Results - Concordance Between the Findings of Epidemiological Studies and Random...Results - Concordance Between the Findings of Epidemiological Studies and Randomized Trials in Nutrition: An Empirical Evaluation and Citation Analysis
- Very-low-carbohydrate ketogenic diet v low-fat diet for long-term weight loss: a...Very-low-carbohydrate ketogenic diet v low-fat diet for long-term weight loss: a meta-analysis of randomised controlled trials - Database of Abstracts of Reviews of Effects (DARE): Quality-assessed Reviews
- GeneReviews Advanced Search Help - GeneReviews®GeneReviews Advanced Search Help - GeneReviews®
- Discussion - Systematic Review of Head Cooling in Adults After Traumatic Brain I...Discussion - Systematic Review of Head Cooling in Adults After Traumatic Brain Injury and Stroke
- Assessing the Accuracy of Google Translate to Allow Data Extraction From Trials ...Assessing the Accuracy of Google Translate to Allow Data Extraction From Trials Published in Non-English Languages
Your browsing activity is empty.
Activity recording is turned off.
See more...