Artificially defined transformation methods for obtaining dose points in dose-response meta-analysis should be more standardized: a cross-section study.

Article type
Authors
Lao Y1, Lu Z2, Yong R2, Wei Q2, Ma Y1, Wang Y2, Fan J1, He G1, Yao L3, Yang K4
1Second Clinical Medical College, Lanzhou University, Lanzhou
2First Clinical Medical College, Lanzhou University, Lanzhou
3Health Research Methodology, Department of Health Research Methods, Evidence and Impact, McMaster University
4Evidence-Based Medicine Center, School of Basic Medical Sciences, Lanzhou University, Lanzhou
Abstract
Background:
Dose-response meta-analyses (DRMAs) had been published increasingly over the past few years. The basic purpose of DRMAs is to reveal the relationship between disease risk and exposure dose. In order to meet the data condition for dose-response estimation, open-ended dose interval reported in original studies should be transformed to a certain dose point. And multiple units of exposure/intervention reported in original studies were also often unified to one. However, transformation methods artificially defined by DRMAs authors were not always the same which might cause misleading results.
Objectives:
To investigate how DRMAs authors defined transformation methods to obtain dose points for dose-response estimation.
Methods:
We searched PubMed for meta-analyses published in 2019 that explicitly combined dose-response estimates from multiple original studies and reported the results of dose-response meta-analyses. Paired authors selected eligible studies and extracted related information independently. Disagreements between two reviewers were resolved by discussing with a third reviewer.
Results:
247 DRMAs were finally included. The main outcomes were the risk of cancer and cardiovascular disease. As for open-ended interval, 32.8% studies reported methods to obtain the lower boundary of the lowest category. The two most common statements were: width of the open-ended interval was the same as adjacent category (58%) and the lowest category was assumed to zero (29.6%). Additionally, 4.9% articles assumed the lowest boundary to a certain nonzero dose based on specific clinical background. 7.5% studies reported another six different methods to obtain the lower boundary. Furthermore, 39.3% studies reported methods to obtain the higher boundary of the highest open-ended category. In this case, 71.2% studies assumed the width of the highest category based on different multiples of adjacent category: equality (65%), 1.5 times (5.2%), or half (1%). 27.8% studies assumed higher boundary of the highest category based on different multiples of the lower bound of the same interval: 1.2 times (13.4%), 1.5 times (9.3%), 1.25 times (2.1%), 1.4 times (1%), 2 times (1%), and equality (1%). Another 1% articles reported a certain nonzero dose based on the specific clinical background. As for unifying multiple units of reported in original studies, 14.6% studies reported transform methods, among which 47.2% intervention/exposure were diet related factors. In studies considering certain diet related factors, 77.3% studies involved transformation methods between nine different units and “gram” while 22.7% studies reported methods between “cup” and “ml”. But these methods are heterogeneous from each other. The overall results were presented in Figure 1.
Conclusions:
Artificial defined transformation methods were frequently used for obtaining dose points in DRMAs. But those methods were heterogenous which might cause misleading results and should be more standardized.
Patient or healthcare consumer involvement:
None.