Sec.2-Ch.2-Subsec.2：Dose–Response Meta-analysis in R
A Practical Guide to Understanding DRMA and Using the dosresmeta Package for Linear and Nonlinear Dose–Risk Modeling in Medical Research
Dose–response meta-analysis (DRMA) is an advanced statistical method used to quantitatively evaluate the relationship between exposure dose and outcome risk. Compared with traditional meta-analyses of binary or continuous data, DRMA can handle data from three or more exposure levels, allowing direct estimation of the relationship between dose and disease risk. By synthesizing multiple original dose–response studies, researchers can obtain a unified effect estimate and construct linear or nonlinear curves, providing more precise and reliable evidence for scientific decision-making.
I. Understanding Dose–Response Meta-analysis
Meta-analysis (MA) is a class of statistical methods for quantitatively synthesizing multiple study results, dating back to the 1970s and 1980s. Initially, meta-analysis was defined as “a quantitative analytical method for synthesizing multiple study results,” a concept proposed by L. V. Hedges in 1985. In 1991, Fleiss provided a more rigorous and precise definition, considering meta-analysis as a class of statistical methods for comparing and synthesizing research results addressing the same scientific question, with the validity of conclusions depending on whether included studies met certain conditions. Over time, institutions and scholars such as the Cochrane Collaboration, the U.S. National Library of Medicine (NLM), the Himmelfarb Health Sciences Library, and Gene Glass have proposed their own definitions. Thus, meta-analysis is fundamentally a statistical method that can combine different study results to obtain more precise and statistically powerful conclusions. However, as the theoretical framework and methods of meta-analysis have continuously improved, new meta-analysis methods have emerged, among which dose-response meta-analysis (DRMA) is a typical representative.
Dose-response meta-analysis has become an indispensable method in modern epidemiology and clinical research primarily because it can quantitatively describe the relationship between exposure levels and disease risk, compensating for the limitations of traditional binary or simple continuous variable meta-analyses. Traditional meta-analyses typically only compare effect differences between exposed and non-exposed groups, unable to reveal the specific impact of dose changes on risk, and difficult to answer the core question of “how much risk change will increasing or decreasing exposure bring.” In real-world research, many exposure factors (such as smoking amount, drug dosage, nutrient intake, etc.) affect health outcomes not simply by presence or absence, but rather change linearly or nonlinearly with increasing dose. Therefore, relying solely on binary analysis tends to underestimate or overestimate the true impact of exposure, or even miss critical dose thresholds or saturation points. Through dose-response meta-analysis, researchers can integrate data from different dose groups across multiple studies, quantitatively estimate risk at each dose level, and plot dose-response curves through linear or nonlinear models, thereby revealing potential dose-dependent effects. This analytical method not only improves statistical power, enabling comprehensive utilization of results from small-sample studies, but can also explore nonlinear relationships and dose thresholds, providing scientific evidence for developing public health guidelines, drug dosage regimens, and clinical intervention strategies. Furthermore, dose-response meta-analysis can identify heterogeneity between studies, assess the impact of confounding factors, and improve the robustness of conclusions through sensitivity and subgroup analyses, thereby playing an important role in evidence-based medicine and policy-making.
