Sec.2-Ch.2-Subsec.3：Continuous Dose-Response Analysis and Prediction with R
A Practical Guide to Fixed-Effects Dose-Response Analysis Using the dosresmeta Package in Epidemiological Research
In epidemiology and clinical research, understanding the relationship between exposure dose and disease risk is a fundamental basis for developing public health strategies and clinical intervention plans. Traditional case–control or cohort studies often compare only the high-exposure group with the reference group, ignoring the continuous influence of different dose levels on disease risk. As a result, detailed quantification of risk changes is limited.
A typical fixed-effect dose–response analysis provides a methodological solution by integrating risk estimates at each dose level into a continuous dose–risk trend using regression models. This approach quantifies how much risk changes with each unit increase in dose. Under this framework, the dose–effect relationship within the study is assumed to be unique and fixed, without between-study heterogeneity. Thus, model parameters accurately reflect the average effect across dose groups within that study. This method fully utilizes information from all exposure categories, improves statistical efficiency and trend detection, and lays the foundation for subsequent dose–response meta-analysis.
I. Understanding Classical Fixed-Effect Dose–Response Analysis
Fixed-effect dose–response analysis and dose–response meta-analysis are methodologically related but differ in scope and purpose. Fixed-effect analysis typically applies to grouped data within a single study, fitting the relationship between continuous dose and outcome through regression modeling. It assumes that the internal dose–effect relationship is true and unique, with no heterogeneity.
For example, in a case–control study, log odds ratios (log OR) or log relative risks (log RR) can serve as dependent variables, while dose level serves as the independent variable. Linear or nonlinear trend models quantify how risk changes per unit increase in dose. Unlike simple high-versus-low comparisons, this method uses information across all exposure groups, improving statistical efficiency and trend detection.
