Meta-Analysis for Linear and Nonlinear Dose-Response Relations: Examples, an Evaluation of Approximations, and Software

Abstract

Two methods for point and interval estimation of relative risk for log-linear exposure-response relations in meta-analyses of published ordinal categorical exposure-response data have been proposed. The authors compared the results of a meta-analysis of published data using each of the 2 methods with the results that would be obtained if the primary data were available and investigated the circumstances under which the approximations required for valid use of each meta-analytic method break down. They then extended the methods to handle nonlinear exposure-response relations. In the present article, methods are illustrated using studies of the relation between alcohol consumption and colorectal and lung cancer risks from the ongoing Pooling Project of Prospective Studies of Diet and Cancer. In these examples, the differences between the results of a meta-analysis of summarized published data and the pooled analysis of the individual original data were small. However, incorrectly assuming no correlation between relative risk estimates for exposure categories from the same study gave biased confidence intervals for the trend and biased P values for the tests for nonlinearity and between-study heterogeneity when there was strong confounding by other model covariates. The authors illustrate the use of 2 publicly available user-friendly programs (Stata and SAS) to implement meta-analysis for dose-response data.

Association between antidepressant medication use and epithelial ovarian cancer risk: a systematic review and meta‐analysis of observational studies

“…Furthermore, the study-specific OR was summarized for each 1-year-increment for the duration of antidepressant use. A study-specific trend reflecting correlated log OR values across different durations of drug use was computed using the generalized least-squares trend estimation method developed by Greenland et al [40] and Orsini et al [41]. Additionally, a potential nonlinear dose-response relationship between the duration of antidepressant use and EOC risk was modelled using restricted cubic splines with three knots at fixed percentiles (10%, 50% and 90%, respectively) of the distribution of exposure.…”

Dietary total fat and fatty acids intake, serum fatty acids and risk of breast cancer: A meta‐analysis of prospective cohort studies

“…If the upper boundary of the highest category was not provided, we assumed that the boundary had the same amplitude as the adjacent category. A two-stage random-effects dose-response meta-analysis 17 was performed to compute the trend from the correlated log RR estimates across levels of dietary total fat and fatty acids intake, respectively, taking into account the between-study heterogeneity. In the first stage, a restricted cubic spline model with three knots at the 25th, 50th and 75th percentiles of the levels of dietary total fat and fatty acids intake was estimated using generalized least square regression taking into account the correlation within each set of published RRs.…”

Sex‐Specific Associations Between Alcohol Consumption and Incidence of Hypertension: A Systematic Review and Meta‐Analysis of Cohort Studies

“…Using studies that reported data for 4 or more alcohol intake groups, we conducted 2‐stage restricted cubic spline regression in multivariate metaregression models taking into account the variance‐covariance matrix for risk estimates derived from 1 reference group40, 41 to calculate continuous nonlinear dose‐response curves for total alcohol consumption (g/d) in relation to abstainers. All meta‐analytical analyses were conducted on the natural log scale of the RRs (and hazard ratios) in Stata statistical software version 14.2 (Stata LP, College Station, TX).…”