Reporting and interpretation of subgroup analyses in orthodontic meta-analysis; a meta-epidemiological study

In systematic reviews (SRs) with a meta-analysis, a subgroup analysis can be utilised to explore heterogeneity of treatment effects across patient characteristics and assess how patient characteristics, types of intervention, or trial characteristics modify the treatment effect. The aim of this meta-epidemiological study was to investigate the reporting and interpretation of subgroup analyses in the orthodontic literature.

SRs published between 1st January 2017 - 30th June 2024 in five orthodontic journals and the Cochrane Database of Systematic Reviews (CDSR) were hand searched from journal websites. We assessed the presence of testing for subgroup differences (interaction test) and the interpretation and reporting of subgroup analysis limitations were also reviewed. Descriptive statistics of the SR characteristics and univariable cross-tabulation to detect associations between journal type and variables affecting interpretation of subgroup analyses were undertaken. On an exploratory basis, exact logistic regression was used to examine the effect of publication year on the correct interpretation of subgroup results.

Five thousand one-hundred and one (5101) articles were screened, 298 SRs were identified of which 70 met the inclusion criteria. In the subgroup analysis, interpretation of the p-value was undertaken in just over half of cases (n = 40, 57.1%). Only 32.9% (n = 23) of subgroup analyses were deemed correctly interpretated. Most commonly, no subgroup limitations were reported by the authors (n = 48, 68.6%) even when such limitations existed. There was no association between year of publication and correct interpretation of subgroup analyses (OR:0.93, 95% CI: 0.70, 1.22, p = 0.62).

There appears to be a lack of awareness of how to correctly interpret subgroup analyses from forest plots. Only a third of orthodontic subgroup analyses assessed in this study were correctly interpreted. Misinterpretation of the effect of covariates (e.g. patient characteristics) on the effect of a treatment intervention could potentially have a negative impact on patient healthcare decisions. Recommendations to improve the interpretation of subgroup analyses are proposed.