Assessment of early exaggerated treatment effects in orthodontic interventions using cumulative meta-analysis.

BACKGROUND

The reported initial strong treatment effects reported in early trials that are refuted in subsequent future studies assessing the same interventions have been attributed to novelty bias. The aim of this study was to determine whether there is any evidence of novelty bias in the reported treatment effects of orthodontics interventions.

MATERIALS AND METHODS

Relevant orthodontic systematic review (SRs) topics containing at least one meta-analysis on either binary or continuous outcomes with a minimum of three trials considered important areas in the field of orthodontic practice were identified. SR, meta-analysis, and primary study-level characteristics were extracted. Descriptive statistics were calculated at the SRs, meta-analysis, and at the individual study level. All SR and trial-level data were imported into the statistical software and all meta-analyses were replicated using the cumulative random-effects meta-analysis approach. Changes in the size and direction of the estimates between the first trial and the cumulative effect over time were recorded.

RESULTS

Forty-seven meta-analyses were included. The total number of primary studies included within these meta-analyses was 408 (N = 408). Overall, the final effect size estimate decreased in 29 (61.7%, N = 29/47) cumulative meta-analyses whilst it increased in the remaining 18 (38.3%, N = 18/47). No association between the level of risk of bias and the cumulative absolute effect size was evident (OR 1.00; 95% CI: 0.98, 1.03; P = 0.717) after adjusting for year of the primary study (P = 0.22).

CONCLUSIONS

Clinicians should be wary of the results of trials reporting the effectiveness of new interventions as there is a possibility that the reported effect size will be often exaggerated.