Do all individuals benefit equally from non-surgical periodontal therapy? Secondary analyses of systematic review data

Abstract

Aims: This study aimed to assess the variability and treatment effect heterogeneity in response to non-surgical periodontal therapy (NSPT).

Methods: Data from randomized controlled trials included in two recent systematic reviews on the effect of NSPT on mean clinical attachment loss (CAL), mean probing pocket depth (PPD), percentage of sites with bleeding on probing (%BOP), PPD ≤3 mm (%PD ≤3 mm), and C-reactive protein levels (CRP) at 3-12-month follow-up among adults with systemic diseases or conditions were used. In these trials, the control arms received no treatment, hygiene advice, or supragingival scaling. The Bayesian meta-regression models were utilized to assess the variability ratios between NSPT and control groups.

Results: Data from 36 trials on mean PPD, 32 trials on mean CAL, eight trials on %PD ≤3 mm, 31 trials on %BOP and 19 trials on CRP were used. Variability in mean CAL and CRP was approximately 10% higher in the NSPT arms than in the control arms, hinting that there may be room for treatment effect heterogeneity. Instead, variability in mean PPD, %BOP, and %PD ≤3 mm was lower in the NSPT arms than in the control arms.

Conclusion: Potential treatment effect heterogeneity in response to NSPT was observed for CRP and mean CAL. However, substantial measurement error in CAL and natural variation in CRP may contribute to these findings. Conversely, treatment effect heterogeneity appears less pronounced for mean PPD, %BOP, and %PD ≤3 mm, potentially due to greater treatment effects in patients with more severe periodontitis and reduced measurement error in these parameters.

Keywords: Bayesian; heterogeneity; non‐surgical periodontal treatment; periodontal diseases; periodontal therapy; periodontitis.

FIGURE 1

Variability in continuous outcome in control and intervention arms at the end of follow‐up in six hypothetical randomized controlled trials involving eight participants. Adapted from Cortes et al. Lines between the arms indicate the non‐observable effect for each individual. Densities are the potential distributions of the outcome in each arm. A represents a situation where there is no treatment effect and outcome variability does not change; B represents a situation where treatment effect is constant and outcome variability does not change; C represents a situation where there is treatment‐by‐subgroup interaction, and treatment effect is constant within the two subgroups, outcome variability increases by treatment; D represents a situation where treatment effect varies by each patient and outcome variability increases by treatment; E represents a situation where there is treatment‐by‐subgroup interaction, and treatment effect is constant within the two subgroups, however, outcome variability decreases by treatment; and F represents a situation where treatment effect varies by each patient, and outcome variability decreases by treatment.

FIGURE 2

Estimates of the standard deviation of individual treatment effects for the five outcomes assuming different strenghts of correlation between the endpoint outcome value under control intervention and the treatment effect. Lower and upper lines represent 2.5th and 97.5th percentiles, respectively.

FIGURE 3

Distribution of individual treatment effects for the five outcomes. Vertical lines represent the estimated average treatment effect. The average estimated standard deviation of individual treatment effects was calculated assuming zero correlation between the endpoint outcome value under the control intervention and the treatment effect.