Clinical Evidence Dashboard

Dexmedetomidine Dosing and Timing in Postoperative Delirium Prevention

Designed for rapid clinical interpretation of how dexmedetomidine was administered across eligible trials.

Included comparisons: Dexmedetomidine versus placebo or saline

Outcome considered: Delirium incidence

Source review: Effectiveness of drug interventions to prevent delirium after surgery for older adults: systematic review and network meta-analysis of randomised controlled trials.
Luney et al., 2026. DOI: 10.1136/bmj-2025-085539 (BMJ article page)

- Filtered Trials

- Pre-operatively

- Intra-operatively

- Post-operatively

- Participants

How to Read This Page

The charts display number of trials. Use timing filters to isolate when dexmedetomidine was given: Pre-operatively, Intra-operatively, or Post-operatively.

Filter Trials

You can select multiple categories in each filter.

Infusion Dose Bands

Timing of Dexmedetomidine Administration

Bayesian Meta-Analysis

Dots depict the median Observed OR. Purple curves depict study-specific shrinkage marginal posterior distributions. Vertical lines show the pooled median and 95% CrI limits. Credible intervals (CrI) are reported as highest-density intervals.

This panel reuses one Bayesian model fitted on the full eligible dataset. Filters update which trial rows are displayed in the forest plot, while the pooled effect remains the same full-dataset estimate.

Explanation about the Bayesian model

We adapted Model 4 from Jackson et al.: A comparison of seven random-effects models for meta-analyses that estimate the summary odds ratio (DOI: 10.1002/sim.7588).

Model structure:

\[ \begin{aligned} y_{ij}\sim \mathrm{Binomial}(n_{ij}, p_{ij}), \\ \mathrm{logit}(p_{ij})=\gamma_i + \theta\,t_{ij} + u_i\,z_{ij} \end{aligned} \]

\[ z_{ij}=t_{ij}-0.5,\qquad u_i\sim\mathcal{N}(0,\tau^2),\qquad \log(\mathrm{OR}_i)=\theta + u_i \]

We fitted this in the Bayesian R package brms using the binomial likelihood with:

events | trials(total) ~ 0 + study + treat + (treat12 - 1 | study)

study: fixed study-specific baseline log-odds (\(\gamma_i\)).
treat: treatment indicator (0 = control, 1 = dexmedetomidine), with pooled log-OR (\(\theta\)).
treat12: centered treatment coding (\(treat - 0.5\)), so control = \(-0.5\), dexmedetomidine = \(+0.5\).
(treat12 - 1 | study): study-level random treatment deviation (\(u_i\)), with SD \(\tau\).

Priors used in the fitted model:

\[ b_{\text{study}}\sim\mathcal{N}(0,1.5^2),\qquad b_{\text{treat1}}\sim\mathcal{N}(0,0.82^2),\qquad \tau\sim\mathrm{Log\text{-}Normal}(-1.855,\,0.87^2) \]

For heterogeneity, we used bayesmeta::TurnerEtAlPrior(outcome = "cause-specific mortality / major morbidity event / composite (mortality or morbidity)", comparator1 = "pharmacological", comparator2 = "placebo / control") from Turner et al. (2015): Predictive distributions for between-study heterogeneity and simple methods for their application in Bayesian meta-analysis (DOI: 10.1002/sim.6381).

Shrinkage study-specific effects are computed from posterior draws as \(\mathrm{OR}_i=\exp(\theta+u_i)\) and summarized with median and 95% CrI. Observed OR values are unshrunken two-by-two estimates from metafor::escalc(measure = "OR"), shown as point estimate and Wald 95% CI.

Trial-Level Details

You can select multiple categories in each table filter.

Study	Country	Risk of Bias	Bolus	Infusion	Timing of intervention	Route