7 Méta-recherche

     

Aucune étude de méta-recherche n’a été trouvée. Le nombre d’essais bayésiens publiés jusqu’à présent (novembre 2021) semble restreint (une recherche non exhaustive exhaustive rapporte quelques dizaines d’essais contrôlés randomisés) [1 , 2 , 4 , 5 , 6 , 8 , 9 , 10 , 11 , 12 , 13] . Cette approche a été utilisée à plusieurs reprises pour des traitements de la COVID. Antérieurement la majorité des essais concernait des dispositifs médicaux.

La plupart de ces essais a utilisé des « apriori » non informatif et des seuils de probabilité à postériori d’efficacité d’au moins 97.5%, avec fréquemment un ajustement pour la multiplicité :

Essai [ref]

Choix du prior (dans matériel et méthodes)

Early Convalescent Plasma for High-Risk Outpatients with Covid-19 [12]

The prior probability of outcome for each treatment group was assumed to follow a noninformative beta distribution, which yielded a beta distribution for the posterior probability when a binomial likelihood was assumed for the outcome.

Effect of Tocilizumab vs Usual Care in Adults Hospitalized With COVID-19 and Moderate or Severe Pneumonia [11]

For the day 4 outcome, we used a beta prior distribution with parameters 1 and 1 for the proportion in each arm (eFigure 1 in Supplement 2). For the day 14 outcome, we used a Gaussian prior distribution with a mean of 0 and variance of 106 for the log hazard ratio (HR)

For the primary analyses, a non-informative flat prior distribution for the log HR was used, as a Gaussian distribution with mean 0 and variance 106

Effect of anakinra versus usual care in adults in hospital with COVID-19 and mild-to-moderate pneumonia (CORIMUNO-ANA-1) [4]

For the day 4 outcome, we used a β prior distribution with parameters 1 and 1 for the proportion in each treatment group. For the day 14 outcome, we used a Gaussian prior distribution with a mean log hazard ratio (HR) of 0 and variance of 1 × 10⁶ for the log HR.

Interleukin-6 Receptor Antagonists in Critically Ill Patients with Covid-19 [10]

Prior distributions for individual treatment effects were neutral

Pas plus de précision dans l’article

REMAP-CAP protocol [7]

REMAP-CAP launches with no prior assumptions regarding which interventions are superior, akin to a typical RCT design.

Azithromycin for community treatment of suspected COVID-19 in people at increased risk of an adverse clinical course in the UK (PRINCIPLE): a randomised, controlled, open-label, adaptive platform trial [6]

The log hazard ratio for treatment has the weak informative prior j N(0; 0:32); and is assumed to be constant over time. The weak informative prior for the log hazard ratio places the prior mass of the HR between 0.5 and 2.0, which in line with clinical expectations for potential therapies, and also will be quickly overwhelmed with accruing data.

Therapeutic Anticoagulation with Heparin in Noncritically Ill Patients with Covid-19 The ATTACC, ACTIV-4a, and REMAP-CAP Investigators [5]

The primary model incorporated weakly informative Dirichlet prior distributions for the number of days without organ support