French National Research Agency – Generic Projects 2022

Integrating multiple sources of evidence for optimized comparative effectiveness research (OptiCER)

Principal Investigator: Anna Chaimani

Participants: Silvia Metelli, Raphaël Porcher, Isabelle Boutron

Host Institution: METHODS Team, CRESS

The systematic identification and synthesis of all available evidence on a specific clinical question is an integral part of the medical decision-making process and the planning of future research. For most medical conditions, a plethora of healthcare interventions are available and, thus, network meta-analysis (NMA) – the only tool enabling simultaneous comparisons between many interventions – is currently among the highest levels of evidence for the development of clinical guidelines. However, current NMA methodology has important limitations for properly synthesizing different study designs (randomised, cohort, case-control, etc.), which could increase the external validity of the findings. These limitations impose restrictions in most NMAs by usually allowing the synthesis of a selective sample of randomised controlled trials only, rendering their results rarely applicable in real settings of healthcare practice. The aim of the OptiCER project is a) to develop and evaluate new synthesis methods for integrating all different sources of evidence available for addressing a common research question and b) to develop new presentation and communication tools for presenting the findings of these ‘all-evidence’ NMAs targeting the needs of different stakeholders (clinicians, regulators, guideline developers). All new statistical models and tools will be accompanied by open-access and user-friendly software. The new methods will be used for the synthesis of the COVID-19 studies but will also be freely available for application in other medical fields. In this way, clinical recommendations will be based on a global view and synthesis of any piece of evidence rather than on a partial approach.

Marie Skłodowska-Curie Individual Fellowship

Dynamic Comparative Effectiveness Research for health care interventions (DyCER)

Individual Fellow: Silvia Metelli

Supervisor: Anna Chaimani

Host Institution: METHODS Team, CRESS

Many health conditions are complex and dynamic in nature. For instance, chronic diseases often require the individuals to undergo treatments for long periods of time, thus involving a series of progressive treatment decisions to be made. Understanding how treatment effectiveness varies over time is crucial to better inform clinical decisions. So far, comparative effectiveness research has mainly taken a static perspective both in primary research and in evidence synthesis. A first attempt to move towards a dynamic approach in primary research was provided recently with just-in-time adaptive interventions (JITAIs), which are mobile technology interventions where both treatment and response occur intensively over time (see also dynamic treatment regimes); while for evidence synthesis a dynamic approach has never been considered. DyCER aims to contribute to bridge this gap in two different, yet related, ways: (i) investigating dynamic regimes and providing methods for JITAIs personalised at patient-level and (ii) developing methodology for the dynamic evolution of networks of interventions. In both cases, the project will focus on the case of complex interventions, which are interventions composed of multiple, possibly interactive, components (examples include psychological interventions for mental illness or smoking cessation support). Overall, this holds the potential to yield more effective planning of future research, and in the case of JITAIs, more personalized decision-making.

IdEx Université Paris 2019 Fellowship

Dealing with sparse data in network meta-analysis

Doctoral Fellow: Theodoros Evrenoglou

Supervisor: Anna Chaimani

Host lab: METHODS Team, CRESS

Background: A major issue when analyzing networks of interventions is the absence of sufficient data to draw useful conclusions. This phenomenon has two aspects. First, at the trial-level the outcomes under investigation are often rarely observed (e.g. different types of adverse events). Results from conventional meta-analyses and NMA models are based on large sample approximations which in the presence of rare events are often implausible. As a result, using such models to synthesize studies with rare events threatens the validity and accuracy of the results of meta-analyses and NMAs. Second, at the comparison-level networks of interventions often appear rather sparse; hence only a handful of studies per comparison is available and interventions are weakly connected with each other. Obtaining meaningful and interpretable results from such networks is quite challenging since a) clinical and statistical assumptions of NMA cannot be evaluated properly and b) the relative effect estimates are usually imprecise and fail to answer the clinical questions with respect to the effectiveness of interventions.

Objectives: The aim of this doctoral fellowship is to develop new statistical methods that will allow drawing useful and meaningful conclusions from NMAs with sparse data both at trial- and comparison-level. Specifically,

  1. to develop new methods for dealing properly with rare events in meta-analysis and NMA
  2. to develop new NMA models that will allow sharing information across different networks that target relevant questions
  3. to illustrate the use of the above methods using real clinical examples

Other projects we participate

More info

More info

Assistance Publique Hôpitaux de Paris (Clinical Research and Development Department)

Live cumulative network metaanalysis: Systemic pharmacological treatments for chronic plaque Psoriasis

Principal Investigator: Emilie Sbidian

Completed projects

Université Sorbonne Paris Cité Chaire d’Excellence Fellowship – Methods for Evidence Synthesis

Principal Investigator: Anna Chaimani

Host Institution: METHODS Team, CRESS

Washington University Institute of Clinical and Translational Sciences – Career Development Awards Program

Principal Investigator: Sonal Patil