Stefan Elbers

71 Living systematic reviews in rehabilitation science BACKGROUND Systematic reviews are considered to be the foundations of evidence-based practice (Chandler & Hopewell, 2013; Shojania et al., 2007; Stevens, 2001). The structured method of identifying, appraising, evaluating and synthesizing primary research findings facilitates clinical decision making based on the available evidence at a given time (Boutron et al., 2020; Simmonds et al., 2017). Despite its widespread success, researchers have identified several problems in the current conduct of systematic reviews. First, the average duration of performing a systematic review is almost a year from initial search to publication (Chandler & Hopewell, 2013). Furthermore, only a minority of the published reviews are updated and the median estimated time to update is approximately 3 years for Cochrane reviews and 5 years for non-Cochrane reviews (Bashir et al., 2018; Rombey et al., 2020). Given the continuously increasing flow of new trials, many systematic reviews become rapidly out of date (Bastian et al., 2010; Boutron et al., 2020). Second, poorly prepared systematic reviews are susceptible to bias, such as incomplete search strategies (Salvador-Oliván et al., 2019), errors in data extraction (Horton et al., 2010), and inter-rater disagreement in risk of bias assessments (Hartling et al., 2009). Although the effect on outcomes may be limited (Jones et al., 2005), the current system does not easily allow for corrections or updated practices after publication of the initial results. Living systematic reviews (LSRs) are proposed as an alternative method to traditional systematic reviews that may offer a solution to these challenges. A systematic review is considered “living” when it includes a system by which it continuously incorporates newly available evidence into the analysis (Elliott et al., 2017). Generally, this system consists of a predefined cyclic workflow with planned updates at intervals ranging from days to months. Assisted by the availability of recent software developments that provide (semi) assisted solutions for the review process, these updates take only a fraction of the time of conducting a base review (Créquit et al., 2020; Thomas et al., 2017). The chosen timeframe is at the discretion of the researchers and generally depends on the speed at which a topic of interest is expected to develop. Although LSRs have specifically been designed for rapid updating cycles of emerging topics, such as mapping the COVID-19 trials (COVID-NMA), protocols have been published that include updating cycles of 12 months (De Zeeuw et al., 2020; Pollock & van Wijck, 2019; Winters et al., 2018). The aim of this commentary is to illustrate the process of transitioning a regular review into a LSR and to propose a workflow for updating LSRs over time. We will use a recently performed research project where we transitioned a regular systematic review into a LSR as an example to discuss the required steps and key issues.