SI: CHALLENGES AND CONTROVERSIES IN TRANSLATIONAL STROKE RESEARCH How to Measure Recovery? Revisiting Concepts and Methods for Stroke Studies Marc Hommel 1,2 & Olivier Detante 3 & Isabelle Favre 3 & Emmanuel Touzé 4 & Assia Jaillard 1,2 Received: 11 February 2016 /Revised: 3 July 2016 /Accepted: 21 July 2016 /Published online: 8 August 2016 # Springer Science+Business Media New York 2016 Abstract In clinical trials, assessing efficacy is based on val- idated scales, and the primary endpoint is usually based on a single scale. The aim of the review is to revisit the concepts and methods to design and analyze studies focused on resto- ration, recovery and or compensation. These studies are be- coming more frequent with the development of restorative medicine. After discussing the definitions of recovery, we ad- dress the concept of recovery as the regain of lost capabilities, when the patient reaches a new equilibrium. Recovery is a dynamic process which assessment includes information from initial and final status, their difference, the difference between the final status of the patient and normality, and the speed of restoration. Finally, recovery can be assessed either for a spe- cific function (focal restoration) or for a more global restora- tion. A single scale is not able to assess all the facets of a skill or a function, therefore complementary information should be collected and analyzed simultaneously to be tested in a single analysis. We are suggesting that recovery should be consid- ered as a latent variable and therefore cannot be measured in pure form. We are also suggesting to customize the data col- lection and analysis according to the characteristics of the subjects, the mechanisms of action and consequences of the intervention. Moreover, recovery trials should benefit from latent variable analysis methods. Structural equation modeling is likely the best candidate for this approach applicable in pre- clinical and clinical studies. Keywords Methodology . Intervention evaluation . Study design . Clinical scale . Latent variable . Modeling Introduction Many definitions of recovery have been proposed, from neu- ronal rewiring and anatomical restitution to improvement of performance or coping and social integration. From the perspective of the conceptual framework of the WHO International Classification of Functioning (ICF) [1, 2], recovery refers to health condition, at different levels of body function and structure, activities, and participation. In order to determine the most important indicators of successful rehabil- itation outcome, an attempt to classify instruments according to ICF was proposed such as the Fugl-Meyer scale [3] to assess impairment at the body structure level, the Barthel Index [4] and the mRankin scale [5] at the activities level, and the Euroquol-5D [6] at the participation level [7–9]. Furthermore, the term Btrue recovery^ has been used to define the complete restitution of neuronal networks [10, 11]. However, there is still some ambiguity in defining true recovery, since this term is used to refer to both improved performances and return to normal clinical patterns [12]. Focusing on motor system, some motor tests based on video assessment have been developed to quantify recovery [10], assuming restitution (true recovery) is present if recovered movements have the same quality (patterns) as normal move- ments [10, 12, 13]. In contrast, impaired performance would reflect incomplete recovery that may be due to compensation [10, 11]. * Marc Hommel marc.hommel@ujf-grenoble.fr 1 University Grenoble Alpes, AGEIS EA 7407, Grenoble, France 2 CHU Grenoble, Pôle Recherche, Grenoble, France 3 CHU Grenoble, Pôle Psychiatrie Neurologie, Stroke unit, Grenoble, France 4 INSERM U919, Stroke unit, CHU, University of Caen Basse Normandie, Caen, France Transl. Stroke Res. (2016) 7:388–394 DOI 10.1007/s12975-016-0488-0