Muscle Trajectories in Acute Stroke Patients

Overview

The aim of this study is to explore time-related trajectories of muscle alterations and inflammation in acute hospitalized stroke patients. Furthermore, the researchers want to gain insight in the predictive values of these time-related trajectories towards gait recovery in the acute stroke population.

STUDY DESIGN: Two longitudinal prospective cohort studies will be conducted in which non-ambulatory (cohort 1) and ambulatory (cohort 2) acute stroke patients will participate. PATIENT RECRUITMENT: The investigators aim to recruit 200 subjects (100/cohort). Patients will be recruited at the Neurology ward of the UZ Brussel. PROCEDURE: All stroke survivors admitted to the Neurology ward of UZ Brussel will be screened for eligibility. Afterwards, an informed consent will be conducted for all subjects who met the inclusion criteria. Baseline assessments (T0) of gait recovery outcomes on the one hand and predictors for gait recovery on the other hand will be measured. To predict gait recovery, researchers will observe two novel biomarkers: stroke-induced muscle wasting and inflammation. Furthermore, the investigators will also assess relevant known predictors for gait recovery to compare the relevance of the novel markers. T0 assessments will start preferably within 3 days post-stroke. To assess time-related trajectories of muscle alterations and inflammation, follow-up assessments of these predictors will be performed 3 days after baseline assessments (T1), at discharge (T2) and 3 months follow-up (T3). T1 follow-up measurements will only be possible for patients with motor impairments post-stroke since they have a longer stay at the hospital compared to patients without motor impairments after stroke (mean hospital stay of 5 to 8 days for patients without or with motor impairments post-stroke respectively). The assessments of the gait recovery outcome measures will be repeated at discharge (T2) and 3 months follow-up (T3). MATERIALS: To measure gait recovery in acute stroke survivors, the researchers will make use of wearable gait sensors (Physiolog®, Gait Up SA, Switzerland) to register gait speed and a lightweight chest carrying gas analysis system (Metamax 3B, Cortex, Germany) to measure cardiorespiratory parameters. For the predictors, investigators will use handheld dynamometers (MicroFET2 and Martin Vigorimeter) to assess muscle strength, grip strength and muscle fatigue. Furthermore, researchers need a Bioelectrical Impedance analysis (BIA) device (Bodystat® QuadScan 4000, UK) to assess the muscle mass of our subjects and a portable ultrasound system (Viamo SV 7 with linear-array transducer, Canon Medical Systems, Netherlands) to assess muscle architecture. STATISTICAL ANALYSIS: Various biomarkers will be observed at each of the planned time points. Because the aim is to make correct predictions based on any information that is available at the early stages, the observations will not only be considered as such, but also summarized in terms of their time- related characteristics, such as steepest drop, frequency of improvement, or any other characteristic that may reveal itself as distinguishing. These predictors will be combined into predictive models such as random forests and boosting to establish the best combinations for making good predictions while accommodating inter-predictor correlations. The quality of the models will be established with cross-validation. The large set of observations and their summarized temporal characteristics will be used to determine whether different types of trajectories emerge. Hierarchical cluster analysis will label patients for each of the cluster solutions and the usefulness of patient labelling will be evaluated by their predicted gait performance. The extracted patient type will be included in random forests or boosting to evaluate its importance. The extracted patient type could also be used jointly with other predictors suggested as important in either a linear/logistic (mixed) model or an extended cox proportional hazard model, which are more traditional statistical approaches. While predictive performance remains the key goal, such models would be more interpretable on the potential underlying mechanism, with parameter estimates and confidence intervals.

Conditions

Interventions

Eligibility

Locations (1)

Outcomes

Central Contacts