Introduction

Beyond 13 million people are affected by stroke annually, which is among the leading causes of incapacity and mortality, placing a significant psycho-socio economic burden worldwide [1]. Through the advent of medical techniques, mortality rates following stroke have dropped remarkably, leading to higher disability-adjusted life years (DALY) [2]. Approximately 26% of patients with stroke (PwS) will face levels of dependency in their daily activities up to six months after the stroke onset [3]. However, the motor function is not the mere focus of disabilities among PwS; patients’ cognitive profile and affect are also altered by stroke [4, 5].

Acute stroke is referred to the acute onset of focal neurological findings in a vascular territory. There are two main types of stroke, following a vascular occlusion resulting in an infarct, regarded as ischemic stroke (85%), or caused by a sudden burst of a blood vessel, regarded as hemorrhagic stroke (15%). In response to this cerebrovascular event, CNS uses an ability to adapt and reorganize the damage known as “Neuroplasticity”. A group of signaling molecules known as neurotrophins play essential roles in synaptic plasticity, neurogenesis and survival regulations in neural cells [6,7,8]. In particular, many studies have investigated Brain-Derived Neurotrophic Factor (BDNF) as a member of the neurotrophins family for its role in the nervous system [9,10,11]. Extensive evidence demonstrated some associations between BDNF level and various neurological disorders; for instance, low concentrations of BDNF has been reported in neurodegenerative and neuropsychiatric disorders such as Alzheimer’s disease, Parkinson’s disease, depression, autism spectrum disorders, and post-traumatic stress disorder [9,10,11,12].

Keen interest has been raised to find the relationship between the altered level of BDNF and stroke; studies hypothesize that BDNF level is correlated with the size of the stroke legion, cognitive profile, and functional outcome [13,14,15]. Based on the substantial impact of stroke, a systematic review and meta-analysis are essential to address the precise association of circulating BDNF in PwS in various conditions.

Materials and method

Literature search and selection criteria

We conducted a comprehensive systematic search by using the keywords “stroke” OR “ischemic stroke” OR “hemorrhagic stroke” OR “Cerebrovascular Accident” OR “Brain Vascular Accident” AND “Brain-Derived Neurotrophic Factor” OR “BDNF” on PubMed, Web of Science and Scopus in January 2021. Original articles were included by the following criteria; (1) a study measuring serum or plasma levels of BDNF in PwS. Studies (a) with animal subjects, (b) review or meta-analysis studies, (c) genetic investigation were excluded from the study (d) using other measuring methods rather than ELISA, like western blot, mass spectrometry or immunohistochemistry.

Additionally, the reference list and citations of all included articles were checked to find potential eligible articles, resulting in 5 eligible articles. We updated our search, using the exact keywords, on August 7th 2021, which resulted in the inclusion of one study [16]. Finally, a total number of 62 articles were entered into our systematic review (Fig. 1).

Fig. 1
figure 1

PRISMA flow diagram

Full size image

Data extraction

The following data were extracted from each included publication; the name of the first author, year of publication, location of study (country), type of specimen (serum, plasma or CSF), number of subjects in each study group, demographic characteristics (i.e., age and gender) of both groups, stroke type (ischemic or hemorrhagic), stroke site, patients’ score for National Institutes of Health Stroke Scale (NIHSS), post-stroke duration at the time of sample collection, mean and standard deviation (SD) of the peripheral BDNF levels, and the assay used for BDNF measurement.

Two independent researchers extracted data from each manuscript, and the mismatch was discussed and resolved afterwards. In the case of figure illustration, the data were interpreted from the figure using an online interpreter [17]. The corresponding/first authors were contacted and invited to share their results if the two prior approaches were unsuccessful. An excel spreadsheet was made from the relevant data and is available on request. Characteristics of the included studies is illustrated in Table 1.

Table 1 Characteristics of the included studies
Full size table

 

Quality assessment

Two authors performed the methodological quality assessment independently to evaluate each included study's risk of bias and applicability by QUADAS-2 tool [18]. This tool is designed to assess the risk of bias in four main domains, including patient selection, index test, reference standard and, flow and timing. Concerns regarding the applicability were also apprised in the first three domains. In each domain, there are three possible rankings; studies categorized as “high” represent the lowest quality with the highest risk of bias and concerns, while studies categorized as “low” indicate the highest quality with the lowest risk of bias and considerations. The “unclear” category was used for studies that reported insufficient data in each domain. Any disagreement was resolved by discussion or consultation with another author (Table 2 and Fig. 2).

Table 2 QUADAS
Full size table
Fig. 2
figure 2

Proportion of risk of bias assessment by each domain

Full size image

Quantitative analysis

All statistical analyses were performed using “meta” (version 4.17–0), “metafor” (version 2.4–0), and “dmetar” (version 0.0–9) packages, R (R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria) and STATA. A p-value of < 0.05 was considered statistically significant. 

Meta-analysis

We reported standardized mean difference (SMD) (Hedge's g) and 95% confidence interval (CI) for each between-group comparison. The SMD of \(\le\) 0.2, 0.2–0.8, and 0.8 \(\le\) represented small, moderate, and large effect sizes, respectively.

The Cochran’s Q test and the I2-index were used to investigate heterogeneity between studies. The I2-indices of \(\le\) 25%, 26–75%, and 75% ≤ represented low, moderate, and high degrees of heterogeneity, respectively [19]. Fixed effects models were applied if the results were homogeneous (I2 < 40% and p-value > 0.05), and random effect models according to the DerSimonian and Laird method [20] were used if these results were heterogeneous (I2 ≥ 40% or p-value ≤ 0.05) [21].

Our primary aim was to compare BDNF levels between PwS and healthy controls (HC). Additionally, we aimed to address (1) the longitudinal changes, (2) the effects of physical training, (3) repeated transcranial magnetic stimulation (rTMS), and (4) clinical diagnosis of post-stroke depression (PSD) on the circulating BDNF levels in PwS.

Within the groups undergoing additional analysis, some concerns existed in the physical training group. First, various training modalities were implemented, ranging from regular post-stroke physiotherapy sessions to different exercise protocols in addition to the variation in each session’s duration. Lastly, sample collection time varied between studies immediately after the session completion to weeks and months after the whole program was terminated.

To overcome this diversity and reach a reasonable grouping modality with maximum homogeneity, this is besides the heterogeneity and sensitivity analysis, we divided studies into two primary groups according to the time of sample collection, those who collected samples (1) immediately after the exercise (within 20 min to 24 h) and (2) with a time delay after the training was completed (within two weeks to a couple of months). The variety of the training protocols was also an interesting point throughout the studies. Thus we conducted a subgroup analysis between routine physiotherapy protocol and other exercise protocols. Although there were several specifications for the exercising routines like High-Intensity interval training (HIIT), we could not devote a specific subgroup for each due to the inadequate number of observations.

It is noteworthy that some of the included studies had applied two or more physical interventions, such as routine physiotherapy, aerobic exercise, and non-aerobic exercise, into different study groups, leading to more observational groups being included from the study. Similarly, many studies have reported both immediate and delayed levels so we could extract more than one observational groups from a single study. This clearly describes repeated study names in our meta-analysis results. Table 3 illustrates a description of the included studies in our physical training group in detail.

Table 3 PT protocols
Full size table

We had no concerns about the categorization of the other additional group analysis.

Regression and sensitivity analysis

To further assess causes of heterogeneity, we conducted separate regression of mean age and mean NIHSS2 on SMD whenever the required data was available for a number ten or more studies. Moreover, for the meta-analyses including ten or more studies that had significant heterogeneity, we conducted sensitivity analysis to identify influential articles. We performed Leave-One-Out analysis by omitting one study each time and recalculating the effect size. We also applied the diagnostic procedure developed by Viechtbauer and Cheung [22] to detect influential cases and Baujat Plot [23] to investigate each study's effect on the effect size and heterogeneity.

Meta bias

We assessed publication bias by visual inspection of the degree of funnel plot asymmetry. Egger bias test [24] and Begg-Mazumdar Kendall's tau [25] were also performed to confirm the visual perception from the funnel plot objectively. A p-value < 0.1 was considered as evidence of publication bias.

Results

Our initial database search revealed a total number of 6243 citations. We removed 2495 duplicate citations, and 3653 additional citations were excluded through title and abstract screening. We retrieved 95 articles for full-text screening, from which 29 studies were excluded for the following reasons: animal sample size, review article, duplicated articles from one same study, in vitro study, and reporting one single measure in one single study group at only one-time point. An additional number of 5 articles were removed during the data extraction because of: insufficient data despite author contact [22, 26], reporting BNGF instead of BDNF [23], unclear data from an extensive cohort study (Framingham [3]), reporting peripheral BDNF in association with another biomarker instead of pure measures [27].

Finally, a total number of 62 articles were entered into our systematic review. We could not enter 16 articles to any meta-analysis group for the following reasons; providing delta BDNF instead of exact measures [28], reporting peripheral BDNF as a proportion to the control group [29], the full text was not found. However the abstract was sufficient to extract the descriptive data [30], unsatisfactory data and irresponsive author (missing SD despite reporting the mean) [31], single study group with no comparison or specific grouping which did not fall into our meta-analysis categorizations [14, 32,33,34,35,36,37,38,39,40,41,42].

Serum BDNF levels in PwS vs. HC

Seventeen studies assessed the differences between serum BDNF levels in PwS (n = 1191) and HC (n = 1856) [13, 16, 43,44,45,46,47,48,49,50,51,52,53,54,55,56,57]. PwS had significantly lower serum BDNF levels than HC (SMD [95%CI] = − 1.02 [− 1.47 to − 0.57], p-value < 0.001, I2 = 96%, p-value < 0.001) (Fig. 3). No publication bias was detected (Fig. 4). Sensitivity analysis showed that Align et al. [58] and Chan et al. [54] were influential (Additional file 1: Fig. S1). After excluding these studies, the meta-analysis of the remaining fifteen studies showed significantly lower serum BDNF levels in PwS than HC as well (− 0.92 [− 1.35 to − 0.50], p-value < 0.001, I2 = 96%, p-value < 0.001). To further assess sources of heterogeneity, we performed meta-regression. The age partially explained the high heterogeneity (correlation coefficient = − 0.11, R2 = 62.81%, p-value = 0.000).

Fig. 3
figure 3

Forest HC stroke

Full size image
Fig. 4
figure 4

Funnel HC stroke

Full size image

Longitudinal investigation within stroke population

To address the pattern of serum BDNF level through the time of stroke onset, 23 studies longitudinally assessed their samples. A total number of 3678 patients were included in 39 observations. Time points for sample collection varied extremely between the studies. After the precise data extraction for the time of sample collection regarding the stroke onset, we grouped the observation into five general groups. (1) Observations which compared baseline measures to the first day of the stroke, baseline vs day one or acute that covered four observations and 202 participants [32, 39, 59, 60]; (2) Observations comparing the baseline measures to the peripheral level of BDNF within the first week of stroke onset (mostly 3–7 days) baseline vs week one or sub-acute, consisting of 14 observations and 1366 patients [32, 39, 42, 44, 54, 56, 59, 61,62,63,64,65,66,67]; (3) Observations investigating the difference between baseline levels of BDNF and day 30 or above, baseline vs. over one month or chronic, that included 11 observations and 1093 participants [16, 39, 43, 54, 56, 65,66,67,68,69]; (4) Comparison of the first day of the stroke to the first week after the onset, Day 1 vs Week 1 consisting three observations and 162 participants [32, 39, 59]; (5) Lastly the seven observations investigating the BDNF pattern, in 855 participants, within the first week of stroke and over the one months of affecting the patients [39, 54, 56, 65,66,67, 70].

Based on the high heterogeneity level detected with I2 (over 74%) in every group, we conducted a random effect model analysis separately for each above mentioned observational group. No significant difference were addressed in the SMD in any of the included groups. Nor any publication bias was found (Additional file 2: Fig. S2, Additional file 3: Fig. S3, Additional file 4: Fig. S4, Additional file 5: Fig. S5, Additional file 6: Fig. S6, Additional file 7: Fig. S7, Additional file 8: Fig. S8, Additional file 9: Fig. S9, Additional file 10: Fig. S10, Additional file 11: Fig. S11).

Effect of depression on serum BDNF levels in PwS

Six trials with 411 patients and 245 patients without depression compared serum BDNF between these two groups [43, 44, 71,72,73,74]. All the studies followed Diagnostic and Statistical Manual of Mental Disorders 4th version (DSM-IV) criteria to assess their sample for depression. Hamilton Depression Rating Scale (HDRS) was also used by four studies [43, 44, 73, 74]. While every included article assessed PSD, only two addressed the history of depression in their patients before the stroke [72, 73]. Patients with depression had significantly lower levels of BDNF than the participant in the non-depressed group (SMD [95%CI] = − 0.60 [− 1.10 to − 0.10], p-value < 0.001, I2 = 88%, p-value < 0.001) (Fig. 5). No publication bias was detected. Due to the low number of included studies, we could not conduct meta-regression or sensitivity analysis (Fig. 6).

Fig. 5
figure 5

Forest depression

Full size image
Fig. 6
figure 6

Funnel depression

Full size image

Effect of physical training on serum BDNF levels in PwS

Sixteen studies with 738 patients provided original data on BDNF concentration before and after applying a specific physical training protocol. The training modality included regular physiotherapy sessions [75,76,77], aerobic exercises [78,79,80], high-intensity interval training (HIIT) [81], and proprioceptive neuromuscular facilitation exercise (PNF) [82] (Details are illustrated in Table 3). In addition, the number of sessions varied among the included studies starting from only one session [78] to as many as 56 sessions [16]. Also, the duration of sessions ranged from 12 min [83] to 120 min [75]. 13 observational groups were included in the immediate analysis, and eleven observational groups were in the delayed analysis.

Sample collection immediately after physical training

Analysis for the immediate group showed a positive effect of physical training in general on BDNF level immediately after the intervention with (SMD [95%CI] = 0.49 [0.09 to 0.88]), p-value = 0.02, I2 = 85%, p-value < 0.001) (Fig. 7). Within the immediate group, we did a subgroup analysis on different training modalities. Eight observations that performed any sort of exercise fell into our exercise subgroup, and five observations applying regular rehabilitation or physiotherapy were categorized as rehabilitation subgroup. The subgroup analysis showed that only in the exercise group BDNF levels significantly increased immediately after physical training (SMD [95%CI] = 0.75 [0.25 to 1.25], p-value = 0.003). No subgroup differences were detected. Sensitivity analysis showed that the study of Anjum et al. [84] was influential (Additional file 12: Fig. S12). After omitting this record, the overall effect size did not remain significant (SMD [95%CI] = 0.34 [− 0.03 to 0.71], I2 = 82%) (Additional file 13: Fig. S13). Lastly, publication bias was not observed between the included studies (Fig. 8).

Fig. 7
figure 7

Forest immediate

Full size image
Fig. 8
figure 8

Funnel immediate

Full size image

Sample collection with a delay after physical training

The analysis of the delayed group consisting of eleven observations showed no significant effect of the intervention at this delayed phase of sample collection (SMD [95%CI] = 0.02 [− 0.43 to 0.47], I2 = 83%) (Fig. 9). Similarly, BDNF levels did not change significantly in any of the subgroups. Leave-one-out analysis showed that after omission of Wang et al., 2021 study [16], the I2 index reduced to 68% while the overall effect size remained not significant (Additional file 14: Fig. S14 and Additional file 15: Fig. S15). No publication bias in either group was noted (Fig. 10).

Fig. 9
figure 9

Forest delayed

Full size image
Fig. 10
figure 10

Funnel delayed

Full size image

Repeated transcranial magnetic stimulation (rTMS)

Four studies measured the BDNF levels in 105 PwS after receiving rTMS; Lu 2015 used plasma while the other three used serum samples [56, 75, 77, 85]. The overall meta-analysis of the studies revealed no difference in the BDNF levels between the rTMS and the sham stimulation group (SMD [95%CI] = 0.00 [− 0.27 to 0.27]). No significant heterogeneity was found (I2 = 0%, p-value = 0.48) (Fig. 11). Meta bias in the included studies is presented visually in the funnel plot (Fig. 12), with Begg’s and Egger’s tests of 0.0603 and 0.081, respectively.

Fig. 11
figure 11

rtms stroke corrected

Full size image
Fig. 12
figure 12

Funnel rtms

Full size image

Discussion

With this study, we confirm that the BDNF level is significantly lower in the patients with stroke than in the healthy controls. Interestingly, it is also lower in patients with post-stroke depression compared to PwS without depression. These results are compatible with the hypothesis that a lower level of BDNF is associated with the pathogenesis of neural loss in several neuropsychiatric disorders and cognitive deficits [12, 86,87,88,89,90,91]. Although the severity and type of stroke, either ischemic or hemorrhagic, can possibly correlate with the BDNF level as addressed by Chaturvedi et al. [82], the lack of reporting data from the included studies prohibited us from further regression analysis of these influential factors.

We assume that neural rehabilitation following a stroke measured with BDNF happens in an insignificant amount in the absence of additional intervention. Individual differences in the neural recovery have made some fluctuations in the level of BDNF days, weeks and months post-stroke, but have led to no significant differences between various time points in this meta-analysis; the overall BDNF pattern was neither increasing nor decreasing through time. This implies that stroke can lead to an irreversible decrease in the level of BDNF or the rate of neural recovery.

In addition, our analysis also confirmed a positive effect of physical training, regardless of performing exercise or routine physiotherapy or rehabilitation, on the BDNF level immediately after the intervention. In contrast, no significant effect was detected when the BDNF samples were collected after a time point from the intervention (delayed). Moreover, the analysis of the effect of rTMS on PwS showed zero effect on the circulating BDNF.

BDNF protein is a member of nerve growth factors, discovered in 1982 [92]; since then, a strong body of evidence has suggested important roles of BDNF in neurogenesis and synaptic plasticity along with suppressive effects on apoptosis in the central and peripheral nervous system. BDNF works through different signaling mechanisms by binding to its high-affinity receptor, known as tropomyosin receptor kinase B (TrkB) [90, 93]. It appears that three main cascades are involved in BDNF signaling pathways; (1) Ras/MAPK/ERK pathway; (2) IRS-1/PI3K/AKT pathway, and (3) PLC/DAG/IP3 pathway. These pathways result in the regulation of the encoding of various proteins, which are associated with different processes such as neurogenesis, synaptic plasticity, and cell survival [90, 94,95,96]. All of which are essential for synaptogenesis, restoring the ability of recovery and survival after a neural insult such as stroke [97]. Circulating BDNF is detected and measured chiefly by ELISA kits; However, these kits demonstrate a reasonable difference in mean BDNF measures in the same samples; they differ in the number of mature vs pro BDNF detection. Among the existing kits, the Biosensis and Aviscera-Bioscience, measuring total BDNF and mature BDNF, respectively, are the most performant and recommended for clinical studies. Moreover, plasma samples are sensitive to preparation procedures such as room temperature and varying time needed for platelet lysis, making the pre-procedure time-consuming and leading to different reported measures with changing operators. Thus, it is advised to measure serum BDNF concentration minimally affected by extrinsic factors and about 100 folds higher than plasma levels [98].

Many studies have focused on the alternation of the peripheral BDNF level in different situations and reported BDNF changes in neurodegenerative disorders, brain insults, and psychiatric disorders [12, 89,90,91, 93]. Stroke as the second leading cause of death and morbidity worldwide is investigated remarkably, specifically in regards to post-stroke recovery measures. A notable number of studies have investigated the impact of numerous modalities such as oral or intravenous medications, physical interventions, and brain stimulation on the BDNF level as an indicator of plasticity in PwS. Within medications, rtPA, antioxidants such as Saffron, statins, and Cytoflavin, have shown promising effects on increasing the BDNF level in each individual study [39, 42, 52, 61], which did not meet the meta-analysis threshold in the current research article (total number of 3 studies). Additionally, specific brain stimulation using rTMS which was assessed by 4 independent studies showed zero effect on BDNF in our meta-analysis.

Post stroke depression (PSD)

The current study reveals decreased levels of BDNF among the PSD group compared to PwS with no clinical depression. This implies that a decreased BDNF level can be an early predictor of depression in PwS. Since the BDNF is also decreased generally in PwS, there should exist a definite criteria, either a threshold or a difference amount of drop in BDNF level, on which we consider a PwS susceptible to develop PSD. Suggested amount in the current literature are very wide and inconsistent for this matter. Yang et al. [44] mentioned a 28 fold higher risk of depression in PwS with BDNF levels lower than 5.86 ng/ml, while Li et al. [99] report an 11.5-fold increase in the risk of post-stroke depression in case of BDNF < 10.2 ng/ml. An increasing response to BDNF to antidepressant consumption in PwS diagnosed with PSD also suggests the effect of BDNF among this population [100, 101]. Altogether may open the door to the proposal of new therapeutic options in PwS.

Physical training

We confirm that exercises and rehabilitation programs promote BDNF production and motor recovery following stroke [102, 103]. Literature shows the impacts of training on post-stroke circulating BDNF levels vary based on the program’s duration and intensity [102, 104]. Programs with higher intensity resulted in higher BDNF levels as confirmed by de Morais et al., a significant increase in serum BDNF levels were detected in their moderate exercise group (64–76% of maximum heart rate), not their mild group (50–63% of maximum heart rate) [105]. Moreover, gradually increasing intensity results in a significantly higher increase in serum BDNF [102]. Similar to the exercise, the degree of the rehabilitation program correlated significantly with the BDNF level and motor recovery, supposing a critical threshold in the intensity that the impact could not be detected below that level [106]. As defined in mentioned studies, a threshold exists for the severity of the training that zero benefit is noticed below the point. This makes us assume that the duration of each session would have the same pattern of effect, which is awaiting to be discovered.

Although the duration of the training can potentially alter BDNF levels, none of our included studies has performed a specific, clear comparison for this matter (e.g., comparing the numbers of sessions or the session duration). In regards to comparing different exercise protocols with rehabilitation modalities, El-tamawy et al. suggest that implementing physiotherapy in combination with aerobic exercise leads to higher BDNF levels compared to the routine physiotherapy group [76]. Lastly, the effect of time points in data collection was of remarkable significance. Our analysis confirms a promising short-term effect of exercise on motor recovery by higher levels of circulatory BDNF when assessed immediately after the session, while low to zero long-term impact is detected on BDNF level following exercise. The problem is that the current protocols of treatment are failing to have a sustained effect and the termination of the intervention leads to a decrease in overall benefit. At this point, we need to come up with more effective treatment protocols for the physical training, and this does not imply that the exercise per se is ineffective; rather our existing modality is.

Conclusion

In conclusion, stroke significantly affects the level of BDNF, which is positively correlated with neural plasticity and post-stroke recovery in various domains such as cognition, affect, and motor function. Our study also confirms the significantly lower levels of BDNF in patients with stroke than the healthy controls, patients with the clinical diagnosis of depression to non-depressed patients and significantly higher levels of BDNF in patients who do moderate to severe physical training. Furthermore, literature imposes different modalities to overcome the effect of BDNF reduction, from which our study confirms the short-term effect of moderate to intense exercise or rehabilitation. We believe that BDNF could be regarded as a valuable diagnostic biomarker for acute stroke and a potential screening factor to observe the effectiveness of treatment. Existing modalities lack a definite protocol by which a maximum benefit is reached. Thus, further studies should focus on addressing the exact dosage of proposed drugs such as antioxidants, antidepressants, statins and a definite intensity and duration for the physical training. Lastly, investigating the role of BDNF supplementation for severe stroke patients could be of great value.