The effects of interval training on peripheral brain derived neurotrophic factor (BDNF) in young adults: a systematic review and meta-analysis

The aim of the current meta-analysis was to determine the effects of acute and chronic interval training (IT) on serum and plasma BDNF concentrations in healthy young adults. A literature search was performed using six databases until February 2020. The TESTEX scale was used to assess the quality of studies. Effect sizes (ES) were computed and two-tailed α values < 0.05 and non-overlapping 95% confidence intervals (95% CI) were considered statistically significant. Heterogeneity, inconsistency (I2), and small-study effects using the Luis Furuya–Kanamori (LFK) index were examined. Fifteen studies (n = 277 participants, age = 24 ± 3 years) were included. The overall effects of IT on circulating BDNF concentrations were moderate and significant (ES = 0.62, 95% CI 0.00, 1.24, heterogeneous (p < 0.001), highly inconsistent (I2 = 90%), and with major asymmetry (LFK index = 2.76). The acute effect of IT on peripheral BDNF levels was large and significant (ES = 1.10, 95% CI 0.07, 2.14), heterogeneous (p < 0.001), highly inconsistent (I2 = 92%), and with major asymmetry (LFK index = 3.34). The chronic effect of IT on circulating BDNF was large and significant (ES = 0.93, 95% CI 0.40, 1.46), heterogeneous (p < 0.001), with moderate inconsistency (I2 = 70%), and minor asymmetry (LFK index = 1.21). Acute and chronic IT elicited a moderate increase in serum and plasma BDNF concentrations in a healthy young population.

Eligibility criteria. Studies that met the following criteria were included: (1) randomized controlled trials (RCT) and controlled trials without randomization (pre-test), (2) healthy normal-weight participants (as determined by a body mass index (BMI) between 20 to 24 kg/m 2 or a body fat mass < 20% for men and < 28% for women, (3) young adults (18 to 40 yr. old), (4) male and female of different ethnic groups, (5) interventional studies, (6) serum and plasma circulating BDNF, (7) studies including participants free of any pharmacological prescription medication or drug, or recreational smoking, (8) studies using the enzyme-linked immunosorbent assay (ELISA) method to determine circulating BDNF. Studies that met the following criteria were excluded: (1) studies involving overweight and obese participants (BMI > 25 kg/m 2 ), (2)  Search strategy. Search strategies were developed using text words as well as Medical Subject Headings associated with the effects of exercise on BDNF. The search strategy included the following key words in English language: interval training, BDNF, intermittent training, high intensity intermittent training, interval running, brain-derived neurotrophic factor, high-intensity interval training, HIIT, sprint interval training, SIT, CrossFit, Tabata. Boolean operators AND, OR, NOT OR Mesh option were used to concatenate the search terms (key words). A secondary search was performed by screening the reference list of the selected studies and relevant review articles. Finally, a forward citation tracking of the selected studies was conducted through Scopus. An example of the search strategy for one of the databases searched (PubMed) is shown in supplementary Fig S1 online.

Study records and selection.
All studies to potentially be screened were imported into Mendeley software, version 1.19.3 (Elsevier Inc., New York, NY, USA). One author then removed duplicates both electronically and manually. A copy of the database was then provided to two authors for duplicate screening. The two authors selected all studies, independent of each other. The full report for each article was obtained for all titles and abstracts that appeared to meet the inclusion criteria or where there was any uncertainty. Reasons for exclusion were coded as one or more of the following: (1) duplicates (2) missing or incomplete descriptive statistics (3) inappropriate research design (4) language different to English (5) abstracts only and (6) animal model. Upon completion, the two authors met and reviewed their selections. Given the small number of studies selected, discrepancies were reached by consensus. Based on the final number of studies to be included, the overall precision of the searches was calculated by dividing the number of studies included by the total number of studies screened after removing duplicates. The number needed to read (NNR) was then calculated as the inverse of the precision 49 .
Data extraction. Titles and/or abstracts of studies retrieved using the search strategy and those from additional sources were screened independently by two review authors (PCGS and AJM) to identify studies that potentially met the inclusion criteria outlined above. The full text of these potentially eligible studies was retrieved and independently assessed for eligibility by two review team members. Any disagreement between them over the eligibility of particular studies was resolved through discussion with a third reviewer (IR). , New York, NY, USA) and exclusion reasons were recorded. Data were exported to a standardized, pre-piloted Excel spreadsheet used to extract data from the included studies for assessment of study quality and evidence synthesis. The extracted information included publication year, participant demographics and baseline characteristics (e.g., gender, age, cardiorespiratory fitness level), details of the intervention (e.g., exercise frequency, intensity, duration, session duration, total duration of the intervention, dropouts) and control conditions, outcomes (i.e., serum and plasma BDNF) (mean and standard deviation). Two review authors extracted data independently and discrepancies were identified and resolved through discussion with a third author. Missing data were requested from study authors.
Primary outcome. The primary outcome was the change in peripheral BDNF concentration between control and experimental conditions (i.e., repeated measures design) or groups (i.e., independent group design). It is worth noting the first post-exercise BDNF measure was considered for analysis.

Risk of bias assessment in individual studies.
Two review authors independently assessed the risk of bias in included studies by using the Tool for the Assessment of Study Quality and Reporting in Exercise (TES-TEX) 50 . The TESTEX is a 12-item (5 points for study quality and 7 points for reporting) and 15-point scale (5 points for study quality and 10 points for reporting) developed to facilitate a comprehensive review of exercise training trials. Disagreements between the review authors over the risk of bias in particular studies were resolved by discussion, with involvement of a third review author where necessary.
Data synthesis and calculation of effect sizes. The effect size (ES) was calculated as the difference between means according to the methodology proposed by Borenstein, Hedges, Higgins, and Rothstein (2009) 51 . For the calculation, the initial score (pre-test) of BDNF was compared with the final score (post-test) after an intervention (exercise). The ES was subsequently adjusted to take into account the bias introduced by small samples 52 . For the analysis, the random effects model was used, which assumes that ESs vary between studies 51,53 . In this study, ES was interpreted as trivial (0 to 0.19), small (0.20 to 0.49), moderate (0.50 to 0.79) and large (≥ 0.80) 54 . ANOVA and independent samples t-test were used to determine mean ES differences between categorical moderator variables.
The degree of heterogeneity of the studies was analyzed through Cochran's Q test 57 and the degree of consistency between the studies was calculated through the I 2 test 58 . The I 2 statistic ranges from 0 to 100%, and is interpreted as low (≤ 25%), moderate (26-74%) and high (≥ 75%) 58 . The effect of the studies with small samples was determined by the Doi plot and LFK index 55 . LFK index values outside the interval between −1 and + 1 are considered consistent with asymmetry (i.e. publication bias) 59 . An α level ≤ 0.05% and 95% confidence intervals (95% CI) that did not include zero (0) were considered to represent statistically significant small-study effects.
Software used for data synthesis. All data were analyzed using IBM SPSS Statistics for Windows, Ver- activity monitoring in control groups (73%), (11) relative exercise intensity remained constant (10%), and (12) exercise volume and energy expenditure (93%). Given the inability to truly blind participants in exercise intervention trials, all studies (100%) were considered to be at a high risk of bias for the categories "allocation concealment" and "blinding of assessor". In addition, 87% of the studies did not report adverse effects and 13% of the studies reported adherence to exercise interventions.
Categorical moderator variable analysis on acute and chronic exercise interventions showed that there is no statistically significant subgroup effect for fitness level, type of training, and medium (serum vs. plasma) for acute and chronic IT ( Table 2). There was a strong tendency (p = 0.052) for gender in chronic IT analysis; a higher ES was observed in females during chronic IT intervention compared with males ( Table 2). For continuous moderators, no significant correlations were found between ES and age for acute (r = -0.18, p = 0.534) and chronic (r = -0.09, p = 0.805) exercise. No significant correlations were found between ES and sample size for acute (r = -0.20, p = 0.501) and chronic (r = -0.45, p = 0.192) exercise.

Discussion
The present study was designed to systematically-review and meta-analyze the effects of acute and chronic IT on circulating BDNF concentration in young adults. Overall, acute and chronic IT increased peripheral BDNF concentration. In chronic IT, females showed greater increases in BDNF compared with males. Finally, the study showed that the fitness levels did not regulate the BDNF response after IT, at least in the studied population (apparently healthy young adults).
The data of the current study are in agreement with the previous report focused on the impact of aerobic exercise on peripheral BDNF 60 . In Dinoff 's study, the exercise protocols were longer than the interventions analyzed in the current meta-analysis (≤ 30 min/session). This finding suggests that IT is an effective treatment to improve brain health with more time efficiency than MICT. The latter condition is concordant with peripheral adaptations induced by IT (e.g., oxidative capacity in muscle, cardiometabolic markers) 31,37,61 .
IT is characterized by lactate accumulation in blood 45,[62][63][64] . Studies in rodents have demonstrated that blood lactate (BLa) produced during exercise reaches the brain and enhances expression of genes associated with cognition (i.e. Bdnf) 65,66 . Although in humans this response has not been completely demonstrated, authors suggested a similar effect of BLa in brain 45,63,67,68 . Resulting in diverse improvements in executive function 63 . Unfortunately, in the current meta-analysis, there were not enough studies that reported blood lactate changes; consequently, it was not possible to run meta-regressions to identify the role of this metabolite in the BDNF response.
Non-statistical differences were found among BDNF changes in plasma and serum (Table 2). While some studies did not find statistical differences between BDNF changes in serum and plasma following physical exercise 27 , others reported significant changes in circulating BDNF in plasma compared with serum 60 . In previous studies, aerobic, strength, and concurrent training were analyzed, whereas, in the current meta-analysis, IT interventions were examined. Circulating BDNF changes are sensitive to training modality 45,69 ; therefore, it was not possible to compare our data with other systematic and meta-analytic works 27,60 .
In serum, BDNF concentration is > 50 fold higher than plasma 18,70,71 . In the periphery, platelets store BDNF; therefore, these cells are considered the major reservoir of circulating BDNF 71,72 . Once activated, platelets release BDNF 18,71 . This process is considered the main mechanism to explain differences between serum and plasma concentrations 73,74 . The evidence suggests that chronic training improves the capacity of platelets to release BDNF 18,75 . Concerning this, we did not discard that the length of interventions examined in the current study was insufficient to modify the platelet's capacity in the BDNF secretion; thus, further studies are necessary to elucidate this hypothesis. In addition, it is known that IT is an exercise modality that increases muscle damage 76 . We www.nature.com/scientificreports/ believe that this condition could be present in the participants and consequently will generate platelet activation 71 , releasing BDNF to repair muscle injuries 77 . This physiological response might explain the lack of differences among the BDNF changes in serum and plasma (Table 2). Finally, we did not discard that the small numbers of studies included in the current meta-analysys can explain the lack of differences among the blood mediums. Furthermore, it is worth noting that plasma volume (PV) changes should be considered in studies that assess the impact of exercise on biomarkers such as neurotrophins. It is known that exercise modifies PV 78 , which can increase biomarker concentrations. Thus, the results of studies neglecting to measure PV changes should be viewed with caution [78][79][80] . In one study examined, there was no effect on circulating BDNF with IT when PV was not adjusted. In contrast, DiBatista et al. showed that IT increased BDNF levels following PV adjustment. Finally, in work conducted by Reycraft and colleagues,PV was not adjusted and the authors reported a significant effect of IT on BDNF. The results of these studies show that PV should be considered when evaluating the effects of IT on circulating BDNF levels. Moreover, studies where BDNF was assessed in plasma were fewer than studies where the biomarker was measured in serum. Thus, unequal distribution can be a confounding variable to find statistical differences.
Similarly, to the medium, fitness level did not significantly affect the BDNF response to IT. These findings are contrary to previous reports 18,44,60,75,81 . Despite the established negative correlation between fitness level and BDNF response during exhaustive or aerobic exercise 44,60 , biochemical and physiological mechanisms are not fully understood. One hypothesis suggests that well-trained participants have higher BDNF receptor levels in peripheral organs (e.g., skeletal muscle) which could attenuate circulating BDNF changes during exercise 82 . Once it activates the peripheral TrkB receptor, BDNF participates in the repair of skeletal muscle 77 . As indicated above, Figure 5. Interval training increases circulating BDNF levels in healthy adults (upper right). During this response, the brain (hippocampal region) seems be the main BDNF source; nevertheless, other tissues function as BDNF synthesizers. The mechanism of activation during IT has not elucidated yet (above right). In brain, BDNF synthesis is activated by an increase of calcium (Ca2+) concentrations in the cytosol. Inside neurons, Ca2+ activates calmodulin dependent kinase II (CaMKII), triggering activation of the MAPK/ERK/MSK cascade resulting in an increase in the expression and phosphorylation of cAMP response element-binding protein (CREB). CREB initiates BDNF transcription resulting in increased BDNF synthesis and release (left). Once secreted, the neurotrophin regulates molecular mechanisms associated with neuronal growth, cognition, and neuron survival (above left). Finally, scientific evidence suggests that other circulating molecules such as lactate and estrogen enhance BDNF synthesis in brain (center). The putative mechanism indicate that lactate increases calcium current in the neurons, and estrogens activates nuclear estrogen receptors and membrane estrogen receptors that enhance the BDNF synthesis. Figure made with adobe illustrator cs6. https:// www. adobe. com/ produ cts/ illus trator/ free-trial-downl oad. html. Figure conceived and  www.nature.com/scientificreports/ IT induces muscle damage in well-trained and untrained participants 76 ; therefore, we did not discard that the low peripheral BDNF levels were induced by muscular damage after IT. That condition could partially explain the lack of significant differences in BDNF changes between athletes and untrained participants ( Table 2). Another hypothesis suggests that trained participants show better cognitive performance than sedentary people 44,83 . Indeed, athletes and well-trained individuals have more efficient uptake and utilization of BDNF which has been shown to improve neural plasticity and improve performance in cognitive tasks compared to untrained participants 44,83 . The extensive utilization of BDNF in brain reflects a lower peripheral BDNF in athletes and well-trained participants with respect to untrained people 44,83 . Therefore, we do not discard the possibility that active participants show a high capacity to uptake BDNF in brain after IT compared with sedentary participants ( Table 2). In contrast, sedentary participants have lower synthesis and release of BDNF. Both conditions combined resulted in a non-significant statistical effect among active and sedentary ( Table 2). The null findings observed in sedentary and active participants after acute IT can be explained by stress hormone activity. Specifically, IT is perceived as difficult and vigorous in well-trained and untrained population 45,76,[84][85][86] . In agreement with this, IT increases systemic cortisol concentrations in athletes and untrained participants 47,[87][88][89] . Cortisol is a hormone that decreases BDNF synthesis 90 . Therefore, higher cortisol levels could be present in the participant (sedentary, active, and athlete participants) after IT, reducing differences in BDNF changes (Table 2). Finally, we do not exclude the possibility that the small numbers of studies included in the current meta-analyses can explain the lack of differences among fitness levels.
We found a high ES (strong tendency) for females compared with males; a difference shown principally in chronic IT (Table 2). This may be explained by the role of steroid hormones since it is known the positive effect of 17β estradiol on BDNF synthesis in the brain [91][92][93][94][95] . The estrogen hormone concentrations change during the menstrual cycle 96 ; particularly, high levels of estrogen are found during the late follicular phase 97 . In the studies analyzed in the current meta-analysis, the menstrual cycle was not coded; therefore, we do not discard the possibility that some of the blood collection made in females was performed during the follicular phase, resulting in an enhancement effect of estrogen to IT impact on BDNF changes compared to males. Additionally, as discussed previously, platelets store and release BDNF 71,72 . In this sense, classic and emerging studies show that women have higher platelet content than men [98][99][100] . In light of this, we do not discard that platelet count could contribute to a higher BDNF response in women compared with men (Table 2). Additionally, authors have previously suggested that skeletal muscle uptake BDNF; once captured the neurotrophin regulates metabolic and neuromuscular responses [101][102][103] . In females, muscle mass is lower than males [104][105][106] . Therefore, it is possible that differences skeletal muscle mass among sex, can explain the larger ES in women compared with men ( Table 2).
The current meta-analysis highlights that IT is an effective strategy to increase peripheral BDNF concentrations in young healthy adults. Our findings are in agreement with prior meta-analysis focused on assessing the impact of physical exercise (e.g., aerobic and strength exercise) on circulating BDNF in young adult and healthy population 27,60,107 (Fig. 5). This finding adds relevant information to previous studies reporting a positive impact of IT on fitness levels [108][109][110] , and hemodynamic variablesy 111 . Therefore, the state of the art, based on quantitative analysis suggests that IT may be considered an adequate physical exercise modality to strengthen the health (brain and peripheral physiological functions) in an apparently healthy young adult population.

Data availability
The data that support the findings of this study are available from the corresponding author on request.