This was a 22-week randomized controlled trial. Physical and cognitive assessments were conducted over a 4-week period: 2 weeks before the intervention and 2 weeks after the intervention, to collect pre- and post-intervention data. Additionally, 2 weeks were dedicated to familiarizing the participants with the exercises, and 16 weeks were allocated for the implementation of the training protocols (Figure 1).

The participants were recruited from the community using leaflets and flyers distributed around the Federal University of Sergipe, through a non-probability sampling method (Figure 2). To detect between-group differences with an alpha level of 0.05 and assuming a large effect size, we would need a total of 45 participants in our study to achieve a power of 0.80. Notably, we considered a 20% sample loss during the intervention, totaling 54 participants. We randomized the participants based on the residual value of the means of all performed tests, which were organized in ascending order for every three values using Microsoft Excel software.

The selection process for participants followed several inclusion criteria: 1) female sex; 2) physically independence; 3) being literate; 4) not involved in any hormone replacement therapy; 5) scoring >8 and <26 points on the Montreal Cognitive Assessment (MoCA) (Malek-Ah et al., 2024); 6) reporting complaints of memory problems (Gifford et al., 2014); and 7) without diagnosed neurological diseases (dementia, Parkinson’s disease, Alzheimer’s disease, multiple sclerosis, and depression), recent surgeries, vision, hearing, musculoskeletal, and cardiorespiratory problems that would prevent the practice of high-intensity physical exercise. Participants with an attendance rate below 70% in the training sessions and those who did not attend the final assessment were excluded from the study analyses.

After the application of the eligibility criteria, the participants signed the Free and Informed Consent Form. This study was conducted in accordance with the Declaration of Helsinki and approved by the institution’s Research Ethics Committee (CAAE: 76026223.4.0000.5546; No. 6,598,301). Furthermore, it is registered in the Brazilian Registry of Clinical Trials under protocol RBR-89svjw3 (available at https://ensaiosclinicos.gov.br/rg/RBR-89svjw3).

Three weekly sessions were conducted, each lasting approximately 50 min, over a period of 16 weeks, totaling 48 trainings sessions, with a minimum interval of 48 h between sessions. Both training protocols were supervised by experienced physical training professionals for older adults. There was one professional for every five participants to ensure safety and correct exercise execution.

Initially, the participants underwent an eligibility assessment, and information such as age, body weight, height, address, motivation, medication use, and chronic diseases was recorded. For the cognitive and physical tests, all measurements were performed by trained and experienced professionals in the data collection procedures, which always occurred in the morning period. The assessors were blinded to the physical training protocols previously performed by the participants.

The initial sample consisted of 89 older women (mean age = 67.1 ± 4.7 years; BMI = 29.29 ± 4.94 kg/m²) at risk for MCI (MoCA: 19.32 ± 3.73 points), independent in activities of daily living (IADL: 20.61 ± 0.86 points), and with a low indicative risk of depression (GDS-15: 2.46 ± 1.39). Of these, 68 completed all stages of the intervention, and the experimental groups showed a 91% adherence rate to the training sessions (Table 3).

In FT, the volume-load demonstrates that the training load was progressive throughout the 16 weeks (Figure 3A). Conversely, the monitoring of HRres demonstrates that an intensity between 50% and 80% was maintained during the AT sessions (Figure 3B).

Following the 16-week intervention period, we observed a significant group*time interaction for cognitive status (MoCA) (F (2, 65) = 3.16; η² = 0.010; p = 0.035), with no significant group effect (F (2, 65) = 0.910; η² = 0.020; p = 0.408) but a significant time effect (F (2, 65) = 89.68; η² = 0.143; p ≤ 0.001). Upon comparing time points, we detected a large effect size increase in both the FT and AT groups compared to baseline values (FT: d = 0.99; p ≤ 0.001/AT: d = 0.97; p ≤ 0.001). However, the CG showed no significant difference despite a moderate effect size (d = 0.51; p > 0.05) (Figure 4A).

Regarding the peripheral quantification of BDNF, we found a significant group*time interaction (F (2, 37) = 9.07; η² = 0.102; p ≤ 0.001), with no significant group effect (F (2, 37) = 0.518; η² = 0.018; p = 0.606) and no significant time effect (F (2, 37) = 0.079; η² = 0.002; p = 0.790). Upon further exploration of the results, we only detected a large effect size increase in the FT group compared to baseline values (d = 0.95; p < 0.011). Furthermore, when comparing post-test time points, we observed a difference with a large effect size between the FT and CG groups (d = 0.96; p = 0.026) (Figure 4B).

Regarding cognitive function, for which the analyzed variables may influence performance, we observed no significant differences between the groups for short-term memory (DSF), working memory (DSB), semantic memory (FV_QAni), and executive function (FV_QAgru). However, FT showed an improvement with a moderate to large effect size compared to baseline values in semantic memory (d = 0.95; p < 0.001) and executive function (d = 0.63; p = 0.043). In contrast, AT demonstrated an improvement with a large effect size compared to baseline values only for semantic memory (d = 0.97; p < 0.001). The CG showed no adaptations for any of the cognitive variables (Table 4).

Regarding cardiorespiratory fitness (6MWT), we observed differences between the groups (FT/CG: p = 0.008; AT/CG: p = 0.041). When compared to baseline, we found an increase in the distance covered for both the FT and AT groups with a moderate effect size (FT: d = 0.66; p = 0.002/AT: d = 0.56; p = 0.036). In the CG, we observed a non-significant reduction with a small effect size (d = 0.35; p = 0.464). Regarding upper limb strength (ULS), we observed no significant differences between the groups. However, when considering changes from baseline, only the FT group showed an increase in handgrip strength in kilogram-force with a moderate effect size (d = 0.64; p ≤ 0.001) (Table 5).

For the tests assessing functional fitness involving the stand up ability to rise from a chair (FTSST) and gait speed (GS), we observed a difference between the groups only for GS (FT/CG: d = 1.25; p = 0.003). When compared to baseline values, both intervention groups improved the time to complete the task in the FTSST (FT: d = 0.61; p ≤ 0.001/AT: d = 0.41; p = 0.021) and GS (FT: d = 0.97; p ≤ 0.001/AT: d = 0.59; p ≤ 0.001), with effect sizes ranging from small to large. The CG showed no adaptations following the intervention (Table 5).

Consistently, regarding walking ability, TUG and 10MWT, we observed no significant differences between the groups. When compared to baseline values, both intervention groups reduced the time to complete the task in the TUG (FT: d = 0.69; p ≤ 0.001/AT: d = 0.54; p = 0.022) and similarly for the 10MWT (FT: d = 0.56; p ≤ 0.001/AT: d = 0.67; p = 0.011), with moderate effect sizes. The CG showed no adaptations following the intervention (Table 5).

Finally, regarding the tests related to dexterity, GJST and DTTS, we observed differences between the groups only for GJST (FT/CG: d = 1.04; p = 0.007) with a large effect size. When compared to baseline values, only the FT group reduced the time for task completion in the GJST (d = 0.58; p = 0.004) and similarly for the DTTS (d = 0.69; p ≤ 0.001), both with a moderate effect size. The AT and CG showed no adaptations following the intervention (Table 5).

The present study aimed to evaluate the effects of 16 weeks of FT and AT on cognitive function, functional fitness, and serum BDNF levels in older women with mild cognitive impairment. Our main finding was that both training protocols promoted consistent improvements in cognitive state and semantic memory. However, only FT improved executive function and increased peripheral BDNF concentration. Regarding functional fitness, we observed larger effect sizes in the ability to stand up and in tasks involving manual dexterity, as well as specifically in handgrip strength, thus confirming our initial hypothesis.

The multisystem adaptations resulting from the combination of multicomponent stimuli with movement specificity for daily activities, together with the individualization and progression of the training load, have been previously evidenced by our group (Pantoja-Cardoso et al., 2023; Vasconcelos et al., 2022; Carrasco-Poyatos et al., 2019). The present study provides a consistent investigation into the effects of physical training on cognitive function and quantification of the training load, which suggests the absence of adaptive stagnation throughout the intervention.

Regarding cognitive status, significant effects with a large effect size (FT: d = 0.99; AT: d = 0.97) were observed for both training protocols. These findings likely stem from the positive adaptation of cardiorespiratory capacity, which is directly linked to enhanced cerebral vascularization (Ferrer-Uris et al., 2022; Dupuy et al., 2015; Tsai et al., 2014). This adaptation promotes increased connectivity between brain regions (Holzschneider et al., 2012), enhancing learning (Ferrer-Uris et al., 2022) and attenuate cognitive decline (Hoffmann et al., 2016). Our results are consistent with previous studies in older adults with MCI (Vega-Ávila et al., 2022; Yong et al., 2021). Furthermore, improvements caused by the FT protocol may be related to visuospatial stimuli (Lee et al., 2024) and the motor complexity of the exercises (Fontes et al., 2020). These stimuli may positively favor the availability of oxygen and nutrients to neural cells (Crichton et al., 2014), with continuous activation of areas related to movement planning and execution (Lee et al., 2024).

Executive function is crucial for problem-solving and the planning of cognitive or motor tasks. It comprises working memory, cognitive flexibility, and inhibitory control, all of which independently contribute to executive function (Diamond, 2013). Only the FT protocol showed significant effects on executive function with a moderate effect size (FT: d = 0.63), which can be attributed to the motor demands of the neuromuscular-dominant block, combined with the high volume of multi-joint and multi-planar exercises. This requires more planning when compared to the rhythmic movements used in the AT protocol with a small effect size (AT: d = 0.46) (La Scala Teixeira et al., 2019). Consequently, greater activation of areas responsible for motor planning, such as the supplementary motor area, premotor cortex, cerebellum, and basal ganglia, is expected (Vints et al., 2024; Kušleikien et al., 2025; Barha et al., 2022).

Specifically, regarding short-term memory (DSF) (FT: d = 0.05; AT: d = 0.43) and working memory (DSB) (FT: d = 0.43; AT: d = 0.39), no significant adaptations were observed for any of the groups, with effect sizes ranging from trivial to small. The ineffectiveness of both FT and AT may stem from the absence of specific short-term memory stimuli (2–10 min) for subsequent recall and manipulation. Possibly, the addition of motor-cognitive dual tasks could provide effective adaptations for working memory (Norouzi et al., 2019). This is because the inclusion of dual tasks will generate greater activation of the prefrontal cortex, which is related to improved short-term and working memory performance (Marvel et al., 2019). Still, some studies that did not use dual tasks found effects with longer intervention durations, as well as the short duration of the current intervention, given that these effects have been observed in studies lasting 24 weeks (Spruill, 2010; Zhidong et al., 2021).

In terms of semantic memory, both groups showed substantial improvement. These adaptations may be attributed to the moderate to high intensity of the sessions, which suggests an increase in cortical vascularization (O’Callaghan et al., 2020), followed by greater excitability of prefrontal and temporal regions (Nocera et al., 2017) particularly the hippocampus (Rivas-Campo et al., 2023), contributing to the reorganization of neural circuits (Smith et al., 2013). Both findings are consistent with the existing literature, demonstrating that multicomponent and aerobic stimuli can promote adaptations in semantic memory (Rivas-Campo et al., 2023).

Pertaining to peripheral BDNF concentration, only the FT protocol showed significant increases compared to both baseline values, with a large effect size (d = 0.95) and the CG (d = 0.96). BDNF is essential for neural plasticity and is highly expressed in brain areas involved in cognitive processing. BDNF improves memory storage and dendritic spine plasticity. The action of BDNF is primarily mediated by the tyrosine kinase receptor B (TrkB) and initiates signaling pathways that promote neuroprotection (Wang et al., 2020; Zhao et al., 2022). Cognitive function and BDNF are modulated by the combination of cardiorespiratory, neuromuscular, and motor stimuli (Venegas-Sanabria et al., 2022; Yan et al., 2023).

Furthermore, at the peripheral level, maintaining the motor complexity in multi-joint and multi-planar exercises likely exerts greater neuromuscular activation, improving stability between the neuromuscular junction and the motor endplate (Sepulveda-Loyola et al., 2025). These alterations may favor an increase in BDNF synthesis within skeletal muscle and its subsequent release into the bloodstream (Imaizumi et al., 2025).

It has been demonstrated that aerobic exercise significantly increases serum BDNF levels, and a significant relationship between BDNF levels and physical activity levels has been reported (Coelho et al., 2014). The lack of response in AT may be related to the maintenance of the training load throughout the intervention period, as progressively applied intensity in aerobic exercise influences the modulation of BDNF concentration (Huang et al., 2024; Babaei et al., 2013). Additionally, genetic factors that can negatively impact BDNF synthesis, such as the Val66Met gene polymorphism and APOEε4 (ε4), were not identified in the groups (Enette et al., 2020; Lemos et al., 2016; Allard et al., 2017). This may have affected the synthesis and increased peripheral BDNF concentration in AT.

For cardiorespiratory fitness, both groups showed moderate improvement (FT: d = 0.66; AT: d = 0.56). In fact, it is widely known that AT applied at moderate to high intensity promotes cardiovascular and respiratory adaptations that favor the enhancement of this physical capacity (Alemayehu and Teferi, 2023). While in FT, the use of sprints, changes of direction, and other movements performed at high intensity (de Resende-Neto et al., 2021; Rivas-Campo et al., 2023), combined with the circuit exercise structure (Gbiri and Amusa, 2020), may lead to positive changes in central and peripheral mechanisms that favor oxygen transport and utilization.

Concerning the ability to stand up, assessed by the FRT (FT: d = 0.97; AT: d = 0.59) and FTSST tests (FT: d = 0.61; AT: d = 0.41), both the AT and FT protocols showed significant improvements with effect sizes ranging from small to large. The rhythmic and high-intensity interval part applied in AT may stimulate muscle power and motor coordination, thus leading to improvements in this ability (Li et al., 2024). In addition, FT is predominantly neuromuscular, with training session parts focused on stimulating power and muscle strength of the lower limbs, in addition to including movements with a motor pattern similar to the functional tests analyzed (Vasconcelos et al., 2020; Carrasco-Poyatos et al., 2019).

As for walking ability, mimicked by the TUG (FT: d = 0.69; AT: d = 0.54) and 10MWT (FT: d = 0.56; AT: d = 0.67) tests, both the AT and FT protocols showed significant improvements with moderate effect sizes. The stimuli targeting velocity and dynamic balance during acceleration and deceleration actions, present in both protocols, stimulate muscle power (Li et al., 2024; Carrasco-Poyatos et al., 2019). It is already known that there is a strong correlation between lower limb power and dynamic balance, as well as usual gait speed (Yee et al., 2021). These stimuli constantly perturb the postural control system and activate stabilizing muscles, promoting adaptations in dynamic balance (Shahtahmassebi et al., 2019).

With respect dexterity, which was analyzed using the GJST (FT: d = 0.58; AT: d = 0.51) and DTTS (FT: d = 0.69; AT: d = 0.23) tests, only the FT protocol showed a significant improvement with moderate effect sizes for both tests. The absence of significant adaptation in the AT group may be due to the lack of task-specific exercises involving the upper limbs. FT presents a result similar to other studies, in which the quantity of exercises targeting the upper limbs, applied with pushing and pulling patterns, may favor dexterity (Aragão-Santos et al., 2024). Other studies involving strength training have highlighted that performing movements with large amplitudes favors gains in joint mobility (Carrasco-Poyatos et al., 2019; DE Matos et al., 2017).

Finally, regarding Handgrip Strength, which is considered a strong predictor of mortality and disability, a significant improvement was observed only for FT (d = 0.64) with moderate effect sizes (Taekema et al., 2010). These adaptations in FT may be attributed to the manipulation of equipment (kettlebell, Bulgarian Bag, Weight Plates) during the strength exercises. This result is consistent with the current literature (Aragão-Santos et al., 2019).

Our findings provide important contributions regarding the effects of different interventions on functional fitness, with particularities not yet evidenced in the scientific community. One limitation of this study is the lack of intensity progression in AT, which may have influenced the absence of significant changes in BDNF and could potentially yield greater effects on functional fitness. However, the percentage of HRR was monitored to maintain moderate to high intensity during the sessions, as suggested in other studies to promote cardiorespiratory adaptations that are associated with improved cognitive function. Another limitation is that the ICC was not calculated; however, according to similar studies of our research group, the tests used exhibit sensitivity above 75% and an ICC greater than 0.80. Finally, the absence of intention-to-treat (ITT) analysis is considered a limitation of this study.

Based on our findings, future research should consider other cognitive subdomains, utilizing more sensitive and specific tests, and also verify the activation of different brain areas during interventions or assessments. We recommend comparisons between men and women, with and without mild cognitive impairment, regarding cognitive and physical performance, verifying the impact of the interventions on quality of life. Furthermore, it is essential to investigate the integration of cognitive and motor stimuli over time, analyzing the residual effect of physical exercise on cognition.

Sixteen weeks of FT and AT promoted improvements in cognitive status and semantic memory, but only FT enhanced overall executive function and BDNF serum levels. Regarding functional fitness, both training modalities improved cardiorespiratory fitness, as well as the ability to rise and walking ability. However, only FT positively altered dexterity and handgrip strength in older women with mild cognitive impairment. These results highlight the importance of structured, targeted training programs that are low-cost and offer several benefits for this specific population, making both protocols feasible for implementation in community or clinical settings that require physical training programs to enhance the autonomy and quality of life of the aging population. Thus, this study contributes evidence supporting non-pharmacological interventions aimed at mitigating cognitive and functional decline in older adults, specifically in women, who are disproportionately affected by MCI.