The study, conducted by Yixuan Wu and Yin Wu of Soochow University, is the first to simultaneously compare the cognitive effects of specific exercise modalities in older adults with mild cognitive impairment, a transitional state between normal aging and dementia that affects a surging share of the world’s aging population. Because no disease-modifying drugs currently exist for dementia, identifying interventions that work during this early window has become a global public health priority, and physical exercise has emerged as the cornerstone of prevention efforts.

To build their treatment hierarchy, the researchers sifted through 6,737 records from PubMed, Embase, Web of Science, and the Cochrane Library, ultimately including 16 randomized controlled trials involving 1,208 participants with a mean age over 65. The trials, conducted across China, South Korea, Pakistan, Brazil, Turkey, Poland, and Japan, lasted 12 to 24 weeks with two to three sessions per week. Global cognition was measured with standardized instruments, the Montreal Cognitive Assessment or the Mini-Mental State Examination, allowing effects to be pooled across studies using standardized mean differences.

Network meta-analysis is a statistical technique that goes beyond traditional pairwise comparisons. Rather than asking only whether exercise A beats a control group, it synthesizes both direct head-to-head evidence and indirect evidence across an entire network of interventions, even when two treatments have never been tested against each other in the same trial. The researchers used a frequentist framework in Stata, checked for consistency between direct and indirect evidence with the design-by-treatment interaction model, and ranked interventions using SUCRA values, which express the probability that a treatment is the best option.

The results were striking. Compared with an active control group, combined exercise, which integrates physical exertion with concurrent cognitive tasks, produced the largest improvement in cognitive scores, with a standardized mean difference of 3.78 and a 68.0 percent probability of being the single most effective intervention. Multicomponent exercise, which blends aerobic and resistance elements, showed a significant effect of 1.71, and plain aerobic exercise a significant effect of 1.47. Tai Chi and coordination training showed positive trends but their confidence intervals crossed the line of no effect, likely because those arms contained fewer participants.

The ranking probabilities tell the story in sharper relief. Tai Chi came second with a 15.9 percent probability of being the best treatment, followed by coordination training at 13.3 percent, while passive and active controls consistently ranked at the bottom. The pattern points to a unifying principle: interventions demanding high cognitive and neural engagement, whether through dual-task loads or complex motor sequencing, confer superior benefits over those requiring purely physical exertion.

The mechanism, the authors argue, is neuroplasticity operating on two fronts at once. Physical exertion alone increases cerebral blood flow and upregulates brain-derived neurotrophic factor, a protein that supports the survival and growth of neurons. Adding a simultaneous cognitive load forces the brain to allocate attention and executive resources efficiently, and this synergistic stimulus appears to drive deeper structural and functional adaptations in the prefrontal cortex and hippocampus, the very regions most affected in mild cognitive impairment. In effect, synapses generated by physical activity are immediately pressed into service by cognitive demand, maximizing cognitive reserve.

Tai Chi’s strong showing provides quantitative backing for traditional mind-body practices. Unlike the repetitive, automated motor patterns of treadmill walking, Tai Chi demands sustained focused attention, sequential memorization of intricate movement routines, and deep mindful breathing, effectively embedding a physical-cognitive dual task within a single activity. The authors also highlight a second pathway: mind-body exercises modulate the autonomic nervous system, downregulating sympathetic hyperactivity and reducing cortisol, a stress hormone known to accelerate hippocampal atrophy. Tai Chi may thus stimulate neural networks while simultaneously shielding the brain from stress-induced damage.

The analysis is not without caveats. Many included trials failed to report precise exercise intensity data, such as target heart rate zones or perceived exertion ratings, preventing a definitive dose-response analysis. Some network nodes, notably coordination training, contained few direct comparisons, producing wider confidence intervals, and the studies varied in how faithfully interventions were delivered and how strictly participants adhered to them. The authors call for larger multi-arm trials directly comparing combined training with Tai Chi and aerobic exercise, and for stricter adherence to reporting standards such as the CERT consensus.

Still, the consistency checks were reassuring: no significant global inconsistency emerged between direct and indirect evidence, and a comparison-adjusted funnel plot showed no evidence of publication bias. For clinicians, physical therapists, and the growing population of older adults worried about memory, the practical message is clear. A brisk walk is good, but a workout that also makes you think, whether a dual-task class, a coordination drill, or a Tai Chi session, may be the most powerful prescription available for slowing cognitive decline before it becomes dementia.

Subject of Research: Comparative efficacy of exercise modalities for global cognition in older adults with mild cognitive impairment

Article Title: Comparative efficacy of exercise modalities for mild cognitive impairment in older adults: A network meta-analysis

Article References: Wu, Y., & Wu, Y. (2026). Comparative efficacy of exercise modalities for mild cognitive impairment in older adults: A network meta-analysis. iScience, 29(11), Article 117561. https://doi.org/10.1016/j.isci.2026.117561

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117561

Keywords: mild cognitive impairment, network meta-analysis, combined exercise, Tai Chi, aerobic exercise, coordination training, neuroplasticity, BDNF, dual-task training, dementia prevention, cognitive reserve, older adults

Rotator cuff tears are a leading cause of shoulder pain and disability, and while arthroscopic rotator cuff repair can surgically restore the torn tendon to bone, the operation is only half the story. Recovery depends heavily on the quality, timing and progression of rehabilitation. Patients routinely face weeks of protective immobilization followed by gradual loading, and along the way they may contend with joint stiffness, scar adhesion, muscle atrophy and persistent pain. What is more, successful recovery is not simply a matter of tissue healing. Proprioception — the body’s sense of joint position — is often impaired after a cuff tear, and psychological factors such as kinesiophobia, the fear of movement, can quietly sabotage a patient’s willingness to engage fully in therapy. The research team, led by Liang Chen and Peng Chen of the Department of Sports Medicine, together with Zi-Hao Jia of Fuzhou University, set out to test whether cognitive-motor dual-task training could address these intertwined physical and neurological barriers simultaneously.

The rationale comes from dual-task theory, which holds that when people perform a motor task and a cognitive task at the same time, the two compete for a limited pool of attentional resources. Training under this divided-attention condition is thought to improve the efficiency of attentional allocation, reduce interference costs, and help transfer motor control from the clinic to real-world settings — where movements are rarely performed with full, undivided attention. In daily life, after all, we reach, lift and carry while conversing, navigating, or planning our next action. Dual-task training has already shown promise in older adults, stroke patients and people recovering from hip replacement or elbow fracture surgery, but rigorous randomized evidence in patients recovering from rotator cuff repair had been lacking.

To fill that gap, the team enrolled 72 adults with full-thickness rotator cuff tears involving at least the supraspinatus tendon who had undergone arthroscopic repair. Following surgery, participants were randomly assigned in equal numbers to a control group or a dual-task group. Both groups received the same conventional rehabilitation program, including wall-climbing exercises in the coronal and sagittal planes, back extension work, and closed-chain exercises such as door-pushing. The dual-task group, however, performed a serial subtraction-by-3 task — counting backward aloud from 100 in steps of three — throughout their exercise sets. Training began at postoperative week 6 and ran for 12 weeks, with two supervised sessions per week. The trial was prospectively registered with the Chinese Clinical Trial Registry, and all outcome assessments were carried out by a physical therapist blinded to group assignment, since neither patients nor treating therapists could be blinded to the intervention.

The intervention was carefully phased to match the biology of tendon healing. During postoperative weeks 6 to 12, the cognitive task was layered onto active and active-assisted shoulder exercises and low-load scapular control training. From weeks 12 to 18, dual-task training was integrated into light-resistance strengthening. Progression was criterion-based rather than purely time-based: exercises advanced only when patients could complete the prescribed repetitions with good movement quality, proper scapulohumeral coordination, and no compensatory patterns or symptom flare-ups. Therapists monitored cognitive engagement closely, calculating arithmetic accuracy in each session; mean accuracy across the trial was 92.0 percent, indicating that patients genuinely sustained the mental task rather than letting it lapse. Notably, every single participant completed the 12-week program, an adherence figure that speaks to the feasibility of the approach.

The outcomes were assessed at postoperative weeks 6 and 18 using an unusually broad battery. The primary clinical measure was the Constant-Murley Score, a composite index spanning pain, activities of daily living, range of motion and muscle strength. The primary neural measure was far more novel: an electroencephalography-derived index the authors call β-ΔwPLINBS. To compute it, patients wore a 64-channel EEG cap while performing shoulder abduction, with signals sampled at 1024 Hz and meticulously preprocessed using bandpass filtering, artifact removal via independent component analysis, and source localization through minimum norm estimation projected onto a standard brain atlas. The researchers then calculated the weighted phase lag index — a measure of functional connectivity between brain regions — and used network-based statistics to identify subnetworks that differed between patients and healthy controls. The β-ΔwPLINBS quantifies how far a patient’s beta-band sensorimotor connectivity deviates from the healthy-control average; a smaller value means the brain network looks more like that of a healthy person. The beta band was chosen deliberately, as beta oscillations are among the most characteristic neural rhythms of the sensorimotor system, intimately involved in motor control and sensory integration.

The results were striking. On the Constant-Murley Score, the dual-task group improved significantly more than controls, with a large between-group effect size of d = 1.93 and a highly significant time-by-group interaction. Pain, measured on the Visual Analog Scale, fell more steeply in the dual-task group (d = 1.56), and kinesiophobia, assessed with the 11-item Tampa Scale, also improved substantially more (d = 0.88). Crucially, the brain followed suit: the dual-task group showed a significantly greater reduction in β-ΔwPLINBS (d = 0.90), meaning their beta-band sensorimotor network connectivity shifted closer to the healthy-control reference than it did in patients doing exercise alone. Not every measure moved in lockstep, however. Absolute angular error, a test of proprioceptive accuracy in which blindfolded patients reproduce a 30-degree abduction target using a wrist-mounted angle sensor, improved in both groups but without a significant added benefit from dual-tasking. The same was true of the DASH questionnaire measuring perceived disability in daily activities.

These divergent findings are, in some ways, the most intellectually interesting part of the study. The authors suggest that dual-task training may act first on central integration — attentional allocation, sensorimotor integration and network-level coordination — rather than on peripheral joint position sense, which may lag behind. Similarly, the improvement in kinesiophobia without a parallel gain in DASH scores hints that the training readily touches psychological and control-related dimensions of recovery, such as confidence in movement, while the translation into patients’ subjective appraisal of complex daily activities may simply require more time. The team also candidly notes methodological caveats: CMS contains performance-based components that could be influenced by the inability to blind participants and therapists, and a single arithmetic task cannot represent the full family of dual-task paradigms, from inhibitory control to visual search. The β-ΔwPLINBS index itself is study-specific, not yet a validated biomarker, and no mediation analyses were performed to confirm a direct neural-to-clinical pathway.

Still, the implications are considerable. If adding nothing more than a subtraction task to standard exercises can meaningfully amplify functional recovery, reduce pain, ease fear of movement and nudge brain networks toward a healthier configuration — all at essentially no cost and with perfect adherence — the barrier to adoption in clinical practice is low. The findings align with a growing body of work showing that postoperative outcomes after rotator cuff repair are determined not only by the structural success of the repair but by patients’ fear responses, attentional capacities and engagement patterns. They also underscore the authors’ call for a multidimensional outcome framework in rehabilitation research, one that integrates clinical scales, psychological measures and neural network indices rather than relying on any single metric.

The trial’s single-center design, its week-6 start point amid ongoing tissue healing, and its 18-week follow-up all leave open questions about long-term maintenance, return to work and sport, and reinjury risk. Future studies, the authors argue, should employ larger samples, longer follow-up, adaptive cognitive-task progression and formal mediation analyses to trace how psychological factors, brain network reorganization and functional recovery drive one another over time. But the central message stands: rehabilitation after shoulder surgery may be as much a task for the brain as for the tendon, and a little mental arithmetic during physiotherapy might be a surprisingly powerful way to train both.