Non-invasive neuromodulation in Alzheimer's disease: toward brain-heart coupling as a candidate framework for biomarker-guided adaptive intervention.

Alzheimer's disease (AD) has entered an era of biologically informed, disease-modifying therapy, yet durable clinical benefit remains limited, highlighting the need for complementary interventions. Noninvasive neuromodulation has emerged as a promising strategy, but the field is still largely organized by device categories, stimulation targets, and short-term cognitive outcomes. This framing alone does not readily account for the marked heterogeneity, stage dependence, and limited durability reported across studies. This narrative review synthesizes clinical, preclinical, and mechanistic evidence across repetitive transcranial magnetic stimulation, transcranial electrical stimulation, sensory gamma entrainment, focused ultrasound, photobiomodulation, and transcutaneous vagus nerve stimulation. We propose that device classification is a useful starting point but an incomplete explanatory framework, because different modalities have been linked, with varying levels of evidence, to overlapping processes involving network reconfiguration, autonomic regulation, neurovascular function, inflammatory signaling, and clearance-related biology. Against this background, we examine the brain-heart axis and the central autonomic network (CAN) as a candidate intermediate level at which central neural and peripheral physiological responses to stimulation might be jointly characterized. Building on this view, we propose a biomarker hierarchy that extends from heart rate variability to directional and time-varying brain-heart coupling measures, which could be tested as candidate markers of physiological state, target engagement, and treatment responsiveness. At present, however, there is no prospective evidence that brain-heart coupling causally mediates the therapeutic effects of neuromodulation in AD. We, therefore, present brain-heart coupling as a testable candidate framework for studying state dependence, target engagement, and, ultimately, biomarker-guided adaptive neuromodulation.