Purpose <p>Obstructive sleep apnea (OSA) is associated with neurocognitive impairment and central nervous system dysfunction. Long-latency reflexes (LLRs) provide a practical neurophysiological probe of cortical sensorimotor integration. This study investigated whether OSA selectively alters upper-limb LLR components.</p> Methods <p>Thirty-nine patients with polysomnography-confirmed OSA and 21 healthy controls underwent Epworth Sleepiness Scale assessment and standardized median nerve stimulation at the wrist, with recordings from the abductor pollicis brevis during sustained contraction. H-reflex and LLR1–LLR3 onset latencies and response detectability were evaluated. Patients were stratified by OSA severity and REM/positional phenotype. Associations with polysomnographic indices, including minimum oxygen saturation, were explored.</p> Results <p>H-reflex, LLR1, and LLR3 latencies did not differ significantly between groups. LLR2 latency was prolonged in moderate-to-severe OSA compared with controls in the unadjusted analysis (53.1 vs. 50.4 ms, <i>p</i> = 0.014), and the H–LLR2 interval was also longer (27.1 vs. 25.5 ms, <i>p</i> = 0.029). After adjustment for age and BMI, the group effect on LLR2 latency remained significant (F = 3.686, <i>p</i> = 0.032), whereas the adjusted group effect for the H–LLR2 interval was no longer significant. In an exploratory subgroup analysis of mild OSA, unadjusted inverse correlation was observed between LLR2 latency and minimum oxygen saturation (<i>r</i> = − 0.845, <i>p</i> = 0.008); however, this association did not persist after adjustment for age and BMI (partial <i>r</i> = 0.188, <i>p</i> = 0.378).</p> Conclusions <p>Moderate-to-severe OSA may be associated with selective prolongation of LLR2 latency, compatible with altered long-loop sensorimotor processing. LLR testing may represent a simple exploratory neurophysiological approach to complement conventional sleep metrics, pending validation in larger prospective cohorts.</p>