Background: Acute mountain sickness (AMS) is a common condition at high altitude, presenting with symptoms such as headache, fatigue, and, in severe cases, life-threatening pulmonary or cerebral oedema. Predicting the occurrence of AMS using standard clinical metrics remains challenging. Network physiology is a multidisciplinary field that explores the collective behaviour of physiological systems and may offer a novel approach to understanding AMS by visualising and quantifying the complex interactions and information exchange between these systems.

Aim: This study aims to explore how physiological networks change during 11 hours of normobaric hypoxia and whether causal information flow within the cardiorespiratory system differs between individuals who develop symptoms of AMS and those who do not.

Method: All experimental procedures were approved by the Science and Health Faculty Ethics Committee of The University of Portsmouth (project no. SHFEC 2024-034). A convenience sample of ten individuals provided their written informed consent to participate in this study (9 male and 1 female, age: 26.5±5.9).  The participants underwent 11 hours of normobaric hypoxia (FiO₂= ~0.12) while the following physiological signals were recorded: heart rate (HR), respiratory rate (RR), tidal volume (VT), minute ventilation (VE), end-tidal CO₂ (PetCO₂), end-tidal O₂ (PetO₂), and capillary oxygen saturation (SpO₂). Transfer entropy was used to construct causal network maps, measuring the directed information exchange between time-series variables (1). The Lake Louise Acute Mountain Sickness Questionnaire (LLS) was used to assess the development of AMS. An LLS score of ≥ 3, in the presence of headache and at least one other symptom, indicates AMS (n = 5). Some participants exited early because they were unable to complete the 11 hours of hypoxia.

Results: Two-way ANOVA results (n=10) showed no significant difference between mean values of measured physiological variables between AMS and Non-AMS group throughout hypoxia hours, suggesting that cardiorespiratory signals alone may fail to predict AMS onset. Network analysis revealed some topological differences between groups; Qualitatively, non-AMS group exhibited a more connected network with highly integrated information exchange, suggesting better physiological adaptation. The AMS group showed a progressive decoupling of physiological nodes while this was potentially confounded by participants’ attrition as they developed AMS and exited at different hours. Specifically, SpO₂ and HR nodes, main information receiving (indegree) nodes at the baseline hour (FiO₂=21%), experienced a notable decrease as participants entered hypoxia for both groups. To account for different AMS participants’ exit time due to AMS onset, Event–Locked Analysis was performed, resetting the baseline to AMS event onset. Results showed qualitatively that the system underwent significant uncoupling in the hours preceding AMS, characterized by SpO₂ and HR’s visible indegree reduction, suggesting less information exchange under prolonged hypoxia. The uncoupling may be further supported by the decrease in information sent between all variables in hours before AMS compared to baseline. There was no statistical difference in indegree and outdegree between the groups.

Conclusion: Physiological network mapping has the potential to visualise adaptive responses to physiological stressors, such as prolonged hypoxia, in order to develop physiological markers for the early detection of AMS.