Search

We searched titles and abstracts from PubMed (MEDLINE) and Scopus from inception to July 2022 for potential research studies using the search terms and Boolean operators: (hypoxia or hypoxic) AND (hangover OR recovery) AND (cognition OR cognitive OR "cognitive performance" OR “flight performance” OR mood OR sleepiness OR symptoms OR effort OR physiology OR physiological OR oxygenation OR "oxygen saturation" OR "heart rate" OR "heart rate variability" OR “blood pressure” OR ventilation OR “respiratory rate” OR respiration OR “blood gases”) NOT (animal OR murine OR rodent OR mice OR porcine OR dog OR fish OR cell). Furthermore, additional articles known to the first author (DMS) that were not identified in the literature search were added and titles (only) of the following were screened by DMS: 1) reference lists of eligible articles; 2) forward citation tracking using Google Scholar of eligible articles; and 3) searching key journals and Google Scholar using a combination of the terms: hypoxia, hypoxic, recovery and hangover. All potential articles identified through other sources were screened by DMS immediately following retrieval. Full details of the screening process are displayed in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow chart.

https://doi.org/10.1371/journal.pone.0289716.g001

Inclusion and exclusion criteria

We included studies fulfilling the following criteria: 1) between- or within-subject experimental trials; 2) human participants of any age with no known medical conditions; 3) an objective measure of cognition or physiology during recovery from an acute hypoxic exposure; and 4) peer reviewed full text original research studies published in English. Acute hypoxia was defined as <300 min of breathing hypoxic air that lowered peripheral blood haemoglobin-oxygen saturation (SpO₂) to <90% (i.e. normobaric or hypobaric hypoxia) and recovery was defined as the period of breathing air that increased and normalised peripheral blood SpO₂ to >95% following an acute hypoxic exposure. We excluded studies if: 1) the control or baseline group (or condition) were not administered air able to maintain SpO₂ >95%; 2) other interventions known to effect cognition were included in conjunction with oxygen manipulation, except for carbon dioxide; 3) single or accumulated repeated hypoxia exposures were administered beyond 300 min; 4) hypoxia was repeatedly interrupted by normoxic breathing; and 5) cognitive and/or physiological data were not adequately reported. As we were specifically interested in physiological responses that can be readily measured in military aviation training and operational settings, we restricted our physiological indices of interest to ventilatory, cardiovascular, regional cerebral tissue haemoglobin-oxygen saturation (rSO₂), SpO₂, and cardiac autonomic responses. Measures of less interest and low feasibility in the training and operational setting (e.g. intra-ocular pressure, limb blood flow, and vascular resistance) were omitted. Review articles, unpublished abstracts, theses, and dissertations were also excluded.

Screening

Literature search results were entered into Mendeley, which automatically removed duplicates, then exported to Rayyan (i.e. online systematic review software). Two authors (DMS and AB) independently screened titles and abstracts for suitability. The full text for studies of interest were retrieved and independently evaluated for suitability by the same two authors. Disagreements between authors’ decisions were resolved via discussion and consensus.

Risk of bias/quality assessment

Two authors (DMS and PMB) independently performed the assessment of risk of bias in the included studies using the Rosendal Scale [9]. This scale combines the PEDro scale [10], Jadad scoring system [11] and Delphi list [12]. This Rosendal scale was selected as the PEDro scale, Jadad scoring system and Delphi list have been extensively evaluated and validated. A Rosendal score of 60% is considered as excellent methodological quality [9]. Two checklist items were removed, including to whether the researchers were blinded as it was deemed inappropriate (i.e. due to safety reasons) and reporting of methods used to report adverse effects as the responses to acute hypoxia are inherently deemed adverse. An additional checklist item assessing whether treatment order was counterbalanced was included. No studies were excluded based on quality assessment results. Disagreements were resolved by discussion.

Overview of included studies and study quality

The literature search initially identified 2698 studies for screening. The full text of 32 studies were assessed and 22 were considered eligible. An additional 8 studies from other sources were identified and considered eligible, giving a total of 30 full text articles included in this review. Table 1 provides an overview of hypoxia and recovery interventions for the included studies. 20 studies utilised poikilocapnic hypoxia [5, 6, 13–30], 9 studies utilised isocapnic hypoxia [31–39] and 1 study utilised hypercapnic hypoxia [40]. The duration of hypoxia was typically <30 min and depended on the severity of hypoxic breathing; for instance, studies assessing ~25,000 ft (7,600 m) equivalent (i.e. ~7–8% oxygen) were ~5 min or less, whereas, studies assessing ~18,000 ft (5,486 m) equivalent (i.e. ~10–11% oxygen) were ~25–30 min. The severity of hypoxic breathing was generally >18,000 ft equivalent or <10–11% oxygen, and some studies titrated the F_IO₂ to maintain a specific end-tidal partial pressure of oxygen (P_ETO₂) or SpO₂. Some studies assessed different hypoxic durations and severities. For recovery, 19 studies only administered normoxia (i.e. 21% oxygen) [14, 15, 20–27, 30, 32–34, 36–40], 2 studies only administered hyperoxia [18, 29], 5 studies administered both normoxia and hyperoxia [6, 13, 17, 28, 35], and 4 studies compared differences between normoxic and hyperoxic interventions [5, 16, 19, 31]. We included measures of cognitive performance, SpO₂, heart rate (HR), heart rate variability (HRV; specifically, root mean square of successive differences between normal heartbeats [rMSSD], standard deviation of NN intervals [SDNN] and low-frequency [LF]/high-frequency [HF] ratio of spectral domains), blood pressure (BP; specifically mean arterial pressure [MAP], diastolic blood pressure [DBP] and systolic blood pressure [SBP]), rSO₂, ventilation and electroencephalogram (EEG) measurements. Methodological quality assessment yielded an average Rosendal score of 58 ± 15% (range, 30–87%). Full results of the quality assessment can be found in S1 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of included studies reporting recovery of cognitive and/or physiological responses following a hypoxic exposure.

https://doi.org/10.1371/journal.pone.0289716.t001

Cognition

Eleven studies measured cognition during recovery from hypoxia, including measures from standardised tests and simulated flight performance (Table 2). One study measured composite scores of different cognitive domains within the synthetic work environment (SYNWIN) test battery and reported normalisation at ~20 min [13]. Three studies measured simple and choice reaction time; 1 study reported normalisation at 60 min [17] and 2 studies reported continued impairments, 1 at 10 min [21] and 1 at 24 hours [20]. Two studies measured attention; 1 study reported normalisation of reaction time and lapses (using the psychomotor vigilance task [PVT]) at 60 min [5] and 1 study reported normalisation of commission errors (using the Conners’ continuous performance test) at 13–16 min [27]. One study measured sequential number reading and reported normalisation at 5 min [25] and 1 study measured mathematical processing using the Paced Auditory Serial Addition Task (PASAT), a proxy of attention and working memory, and reported normalisation at 90 sec [19]. Two studies measured simulated flight performance and reported impairments for 10 min [6, 28]. Another study also measured simulated flight performance, in conjunction with a time-estimation task, but due to methodological issues, could not make comparisons to baseline; instead, comparisons between 21% and 13.1% oxygen recovery revealed no difference in flight performance errors, and time-estimation task lapses and variance at 35 min [22]. Not all hypoxic protocols impaired all measures of cognition performance [5, 20, 27] and 21% versus 100% oxygen did not elicit clear or persistent differences in cognitive performance [5, 19].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Studies reporting recovery of cognitive responses following a hypoxic exposure.

https://doi.org/10.1371/journal.pone.0289716.t002

Cardiovascular responses

Table 3 provides an overview of the recovery of SpO₂, HR, HRV and BP during recovery from hypoxia. Eighteen studies measured SpO₂; SpO₂ normalised within 1–10 minutes in almost all studies [5, 14, 15, 18, 19, 21, 24–26, 29, 34–37] and three studies reported SpO₂ normalised by 13–35 min [23, 27, 30]. One study could only report SpO₂ recovery compared to a hypoxic condition [22]. Fifteen studies measured HR; HR recovered in 1–10 min in nine studies [5, 19, 24, 29, 33, 35, 36, 38, 39] and 11–20 min in three studies [26, 27]. HR was lower at 7 min in two studies [14, 15] and one study could only report HR recovery compared to a hypoxic condition [22]. Five studies measured HRV indices; HRV normalised at 3–7 min in three studies [14, 15, 19] and at 20 min in one study [23], whereas one study reported some perturbations of HRV persisted at ~137 sec [29]. Seven studies measured BP indices; BP normalised at 5–10 min in 3 studies (i.e. SBP [33, 38] and MAP and SBP [39]), and SBP and DBP normalised at 20 min in one study [30]. However, SBP remained increased after an unspecified time in one study [35], DBP and SBP remained increased at 5 min in one study [36], and MAP remained increased at 11 min in one study [26]. HR, HRV and BP indices were not altered by all hypoxic interventions, and 100% oxygen breathing during recovery normalised SpO₂ [5, 19] and HR [5] only marginally faster (seconds) than 21% oxygen.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Studies reporting recovery of cardiovascular responses following a hypoxic exposure.

https://doi.org/10.1371/journal.pone.0289716.t003

Ventilatory responses

Ten studies measured ventilatory indices during recovery from hypoxia (Table 4). Minute ventilation (VE) normalised within seconds to 15 min in almost all studies [16, 27, 31–33, 35, 36, 38, 39]; however, one study reported VE was reduced after 5 min [40] and one study reported VE during hyperoxic recovery transiently undershot hyperoxic baseline following 3 and 5 min of hypoxia (but not shorter hypoxia exposures) [16].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Studies reporting recovery of ventilatory responses following a hypoxic exposure.

https://doi.org/10.1371/journal.pone.0289716.t004

Neurophysiological responses

Cerebral tissue oxygenation.

Four studies measured rSO₂ during recovery from hypoxia (Table 5). rSO₂ normalised within 10–13 min in two studies [27, 39], whereas one study reported rSO₂ did not normalise after 10 min [20] and one study reported rSO₂ normalised at 24 hours, but not at 2 hours [20].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Studies reporting recovery of regional cerebral tissue oxygenation responses following a hypoxic exposure.

https://doi.org/10.1371/journal.pone.0289716.t005

EEG.

Two studies measured EEG indices during recovery from hypoxia (Table 6). One study reported differences in EEG indices persisted for up to 4 hours with no difference between 21% and 100% oxygen recovery [5], whereas another study reported rapid normalisation (minutes) of EEG indices albeit a transient (seconds) worsening with 100% oxygen breathing during recovery compared with 21% oxygen [19].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Studies reporting recovery of EEG responses following a hypoxic exposure.

https://doi.org/10.1371/journal.pone.0289716.t006

State and limitations of evidence

The purpose of this systematic review was to consolidate the available evidence on the recovery of cognitive and basic physiological responses following an acute hypoxic exposure. This was to improve our understanding of how hypoxia could impair performance and compromise safety despite returning to normoxic (or hyperoxic) air breathing. Currently, there are insufficient published articles to accurately quantify post-hypoxia recovery profiles of cognitive and physiological indices, and their influential factors, which prevented meta-analysing the data and restricted this article to a systematic review. Some studies also employed methodologies that made it difficult to extract true recovery durations and, therefore, the time point for full recovery could not be determined. This was often due to articles not reporting time course profiles (i.e. repeated measures) and only reporting a single value that represented a range of time. The majority of research (19 studies) included men and women to allow findings to be applicable to both sexes; however, there was still a tendency to favour males. Study quality was moderate (average Rosendal score of 58 ± 15) and methodological differences made it difficult to compare between studies. Nevertheless, there is sufficient evidence supporting the presence of a ‘hypoxia hangover’ and to help to inform post-hypoxia restrictions.

Recovery of cognitive functions

Hypoxia exponentially degrades cognitive functions with increasing hypoxaemia until loss-of-consciousness [4]. During hypoxia, there is preferential blood flow to posterior regions of the brain that are essential for regulating vital functions (e.g. breathing) [41] but are not highly involved in complex cognitive functions. Therefore, the anterior regions are more at risk of reduced oxygen delivery and, since they are involved in higher order and more complex cognitive processes, are easily degraded by hypoxia. This may be exacerbated by increased oxygen requirement of active brain regions and higher neuronal sensitivity to oxygen deprivation. In the present review, studies included a range of simple and complex cognitive tasks, and since some of these were not sensitive to the effects of hypoxia, they were unable to provide insight into the recovery period. Nonetheless, there appears to be persistent cognitive impairments for some standardised cognitive tasks (or domains) and more complex simulated flight performance tasks.

The most informative study designs used repeated measures of simple tasks that have a short temporal resolution, such as reaction speed and attention (refer to Table 2). Initially, Phillips et al. (2009) reported choice reaction time was slower 10 min into recovery, with some participants exhibiting a slower reaction time compared to hypoxia [21]. The same researchers later demonstrated simple and choice reaction speed were slower after 1 and 2 hours into recovery, and normalised at 24 hours [20]; however, there were no measures between 2 and 24 hours, with full recovery likely occurring earlier. More recently, Blacker & McHail (2021) reported vigilant attention (i.e. mean reaction time using a 10 min psychomotor vigilance task) was reduced 20 min into recovery and normalised after 1 hour, with no differences between breathing 21% or 100% oxygen during recovery [5]. This was similar to an earlier study that reported total response time for simple and choice reaction time tasks normalising by 60 min [17]. These observations suggest that performance of both simple reaction time and more demanding vigilance tasks can be impaired in the hours following a hypoxic exposure.

Performance of complex tasks were less informative as these are more difficult to measure. Since they typically take longer to assess, repeated measures are also difficult to ascertain. Similarly, studies reporting cognitive function for a single timepoint or range of time proximal to the cessation of hypoxia did not provide much insight into cognitive recovery profiles. Nevertheless, auditory serial addition task performance (a measure of attention and working memory) normalised at 90 sec [19], and task time and errors for serial number reading using the King-Devick test normalised at 5 min [25], suggesting rapid recovery. There may also be differences in sensory processing following hypoxia. For example, auditory monitoring tasks may recover at a slower rate compared to visual monitoring, memory and mathematical processing [13]. Future studies should aim to corroborate these findings and discern if there is a difference in post-hypoxia recovery profiles between simple and complex cognitive tasks, and how these are influenced by the hypoxic dose and recovery procedures.

Flight performance tasks are more reflective of the real-world implications posed by post-hypoxia cognitive impairments. Robinson et al. (2018) assessed simulated flight performance in non-pilots concurrently with a time-estimation task during successive hypoxic exposures, but only comparisons between different recovery modalities were possible. Following hypoxic exposure (i.e. normobaric 6.5% oxygen for 5 min), there were no differences in flight and time-estimation task performance between 21% and 13.1% oxygen breathing after 35 min [22]. However, flight performance errors and time-estimation task lapses during recovery with 13.1% oxygen were higher when preceded by hypoxia compared with normoxia, suggesting there was an additive effect from the prior hypoxic exposure [22]. Further, Varis et al. (2019 & 2022) assessed simulated flight performance during a return-to-base landing following an inflight hypoxic emergency on trained Hawk fighter pilots in two studies and reported impairments persisted for 10 min following exposure to 6% oxygen, alongside impaired situational awareness and adverse subjective feelings such as light headedness, visual impairments and dizziness [6, 28]. Some participants also reported incidents when driving home following testing [6].

Recovery of physiological and neurophysiological status

Hypoxia elicits an integrated physiological response that stems primarily from peripheral chemoreceptors sensing hypoxaemia [42]. This chemoreflex is initiated from receptors predominantly located in the carotid body, which increase activity via the carotid sinus nerve that projects to the lower brainstem and nucleus tractus solitarius. This increases ventilation and autonomic sympathetic activity to raise HR and BP, and redistribute blood flow to critical tissues where vasodilation occurs, such as the brain [43]. The subsequent baroreflex initiates a vagal response and there seems to be a baroreflex resetting to shift baroreceptor activity to a higher threshold that allows for sustained and increased sympathetic innervation [36]. In severe hypoxia, compensatory responses are insufficient to maintain brain tissue oxygenation [44], which results in widespread slowing of EEG indices [19]. During normoxic recovery, HR and SpO₂ recover within seconds and only marginally faster when breathing 100% oxygen (refer to Table 3). This withdraws the primary chemoreceptor stimulus, yet some physiological perturbations persist.

Autonomic nervous system activity can be inferred using cardiac indices, such as HRV. During hypoxia, HRV indices tend to decline, including SDNN, rMSSD and LF and HF spectral domains (refer to Table 3), indicating greater sympathetic innervation and parasympathetic withdrawal; however, these seem to normalise within 3–20 min [14, 15, 19, 23]. Sex and susceptibility to hypoxia also appear to influence the HRV response. For example, Botek et al. (2015) reported greater vagal withdrawal (i.e. lower HF spectral domain) during hypoxia and at 7 min into recovery in participants exhibiting a lower SpO₂ [14]. The same researchers later reported males had a relatively higher sympathetic response to hypoxia exposure compared with females (i.e. higher natural logarithm of SDNN/rMSSD and LF spectral domain), but differences did not persist at 7 min into recovery [15]. Therefore, HRV indices suggest the increase in sympathetic activity during hypoxia normalises minutes into recovery.

Increased sympathetic activity may, in fact, persist for longer than indicated by HRV indices. For example, several included articles also reported increased MSNA for at least 15–20 min into recovery [26, 33, 35, 36, 38]. Although MNSA was not included in this review due to it being a difficult measure to ascertain, it provides a direct measure of sympathetic activity and highlights that increased sympathetic activity during hypoxic recovery may not be fully captured by HRV. The reasons for this are uncertain and despite elevated post-hypoxia MNSA being reduced by periods of hyperoxic breathing to suggest reduced chemoreflex activity [35], the rapid normalisation of SpO₂, end-tidal gases and VE indicates chemosensitivity is unlikely underpinning persistent elevated MSNA. Rather, increased MSNA is more likely due to other reasons, such as long-term potentiation of post-ganglionic nerves [35].

The cardiovascular response to hypoxia increases BP, particularly SBP, which appears to normalise within 5–20 min in most, but not all, studies (refer to Table 3). Increased BP initiates the baroreflex to increase vagal activation and during hypoxia there is an attenuated cardiac baroreflex to allow for vagal adaptation. For example, Roche et al. (2002) demonstrated the increased hypoxic sympathetic excitation can be followed by an upregulated parasympathetic drive stemming from overactivity of the baroreflex to cause relative bradycardia during recovery [23]. This reduction in post-hypoxia HR was also shown by Botek et al. (2015 & 2018) [14, 15]. The interaction of arterial carbon dioxide and hypoxia may also alter baroreceptor resetting, with poikilocapnic hypoxia potentially having less influence on baroreceptor resetting compared to isocapnic hypoxia [45]. These effects highlight a complex interplay of chemoreflex and baroreflex regulation during hypoxia and recovery.

Brain tissue oxygenation (e.g. rSO₂) is a more informative measure than peripheral indices (e.g. SpO₂) of cognitive function [44]. However, studies measuring post-hypoxia rSO₂ using near-infrared spectroscopy (fNIRS) report conflicting findings. Phillips et al. (2009 & 2019) demonstrated rSO₂ remained reduced at 10 min [21] and 2 hours [20], and normalised by 24 hours into recovery [20]. These perturbations in rSO₂ mirrored performance impairments for SRT and CRT tasks [20, 21]. Nonetheless, considering no measures were taken between 2 and 24 hours, rSO₂ probably normalised before 24 hours and, as these studies were not adequately controlled, the authors described the findings as preliminary. In comparison, Steinback et al. (2012) and Uchida et al. (2020) reported rSO₂ normalised 10–16 min into post-hypoxia recovery [27, 39]. These differences are difficult to explain but are likely due to methodological differences in hypoxic and recovery interventions, and fNIRS measurement techniques, which may not necessarily accurately reflect brain tissue oxygenation [46]. Further, measures of frontal/prefrontal cortex oxygenation represents a tissue average and may not be sensitive to regional differences that can occur during hypoxia.

The ventilatory increase during hypoxia peaks after 5 min then declines over 20–30 min to a steady-state above pre-hypoxic levels [26]. This hyperventilation can cause hypocapnia if carbon dioxide is not administered within the breathing gas (i.e. poikilocapnic hypoxia), which may be an important physiological determinant in studies measuring flight performance during hypoxic recovery [28]. Ventilation typically normalised within 15 min following cessation of hypoxia (refer to Table 4) and, in some studies, this included an initial transient undershoot to below pre-hypoxic levels after returning to normoxia [40] and hyperoxia (compared to hyperoxic baseline) [16]. The ventilatory response to repeated hypoxic exposures may also be attenuated [31], which suggests central chemosensitivity is reduced, which could increase susceptibility to hypoxia by exacerbating hypoxaemia.

EEG measures provide insight into brain signalling activity. Malle et al. (2016) measured EEG waveforms whilst performing a demanding cognitive task (i.e. PASAT) and showed an increase in SEF95 during hypoxia (i.e. the frequency below which 95% of total EEG power was contained), suggesting an increase in fast-wave activity [19]. However, recovery with 100% oxygen breathing generated a robust EEG slowing for ~30 sec (i.e. increase in theta activity and decrease in SEF95) [19], suggesting hyperoxia may elicit an initial harmful effect on the brain. Whereas, Blacker & McHail (2021) measured passive elicited event-related potentials that assess auditory processing, and demonstrated a continued decline in mismatch negativity (MMN) amplitude during post-hypoxia recovery, which normalised after 120 min, and a delayed response MMN peak latency, with shorter latencies that normalised after 240 min [5]. Whereas, P3a, a measure of attention, normalised immediately during recovery, and 100% compared with 21% oxygen breathing had no effect on recovery of EEG indices [5]. Therefore, pre-conscious auditory processing may require up to 2 hours to recover following a hypoxic exposure, which could be connected with the persistent impairment in auditory monitoring previously reported [13].

Perspectives and conclusion

Understanding recovery from hypoxia and its practical implications is critical for various populations. Current research suggests there may be lagging effects for the recovery of some cognitive and physiological indices, but temporal profiles are unclear. Generally, most research has focussed on measuring responses to hypoxia rather than during recovery following return to normoxic and/or hyperoxic breathing. The effects of hypoxia have largely been established and although there is a better need to understand how these affect behaviour and decision making within various real-world situations, further research also needs to consider the post hypoxic period. Impaired cognitive functions can compromise performance and safety, and there is emerging evidence to suggest complex real-world skills, such as piloting an aircraft, are impaired during the immediate minutes following a hypoxic exposure. Thus, if there is the assumption that recovery is rapid due to SpO₂ normalising within seconds-to-minutes, then individuals are likely to expose themselves to unnecessary risk during the post-hypoxia period that could result in serious or fatal incidents. This could occur following numerous situations, such as an inflight decompression at high-altitude, failure of oxygen supply systems to provide sufficient oxygen, hypoxia recognition training (i.e. standard military aircrew training), or after hypoxia research studies. However, the duration of potential impairments and the recovery profiles are unclear.

The mediating effects of the hypoxic dose and level of oxygen supplied during recovery is also uncertain. It seems plausible that the longer and more severe a hypoxic exposure, the longer the recovery period. However, whether interactions of F_IO₂ (and/or barometric pressure) and duration that elicit similar hypoxic doses cause different recovery profiles should also be assessed. For example, studies within this review included prolonged moderate-hypoxia (e.g. ~18,000 ft [5,486 m] equivalent or ~10–11% oxygen) and short severe-hypoxia (e.g. ~25,000 ft [7,620 m] equivalent or ~7–8% oxygen) interventions, but it is unclear which elicits greater cognitive and physiological effects during the post-hypoxic recovery. With poikilocapnic hypoxia, there is also an increased risk of hypocapnia resulting from hyperventilation, which causes cerebral hypoperfusion to exacerbate the effects of hypoxia. Therefore, the mediating effect of adding carbon dioxide to the recovery gas also needs to be examined.

Hyperoxic breathing is currently a focal area of research, particularly in military aviation, and appears to have a beneficial effect on aspects of cognition [47]. Nonetheless, the use of 100% oxygen for recovery from hypoxia did not demonstrate a beneficial effect compared to normoxia. Rather, 100% compared with 21% oxygen breathing during recovery seemed to cause an EEG slowing and impaired cognitive performance during the initial seconds following hypoxia [19], but this was not reported for all studies [5]. This suggests hyperoxic recovery may be harmful, which has previously been eluded to in the military aviation context [4], yet remains standard practice. There may also be a paradoxical effect whereby hyperoxic or nomoxic breathing post-hypoxia elicits a transient worsening of cognitive functions and symptoms, termed the ‘oxygen paradox’ [48]. This is potentially due to a transient vasoconstrictive effect of hyperoxia that reduces cerebral perfusion and a drop in arterial blood pressure [49], thus exacerbating tissue hypoxia. However, to the authors’ knowledge, there is no published literature demonstrating the effects of hyperoxia compared with normoxia on post-hypoxia cerebral perfusion and tissue oxygenation. In fact, there is scant research investigating the oxygen paradox, including cognitive, physiological and perceptive responses.

Whether hypoxia is induced by a reduction in barometric pressure or F_IO₂ is critical to manage the risk of decompression sickness (DCS). Ascending above 18,000 ft (5,485 m), which is approximately half the atmospheric pressure at sea-level, is associated with an increased risk of venous gas emboli and DCS [50]. Although we did not include DCS risk within our review, this is an important consideration in the aviation context and pose a greater risk than hypoxia alone.

In summary, this systematic review suggests there is a need for post-hypoxia restrictions to minimise the risk of potential incidents and accidents due to lagging cognitive and physiological effects. As such, current evidence supports the presence of a ‘hypoxia hangover’ but the severity and duration of impairments are difficult to quantify and likely depend on several factors. Future research should aim to systematically assess a range of cognitive and physiological responses that persist into recovery following a hypoxic exposure, and the influence of different conditions (e.g. hypoxic dose and recovery procedures). This review suggests that recovery of SpO₂ and HR may only indicate partial recovery, and normalisation of other physiological indices can require to up to ~2–4 hours to return to levels before the hypoxic exposure. It therefore seems appropriate that safety measures are implemented following acute hypoxic exposures to mitigate the risks imposed by persistent cognitive impairments and physiological perturbations.