MEDEX 2015: Heart Rate Variability Predicts Development of Acute Mountain Sickness

Abstract

Sutherland, Angus, Joseph Freer, Laura Evans, Alberto Dolci, Matteo Crotti, and Jamie Hugo Macdonald. MEDEX 2015: Heart rate variability predicts development of acute mountain sickness. High Alt Med Biol. 18: 199–208, 2017.

Aims:

Acute mountain sickness (AMS) develops when the body fails to acclimatize to atmospheric changes at altitude. Preascent prediction of susceptibility to AMS would be a useful tool to prevent subsequent harm. Changes to peripheral oxygen saturation (SpO₂) on hypoxic exposure have previously been shown to be of poor predictive value. Heart rate variability (HRV) has shown promise in the early prediction of AMS, but its use pre-expedition has not previously been investigated. We aimed to determine whether pre- and intraexpedition HRV assessment could predict susceptibility to AMS at high altitude with better diagnostic accuracy than SpO₂.

Methods:

Forty-four healthy volunteers undertook an expedition in the Nepali Himalaya to >5000 m. SpO₂ and HRV parameters were recorded at rest in normoxia and in a normobaric hypoxic chamber before the expedition. On the expedition HRV parameters and SpO₂ were collected again at 3841 m. A daily Lake Louise Score was obtained to assess AMS symptomology.

Results:

Low frequency/high frequency (LF/HF) ratio in normoxia (cutpoint ≤2.28 a.u.) and LF following 15 minutes of exposure to normobaric hypoxia had moderate (area under the curve ≥0.8) diagnostic accuracy. LF/HF ratio in normoxia had the highest sensitivity (85%) and specificity (88%) for predicting AMS on subsequent ascent to altitude. In contrast, pre-expedition SpO₂ measurements had poor (area under the curve <0.7) diagnostic accuracy and inferior sensitivity and specificity.

Conclusions:

Pre-ascent measurement of HRV in normoxia was found to be of better diagnostic accuracy for AMS prediction than all measures of HRV in hypoxia, and better than peripheral oxygen saturation monitoring.

Introduction

Acute mountain sickness (AMS) is a complex syndrome presenting at altitudes above 2500 m and will affect ∼50% of people above 4000 m altitude (Maggiorini et al., 1990). It is characterized by headache, gastrointestinal symptoms, fatigue, insomnia, and neurological symptoms diagnosed clinically using the Lake Louise Score (Lake Louise Consensus on the Definition and Quantification of Altitude Illness, 1991; Honigman et al., 1993). Accurate prediction of AMS susceptibility is of importance to enable enhanced prevention measures through increased observation, behavioral changes (e.g., slower ascent profile), and/or prophylactic medication. However, a widely accepted method for prediction of AMS currently does not exist.

The exact mechanism causing AMS is unknown, but it is thought to be related to neurohumoral or hemodynamic processes within the central nervous system (Hackett, 1999). Reduced arterial oxygen saturation is presumably the ultimate upstream cause and is known to be associated with increased reporting of AMS symptoms (Oliver et al., 2012). Nevertheless, participants with similar peripheral oxygen saturation (SpO₂) may or may not present with AMS (Luks and Swenson, 2011). As SpO₂ decreases during ascent to high altitude, heart rate increases, presumably as a result of increased sympathetic activity and to maintain oxygen delivery (Brierley et al., 2012). Although the role of the autonomic nervous system in AMS is poorly understood, it is conceivable that autonomic dysregulation could play an important role in its genesis (Duplain et al., 1999).

Despite the mechanism of AMS remaining elusive, many authors have attempted to develop tools to predict AMS susceptibility. The most commonly investigated tool remains SpO₂ monitoring (Burtscher et al., 2008). Unfortunately, such predictive tools have not shown high diagnostic accuracy and therefore have not been widely accepted (Burtscher et al., 2008; Luks and Swenson, 2011). Therefore, alternative or additional measures to improve AMS prediction are of interest. In this regard, Burtscher et al. (2008) demonstrated a small improvement in AMS prediction when combing SpO₂ monitoring with the blood lactate response to hypoxia, but this invasive approach requires technical skills and is time-consuming. If autonomic dysregulation in response to hypoxia has a role in the genesis of AMS, then measures of peripheral autonomic activity such as heart rate variability (HRV) may provide an alternative additional predictive tool.

HRV is the beat-to-beat variation in R-R intervals in an electrocardiogram and it has been suggested that it can be used to monitor the activity of the autonomic nervous system (Billman, 2011). The R-R interval data are analyzed and the power of HRV is quantified by determining the areas of the spectrum in different component widths: high frequency (HF; 0.15–0.4 Hz), low frequency (LF; 0.04–0.15 Hz), very LF (VLF; ≤0.04 Hz), and ultra-LF (ULF; ≤0.0033 Hz) (Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, 1996; Camm et al., 1996).

HF is widely accepted as an indicator of vagal tone. It is currently considered a good estimator of parasympathetic activation of the autonomic nervous system because the sympathetic nervous system accounts for only 10% of HF variability (Malliani et al., 1991; Billman, 2013). In contrast, it has been suggested that sympathetic and parasympathetic nervous system activation explains roughly 50% and 25% of the LF variability, respectively (De Vito et al., 2002; Shaffer et al., 2014). In addition, LF currently is not considered a valid marker of autonomic nervous activity per se as it is influenced by the activity of baroreceptors in addition to those of the autonomic nervous system (De Vito et al., 2002; Rahman et al., 2011; Billman, 2013). A ratio of LF/HF has also been widely used as an indicator of sympathovagal balance. Taking into account what is known about LF, the LF/HF ratio may be more usefully described as a mutual relationship between baroreflex sensitivity and parasympathetic modulation (Cohen and Taylor, 2002; Rahman et al., 2011; Billman, 2013). Finally, a physiological explanation and the mechanism involved in the generation of VLF and ULF have not been clarified. Nevertheless, some authors suggest that circadian rhythms, core body temperature, metabolism, hormones, and intrinsic rhythms generated by the heart could be the cause of VLF and ULF fluctuations (Shaffer et al., 2014).

In clinical settings, HRV has been shown to be useful in postmyocardial infarction risk stratification and in prediction of anxiety disorders (La Rovere et al., 1998; Chalmers et al., 2014). At altitude, a number of studies have investigated the dysregulation of the autonomic nervous system using HRV. These studies have shown a transient reduction in parasympathetic and increased sympathetic activity during acute exposure to hypobaric hypoxia (Kanai et al., 2001). However, there are few studies linking these alterations in the autonomic nervous system with AMS symptoms. Karinen et al. (2012) investigated the potential predictive value of HRV assessed during the actual ascent in determining later AMS susceptibility, showing good sensitivity and specificity in diagnosing AMS, but if HRV is to become a useful applied tool in the prediction and management of AMS, its utility in assessment before expedition to altitude should be investigated. In this regard, Zhang et al. (2014) explored the response of the autonomic system when exposed to short-term normobaric hypoxia and showed that even on acute short-term exposure, a decrease in HRV is seen. Taken together, the query remains of the possible predictive valve of HRV in normobaric hypoxia to predict the development of AMS symptoms on a subsequent expedition to altitude. Advances in heart rate telemetry monitoring systems have made it possible to evaluate HRV noninvasively and in a real-time manner. So, if successful, HRV may offer a low-cost noninvasive tool to help identify those who are at risk of AMS before ascent (Burtscher et al., 2004).

The aim of this study was to determine whether HRV assessment before (acute normobaric normoxia or hypoxia) and during (chronic hypobaric hypoxia) an expedition could predict susceptibility to AMS at high altitude with similar or better diagnostic accuracy to peripheral oxygen desaturation during acute normobaric hypoxia or chronic hypobaric hypoxia. We hypothesized that LF, HF, and their ratio assessed during 15 minutes of normobaric hypoxia (FiO₂ = 11.5%) would have acceptable and the best diagnostic accuracy to predict AMS susceptibility (defined as a Lake Louise Score of headache ≥1 and total score ≥3) during a subsequent trek to high altitude.

Methods

Participants

The study group consisted of 44 healthy participants (age range 24–62 years; 26 males). Participant demographics are provided in Table 1. Participants' maximum living altitude was 450 m and none of the participants had been above 3000 m during the month before the expedition. None of the participants were taking prophylactic medications for AMS, and none were taking medication known to increase HRV (beta-adrenoreceptor antagonists, statins, and angiotensin-converting enzyme inhibitors; Bilchick and Berger 2006). All participants gave written informed consent to participate and the study was approved by the National Institute for Social Care and Health Research Wales Research Ethics Service (14/WA/1260).

Table 1.

Participant Demographics

	Participants included		Participants excluded
	Male	Female	Male	Female
Gender (n)	17	11	9	7

	Mean	SD	Mean	SD
Age (year)	34.0	12.7	42.9	13.4
Height (cm)	172	10	171	10
Body mass (kg)	66.3	12.0	70.9	13.6
VO₂max (mL/kg/min)	46.7	6.0	44.5	9.7
Living altitude (m)	74	118	85	100
Days mountaineering in past year (n)	9	12	5	9
Habitual physical activity score (a.u.)	5.3	1.5	4.2	2.5

	Participants included	Participants excluded
	Yes	Yes
AMS susceptible on the expedition (n)	20	6
Previous experience of AMS (n)	20	13
Previous experience of high altitude pulmonary edema (n)	1	0
Previous experience of high altitude cerebral edema (n)	0	0
Smoker (n)	1	1
Consumes alcohol (n)	27	15
History of headache/migraine (n)	3	5
Consumes dietary supplement (n)	2	3
Previous history of travel above 3000 m (n)	26	15

AMS, acute mountain sickness.

Study design

Baseline data were collected in the laboratories of Bangor University, North Wales (65 m above sea level). One month later, participants completed an expedition that trekked on foot to a field laboratory at 3841 m and on to a maximum altitude of 5085 m (Manaslu circuit in the Nepali Himalaya). Participants were divided into five groups in which ascent profiles were essentially similar (Fig. 1).

FIG. 1.

Ascent profile for all groups.

Baseline data collection

To reduce variability in HRV measures, participants were required to avoid drinking caffeine, alcohol, or smoking for at least 12 hours and to avoid eating and drinking for at least 1 hour before the measurements. Participants presented at the laboratory and were fitted with a commercially available heart rate monitor capable of obtaining HRV data (RS800CX; Polar Electro, Warwick, United Kingdom) and a peripheral oxygen saturation monitor (Onyx II 9550; Nonin, Minnesota). Participants first lay supine in normoxia for 15 minutes with their eyes open and were asked to breathe normally and regularly. Paced breathing was not required of participants because it was deemed impractical for clinical settings and is not required to obtain reliable HRV data (Kobayashi, 2009; Kox et al., 2011). HRV data were collected for the last 5 minutes as previously recommended (Camm et al., 1996) and SpO₂ was obtained in the final 30 seconds. Participants were then taken into a hypoxic chamber set at FiO₂ of 11.5% (equivalent to ∼5000 m) for 15 minutes where the protocol above was repeated.

Data collection during ascent

Under the supervision of each trekking group's research lead, participants kept a daily diary of various parameters, including self-reported AMS symptoms by Lake Louise Score and SpO₂ by fingertip pulse oximetry (as above). The same time of day, setting (for control of background ambient light), and finger were used each day for the measurement of SpO₂ and each participant had a partner who obtained readings to ensure participants remained blinded to their results. Participants had warm hands and obtained measures following overnight fast (Luks and Swenson, 2011). Preceding the overnight fast (∼8 hours), normal habitual caffeine intake was allowed.

In addition, at 3841 m, the HRV and SpO₂ were again obtained using the same data collection methods as at baseline, but while breathing ambient air only.

Data analysis

Data processing was performed by Polar Precision Performance software (version 3.02.007) and Kubios software (version 2.2) (Tarvainen et al., 2013). An initial examination of the HRV data excluded participants from analyses if their data were not considered physiologically plausible or if the tachogram presented evident artifacts exceeding 5% of the total R-R intervals (Dos Santosa et al., 2013) (Fig. 2). A low-level correction was then applied if the R-R Interval was lower than 0.35 seconds (refractory time) or R-R intervals that differed more than 20% from the preceding data point were detected. If the low-level correction detected more than 5% artifact in the HRV recording, then the tachogram was also excluded from analyses (Lippman et al., 1994; Wessel et al., 2000; Dos Santos et al., 2013; Liu et al., 2013).

FIG. 2.

Participant flow through the study.

The power spectra were quantified by measuring the area in two frequency bands over a 5-minute epoch. The LF power was calculated for frequency band 0.04–0.15 Hz and the HF power for frequency band 0.15–0.40 Hz. The ratio of LF power to HF power (LF/HF) was also determined.

SpO₂ data were expressed as the drop in SpO₂ (%) following exposure to hypoxia. AMS was diagnosed when participants were above 2500 m and had presence of headache and at least one other symptom by Lake Louise Score. These participants were classified as AMS susceptible. Participants not meeting these criteria were classified as AMS resistant. Total symptom score was also presented.

Statistical analyses

Diagnostic accuracy analysis was performed to assess the ability of HRV and SpO₂ to predict AMS susceptibility during consequent exposure to altitude. Receiver operating characteristic (ROC) curve analysis to predict AMS susceptibility was applied to the HRV data obtained in three different environmental conditions: normobaric normoxia and normobaric hypoxia (both obtained at baseline) and hypobaric hypoxia (obtained during the trekking ascent at 3841 m). Moreover, ROC curve analysis to predict AMS susceptibility was also applied to the drop in SpO₂ between normoxia and normobaric hypoxia (after 15 minutes, FiO₂ = 11.5%) and the drop in SpO₂ between normoxia and hypobaric hypoxia (after 12 trekking days, 3841 m). All the diagnostic accuracy analyses were performed using R 3.2.5 (R Foundation, www.r-project.org). OptimalCutpoint package was installed in R and it was used to obtain the area under ROC curve (AUC), the optimal cut point, the sensitivity, the specificity, and the ROC curve plots (López-Ratón and Rodríguez-Álvarez, 2014). Epi and pROC packages were also installed in R and used to confirm the results concerning AUC generated by the OptimalCutpoint package. As results were the same, only data from the OptimalCutpoint package are presented herein. The pROC package was also used to calculate the confidence intervals (CIs) of the AUC. An AUC greater than 0.7 was considered fair and an AUC greater than 0.8 was considered moderate (Obuchowski et al., 2004); in both cases, it was required that the 95% CI was greater than 0.5 (www.r-project.org).

Results

Of the 44 participants recruited at the beginning of the study, 28 provided acceptable HRV, SpO₂, and AMS data at all required time points. Therefore, only the data regarding these participants have been used for the data analysis and statistical analysis. The flow of participants through the study is shown in Figure 2. The mean HRV and SpO₂ responses to various environmental conditions are shown in Table 2 and the individual responses for selected variables are shown in Figure 3.

FIG. 3.

Individual responses to hypoxic exposure for selected heart rate variability and peripheral oxygen saturation variables assessed before and during a high altitude expedition. (A) LF/HF ratio after 15-minute exposure to normobaric normoxia; (B) LF after 15-minute exposure to normobaric hypoxia; (C) drop in peripheral oxygen saturation between normoxia and normobaric hypoxia (after 15 minutes, FiO₂ = 11.5%); (D) drop in peripheral oxygen saturation between normoxia and hypobaric hypoxia (after 12 trekking days, 3841 m). HF, high frequency; LF, low frequency; Δ SpO₂ drop in peripheral oxygen saturation between normoxia and hypoxia.

Table 2.

Heart Rate Variability Response and Peripheral Oxygen Saturation Response to Hypoxic Exposure Assessed Before and During a High Altitude Expedition

Environmental condition	Variable	Mean	SD
After 15-minute exposure to normobaric normoxia	HRV ratio (a.u.)	1.99	1.35
	HRV LF (ms²)	1380	1096
	HRV HF (ms²)	1088	1953
After 15-minute exposure to normobaric hypoxia	HRV ratio (a.u.)	3.91	3.13
	HRV LF (ms²)	1031	845
	HRV HF (ms²)	347	273
	Δ SpO₂ (%)	21.1	7.2
After 12 days of exposure to hypobaric hypoxia	HRV ratio (a.u.)	2.78	3.44
	HRV LF (ms²)	1246	1604
	HRV HF (ms²)	763	974
	Δ SpO₂ (%)	9.7	2.5

HF, high frequency; HRV, heart rate variability; LF, low frequency; Δ SpO₂, drop in peripheral oxygen saturation between normoxia and hypoxia.

During the expedition, 20 of the 28 participants showed AMS symptoms, giving an AMS incidence of 71.4%. The length of time with AMS was mean (SD) 3.85 (2.23) days. Participants who presented with clinically defined AMS reported a total accumulative symptom score of 30.3 (14.5); participants who did not present with clinically defined AMS reported a total accumulative symptom score of 12.3 (6.4).

Of the nine HRV predictor variables considered in this study, three variables presented fair diagnostic accuracy to predict AMS susceptibility (normoxia LF, chronic hypobaric hypoxia ratio, and chronic hypobaric hypoxia LF) and two presented moderate diagnostic accuracy (normoxia ratio and acute normobaric hypoxia LF) (Table 3). Of these two moderately accurate predictors, the normoxia ratio had the best diagnostic accuracy, having a higher AUC (0.88 vs. 0.80) and better sensitivity and specificity, correctly identifying 85% of true positives (AMS-susceptible participants) and 88% of true negatives (AMS-resistant participants) (Fig. 4). Notably, LF had better diagnostic accuracy than HF in all environmental conditions, having AUC values greater than 0.7 (AUCs of HF variables were all lower than 0.7). Moreover, LF showed similar or better sensitivity and specificity than HF. In contrast, SpO₂ variables considered in this study showed a poor diagnostic accuracy (AUC <0.7) with poor values of sensitivity and specificity (Table 4 and Fig. 4).

FIG. 4.

Receiver operating characteristic curve for selected heart rate variability and peripheral oxygen desaturation variables assessed before a high altitude expedition to determine susceptibility to acute mountain sickness. Hypoxia LF, LF after 15-minute exposure to normobaric hypoxia; Normoxia Ratio, LF/HF ratio after 15-minute exposure to normobaric normoxia; SpO2 after 15′, drop in peripheral oxygen saturation between normoxia and normobaric hypoxia (after 15 minutes, FiO₂ = 11.5%).

Table 3.

Diagnostic Accuracy of Heart Rate Variability Assessed Before and During a High Altitude Expedition to Determine Susceptibility to Acute Mountain Sickness

Environmental condition	HRV variable	AUC (CI)	Sensitivity %	Specificity %	Optimal cutoff
After 15-minute exposure to normobaric normoxia	Ratio	0.875 (0.741–1.000)	85	87.5	≤2.28 a.u.
	LF	0.781 (0.591–0.972)	75	75	>552 ms²
	HF	0.506 (0.235–0.753)	50	55	>1252 ms²
After 15-minute exposure to normobaric hypoxia	Ratio	0.638 (0.380–0.895)	70	62.5	≤3.13 a.u.
	LF	0.800 (0.627–0.973)	70	75	>275 ms²
	HF	0.694 (0.424–0.963)	70	75	>740 ms²
After 12 days of exposure to hypobaric hypoxia	Ratio	0.731 (0.532–0.931)	70	75	≤1.91 a.u.
	LF	0.725 (0.503–0.947)	80	62.5	>276 ms²
	HF	0.625 (0.383–0.867)	60	62.5	>777 ms²

AUC (CI), area under receiver operating characteristic curve (95% confidence interval).

Table 4.

Diagnostic Accuracy of Peripheral Oxygen Desaturation Assessed Before and During a High Altitude Expedition to Determine Susceptibility to Acute Mountain Sickness

Environmental condition	AUC (CI)	Sensitivity %	Specificity %	Optimal cutoff
After 15-minute exposure to normobaric hypoxia	0.688 (0.478–0.897)	65	75	>21%
After 12 days of exposure to hypobaric hypoxia	0.688 (0.449–0.926)	65	75	>10%

Discussion

The aim of this study was to assess the diagnostic accuracy of pre- and intraexpedition HRV and SpO₂ measurements to predict AMS susceptibility at high altitude.

The study's main findings were that LF/HF ratio in normoxia and LF following 15 minutes of exposure to normobaric hypoxia are promising variables to predict AMS during subsequent altitude exposure, showing moderate diagnostic accuracy (AUC >0.8). Surprisingly, LF/HF ratio in normoxia had the best diagnostic accuracy (having the highest AUC) and the best sensitivity (85%) and specificity (88%) of all predictors. Three other variables had fair diagnostic accuracy (AUC >0.7) to predict AMS susceptibility: normoxia LF, chronic hypobaric hypoxia ratio, and chronic hypobaric hypoxia LF. Generally, LF band HRV had better diagnostic accuracy than HF bands. In contrast, fall in SpO₂ following 15 minutes of exposure to normobaric hypoxia or following 12 days of exposure to hypobaric hypoxia had poor diagnostic accuracy (AUC <0.7).

We further found that at rest in normobaric normoxia, the mean LF/HF ratio was higher in the group of participants who did not subsequently develop AMS than in those who developed AMS. Despite this finding and the generally better diagnostic accuracy for this parameter compared with change in SpO₂, scatterplots of individual data reveal that there is still significant overlap of data points for LF/HF ratio between the AMS and non-AMS groups (Fig. 3).

Heart rate variability

The relationship between hypoxia and HRV is complex and thus it is unsurprising that there are inconsistent results in the literature (Loeppky et al., 2003; Lanfranchi et al., 2005; Chen et al., 2008; Huang et al., 2010; Koehle et al., 2010; Karinen et al., 2012; Wille et al., 2012; Mairer et al., 2013). Generally, hypoxia is thought to potently stimulate the sympathetic nervous system, causing a reduction of HRV variability (LF and HF components) and an increase in the LF/HF ratio (Zhang et al., 2014). The results reported herein are consistent with this commonly observed response to hypoxia, showing a reduction in HRV upon acute exposure that was somewhat normalized with chronic exposure.

The present study extends previous findings by identifying that an LF/HF ratio of less than 2.28 a.u. in normoxia could identify AMS-susceptibility with reasonable diagnostic accuracy. The only other study to have described risks of developing AMS by using an HRV cutoff is by Huang et al. (2010), who found a sevenfold increase in the odds of developing AMS in participants with an LF/HF ratio greater than 1.3 a.u. measured at 1317 m. Although the present results do not concur with this finding, the measurements in Huang et al.'s study were taken in field conditions and were used to predict AMS at moderate altitude (3440 m). In contrast, the present data were taken pre-expedition and were used to predict AMS at a much higher altitude (5085 m). Comparisons between the present and previous studies are also difficult because of differences in data analysis. For example, to our knowledge, this is the first study to have employed strict a priori agreed exclusion criteria regarding implausible data and artifact and the first study to include a diagnostic accuracy analysis to determine performance of HRV data obtained pre-expedition to predict AMS susceptibility.

In the present study, LF/HF ratio in normoxia had the best diagnostic accuracy, while the LF band demonstrated better diagnostic accuracy than HF under all conditions. As mechanisms underlying HRV are complex, the exact physiological explanation of these findings is unclear. The traditional understanding that LF and HF components reflect sympathetic and parasympathetic activities, respectively, is now seen as overly simplistic (Scheff et al., 2014). It has recently been appreciated that other important factors in modulating HRV are mechanoreceptors and ventilatory response (Scheff et al., 2014) and that autonomic changes that affect heart rate tend to alter both LF and HF components (Chiu et al., 2003). Nevertheless, it is tempting to speculate that since LF/HF ratio and LF take into account both sympathetic and parasympathetic cardiac supply and may be related to baroreflex function, these HRV variables are better overall reflections of baroreflex activity, cardiac tone, and sympathovagal balance (De Vito et al., 2002; Shaffer et al., 2014). Hence, their performance to predict AMS susceptibility may be superior because they reflect the interaction of multiple physiological factors that cause AMS.

A surprising finding was the superior diagnostic accuracy of measurements taken in normoxia compared with hypoxia. Many studies have considered pre-expedition measurements in hypoxia to be a proxy for hypoxic conditions at true altitude, but it may be that individuals' physiological response to prolonged hypoxia, when given time to acclimatize, is different to acute hypoxic exposure. In this regard, Zhang et al. (2014) found a significant increase in cardiorespiratory phase synchronization on exposure to acute hypoxia, which was associated with a lower HRV. Thus, variable ventilatory responses to acute hypoxia may confound the measurements of HRV if also obtained in acute hypoxia.

A further problem when comparing studies is that models of the effect of altitude on the autonomic nervous system often only consider the hypoxia variable, although both temperature and barometric pressure may have an independent effect on physiology at altitude (Conkin and Wessel, 2008; Coppel et al., 2015). Moreover, AMS symptoms are less severe in normobaric hypoxia (Roach et al., 1996). Interestingly, the present finding that the best predictor for subsequent AMS is LF/HF ratio in normobaric normoxia suggests that there are physiological differences between AMS-susceptible and AMS-resistant individuals, which are observable even at sea level. The protective nature of a high HRV and a sympathovagal balance shifted to sympathetic dominance may be protective per se, or may reflect some other parameter, such as physical fitness or autonomic activity responsiveness, in particular to stressors such as hypoxia. Other sea level physiological differences have been demonstrated previously, such as gag reflex (Austin and Sleigh, 1995) and resting blood pressure (Lanfranchi et al., 2005). In support of our findings, Akselrod et al. (2001) found that individuals with higher baseline autonomic activity were more resistant to desaturation on exposure to acute hypoxia. Although a recent study by Botek et al. (2015) found no correlation between the LF/HF ratio and delta SpO₂, it was not determined whether LF/HF ratio or delta SpO₂ predicted subsequent AMS, so the lack of correlation may reveal the poor predictive value and the variable nature of the SpO₂ response to acute hypoxia (rather than invalidating HRV measurements for field use).

Oxygen saturations

Measurement of SpO₂ at rest has previously been proposed as an indicator for inadequate acclimatization (Karinen et al., 2010), with several studies finding an inverse relationship between SpO₂ and AMS susceptibility (Karinen et al., 2010; Mairer et al., 2013). More severe desaturation in hypoxia in AMS-susceptible participants is probably due to differences in ventilatory response (Burtscher et al., 2004) or a ventilation/perfusion mismatch (Loeppky et al., 2008). However, measurement of SpO₂ for 15 minutes in acute normobaric hypoxia and chronic hypobaric hypoxia in the present study showed poor diagnostic accuracy. Indeed, other studies have found no association (Wille et al., 2012). Although some authors have proposed that inconsistencies may be due to differences in the time of measurement (SpO₂ measured after >20 minutes in hypoxia, but within the first hour, appears to have superior predictive value than later measurements (Burtscher et al., 2008; Loeppky et al., 2008)), it is an unlikely explanation in the present study because data were also taken after chronic exposure to hypobaric hypoxia and still showed poor diagnostic accuracy. It is conceivable that the peripheral oxygen saturation method may be unreliable even though the present methodology was standardized and likely to be more reliable than if used outside of the research setting.

Strengths and limitations of the study

The rate of ascent and previous acclimatization can affect the likelihood of developing AMS (Bartsch and Saltin, 2008). The ascent profiles for all participant groups were similar and all participants climbed the same route (Fig. 1). None of the participants had been to altitude within the month preceding the expedition. All data were collected at the same time after reaching a given altitude with control of variables such as caffeine, smoking, and eating before collection of HRV data. We also gave advice to ensure that participants stayed well hydrated during the expedition. Thus, many of the variables that can influence AMS and HRV were well controlled (enhancing internal validity). We note that uncontrolled variables such as age and sex can influence HRV; however, the outcome measure of AMS was obtained in the field in a group of heterogeneous participants representative of typical high altitude sojourners (enhancing external validity). We have also provided detailed explanation of the criterion used to exclude HRV data due to artifact or physiologically implausible data. Combined with the use of diagnostic accuracy statistics and individual data plots, the study thus allows a meaningful interpretation of the utility of HRV at different time points (before or during the expedition) to predict AMS.

A major concern regarding the use of HRV to predict AMS susceptibility is that of 35 complete data sets, 6 participants' data had to be excluded due to artifacts or implausible data (Fig. 2), despite controls such as lying still and not talking. Thus, until the data collection equipment is improved or simplified, there will be participants for whom prediction of AMS susceptibility is currently not possible. It is also possible that HRV simply reflects some other unidentified predictor of AMS (but exploratory analyses of other independent variables could not identify any such predictor).

Finally, it is well known that respiratory rate affects HRV and peripheral oxygen saturations. Bilman (2011) discussed the effect of respiratory parameters on HRV, suggesting that both respiratory rate and tidal volume alter HRV, including both the HF and LF components. Lehrer and Gevirtz (2014) suggest that a respiratory rate between 4.5 and 7 breaths per minute improves autonomic homeostasis through HRV biofeedback. They further suggest that breathing at the rate of the cardiovascular system's resonance frequency can enhance performance. However, respiratory rate is grossly affected by altitude (Richard et al., 2014); controlling and thus changing participants' respiratory rate would not reflect their physiological state during the rest of the expedition. Furthermore, it is practical to control respiratory rate in normobaric normoxia, but it may be impossible to control respiratory rate in normobaric or hypobaric hypoxia. Instead, we told participants to lie awake and breath normally, thus avoiding periodic breathing. As for peripheral oxygen saturations, some variability in this measure will have arisen due to using only acute exposure to normobaric hypoxia. Such variability may lead to an underestimation of the diagnostic accuracy of SpO₂. However, previous observations by our group suggest that this variability is considerably reduced at 15 minutes compared with initial exposure. Thus, for both HRV and peripheral oxygen saturation, we chose to make a compromise between scientific rigor and practically implementable methods that would most likely be adopted in the field.

Implications for practice

In the present study, an accurate predictor for AMS susceptibility was identified using a simple tool implementable at sea level. This is important because for a diagnostic tool for AMS to be clinically useful, results must be available in time to make adjustments to prevent disease. For example, lengthening the ascent profile or taking prophylactic mediation. A recent study by Karinen et al. (2012) found root mean square differences of HRV at 2400 m had 92% sensitivity for detecting AMS if ascent continued to 3000–4300 m without acclimatization. However, the need for a field laboratory and technicians at 2400 m, at which point some individuals may already be suffering from AMS (Burtscher et al., 2008) and from which altitude it may be too late to change the itinerary, makes these measurements unwieldy and unsuitable as methods of predicting disease (Song et al., 2013). Even studies finding potentially clinically useful pre-expedition predictors usually require long exposure to simulated altitude (Mairer et al., 2013), which is demanding on time, technology, and finances.

LF/HF ratio of less than 2.28 in normoxia was recognized as being the best cut point to identify individuals at risk of AMS during altitude exposure. This has interesting practical implications because it suggests that hypoxic exposure pre-expedition may not be required to predict AMS susceptibility. This finding is consistent with the results of Rossetti et al. (2016; in press), who found that normoxic VO₂ max predicts altitude performance without measures of hypoxic sensitivity. Our findings also give an insight into cardiac autonomic responses in regard to AMS development and might help to clarify the question of whether HRV modulations appear as a consequence and/or cause of AMS.

Currently, however, postprocessing of the HRV data is still required, making interpretations of the results difficult without some technical knowledge. In addition, until equipment is improved, any HRV test should last at least 10 minutes to allow for removal of artifact from the recording. Modification of existing heart rate monitors should be considered, to give information to users pre-expedition on their susceptibility to AMS at true altitude, by measuring LF/HF ratio in normoxia. A simple diagnostic test incorporated into a watch has the potential to reduce AMS by allowing implementation of preventative measures, but should be used with caution to prevent participants who are predicted to be AMS resistant from ignoring general advice on AMS prevention. Finally, at present, some overlap in HRV responses between AMS-susceptible and AMS-resistant individuals exists. Therefore, further research is needed to validate our findings with an increased number of more heterogeneous participants, with varying degrees of AMS, to determine the true utility of HRV data to predict AMS.

Conclusion

This study showed that HRV has better diagnostic accuracy for AMS prediction than peripheral SpO₂ saturation monitoring. These results show that the measurement with the highest diagnostic accuracy is normoxic LF/HF ratio, measured pre-expedition. Further confirmatory studies are needed to validate these measurements, but they suggest an alternative approach to obtain a more accurate simple method of predicting AMS.

Footnotes

Acknowledgments

The authors would like to thank all research participants who took part in the MEDEX 2015 expedition.

Author Disclosure Statement

No competing financial interests exist.

References

Akselrod

, Barak

, Ben-Dov

, Keselbrener

, and Baharav

. (2001). Estimation of autonomic response based on individually determined time axis. Auton Neurosci, 90:13–23.

Austin

, and Sleigh

. (1995). Prediction of acute mountain sickness. BMJ, 311:989–990.

Bartsch

, and Saltin

. (2008). General introduction to altitude adaptation and mountain sickness. Scand J Med Sci Sports, 18 Suppl 1:1–10.

Billman

. (2011). Heart rate variability—A historical perspective. Front Physiol, 2:86.

Billman

. (2013). The LF/HF ratio does not accurately measure cardiac sympatho-vagal balance. Front Physiol, 4:26.

Botek

, Krejci

, De Smet

, Gaba

, and Mckune

. (2015). Heart rate variability and arterial oxygen saturation response during extreme normobaric hypoxia. Auton Neurosci, 190:40–45.

Brierley

, Parks

, and Wolff

. (2012). The relationship of acute mountain sickness to arterial oxygen saturation at altitudes of 3324 to 5176 m. In: Oxygen Transport to Tissue XXXIII. Wolf

, Bucher

, Rudin

, Van Huffel

, Wolf

, Bruley

, and Harrison

, eds. Springer New York, New York. pp. 207–212.

Burtscher

, Flatz

, and Faulhaber

. (2004). Prediction of susceptibility to acute mountain sickness by SaO2 values during short-term exposure to hypoxia. High Alt Med Biol, 5:335–340.

Burtscher

, Szubski

, and Faulhaber

. (2008). Prediction of the susceptibility to AMS in simulated altitude. Sleep Breath, 12:103–108.

10.

Camm

, Malik

, and Bigger

. (1996). Heart rate variability: Standards of measurement physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Circulation, 93, 17:354–381.

11.

Chalmers

, Quintana

, Abbott

, and Kemp

. (2014). Anxiety disorders are associated with reduced heart rate variability: A meta-analysis. Front Psychiatry, 5:80.

12.

Chen

, Lin

, Shiao

, and Chang

. (2008). Effect of rapid ascent to high altitude on autonomic cardiovascular modulation. Am J Med Sci, 336:248–253.

13.

Chiu

, Wang

, Huang

, Tso

, and Kao

. (2003). The influence of mean heart rate on measures of heart rate variability as markers of autonomic function: A model study. Med Eng Phys, 25:475–481.

14.

Cohen

, and Taylor

. (2002). Short-term cardiovascular oscillations in man: Measuring and modelling the physiologies. J Physiol, 542:669–683.

15.

Conkin

, and Wessel

. (2008). Critique of the equivalent air altitude model. Aviat Space Environ Med, 79:975–982.

16.

Coppel

, Hennis

, Gilbert-Kawai

, and Grocott

. (2015). The physiological effects of hypobaric hypoxia versus normobaric hypoxia: A systematic review of crossover trials. Extrem Physiol Med, 4:2.

17.

De Vito

, Galloway

, Nimmo

, Maas

, and Mcmurray

. (2002). Effects of central sympathetic inhibition on heart rate variability during steady-state exercise in healthy humans. Clin Physiol Funct Imaging, 22:32–38.

18.

Dos Santos

, Barroso

, Macau

, and De Godoy

. (2013). Application of an automatic adaptive filter for heart rate variability analysis. Med Eng Phys, 35:1778–1785.

19.

Duplain

, Vollenweider

, Delabays

, Nicod

, Bartsch

, and Scherrer

. (1999). Augmented sympathetic activation during short-term hypoxia and high-altitude exposure in subjects susceptible to high-altitude pulmonary edema. Circulation, 99:1713–1718.

20.

Hackett

. (1999). The cerebral etiology of high-altitude cerebral edema and acute mountain sickness. Wilderness Environ Med, 10:97–109.

21.

Honigman

, Theis

, Koziol-Mclain

, Roach

, Yip

, Houston

, Moore

, and Pearce

. (1993). Acute mountain sickness in a general tourist population at moderate altitudes. Ann Intern Med, 118:587–592.

22.

Huang

, Tseng

, Fan

, Yen

DHT

, Kao

, Chang

, Kuo

TBJ

, Huang

, and Lee

. (2010). Alternations of heart rate variability at lower altitude in the predication of trekkers with acute mountain sickness at high altitude. Clin J Sport Med, 20:58–63.

23.

Kanai

, Nishihara

, Shiga

, Shimada

, and Saito

. (2001). Alterations in autonomic nervous control of heart rate among tourists at 2700 and 3700 m above sea level. Wilderness Environ Med, 12:8–12.

24.

Karinen

, Peltonen

, Kahonen

, and Tikkanen

. (2010). Prediction of acute mountain sickness by monitoring arterial oxygen saturation during ascent. High Alt Med Biol, 11:325–332.

25.

Karinen

, Uusitalo

, Vaha-Ypya

, Kahonen

, Peltonen

, Stein

, Viik

, and Tikkanen

. (2012). Heart rate variability changes at 2400 m altitude predicts acute mountain sickness on further ascent at 3000–4300 m altitudes. Front Physiol, 3:336.

26.

Kobayashi

. (2009). Does paced breathing improve the reproducibility of heart rate variability measurements?. J Physiol Anthropol, 28:225–230.

27.

Koehle

, Guenette

, and Warburton

. (2010). Oximetry heart rate variability and the diagnosis of mild-to-moderate acute mountain sickness. Eur J Emerg Med, 17:119–122.

28.

Kox

, Pompe

, Van Der Hoeven

, Hoedemaekers

, and Pickkers

. (2011). Influence of different breathing patterns on heart rate variability indices and reproducibility during experimental endotoxaemia in human subjects. Clin Sci, 121:215–222.

29.

La Rovere

, Bigger

Jr , Marcus

, Mortara

, and Schwartz

. (1998). Baroreflex sensitivity and heart-rate variability in prediction of total cardiac mortality after myocardial infarction ATRAMI (Autonomic Tone and Reflexes After Myocardial Infarction). Investigators Lancet, 351:478–484.

30.

Lake Louise Consensus on the Definition and Quantification of Altitude Illness. (1991). Hypox Mount Med. Proceedings of the 7th International Hypoxia Symposium, Lake Louise, Canada, February.

31.

Lanfranchi

, Colombo

, Cremona

, Baderna

, Spagnolatti

, Mazzuero

, Wagner

, Perini

, Wagner

, Cavallaro

, and Giannuzzi

. (2005). Autonomic cardiovascular regulation in subjects with acute mountain sickness. Am J Physiol Heart Circ Physiol, 289:H2364–H2372.

32.

Lehrer

, and Gevirtz

. (2014). Heart rate variability biofeedback: how and why does it work?. Front Psychol, 5:756.

33.

Lippman

, Stein

, and Lerman

. (1994). Comparison of methods for removal of ectopy in measurement of heart rate variability. Am J Physiol, 267:H411–H418.

34.

Liu

, Batchinsky

, Cancio

, and Salinas

. (2013). The impact of noise on the reliability of heart-rate variability and complexity analysis in trauma patients. Comput Biol Med, 43:1955–1964.

35.

Loeppky

, Icenogle

, Charlton

, Conn

, Maes

, Riboni

, Gates

, Melo

, and Roach

. (2008). Hypoxemia and acute mountain sickness: Which comes first?. High Alt Med Biol, 9:271–279.

36.

Loeppky

, Icenogle

, Maes

, Riboni

, Scotto

, and Roach

. (2003). Body temperature autonomic responses and acute mountain sickness. High Alt Med Biol, 4:367–373.

37.

López-Ratón

, Rodríguez-Álvarez

. OptimalCutpoints: computing optimal cutpoints in diagnostic tests R package version 11–13 2014. http://CRANR-projectorg/package=OptimalCutpoints (last accessed November 19, 2016).

38.

Luks

, and Swenson

. (2011). Pulse oximetry at high altitude. High Alt Med Biol, 12:109–119.

39.

Maggiorini

, Bühler

, Walter

, and Oelz

. (1990). Prevalence of acute mountain sickness in the Swiss Alps. BMJ, 301:853–855.

40.

Mairer

, Wille

, Grander

, and Burtscher

. (2013). Effects of exercise and hypoxia on heart rate variability and acute mountain sickness. Int J Sports Med, 34:700–706.

41.

Malliani

, Pagani

, Lombardi

, and Cerutti

. (1991). Cardiovascular neural regulation explored in the frequency domain. Circulation, 84:482–492.

42.

Obuchowski

, Lieber

, and Wians

Jr . (2004). ROC curves in clinical chemistry: Uses misuses and possible solutions. Clin Chem, 50:1118–1125.

43.

Oliver

, Sanders

, Williams

, Smith

, Lloyd-Davies

, Roberts

, Arthur

, Hardy

, and Macdonald

. (2012). Physiological and psychological illness symptoms at high altitude and their relationship with acute mountain sickness: A prospective cohort study. J Travel Med, 19:210–219.

44.

Rahman

, Pechnik

, Gross

, Sewell

, and Goldstein

. (2011). Low frequency power of heart rate variability reflects baroreflex function not cardiac sympathetic innervation. Clin Auton Res, 21:133–141.

45.

Richard

, Sahota

, Widmer

, Ferguson

, Sheel

, and Koehle

. (2014). Acute mountain sickness chemosensitivity and cardiorespiratory responses in humans exposed to hypobaric and normobaric hypoxia. J Appl Physiol, 116:945–952.

46.

Roach

, Loeppky

, and Icenogle

. (1996). Acute mountain sickness: Increased severity during simulated altitude compared with normobaric hypoxia. J Appl Physiol, 81:1908–1910.

47.

Rossetti

GMK

, Macdonald

, Smith

, Jackson

, Callender

, Newcombe

, Storey

, Willis

, van den Beukel

, Woodward

, Pollard

, Wood

, Newton

, Virian

, Haswell

, and Oliver

. MEDEX2015: Greater sea-level fitness is associated with lower sense of effort during Himalayan trekking without worse Acute Mountain Sickness. High Alt Med Biol. In Press; DOI: 10.1089/ham.2016.0088

48.

Scheff

, Griffel

, Corbett

, Calvano

, and Androulakis

. (2014). On heart rate variability and autonomic activity in homeostasis and in systemic inflammation. Math Biosci, 252:36–44.

49.

Shaffer

, Mccraty

, and Zerr

. (2014). A healthy heart is not a metronome: An integrative review of the heart's anatomy and heart rate variability. Front Psychol, 5:1040.

50.

Song

, Ke

, Luo

W-J

, and Chen

J-Y

. (2013). Non-high altitude methods for rapid screening of susceptibility to acute mountain sickness. BMC Public Health, 13:902.

51.

Tarvainen

, Niskanin

, Lipponen

, Ranta-aho

, Karjalainen

. (2014). Kubios HRV-Heart rate variability analysis software. Comput Methods Programs Biomed, 113:210–220.

52.

Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. (1996). Heart rate variability. Standards of measurement physiological interpretation and clinical use. Eur Heart J, 17:354–381.

53.

Wessel

, Voss

, Malberg

, Ziehmann

, Voss

, Schirdewan

, Meyerfeldt

, and Kurths

. (2000). Nonlinear analysis of complex phenomena in cardiological data. Herzschrittmacherther Elektrophysiol, 11:159–173.

54.

Wille

, Mairer

, Gatterer

, Philippe

, Faulhaber

, and Burtscher

. (2012). Changes in cardiac autonomic activity during a passive 8 hour acute exposure to 5 500 m normobaric hypoxia are not related to the development of acute mountain sickness. Int J Sports Med, 33:186–191.

55.

Zhang

, She

, Zhang

, and Yu

. (2014). Effects of acute hypoxia on heart rate variability sample entropy and cardiorespiratory phase synchronization. Biomed Eng Online, 13:73.