Objective Versus Self-Reported Sleep Quality at High Altitude

Introduction

Sleep disruption is virtually universal for those arriving at high altitude (>7500 ft), making it one of the most prominently reported symptoms of acclimatization to hypobaric hypoxia. The question on self-reported sleep quality is one of five questions on the Lake Louise Symptom Questionnaire (LLSQ), along with questions about symptoms of headache, gastrointestinal distress (nausea), dizziness, and fatigue. For the sleep quality question, subjects report whether they slept as well as usual (0 points), did not sleep as well as usual (1 point), woke many times/had a poor night's sleep (2 points), or did not sleep at all (3 points). In the presence of a headache, acute mountain sickness (AMS) has classically been diagnosed for Lake Louise Score (LLS) ≥3, but some research uses LLS ≥4 or LLS ≥5 to diagnose AMS (Nussbaumer-Ochsner et al., 2011; Frühauf et al., 2016). High-altitude research studies that use AMS (diagnosed by LLSQ) as the primary outcome depend on self-reported sleep problems to calculate the LLS.

In low altitude settings, Weaver et al. reported poor correlations between polysomnographic data (e.g., apnea hypopnea index [AHI]) and self-reported symptoms such as sleepiness, general health, and mental health (Weaver et al., 2004). Lauderdale et al. (2008) underscored this finding by showing that self-reported sleep duration has only a modest correlation with sleep duration measured by actigraphy.

In high-altitude settings, Nussbaumer-Ochsner et al. also found no correlation between actigraphy data and subjective reporting of sleep quality and sleep duration (Nusbaumer-Ochsner et al., 2011). Furthermore, several high-altitude sleep studies have been unable to demonstrate a consistent connection between disrupted sleep measured by polysomnography and the development of AMS using the LLSQ. (Burgess et al., 2004; Anderson et al., 2015; Tseng et al., 2015). MacInnis et al. (2013) directly evaluated the internal consistency of the LLSQ and concluded that the sleep quality item is only weakly related to other items of the LLSQ and that it may be better to consider sleep problems separately from other symptoms of high-altitude illness. For these and other reasons, some have suggested that the sleep question be removed from the LLSQ altogether (Milledge, 2014). To directly explore the relationship between objective and subjective sleep quality at high altitude, we reanalyzed data collected during our previous study of sleep disruption and AMS in workers rapidly transported to the South Pole (Anderson et al., 2015).

Methods

We reanalyzed sleep data and LLSQ responses for the 38 individuals examined in our 2015 study, none of whom reported taking acetazolamide, and also included an additional 25 individuals who reported taking acetazolamide during the study period for a total of 63 subjects. Subject recruitment, IRB approvals, and acetazolamide usage have been described in detail elsewhere (Anderson et al., 2011). In brief, subjects completed a 3-hour flight from McMurdo Base (sea level) to the Amundsen-Scott South Pole station (3200 m, 9800 ft) and completed overnight polysomnography on their first night at high altitude using the Vivometrics Lifeshirt, which is a field polysomnography device with EEG leads, chest and abdominal bands, and finger oximetry (Grossman, 2004; Heilman and Porges, 2007). Acetazolamide usage was optional. We compared overnight sleep study results with sleep question responses on the LLSQ from day 1, the morning immediately after their sleep study was conducted.

We used self-reported sleep quality score (0 = no problems, 1 = some problems, 2 = poor sleep, 3 = did not sleep at all) to create comparison groups for key polysomnographic outcomes. Categorical data were described as count (percentage). Continuous data were described as mean ± standard deviation or median (25th percentile, 75th percentile), based on the absolute value of the skewness coefficient being ≤1 versus >1. Demographic data were described and compared according to the LLSQ result, by the chi-square, one-way ANOVA, or by the Kruskal–Wallis test. Polysomnographic data were described and compared, both by LLSQ result and by high-altitude medication (acetazolamide) use. Hypothesis testing was carried out for polysomnographic data, and compared by acetazolamide use, using Wilcoxon rank sum test. Comparison of the three key outcomes between pairs of LLSQ sleep quality groups was done with univariate ordinal logistic regression and univariate logistic regression, as appropriate.

Results

Of the 63 participants, 22 (35%) reported no problems with sleep, 20 (32%) reported some problems, and 21 (33%) reported a poor sleep. Only one of our participants reported that he could not sleep at all (sleep question score = 3) so he was included in the poor sleep group (no sleep, n = 1). Twenty-five (25/63, 40%) of 63 participants took acetazolamide, with 10 (40%), 9 (36%), and 6 (24%) participants reporting no problems, some problems, or poor sleep, respectively. Overall, participants were predominantly male (n = 41, 65%) and had healthy mean body mass index (25.4 ± 2.6 kg/m²) and healthy body fat% (median 19.9%, interquartile range [IQR] 13.9, 23.1). The majority of subjects had lived <5000 ft. during the previous 3 months (n = 45, 71%). More than half of the subjects had never been to the South Pole before (n = 36, 57%) and more than half had previous problems with illness at high altitude (n = 35, 55%). In general, we found no statistically significant differences (p < 0.05) in demographic data or baseline physical characteristics between the three respondent groups (Table 1).

Discussion

Our data reveal that subjects report varying degrees of sleep disturbance, but when analyzed by these responses, they show no statistically significant differences in selected objective sleep outcomes (AHI, overnight oxygen saturation, and sleep efficiency). This finding is relevant for high-altitude studies that use AMS (scored by Lake Louise Criteria) as a primary outcome, and especially those that examine the relationship between polysomnography and subsequent development of AMS.

Subjective sleep disturbances at high altitude are very common and were one of the most frequently reported symptoms on the LLSQ during our initial study of acclimatization to high altitude at the South Pole (Anderson et al., 2011). This was confirmed by a study using more robust sleep questionnaires (Pittsburgh Sleep Quality Index, PSQI; Athens Insomnia Scale, AIS-8) that also reports decreased subjective sleep quality, especially reduced general sleep quality, and prolonged sleep induction (Syzmczak et al., 2009).

Objective sleep disturbance is also prominent during acclimatization. A recent minireview of the high-altitude sleep literature concludes that during rapid ascent to high altitude, there is a reduction in total sleep time, sleep efficiency, deep sleep (non-rapid eye movement [NREM] stages 3 and 4), and a significant increase in arousals and periodic breathing (Bloch et al., 2015). These effects are likely high altitude dependent and the effects tend to moderate with acclimatization (Bloch et al., 2010).

Even though objective sleep quality and subjective sleep quality are degraded at high altitude, attempts to find a correlation between objective and subjective measures have failed to find a connection. One study of sleep in mountaineers with AMS at 3250 m revealed a minor reduction in rapid eye movement (REM) sleep but an otherwise similar sleep structure when compared with those without AMS (Tseng et al., 2015). In our previous study of sleep and AMS in workers rapidly transported from sea level to the South Pole, we found no statistically significant differences in polysomnography between polar workers with AMS and those without AMS (Anderson et al., 2015). Nussbaumer-Ochsner et al. also found that subjective reporting of sleep quality and sleep duration did not correlate with objective measures such as actigraphy and polysomnography (Nussbaumer-Ochsner et al., 2011). The absence of a strong correlation between polysomnography and AMS status in these studies indicates that the self-reported sleep question on the LLSQ likely creates a nondifferential misclassification bias, which pushes results toward the null hypothesis.

Our Antarctic data may be limited somewhat by the fact that the Vivometrics LifeShirt has not been validated for full polysomnography. Scoring a clinical polysomnogram is itself an imprecise process and interscorer agreement has been estimated at 85%, which sets an upper limit on the requirements for reliability during validation (Kelly et al., 2012). Our results were overread by board-certified sleep specialists for accuracy. Another limitation might be the use of only three parameters (AHI, sleep efficiency, and overnight oxygen saturation) to evaluate sleep quality and the question of whether or not these measures provide an appropriate appraisal of sleep quality. The two nonhigh-altitude studies from the sleep medicine literature that evaluate the relationship between subjective sleep quality and objective sleep quality both use these measures of sleep disturbance (AHI) or sleep duration (Sleep efficiency) in their analysis (Weaver et al., 2004; Lauderdale et al., 2008).

As has been observed in both high-altitude and low altitude settings, our data indicate that self-reported sleep is a poor proxy for objectively measured sleep at high altitude. In addition, the self-reported sleep question on the LLSQ may create a nondifferential misclassification bias in analyses that explore the relationship between objective sleep disruption and AMS because individuals who reported no problems sleeping would be less likely to meet criteria for AMS even though they may have had disruptive sleep by objective measures. Macinnis et al. list a number of circumstances that make the sleep quality question a problematic outcome for high-altitude research and our results provide additional objective data to support these observations (Macinnis et al., 2013; Milledge, 2014). Those conducting future research on the relationship between polysomnography and AMS (using the LLSQ) should exclude the sleep question on the LLSQ from AMS scoring.