Phenotyping interindividual variability in obstructive sleep apnoea response to temazepam using ventilatory chemoreflexes during wakefulness

Authors

  • DAVID WANG,

    1. Department of Respiratory and Sleep Medicine, Royal Prince Alfred Hospital, Camperdown, Sydney, NSW, Australia
    2. Woolcock Institute of Medical Research, University of Sydney, Glebe, NSW, Australia
    Search for more papers by this author
  • NATHANIEL S. MARSHALL,

    1. Woolcock Institute of Medical Research, University of Sydney, Glebe, NSW, Australia
    2. Centre for Integrated Research and Understanding of Sleep (CIRUS), University of Sydney, Sydney, NSW, Australia
    Search for more papers by this author
  • JAMES DUFFIN,

    1. Department of Anaesthesia and Physiology, University of Toronto, Toronto, ON, Canada
    Search for more papers by this author
  • BRENDON J. YEE,

    1. Department of Respiratory and Sleep Medicine, Royal Prince Alfred Hospital, Camperdown, Sydney, NSW, Australia
    2. Woolcock Institute of Medical Research, University of Sydney, Glebe, NSW, Australia
    3. Centre for Integrated Research and Understanding of Sleep (CIRUS), University of Sydney, Sydney, NSW, Australia
    Search for more papers by this author
  • KEITH K. WONG,

    1. Department of Respiratory and Sleep Medicine, Royal Prince Alfred Hospital, Camperdown, Sydney, NSW, Australia
    2. Woolcock Institute of Medical Research, University of Sydney, Glebe, NSW, Australia
    3. Centre for Integrated Research and Understanding of Sleep (CIRUS), University of Sydney, Sydney, NSW, Australia
    Search for more papers by this author
  • NARGIS NOORI,

    1. Woolcock Institute of Medical Research, University of Sydney, Glebe, NSW, Australia
    Search for more papers by this author
  • SUSANNA S. W. NG,

    1. Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, NT, Hong Kong, China
    Search for more papers by this author
  • RONALD R. GRUNSTEIN

    1. Department of Respiratory and Sleep Medicine, Royal Prince Alfred Hospital, Camperdown, Sydney, NSW, Australia
    2. Woolcock Institute of Medical Research, University of Sydney, Glebe, NSW, Australia
    3. Centre for Integrated Research and Understanding of Sleep (CIRUS), University of Sydney, Sydney, NSW, Australia
    Search for more papers by this author

Dr David Wang, Department of Respiratory and Sleep Medicine, Royal Prince Alfred Hospital, Camperdown, Sydney 2050, NSW, Australia. Tel.: +61-2-9114-0446; fax: +61-2-9114-0014; e-mail: david.wang@sydney.edu.au

Summary

Centrally active agents have a variable impact in patients with obstructive sleep apnoea (OSA) that is unexplained. How to phenotype the individual OSA response is clinically important, as it may help to identify who will be at risk of respiratory depression and who will benefit from a centrally active agent. Based on loop gain theory, we hypothesized that OSA patients with higher central chemosensitivity have higher breathing instability following the use of a hypnosedative, temazepam. In 20 men with OSA in a double-blind, placebo-controlled cross-over trial we tested the polysomnographically (PSG) measured effects of temazepam 10 mg versus placebo on sleep apnoea. Treatment nights were at least 1 week apart. Ventilatory chemoreflexes were also measured during wakefulness in each subject. The patients (mean ± standard deviation; 44 ± 12 years) had predominantly mild-to-moderate OSA [baseline apnoea–hypopnoea index (AHI) = 16.8 ± 14.1]. Patients’ baseline awake central chemosensitivity correlated significantly with both the change of SpO2 nadir between temazepam and placebo (r = −0.468, = 0.038) and oxygen desaturation index (ODI; = 0.485, = 0.03), but not with the change of AHI (= 0.18, = 0.44). Peripheral chemosensitivity and ventilatory recruitment threshold were not correlated with the change of SpO2 nadir, ODI or AHI (all > 0.05). Mild–moderate OSA patients with higher awake central chemosensitivity had greater respiratory impairment during sleep with temazepam. Relatively simple daytime tests of respiratory control may provide a method of determining the effect of sedative–hypnotic medication on breathing during sleep in OSA patients.

Introduction

Obstructive sleep apnoea (OSA) is a common disorder for which there is no current established pharmacological therapy (Hedner et al., 2008). Given the variable effectiveness of continuous positive airway pressure (CPAP) therapy, identification of pharmacological treatment for OSA remains an important objective (Weaver and Grunstein, 2008). Previous studies investigating the effects of a range of pharmaceutical agents in OSA have shown marked interindividual variation between patients. The notable recent example of this has been the disparate trial findings regarding the effects of the anti-depressant, mirtazapine, which has been reported to improve (Carley et al., 2007), worsen and have no effect (Marshall et al., 2008) on OSA severity. Similar divergent results have been found with a range of pharmaceutical agents and even low flow oxygen (Dempsey et al., 2010; Grunstein et al., 1994; Hedner et al., 2008; Wellman et al., 2008).

Hypnosedatives and other central nervous system (CNS) depressants are generally considered to produce respiratory depression during sleep. However, in patients with milder OSA, wide unexplained interindividual variability in the effect of centrally active agents on breathing during sleep have been reported (Hoijer et al., 1994; Rosenberg et al., 2007). This was exemplified by one randomized cross-over trial study where patients were given nitrazepam 5 and 10 mg and placebo on three separate nights. The drug appeared to improve OSA in some patients and worsen it in others (Hoijer et al., 1994). This may suggest there are different functional phenotypes of sleep apnoea that are associated with varying responses to hypnosedatives.

The ability to phenotype this response will be clinically important, as it may help to identify patients at risk of respiratory depression and others who may benefit from the use of such a pharmaceutical agent. To our knowledge, no technique has been used to clinically phenotype the interindividual response to CNS-active pharmaceutical agents in OSA patients. A number of factors could influence this response, including gender, race, obesity, ageing, craniofacial structure, ventilatory and upper airway function, sleep regulation and others (Riha et al., 2009). It is likely that the variability in response is influenced mainly by individual upper airway function and the compensatory control responses to airway narrowing, such as chemical drives and thresholds including arousal threshold. Upper airway phenotyping techniques such as magnetic resonance (MR) imaging or critical airway pressure measurements involve complex methodology and are problematic. First, OSA severity can vary considerably among patients with the same degree of pharyngeal collapsibility (Isono et al., 1997; Younes, 2003). Secondly, a number of fixed skeletal and soft tissue features have been identified that may account for the differences in passive mechanical behaviour: those features may change in the same individual in response to changes in body and neck position, lung volume, oedema, vascularity and surface tension properties (Isono et al., 1997; Younes, 2003). Thirdly, current techniques for testing airway patency are complicated, costly and labour-intensive and are not yet suitable for widespread general clinical use (Eckert et al., 2009).

The compensatory physiological responses to an abnormally collapsible airway might determine OSA more powerfully than any anatomic problem (Isono et al., 1997; Loewen et al., 2009; Younes, 2003). A collapsible airway may allow for one significant airway narrowing or obstruction, but any subsequent repetitive cycling behaviour in ventilation has to depend upon an altered neurochemical control mechanism (Dempsey et al., 2004; White, 2005; Younes, 2003). Recent discussion of the ‘loop gain’ theory of OSA pathogenesis suggests that an obstructed airway can stimulate an increased loop gain/controller gain/chemical drive. In turn, this can create ventilation ‘overshoot’/instability and predispose to a cyclical pattern of breathing (Hudgel et al., 1998; Khoo, 2000; Wang et al., 2007; White, 2005; Younes et al., 2001). Conversely, the increased chemical drive may be protective and serve as a compensatory mechanism for the obstructed airway (Younes, 2008). If upper airway compromise is not resolved and the compensatory increased chemical drive is depressed, then patients will be at risk of hypoventilation and impaired gas exchange. We therefore speculate that there could be an intrinsic linkage between individual response to CNS-sedative agents and awake ventilatory chemosensitivity in OSA patients. In addition, the ventilatory recruitment threshold may also influence the variability, because the chemical drive recruitment threshold (TER) during sleep has been identified as a key factor for abnormal pharyngeal oscillation (Dempsey et al., 2004; Younes, 2008). The relationship between TER and arousal threshold plays a particularly important role in ventilator instability (Younes, 2008). While arousal threshold is unlikely to be tested easily during awake, TER can be estimated by assessing ventilatory recruitment threshold as a part of our awake chemoreflex test (Mateika et al., 2004).

Following the above theoretical framework, we hypothesize that individual OSA patient response to temazepam is predicted by ventilatory chemosensitivities and/or ventilatory recruitment threshold. The apnoea–hypopnoea index (AHI) was the primary outcome variable, but oxygenation-related PSG parameters were also assessed carefully. If the hypotheses can be verified, then ventilatory chemoreflex testing may become a simple daytime tool to help identify those OSA patients at risk from CNS depressants and those who may actually benefit from them.

Methods

The study was conducted at the clinical sleep laboratory of Royal Prince Alfred Hospital (RPAH), a major teaching hospital of the University of Sydney, Australia. The study protocol was approved by the Sydney South West Area Health Service (SSWAHS) Ethics Review Committee that supervised the RPAH (Protocol no.: X08-0079). Written consent forms were signed by all patients prior to the study. The Australian Clinical Trial Registry number is ACTRN12608000318381.

Patients

Twenty-two men with predominantly mild-to-moderate OSA were recruited from the sleep clinics of the Royal Prince Alfred Hospital and the associated Woolcock Institute of Medical Research. Patients were either diagnosed previously with mild-to-moderate OSA (AHI 5–30 h−1) or were referred to the sleep laboratory with a preliminary diagnosis of ‘mild-to-moderate OSA’ (based on referring sleep physicians’ clinical judgement that a PSG was required). Only men were included, due to potential ventilatory chemoreflex changes in women with the menstrual cycle (White et al., 1983). Patients were excluded if they were suffering from any uncontrolled concurrent medical or psychiatric illness or if they were taking any concurrent medications that are known to affect sleep respiration or interact with temazepam; had medical conditions that would contraindicate temazepam; or had other major sleep disorders such as periodic limb movement syndrome (PLMS) or irregular sleep patterns, such as shift-workers. All patients underwent a thorough physical examination by a respiratory/sleep physician prior to entering the study.

Protocol

We used a randomized (1 : 1 ratio) double-blind, placebo-controlled cross-over design to investigate the effects of 10 mg temazepam compared to placebo on a single night of sleep measured by PSG with at least a 1-week washout between arms. Before the first sleep study and before administration of either drug, all patients were tested for their ventilatory chemoreflexes between 15:00 and 16:00 h. They were asked to take either drug at 21:00 h, and have their sleep monitored from 22:00 h (lights-off time). The randomization sequence was generated by a research pharmacist at Royal Prince Alfred Hospital Pharmacy, who also dispensed treatment and who never met any patient and only disclosed the sequence to us after we had completed data collection, processing including all sleep scoring. The data analyst was not blinded to treatment allocation.

Ventilatory chemoreflex testing

Assessment of ventilatory chemoreflexes provides important information on the function of central and peripheral chemoreceptors, the control system for maintaining PaCO2 and PaO2 homeostasis. The theoretical basis (Lloyd, 1966) and practical application (Read, 1967) have been described elsewhere. Recently, an improved modified chemoreflex test (Duffin et al., 2000) has been introduced which has several advantages: (i) direct testing of ventilatory recruitment threshold, which is also a key parameter in ventilatory chemoreflex (Duffin, 2007; Duffin et al., 2000) and (ii) employing iso-oxia (holding oxygen constant) throughout the test, which can avoid a confounding response from the change of oxygen level in the original Read method (Duffin et al., 2000). We adopted this testing method and built a fully computerized system which can test both ventilatory recruitment threshold and central and peripheral chemosensitivities (Duffin et al., 2000). Central chemosensitivity was determined by testing iso-oxic hyperoxic ventilatory response to CO2. It was a 10-min test, including 5 min of hyperventilation and 5 min of rebreathing through a closed circuit. During the 5-min hyperventilation, end-tidal PCO2 was controlled between 19 and 25 mm Hg. The computer then switched the valve and the patient rebreathed for 5 min through a bag containing a mix gas of 6% of CO2 and 94% O2. The PO2 in the circuit was held constant at 150 mm Hg. The computer analysed O2 consumption continuously over the past three breaths and used a prediction model to determine how much O2 to feed into the circuit. The ventilatory recruitment threshold (PCO2) and central chemosensitivity (the slope of PCO2 plotted against minute ventilation) were analysed through purpose-built software (Fig. 1). After 45 min of sedentary rest, patients were retested for peripheral chemosensitivity: the iso-oxic hypoxic ventilatory response to CO2. The testing procedure was otherwise identical to the first test, except that the gas mix was 6% of CO2 and 94% N2, and the PO2 in the circuit was held constant at 50 mm Hg (Fig. 1).

Figure 1.

 An example of the ventilatory chemoreflex analyses from our data. Slope (a) is the ‘central chemosensitivity’. Slope (b) is the ‘mixed chemosensitivity’. Slope (b) − Slope (a) = ‘peripheral chemosensitivity’. (T1) is the ‘CO2 recruitment threshold’ used for data analyses. (a) Hyperoxic Ventilatory response to CO2; (b) Hypoxic Ventilatory response to CO2.

Overnight PSG

Attended full PSGs (Alice 5; Respironics/Philips, Andover, MA, USA) were monitored using standard methods. Measurements include four channels of electroencephalogram (EEG), two channels of electro-oculogram (EOG), chin electromyogram (EMG), anterior tibial EMG, electrocardiogram (ECG), body position, nasal pressure, chest and abdomen movements and SpO2. All PSG studies were scored blindly by an experienced sleep technologist using Rechtschaffen and Kales (1968) criteria. Respiratory events and arousals were scored according to standard Chicago and American Sleep Disorders Association (ASDA) criteria, respectively (AASM Task Force, 1999; American Sleep Disorders Association, 1992). AHI was calculated by dividing the total number of apnoeas and hypopnoeas by the total sleep time (h). Oxygen desaturation index (ODI) was calculated by dividing the total number of ≥ 3% SpO2 drop by the total sleep time (h).

Statistical analyses

For sample size calculation, in the absence of previous data on which to base the sample size estimate for our primary analysis, we based our power calculation on analysis of the treatment effect of temazepam on sleep apnoea, as measured by the AHI. Assuming a standard deviation of the difference in AHI of 5.0 (calculated from Table 3 of Hoijer et al., 1994) with 20 patients in a cross-over study would be able to detect a difference in AHI of 3.3 events h−1, with a two-tailed significance level of 0.05 and 80% power.

Descriptive data were expressed as mean ± standard deviation (SD), unless stated otherwise. Comparisons between temazepam and placebo were tested with paired t-tests. However, the primary outcomes in this trial were not head-to-head comparisons of drug and placebo, but rather correlations between baseline wake phenotyping of respiratory responses and the placebo-adjusted responses to the drug during sleep (temazepam night – placebo night). These correlations were tested by either Pearson’s or Spearman’s tests, as appropriate. We used linear regression to examine whether changes in rapid eye movement (REM) sleep time caused by temazepam mediated the chemosensitivity to OSA relationships. Analyses were performed using spss version 17 (SPSS, Chicago, IL, USA). A P-value of < 0.05 was considered significant.

Results

We randomized 22 patients and 20 completed the protocol. One patient dropped out due to moving interstate. Another patient was excluded due to significant PLMS that became evident only on the first study after randomization. Patients’ average age was 44 ± 11.6 (range 19–63) years. Their average body mass index (BMI) was (mean ± SD) 27.5 ± 7.5 kg m2. Thirteen had an AHI < 15, four had an AHI between 15 and 30 and three had an AHI > 30. Epworth Sleepiness Scale scores (ESS) averaged 9.4 ± 4.3, with 12 having ESS ≤ 10, six an ESS between 11 and 15 and two an ESS ≥ 16.

Baseline central chemosensitivity averaged 2.13 ± 0.90 l min−1 mm Hg−1, peripheral chemosensitivity 2.36 ± 1.53 l min−1 mm Hg−1, and ventilatory recruitment threshold (PCO2) 41.25 ± 5.29 mm Hg. An example of the ventilatory chemoreflex analysis is shown in Fig. 1.

PSG data from active and placebo temazepam nights are compared in Table 1. After taking 10 mg temazepam, patients had significantly reduced REM sleep and tended to have reduced Stage 3 sleep and increased Stage 2 sleep. Overall, temazepam did not change patients’ AHI, SpO2 nadir or ODI. However, the high SDs indicated the expected large interindividual variability (see Fig. 2). SpO2 nadir is notable, in that there are individuals with both large increases and large decreases after using temazepam, adjusted for placebo.

Table 1.   Comparison of the effects of temazepam and placebo on polysomnographically measured sleep characteristics
 TemazepamPlaceboP
  1. TST, total sleep time; S1%, Stage 1%; REM%, rapid eye movement sleep %; AHI, Apnoea–Hypopnoea Index; ODI, Oxygen Desaturation Index.

TST (min)365.28 ± 84.02370.35 ± 62.680.77
Sleep latency (min)22.19 ± 24.1022.76 ± 22.200.90
Sleep efficiency (%)77.55 ± 15.7477.13 ± 11.690.90
S1%4.37 ± 3.284.76 ± 4.550.54
S2%70.38 ± 7.7266.21 ± 7.690.06
S3%5.80 ± 3.847.56 ± 3.520.06
S4%9.31 ± 9.238.73 ± 7.210.67
REM%8.41 ± 7.1112.06 ± 6.880.047
AHI18.02 ± 14.2716.84 ± 14.060.44
Apnoea Index5.55 ± 14.475.56 ± 11.780.99
Hypopnoea Index12.41 ± 7.6811.17 ± 4.830.39
Hypopnoea ratio (%)0.83 ± 0.230.84 ± 0.230.71
Arousal Index18.44 ± 9.8417.10 ± 6.470.29
SpO2 nadir (%)87.40 ± 8.7485.70 ± 8.820.25
ODI6.53 ± 9.406.56 ± 8.300.98
Figure 2.

 The apparent ‘random’ effects of temazepam on AHI and SpO2 nadir (= not significant). TAHI/TSpO2/TODI = AHI/SpO2/ODI on active temazepam night.

Primary analyses

Baseline awake central chemosensitivity correlated positively with both the placebo-adjusted change of SpO2 nadir (r = −0.468, = 0.038; Fig. 3) and oxygen desaturation index (ODI; = 0.485, = 0.03). It did not correlate with the change of AHI (= 0.18, = 0.44). However, the change of hypopnoea ratio [calculated as the percentage of hypopnoea/(hypopnoea + apnoea)] correlated significantly with both the change of SpO2 nadir (= 0.649, = 0.002; Fig. 4) and ODI (r = −0.686, = 0.001).

Figure 3.

 Association between central chemosensitivity and the changes of SpO2 nadir (r = −0.468, P = 0.038) after using temazepam (Active night – placebo night).

Figure 4.

 Association between the changes of SpO2 nadir and the change in hypopnoea ratio (r = 0.65, P = 0.002) after using temazepam (Active night – placebo night).

We used two linear regression models to examine whether the change in REM sleep time mediated the relationship between central chemosensitivity and the changes in SpO2 nadir and ODI. The REM change was not associated with SpO2 nadir (t = −0.038, = 0.97) or ODI (t = −1.01, = 0.92). The strength of the association between chemosensitivity and SpO2 nadir and ODI was unaffected by the presence of REM change in the model. In addition, no significant simple bivariate correlation was found between the change of REM sleep time and the change of SpO2 nadir (= 0.08, = 0.75), ODI (r = −0.1, = 0.67), AHI (r = −0.02, = 0.95) and baseline central chemosensitivity (r = −0.19, = 0.42).

Peripheral chemosensitivity and ventilatory recruitment threshold were not correlated with the change of SpO2 nadir, ODI or AHI (all > 0.05). Baseline central chemosensitivity was not correlated significantly with either placebo night AHI or SpO2 nadir (> 0.05).

Discussion

We observed a correlation between baseline awake central chemosensitivity and worse oxygenation during sleep with use of the hypnosedative agent temazepam. In accord with a previous study of another benzodiazepine, nitrazepam (Hoijer et al., 1994), temazepam did not worsen sleep-disordered breathing, but we observed the expected wide interindividual variability in response to the drug. These data suggest that waking ventilatory chemoreflexes may potentially be useful in phenotyping those individuals with OSA who are prone to develop respiratory depression following CNS depressant use. While chemoreflex testing during sleep might provide a more direct assessment of processes underlying OSA, we evaluated daytime testing as a potential simple assessment that could be useful in clinical practice.

Genotyping has been considered as an obvious avenue for subtyping OSA, and it has been suggested that hereditary factors explain 40% of the variance of OSA in the population, with the rest attributable to environmental factors (Casale et al., 2009). Studies are yet to identify strong specific genes associated with OSA (Casale et al., 2009; Redline and Tishler, 2000; Riha et al., 2009). This is not unexpected, given that there are so many interrelated pathologies that converge to cause OSA. As genotype and environment together create phenotype, we believe that a direct clinical phenotyping parameter may have a better association with individual response to various CNS active agents.

One way to phenotype OSA would be to measure the compensatory mechanisms for upper airway obstruction. Without compensatory mechanisms, even the mildest hypopnoea would result in near-doubling of arterial PCO2 and severe desaturations (Younes, 2008). There are three major compensatory mechanisms identified for the obstructed airway: (i) arousal threshold/chemical drive recruitment threshold: (ii) compensation via an increase in inspiratory duty cycle; and (iii) compensation via increased chemical drive (Loewen et al., 2009; Younes, 2008). Our study demonstrated that if a high compensatory chemical drive is suppressed by a CNS depressant, OSA patients will tend to have decreased oxygenation and/or a change of respiratory pattern from hypopnoea to apnoea. The relationship has been demonstrated by both the change of SpO2 nadir and ODI. In this instance, the overall AHI may not increase with greater respiratory depression, as AHI is only a measurement of the respiratory event cycling frequency and not the severity of each event measured. We quantified the change in the types of events making up the AHI using a ‘hypopnoea ratio’. We employed this instead of ‘hypopnoea duration’ measurement mainly because there are no standard criteria for the measurement of hypopnoea duration; it is hard to define precisely the beginning and the end of a long hypopnoea event. It also fails to account for any reduction in the number or severity of the classic apnoeas.

Our data are consistent with a previous study investigating respiratory phenotypes that predict the treatment effect of oxygen in OSA using a method to measure loop gain (controller gain/ventilatory chemosensitivity) during sleep (Wellman et al., 2008). This study compared the effect of supplemental oxygen on AHI in six OSA patients with high loop gain and six with a low loop gain, and found that supplemental oxygen reduced the AHI significantly in those OSA patients with high loop gain, but not in those with low loop gain (Wellman et al., 2008). This important study highlighted the potential of respiratory phenotyping as a method to ‘personalize’ treatment alternatives in individual patients. One limitation of this work is that the loop gain measurement required an overnight sleep study with a CPAP mask in position, and thus is not a simple phenotyping tool. The loop gain measurement required also did not measure ventilatory recruitment threshold, which may be important in understanding respiratory response heterogeneity in drug effects on OSA. In contrast, our study showed there is a relationship between an awake respiratory measurement and sedative effects on OSA with a 10-min central chemosensitivity test.

It is important to note that not all the OSA patients had a higher chemical drive, as we did not find a positive correlation between baseline central chemosensitivity and placebo–night severity of OSA. We believe that for OSA patients with a lower chemical drive there must be other functional compensatory mechanisms to keep their blood gas normal. In those situations, a mild CNS depressant may not cause notable respiratory depression through depressing chemical drives. A recent study stratified 25 OSA patients into three groups based on the level of pharyngeal closing pressure (Pcrit). They found that the significant correlation between loop gain and AHI existed mainly in the group with pharyngeal closing pressure near atmospheric levels (Pcrit between −1 and +1 cm H2O; Wellman et al., 2004). Similarly, a theoretical modelling study (Longobardo et al., 2008) found that an increased loop gain caused OSA when the upper airway was moderately collapsible, and that lowering the loop gain with supplemental oxygen eliminated the obstruction. This effect was less apparent when the airway was made more collapsible. In our study, we did not measure Pcrit because the aim of our study was to examine a practical clinical phenotyping method. Importantly, our data suggest that although not all the OSA patients have increased chemical drive, patients that do have increased central chemosensitivity are vulnerable to CNS depressant-related respiratory depression.

Another compensatory mechanism for OSA is related to TER and arousal threshold. TER measured during sleep has been considered as ‘the single most important variable that impacts the tendency to oscillate in the presence of abnormal pharyngeal mechanics’ (Younes, 2008). However, our measurement of awake ventilatory recruitment threshold did not correlate with the change in oxygenation. As our sample size is relatively small and has a restricted range of OSA severity, we cannot exclude the possibility that a correlation exists with an increased sample size or wider severity of OSA. We designed the study to test effects in a predominantly mild-to-moderate OSA sample. In this group there is unlikely to be major pathological variability in ventilatory recruitment thresholds. In more severe hypercapnic conditions, such as obesity hyperventilation syndrome or overlap syndrome [OSA plus chronic obstructive pulmonary disease (COPD)], patients often have significantly elevated waking ventilatory recruitment thresholds and blunted central chemosensitivity (Verbraecken et al., 1995; Younes, 2008). A measurement of ventilatory recruitment threshold would be more important in those syndromes.

In summary, OSA patients with higher awake central chemosensitivity had worse respiratory depression during sleep with the use of a mild CNS depressant (temazepam). As measuring central chemosensitivity requires only a 10-min daytime test, it has the potential to become a time-efficient tool for phenotyping OSA patients.

Declarations of Interest

None.

Acknowledgements

Dr David Wang is supported by NHMRC Health Professional Research Fellowship and Sydney Medical School Early Career Researcher/New Staff Award. Professor Ronald Grunstein is supported by NHMRC Practitioner Fellowship. Dr Nathaniel Marshall is supported by NHMRC funded CIRUS fellowship.

Ancillary