Sympathetic activation by obstructive sleep apnea: a challenging ‘off-label’ meta-analysis

A single occurrence of obstructive apnea during sleep will trigger sequential hemodynamic, chemical and cortical perturbations that collectively elicit surges in central sympathetic outflow, neural norepinephrine release and blood pressure; these ebb once breathing resumes [1–3]. Left untreated over months or years, the cumulative burden of scores or hundreds of such events nightly elicits sustained autonomic after-effects, apparent during wakefulness [3]. Relative to healthy individuals, vagal heart rate modulation of patients with obstructive sleep apnea (OSA) is reduced [4], their blood pressure variability is increased [4] and their muscle sympathetic nerve activity (MSNA) is substantially greater than predicted by their age, sex, or body mass index (BMI) [3–6], with sympathetic firing incidence during quiet rest directly proportional to the frequency of apneas and hypopneas per hour of sleep (Apnea-Hypopnea Index, AHI) [7], a conventional measure of OSA severity [8]. In a randomized trial in which continuous positive airway pressure abolished OSA, 1 month of treatment eliminated this daytime sympathetic excess [9]. This upward setting, by chronic untreated OSA, of MSNA during wakefulness has been attributed to long-term structural or functional neural adaptations to hypoxia and oxygen free radical generation evoked by recurring apneas and hypopneas during sleep. Such autonomic neuroplasticity involves both oxygen sensing components of the peripheral chemoreceptor and specific cortical and brain-stem regions participating in the generation or modulation of efferent vagal and sympathetic nerve discharge [10–17]. Peripheral chemoreceptor sensitivity to hypoxia is augmented [10]. The left dorsal posterior insula thins, proportionate to hypoxia severity, whereas the left mid-cingulate cortex and both thalami thicken; the magnitude of such thickening correlates positively with MSNA burst incidence during wakefulness [17]. Current published literature concerning the MSNA of patients with OSA comprises case–control, mechanistic or interventional studies of relatively small cohort size. Many originate from the same laboratory, with each addressing a singular hypothesis. In a recent issue of the journal, Quarti-Trevano and colleagues [18] offered a systematic review and meta-analysis of such data, with three specific aims: to determine whether such daytime sympathetic activation coheres to OSA rather than concomitant co-morbidities also characterized by autonomic perturbations; if the magnitude of sympathetic activation is directly proportional to the AHI; and whether heart rate adds clinical value, as a surrogate to microneurography (essentially a research probe) as a signal of both sympathetic activation and OSA severity. Although the frequency and severity of apnea-induced hypoxia [19] and the frequency of arousal from sleep [20] are arguably more causal of chronic sympathetic excitation during wakefulness, the majority of microneurographic studies report only the AHI. Quarti-Trevano et al.[18] accomplished admirably all three aims, demonstrating: an incremental increase in daytime MSNA burst incidence with each grade of OSA severity; a significant positive correlation between MSNA and the AHI, independent of body weight or BMI; and direct significant correlations between heart rate and both MSNA and the AHI. They concluded that OSA is itself a state of chronic sympathetic activation, regardless of co-morbidities and therefore an appropriate target of interventions designed to restore autonomic equilibrium. These authors merit congratulation not only for establishing these concepts convincingly but in particular for their thoughtful subsequent management of the challenges encountered in the course of this endeavor [21]. Meta-analyses were initially designed, introduced to the clinical literature and adopted enthusiastically by practitioners because of their capacity to estimate effect sizes and confidence intervals of interventions evaluated in randomized controlled trials of cohort sizes that individually might have been insufficient to reject equipoise. Fundamental principles included the uniqueness of study participants and randomization, to minimize the influence of unknown but potentially important confounding variables. Later, non-randomized studies were introduced to meta-analyses, with the caveat that ensuing risk of bias was amplified and often inestimable. Quarti-Trevano et al.'s systematic review of the English language literature identified 16 pertinent cross-sectional studies, 7 prospective studies and 3 clinical trials, comprising 528 patients with sleep apnea and 268 healthy controls [18]. However, in the analysis and synthesis of such data, it transpired that these numbers were not singular. More often than not, the same patient and control data were presented in two or more studies (in once instance, the same 25 OSA patient data appeared in 5 publications concerning different mechanistic hypotheses; Vaughan Macefield, personal communication) – but the authors had no means of establishing such precisely and, accordingly, eliminating each redundant entry. To lessen the likelihood of duplication, they then excluded 12 of the 26 first-pass articles [21]. According to their Table S4 [21], if one sums the number of participants reported for each, the revisited meta-analysis comprises 365 OSA patients and 165 controls. But are all data in this smaller pool derived from unique individuals? Since several laboratories were represented by more than one publication (often reporting rather similar cohort mean data), likely not. The magnitude of redundancy within this residue cannot be established without access to primary source participant-level data. According to Table S4 in their updated work [21], the mean MSNA burst incidence in those with OSA (mean AHI 36.2 events/h) was 63.6 bursts/100 heart beats whereas in the control group (age difference) (mean AHI stated to be 1 event/h) this was 45.8 bursts/100 heart beats. Assuming, as a thought experiment, that these group means are not weighted by double or triple counting of specific individuals’ data points, let us then ask: is this differential excess of ∼40% on average an overestimate, an underestimate, or a fair approximation of the magnitude of central upregulation of sympathetic outflow induced by months or years of quotidian cycles of apnea, hypoxia, hypercapnia, and arousal during sleep? Because the control group was on average 5.4 years younger, normative data concerning the influence of age on MSNA [6] would posit an overestimate. Also, at their average age, any overestimate would be augmented if there were disproportionately more women in the control population than in the OSA cohort [6]. Biffi et al.[21] identified participant sex in their review of primary sources, but did not report separately the net number of men and women represented by these 14 publications. Conversely, considering the positive linear relationship illustrated in the authors’ revisited Figure 5 [21], had these controls more than 1 apneic event/h on average, the impact of OSA might have been underestimated. Of the 14 studies compiled, approximately 1/4 reported no controls; 1/4 selected controls based on good health and the absence of symptoms suggestive of OSA; 1/4 performed polysomnography on all participants but did not report the actual AHIs of those classified as controls, and 1/4 performed polysomnography on all participants and reported the AHIs of all participants. In the latter, a surprising number of community-based volunteers who had no symptoms or foreknowledge of sleep apnea had mild, moderate, or even severe OSA. Thus, 2 studies from 1 laboratory reported, for controls, mean AHIs of 7 [22] and 8 [23] events/h and mean arousal indices of 21 [22] and 21 [23] events/h; 2 publications from our program (with considerable participant redundancy) reported, for controls, mean AHIs of 6 [7] and 8 events/h [20] and a mean arousal index of 20 events/h [20]. On the basis of this finding of non-zero AHI from healthy asymptomatic individuals from two independent laboratories, it is extremely unlikely that the AHI of all of the control subjects, whether polysomnography was not performed, or performed, but the findings not disclosed, was zero. If the mean AHIs of these control cohorts were indeed >0, then both meta-analyses [18,21] may have underestimated the true impact of OSA on MSNA. Importantly, the present data are specific to the resting state. Combining blood oxygen level-dependent (BOLD) magnetic resonance imaging and microneurography, our group detected no difference between individuals with no or mild OSA and with moderate or severe OSA [7] in the magnitudes of increases in BOLD signaling in cortical autonomic regions or in sympathetic nerve firing when obstructive apnea was simulated by a brief series of Muller maneuvers [24]. These considerations aside, the four key insights from this arduous scholarly effort are: OSA is a condition of sympathetic excess during wakefulness, regardless of patient co-morbidities; the magnitude of daytime sympathetic excitation is directly proportional to the frequency of apneas plus hypopneas during sleep; it should not be presumed, without validation by polysomnography, that lean individuals, self-reported as healthy, are without OSA; and, when assembling the findings of multiple studies from the same laboratory, addressing, appropriately, distinctly different questions, the high risk of aggregating data that do not originate from unique individuals should serve as a caution for those embarking on similar meta-analyses. ACKNOWLEDGEMENTS Conflicts of interest There are no conflicts of interest.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.