The long and short of respiratory events in OSA
Author: Omar Mesarwi, MD
Date: April 1, 2019
AHI. It’s often the only number that practitioners use to characterize sleep apnea severity, which is kind of amazing when you think about it, since polysomnograms and even home sleep apnea tests record huge amounts of data. At our institution, even the report summarizing a polysomnogram is eight pages long, and yet study results are so frequently distilled into a single value which is meant to summarize the entirety of the heterogeneity of OSA pathophysiology as it applies to a patient. Obviously, a lot is getting lost in this process. For years, researchers have looked beyond the AHI to determine what other measures which we obtain from routine testing may lead to adverse health outcomes, and today’s paper takes on that role too.
If you had to guess whether short or prolonged respiratory events in OSA leads to worsened outcomes, which would you pick? If you adjust for AHI, conventional wisdom suggests that longer events should do more damage, right? Patients are exposed to hypoxia and hypercapnia for more prolonged periods. Seems like that’s not a great thing. Respiratory event duration is a function of the arousal threshold in OSA, however, and one might argue that with respiratory events of shorter duration, patients are more prone to sleep fragmentation and chemosensitivity, leading to worse outcomes. In a recent study from Butler and colleagues in the Blue Journal, authors sought to examine the relationship between respiratory event duration and mortality using the Sleep Heart Health database, a massive cohort of 6400 individuals with well characterized sleep and cardiovascular parameters. Among this larger subset, the authors analyzed available mortality data on 5712 subjects. They categorized patients into quartiles of respiratory event duration, and looked for mortality associations, in three models. (Model 1 included covariates of age, gender, race, smoking status, and BMI; Model 2 added AHI; and Model 3 further added prevalent hypertension, diabetes, stroke, coronary heart disease, and heart failure.)
Around two thirds of the overall group had OSA of some severity, and about a third of the overall group were moderate to severe. (The irony, by the way, is not lost on me that I’m using just the AHI to characterize OSA severity here.) Interestingly, subjects with the longest respiratory events had the highest AHI. I’m not sure I would have predicted that. Mortality was significantly higher in the shortest respiratory event quartile than in the longest (27.2 vs 19.5 deaths per 1000 person-years, p<0.001 for the ANOVA), after multivariate adjustment. This difference in mortality held irrespective of model used, and also when respiratory event duration was made a continuous variable rather than categorical (i.e., when quartiles weren’t used).
In my practice, I look at markers of hypoxemia (T90, nadir saturation) in addition to AHI, and I look at the relative proportion of hypopneas versus apneas. I also look at positional variation in certain instances, and sometimes at REM-related OSA, but this will paper push me to look beyond even these variables. We can’t get by ignoring the majority of the data we obtain from a sleep study any longer.
