Reply: “Obstructive Sleep Apnea, Sleep Duration, and Fasting Glucose” and “The Impact of Obesity on the Association between Obstructive Sleep Apnea and Glucose Metabolism”

From the Authors:

Dr. Kawada expresses concern about our use of actigraphy to quantify habitual sleep duration (1). We did use the medium threshold (40 counts per epoch) to differentiate wake from sleep. Prior studies have not found clear benefits in terms of agreement with polysomnography by using either a higher or lower threshold in adults (2, 3). Although actigraphy is known to underestimate wake after sleep onset as a result of scoring of quiet wakefulness as sleep, there is little evidence to suggest this offset leads to systematic bias. In particular, a prior study found similar performance of actigraphy in quantifying sleep in those with and without sleep apnea (4).

Dr. Kawada is also concerned by our categorization of sleep duration. The decision to define short sleep duration as ≤5 hours/night was made to parallel the categorizations used in prior cohort studies; in particular, the Sleep Heart Health Study (5) and the Nurses’ Health Study (6). In addition to these categorical analyses, we assessed the effect of sleep duration as a continuous measure modeled both in a linear fashion and with a quadratic term to allow for a U-shaped relationship. The results from these models were consistent with those from the categorical model presented.

Dr. Kawada contrasts our findings with the recent report by Twedt and colleagues of an association between short sleep duration and hyperglycemia in a cohort of women with gestational diabetes (7). We believe their findings are in line with our work, which also found an association between short sleep and abnormal fasting glucose. This association weakened after accounting for sleep apnea severity. As we note, it is unclear whether this reflects confounding or a loss of statistical power. The effect of sleep apnea on the short sleep/glycemia relationship was not addressed by Twedt and colleagues, as it was not a focus of their study. Further research on this topic is clearly needed.

We agree with Dr. Kawada that the effect of mental health in moderating the relationship between sleep and metabolic disease is an important topic for future research; however, this topic was beyond the scope of our recent work.

Finally, Dr. Song draws attention to the potential for obesity to modify the association between sleep apnea and glycemia. To investigate this, we first constructed a model including an obesity × sleep apnea interaction. Although both sleep apnea and obesity were predictors of glycemic status, we found no evidence of an obesity × sleep apnea interaction (P = 0.68). Furthermore, the ethnicity × sleep apnea interaction persisted (P = 0.03) after adding the obesity × sleep apnea interaction, suggesting any obesity × sleep apnea effect does not explain the effect modification by ethnicity that we found. Finally, we found no evidence that the ethnicity × sleep apnea interaction varies by obesity, although we were almost certainly underpowered to detect such a three-way interaction.