Obstructive sleep apnea symptom phenotypes and physical health approximately 5 years later in the Sleep Heart Health Study.

Objective

To examine whether four established obstructive sleep apnea (OSA) symptom phenotypes were associated with physical health and functioning approximately 5 years later.

Methods

We analyzed 1207 participants with OSA from the Sleep Heart Health Study (SHHS). Four phenotypes were reconstructed by latent class analysis using 14 binary symptom indicators and a categorized Epworth Sleepiness Scale score. Physical Component Summary (PCS) and Physical Functioning (PF) scores at follow-up were evaluated using multivariable linear regression with adjustment for the corresponding baseline score and prespecified covariates. Longitudinal mixed-effects models tested phenotype-by-time interactions. Declines of at least 5 points were evaluated using modified Poisson regression.

Results

The cohort included 395 minimally symptomatic, 149 disturbed sleep, 463 moderately sleepy, and 200 excessively sleepy participants. Phenotype was associated with follow-up PCS (overall P = 0.005) and PF (overall P = 0.021). Compared with the minimally symptomatic phenotype, PCS was lower in the excessively sleepy phenotype (adjusted difference, -3.77; 95% confidence interval [CI], -6.09 to -1.44; Holm-adjusted P = 0.009) and the moderately sleepy phenotype (-2.31; 95% CI, -3.91 to -0.72; Holm-adjusted P = 0.023). PF was lower in the moderately sleepy phenotype (-5.36; 95% CI, -9.22 to -1.50; Holm-adjusted P = 0.039). The excessively sleepy phenotype had a higher risk of a PCS decline of at least 5 points (risk ratio, 1.73; 95% CI, 1.17 to 2.56; Holm-adjusted P = 0.019). Phenotype-by-time interactions were not significant for PCS (P = 0.646) or PF (P = 0.461).

Conclusions

Sleepy OSA symptom phenotypes were associated with lower baseline-adjusted physical health approximately 5 years later. The moderately sleepy phenotype also had lower PCS and PF, suggesting that a simple sleepy or nonsleepy classification may not capture clinically relevant symptom heterogeneity.