Sleep-Disordered Breathing Is Associated With Recurrent Ischemic Stroke

Brown, Devin L. MD; Shafie-Khorassani, Fatema MPH; Kim, Sehee PhD; Chervin, Ronald D. MD, MS; Case, Erin BA; Morgenstern, Lewis B. MD; Yadollahi, Azadeh PhD; Tower, Susan MD; Lisabeth, Lynda D. PhD

Disclosures

Stroke. 2019;50(3):571-576. 

In This Article

Results

Of the 947 ApneaLink Plus studies performed on ischemic stroke subjects, 878 were successful and yielded a result. Of the successful sleep apnea studies, 842 were associated with an index ischemic stroke. Baseline characteristics of the 842 unique subjects are found in Table 1. The median age was 65 (interquartile range [IQR], 57–76), 47% were female, 58% were MA, 34% were NHW, and the remainder were other race/ethnicities. Of the entire group, 526 (63%) had SDB defined by an REI score ≥10; the median REI score was 14 (IQR, 6–26). SDB was associated with male sex, MA ethnicity, being insured, nonsmoking status, diabetes mellitus, hypertension, lower educational attainment, and higher body mass index (Table 1). The median time from stroke symptom onset to SDB assessment was 13 days (IQR, 6–22). Of the 842 unique individuals, 778 were MA or NHW and constitute the analytic sample for the primary analysis. Among MAs, the median REI score was 16 (IQR, 7–27); among NHWs, it was 11 (IQR, 6–24), P=0.001.

Events

Over the study time period (August 2010 to December 2015), 85 (11%) recurrent ischemic strokes and 104 (13%) deaths occurred among 775 MA or NHW subjects with an index ischemic stroke and no missing data. The median follow-up time was 591 (IQR, 245–1175) days.

Associations Between REI and Recurrent Ischemic Stroke

In unadjusted analysis, REI was associated with recurrent ischemic stroke (hazard ratio [HR], 1.02 per one-unit increase in REI score [95% CI, 1.01–1.03]). The relationship was similar in the fully adjusted model (HR, 1.02 [95% CI, 1.01–1.03]; Table 2). For a 20-unit increase in REI scores, the association between REI and ischemic stroke recurrence in the fully adjusted model was HR, 1.51 (1.16–1.97). In a separate fully adjusted model, an added interaction term between REI and ethnicity was not significant (P=0.11). Among NHWs, a 20-unit increase in REI score was associated with a 2-fold higher rate of recurrent ischemic stroke (HR, 2.04 [95% CI, 1.31–3.17]). Among MAs, the association was not as strong and did not reach significance (HR, 1.33 [95% CI, 0.98–1.83]).

In the fully adjusted model excluding REI, MA ethnicity was associated with a nonsignificant increased rate of ischemic stroke recurrence, with an HR of 1.38 (95% CI, 0.84–2.25). When REI was included in the model, the ethnic association showed little change, with an HR of 1.31 (95% CI, 0.80–2.14). This represents a 5% change in the HR. Thus, REI does not appear to confound the ethnicity-recurrent stroke association.

Associations Between REI and Death

In unadjusted analysis, REI was not associated with death (HR, 1.01 per one-unit increase in the REI score [95% CI, 1.00–1.02]). The relationship was similar in the fully adjusted model (HR, 1.00 [95% CI, 0.99–1.02]; Table 3). In a separate fully adjusted model, an interaction term between REI and ethnicity was not significant (P=0.13). A 20-unit change in REI was not associated with mortality among NHWs (HR, 0.73 [95% CI, 0.42–1.25]) or MAs (HR, 1.16 [95% CI, 0.88–1.52]).

In the fully adjusted model that excluded REI, MA ethnicity was associated with an increased rate of mortality, with an HR of 1.61 (95% CI, 1.04–2.50). When REI was included in the model, the ethnic association was similar (HR, 1.61 [95% CI, 1.04–2.50]). Thus, REI does not appear to confound the ethnicity-mortality association.

Adequacy of the Models

No violation of the proportional hazards assumption occurred for the primary outcome of recurrence. However, the proportional hazards assumption was violated for REI (P=0.008) and National Institutes of Health Stroke Scale (P=0.029) in the cause-specific mortality model. Because the proportional hazards assumption, based on a graphical assessment by the standardized martingale residuals, was violated across a short duration and only early in the time period, at which point few events had occurred, time-stratified models were unstable and therefore not pursued. The linearity assumption was satisfied for all continuous variables in all models.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....