Nonfatal and fatal cardiovascular disease events in CPAP compliant obstructive sleep apnea patients

We compared the occurrence of nonfatal and fatal CVD events in patients with long-term CPAP treatment (CPAP-treated patients) to patients who despite of the doctor’s advice discontinued the CPAP treatment (control patients). The groups were pairwise matched for gender, age, and AHI.

Using the discharge database of the Turku University Hospital (Turku, Finland), we identified 1116 patients who had commenced CPAP treatment due to OSA during the years 2002–2006 and continued the treatment at least for 5 years (Fig. 1). Patients without available cardiorespiratory polygraphy data were excluded (N = 52). From the eligible patients, 1030 had at least four follow-up visits during the treatment allowing the mean user hours to be computed for each patient with confidence. These 1030 patients formed the group of CPAP-treated patients without any further selection.

The control group was identified among those OSA patients who had commenced CPAP treatment during the years 2002–2009 (N = 2359), but despite of the doctor’s advice discontinued the treatment within 5 years. Patients without available cardiorespiratory polygraphy data were excluded (N = 83). The remaining patients (N = 2276) had used CPAP for a median of 8.9 (IQR 28.6) months. Among these patients, K-means clustering algorithm with Manhattan distance (absolute value distance) was used to find the best-matched control for each CPAP-treated patient in regard to gender, age and AHI. The average difference between a CPAP-treated patient and a matching control patient was 2.8 years in age and 3.5/h in AHI reported as absolute values. The pairwise matched control patients (N = 1030) had used CPAP for a median of 4.0 (IQR 16.0) months prior to the commencement of the study.

All included patients underwent cardiorespiratory polygraphy prior to initiation of CPAP treatment. OSA was diagnosed either by a cardiorespiratory polygraphy in the hospital (Embla, Somnologica Software, Flaga hf. Medical Devices, Reykjavik, Iceland) or by an ambulatory device at home (Embletta, Somnologica Software, Flaga hf. Medical Devices, Reykjavik, Iceland). A pulmonologist or a clinical neurophysiologist scored the sleep studies. Categories of OSA severity were defined according to the AHI: AHI < 5/h (normal), AHI ≥ 5, and < 15/h (mild OSA), AHI ≥ 15 and < 30/h (moderate OSA), and ≥ 30/h (severe OSA) [14]. Generally, CPAP treatment was recommended for patients with AHI ≥ 15/h. In patients with AHI < 15/h, assessment of the need for treatment was made individually according to patients’ symptoms and clinical findings strongly suggestive for OSA.

The scores of the self-administered Epworth Sleepiness Scale (ESS) [15] and General Health Questionnaire (GHQ-12) [16] were determined at the commencement of CPAP treatment and entered in the electronic medical records.

The following baseline variables were examined from the electronic medical records: age, gender, body mass index (BMI), serum cholesterol concentration, and smoking status (defined as current smoker, ex-smoker or never smoker). Hypertension was defined as blood pressure (BP) greater than 140/90 mmHg and/or use of antihypertensive medication. CVD included coronary, cerebral, or peripheral artery diseases. An internist diagnosed coronary artery disease (CAD) by clinical symptoms and findings and/or exercise testing or coronary angiography. A neurologist confirmed the diagnosis of stroke by computerized tomography (CT) or magnetic resonance imaging (MRI). Peripheral artery disease was assessed by ankle-brachial index. Diagnosis of type 2 diabetes (T2D; fasting glucose of at least 7.0 mmol/l, use of insulin and/or oral antidiabetics) or impaired fasting glucose (IFG; fasting glucose of 6.1–6.9 mmol/l) was determined. Information of chronic obstructive pulmonary disease (COPD) diagnosed by a pulmonologist was derived from the electronic hospital records.

The study end point was a composite of nonfatal (primary or secondary) and fatal CVD events. In the Cox regression model, the time to the first event was analyzed. Nonfatal CVD event was a composite of angina pectoris (stable or unstable), CAD diagnosed by angiography, and invasively treated, nonfatal myocardial infarction and nonfatal stroke. CVD events were diagnosed in special health care by cardiologists or internists according to the international guidelines. The Turku University Hospital is the only hospital in the region performing surgical and interventional cardiology procedures and having emergency department and intensive cardiac care unit. Basic cause of death included all ICD10 I-category diagnoses (I00–I99) confirmed by autopsy or documented disease history. Cause of death was derived from the national registry maintained by Statistics Finland.

Among CPAP-treated patients, the follow-up time started from the commencement of CPAP treatment and ended to the first nonfatal or fatal CVD event, all-cause death, CPAP withdrawal, or the last CPAP follow-up visit before the end of the year 2014. The median follow-up time was 8.7 (IQR 2.8) years. The control patients were followed from the last CPAP follow-up visit until the first nonfatal or fatal CVD event, all-cause death, or until the end of the year 2014. The median follow-up time was 6.2 (IQR 4.1) years.

CPAP follow-up visits occurred approximately three times during the first year and after habituation every second year. Mean number of visits during the follow-up period was 8.4 ± 2.2 per CPAP-treated patient. Treatment pressure (cmH₂O) and CPAP usage hours (h/day) were recorded by inbuilt counter clock of the CPAP device. The use of CPAP was determined from the device at every visit undertaken during the study and entered in the electronic medical records. For each patient, the mean usage hours across the CPAP treatment period were determined. CPAP-treated patients with a median CPAP usage of at least 4 h/day were considered compliant to the treatment. Patients engaging in bariatric surgery and controls commencing mandibular advancement device (MAD) treatment during the study follow-up were identified.

Data analyses were performed using the IBM SPSS Statistics 22.0 (Armonk, NY, USA: IBM Corp.) software package. Continuous variables with normal distribution were presented as mean values and standard deviations (SD), whereas variables not normally distributed were reported as median values and interquartile ranges (IQR). Means and medians were compared by using the independent-samples T test and the Mann–Whitney U test, respectively. Categorical variables were expressed as frequencies and percentages, and the comparisons were performed with the χ² test. P values of < 0.05 were considered significant.

Cox regression model was used to evaluate the association between CPAP treatment and the occurrence of nonfatal and fatal CVD events. The adjusted model was adjusted for the following confounding covariates: gender, age, AHI, BMI, IFG/T2D, COPD, hypertension, and CVD at baseline. In the unadjusted model, each covariate was computed separately in the model. Hazard ratios (HR) and 95% confidence intervals (CI) were used to determine the associations between the studied covariates and the time to the study end point. Nonadjusted Kaplan–Meier curves with data censored at the time of the end of follow-up were used to visually compare the risk of nonfatal or fatal CVD event across the groups.

Each CPAP-treated and corresponding control patient was pairwise matched for age, gender, and AHI at baseline (Table 1). Severe OSA was diagnosed in 498 (48.3%) of the CPAP-treated and in 454 (44.1%) of the control patients (p = 0.052). Moderate–severe OSA was diagnosed in 76.2% of the CPAP-treated and 76.5% of the control patients, whereas 23.8% and 23.5% of the patients had an AHI less than 15/h, respectively (p = 0.876). Majority of the CPAP-treated and control patients were male and obese (BMI > 30 kg/m², 67.1% vs. 60.3%, p = 0.001, respectively). Excessive daytime sleepiness (ESS ≥ 10) was reported in 44.5% of the CPAP-treated and 37.2% of the control patients (p = 0.001).

Based on CVD risk factors the two groups differed from each other. The strongest risk factor, previous CVD, was diagnosed in 5.4% (N = 56) of the CPAP-treated and 12.6% (N = 130) of the control patients at baseline (p < 0.001). CAD was the most common CVD at baseline (3.3% in CPAP-treated and 6.7% in control patients, p < 0.001).

Among majority of the CPAP-treated patients, the compliance to CPAP was extremely good, median 6.4 (IQR 2.3) h/day across the follow-up period of 8.7 (IQR 2.8) years (Table 1). Only in 11.4% of the CPAP-treated patients, the median use was less than 4 h/day (median use 2.9 h/day, IQR 1.3). All patients in the control group were advised to commence CPAP treatment at the same clinic with same principles, but the majority of them stopped early on the treatment (median 4.0 months, IQR 16.0) and the compliance was poor (median use 0.7 h/day, IQR 2.6).

The study end point occurred in total of 14.4% (N = 148) of the CPAP-treated and in 18.8% (N = 194) of the control patients (p = 0.006) (Table 2). Similar trend was shown when fatal CVD events were analyzed separately (N = 22, 2.1% vs. N = 79, 7.7%, p < 0.001, respectively), but among nonfatal CVD events, no difference between the groups was observed (N = 126, 12.2% vs. N = 115, 11.2%, p = 0.451, respectively). Ischemic heart disease was the most common cause of CVD death (N = 52/101, 51.5%). More detailed data on CVD diagnostics and treatment are shown in online resource (Online Resource 1).

CPAP treatment was associated with a decreased risk of the study end point compared to controls (HR 0.56, CI 95% 0.5–0.7, p < 0.001) (Table 3). The association remained significant after adjusting the model with other known CVD risk factors (HR 0.64, CI 95% 0.5–0.8, p < 0.001). In the adjusted model, the strongest risk factor was previous CVD (HR 4.09, CI 95% 3.1–5.3, p < 0.001). Also, male gender, older age, hypertension, and IFG/T2D remained independent risk factors while AHI did not. The association between CPAP treatment and reduction in CVD events was also depicted in nonadjusted Kaplan–Meier curves (Fig. 2). The association became even more significant when all CVD deaths (N = 33, 3.2% in CPAP-treated and N = 106, 10.3% in control patients, p < 0.001) were analyzed separately (HR 0.23, CI 95% 0.2–0.3, p < 0.001) (Online Resource 2).

Of the CPAP-treated patients with median CPAP usage of more than 6 h/day (N = 598), 13.7% had a CVD event, whereas among patients with lower CPAP usage (N = 432), the corresponding incidence was 15.3% (p = 0.480). A significantly decreased risk of CVD event was not demonstrated among CPAP-treated patients with CPAP usage of more than 6 h/day compared to patients with lower CPAP usage (HR 0.82, CI 95% 0.6–1.1, p = 0.240).