Retrospective Cohort Study Design for CPAP and Cardiovascular Event Reduction in OSA
Primary Recommendation
Design a retrospective cohort study comparing cardiovascular events in OSA patients with documented CPAP adherence (≥4 hours/night on ≥70% of nights) versus untreated or non-adherent patients, focusing on hard cardiovascular endpoints rather than surrogate markers, as this addresses the critical evidence gap between observational studies showing benefit and RCTs showing no effect. 1, 2
Study Design Framework
Population Selection
Include adults with moderate-to-severe OSA (AHI ≥15 events/hour) documented by polysomnography or home sleep testing, as most evidence for cardiovascular effects comes from this severity range 1
Stratify by baseline sleepiness status (Epworth Sleepiness Scale >10 vs ≤10), as sleepy patients demonstrate greater arterial stiffness improvement and may have different cardiovascular risk profiles than non-sleepy patients 3, 1
Exclude patients with established cardiovascular disease at baseline if studying primary prevention, or create a separate cohort for secondary prevention, since one well-designed RCT showed CPAP plus usual care did not prevent cardiovascular events in patients with moderate-severe OSA and established CVD 2
Consider age stratification, particularly including elderly patients (≥65 years), as this population shows 94.4% survival with CPAP versus 78.4% without treatment over 5 years of follow-up 4
Exposure Definition (CPAP Adherence)
Define adherence objectively using device download data showing ≥4 hours/night on ≥70% of nights, as patients typically overestimate their adherence and objective measurement is critical 5, 6
Track adherence monthly and sum over the study period (minimum 12 months, ideally 24-25 months), as protective effects become significant at 6-12 months but not at 3 months 6
Document residual AHI on CPAP therapy, though cardiovascular mortality appears similar between CPAP (residual AHI 4.5±2.3/h) and oral appliance therapy (residual AHI 16.3±5.1/h), suggesting adherence may matter more than complete AHI normalization 7
Primary Outcome Measures
Use hard cardiovascular endpoints as primary outcomes: major adverse cardiovascular events (MACE), cardiovascular mortality, myocardial infarction, stroke, and cardiac procedures 2
Avoid surrogate markers that have not translated to clinical benefit in RCTs, as the divergence between observational studies and RCTs suggests surrogate endpoints may be misleading 2
Track all-cause mortality as a co-primary endpoint, given that observational data shows cardiovascular death rates of 0.56 per 100 person-years with CPAP versus 2.1 per 100 person-years without treatment 7
Secondary Outcome Measures
Blood pressure changes (24-hour ambulatory monitoring preferred), as CPAP produces clinically significant reductions in nocturnal, daytime, and 24-hour systolic and diastolic BP, with largest effects on nocturnal measurements 1
Incident diabetes, as death in older OSA patients is associated with both cardiovascular disease and diabetes 4
Quality of life measures using validated sleep-related QOL instruments 2
Arterial stiffness (carotid-femoral pulse wave velocity), particularly in sleepy patients who show significant cf-PWV reduction after CPAP (p=0.012) versus no change in non-sleepy patients 3
Critical Covariates and Confounders
Baseline Cardiovascular Risk Factors
Hypertension status (normotensive, hypertensive, resistant hypertensive), as BP reduction is clinically significant in hypertensive groups but stratification is essential 1
Body mass index and weight changes, as weight loss improves AHI scores and OSA symptoms independently 5
Diabetes, dyslipidemia, and metabolic syndrome, which are strongly associated with arterial stiffness and cardiovascular outcomes 3, 4
Age, as it is strongly related to arterial stiffness and cardiovascular risk 3
OSA Severity Parameters
Baseline AHI, recognizing that sleepy patients tend to have more severe OSA (AHI 44.8±19.0 vs 29.7±15.7 events/h in non-sleepy patients) 3
Nocturnal oxygen saturation metrics, particularly minimum oxygen saturation and time below 90%, as these relate to cardiovascular risk 8
Arousal index, as this may independently affect cardiovascular outcomes 9
Sociodemographic Factors
- Race, sex, and socioeconomic status, though the protective effect of CPAP appears consistent across these subgroups in Medicare populations 6
Analytical Approach
Statistical Methods
Use propensity score matching to balance covariates between CPAP-adherent and non-adherent/untreated groups, as demonstrated effective in showing mean MBP difference of -1.97 (95% CI: -14.03, -9.92; p<0.0001) 9
Employ Kaplan-Meier survival analysis for time-to-event outcomes, as this clearly demonstrates survival differences (94.4% vs 78.4%) 4
Apply generalized estimating equations for repeated measures and to account for within-subject correlation over time 6
Conduct stratified analyses by sleepiness status, OSA severity, and cardiovascular risk profile, as effects differ significantly between these subgroups 3, 1
Follow-up Duration
Minimum follow-up of 12 months, as protective effects are not significant at 3 months but become evident at 6-12 months 6
Optimal follow-up of 5-8 years to capture long-term cardiovascular mortality differences, as demonstrated in cohorts showing 5.6% mortality with CPAP versus 21.6% without treatment over mean 5-year follow-up 4
Common Pitfalls to Avoid
Adherence Misclassification
Never rely on patient self-report of CPAP use, as patients consistently overestimate adherence and this will bias results toward the null 5
Account for time-varying adherence, as patients may start adherent but become non-adherent over time, requiring monthly tracking rather than single time-point assessment 6
Population Heterogeneity
Do not combine sleepy and non-sleepy patients without stratification, as there is significant interaction between time (CPAP) and sleepiness (p=0.033) for arterial stiffness outcomes 3
Separate primary and secondary prevention populations, as patients with established CVD show different treatment responses than those without baseline cardiovascular disease 2
Outcome Selection Bias
Avoid using only blood pressure as the primary outcome, as the 2-3 mm Hg reductions seen with CPAP may not translate to hard cardiovascular event reduction in RCTs despite being statistically significant 2
Do not exclude patients who develop cardiovascular events early in follow-up, as this introduces immortal time bias 6
Evidence Quality Considerations
The critical divergence between observational studies (showing benefit) and RCTs (showing no benefit) for cardiovascular outcomes creates a unique methodological challenge 1, 2. Your retrospective cohort must address why observational data conflicts with RCT data, likely through:
Better adherence measurement than achieved in RCTs, where adherence is typically suboptimal 5
Longer follow-up periods than most RCTs (5-8 years vs 2-3 years) 4
Careful stratification by sleepiness and baseline cardiovascular risk, as RCTs may have diluted effects by including heterogeneous populations 3, 1
Focus on populations most likely to benefit (sleepy patients, those without established CVD, adequate CPAP adherence) rather than intention-to-treat analyses that include non-adherent patients 2, 6
The American Academy of Sleep Medicine explicitly states there is insufficient evidence to recommend CPAP for cardiovascular event reduction in non-sleepy adults, making this the highest-priority population to study with rigorous observational methods 1, 2.