Table 2.

Effect of statin adherence on the hazard of an acute cardiovascular event

				Adherents vs non-adherents
Model	Type of model	Variables in adherence model	Variables in outcome model	HR	95% CI
1	Discrete-time hazards model	Not applicable	Time-dependent adherence	0.82	0.68 to 0.98
2	Discrete-time hazards model	Not applicable	Time-dependent adherence and baseline confounders*	0.76	0.63 to 0.91
3	MSM	Adherence history, baseline* and lagged time-dependent† confounders	Adherence during the previous assessment year	0.78	0.65 to 0.94
4	MSM, weights truncated at 1st and 99th centiles	Adherence history, baseline* and lagged time-dependent† confounders	Adherence during the previous assessment year	0.78	0.65 to 0.94

*Baseline confounders as presented in online supplementary table S5.

†Lagged time-dependent confounders: marital status, income, labour market status, increase in intensity of statin therapy, diabetes, use of insulin, hypertensive diseases, heart failure or chronic cardiac insufficiency, cardiac arrhythmia, dysfunctions of lipid metabolism, number of concurrent cardiovascular medications, Charlson Comorbidity Index, chronic CHD, chronic CHD hospitalisation, medical procedure related to CHD, chronic cerebrovascular diseases and transient ischaemic attack, atherosclerosis, use of nitrates, rheumatoid arthritis, cancer, mental disorders, depression, respiratory diseases, alcohol-related diseases, use of non-steroidal anti-inflammatory medications, anxiolytics, hypnotics and sedatives, corticosteroids for systemic use, and hormone therapy, total number of concurrent medications, number of in-hospital days.

CHD, coronary heart disease; MSM, marginal structural model.