Study population and design

All patients admitted to the Department of Cardiology at La Timone Hospital in Marseilles between September 2008 and June 2009 were considered eligible to enter the study if they had: chronic therapy with aspirin 75 mg for at least 1 week; undergone coronary stenting for non-ST-segment elevation acute coronary syndrome; a good in-hospital aspirin response defined by AA-Ag less than 30%.

The exclusion criteria were as follows: history of bleeding diathesis; acute coronary syndrome less than 4 days; glycoprotein IIb/IIIa antagonist less than 48 h; New York Heart Association class IV; contraindications to antiplatelet therapy; platelet count less than 100 G/L; creatinine clearance less than 30 mL/min; and low response to aspirin during the hospital phase (AA-Ag > 30%).

Patients received non-enteric coated aspirin 75 mg daily as a directly-observed therapy administered by a nurse during hospitalization, to minimize the risk of non-adherence associated with daily clopidogrel 150 mg. After 3 days, the ‘in-hospital aspirin response’ was measured within 12 hours after each aspirin intake using AA-Ag. Patients were discharged with a prescription of aspirin 75 mg and clopidogrel 150 mg daily and were provided with educational sessions highlighting the importance of patient adherence to physicians’ recommendations. One month after hospital discharge, patients were admitted to our Antiplatelet Monitoring Unit and were asked if they were actually taking their medication. Assessment of ‘outpatient response to aspirin’ with AA-Ag was then performed between 1 and 12 h after each aspirin intake. Patients identified as non-responders received directly observed aspirin therapy 75 mg before reassessment on the same day, 1 to 12 h after administration, in order to exclude bioavailability problems and to properly identify non-adherent patients ( Fig. 1 ). Patients gave written informed consent for participation.

Study design.

Platelet variables

Blood samples were drawn from a peripheral venous catheter. The platelet count was determined in the platelet-rich plasma sample and adjusted to 2.5 × 10 ⁸ mL ⁻¹ with homologous platelet-poor plasma. Platelets were stimulated with AA (0.5 mg/mL) and aggregation was assessed with a PAP4 aggregometer (Biodata Corporation, Wellcome, Paris, France). Aggregation was expressed as the maximal percentage change in light transmittance from baseline with the platelet-poor plasma as reference. Here we report data on maximal intensity of platelet aggregation with AA concentration. The coefficient of variation of maximal intensity of platelet aggregation with AA was 6%. Non-response to aspirin was defined by AA-Ag greater than 30%, as described previously .

Statistical analysis

Statistical analyses were performed using the SPSS software program, version 15.0 (SPSS Inc., Chicago, IL, USA). Potential risk factors and patients’ social and clinical characteristics were screened for inclusion in the model by testing each independently for any significant association with non-adherence, using univariate logistic regression. Variables that achieved a liberal significance level of p ≤ 0.25 in the univariate analysis were included in the multivariate model. For the multivariate analysis, a logistic regression based on a backward elimination approach was used and variables were considered to be significantly associated with the outcome if the p value was less or equal to 0.05. A good way of assessing a binary logistic regression model’s ability to accurately classify observations is to use a receiver operating characteristic (ROC) curve. A ROC curve is constructed by generating several classification tables for cut-off values ranging from 0 to 1, and calculating the sensitivity (proportion of truly non-adherent patients who were correctly identified as such) and specificity (proportion of truly adherent patients who were correctly identified as such) for each threshold value. Sensitivity is plotted against 1 − specificity (i.e. one minus specificity) to create a ROC curve.

The area under the ROC curve (AUC) is commonly used as a summary measure of the receiver operating characteristic curve and provides a measure of discrimination: a model with a large area under the ROC curve suggests that the model is able to accurately predict the value of an observation’s response. Hosmer and Lemeshow have provided general rules for interpreting AUC values : AUC = 0.5, no discrimination; 0.7 ≤ AUC < 0.8, acceptable discrimination; 0.8 ≤ AUC < 0.9, excellent discrimination; AUC ≥ 0.9, outstanding discrimination (this is extremely rare).