Clinical predictors of dual aspirin and clopidogrel poor responsiveness in stable cardiovascular patients from the ADRIE study


  • Members of the ADRIE study group are listed in the Appendix.

Pierre Fontana, Division of Angiology and Hemostasis, University Hospitals of Geneva, 4, Rue Gabrielle-Perret-Gentil, CH-1211 Geneva 14, Switzerland.
Tel.: +41 22 379 59 38; fax: +41 22 372 92 99.

Jean-Luc Reny, Division of Internal Medicine, Béziers Hospital, 2, Rue Valentin Hau, BP 740, 34525 Béziers, France.
Tel.: +33 4 67 35 77 26; fax: +33 4 67 35 77 54.


Summary. Background: Poor response to both aspirin and clopidogrel (dual poor responsiveness [DPR]) is a major risk factor for recurrent ischemic events. Objectives: The aim of this study was to identify factors associated with DPR, defined with specific tests, and derive a predictive clinical score. Methods: We studied 771 consecutive stable cardiovascular patients treated with aspirin (n = 223), clopidogrel (n = 111), or both drugs (n = 437). Aspirin responsiveness was evaluated by serum thromboxane (Tx)B2 assay, and clopidogrel responsiveness by calculating the platelet reactivity index (PRI) on the basis of the phosphorylation status of the vasodilator phosphoprotein. The analysis was focused on patients treated with both drugs, and on independent predictors of DPR. Results: Among patients on dual therapy, there was no relevant correlation between TxB2 levels and PRI values (r = 0.11). Sixty-seven patients (15.4%) had DPR. Diabetes [odds ratio (OR) 1.89, 95% confidence interval (CI) 1.06–3.39], high body weight (> 86 kg vs. < 77 kg, OR 4.74, 95% CI 2.49–9.73), low aspirin dose (75–81 mg vs. ≥ 160 mg, OR 0.12, 95% CI 0.09–0.93) and high C-reactive protein (CRP) level (> 1.6 mg L−1 vs. < 0.6 mg L−1, OR 3.66, 95% CI 1.74–8.72) were independently associated with DPR, via increased TxB2 levels, increased PRI, or both. These associations with TxB2 and PRI were reproduced across the whole population. With use of a factor-weighed score (c-index = 0.74), the predicted prevalence of DPR was 57% in the highest strata of the score as compared with < 4% for the lowest strata. Conclusions: Diabetes, body weight, the aspirin dose and CRP levels are readily available independent predictors of DPR, and some are potential targets for reducing its prevalence.

Antiplatelet drugs form part of the first-line treatment for atherothrombosis [1]. The benefit of low-dose aspirin in cardiovascular patients is related to permanent inactivation of cyclooxygenase-1 and subsequent inhibition of thromboxane (Tx)A2 production, whereas clopidogrel targets the platelet ADP amplification pathway by inhibiting the P2Y12 receptor [2]. Biological antiplatelet drug responsiveness is highly variable, and is generally reported to be associated with cardiovascular outcomes [3]. Dual poor responsiveness (DPR), i.e. poor responsiveness to both aspirin and clopidogrel in the same patient, was recently identified as a major risk factor for recurrent ischemic events after coronary stent placement [4], but few data are available on this phenotype. The term ‘resistance’ is sometimes used to define those patients with inadequate biological responses to antiplatelet drugs. However, as there is no validated universal cut-off value for the definition of ‘resistance’, the term ‘poor responsiveness’, which is more in line with the continuous biological response, will be used hereafter.

Several assays can be used to evaluate antiplatelet drug responsiveness. They differ in their specificity for the targets of aspirin and clopidogrel. Quantification of serum TxB2, the stable breakdown product of TxA2, and quantification of the phosphorylation status of the vasodilator phosphoprotein (VASP), are highly specific for platelet-derived TxA2 inhibition by aspirin and P2Y12 receptor inhibition by clopidogrel, respectively [3,5]. The few existing studies on DPR have often involved less specific, aggregation-based assays, and have usually shown a substantial positive correlation between poor responsiveness to aspirin and poor responsiveness to clopidogrel [4,6–8]; moreover, antiplatelet drug responsiveness was evaluated in patients with recent ischemic events, which might affect platelet responsiveness [9].

Here, we evaluated aspirin and clopidogrel responsiveness by using specific assays in 771 stable cardiovascular patients treated with one or both antiplatelet drugs.



The Antiplatelet Drug Resistances and Ischemic Events (ADRIE) study ( identifier NCT00501423) is a prospective study focusing on the clinical relevance of the platelet response to aspirin and/or clopidogrel in stable cardiovascular patients. The primary objective is to determine whether platelet reactivity is predictive of clinical ischemic events over an ongoing 3-year follow-up period. Predefined secondary objectives include the identification of predictors of platelet reactivity. The present analysis, pertaining to this secondary objective, is aimed at identifying clinical predictors of DPR to aspirin and clopidogrel in patients taking both drugs, and to derive an easy-to-use predictive score.

Between June 2006 and December 2008, consecutive patients with symptomatic documented ischemic atherothrombotic disease (coronary artery disease [CAD], ischemic cerebrovascular disease and/or peripheral artery disease) treated with non-enteric-coated aspirin and/or clopidogrel for < 5 years were included in the university hospitals of Geneva, Switzerland (n = 522), Béziers, France (n = 374) and Montpellier, France (n = 44). The exclusion criteria were severe thrombocytopenia (< 100 × 109 L−1), thrombocytosis (> 400 G L−1), known thrombopathy, anticoagulant treatment, non-steroidal anti-inflammatory drug (NSAID) intake more than three times a month, ticlopidine or dipyridamole treatment, pregnancy, and active cancer. One hundred and sixty-nine patients did not attend the outpatient visit, mainly owing to prescription of anticoagulant treatment after hospital discharge or because they eventually declined to participate, leading to the effective inclusion of 771 patients (416 in Geneva, 327 in Béziers, and 28 in Montpellier). Of these 771 patients, 437 patients were on dual aspirin–clopidogrel therapy, 223 were treated with aspirin alone, and 111 were treated with clopidogrel alone.

The patients attended as outpatients at least 1 month after the last acute ischemic event and at least 10 days after their last NSAID intake, for a physical examination and blood sampling. Laboratory analyses included platelet function tests, fibrinogen assay, von Willebrand factor (VWF) ristocetin cofactor assay, both performed with an automated method (Siemens, Marburg, Germany), and high-sensitivity C-reactive protein (CRP) assay (Beckman Coulter, Paris, France). The patients were specifically questioned about adherence to treatment during a face-to-face interview. If patients omitted to take at least one dose of either antiplatelet drug at least once a week, they were categorized as non-adherent.

The patients gave their written informed consent, and the study protocol was approved by the Central Ethics Committee of the University Hospitals of Geneva (Geneva center) and the Ethics Committee of Montpellier Saint-Eloi (Béziers and Montpellier centers).

Data were recorded in the individual institutions and entered via password-protected secure Internet access (https) into a database, with a double independent entry procedure.

Blood collection

Venous blood was collected in resting (> 15 min) patients with a 21-gauge needle and no tourniquet, after an overnight fast, in tubes containing EDTA, lithium heparin, 0.105 m sodium citrate (1 : 9, v/v) or no anticoagulant (BD Vacutainer; Becton Dickinson, Meylan, France).

Platelet function evaluation

Evaluation of platelet-derived TxA2 production and definition of aspirin poor responsiveness  TxA2 production in response to endogenous thrombin was evaluated by allowing a 6-mL tube of whole blood to clot at 37 °C for 1 h, as previously described [10]. Serum was stored at − 80 °C, and TxB2, the stable breakdown product of TxA2, was assayed within 3 months after collection with an ELISA kit for TxB2 (GE Healthcare, Glattbrugg, Switzerland), blindly to all other test results. The TxB2 assay was performed with two dilutions, each in duplicate, and samples with a coefficient of variation of > 8% were retested until this parameter was ≤ 8%. The TxB2 assay was centralized and performed by the same technician in Geneva.

There is no validated cut-off for aspirin poor responsiveness based on TxB2 serum levels [11]. To define a cut-off value above which patients would be identified as aspirin poor responders, we used data and a receiver operating characteristic (ROC) analysis procedure in a population treated with low-dose aspirin for 2 weeks [12]. On the based of this ROC curve, a TXB2 cut-off of > 12 ng mL−1 was selected to define aspirin poor responsiveness: sensitivity 98% (95% confidence interval [CI] 88.9–99.9); and specificity 100% (95% CI 92.6–100). It is of note that this cut-off is similar to the one identified in an independent population [13].

Evaluation of P2Y12 receptor inhibition and definition of clopidogrel poor responsiveness  VASP phosphorylation status and the platelet reactivity index (PRI) were evaluated in citrated blood with a standardized assay (Platelet VASP/P2Y12; Biocytex, Marseille, France) [14] on a FACStrack flow cytometer (Becton Dickinson) within 24 h after blood collection. Flow cytometry was performed in two centers (Montpellier for the patients recruited in Béziers and Montpellier, and Geneva) with reagents from the same batch, following the manufacturer’s instructions. Clopidogrel poor responsiveness was defined by a PRI of >50% [15,16].

The biologists were blinded to the patients’ clinical status, and the clinicians were blinded to the platelet function results. Data on either TxB2 or PRI were missing for two patients treated with both aspirin and clopidogrel. Thus, only 435 patients were analyzed for DPR.

Statistical analysis

Data are reported as means ± standard deviations or as medians and interquartile range (IQR). Categorical variables are expressed as percentages. Kendall’s statistic was used to evaluate the correlation between TxB2 levels and PRI values, and also between these variables and other quantitative variables. Univariate analysis of the association between DPR and relevant variables that could confound the relationship of interest was based on the non-parametric Kruskal–Wallis test for quantitative variables and the chi-square test or Fisher’s exact test for qualitative variables. Trends across ordered categories of antiplatelet drug poor responsiveness status (poor responsiveness to neither drug, one drug or both drugs) were tested with non-parametric trend test by Cuzick [17] for quantitative variables, and with the Cochrane–Armitage test for qualitative variables. Variables with P-values ≤ 0.1 (patients with DPR vs. other patients) were preselected for multivariate analysis as described elsewhere [18]. When deciles of quantitative variables were not linearly correlated with DPR (weight, body mass index [BMI], age and CRP), the variable was entered in the multivariate model as an ordinal variable stratified in quartiles. When two contiguous quartiles yielded similar odds ratios (ORs), they were pooled in the multivariate analysis to improve the stability of the model. An unconditional logistic regression model with backward selection of variables was then used to identify independent predictors of DPR as compared with the rest of the population. Adjusted ORs and 95% CIs were derived from the estimated regression coefficients. The Hosmer–Lemeshow test was used to assess the goodness-of-fit of the multivariate model. The performance of the model was assessed by computing the area under the curve (AUC), usually referred to as the c-index. The AUC of the final model was validated by using booststrap analysis. Regression coefficients represent the respective weight of each independent predictor present in the final analysis for predicting DPR. These values were then rounded and used to construct a weighted score. The performance of the score was assessed by computing the AUC. The independent predictors of DPR were further tested against TXB2 levels in the whole population treated with aspirin, and against the PRI values in the whole population treated with clopidogrel. Data were analyzed with r software, version 2.10.1 (R Foundation for Statistical Computing, Vienna, Austria; and the sas package of programs, version 9.1 (SAS, Cary, NC, USA).


Figure 1 shows the flowchart of the study, and Table 1 summarizes the characteristics of the whole population according to antiplatelet drug therapy. As expected, most patients on dual treatment had CAD and a history of stenting. There was no other major difference between the antiplatelet treatment subgroups, except for the prescription of β-blockers, statins and angiotensin-converting enzyme inhibitors, reflecting an overall better implementation of secondary prevention medications in patients with recent stenting. Sixteen patients (2.2%) were identified as non-adherent among the whole cohort.

Figure 1.

 Flowchart of the study. TxB2, thromboxane B2; VASP, vasodilator phosphoprotein.

Table 1.   Characteristics of the study population according to antiplatelet therapy
 Total populationPatients with aspirin onlyPatients with clopidogrel onlyPatients with dual treatmentP
  1. ACEI indicates angiotensin-converting enzyme inhibitor; BMI, body mass index; CAD, coronary artery disease; CRP, C-reactive protein; ICD, ischemic cerebrovascular disease; IGR, interquartile range; PAD, peripheral artery disease; PPI, proton pump inhibitor; SD, standard deviation; SRI, serotonin reuptake inhibitor; VW:RCo, von Willebrand ristocetin cofactor.

Age (years), mean ± SD62.9 ± 12.165.3 ± 11.667.1 ± 11.260.7 ± 12.1< 0.001
Sex (% female)18.921.
Qualifying event (%)
 CAD66.348.928.884.6< 0.001
History of stent implantation (%)67.134.542.389.9< 0.001
Hypertension (%)56.757.864.054.20.166
Hypercholesterolemia (%)60.757.266.361.10.300
Current smokers (%)
Diabetes (%)21.722.921.621.10.866
Creatinine clearance (mL min−1), median (IQR)83.5 (67.5–98.5)83.9 (67.4–102.2)82.8 (66.8–96.5)83.2 (68.9–97.5)0.824
Weight (kg), mean ± SD77.0 ± 14.575.7 ± 16.275.5 ± 12.578.2 ± 13.90.022
BMI (kg m−2), mean ± SD26.4 ± 4.226.3 ± 4.726.0 ± 3.826.4 ± 3.90.240
ACEIs (%)48.429.144.159.3< 0.001
β-Blockers (%)63.946.634.280.3< 0.001
Statins (%)88.778.587.494.3< 0.001
Calcium channel blockers (%)16.718.821.614.40.117
Diuretics (%)23.127.423.420.80.169
Compliance (%)97.897.
Aspirin dose (%)
 < 100 mg d−
 100 mg d−162.465.560.9 
 > 100 mg d−16.510.34.6 
Clopidogrel dose 75 mg d−1 (%)96.798.296.30.549
Platelet count (g L−1), median (IQR)224 (190–269)231 (189–275)222 (185–274)221 (192–265)0.414
Mean platelet volume (fL), median (IQR)9 (8.3–9.7)9 (8.4–9.6)8.7 (8.1–9.4)9 (8.3–9.8)0.013
Fibrinogen (g L−1), median (IQR)3.5 (3.0–4.1)3.7 (3.1–4.5)3.5 (3.0–4.0)3.4 (3.0–4.0)0.003
VW:RCo (%), median (IQR)141 (104–177)142 (109–166)144 (104–177)138 (101–180)0.650
Creatinine (μm), median (IQR)82 (72–96)81 (69–95)81 (73–98)83 (72–96)0.291
CRP (mg L−1), median (IQR)1.9 (0.8–4.7)2.4 (1.2–6.1)2.2 (0.9–5.1)1.6 (0.6–4.2)<0.001

Table 2 summarizes the characteristics of the patients on dual therapy, according to antiplatelet drug responsiveness. Aspirin doses ranged from 75 to 160 mg d−1, whereas most patients were on the standard dose of clopidogrel (75 mg d−1).

Table 2.   Characteristics of the patients on dual aspirin-clopidogrel treatment according to drug response
 RespondersPR1DPRP (DPR vs. others)P for trend
  1. ACEI, angiotensin-converting enzyme inhibitor; BMI, body mass index; CAD, coronary artery disease; CRP, C-reactive protein; CYP3A4 statins are atorvastatin and simvastatin; CYP2C19 PPIs are omeprazole, lansoprazole and esomeprazole; DPR, dual poor responsiveness; ICD, ischemic cerebrovascular disease; IQR, interquartile range; PAD, peripheral artery disease; PPI, proton pump inhibitor; PR1, poor response to one drug; SD, standard deviation; SRI, serotonin reuptake inhibitor; VW:RCo, von Willebrand ristocetin cofactor.

Age (years), mean ± SD62.2 ± 12.360.6 ± 12.257.5 ± 11.10.0120.008
Sex (% female)21.818.210.40.0980.050
Qualifying event (%)
Hypertension (%)52.157.650.70.5930.847
Hypercholesterolemia (%)59.761.761.50.9720.743
Current smokers (%)22.418.719.40.9860.473
Diabetes (%)16.420.734.30.0070.005
Creatinine clearance (mL min−1), median (IQR)82.7 (70.5–97.4)83.8 (66.4–96.8)85.0 (74.1–106.9)0.1780.380
Weight (kg), mean ± SD74.0 ± 12.0179.1 ± 13.786.1 ± 15.3< 0.001< 0.001
Weight categories (%)
 ≤ 77 kg61.849.323.9< 0.001< 0.001
 77–86 kg24.224.628.40.571
 > 86 kg13.926.147.8< 0.001
BMI (kg m−2), mean ± SD25.5 ± 3.626.9 ± 3.728.8 ± 4.4< 0.001< 0.001
BMI ≥ 25 kg m−2 (%)49.765.573.10.033< 0.001
ACEIs (%)63.654.264.20.45520.586
Statins (%)93.995.192.50.5650.851
CYP3A4 statins (%)66.565.866.10.8880.936
Calcium channel blockers (%)13.914.814.90.9390.815
PPIs (%)27.931.534.30.56130.297
CYP2C19 PPIs (%)84.882.873.90.3690.310
SRIs (%)6.76.961.0000.897
Insulin (%)25.933.330.40.79360.708
Compliance (%)98.798.595.30.71850.146
Last antiplatelet drug intake
 On day of blood collection (%)61.760.543.50.0150.0373
Aspirin dose
 < 100 mg d−1 (%)32.130.550.70.0060.034
 100 mg d−1 (%)60.066.547.80.316
 > 100 mg d−1 (%)7.931.50.013
Clopidogrel dose 75 mg d−1 (%)96.496.1971.0000.885
Platelet count (g L−1), median (IQR)222 (192–252)223 (193–267)213 (192–284)0.9340.610
Mean platelet volume (fL), median (IQR)9 (8.4–9.7)9.1 (8.4–9.8)8.9 (8.2–9.77)0.4180.760
Fibrinogen (g L−1), median (IQR)3.2 (2.9–3.9)3.5 (3.0–4.0)3.6 (3.1–4.1)0.1000.01
VW:RCo (%), median (IQR)145 (105–180)131 (94–171)137 (107–199)0.3060.810
Creatinine (μm), median (IQR)84 (70–96)83 (74–98)83 (71–93)0.6230.970
CRP (mg L−1), median (IQR)1.2 (0.4–3.83)1.6 (0.69–4)2.40 (1.35–4.6)0.003< 0.001
CRP categories (%)
 ≤ 0.6 mg L−134.822.79.00.001< 0.001
 0.6–1.6 mg L−124.427.623.90.8705
 > 1.6 mg L−140.949.867.2< 0.001

Serum TxB2 ranged from 0 to 465 ng mL−1 (median 7 ng mL−1; IQR 3–12 ng mL−1). Patients with serum TxB2 > 12 ng mL−1 (n = 110) were considered to be poor responders to aspirin. The PRI ranged from 2% to 90% (median 52%; IQR 38–62%), and 227 patients (52%) had a PRI above 50% and were therefore considered to be poor responders to clopidogrel. TxB2 and PRI cut-offs were derived from previous independent clinical studies, as detailed in the methods. Overall, 67 patients (15.4%) were considered to have DPR.

We first examined the association between aspirin and clopidogrel responsiveness in patients treated with both drugs (n = 435). Figure 2 depicts the scatter plot of PRI values against serum TxB2 levels, and shows no relevant association (r = 0.11).

Figure 2.

 Scatter plot analysis of the platelet reactivity index (PRI) and thromboxane B2 (TxB2) levels. Kendall statistics yielded a correlation coefficient of 0.11 (P < 0.001). For readability, the figure focuses on TxB2 levels up to 100 ng mL−1 (> 95% of the population).

In Table 2, patients with DPR are compared with patients with single-agent poor responsiveness and with responders to either drug. DPR patients were younger, were more often diabetic and had higher weight values than patients with poor responsiveness to either or neither drug. They were more often treated with low-dose aspirin (< 100 mg d−1). The time of the last antiplatelet drug intake (the morning of blood collection or the day before) influenced the frequency of single poor responsiveness or DPR in univariate analysis. As all but four of the patients took aspirin and clopidogrel at the same time, we did not stratify this latter analysis for the nature of antiplatelet drug treatment. It is of note that the time of the last antiplatelet drug intake was not available for 42 patients. CRP values were higher in DPR patients than in the other subgroups. Further analysis, including stratifications of statins according to CYP3A4 specificity, proton pump inhibitors according to CYP2C19 specificity, or diabetes according to the type of treatment (insulin vs. no insulin), did not significantly affect the result of the univariate analysis (Table 2). Almost all of the significant variables rose or fell gradually according to drug poor responsiveness status, as in a ‘dose–response’ pattern (poor responsiveness to neither drug, one drug or both drugs). This pattern was similar when the results of the ‘poor responsiveness to a single antiplatelet drug’ group corresponded to either poor responsiveness to aspirin only or to poor responsiveness to clopidogrel only. For example, patients with poor responsiveness to aspirin only (n = 43) had a median CRP level of 2.0 mg L−1 (IQR 0.9–4.5), and those with poor responsiveness to clopidogrel only had a median CRP level of 1.6 mg L−1 (IQR 0.6–4.0, n = 160, P = 0.67)). Similarly, patients with poor responsiveness to aspirin only had a median body weight of 80 kg (IQR 73–90), and those with poor responsiveness to clopidogrel only had a median body weight of 77 kg (IQR 69–87, P = 0.14).

To identify independent predictors of DPR, we constructed a multivariate regression model with selected variables. The variables were age, sex, diabetes, body weight, BMI, time of the last antiplatelet drug intake, aspirin dose, and CRP. As BMI is derived from the body weight, two independent multivariate models were used, one that included body weight only and another that included BMI only, along with the other selected variables. Comparison of these models with the Akaike Information Criterion [19] led to the selection of body weight for inclusion in the score.

This multivariate model showed that diabetes, high body weight, low-dose aspirin and CRP level were independently and significantly associated with DPR (Table 3). The Hosmer–Lemeshow test indicated a good fit of the final multivariate model (= 0.554). The corresponding AUC was 0.77. Following bootstrapping, the corrected AUC was 0.73.

Table 3.   Predictors of dual poor responsiveness identified in multivariate analysis: reference categories were < 77 kg (median value) for body weight, < 100 mg d−1 for aspirin dose, and < 0.6 mg L−1 (first quartile) for C-reactive protein
VariablesEstimate (β)Adjusted OR95% CIP
  1. CI, confidence interval; CRP, C-reactive protein; OR, odds ratio.

Body weight (kg)
 < 77 1< 0.001
 > 861.554.742.49–9.03
Aspirin dose (mg d−1)
 < 10010.009
 100− 0.710.490.29–0.85
 > 100−–0.93
CRP (mg L−1)
 < 0.610.013
 > 1.61.303.661.54–8.72

These four variables were used to create a score (dual non-responsiveness to aspirin and clopidogrel [DRAC] score) in order to determine its predictive value for DPR. The DRAC score ranged from 0 to 55 (Table 4), and the estimated probability of DPR increased gradually with the score value (Fig. 3), reaching 57% for the highest score. The OR for DPR was 16.5 (95% CI 3.8–71.7) in patients with a score ≥ 45 as compared with patients with a score ≤ 20 (Fig. 3, insert). The AUC or c-index, indicative of the discriminative capacity of the DRAC score, was 0.74 (95% CI 0.67–0.81).

Table 4.   Calculation of a score for predicting dual poor responsiveness; the score for each variable was weighted by its β-coefficient in the logistic regression model (described in Table 3)
  1. CRP, C-reactive protein.

Body weight (kg)
 < 770
 > 8615
Aspirin dose (mg d−1)
 < 10020
 > 1000
CRP (mg L−1)
 < 0.60
 > 1.615
Figure 3.

 Prevalence (%) of DPR in the population according to the dual non-responsiveness to aspirin and clopidogrel (DRAC) score values. As several combinations of predictors are possible for a given DRAC value, the prevalence corresponding to the different combinations is displayed for each score value. The logistic multivariate regression model for four DRAC score categories are presented in the table (insert). CI, confidence interval; DPR, dual poor responsiveness; OR, odds ratio.

The specific relationships of these variables with TxB2 levels and PRI values are described in Table 5A. The association between diabetes and DPR was mainly related to the association with PRI, whereas the association between low-dose aspirin and DPR was related, as expected, only to the association with TxB2 level. High body weight and CRP levels were associated with both high TxB2 levels and PRI values. Analysis of the whole population according to antiplatelet drug treatment (Table 5B) reproduced the associations shown in Table 5A. It is noteworthy that, in this latter analysis, diabetes was significantly associated with both TxB2 and PRI values, in aspirin-treated (n = 660) and clopidogrel-treated (n = 548) patients, respectively.

Table 5.   Relationship between predictors of dual poor responsiveness and serum thromboxane B2 (TxB2) levels and platelet reactivity index (PRI) values in: (A) the population treated with both drugs; and (B) the population treated with aspirin ± clopidogrel (for the association with TxB2, n = 660) or clopidogrel ± aspirin (for the association with PRI, n = 548)
VariablesTxB2 (ng mL−1)PPRI (%)P
  1. CRP, C-reactive protein. Data are expressed as median and interquartile range.

  Yes8 (4–19)0.08556 (46–64)0.003
  No7 (3–12)50 (37–61)
 Body weight (kg)
  < 775 (2–9)< 0.00146 (36–59)< 0.001
  77–868 (4–14)51 (39–61)
  > 8610 (6–21)56 (47–67)
 Aspirin dose (mg d−1)
  < 10010 (6–20)< 0.00150 (37–64)0.680
  1005 (3–10)52 (39–61)
  > 1005 (2–7)44 (38–64)
 CRP (mg L−1)
  < 0.64 (2–8)< 0.00147 (36–56)0.003
  0.6–1.67 (4–12)52 (37–63)
  > 1.68 (4–16)54 (41–65)
  Yes8 (4–16)0.04254 (41–64)0.02
  No6 (3–12)49 (36–61)
 Body weight (kg)
  < 775 (3–9)< 0.00146 (34–59)< 0.001
  77–868 (4–14)50 (37–61)
  > 869 (5–21)56 (46–67)
 Aspirin dose (mg d−1)
  < 10010 (6–19)< 0.00150 (37–64)0.680
  1006 (3–10)52 (39–61)
  > 1004 (2–6)44 (38–63)
 CRP (mg L−1)
  < 0.65 (3–9)< 0.00147 (35–56)0.035
  0.6–1.67 (4–12)52 (37–64)
  > 1.67 (4–14)52 (39–64)


This is the first study to address the issue of dual antiplatelet drug poor responsiveness in an outpatient population with stable cardiovascular disease, based on specific assays of aspirin and clopidogrel responsiveness. No correlation was found between the response to aspirin and the response to clopidogrel. DPR, identified in 15% of the patients, was most frequent in patients with diabetes, high body weight, elevated CRP values, and lower aspirin doses. When these four independent predictors were present, the probability of DPR was 57%, as compared with < 4% in patients with none or only one of these characteristics. Some of these determinants, such as diabetes and high CRP levels, are known to be associated with poor cardiovascular outcomes.

High levels of residual TxA2 breakdown products and poor P2Y12 inhibition, respectively, are predictors of ischemic events in aspirin-treated and clopidogrel-treated patients [15,20]. When evaluated with platelet aggregation tests, DPR is a particularly strong risk factor for recurrent ischemic events [4]. However, few studies have addressed the predictors of this DPR phenotype. Assay of serum TxB2 (the stable metabolite of TxA2) and the VASP assay are highly specific tests for evaluating aspirin and clopidogrel responsiveness [5]. Overall, the use of these highly specific assays is the main original feature of the ADRIE study, allowing more precise targeting of the biological effect of aspirin and clopidogrel in stable cardiovascular patients.

Unlike previous studies using non-specific tests [4,6–8], we found no relevant correlation between aspirin and clopidogrel responsiveness (r = 0.11). Indeed, Gori et al. [4] acknowledged that the lack of test specificity contributed to the correlation that they observed between ADP-induced and arachidonic acid-induced platelet aggregation, which were used to evaluate clopidogrel and aspirin potency, respectively.

Determinants of DPR

Diabetes was an independent predictor for DPR in our study population, and may affect aspirin and clopidogrel potency in several ways [21]. We found that diabetes was associated with DPR through an increase in PRI values and a trend towards increased serum TxB2 values (Table 5A). When evaluated in the whole population treated with aspirin, the association between serum TxB2 values and the presence of diabetes became significant (Table 5B).

Heavier patients were also more likely to have DPR. This is consistent with the reported association between body weight and TxB2 in aspirin-treated patients [22] and healthy subjects [23], and between body weight and PRI in clopidogrel-treated patients [24]. Aspirin bioavailability may be lower in heavier patients [22,23], and impaired activation of cyclic nucleotide–specific kinase–VASP pathways has been reported in those patients [25]. Although BMI is often used to address this latter issue, few studies have compared the relative influence of weight and BMI on platelet responsiveness. In an independent study on stable cardiovascular patients tested with the VASP assay, weight had a more marked effect on clopidogrel responsiveness than BMI [24]. This is supported by pharmacologic data indicating that aspirin and the active metabolite of clopidogrel are hydrophilic (logP of the active metabolite of clopidogrel = 1.96, ChemOffice Pro; CambridgeSoft, Cambridge, MA, USA). Thus, as compared with body weight, a parameter that is dependent on body fat such as BMI [26] may not be the most appropriate with respect to pharmacologic data.

A high CRP level was independently associated with DPR. This is consistent with reports of an association between high CRP levels and platelet reactivity in aspirin-treated or clopidogrel-treated cardiovascular patients [27,28].

Finally, the aspirin dose was also associated with DPR. As expected, this association was driven by serum TxB2 levels only. An association between low-dose aspirin and high serum TxB2 levels (measured ex vivo) has previously been reported in healthy subjects [29], as well as in cardiovascular patients [30], and is in line with the aspirin concentration-related reduction in TxB2 production observed in vitro [31]. This latter predictor of DPR could easily be overcome by increasing the aspirin dose in this putatively very high-risk population. Ongoing intervention studies such as ARCTIC (NCT00827411) may help to answer this question.

Rapid identification of patients with DPR remains a challenge. We therefore constructed a score combining the variables identified in our logistic regression model. The discriminative capacity of the DRAC score was good, with a c-index of 0.74, validated by bootstrapping. Indeed, a c-index for a prognostic model is typically between 0.6 and 0.85 [32]. As a comparison, widely used scores to predict thromboembolism in people with non-valvular atrial fibrillation, such as the CHADS2 score, had c-indexes ranging from 0.56 to 0.62 in a recent prospective study [33].

Diabetes was also included in the Residual Platelet Aggregation after Deployment of Intracoronary Stent (PREDICT) score, developed elsewhere to assess high residual platelet aggregation (ADP-induced) after a clopidogrel loading dose in patients undergoing coronary stenting for stable angina and acute coronary syndromes [34]. However, age, creatinine level and left ventricular dysfunction were included in this score, whereas they were not associated with DPR in our study. This may reflect differences in the methods used to assess platelet function, in the endpoints (high residual platelet aggregation vs. DPR), and in the study populations.


Our study also has some limitations. First, the lack of treatment randomization does not allow firm conclusions to be drawn, particularly regarding the association between DPR and the aspirin dose. However, this association remained highly significant after adjustment for several clinical and biological variables, and is consistent with previous reports [29–31]. Second, the frequency of treatment non-adherence was relatively low. Our patients were probably more aware of the importance of adherence to antiplatelet drug treatment, owing to the stated focus of the study and to a relatively recent stent implantation for an acute coronary syndrome in a majority of patients. This low rate of non-adherence to antiplatelet drug treatment is consistent with that described in the recently published POPULAR study [35], in which more than 95% of patients were adherent to antiplatelet drug treatment 6 months after enrollment, as assessed by pharmacy refill data. In any case, misclassification of non-compliant patients in the group of poor responder patients would have tended to bias the results towards the null hypothesis, e.g. a poorly predictive score. Third, although the performance of the multivariate model is confirmed by the bootstrap analysis, the DRAC score should be validated in an independent cohort. Fourth, DPR may be influenced by other factors that we did not take into account. Indeed, the cytochrome P450 2C19 (rs4244285) polymorphism is associated with increased PRI values in subjects treated with clopidogrel [36], and may thus be associated with DPR. Our aim, however, was to identify readily accessible biological and clinical predictors; moreover, this polymorphism accounted for only 12% of the variation in clopidogrel responsiveness in a healthy population [37] and for 5.2% in a population of cardiovascular patients [38]. It is thus unlikely that this genotype would have a major additional impact on the present results on predictors of DPR. Finally, a major limitation is that our population was heterogeneous in terms of timing of last dose intake, and this may have interfered with antiplatelet drug response classification, owing to differences in platelet turnover. However, the timing of the last antiplatelet drug dosing was taken into account, and was no longer significant in the multivariate analysis, which may reflect a limited impact of this parameter in our population.


This study, based on highly specific assays in stable cardiovascular patients, shows that aspirin and clopidogrel responsiveness are independent phenotypes, and that diabetes, body weight, aspirin dose and CRP level are independently associated with DPR. The DRAC score based on these readily available factors identifies a subpopulation of patients with a high prevalence of DPR. Such patients may benefit from more aggressive antiplatelet therapy, but intervention studies are warranted to confirm this hypothesis in stable cardiovascular patients.


P. Fontana, P. Berdagué and J.-L. Reny: study design, data collection and data analysis; C. Castelli and P. Fabbro-Peray: database design and statistical analysis; F. Mach, H. Bounameaux, P. de Moerloose, J.-F. Schved and S. Nolli: study design and data collection. All authors contributed to and approved the final version of the article. P. Fontana is the guarantor of data from Switzerland, and J.-L. Reny is the guarantor of data from France.


The authors thank N. Beteta for excellent technical assistance and J. Mascarini for patient recruitment in Geneva. P. Fontana and H. Bounameaux have received Evolva grant research support. This work was supported by the Swiss National Science Foundation (grant no. 3200B0-116843, P. Fontana), the Swiss Heart Foundation (P. Fontana), the Swiss Society of Angiology (P. Fontana), the Internal Medicine Department of Geneva hospitals (grant no. PRD-06-II-10, P. Fontana), and the French ministry of health Programme Hospitalier de Recherche Clinique (grant no. PHRC APR 2005-14R-11, J.-L. Reny).

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.


The ADRIE group includes, in addition to the authors of the present article, C. Arquizan. F. Becker, G. Böge, E. Boissier, R. Bonvini, Y. Boukriche, S. Cazaban, C. Dumas, S. Gueddi, N. Joan, C. Marty-Ané, P. Nangou, E. Oziol, A. Perrier, I. Quéré, H. Rabesandratana, G. Reber, H. Robert-Ebadi, M. Righini, S. Sedighian, B. Simorre, and R. Sztajzel.