Jurrien M. ten Berg, Department of Cardiology, St Antonius Hospital, P.O. Box 2500, 3435 CM Nieuwegein, the Netherlands. Tel.: +31 306099111; fax: +31 306034420. E-mail: email@example.com
Summary. Background: High on-aspirin platelet reactivity (HAPR) is associated with atherothrombotic events following percutaneous coronary intervention (PCI). The aim of the present study was to identify the platelet function test sensitive for platelet cyclooxygenase-1 inhibition that best predicts atherothrombotic events. Methods and results: Nine hundred and fifty-one consecutive patients on dual antiplatelet therapy undergoing elective PCI were enrolled. On-aspirin platelet reactivity was measured in parallel by arachidonic acid (AA)-induced light transmittance aggregometry (AA-induced LTA), the VerifyNow® Aspirin Assay (VerifyNow® Aspirin Assay), the arachidonic acid prestimulated IMPACT-R (IMPACT-R AA) and the PFA-100 collagen/epinephrine cartridge (PFA COL/EPI). Cut-offs for HAPR were established by receiver-operator characteristic curve analysis. At 1-year follow-up, the composite of all-cause death, non-fatal acute myocardial infarction, stent thrombosis and ischemic stroke occurred more frequently in patients with HAPR when assessed by LTA [10.1% vs. 6.0%, P = 0.020 (n = 925)] and VerifyNow® [13.3% vs. 5.9%, P = 0.015 (n = 422)]. The VerifyNow® ASA assay (AUC = 0.78) and, to a lesser extent, AA-induced LTA (AUC = 0.73) added significantly to a model consisting of clinical and procedural risk factors in predicting atherothrombotic events. In contrast, the IMPACT-R (n = 791) and the PFA Collagen/Epinephrine (n = 719) were unable to discriminate between patients with and without primary endpoint at 1-year follow-up. None of the platelet function tests was able to identify patients at risk for bleeding. Conclusions: AA-induced LTA and the VerifyNow® ASA test were able to identify aspirin-treated patients undergoing PCI with stenting at risk for atherothrombotic events. The VerifyNow® Aspirin Assay had the highest predictive accuracy. None of the tests was able to identify patients at higher risk of bleeding.
Aspirin is the most widely used drug and exerts its effects by the irreversible inhibition of platelet cyclooxygenase (COX)-1, a key enzyme in the conversion of arachidonic acid (AA) to the potent platelet agonist thromboxane (TX)A2 . However, throughout the last two decades, it has become apparent that the individual biological response to low-dose aspirin is heterogeneous [2,3]. Patients with a low response to aspirin as measured with an aspirin-sensitive laboratory test have been termed ‘aspirin-resistant’, but this somewhat misleading term should be strictly reserved for exceptional situations in which the drug is unable to hit its pharmacologic target . High on-aspirin platelet reactivity (HAPR) is a much more appropriate term for high platelet reactivity status despite aspirin therapy in an individual patient. Moreover, it should be emphasized that a growing body of evidence shows a clear relationship between HAPR as measured with laboratory tests sensitive for platelet COX-1 inhibition and the occurrence of atherothrombotic events [5–15].
In the last couple of years, several platelet function tests have been developed, providing direct and indirect measurements of COX-1 inhibition . However, as yet, there is no consensus on the optimal method for assessing the magnitude of on-aspirin platelet reactivity . Therefore, the aim of the present study was to assess which platelet function test specific for platelet COX-1 inhibition is best at predicting atherothrombotic events, including stent thrombosis, in patients undergoing an elective percutaneous coronary intervention (PCI) with coronary stent implantation.
Materials and methods
The POPular study (The Do Platelet Function Assays Predict Clinical Outcomes in clopidogrel Pretreated patients undergoing elective PCI study) was a prospective, observational study that included consecutive patients with established coronary artery disease scheduled for elective PCI with stent implantation. The entry and exclusion criteria were given in the original publication, which described platelet function tests specific for clopidogrel . All patients were on dual antiplatelet therapy with clopidogrel and low-dose aspirin (80–100 mg daily) for at least 10 days. Compliance was verified by a detailed interview upon enrollment (self-reporting) as well as by pharmacy refill data.
Patients using concomitant medication known to affect platelet function other than clopidogrel and aspirin [i.e. non-steroidal anti-inflammatory drugs, dipyramidole, upstream glycoprotein (GP)IIb–IIIa inhibitors] and patients with a known platelet function disorder or a whole blood platelet count < 150 000 μL−1 were excluded.
All coronary interventions were performed according to current guidelines, and the choice of stent type and periprocedural use of GPIIb–IIIa inhibitors was left to the operator’s discretion, but the latter were always administered after blood collection. Written informed consent was obtained before PCI. All data were prospectively collected and entered into a central database. Clinical follow-up was obtained by contacting all patients at 12 months, and a double check was performed on the basis of source documents obtained from medical records from the referring hospitals.
The study was conducted according to the principles of the Declaration of Helsinki and the laws and regulations applicable in The Netherlands. The local institutional review board (Verenigde Commissies Mensgebonden Onderzoek) approved the study.
Follow-up and endpoints
The primary endpoint was defined as a composite of all-cause death, non-fatal myocardial infarction [defined as the occurrence of ischemic symptoms and a spontaneous [i.e. not periprocedural or postprocedural) troponin T value or creatine kinase MB greater than the upper limit of normal], stent thrombosis (definite stent thrombosis according to the Academic Research Consortium criteria) and ischemic stroke (focal loss of neurologic function caused by an ischemic event). The primary safety endpoint was defined as major or minor bleeding according to the modified Thrombolysis In Myocardial Infarction (TIMI) Study Group criteria . An independent committee, blinded for platelet function data, adjudicated all endpoints through review of source documents of medical records.
Before heparinization, whole blood was drawn from the femoral or radial artery sheath. Blood samples were collected into 3.2% citrate Sarstedt tubes for light transmittance aggregometry (LTA) and the IMPACT-R. The VerifyNow® (Accumetrics, San Diego, CA, USA) was performed using 3.2% citrate Greiner tubes, according to the manufacturer’s recommendation. For the PFA-100® System (Siemens Healthcare Diagnostics Products GmbH, Marburg, Germany) 3.8% buffered citrated blood was used, according to the manufacturer’s test protocol. Blood samples for whole blood count were drawn into tubes containing K3-EDTA.
Platelet function measurements
The magnitude of on-aspirin platelet reactivity was quantified using four commercially available platelet function tests that are claimed to be sensitive for platelet COX-1 inhibition: LTA using arachidonic acid (AA) as the agonist, the VerifyNow® Aspirin Assay, the IMPACT-R assay (with AA prestimulation) and the Dade® PFA Collagen/Epinephrine Test Cartridge (COL/EPI). All platelet function measurements were performed between 30 min and 2 h after blood collection.
LTA. LTA was performed in non-adjusted platelet-rich plasma on a four-channel APACT 4004 aggregometer (LABiTec, Arensburg, Germany). Platelet-poor plasma was set as 100% aggregation, and maximal platelet aggregation (%) was measured using AA at a final concentration of 0.5 mg mL−1. In the medical literature, the currently accepted cut-off value for AA-induced LTA to segregate patients with from those without HAPR is 20% aggregation .
VerifyNow® Aspirin Assay. The VerifyNow® system (Accumetrics, San Diego, CA, USA) is a whole blood cartridge-based method for determination of the magnitude of AA-induced platelet agglutination (in a final concentration of 1 mmol L−1) .
After a citrated tube of whole blood is inserted into the cartridge, the platelets become activated by conversion of AA to TXA2 by COX-1. As a result, the activated platelets bind via GPIIb–IIIa receptors to fibrinogen-coated beads and cause agglutination. Infrared light transmittance through the chamber increases as the platelet–bead complexes fall out of the solution. The results are reported in aspirin reaction units (ARU). In the medical literature, the currently accepted cut-off value for the VerifyNow® Aspirin Assay to segregate patients with from those without HAPR is 550 ARU .
IMPACT-R AA. The IMPACT-R device (DiaMed, Cresier, Switzerland) is based on cone and plate(let) analyzer technology . Citrated whole blood samples (130 μL) are, after preincubation with AA (0.32 μm) and gentle mixing (10 r.p.m.) for 1 min, placed in a polystyrene well and subjected to a shear rate of 1800 s−1 for 2 min, using a Teflon cone. When shear stress is applied, von Willebrand factor and fibrinogen are instantly immobilized on the polystyrene surface, serving as a substrate for platelet adhesion and subsequent aggregation. The wells are washed and stained with May–Grünwald stain, and analyzed with an inverted light microscope connected to an image analysis system. Platelet adhesion and aggregation on the surface are evaluated by examining the percentage of total area covered with platelets, designated as surface coverage (SC).
Exposure to AA leads to the formation of microaggregates in patients in whom aspirin does not effectively inhibit platelet function. These microaggregated platelets temporarily lose their adhesive properties. The percentage SC in the AA-prestimulated aliquots is therefore inversely correlated with the magnitude of AA-induced platelet activation. In the medical literature, the currently accepted cut-off value for the IMPACT-R AA to segregate patients with from those without HAPR is 2.5% SC .
PFA-100® System. The PFA-100® System measures platelet function, in particular adhesion and aggregation, in whole blood under high-shear conditions (5000 s−1). The time needed to form a platelet plug occluding the aperture cut into a collagen/epinephrine-coated membrane is determined and reported as closure time (CT) in seconds. The CT inversely reflects the magnitude of platelet reactivity. In the medical literature, the current accepted cut-off value for the PFA COL/EPI to segregate patients with from those without HAPR is 193 s .
Continuous variables are presented as mean (standard deviations). Categorical data are reported as frequencies (%). Categorical variables were compared with the chi-square test. Normally distributed continuous variables were compared with a two-sided unpaired t-test.
To evaluate a platelet function assay’s ability to discriminate between patients with and without the primary endpoint at 1-year follow-up, a receiver operator characteristic (ROC) curve analysis was calculated for each test. The optimal cut-off level was calculated by determining the smallest distance between the ROC curve and the upper left corner of the graph. Patients above this optimal cut-off level were considered to exhibit HAPR. Survival analysis for patients with and without HAPR according to the ROC curve of the specific test was performed with the Kaplan–Meier method, and the differences between groups were assessed by the log-rank test. The measure of effect was the odds ratio (OR), and was estimated from a logistic regression analysis. To correct for overfit of the ROC curve-derived cut-off value, a statistical resampling methodology (bootstrapping) was performed to assess the mean and distribution of the cut-off values derived from the area under the corresponding ROC curve [area under the curve (AUC)]. A total of 1000 replicates of each dataset were created by resampling with replacement; each resampled dataset was the same size as the original. For each platelet function test, the ROC curve was generated and the AUC was computed, and subsequently the cut-off value, along with the 95% confidence interval (CI). Furthermore, the cut-off levels for the four tests were derived from the medical literature and were applied to our clinical outcome data.
Logistic regression modeling was used to identify independent correlates of the primary endpoint. The model included on-treatment platelet reactivity according to the various tests as a categorical variable (patients with HAPR vs. patients without HAPR, using the cut-off defined with the ROC curve analysis) and multiple potential confounders [classic cardiovascular risk factors, renal failure, left ventricular ejection fraction < 45%, total stent length, number of lesions treated, amount of stents implanted, bifurcation lesions, comedication (including use of clopidogrel loading dose, coumadins, proton pump inhibitors, calcium channel blockers, statins or GPIIb–IIIa inhibitors), laboratory parameters (hemoglobin, platelet count and mean platelet volume), and left anterior descending coronary artery or graft stenting]. All univariate variables with a P-value < 0.10 were included in multivariable analysis. Whether a variable made an additional contribution to a logistic regression model without that variable was tested with the likelihood ratio test. The Hosmer–Le Cessie goodness-of-fit test was performed to assess the adequacy of the model. All statistical analyses were performed with r (version 2.9; http://www.r-project.org), and a two-tailed P-value of < 0.05 was considered to be significant.
A total of 1069 consecutive patients were enrolled, of whom 951 were on aspirin for > 10 days. Owing to irregularities in platelet assay supply, as well as technical failure in a minority of platelet function tests, not all platelet function assays were performed in every patient. Furthermore, halfway through the POPular study the VerifyNow® Aspirin cartridge was included. As a consequence, AA-induced LTA was performed in a total of 925 patients; the VerifyNow® Aspirin cartridge in 422 patients; the IMPACT-R in 791 patients; and the PFA COL/EPI in 719 patients.
Baseline characteristics of the cohort are depicted in Table 1. Baseline characteristics of the subpopulations according to the tests performed are summarized in Table S1, demonstrating that the subpopulations tested were well balanced (all P-values > 0.55).
Table 1. Baseline characteristics of the total population
LVEF, left ventricular ejection fraction; PPI, proton pump inhibitors; WBC, white blood cell; LAD, left anterior descending artery. Hypertension: systolic blood pressure > 140 mmHg or diastolic blood pressure > 90 mmHg. Hypercholesterolemia: a fasting low-density lipoprotein cholesterol level ≥ 3.4 mmol L−1 or being on statin therapy at the time of inclusion. Diabetes mellitus: according to the World Health Organization criteria . Family history: one or more first-degree relatives have developed coronary artery disease before the age of 55 years (men) or 65 years (women). Renal insufficiency: serum creatinine > 120 μmol L−1.
64 ± 10.6
LVEF < 45%
Prior myocardial infarction
Platelet count (× 109)
273.4 ± 78.8
WBC count (× 109)
7.7 ± 2.3
Hemoglobin (g dL−1)
13.8 ± 3.5
Mean no. of stents implanted
Minimal stent diameter (mm)
3.1 ± 0.8
Total stent length (mm)
28.3 ± 17.1
Clinical outcome at 12 months was available for 949 (99.9%) of the patients. During 1-year follow-up, a total of 16 (1.7%) patients died, 54 (5.7%) patients had non-fatal acute myocardial infarction, nine (0.9%) patients presented with definite stent thrombosis and 11 (1.2%) patients suffered from non-fatal ischemic stroke. A total of 39 (4.1%) patients presented with bleeding; 23 (2.4%) TIMI major and 19 (1.9%) TIMI minor bleeds.
ROC curve analysis
ROC curve analysis demonstrated that only aggregation-based tests such as AA-induced LTA and the VerifyNow® Aspirin cartridge assay were able to discriminate between patients with and without the primary endpoint at 1-year follow-up (Fig. S1). The estimate of AUC obtained by resampling was almost identical to the parametric estimate, and the bias of the analysis varied between 0.6% and 3.0% (Table 2). In contrast, neither the shear stress-based test IMPACT-R with AA prestimulation nor the PFA COL/EPI Test was able to distinguish between patients with and without postprocedural events. Table 2 shows the AUC and optimal cut-off value for every test, including the estimated cut-off value after bootstrap analysis.
Table 2. Area under the curve (AUC)
Cut-off after bootstrapping
AUC and optimal cut-off values for each test. CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value; LTA, light transmittance aggregometry; AA, arachidonic acid; ARU, Aspirin reaction units; SC, surface coverage.
Baseline characteristics for every test, for patients with and without HAPR, are shown in Table S2. Patients exhibiting HAPR according to AA-induced LTA were significantly older, included a higher proportion of females, had a higher frequency of diabetes mellitus, had a lower platelet count and were less often treated with statins. In patients with HAPR as measured by the VerifyNow® Aspirin cartridge, the proportions of diabetic patients and patients with hypertension were higher. Patients with HAPR according to the IMPACT-R AA were more likely to be female or be non-smokers, had a higher body mass index, suffered less often from hypertension and were treated more often with angiotensin-converting enzyme inhibitors. Patients who exhibited HAPR as defined by the PFA-100 COL/EPI were older, were more often female, had lower hemoglobin levels, more often suffered from diabetes or renal failure and more often had a history of prior myocardial infarction and an impaired ejection fraction.
Logistic regression modeling was used to determine independent predictors for the primary endpoint. The model included on-treatment platelet reactivity according to the various tests as a categorical variable (patients with HAPR vs. patients without HAPR, using the cut-off defined with the ROC curve analysis) and multiple potential confounders. Clinical factors independently predicting the 1-year primary endpoint were age [calculated for an increase of 10 years (OR 1.30; 95% CI 0.99–1.06)], hypertension (OR 2.33; 95% CI 1.04–5.19) and a left ventricular ejection fraction < 45% (OR 1.82; 95% CI 0.91–3.63). Procedural factors independently predicting the primary endpoint were graft stenting (OR 3.31; 95% CI 0.98–11.21) and stenting of a bifurcation lesion (OR 4.77; 95% CI 1.28–17.85). The remaining variables included for multivariate analysis were not found to be independent correlates of the primary endpoint (P > 0.10) and were not included in the model.
The addition of HAPR to this statistical model revealed that HAPR as measured with AA-induced LTA and the VerifyNow® Aspirin cartridge significantly improved the AUC (AUC = 0.73, P = 0.03 and AUC = 0.78, P = 0.02, respectively). Likewise, the likelihood ratio test demonstrated that HAPR according to these tests made an additional contribution to the model (Table 3). The goodness-of-fit test demonstrated that the predicting model was adequate (except for the PFA COL/EPI; P = 0.05, all P-values > 0.50). In contrast, the AUC did not improve when HAPR as measured with the IMPACT-R or the PFA COL/EPI was added to the model.
Table 3. Area under the curve (AUC) of different backward regression models for the prediction of the primary endpoint at 1-year follow-up
*Likelihood tatio test for additional value of high on-aspirin platelet reactivity as measured with different platelet function tests.
†Age, hypertension, left ventricular ejection fraction < 45%.
‡Graft stenting, stenting of a bifurcation lesion.
Model 1: Classical cardiovascular risk factors†
Model 2: Model 1 + procedural risk factors‡
Model 3: Model 2 + residual platelet reactivity
Relationship between HAPR and clinical outcome
The primary endpoint occurred more frequently in patients with HAPR than in patients without HAPR when platelet function was assessed by LTA (10.1% vs. 6.0%, P = 0.02) and the VerifyNow® Aspirin Assay (13.3% vs. 5.9%, P = 0.015), using the cut-off levels of the ROC curve analysis. One-year follow-up for patients with and without HAPR according to each platelet function test is shown in Table 4.
The combined endpoint occurred significantly more often in patients with HAPR when measured with AA-induced LTA and the VerifyNow® Aspirin cartridge than in patients without HAPR, whereas no significant association was detected when platelet function was assessed with the IMPACT-R or with the PFA COL/EPI (Fig. 1).
The occurrence of the primary endpoint was depicted in quintiles according to on-treatment platelet reactivity (Fig. 2). Patients in the higher quintiles according to the AA-induced LTA (P = 0.004) and the VerifyNow® Aspirin Assay (P = 0.008) were at significantly higher risk for the primary endpoint. In contrast, no significant difference in the occurrence of the primary endpoint was observed between quintiles as measured with the IMPACT-R. As the PFA-100® System confines detection of a CT to a 300-s window, the results of the PFA cartridge are depicted as time to aperture closure Kaplan–Meier curves. CTs as measured by the PFA COL/EPI were not significantly different between patients with and without a primary endpoint.
When literature-derived cut-offs were applied to the dataset, none of the tests was able to identify patients at higher risk for an atherothrombotic event, except for AA-induced LTA (the ROC curve analysis-derived cut-off is perfectly consistent with the one provided by the medical literature (Table S3).
Relationship between platelet reactivity and bleeding
A second ROC curve analysis demonstrated that the platelet function tests performed were unable to discriminate between patients with and without bleeding (all AUCs included 0.50 in the CI).
The principal finding of the present study is that HAPR as measured with aggregation-based tests specific for COX-1 inhibition with AA as the agonist is significantly associated with the occurrence of atherothrombotic events. In contrast, the shear stress-based tests IMPACT-R and PFA-100 COL/EPI were not able to identify patients at risk for adverse clinical outcome.
Although previous observational studies have already demonstrated the relationship between HAPR and adverse clinical outcome, these studies were limited by small sample size, the use of clinically non-validated cut-off levels and the availability of only one test per study [5–15].
The present study, which is substantially larger than previous studies, was designed to assess the predictive accuracy of four different platelet function tests in identifying patients at higher risk for atherothrombotic events after PCI.
Among the four tests, the VerifyNow® Aspirin Assay clearly showed the best predictive value for the occurrence of adverse events. Using ROC curve analysis on clinical outcome data, we were able to identify an optimal cut-off of 454 ARU to segregate patients with and without HAPR. This cut-off level is substantially lower than the currently accepted cut-off value of 550 ARU [3,13,22]. However, the commonly used cut-off of 550 ARU is questioned [23,24], as it is not identified by ROC curve analysis and, even more important, it has been determined with a previous design of the VerifyNow® Aspirin Assay that used cationic propyl gallate instead of AA as agonist. The available evidence linking HAPR as measured with the VerifyNow® Aspirin Assay and clinical outcome was obtained with the use of propyl gallate , and our study is the first to show the relationship with AA as the agonist. Remarkably, the VerifyNow® Aspirin Assay loses its predictive ability with the cut-off value of 550 ARU. These findings suggest that a cut-off of 454 ARU might be a more appropriate one for the prediction of clinical outcome with the current design of this cartridge.
In line with previous studies [5,8,25], LTA is also able to discriminate between patients with and without atherothrombotic events. Our ROC curve-derived cut-off of 20% AA-induced aggregation is perfectly consistent with the one derived from the literature . However, the predictive accuracy of LTA is only moderate, and the test might not be suitable for routine use in clinical practice, because of some major limitations, including poor reproducibility, long sample processing time, labor-intensiveness and the need for specialized technicians.
The present study is the first to investigate the association between HAPR according to the IMPACT-R AA and adverse clinical outcome. The cut-off value determined by ROC curve analysis (7.2%) differs from the cut-off value recommended in the literature (2.5%) . The IMPACT-R was not able to segregate patients with and without HAPR or to identify patients at higher risk of atherothrombotic events, using either our cut-off or the one derived from the literature.
The PFA COL/EPI was also unable to discriminate between patients with and without the primary endpoint. Contrasting results have been reported concerning the predictive accuracy of the PFA COL/EPI cartridge. Some studies demonstrated a two-fold to five-fold [9,26,27] higher risk in aspirin-treated patients with a shorter CT, using a cut-off value of either 193 s or 300 s, whereas the largest study thus far demonstrated no association at all between HAPR according to the PFA COL/EPI and adverse clinical outcome .
In accordance with earlier investigations, the incidence of HAPR is highly dependent on the type of platelet function assay used.  In addition, the platelet function tests used in the POPular study are not equally predictive in identifying patients at higher risk of atherothrombotic events. The addition of HAPR according to the aggregation-based COX-1 inhibition-sensitive tests to a model that includes both classic and procedural risk factors moderately, but significantly, enhanced the predictive ability of this model. The VerifyNow® showed the largest increase in predictive value of all tests performed in this cohort (AUC = 0.78), and therefore should be considered as the best platelet function test for identifying high-risk patients. Currently, no data are available concerning the clinical effectiveness of tailoring aspirin therapy on the basis of the results of platelet function testing. Therefore, the correct treatment – if any – of HAPR remains unknown. 
Some limitations merit careful consideration. First, there is a lack of data on serum TXB2, which is, being the stable metabolite of TXA2, considered to be the most specific measurement of platelet COX-1 activity; the availability of this might have given more clarity in the determination of HAPR. Second, not all platelet function tests were performed in every patient. Third, the absence of an association between the magnitude of platelet reactivity and bleeding complications should be interpreted with care, as the present analysis was not powered to identify patients at higher risk of bleeding.
In conclusion, this parallel evaluation of platelet function tests for their ability to predict clinical outcome demonstrates that the VerifyNow® Aspirin Assay is best at identifying patients undergoing elective PCI who are at higher risk for atherothrombotic events. In contrast, none of the platelet function tests is able to identify patients at higher risk of bleeding. As the adequate treatment of HAPR is unknown, the routine use of platelet function testing in clinical practice is not recommended.
We thank the independent committee that adjudicated all events: B. M. Swinkels, W. Dewilde and F. W. A. Verheugt.
Funding/sponsor original study
Siemens Healthcare Diagnostics provided the Dade® PFA Collagen/ADP Test Cartridge and INNOVANCE® PFA P2Y* free-of-charge.
Role of the sponsor
Siemens Healthcare Diagnostics was not involved in the design and conduct of the study, the collection, management, analysis and interpretation of the data, or in the preparation, review or approval of the manuscript.
Disclosure of Conflict of Interests
N. J. Breet has received speaker’s fees from Siemens. J. W. van Werkum has received speaker’s fees from Accumetrics and Siemens, and has acted as a consultant for The Medicines Company. J. M. ten Berg has received a speaker’s fees from Sanofi-Aventis, Eli Lilly, BMS and MSD, and has acted as a consultant for Sanofi-Aventis, Eli Lilly, Schering-plough and Glaxo Smith Kline. The other authors state that they have no conflict of interest.