Oral anticoagulant safety in family practice: prognostic accuracy of Bleeding Risk Scores (from the CACAO study)

Abstract

Background

To assess bleeding risk of patients treated by oral anticoagulants, several scores have been constructed to assist physicians in the evaluation of the benefit risk. Most of these scores lack a strong enough level of evidence for use in family practice.

Objective

To assess the predictive prognostic accuracy of 13 scores designed to assess the risk of major or clinically relevant non-major (CRNM) bleeding events in a French ambulatory cohort receiving Vitamin-K antagonists (VKA) or direct oral anticoagulants (DOACs) in a family practice setting.

Methods

CACAO (Comparison of Accidents and their Circumstances with Oral Anticoagulants) was a multicentre prospective cohort of ambulatory patients prescribed oral anticoagulants. We selected patients from the cohort who had received an oral anticoagulant because of non-valvular atrial fibrillation (NVAF) and/or venous thromboembolism (VTE) to be followed during one year by their GP. The following scores were calculated: mOBRI, Shireman, Kuijer, HEMORR₂HAGES, ATRIA, HAS-BLED, RIETE, VTE-BLEED, ACCP score, Rutherford, ABH-Score, GARFIEL-AF, and Outcomes Registry for Better InformedTreatment of Atrial Fibrillation (ORBIT). Prognostic accuracy was assessed by using receiver operating characteristic curves and c-statistics.

Results

During 1 year, 3,082 patients were followed. All of the scores demonstrated only poor to moderate ability to predict major bleeding or CRNM in NVAF patients on DOACs (c-statistic: 0.41–0.66 and 0.45–0.58), respectively. The results were only slightly better for patients prescribed VKA (0.47–0.66 and 0.5–0.55, respectively) in this indication. The results were also unsatisfactory in patients treated for VTE.

Conclusion

None of the scores demonstrated satisfactory discriminatory ability when used in family practice.

Clinical Trial Registration

ClinicalTrials.gov NCT02376777

Introduction

Prevention is one of the most important parts of family practice. Anticoagulants are widely recommended in the prevention of ischaemic stroke in patients with atrial fibrillation (AF) or of the recurrence of deep vein thromboembolism (DVT) or pulmonary embolism (PE) after a first episode.^1–3 These situations are common in family practice. Patients with AF are seen every week, making up to 4% of patients in their 60s and 15% of those in their 80s, A French GP sees 1 to 3 cases of suspected DVT/PE a month, with confirmation of the diagnosis in 1 to 2 times out of 10.⁴ Widely prescribed, anticoagulants have well-proven effectiveness in those indications but are associated with an increased bleeding risk. In the past, Vitamin-K antagonists (VKA) were prescribed, but direct oral anticoagulant (DOAC) prescriptions are gaining more and more ground totalling twice as many new prescriptions in France in 2018.⁵ Post-marketing surveillance studies have found that these medications lead to bleeding in approximately 1.4–12% of patients per year, depending on the patient’s characteristics, the type of anticoagulant used, the type of bleeding events studied, or collection methods.^5–8 Several bleeding risk scores have been developed and validated, especially in VKA-treated patients. More recently, scores have been developed for DOAC-treated patients, such as the Outcomes Registry for Better Informed Treatment of Atrial Fibrillation (ORBIT), VTE-BLEED, the ACCP score, or that proposed by Rutherford et al.^9–12 Other research teams are constantly adding to this field, such as the recent GAMA_RLSTC score.¹³ However, the various scores do not have the same target populations or definitions of the variables of interest. Moreover, only a few have been prospectively validated, and, to our knowledge, none has been validated in family practice. Nevertheless, some studies have attempted to do this, especially for DOACs, but based on nationwide registry data, which do not necessarily correspond to family practice.^14,15 However, when making shared decisions, GPs need information to share with patients on the benefit-risk balance that takes into account “hard” clinical outcomes together with the patient-perceived treatment burden and patient-reported outcomes such as clinically relevant non-major bleedings (CRNMB). The aim of this study was to validate the use of prognostic scores for major and CRNM bleeding risk in a family practice setting.

Methods

CACAO (Comparison of Accidents and their Circumstances with Oral Anticoagulants) was a multicentre prospective observational cohort study conducted by 463 GPs distributed throughout 47 different counties in France. Consecutive unselected ambulatory patients prescribed an oral anticoagulant were included. Patient enrolment and data collection have been previously described.^16,17 Each investigator enrolled all consecutive eligible patients for 3 months, beginning between 1 April 2014 and 31 December 2014. To be included in the phase 2 follow-up study, patients from phase 1 had to have a nonvalvular AF or thromboembolic indication for anticoagulation and to need anticoagulation for more than 3 months.

Patient demographics, comorbidities, medication history, and the indication for anticoagulant use were collected at inclusion. Then every 3 months for 1 year, GPs collected and entered all available data about bleeding events according to the International Society on Thrombosis and Haemostasis (ISTH) classification,^18,19 thrombotic events, and death, in an e-CRF. Major events (severe bleeding and death) were validated by an independent committee, based on supporting documents. CRNMB events were collected but not analysed by the validation committee. The latter were defined as a non-severe bleeding that required care (medical consultation or hospitalization).

The information collected at the time of patient inclusion was used to calculate the CHA₂DS₂VASc risk of stroke score²⁰ and the following bleeding risk scores: Shireman et al.,²¹ mOBRI,²² Kuijer et al.,²³ HEMMORR₂HAGES,²⁴ HAS-BLED,²⁵ ATRIA,²⁶ RIETE,¹² VTE-BLEED,⁹ ACCP score,¹⁰ Rutherford et al., ABH-Score,¹¹ GARFIEL-AF,²⁷ and ORBIT.¹² Each score was calculated (as far as possible) according to the methods described in their respective creation and validation studies. Functional status was evaluated using the “Timed up and go” test (TUG), a validated tool that can readily be integrated into a family practice consultation.²⁸

Statistical analysis

Characteristics of the study population are reported as numbers and percentages, or mean and standard deviation. Comparisons were performed using Chi-squared or Fisher’s exact tests for qualitative variables and Student’s t-test for continuous variables.

The occurrence of death and bleeding during the year of follow-up was analysed with a Kaplan–Meier method. Incidence rates were calculated with a Cox regression model. For haemorrhagic events, we used the Fine and Gray method to calculate the absolute risk of events, taking into account the competing risk of death. To study bleeding risk scores, we considered only events occurring during the initial exposure to the therapy class (DOAC or VKA) and excluded events occurring after a definitive cessation of anticoagulated treatment or after a switch from one class to the other (VKA to DOAC or DOAC to VKA). Score performance was determined by calculating the area under the ROC curve with its 95% confidence interval (CI). The discriminatory capacity of a score in the low, intermediate or high bleeding risk groups was evaluated by the c-statistic. Bleeding risk thresholds were extracted from the sources for each score. The area under the curve (AUC) for each of the scores was compared using the approach described by DeLong et al.²⁹ A principal component analysis of the data was performed to explore the multidimensional properties of these thirteen scores. No imputation of missing data was done.

We performed the analyses using Stata® version 15.0 (Stata Corp, College Station, Texas). A two-sided P < 0.05 was considered as significant.

Funding and ethics

The CACAO study was supported by the French government via the continuing education program for the healthcare professions. This study was approved by the ethics committee of the University Hospital of Saint-Etienne (IRBN112014/CHUSTE). It has been registered in ClinicalTrials.gov (NCT02376777). All patients received written information about the study, emphasizing their right to refuse participation, or to withdraw at any time. No written informed consent was required for inclusion.

Results

As described previously, 3,082 patients were followed during 1 year (Fig. 1), between April 2014 and December 2015.¹⁷ Patient characteristics at baseline are shown for the whole population with follow-up and according to subsequent major or CRNM bleeding events (Table 1). At the end of the follow-up period (n: 3,048, 2,545 person-years), 124 (4.8%) cases of all-cause death, 49 (1.9%) occurrences of at least one arterial or venous event, 48 (1.9%) major, and 72 (2.8%) CRNM bleedings were reported. All types of bleeding, including minor bleeding such as epistaxis, mono-metrorragia, or haematuria, are detailed in Table 2.

Risk categorization (low, intermediate, or high) of all scores for patients treated with VKAs or DOACs, depending on the therapeutic indication (NVAF or DVT-PE), are detailed in the Supplementary Material Table 1 and Table 2. The occurrence of major and CRNM bleeding according to the bleeding risk category (low or intermediate and high) for each score for patients with NVAF treated with DOAC is shown in Graph 1. All of the pronostic scores showed low discriminative capacity for predicting bleeding risk (c-statistic < 0.7) (Table 3).

The principal component analysis explored two dimensions, as shown in the Supplementary Material Figure 1. The first showed homogeneity in the main dimension. The second seemed to show a comorbid dimension, with on one hand, scores from VTE-BLEED, RIETE, ORBIT, or ATRIA considering bleeding factors in priority (age, previous bleedings, anaemia, or chronic kidney disease), and on the other hand, OBRI, HAS-BLED or the ACCP score considered a broader range of comorbid conditions (alcohol consumption, ischaemic factors, or associated treatment).

Discussion

This study sought to validate the use of bleeding risk scores in a family practice population treated with DOAC or VKA for NVAF or DVT-PE. Based on available practice data, all scores found in the literature were calculated and included in the analysis. All showed a low discriminative capacity for predicting bleeding risk (c-statistic < 0.7) in NVAF patients. They do not meet the criteria of discriminative ability expected for clinical practice use in this population. Risk assessment appears to be better in the DVT-PE indication for scores validated for this indication with VKA-treated patients, but this is also true for those targeting a cardiac indication. These scores included fewer parameters, with an emphasis on advanced age and the presence or not, of active cancer (RIETE, VTE-BLEED, HEMMORR₂HAGES). However, the confidence intervals are too wide to conclude. The inter-score match showed that the different scores did not agree in discriminating the same high-risk patients: none of them succeeded in predicting 3 of the 48 bleedings. Twenty-five bleeding events were predicted by one or two scores. Thirteen bleeding events were clearly predicted by 5 different scores.

Comparison with existing literature

Overall incidence rates were similar to those of cohort studies for any bleeding (4.7% in our study vs. 2.85–6.8% in previous studies), and major bleeding (1.9% vs. 1.12–4.04%).^11,30,31 The 2020 European Society of Cardiology (ESC) Guidelines for the management of NVAF recommend to use the HAS-BLED score,³ whereas the 2019 ESC guidelines for the management of PE did not recommend it.² For NVAF, HAS-BLED has been widely evaluated. Its discriminatory abilities progressively decline over time and studies to extend its use have been attempted.³² As Borre et al. clearly showed, its c-statistic remains at best of only moderate quality in warfarin-treated patients compared to 4 other scores (OBRI, ATRIA, ABC, and HEMORR2HAGES). Moreover, no statistical differences were shown between the scores for VKA-treated patients and, more recently, for DOACs-treated patients.^{14,15,33–35} In our opinion, HAS-BLED use does not meet the expectations of clinical practice, as confirmed here in family practice conditions.

Strengths and limitations

The present study includes much information that was not reported in the other studies. Moreover, thanks to the real-time collection of data available in the GP’s medical files, the CACAO study made it possible to address the use of bleeding score in ‘real life’ conditions. However, the sample size was smaller compared to most other studies. We were unable to collect ethnic characteristics because French law restricts the collection of ethnic data without very strong justification. We noted the low exposure to apixaban in our cohort, probably because this molecule arrived on the French market only a few months before the study. Edoxaban is still not marketed in France. Despite our efforts and contacts with corresponding authors, we were unable to test the clinical performance of a new score, GAMA_RLSTC, developed for NVAF patients treated with DOACs.¹³ The published algorithm does not seem to be reproducible. However, AlAmmari et al. propose a tool, developed in a population of Arab origin, for clinical practice that is now available. We could not test the ABC or IMPROVE scores because of the unavailability in family practice of biomarker data required for the first (an important component of the calculation), or too many variables that are not usually available in family practice for the second.^36,37

For the majority of collected variables, we took care to accurately use them according to the definition used in the scores, when it was available. Despite this, the calculation of some scores had to be slightly adjusted because of the CACAO study design, data availability or adaptation to family practice. For example, alcohol consumption (HEMORR₂HAGES, HAS-BLED, SHIREMAN) was defined as “absent,” “normal,” or “excessive” without precision by GPs; and “labile INR” (HAS-BLED) was only recorded qualitatively by GPs. Other variables, such as the genetic variability of cytochrome P450 2C9 (HEMORR₂HAGES), or platelet count (RIETE and ACCP score), were not available. Abnormal prothrombin time (RIETE) was not used because, at inclusion, patients were already being treated with VKAs, or this parameter was not available for those on DOACs. Another limitation was that some scores were published in their definitive version after the launch of the CACAO study, such as the ACCP score, Rutherford, VTE-BLEED, or GARFIELD-AF scores. As a result, not all the necessary variables were collected, such as chronic obstructive pulmonary disease and hospitalization in the past year (Rutherford), heart rate or carotid artery occlusion (GARFIELD-AF), or the presence of metastatic cancer (ACCP score). These adjustments might have led to a bias in the classification of patients towards a lower bleeding risk and may indicate difficulty of use in family practice.

The way in which scores are used by GPs could render them unreliable in their prognostic capacity. The use of a score requires particular vigilance; all items must be completed without omitting any parameter, otherwise the score loses its discriminatory capacity. Scores must therefore be simple and quick for the practitioner to fill in with commonly used variables that are readily available and adapted to the reality of the field. They must be used in their clinical validation context (therapeutic indication, population, care conditions, for example). Once the score has been calculated in the same conditions as its validation, it remains only a measure at a given moment. Its value, especially if the risk of bleeding is high, should encourage the doctor to discuss the subject with the patient and even to propose treatment. Very often, a patient at high risk of bleeding is also the one who would benefit most, as shown by the high CHAD₂DS₂-VASc score in our population (89.4%). To help GPs in risk assessment, the official website presenting several assessment scores in the simplest manner possible at the same time on the same page were envisaged as COAGUCLIC in France. This information, made comprehensible to patients, would allow a face-to-face discussion of the objective embolic and bleeding risks according to their clinico-biological parameters, therapeutic indication, and context of care. Some initiatives already exist, but they are not easy to use in a shared medical decision context with the patient. Finally, GPs and patients should move toward a risk-reduction perspective by focussing on modifiable risk factors and optimizing their management because not all scores are sufficiently discriminating.

Conclusion

None of the scores demonstrated satisfactory discriminatory ability when used in family practice. More research is needed in the family practice context to better assess bleeding risks in patients on oral anticoagulants and to better support physicians and patients in making an informed and shared decision.

Support

The CACAO study was supported by the French government as part of the continuing education program of healthcare workers.

The members of the CACAO Study Group were as follows (n = 463, all in France): Nathan Abenhaïm, Sophie Ackermann, Maryse Adam Blanpain, Xavier Andreu, Céline Arnould, Audrey Atlan-Cottin, Jean-Pierre Aubert, Isabelle Aubin-Auger, Jacques Aubry, Julien Augueux, Veena Augustin, Annick Bakry, Marine Baldesi, Eric Banoun, Eric Barberet, Rémi Bardet, Florence Barriere, Dan Baruch, Nicolas Baude, Marc Bayen, Sabine Bayen, François Bayle, Yannick Beaufils, Alain Beaupin, Julie Bedel-Chauvaud, Raphaël Bel, Martine Bellier, Farouk Bendamene, Philippe Berard, Cédric Berbé, Christophe Berkhout, Jacques Berland, Charles-Edouard Béthembos, Pierre-Yves Billiard, Olivier Bisch, Aurélie Bizeau, Paul Blanchard, Guy Blanquart, Aurélie Boch, Isabelle Bodein, Emmanuel Boige, Claude Bonin, Anne-Laure Bonis, Marie-Pierre Bonnard, Pascal Bonnet, Pierre-André Bonnet, Gérard Bosselut, Anne Bottet, Philippe Bouche, Bérengère Boucherle, Audrey Bougeard, Serge Bouhana, Mourad Boukeloul, Jean Boulet-Gercourt, Jean-Marie Boulongne, Lionel Bouniol, Jean-Jacques Bourcart, Michel Bourgoin, Véronique Bourguignon-Vartanian, Claire Bouteville, Philippe Boutin, Annelore Boutmy, Evelyne Brenner-Girault, Nicolas Breton, Muriel Briane-Fraysse, Marina Brodbeck, Olivier Brunet, Ariel Buchinger, Anne Buffaz-Sutra, Marc Bur, Philippe Cabourdin, Philippe Cachera, Eric Cailliez, Matthieu Calafiore, Denis Calvet, Pierre Camedescasse, Hervé Canart, Michel Cancade, Christophe Candelier, Christian Capiod, Thierry Carin, Olivier Caron, Elisa Carré, Yannick Carrillo, Nathalie Casagrande, Pierre Causse, Gaëlle Chabert, Frédérick Chabord, Juliette Chambe, Guillaume Chambon, Richard Champeaux, Laurent Charbonnel, Rodolphe Charles, Clément Charra, Samuel Chartier, Julie Chastang, Sophia Chatelard, Eric Chatillon, Laurent Chauvot, Hervé Chelle, Stéphane Chenuet, Nicolas Chevalier, David Chevillot, Jean-Pierre Cibeer, Jacques Cittée, Jean-François Claudel, Yvonnick Clemence, Jean-Paul Clerget, Isabelle Clusier-Jeudy, Marie-Pierre Coispeau, Arnaud Colin, Hervé Collart-Dutilleul, Laurence Compagnon, Marine Compan Malet, Laurent Connan, Hubert Conrad, Bruno Coquillaud, Jean-Luc Cormier, Jean-Charles Couette, Yves Cournoyer, Christian Cousin, Stephen Creton, Jean-Jacques Crignon, André Cros, Michel Cunin, Emmanuel Cussac, Jean-Maurice Dailly, Didier Danvin, Pierre-Marie Darnaut, Bruno Daubin, Thomas De L’Hamaide, Yves De Saint Meleuc, Sabine De Taddeo, Jean-Luc Decker, Anne Decobert, Yannick Delattre, Christine Delavenne, Loïc Delavenne, Bénédicte Delbru, Denis Deleplanque, Nicole Delerive, Laurent Delesalle, François Delforge, Jean-Paul Delgrange, Jean-Pierre Delpierre, Eric Demeulemeester, Nicolas Derain, Fabrice Descombe, Thierry Desmoulins, Marie-Claire Deville-Carollo, Jean-Michel Dherbecourt, Michael Didierjean, Salima Domrane, Brigitte Douzou, Sabine Druart, Xavier Dubeau, Arnaud Dubedat, Résika Dudragne, Christian Duez, Lucien Dufour, Sylvie Duhamel, Lionel Duisit, Nathalie Dumoitier, Hervé Dumond, Julien Dumortier, Emilie Duquesne, David Durand, William Durieux, Philippe Durot, Taous Duss, Bénédicte Eschalier, Eric Espiard, Pierre Eterstein, Joël Etienney, Xavier Faidix, Patrice Famery, David Faria, Renaud Faure, André Ferrer, Jean-Marie Ferrer, Jean Feuillet, Christian Fivel, Christian Flaissier, Julien Fortané, Jenny Forté, Claude Fossé, Joël Foucat, Christophe Fouillard, Vianney Fournier, Virginie Fournier, Déborah Fraizy, Paul Frappé, Isabelle Frenoy-Sansarricq, Bernard Gabbai, Yoann Gaboreau, Fabien Gaillard, Isabelle Garnier, Jean-Michel Garnier, Jean-Claude Gascoin, Yves Gault, Martine Gaultier, Bruno Gay, Benoît Gédon, Christophe Genies, Jean-Marc Géniole, Jean-Luc Gentner, Jean-Louis Ghez, Charles Giraldi, Christian Girard, Philippe Giraud, Sylvain Godart, Pierre Goidin, Aline Gomez, Céline Goncalves, Arnaud Gouget, Marie-Laure Gouget, Sylvain Gournay, Florence Grand, Jean-Claude Granier, Roland Greffe, Sabine Grutter, Catherine Gryb, François-Xavier Guedel, Mathieu Guérin, Julien Guiberteau, Jean-François Guille, Jean-Pierre Guillot, Philippe Guillou, Gilles Gustin, Anne Guyot, Jean-Marie Guyot, Christophe Hardy, David Hassid, Latifa Hayani, Anne Heller, Frédéric Henriot, Thierry Hermouet, Bénédicte Hoenner Hecht, Ferreol Honvoh Senadjro, Gaëtan Houdard, Isabelle Huas-Suarez, Corinne Huber, Patrick Imbert, Pascal Jacques, Cécile Jacquet, Jean-Pierre Jacquet, Denis Jacquiot, Philippe Jacquot, Samy Jaffre, Clémence Jean, Pauline Jeanmougin, Isabelle Jeudy, Jean-Philippe Joseph, Pierre Jouannic, Stéphanie Jousson, André Kastelik, Gérard Klifa, François-René Knockaert, Jean-Philippe Koch, Julia Krotoff, Charlotte Labrune, Véronique Lacaille-Smerilli, François Lacoin, Xavier Lainé, Philippe Lambert, Audrey Lambourg, Dominique Lamy, Myriam Lapenne-Creusot, Jean-Dominique Laporte, Lucie Lartaud, Nadège Lauchet, Dorothée Lavielle, Julien Le Breton, Raphaël Le Diagon, Marie Le Du, Delphine Le Goff, Claire Le Lann, Stéphane Le Mouël, Carole Leblan-Depelsenaire, Frédéric Leclercq, Rudolphe Lécutier, Jean-Nicolas Ledoux, Jean-Marc Lefebvre, Pascal Léger, Robert Lendais, Eric Lengagne, Catherine Sosiewicz-Lengsavath, Sylvie Lenoir, Patrick Lerouge, Benoît Leroy, Ida Leung-Y-Tai, Pascal Ligier, Charlotte Lobel, François Loez, Béatrice Lognos, Baptiste Luaces, Laurence Lucas-Couturier, Anne Lunven, Sigolène Machraoui, André Maciejewski, Loïc Magnen, Georgios Makridis, Jean-Luc Malbrunot, Jean-Marc Mancini, Hervé Mangin, Jean-Marc Maniglier, Philippe Marchant, Julie Marcus, Guillaume Marien, Anne-Laure Martin-Etzol, Christian Mas, Yannick Masset, Pierre Massin, Bruno Masson, Agnès Mattera, Charlotte Matz, François Maufoy, Collette Maury, Laure-Emmanuelle Mavraganis, Olivier Mazin, Lison Mazué, Jean-Philippe Melizan, Dominique Menard, Alain Mercier, Patricia Mercier, Patric-Alain Meyer, Philippe Michellier, Corinne Milleret, Yannick Millot, Françoise Minard, Marie-Léa Miqueu, Dany Mismacque, Catherine Mitifiot, Jean-Louis Moebs, Jean-Michel Monnier, Yves Montariol, Jérémie Montauze, Alain Morand, Gilles Morel, François Morlon, Delphine Mortas, Baptiste Motte, Stéphane Mouget, Luc Munro, Martin Naessens, Bounthanousone Nammathao, Dominique Negre, Charlie Nogrel, Lucile Nouvellet, Ismaël Nureni Banafunzi, Thierry Oger, Jean-François Ortholan, Dominique Osty, Agnès Oude Engberink, Guy-Marc Paillard, Laurent Paillard, Marie-Paule Pautout-Guillaume, Laurent Pech-Gourg, Corinne Perdrix, Véronique Perez, Ségolène Perrillat-Amédé, Christophe Peyrou, Bruno Pichat, Christophe Pigache, Alexis Pinot, Dominique Piquard, Thierry Piquet, Michel Placet, Michel Plauchier, Caroline Pluskota, Maurice Ponchant, Chantal Prat, Martine Prevost, Jean-Pierre Prigent, Thibaut Py, Christian Rafin, Jacques Rambaud, Jean-Paul Rapoud, Anne-Marie Regnier, Sylvain Renaudin, Jean-Michel Rétaux, Amélie Richard, Philippe Richetta, Jean-Michel Rigault, Reinold Rigoli, Anne Ritter-Meinicke, Sarah Robert, Stéphanie Rollin, Didier Rondepierre, Sophie Rosenberg, Mélanie Roth, Fabien Rougerie, Guillaume Royer De Vericourt, Karen Rudelle, Philippe Ruelle, Marcel Ruetsch, Dominique Saillard, Pénélope Saint-Denis, Pietro Sannelli, Philippe Saraidarian, Jean-Pascal Sastourné, Laurent Sauvage, Christian Scellier, Christian Schaal, François-Xavier Schelcher, Daniel Schirlin, Anne Schirrer, Claude Schlienger, Joëlle Schlienger, Philippe Serayet, Denis Serramoune, Marlène Siebler, Jean-Paul Simon, André Soares, Carine Soussotte-Ducasse, Philippe Stefanuto, Marc Steinberger, Marianne Szapiro, Anas Taha, Erol Taluy, Gilles Tanguy, Dominique Tardieux, Michel Tardy, Benoît Tavernier, Jean-Luc Ténédos, Lorène Thelot-Bach, Rémy Tisserand, Audrey Tordoir, François Trillot, Pascal Triouleyre, David Truong, Laurent Turi, Frédéric Vaillant, Hélène Vaillant-Roussel, Pierre Vailler, Josette Vallee, Muriel Vampouille, Jean-Louis Vangi, Fabien Vannier, Simon Varin, Florence Vaugeois, Jean-Charles Vauthier, Virginie Vauthier, Delphine Veillard, Anne-Laure Verjus, Paul-Bernard Verjus, Gilles Verney, Eloïse Vialtel, Fernand Vierling, Graziella Virgone-Rebaud, Marc Vital-Durand, Eric Vittori, Nadège Volcler, Philippe Vorilhon, Pierre Watteau, Christine Weisbecker, Nathalie Wey, Françoise Wilhelm-Nenot, Jean-Louis Wurtz, Patricia Yvon, Claire Zabawa, Jean-Marc Zamboni, Anne-Claire Zipper

Acknowledgments

We wish to thank all patients and GPs who participated in this study.

Funding

The CACAO study was supported by the French government via the continuing education program for the healthcare professions.

Ethical approval

This study was approved by the ethics committee of the University Hospital of Saint-Etienne (IRBN112014/CHUSTE). It has been registered in ClinicalTrials.gov (NCT02376777).

Conflict of interest

None declared.

Data availability

The data underlying this article will be shared upon reasonable request to the corresponding author.

References