Open access Research Persistence with dual antiplatelet therapy after percutaneous coronary intervention for ST-segment elevation acute coronary syndrome: a population- based cohort study in Catalonia (Spain) Aida Ribera,1,15 Ignacio Ferreira-Gonzalez,1,15 Josep Ramon Marsal,1,15 Gerard Oristrell,1,16 Maria Teresa Faixedas,2 Alba Rosas,3 Helena Tizón-Marcos,4 Sergio Rojas,5 Carlos Labata,6 Merida Cardenas,7 Silvia Homs,8 Carlos Tomas-Querol,9 Joan Garcia-Picart,10 Joan A Gomez-Hospital,11 Jose Ignacio Pijoan,12,15 Monica Masotti,13 Josepa Mauri,3,6 David Garcia Dorado,1,16 on behalf of the VESA Study and Codi IAM Investigators To cite: Ribera A, Ferreira- Abstract Gonzalez I, Marsal JR, Strengths and limitations of this study Objectives Guidelines recommending 12-month et al. Persistence with dual dual antiplatelet therapy (DAPT) in patients with ST- antiplatelet therapy after ►► The study describes the trends in persistence with elevation acute coronary syndrome (STEACS) undergoing percutaneous coronary dual antiplatelet therapy (DAPT) during 2010–2015 percutaneous coronary intervention (PCI) were published in intervention for ST-segment in a region-wide comprehensive cohort of patients elevation acute coronary year 2012. We aimed to describe the influence of guideline using administrative data linked to a clinical registry. syndrome: a population- implementation on the trend in 12-month persistence ►► It also evaluates the impact of the DAPT duration based cohort study in with DAPT between 2010 and 2015 and to evaluate its recommended at the percutaneous coronary inter- Catalonia (Spain). BMJ Open relationship with DAPT duration regimens recommended vention hospital discharge on 12-month persistence. 2019;9:e028114. doi:10.1136/ at discharge from PCI hospitals. ►► Limitations of using observational registry data in- bmjopen-2018-028114 Design Observational study based on region-wide registry clude the possibility of coding errors and the inability ►► Prepublication history and data linked to pharmacy billing data for DAPT follow-up. to accurately identify specific contraindications for additional material for this paper Setting All PCI hospitals (10) belonging to the acute treatment or other patient characteristics that might are available online. To view myocardial infarction (AMI) code network in Catalonia be relevant for the study aims. please visit the journal (http://​ (Spain). ►► The use of pharmacy refill data as a proxy of pa- dx.​doi.​org/​10.​1136/​bmjopen-​ Participants 10 711 STEACS patients undergoing PCI 2018-​028114). tients’ adherence and persistence has also limita- between 2010 and 2015 were followed up. tions which have been extensively described. Primary and secondary outcome measures Primary Received 26 November 2018 Revised 19 June 2019 outcome was 12-month persistence with DAPT. Calendar Accepted 2 July 2019 year quarter, publication of guidelines, DAPT duration regimen recommended in the hospital discharge report, PCI. A guideline-based recommendation at PCI hospital baseline patient characteristics and significant interactions discharge was highly associated with full persistence with were included in mixed-effects logistic regression based DAPT. Establishing evidence-based, common prescribing interrupted time-series models. criteria across hospitals in the AMI-network would favour Results The proportion of patients on-DAPT at 12 months adherence and reduce variability. increased from 58% (56–60) in 2010 to 73% (71–75) in 2015. The rate of 12-month persistence with DAPT significantly increased after the publication of clinical Introduction © Author(s) (or their guidelines with a time lag of 1 year (OR=1.20; 95% CI 1.11 The need of dual antiplatelet therapy (DAPT) employer(s)) 2019. Re-use to 1.30). A higher risk profile, more extensive and complex combining aspirin and an ADP-receptor permitted under CC BY-NC. No coronary disease, use of drug-eluting stents (OR=1.90; commercial re-use. See rights blocker for at least 12 months in patients 95% CI 1.50 to 2.40) and a 12-month DAPT regimen and permissions. Published by with ST-elevation acute coronary syndrome recommendation at discharge from the PCI hospital BMJ. (STEACS) undergoing percutaneous coro- (OR=5.76; 95% CI 3.26 to 10.2) were associated with For numbered affiliations see 12-month persistence. nary intervention (PCI) is well established end of article. and was incorporated into clinical guidelines Conclusion Persistence with 12-month DAPT has Correspondence to increased since publication of clinical guidelines. Even in 2012.1 2 Dr Ignacio Ferreira-Gonzalez; though most patients were discharged on DAPT, only 73% Adherence of patients to this strategy is ​nachoferreira@​secardiologia.​es with potential indication were on-DAPT 12 months after crucial to ensure its efficacy. Adherence to Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 1 Open access medication is usually defined3 as the extent to which The registry was completed for the purpose of the patients take medications as prescribed by their healthcare present analysis with retrospective collection of addi- providers and persistence is defined as time from initia- tional specific data: diseased vessels, responsible vessel, tion to discontinuation of a therapy. Patients’ persistence stent type and number of stents. The recommendation with DAPT may be influenced by several factors but will of antithrombotic drugs was also collected ad hoc for the depend strongly on whether they ultimately receive a study from the discharge report. The recommendation of correct prescription from their physicians in the primary DAPT was defined as the recommendation of acetylsali- care setting. Patients may receive recommendations silic acid (ASA) and clopidogrel, prasugrel or ticagrelor from various health providers at different stages of their for specified periods. If the recommended duration of process of care, from the interventionist cardiologist to DAPT was not specified, the discharge recommendation their primary physician. It could be hypothesised that the pattern was classified as ‘unspecified’. A local investigator last would tend to rely on the recommendation of the at each centre performed the specific retrospective data more specialised health professional. Thus, one poten- collection. tial determinant of patients’ persistence with DAPT for History of major haemorrhage, neoplasia, renal at least 12 months is the instructions provided in the disease, heart failure, peripheral arterial disease and discharge report of the hospital where the patient was atrial fibrillation, were obtained from minimum basic attended during the acute phase. data set diagnoses coded in hospitalisation episodes In Catalonia, an autonomous region of Spain, the occurring in the previous 3 months before index hospi- acute care of STEACS is organised through a region-wide talisation. Major haemorrhage was defined as: a diagnosis network, the acute myocardial infarction (AMI) code, to of digestive bleeding in any diagnostic position (primary derive patients with suspected STEACS to 1 of the 10 refer- or secondary) together with a procedure code for endo- ence hospitals with PCI capability. Performance of the scopic treatment or for transfusion of blood products, AMI code is prospectively and exhaustively registered,4 5 or a diagnosis of haemorrhagic stroke, or a diagnosis of providing an appropriate tool for quality evaluation. The intraocular haemorrhage, or a diagnosis of other types of Catalan Health Information System systematically regis- haemorrhage together with a procedure code for trans- ters, among other, data on pharmacy refills. Pharmacy fusion of blood products. Major ischaemic events (AMI billing data, although indirect, is an accepted method or stroke) and major haemorrhage during the 12 months for evaluating persistence with treatment in large patient following the index episode were obtained in the same cohorts.3 6 way. Mortality during the 12 months following the index The aims of the present study were, first, to describe episode was obtained from the insured registry status. persistence with DAPT for at least 12 months in patients Drug treatment during the 12-month postdischarge with STEACS undergoing PCI from 2010 to 2015; second, follow-up was obtained from the pharmacy billing registry. to evaluate the influence guidelines recommendation International Classification of Diseases, Ninth Revision, for a 12-month DAPT schedule on the rate of 12-month Clinical Modification (ICD-9-MC) and ATC codes used persistence along time; and third, to evaluate the asso- for the identification of study variables are listed in the ciation of the DAPT duration recommended at the PCI online supplementary tables 1 and 2. hospital discharge with patients’ persistence with treat- ment for at least 12 months. Data availability statement Additional data are not available due to ethical require- ments of the PADRIS. Methods Data sources Study population Data were obtained through the Public Data Analysis for We included all consecutive patients who survived a Health Research and Innovation Programme (PADRIS). STEACS between January 2010 and December 2015, The PADRIS allows access to information from different received primary or postfibrinolysis PCI in 1 of the 10 sources on public healthcare resources usage for the reference hospitals of the AMI code network, were population of Catalonia linked at the patient level with discharged home or transferred to another hospital and warranted accomplishment of ethical principles. Specifi- survived at least 1 month after AMI. New episodes of cally, for the present study we linked data of the pharmacy STEACS occurring to the same patients during the study billing registry with the AMI code registry. The AMI code period were only accounted as follow-up events. Patients registry was launched in 2010 to evaluate performance of with likely contraindication for DAPT (history of major the AMI code.4 5 Exhaustiveness and quality of data are bleeding or neoplasm in the 3 months prior to the index assessed periodically (see online supplementary methods episode and patients requiring anticoagulation) were for details). The database belongs to the Catalan Depart- excluded. ment of Health and includes demographic, clinical and therapeutic data for each episode of hospitalisation for Persistence with treatment STEACS. It conforms to the ethical and legal require- DAPT was defined as the concomitant use of ASA and a ments for research purposes. P2Y12 antagonist. Persistence with DAPT was estimated by 2 Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 Open access identification of consecutive months with pharmacy refills following the implementation and ‍β3‍ indicates the slope with one container of each agent in the 12-month period change following the implementation (using the interac- after hospital discharge. Because pharmacy billing is regis- tion between time and implementation: TXt). Additional tered in a monthly basis and the exact day of dispensation terms can be added to model the effect of other covaria- is unknown, we considered that a monthly dispensation bles and their interactions with T and Xt and to include until at least month 11 after the index episode would random effects. Note that we set T = 0 at the quarter approximate a 12-month treatment period. If more than where we observed a significant change in the slope at a one container were dispensed in 1 month, the excess lag time after guidelines publication. containers were pulled along the following months. We took into account the clustered structure of data Non-persistence was defined as either discontinuation with patients being treated and, most importantly, with or a break in therapy of at least 2 months after pulling recommendations on DAPT duration being provided in along the excess containers. To describe persistence over different hospitals, by introducing random effects in the the whole study period, we estimated the proportion of logistic regression models. patients alive and within the 12 months after discharge We tested whether models including random intercepts window who were on treatment on each month.7 for hospital and random slopes for each independent The primary outcome was a patient’s persistence with variable were significant using a deviance-based test of DAPT for 12 months following discharge (or in other hypothesis. words, patients withdrawing both agents from the phar- Variable selection for multilevel modelling was based macy until at least month 11). on the bivariate associations with the rate of each depen- dent variable. Candidate individual variables were those Statistical analysis described in tables 1 and 2. We retained in the final We compared baseline characteristics between patients model all variables with a p value <0.2. persisting with DAPT for at least 12 months and patients Plots of predicted probability values were used to show withdrawing DAPT before 12 months with χ2 test or t-test marginal effects of variables of interest and variability when appropriate. We tested for trends in patients’ char- between centres. acteristics along calendar year of discharge for the index procedure with Jonckheere-Terpstra test for differences Sensitivity analyses between ordered categories. Because a substantial proportion of patients were To evaluate the influence of time, guidelines publica- returned to their reference hospital and because it was tion and the DAPT duration recommended in the PCI unknown whether the DAPT duration recommendation hospital discharge report, we modelled logistic regres- was changed at discharge from the second hospital, we sion based interrupted time-series analysis,8 adjusting performed sensitivity analyses excluding these patients. for baseline characteristics. As a first step, because it is Additionally, because ischaemic and haemorrhagic expected that guidelines publication influences practice events occurring during follow-up would change the with a time delay, we plotted the proportion of patients treatment length, sensitivity analyses were also performed persisting on-DAPT for 12 months by year quarter of by excluding patients suffering any vascular event during discharge from the PCI hospital and we tested models follow-up. with a slope change (indicating the start of guideline implementation) at different lag periods after publication Patient and public involvement of the European clinical guidelines (last quarter of 2012). Neither patients nor public were directly involved in the Once the lag period between guidelines publication and study. implementation of recommendations was estimated, we included patient characteristics and second or third level interactions of each characteristic with year quarter and Results moment of implementation. We also included an auto- After excluding patients with likely contraindication correlation term. We coded time (T) as the time elapsed for DAPT (figure 1), we identified 10 711 STEACS since the publication of guidelines plus the lag period patients undergoing PCI who were potential candidates (in quarters) and a dummy variable (Xt) indicating the to receive DAPT for at least 12 months and survived for preimplementation period (coded 0) or the postimple- at least 1 month after discharge. Six hundred and thir- mentation period (coded 1).9 ty-one (5.9%) patients experienced an ischaemic major The standard model specification was the following: event (AMI or stroke) within 12 months after the index ( ) episode, 100 (0.9%) had a major haemorrhage and 280 logit ‍ Yt = β0 + β1 T + β2 Xt + β3 TXt ‍ (2.6%) died between 1 and 12 months after the index episode. After excluding patients who died or were lost to Where ‍β0‍ represents the baseline level at T = 0, ‍β1‍ is follow-up and patients with errors in quarter allocations, interpreted as the change in outcome associated with a 10 262 patients remained for analysis. time unit (quarter) increase (representing the under- Table 1 shows characteristics of study patients lying preimplementation trend), ‍β2‍ is the level change depending on persistence with DAPT. Patients persisting Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 3 Open access Table 1 Characteristics of study patients according to DAPT persistence during follow-up DAPT <12 months DAPT ≥12 months Total (n=3684) (n=6578) (n=10 262) N n (%) N n (%) N n (%) P value Age 3684 61.19±13.21 6578 61.18±12.38 10 262 61.19±12.69 0.578 Gender (female) 3684 765 (20.8%) 6578 1319 (20.1%) 10 262 2084 (20.3%) 0.399 Smoke (Y) 3684 1774 (48.2%) 6578 3042 (46.2%) 10 262 4816 (46.9%) 0.064 Hypertension (Y) 3684 1622 (44%) 6578 3142 (47.8%) 10 262 4764 (46.4%) <0.001 Diabetes (Y) 3684 647 (17.6%) 6578 1298 (19.7%) 10 262 1945 (19%) 0.008 Hypercholesterolaemia (Y) 3684 1291 (35%) 6578 2678 (40.7%) 10 262 3969 (38.7%) <0.001 Polyvascular disease (Y)* 3684 530 (14.4%) 6578 1048 (15.9%) 10 262 1578 (15.4%) 0.038 Previous stroke or transient ischaemic attack (Y) 3684 101 (2.7%) 6578 172 (2.6%) 10 262 273 (2.7%) 0.704 Previous acute myocardial infarction (Y) 3684 265 (7.2%) 6578 570 (8.7%) 10 262 835 (8.1%) 0.010 Previous percutaneous coronary intervention (Y) 3684 194 (5.3%) 6578 482 (7.3%) 10 262 676 (6.6%) <0.001 Previous by-pass surgery (Y) 3684 20 (0.5%) 6578 85 (1.3%) 10 262 105 (1%) <0.001 History of peripheral arteriopathy (Y) 3684 153 (4.2%) 6578 276 (4.2%) 10 262 429 (4.2%) 0.957 Comorbidity (Y)† 3684 462 (12.5%) 6578 885 (13.5%) 10 262 1347 (13.1%) 0.191 Hepatopathy (Y) 3684 48 (1.3%) 6578 56 (0.9%) 10 262 104 (1%) 0.030 History of renal impairment (Y) 3684 185 (5%) 6578 320 (4.9%) 10 262 505 (4.9%) 0.741 History of heart failure (Y) 3684 272 (7.4%) 6578 583 (8.9%) 10 262 855 (8.3%) 0.009 Affected number of vessels ≥2 (Y) 3684 1275 (34.6%) 6578 2731 (41.5%) 10 262 4006 (39%) <0.001 No of treated vessels 3613 1.03±0.21 6516 1.06±0.26 10 129 1.05±0.25 <0.001 No of stents 3681 1.09±0.6 6563 1.21±0.68 10 244 1.16±0.65 <0.001 Drug-eluting stent (Y) 3684 572 (15.5%) 6578 2704 (41.1%) 10 262 3276 (31.9%) <0.001 DAPT score points 3684 1.20±1.20 6578 1.28±1.17 10 262 1.25±1.18 <0.001 Discharged home (Y) 3684 2233 (60.6%) 6578 3672 (55.8%) 10 262 5905 (57.5%) <0.001 Antiplatelet agent at discharge 3684 6578 10 262 <0.001 Clopidogrel 3184 (86.4%) 4872 (74.1%) 8056 (78.5%) Prasugrel 278 (7.5%) 1102 (16.8%) 1380 (13.4%) Ticagrelor 222 (6%) 604 (9.2%) 826 (8%) DAPT recommendation at discharge 3684 6578 10 262 <0.001 1 month 875 (23.8%) 295 (4.5%) 1170 (11.4%) <12 months 385 (10.5%) 173 (2.6%) 558 (5.4%) ≥12 months 1522 (41.3%) 4732 (71.9%) 6254 (60.9%) Unknown 902 (24.5%) 1378 (20.9%) 2280 (22.2%) *Polyvascular disease was defined as presence of at least two of the following conditions: previous myocardial infarction or percutaneous coronary or surgical revascularisation; history of peripheral arteriophaty; history of stroke or transient ischaemic attack. †Comorbidity was defined as presence of one of the following conditions: hepatophaty, history of renal impairment, history of heart failure. DAPT, dual antiplatelet therapy. for at least 12 months had higher prevalence of cardio- 1). The proportion of patients on-DAPT at 12 months vascular risk factors (hypertension, diabetes and hyper- significantly increased from 58% (56–61) in 2010 to cholesterolaemia), higher rate of a previous history 73% (71–75) in 2015. The larger increase in 12-month of cardiovascular disease, more extended coronary persistence was observed between 2014 (64% (62–66)) disease, higher rate of drug-eluting stents implantation and 2015, 2 years after the publication of clinical guide- and slightly higher ischaemic risk (as measured with lines. The proportion of patients with prasugrel or tica- the DAPT score10). Persisting patients were more often grelor instead of clopidogrel started increasing after 2012 transferred to their reference hospital and had had a (online supplementary figure 2). prescription for a longer DAPT period at discharge Some baseline characteristics showed a temporal trend from the PCI hospital. over the study period 2010–2015 (table 2). The prev- The rate of patients on-DAPT after discharge from the alence of cardiovascular risk factors (smoking, hyper- PCI hospital was 91% (95% CI 90 to 91) without relevant tension and hypercholesterolaemia) and comorbidities differences between years (online supplementary figure increased slightly. Likewise, the number of treated vessels 4 Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 Table 2 Characteristics of study patients by year of discharge for the index admission 2010 (n=1537) 2011 (n=1628) 2012 (n=1760) 2013 (n=1779) 2014 (n=1779) 2015 (n=1779) P value Age 61±12.92 61.01±13.09 60.97±12.52 61.27±12.4 61.24±12.72 61.58±12.53 0.325 Gender 310 (20.2%) 311 (19.1%) 372 (21.1%) 359 (20.2%) 352 (19.8%) 380 (21.4%) 0.389 Smoke 686 (44.6%) 754 (46.3%) 811 (46.1%) 839 (47.2%) 840 (47.2%) 886 (49.8%) 0.004 Hypertension 674 (43.9%) 729 (44.8%) 808 (45.9%) 838 (47.1%) 844 (47.4%) 871 (49%) 0.001 Diabetes 296 (19.3%) 319 (19.6%) 340 (19.3%) 327 (18.4%) 325 (18.3%) 338 (19%) 0.438 Hypercholesterolaemia 541 (35.2%) 621 (38.1%) 648 (36.8%) 710 (39.9%) 742 (41.7%) 707 (39.7%) <0.001 Polyvascular disease* 232 (15.1%) 253 (15.5%) 301 (17.1%) 259 (14.6%) 278 (15.6%) 255 (14.3%) 0.370 Previous stroke or transient ischaemic attack 42 (2.7%) 32 (2%) 45 (2.6%) 40 (2.2%) 65 (3.7%) 49 (2.8%) 0.126 Previous acute myocardial infarction 127 (8.3%) 142 (8.7%) 163 (9.3%) 151 (8.5%) 127 (7.1%) 125 (7%) 0.029 Previous percutaneous coronary intervention 94 (6.1%) 104 (6.4%) 123 (7%) 116 (6.5%) 118 (6.6%) 121 (6.8%) 0.485 Previous by-pass surgery 10 (0.7%) 16 (1%) 17 (1%) 19 (1.1%) 18 (1%) 25 (1.4%) 0.056 History of peripheral arteriopathy 51 (3.3%) 70 (4.3%) 93 (5.3%) 66 (3.7%) 86 (4.8%) 63 (3.5%) 0.876 Comorbidity† 155 (10.1%) 190 (11.7%) 233 (13.2%) 243 (13.7%) 266 (15%) 260 (14.6%) <0.001 Hepatopathy 16 (1%) 12 (0.7%) 24 (1.4%) 24 (1.3%) 13 (0.7%) 15 (0.8%) 0.578 Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 History of renal impairment 70 (4.6%) 75 (4.6%) 71 (4%) 86 (4.8%) 89 (5%) 114 (6.4%) 0.009 History of heart failure 80 (5.2%) 118 (7.2%) 154 (8.8%) 159 (8.9%) 187 (10.5%) 157 (8.8%) <0.001 Affected number of vessels ≥2 594 (38.6%) 617 (37.9%) 679 (38.6%) 707 (39.7%) 688 (38.7%) 721 (40.5%) 0.188 No of treated vessels 1.04±0.25 1.03±0.23 1.05±0.26 1.05±0.24 1.05±0.26 1.06±0.24 0.003 No of stents 1.23±0.69 1.17±0.60 1.18±0.63 1.14±0.66 1.12±0.67 1.16±0.66 <0.001 Drug-eluting stent 462 (30.1%) 388 (23.8%) 472 (26.8%) 493 (27.7%) 652 (36.6%) 809 (45.5%) <0.001 DAPT score points 1.24±1.2 1.24±1.19 1.26±1.17 1.26±1.17 1.22±1.16 1.28±1.19 0.435 Discharged home 961 (62.5%) 951 (58.4%) 933 (53%) 1042 (58.6%) 959 (53.9%) 1059 (59.5%) 0.041 Antiplatelet agent at discharge <0.001  Clopidogrel 1528 (99.4%) 1587 (97.5%) 1582 (89.9%) 1384 (77.8%) 1114 (62.6%) 861 (48.4%)  Prasugrel 9 (0.6%) 38 (2.3%) 135 (7.7%) 244 (13.7%) 416 (23.4%) 538 (30.2%)  Ticagrelor 0 (0%) 3 (0.2%) 43 (2.4%) 151 (8.5%) 249 (14%) 380 (21.4%) DAPT recommendation at discharge <0.001  1 month 281 (18.3%) 259 (15.9%) 243 (13.8%) 181 (10.2%) 135 (7.6%) 71 (4%)  <12 months 150 (9.8%) 137 (8.4%) 100 (5.7%) 69 (3.9%) 65 (3.7%) 37 (2.1%)  ≥12 months 792 (51.5%) 871 (53.5%) 975 (55.4%) 1086 (61%) 1154 (64.9%) 1376 (77.3%)  Unknown 314 (20.4%) 361 (22.2%) 442 (25.1%) 443 (24.9%) 425 (23.9%) 295 (16.6%) *Polyvascular disease was defined as presence of at least two of the following conditions: previous myocardial infarction or percutaneous coronary or surgical revascularisation; history of peripheral arteriophaty; history of stroke or transient ischaemic attack. †Comorbidity was defined as presence of one of the following conditions: hepatophaty, history of renal impairment, history of heart failure. DAPT, dual antiplatelet therapy. Open access 5 Open access Figure 2 Observed proportion of patients persisting with DAPT for at least 12 months at each quarter and the interrupted time series model fitted after setting a 1 year lag period from publication to implementation of guidelines. DAPT, dual antiplatelet therapy. procedure, taking prasugrel instead of clopidogrel and a recommendation of DAPT for a longer period at discharge from the PCI hospital. Autocorrelation was not signif- icant. Guideline implementation had a positive effect on persistence: a 20% increase in the odds of 12-month persistence each quarter after a lag of 1 year since publica- tion. The effect of drug-eluting stents was attenuated with time (OR for interaction: 0.96; 95% CI 0.94 to 0.97) while the effect of prescription was attenuated with time after guideline implementation (OR for the interaction 0.86, Figure 1 Patients flow. AMI, acute myocardial infarction; 95% CI 0.79 to 0.94 for a recommendation of ≥12 months DAPT, dual antiplatelet therapy; PCI, percutaneous coronary and 0.88, 95% CI 0.81 to 0.97 for an unknown recommen- intervention; ACO, anticoagulant therapy. dation). The effect of implantation of drug-eluting stents and type of recommendation also varied between hospi- the rate of drug-eluting stents implantation and the rate tals (significant random slopes). of prasugrel or ticagrelor also increased with time. Results of sensitivity analyses, excluding patients with The overall rate of explicit DAPT recommendation for ischaemic or haemorrhagic events during follow-up, at least 12 months in the hospital discharge reports was or excluding patients who were transferred to another 51% (49–53) in 2010 and increased to 77% (75–79) in centre after PCI were similar to the main analysis (online 2015 but it was highly variable between hospitals (online supplementary table 3S). supplementary figure 3). The interaction between drug-eluting stents and time Figure 2 shows the observed proportion of patients can be seen in figure 3A. Because 12-month persistence persisting with DAPT for at least 12 months at each time increased with time in patients without drug-eluting point and the interrupted time series model fitted after stents, the effect of type of stent is attenuated with time. setting a 1 year lag period from publication to implemen- The interaction of the recommendation pattern with time tation of guidelines. and guideline implementation can be seen in figure 3B: Table 3 shows results of the complete cases analysis 12-month persistence increased with time mainly in the (n=10 244) using interrupted time series logistic regres- subgroups with shorter time specification in the discharge sion. Variables showing association with 12-month report and also in patients without a specific recommen- persistence on DAPT were two or more diseased vessels, dation, but this increase started after guideline imple- higher number of stents implanted, receiving drug- mentation (1 year after publication). Figure 3C shows a eluting stents, hypercholesterolaemia, a previous surgical substantial reduction in the variability between centres 6 Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 Open access Table 3 Factors associated with a persistence of at least 12 months as assessed with interrupted time series logistic regression model Fixed effects OR 95% CI P value Drug-eluting stent 1.90 1.50 to 2.40 <0.001 No of stents 1.22 1.13 to 1.32 <0.001 Antiplatelet agent at discharge (Ref. clopidogrel)   Prasugrel 1.59 0.88 to 1.26 <0.001   Ticagrelor 1.05 1.36 to 1.86 0.575 Recommendation at PCI hospital discharge (Ref. 1 month)  <12 months 1.67 0.89 to 3.14 0.110  ≥12 months 5.76 3.26 to 10.2 <0.001   Unknown 2.25 0.84 to 6.01 0.107 Hypercholesterolaemia 1.19 1.08 to 1.31 <0.001 Previous by-pass 1.85 1.09 to 3.14 0.023 surgery Two or more treated 1.21 1.10 to 1.33 <0.001 vessels Drug-eluting stent * 0.96 0.94 to 0.97 <0.001 time (quarter) Guideline 1.20 1.11 to 1.30 <0.001 implementation Recommendation at PCI hospital discharge (Ref. 1 month) * time (quarter) * guideline implementation Figure 3 Temporal trend of interhospital variability in  <12 months 0.90 0.79 to 1.04 0.144 12-month DAPT recommendation at the PCI hospital,  ≥12 months 0.86 0.79 to 0.94 <0.001 measured as the percentage of variance explained by the   Unknown 0.88 0.81 to 0.97 0.007 hospital level (intraclass correlation and 95% CI). Predicted probabilities of 12-month persistence by (A) drug-eluting Random effects Variance 95% CI stent, (B) recommendation pattern and (C) centre, over time. Random—intercept 0.46 0.22 to 1.53 DAPT, dual antiplatelet therapy; PCI, percutaneous coronary Random—slopes intervention. Recommendation at PCI hospital discharge (Ref. 1 month) Discussion According to published guidelines, all STEACS patients  <12 months 0.37 0.17 to 1.23 undergoing PCI without contraindication should be kept  ≥12 months 0.39 0.19 to 1.31 on DAPT for at least 12 months unless an event occurs that   Unknown 1.97 0.93 to 6.58 precludes continuing with this treatment. In this observa- Drug-eluting stent 0.10 0.05 to 0.33 tional region-wide study, we have found an increase in the Adjusted Intracalss 0.085 proportion of patients on-DAPT at 12 months from 58% Correlation Coefficient to 73% in the period from 2010 to 2015, with an accel- PCI, percutaneous coronary intervention. erated rate starting in the fourth quarter of 2013, 1 year after the publication of European guidelines. We also found a high variability between hospitals in mainly due to an increase in the proportion of 12-month the adherence to guidelines when recommending DAPT persistence in patients attended in centres where the for at least 12 months which leads to substantial differ- initial proportion was lower (significant random inter- ences between hospitals in the rate of patients persisting cept and slopes). with the recommended DAPT. The progressive increase Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 7 Open access in the overall rate of 12-month persistence was accompa- of 1 month for patients receiving BMS and 6 months for nied by a substantial reduction of interhospital variability. patients receiving DES’, with class IC and IIb, respectively. Likelihood of patients persisting with DAPT for These messages, which ultimately reflected the lack of 1 year is strongly related to the instructions given at the clinical trials aimed to answer the specific question about PCI hospital discharge, as we observed a lower rate of DAPT duration, could have induced a perception of arbi- 12-month persistence in patients receiving a discharge trariness leading to variability in clinical practice. DAPT recommendation for <12 months. Although a In fact, the optimal duration of DAPT has not yet been causal direct relationship between the established recom- totally established in more contemporary clinical trials. mendation in the more specialised setting and the final The most recent randomised clinical trial conducted prescription at the primary care setting cannot be stated in patients with STEACS aimed to assess the question on the basis of observational data, this finding suggests of 12-month vs a 6-month DAPT duration, showed that that prescribing physicians strongly rely on the first 6-month DAPT duration after primary PCI was non-in- recommendation specified at the discharge report in the ferior to 12-month duration to prevent major cardiovas- PCI hospital. Therefore, hospital cardiologists should be cular events.19 In another trial in the context of ACS, 12 kept aware of their impact and encouraged to be clear months or longer DAPT duration versus 6 months was and specific enough when providing DAPT time recom- not associated with lower major cardiovascular events and mendations in the discharge report form. total mortality.20 The recommendation in clinical guidelines of DAPT Regardless the level of the evidence, one would expect for at least 12 months following STEACS1 2 was based on that a class I recommendation should be uniformly the duration of follow-up of randomised clinical trials followed by clinicians. Moreover, as patient characteristics designed for other purposes11–13 and, although a 12-month did not substantially differed across hospitals, we should treatment seemed reasonable,14 no randomised studies expect a lower variability between hospitals. A large vari- had been performed within the study period aimed ation in individual country practices concerning the specifically at comparing 12-month DAPT with shorter pattern of DAPT duration after ACS has been described, periods in STEACS patients receiving PCI and thus this suggesting that local systems are strong drivers of DAPT recommendation might well be seen as somehow arbi- duration.21 These findings may imply that there is still trary by some prescribers. room for improvement in the quality of care of STEACS In 2015, the need for long-term DAPT was reinforced patients and that quality improvement programmes, by the recommendation of extended DAPT beyond whose efficacy and cost-effectiveness are still under eval- 12 months in patients with ACS receiving drug-eluting uation, could be useful to reduce variability in clinical stents,15–17 but still safety concerns might induce some practice.22 This is of prime importance in the context of prescribers to be reluctant to prolong DAPT, especially in the prescription of DAPT duration after ACS in which the patients with higher complexity.18 Safety concerns might clinician-driven variability in prescription patterns adds also explain the high proportion of discharge reports to the different levels patients’ adherence.21 with non-specified DAPT period, which deserved special Higher atherosclerotic burden and increased ischaemic attention in our analyses. Cardiologists might be reluctant risk was associated to better persistence with DAPT. The to prescribe a specific duration of DAPT maybe fearing need for 12-month DAPT schedules in patients treated about the emergency of events that increase the haemor- with drug-eluting stents is clearly perceived by physi- rhagic risk at some point after discharge, thus relying on cians but the magnitude of this association varies largely the follow-up that will be made at the ambulatory setting. between hospitals. This means that, even in clear indica- Our results showing a high degree of persistence for tions, there are different levels of adoption of emerging patients without a specification of DAPT time point out clinical recommendations in hospitals belonging to the to the fact that this decision is not necessarily ‘incorrect’, same AMI network. and that health providers coming later in the process of It is also apparent from our data that the speed of adop- care are probably doing their job. tion of clinical guidelines is different among hospitals and We might wonder whether the observed high variability that an acceptable and generalised level of adherence between hospitals in the instructions provided about is only reached after 2 years of implementation. Similar DAPT duration actually reflects suboptimal quality of trends have been found in other contexts and earlier care or confusion in the interpretation of international periods23–28 reporting DAPT use between 60% and 80% guidelines. In fact, although 2012 European Society of at discharge and between 25% and 75% at 1 year. In this Cardiology guidelines state that DAPT must be continued sense, together with other quality improvement initia- for 12 months after STEACS with a class of recommen- tives, the use of population-based registries to provide dation I, the level of evidence was established as C.1 audit and feedback could be useful to promote quicker Thus, there was general agreement that a minimum of and smoother adoption of clinical practice guidelines.29 12 months of DAPT is likely to be beneficial but based There are a number of assumptions that might be ques- only on a consensus of experts or observational studies. tionable: a number of factors have been described to Moreover, it is literally stated that the given recommenda- contribute to underprescription.30 The complete process tion on DAPT duration should be ‘with a strict minimum of care and the definite prescription at the ambulatory 8 Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 Open access 4 setting is poorly known for individual patients and has Cardiology Department, Hospital del Mar, and Heart Diseases Biomedical Research not been considered in this study. Changes in treatment Group and IMIM (Hospital del Mar Medical Research Institute), Faculty of Medicine, Universitat Autonoma de Barcelona, Barcelona, Spain prescription might be justified by the patients’ varying 5 Cardiology Department, Hospital Universitari de Tarragona Joan XXIII, IISPV, conditions during follow-up. We assume that hospital Universidad Rovira Virgili, Tarragona, Spain recommendation influences final prescription, and conse- 6 Cardiology Department, Hospital Germans Trias i Pujol, Badalona, Spain 7 quently, final adherence to guidelines, but it can also be Cardiology Department, Hospital Josep Trueta, Girona, Spain 8 that both ‘prescribers’ facing the same patient share the Cardiology Department, Hospital Mútua de Terrassa, Barcelona, Spain 9 Cardiology Department, Hospital Arnau de Vilanova, Lleida, Spain same criteria for prescription. That is, the hospital cardi- 10 Cardiology Department, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain ologist might have decided to recommend DAPT for a 11 Cardiology Department, Hospital de Bellvitge, Barcelona, Spain shorter period to an elderly patient with other comor- 12 Clinical Epidemiology Unit, Hospital Universitario de Cruces. Biocruces Bizkaia bidities and suboptimal quality of life due to mild diges- Health Research Institute, Bilbao, Spain 13 tive symptoms, even if her objective bleeding risk is not Institut Clínic Cardio vascular (ICCV), Hospital Clínic de Barcelona, Barcelona, Spain high; similarly, the primary physician or the cardiologist 15 Centro de Investigación Biomédica en Red en Epidemiología y Salud Pública at the primary care setting might have also decided to be CIBERESP, The Spanish Health Institute (ISCIII), Spain less aggressive for the same reasons, even without being 16 Centro de Investigación Biomédica en Red para Enfermedades Cardiovasculares influenced by the recommendation of the first prescriber. CIBERCV, Spanish Health Institute (ISCIII), Spain This would probably explain a large amount of the strong relationship between hospital recommendation and Collaborators The Codi IAM Registry Investigators (listed below) contribute to the functioning of the Codi IAM and to data acquisition for the Codi IAM Registry: C pharmacy dispensation. Moreover, although effects were Carvajal, P Martínez, RM Lidón, J Bañeras (Hospital Universitari de la Vall d’Hebrón, adjusted for patient characteristics and vascular events Barcelona), J Garcia-Picart (Hospital de la Santa Creu i Sant Pau, Barcelona), during follow-up, there might be other unmeasured M Massotti, G Jiménez, M Hernández, A Regueiro (Hospital Clinic, Barcelona), J reasons for deciding on a shorter DAPT period facing an Basaganyas, P Loma, M Cardenas, E Badosa, A Fageda (Hospital Josep Trueta, Girona) A Ariza, C García-Amigó, A Cequier, A Gómez-Hospital, G Marín, J Maristany, individual patient. V Montoya (Hospital Universitari de Bellvitge, Hospitalet), C Labata, J Mauri, E In addition, the recommendation at PCI hospital Fernández-Nofrerias, X Carrillo, C Garcia-Garcia, C Oliete (Hospital Universitari discharge may not coincide with the final hospital Germans Trias i Pujol, Badalona), A Bardaji, G Bonet-Pineda, S Rojas (Hospital Joan prescription in patients derived to another reference XXIII, Tarragona). Juan F Muñoz, F Padilla, S Homs (Hospital Mútua de Terrassa, Terrassa), B Baquerizo, L Recasens, H Tizon-Marcos (Hospital de Mar IMAS, hospital after PCI. However, results of sensitivity analyses Barcelona), A Bosch Gaya, F Worner, L Barta, M Agustí, A Gené (Hospital Arnau de excluding these patients did not differ substantially from Vilanova, LLeida), X Jimenez, A Mora (Sistema d’Emergencies Mèdiques SEM SA, the results of the main analyses. Barcelona), MT Faixedas, J Jiménez, A Rosas (Health Department, Generalitat de The study was aimed to ascertain influence of guidelines Catalunya). on hospital recommendation and its impact on patients’ Contributors IF-G, ARi and JRM conceived and designed the study. ARi, JRM and MTF participated in the acquisition and analysis of data. HT-M, SR, CL, MC, persistence with DAPT. The impact of persistence on clin- SH, CT-Q, JG-P, JAG-H and MM were responsible for data acquisition in their ically relevant results will be assessed in another article. respective hospitals. ARi, IF-G, JRM, GO, ARo, JIP, JM and DGD were involved in the Similarly, the study was not specifically aimed at a deep interpretation of results. ARi and IF-G wrote the manuscript and all other authors assessment of determinants of adherence. This requires a revised it critically and approved its final version. detailed examination of the social context and a detailed Funding The study was supported by Instituto de Salud Carlos III grand number assessment of individual psychological factors.31 PI13/00399 and the Fundació la Marató de TV3 grand number 430/U/2015. The funders did not have any role in the study design and development. Competing interests None declared. Patient consent for publication Not required. Conclusion Ethics approval The study obtained ethics approval from the Vall d’Hebron Clinical The study shows that 12-month DAPT persistence in Research Ethics Committee (EPA(AG)7/2014(3989)). revascularised patients with STEACS in Catalonia (Spain) Provenance and peer review Not commissioned; externally peer reviewed. has substantially increased between years 2010 and 2015 Data sharing statement The study is based on clinical and administrative data especially since 1 year after the publication of European that belong to the Government administration and is not subject to data sharing. guidelines in 2012. Guideline implementation was also Open access This is an open access article distributed in accordance with the followed by a substantial decrease in variability between Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which centres. We have shown that instructions given at the PCI permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is hospital discharge are strongly associated with persistence. properly cited, appropriate credit is given, any changes made indicated, and the use Thus, establishing common and rational prescribing is non-commercial. See: http://​creativecommons.​org/​licenses/​by-​nc/​4.​0/. criteria between hospitals in the STEACS network may favour patients persistence with scheduled prescriptions and also reduce variability in clinicians’ practices. Author affiliations References 1 1. Steg PG, James SK, Atar D, et al. ESC Guidelines for the Cardiology Department, Vall d'Hebron University Hospital and Research Institute, management of acute myocardial infarction in patients presenting Universitat Autònoma de Barcelona, Barcelona, Spain with ST-segment elevation: The Task Force on the management of 2 Servei Català de la Salut, Generalitat de Catalunya, Barcelona, Spain ST-segment elevation acute myocardial infarction of the European 3 Departament de Salut, Generalitat de Catalunya, Barcelona, Spain Society of Cardiology (ESC). Eur Heart J 2012;33:2569–619. Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114 9 Open access 2. O’Gara PT, Kushner FG, Ascheim DD, et al. 2013 ACCF/AHA 18. Tinetti ME, Bogardus ST, Agostini JV. Potential pitfalls of disease- Guideline for the Management of ST-Elevation Myocardial Infarction: specific guidelines for patients with multiple conditions. N Engl J A Report of the American College of Cardiology Foundation/ Med 2004;351:2870–4. American Heart Association Task Force on Practice Guidelines. 19. Kedhi E, Fabris E, van der Ent M, et al. Six months versus 12 months Circulation 2013;2013:e362–425. dual antiplatelet therapy after drug-eluting stent implantation in 3. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med ST-elevation myocardial infarction (DAPT-STEMI): randomised, 2005;353:487–97. multicentre, non-inferiority trial. BMJ 2018;363:k3793. 4. Carrillo X, Fernandez-Nofrerias E, Rodriguez-Leor O, et al. Early 20. Hahn JY, Song YB, Oh JH, et al. 6-month versus 12-month or ST elevation myocardial infarction in non-capable percutaneous longer dual antiplatelet therapy after percutaneous coronary coronary intervention centres: in situ fibrinolysis vs. percutaneous intervention in patients with acute coronary syndrome (SMART- coronary intervention transfer. Eur Heart J 2016;37:1034–40. DATE): a randomised, open-label, non-inferiority trial. Lancet 5. Regueiro A, Bosch J, Martín-Yuste V, et al. Cost-effectiveness of a 2018;391:1274–84. European ST-segment elevation myocardial infarction network: results 21. Bueno H, Pocock S, Danchin N, et al. International patterns of dual from the Catalan Codi Infart network. BMJ Open 2015;5:e009148. antiplatelet therapy duration after acute coronary syndromes. Heart 6. Swieczkowski D, Mogielnicki M, Cwalina N, et al. Medication 2017;103:132–8. adherence in patients after percutaneous coronary intervention due 22. Flather MD, Babalis D, Booth J, et al. Cluster-randomized trial to acute myocardial infarction: From research to clinical implications. Cardiol J 2016;23:483–90. to evaluate the effects of a quality improvement program on 7. Gislason GH, Rasmussen JN, Abildstrøm SZ, et al. Long-term management of non-ST-elevation acute coronary syndromes: the compliance with beta-blockers, angiotensin-converting enzyme European Quality Improvement Programme for Acute Coronary inhibitors, and statins after acute myocardial infarction. Eur Heart J Syndromes (EQUIP-ACS). Am Heart J 2011;162:700–7. 2006;27:1153–8. 23. Czarny MJ, Nathan AS, Yeh RW, et al. Adherence to dual antiplatelet 8. Kontopantelis E, Doran T, Springate DA, et al. Regression based therapy after coronary stenting: a systematic review. Clin Cardiol quasi-experimental approach when randomisation is not an option: 2014;37:505–13. interrupted time series analysis. BMJ 2015;350:h2750. 24. Green A, Pottegård A, Broe A, et al. Initiation and persistence 9. Bernal JL, Cummins S, Gasparrini A. Interrupted time series with dual antiplatelet therapy after acute myocardial infarction: regression for the evaluation of public health interventions: a tutorial. a Danish nationwide population-based cohort study. BMJ Open Int J Epidemiol 2017;46:348-355. 2016;6:e010880. 10. Yeh RW, Secemsky EA, Kereiakes DJ, et al. Development and 25. Anastasius M, Lau JK, Hyun K, et al. The underutilisation of dual validation of a prediction rule for benefit and harm of dual antiplatelet antiplatelet therapy in acute coronary syndrome. Int J Cardiol therapy beyond 1 year after percutaneous coronary intervention. 2017;240:30–6. JAMA 2016;315:1735. 26. Prami T, Khanfir H, Deleskog A, et al. Clinical factors associated 11. Yusuf S, Zhao F, Mehta SR, et al. Effects of clopidogrel in addition with initiation of and persistence with ADP receptor-inhibiting oral to aspirin in patients with acute coronary syndromes without ST- antiplatelet treatment after acute coronary syndrome: a nationwide segment elevation. N Engl J Med 2001;345:494–502. cohort study from Finland. BMJ Open 2016;6:e012604. 12. Wiviott SD, Braunwald E, McCabe CH, et al. Prasugrel versus 27. Varenhorst C, Jensevik K, Jernberg T, et al. Duration of dual clopidogrel in patients with acute coronary syndromes. N Engl J Med antiplatelet treatment 2007;357:2001–15. with clopidogrel and aspirin in patients with acute coronary 13. Wallentin L, Becker RC, Budaj A, et al. Ticagrelor versus clopidogrel syndrome. Eur Heart J in patients with acute coronary syndromes. N Engl J Med 2014;35:969–78. 2009;361:1045–57. 28. Chung SC, Sundström J, Gale CP, et al. Comparison of hospital 14. Eisenstein EL, Anstrom KJ, Kong DF, et al. Clopidogrel use and long- variation in acute myocardial infarction care and outcome between term clinical outcomes after drug-eluting stent implantation. JAMA Sweden and United Kingdom: population based cohort study using 2007;297:159. 15. Bonaca MP, Bhatt DL, Cohen M, et al. Long-term use of ticagrelor nationwide clinical registries. BMJ 2015;351:h3913. in patients with prior myocardial infarction. N Engl J Med 29. Eisenstein EL, Wojdyla D, Anstrom KJ, et al. Evaluating the impact 2015;372:1791–800. of public health notification: Duke clopidogrel experience. Circ 16. Mauri L, Kereiakes DJ, Yeh RW, et al. Twelve or 30 months of Cardiovasc Qual Outcomes 2012;5:767–74. dual antiplatelet therapy after drug-eluting stents. N Engl J Med 30. Cabana MD, Rand CS, Powe NR, et al. Why don’t physicians follow 2014;371:2155–66. clinical practice guidelines? A framework for improvement. JAMA 17. Levine GN, Bates ER, Bittl JA, et al. 2016 ACC/AHA Guideline 1999.65. Focused Update on Duration of Dual Antiplatelet Therapy in Patients 31. Ferreira-González I, Marsal JR, Ribera A, et al. Background, With Coronary Artery Disease: A Report of the American College incidence, and predictors of antiplatelet therapy discontinuation of Cardiology/American Heart Association Task Force on Clinical during the first year after drug-eluting stent implantation. Circulation Practice Guidelines. J Am Coll Cardiol 2016;68:1082–115. 2010;122:1017–25. 10 Ribera A, et al. BMJ Open 2019;9:e028114. doi:10.1136/bmjopen-2018-028114