The anti-thrombotic and glucose lowering therapy in diabetic patients with coronary artery disease undergoing PCI: a prospective multicenter observational study on drug use and implications for clinical outcomes (ARTHEMIS) study, was an investigator-initiated nationwide prospective observational study, implemented at 12 medium-to-high volume PCI-capable hospitals in Portugal. A detailed description of the clinical endpoints and their definitions is provided in the Supplemental Materials.

Baseline data was collected at the time of inclusion in a dedicated e-CRF. The study was planned and reported according to the STROBE statement for reporting observational studies.12 An external contract research organization ensured the quality and reliability of the data, through regular monitoring.

All consecutive patients with type-2 diabetes undergoing PCI with stent implantation performed in at least one major coronary artery in the context of stable CAD or an acute coronary syndrome were screened and invited to participate. Inclusion could be considered at any time between PCI and hospital discharge. Other than balloon-only PCI, short life expectancy due to significant comorbidity (according to the opinion of the attending physician), type 1 diabetes and unwillingness to participate, no major exclusions existed. Inclusion and exclusion criteria are detailed in the online Supplemental Material.

For screening and inclusion purposes, the diagnosis of type-2 diabetes was adjudicated as self-reported when already known at the time of hospital admission, and according to the 2021 criteria of the American Diabetes Association when newly diagnosed or confirmation was warranted.13 Ischemic and bleeding risks were estimated by the DAPT7 and PRECISE-DAPT Risk Scores.14

After screening, patients were treated according to local practice; PCI technique was left to the discretion of the operator, as well as subsequent medical management. The indication for PCI was reported by the local investigators and was not centrally adjudicated. As per-protocol, DAPT regimen had to be specifically indicated, including the drugs prescribed and specific planned duration. If available (according to local clinical practice in each hospital), HbA1c was registered at the index admission and at each visit.

Taking the sample size of previously published studies that evaluated different strategies of DAPT duration as reference, we arbitrarily estimated that the inclusion of 1000 consecutive patients would allow for a detailed characterization of the diabetic population with CAD and the evaluation of ischemic, heart failure and hemorrhagic events throughout the two-year follow-up.

Results are reported as percentages (for dichotomous variables) and as means and standard-deviation or median and interquartile range for continuous variables, as appropriate. Normality was assessed by the Kolmogorov-Smirnoff test and by visual inspection of histograms and Q-Q plots. When necessary, comparisons were performed using the χ2 test (with Yates correction when appropriate) for categorical variables, the Student t test, the Satterthwaite test or 1-way ANOVA for continuous variables with normal distribution or the Kruskal–Wallis test for continuous variables with a non-normal distribution. When appropriate, 95% confidence intervals were calculated. Two-tailed tests of significance are reported. For all comparisons, a p-value of <0.05 was considered statistically significant. Statistical analysis was performed with IBM SPSS software version 23 (SPSS Inc, Chicago, IL).

The study was undertaken according to good clinical practices and all patients provided written informed consent. The protocol complies with the Declaration of Helsinki and was approved by institutional review boards at each participating site. The study was registered with the Portuguese National Data Protection Committee and on ClinicalTrials.gov (identifier: NCT04481997).

The study was sponsored by the Association for Research and Development of the Faculty of Medicine (AIDFM) – Cardiovascular Center of the University of Lisboa (CCUL) and had the scientific sponsorship of the Portuguese Association of Cardiovascular Intervention (APIC). It was funded as an investigator-initiated study, by an unrestricted grant from AstraZeneca (ESR-19-20188). Study coordinators and local investigation teams were fully responsible for protocol preparation, patient selection, study compliance, and data entry and analysis.

Between January and November 2022, 6557 PCI procedures were performed in the 12 participating centers, of which 2056 (31.4%) were reportedly in patients with diabetes. From the 1457 who were eligible (meeting all inclusion/exclusion criteria), 1000 were included in the study at the time of the procedure (68.6%). Of these, four were subsequently excluded because a diagnosis of diabetes was not confirmed. Of the remaining 996 patients, five died during the index hospital admission and were excluded from the analysis (Figure 1). Overall average inclusion rate was 111 patients/month. A description of inclusion rates per each participating hospital is detailed in Supplemental Material (Supplemental Figure 2 and Supplemental Table 1).

Figure 1.

Flowchart of patient inclusion (for details, see Supplemental Figure 1).

Registry data provides an opportunistic snapshot on how results of randomized trials are translated to everyday practice through guideline uptake. Our registry has the virtue of including centers across the entire country, encompassing a wide range of socio-economic and logistic scenarios thus enabling a solid representation of the real world. Importantly, the information concerning almost 70% of all eligible patients has been captured, although the main reason for non-inclusion was unclear (operator decision or not reported). Interestingly though, baseline characteristics of our patients are remarkably similar to those of patients with diabetes included in the THEMIS (ticagrelor in patients with stable coronary disease and diabetes) randomized trial, except for a lower incidence of prior revascularization and lower ejection fraction.15

It is noteworthy that despite the high-risk profile of our population, over 1/3 of patients were not on any antithrombotic treatment by the time of hospital admission. These patients were less likely to have classical risk factors (such as hypertension and high cholesterol, but not smoking history), clinically overt vascular disease, prior admissions due to heart failure and atrial fibrillation, but there were no differences concerning gender and age.

At hospital discharge, the overall adherence to current recommendations in both chronic coronary syndrome and ACS patients seemed to be quite straightforward, although a significant proportion of patients with ACS were not discharged on a new P2Y12 inhibitor. These patients were older and had a higher bleeding risk, although recent guidelines do not in fact encourage this approach.16 On the other hand, PCI complexity was the main modulator of total prescribed DAPT duration. Trials assessing prolonged DAPT regimens have not been stratified by CAD complexity at baseline, just after PCI, and the DAPT score was developed in patients that have completed a standard duration regimen without events.7 Still, variables related to procedural risk and complexity have some weight in the score and a post-hoc patient-level metanalysis of randomized trials showed a significant interaction between the benefits of prolonged DAPT duration and PCI complexity.17

Trials performed in patients at high risk of bleeding undergoing PCI have suggested that short DAPT regimens are safe, whether or not on a background of direct oral anticoagulants. Despite individual trials have been formally underpowered to specifically assess between-group differences regarding thrombotic events (such as stent thrombosis), safety has been confirmed in large series and metanalysis.18,19 Operators have seemingly welcomed these findings as demonstrated by the high adoption of short protocols (<6 months) particularly in patients with a concomitant indication for anticoagulation, but also – although to a lesser extent – in those with high PRECISE-DAPT score and no-anticoagulation. Ongoing follow-up will further inform on clinical outcomes of these strategies on a real word setting.

Guidelines recommend a HbA1c<7% as a metabolic control target in patients with diabetes and CAD.20 Although glycemic levels are associated with cardiovascular outcomes, the most widely used drugs for diabetes for decades – namely metformin, sulfonylureas and dipeptidyl-peptidase-4 inhibitors – have either not been specifically tested or have failed to consistently reduce ischemic events and mortality, despite improving glucose levels.11,21 Recently, SGLT2 inhibitors and GLP-1 analogues have gained preponderance as guideline-recommended first line therapy, driven by significantly improvements in cardiovascular morbidity and mortality, even in patients without overt diabetes.20,22 Although in our cohort, mean HbA1c was within range of prior randomized trials performed in patients with DM and known CAD,15,23 optimal metabolic control rates (Hb A1c<7%) were worse than those reported both in the first diabetes prevalence study in Portugal (PREVADIAB study)24 and in the First National Health Examination Survey – INSEF 201525 (41% vs. 69% and 63%, respectively). This discrepancy most likely stems from differences in the targeted patients, given that the aforementioned surveys were aimed at studying broader populations with known and unknow diabetes, rather than selected individuals with clinically overt atherosclerotic disease, inherently more likely to have advanced diabetes and poorer metabolic control. Still, less than one third of patients were taking SGLT2 inhibitors by the time of hospital admission, and the proportion of those medicated with GLP-1 analogues was marginal. Despite the improvement at hospital discharge, prescription rates, especially for SGLT2 inhibitors, were still suboptimal. Altogether, this may be a reflection of some unwarranted perception of low tolerability to the drug in the real-life setting, impaired awareness of potential benefits by clinicians, worries about a low absolute benefit at the individual patient level, and, not less importantly, cost constraints.

Our work has the inherent limitations of observational studies. It is not possible to reliably measure the potential impact of the Hawthorne effect and its influence on prescription patterns. Considering that the calculation of risk scores (both hemorrhagic and ischemic) may not always be routine practice, it is possible that their knowledge by operators and/or attending physicians may have induced some behavioral change, by increasing the chance of changing standard DAPT duration (as compared to empirical clinical judgment alone).

Despite recommendations being made to include all consecutive cases meeting study criteria, over one third of eligible patients were not enrolled, mostly for unknown reasons. As data on this group was not systematically collected, it is not possible to firmly assess any potential biases resulting from inadvertent selection. Still, as previously mentioned, the baseline characteristics of our population are similar to other studies focusing on the same patient subset, which is reassuring of the external validity of our study.

Finally, for most variables including the diagnosis and duration of diabetes, investigators relied on self-reported data. Nevertheless, although confirmation of diabetes by laboratory testing was not protocol-mandated, only a minority of patients were excluded because of unconfirmed diabetes.