Incidence of bleeding and performance of the PRECISE-DAPT score in predicting bleeding in patients on dual antiplatelet therapy after treatment for acute coronary syndrome in Kenya

Abstract

Introduction

Dual Antiplatelet Therapy (DAPT) plays an important role in the secondary prevention of ischemic events after treatment for acute coronary syndrome (ACS). The long-term use of DAPT is associated with an increased risk of bleeding, which affects morbidity and mortality. Risk stratification scores have been developed to predict this risk and provide a balance against the risk of ischemic events. The aim of this study was to determine the incidence of bleeding in a cohort of patients in Kenya on DAPT and assess the performance of the PRECISE-DAPT Score in predicting the risk of bleeding.

Methods

This was a retrospective study conducted in three hospitals in Kenya among patients on DAPT after ACS between January 2019 and April 2022. We reviewed medical records for demographic and clinical characteristics and conducted telephone interviews to assess bleeding for patients on DAPT for a minimum period of one year. Bleeding events were categorized according to the TIMI criteria for bleeding, and the PRECISE-DAPT Score was calculated using an online calculator. The cumulative one-year incidence of bleeding was calculated and presented as frequencies and percentages. Receiver operating characteristic (ROC) analysis and C-statistics were used to quantify the ability of the PRECISE-DAPT Score to predict bleeding events, whereas calibration was estimated using the Hosmer‒Lemeshow goodness-of-fit test.

Results

A total of 202 patients were enrolled in the study. The study population was predominantly male (n = 156, 77.2%) and African (n = 141, 69.8%), with a median age of 61 years (IQR 52–72). Majority were admitted with ST-Elevation Myocardial infarction (STEMI) (n = 126, 62.4%) and had a mildly reduced left ventricle ejection fraction (n = 124, 61.4%). Fourteen patients (6.9%) met the TIMI criteria for bleeding, of whom 11 (5.4%) had minimal bleeding and 3 (1.5%) had minor bleeding. There was no incidence of major bleeding. The discrimination and calibration of the PRECISE-DAPT Score was good {ROC curve 0.699 (95% CI: 0.564–0.835)} and the Hosmer–Lemeshow goodness-of-fit test (Chi-square, 6.53; p = 0.588), respectively.

Conclusion

The incidence of bleeding was low, with the majority of patients having minimal bleeding that did not require medical intervention. The PRECISE-DAPT Score performed well in predicting bleeding in patients on DAPT.

Introduction

Dual Antiplatelet Therapy (DAPT) is important for preventing ischemic events after treatment for acute coronary syndrome (ACS) especially percutaneous coronary intervention (PCI). The recommended DAPT after PCI is aspirin and P2Y12 inhibitor for a period of 12 months, followed by long-term aspirin monotherapy [1]. However, clinical decision-making regarding the optimal duration of DAPT after PCI should consider the bleeding risk profile and initial clinical presentation, among other factors [2]. The incidence of TIMI major bleeding ranges from 1 to 4.5% for patients on DAPT for a duration of 12 months [3,4,5].

To tailor the duration of DAPT and reduce the incidence of bleeding, various risk stratification tools are used. The PRECISE-DAPT score is a simple tool that uses five parameters that are easy to derive, and its scores are easy to interpret [4]. It has been validated by various clinical trials, including PLATO [5] and the BernPCI registry. Its performance is equal to that of the other scores. An analysis of the GLOBAL-LEADERS Trial, revealed that there was no difference in the 30-day performance of PRECISE-DAPT compared with the CRUSADE and ACUITY scores [6]. Compared with the BleeMACS, PARIS and REACH scores, the PRECISE-DAPT Score performed moderately well in predicting one-year bleeding risk [7].

The incidence of ACS and the number of patients undergoing PCI or fibrinolytic therapy are increasing in Sub-Saharan Africa [8]. However, despite these recent advances in the treatment of coronary artery disease, bleeding in patients on DAPT and the performance of the PRECISE-DAPT Score in predicting bleeding have not been studied in a cohort of patients in East Africa.

In light of this background, we aimed to establish the incidence of bleeding in a cohort of patients treated for ACS in three tertiary hospitals in Kenya and the performance of the PRECISE-DAPT Score in predicting bleeding. Insights from this study will provide much needed information that can help in the clinical decision-making process when starting and following up of patients on DAPT.

Methods

Study settings and design

This was a multi-centre retrospective study conducted in three tertiary hospitals in Kenya. These hospitals are part of an ongoing registry in the country encompassing four main cardiovascular diseases, namely; acute coronary syndrome, atrial fibrillation, heart failure and pulmonary thromboembolism.

The Aga Khan University Hospital, Nairobi (AKUH, N), is the largest private hospital in Kenya with a bed capacity of 289 and the only hospital in Kenya and the region accredited by the Joint Commission International, Clinical Care Certification Program (CCPC) in the management of ACS. This external validation ensures that the hospital adopts international standards that provide standardization of care for patients with ACS. This is a PCI-capable hospital and manages the largest number of acute myocardial infarct patients in Nairobi. It has a 24-hour cardiac catheterization laboratory and a six-bed coronary care unit with 24-hour coverage by an on-call cardiology fellow and an interventional cardiologist. The inpatient cardiology section is registered with an American College of Cardiology (ACC) suite of cardiovascular data registries, the NCDR-Cath PCI. This registry captures data on demographics, patients’ diagnoses, laboratory parameters, patient-specific treatments, revascularization strategies, and in-hospital outcomes [9]. The data is entered by a designated research nurse and verified by an interventional cardiologist for accuracy and completeness. This tool captures the department’s adherence to ACC/AHA clinical practice guidelines and procedure performance standards in coronary revascularization.

Kenyatta National Hospital (KNH), Nairobi, is the largest public teaching and referral hospital in Kenya, with a bed capacity of 1800 (including 82 beds in the intensive care unit). It has 50 wards and 24 outpatient clinics and attends to about 900,000 inpatients and 800,000 outpatients annually. It has approximately 150 general cardiology beds with an on-call internal medicine resident and an interventional cardiologist, among other clinical staff. It is PCI-capable and manages the highest number of ACS patients in the public sector. However, owing to the delay in presentation and because the cardiac catheterization laboratory does not operate on a 24-hour basis, only a small fraction of patients presenting with ST-Elevation Myocardial Infarction (STEMI) are eligible for primary PCI. Most patients are treated medically, with more than 85% receiving dual antiplatelet therapy [10]. The hospital has both a paper-based and electronic health management information system that captures data on demographics, diagnosis, laboratory parameters, treatment, and in-hospital outcomes, among others.

Moi Teaching Referral Hospital (MTRH), Eldoret, is a multispecialty, 1000-bed, non-PCI capable public teaching and referral hospital located approximately 300 km North West of Nairobi. It is the main referral centre for the western part of Kenya. It has two 50-bed cardiology wards and a 10-bed coronary care unit where ACS patients are managed. These are managed by specialist internal medicine residents, general cardiology fellows and consultant cardiologists. ACS patients are managed medically (including thrombolysis and DAPT) and referred for coronary angiography at PCI-capable units if indicated.

Study population and sample size determination

The study involved extracting data from the hospitals’ paper-based, electronic health records and the NCDR-Cath PCI registry. All patients above the age of 18 years who had completed a minimum of 12 months of DAPT after treatment for ACS between January 2019 and April 2022 were included. Patients who had been on DAPT for other indications, such as infarctive stroke or post coronary artery bypass graft (CABG), those on oral anticoagulants, those with incomplete records, those who could not be reached by phone and those who did not give consent, were excluded from the study.

The index date of enrolment in the study was 7 days after PCI and initiation of DAPT. The 7-day period was to eliminate the peri-procedural bleeding, which may not be related to DAPT. Data on patient follow-up were obtained from the date of enrolment to the date of first bleed, completion of DAPT or the end of the one-year study period. Data on bleeding outcome was collected through telephone interviews via a standardized questionnaire after verbal informed consent was obtained. Bleeding events were classified according to the TIMI criteria for bleeding into major, minor or minimal bleeding episodes [11]. Details of any bleeding event that required a hospital visit were confirmed by checking the records of outpatient and inpatient clinic visits.

The PRECISE-DAPT Score was calculated with a web calculator and used to assess the risk of bleeding at the initiation of DAPT. The patients were divided into two groups, as reported in the derivation and validation cohorts [12], as follows: low and intermediate (PRECISE-DAPT score < 25) and high (PRECISE-DAPT score > 25).

The one-year major and minor bleeding risk according to major outcome trials varies from 1 to 4.5%. The formula below [13] was used to estimate the sample size needed. P is the estimated incidence of bleeding, d is the precision rate, and Z is the statistic corresponding to the confidence level.

$$N\;=\frac{z^2\;P\;(1\;-\;P)}{d^2}$$

Working with a 95% confidence level, an estimated incidence of bleeding of 2%, and a confidence interval (margin of error) of ± 2%, the sample size was estimated at 188.

Data management

Continuous variables were presented as means with standard deviations for normally distributed variables or medians with interquartile ranges for non-normally distributed variables and were compared with the Welch two-sample t test or the Wilcoxon rank-sum test. Categorical variables were presented as frequencies and percentages and were compared with Pearson’s chi-square test or Fisher’s exact test.

The cumulative one-year incidence of bleeding was calculated and presented as frequencies and percentages. Logistic regression models were used to assess the independent risk factors for bleeding. Variables were entered into the multivariable model if P < 0.10 after univariable analysis. Determinants of bleeding were identified via a binary logistic regression model; odds ratios (ORs) with corresponding 95% confidence intervals and P values were reported.

Accurate discrimination and calibration are key distinguishing features in predictive models. Discrimination was assessed by the area under the receiver operating characteristic (ROC) curve. An area of 0.7–0.8 is considered fair, 0.8–0.9 is good, and > 0.9 is excellent. However, in general, a model with a c statistic of > 0.70 is considered to have acceptable discriminatory capacity.

A Hosmer–Lemeshow goodness-of-fit test, which follows a chi-square distribution, was used to evaluate the model fit as well as calibration, with a p value of > 0.05 signifying no evidence of poor fit. Statistical analysis was performed using R version 4.1.3 (2022 03–10).

Results

A review of the medical records and ACS registries in the three hospitals revealed that a total of 445 patients had been treated for ACS between 2019 and 2022. The majority (n = 399, 89.6%) had been treated at the AKUH-N. Out of these, 279 patients were initiated on DAPT. Forty-two (42) patients could not be reached for assessment of bleeding since they did not have a functional telephone numbers, their phones were off, or they did not answer the calls. Thirty-five (35) patients were excluded (21 were on anticoagulants, 13 died before the end of one year, and one declined to give consent). For those who died before the end of one year, the cause of death was not related to bleeding. A total of 202 patients were recruited into the study, as shown in Fig. 1 below. Majority of the patients (n = 185, 91.6%) were from AKUH-N, (n = 12, 5.9%) were from KNH, and (n = 5, 2.5%) were from MTRH.

The baseline and clinical characteristics of the patients included and those excluded from the study were the same.

Fig. 1

Flow diagram illustrating the selection of the study population

Socio-demographic and clinical characteristics of the patients

The median age of our cohort was 61 years (IQR: 52.0 − 70.0). Our cohort was predominantly male (n = 156, 77.2%) and African (n = 141, 69.8%), as shown in Table 1.

Table 1 Socio-demographic characteristics of the cohort

The most common comorbid condition was hypertension (n = 94, 46.5%). Approximately a quarter of the study population were smokers (n = 54, 26.7%) and a third were diabetic (n = 71, 35%). (n = 51, 25.2%) of the patients were on treatment for dyslipidaemia and (n = 29, 14.4%) had a family history of coronary artery disease as shown in Table 2 below.

Table 2 Underlying risk and comorbid conditions

The most common ACS presentation was STEMI (n = 126, 62.4%), and the majority of the patients had mildly reduced left ventricle ejection fraction (n = 124, 61.4%), as illustrated in Table 3.

Table 3 Clinical presentation and diagnosis

Prior to ACS diagnosis, (n = 28, 13.9%) of the patients were on aspirin, (n = 15, 7.4%) were on clopidogrel, and only one patient was on ticagrelor, as illustrated in Table 4.

Table 4 Previous antiplatelet use

The mean white cell count was 9.2 per microliter (SD: 3.5), the mean haemoglobin level was 14.5 g/dl (SD: 2.0), and the mean platelet count was 267.3 per microliter (SD: 76.0). The majority of the patients (n = 183, 90.6%) had a normal haemoglobin level of more than 12 g/dl. The mean creatinine level was 104.7 µmol/l (SD: 61.1), and the mean eGFR was 75.9 ml/min (SD: 22.1). Fifty-seven (28.2%) patients had stage 1 chronic kidney disease (CKD), 101 (50%) had stage 2 CKD, and 37 (18.4%) had stage 3 CKD, as shown in Table 5 below.

Table 5 Laboratory parameters of the cohort

During admission, (n = 62, 30.7%) of the patients received a glycoprotein IIa/IIIb antagonist, Eptifibatide; {(n = 47, 77%) and (n = 15, 23%)} during cardiac catheterization and upstream respectively. Eighteen patients (8.9%) received a thrombolytic agent. All the patients were discharged on aspirin. Regarding the second antiplatelet, (n = 119, 58.9%) were discharged on Clopidogrel, (n = 67, 33.2%) on Ticagrelor and (n = 16, 7.9%) on Prasugrel. Other discharge medications included; beta blockers. (n = 156, 77.2%), angiotensin converting enzyme inhibitors (n = 93,46%), statins (n = 178, 89.4%) and proton pump inhibitors (n = 132, 71.7%), as illustrated in Table 6.

Table 6 Pharmacotherapies during admission and discharge

Incidence and type of bleeding

The overall incidence of bleeding was (n = 14, 6.9%), that of major bleeding was 0%, that of minor bleeding was (n = 3, 1.5%), and that of minimal bleeding was (n = 11, 5.4%). Among the 11 episodes of minimal bleeding, (n = 7, 63.6%) were epistaxis, (n = 2, 18.2%) bled under the skin, one had lower gastrointestinal tract (GIT) bleeding, and one had upper GIT bleeding, as illustrated in Table 7. All the three patients who had minor bleeding had upper GIT bleeding with haemoglobin drop of < 5 g/dl and required transfusion with packed red blood cells. Most of the episodes of bleeding (n = 9, 64.2%) occurred within the first three months of initiation of DAPT.

Table 7 Incidence and type of bleeding

Risk factors associated with bleeding

Table 8 shows the parameter estimates for some of the possible predictors of bleeding. In the univariable analysis, only the estimated glomerular filtration rate (eGFR) had a significant effect on bleeding. {Mean eGFR 76.8 (SD 22.0) vs. 63.1 (SD 20.3), p = 0.029}.

In the multivariable analysis, controlling for all the variables in the logistic regression model, none of the variables had a significant effect on bleeding. However, despite the lack of a significant association, the use of Ticagrelor/Prasugrel and the use of a GP IIb/IIIa inhibitor had greater effect on bleeding (46% and 72%, respectively) than did the use of Clopidogrel and the absence of a GP IIb/IIIa inhibitor.

Table 8 Logistic regression analysis of the determinants of bleeding

Association of PRECISE-DAPT score and bleeding

Using a PRECISE DAPT Score cut-off of 25, with < 25 being a low/intermediate score and > 25 being a high score, there was a significant association between the PRECISE DAPT Score and bleeding {OR 3.05 (0.88–9.49), p = 0.06, as shown in Table 9}. The incidence of bleeding was 2.4% in patients with a PRECISE-DAPT score < 25 and 4.4% in those with a PRECISE-DAPT Score > 25.

Table 9 PRECISE DAPT and bleeding

Discrimination and calibration of PRECISE-DAPT score

The area under the ROC curve (AUC) was calculated to evaluate the predictive value of the PRECISE-DAPT score for bleeding. It was a statistically predictive marker of bleeding (AUC: 0.699, 95% CI: 0.564–0.835). The sensitivity was 0.846 (95% CI: 0.786–0.894), and the specificity was 0.357 (95% CI: 0.128–0.649). The model demonstrated a 70% probability of the PRECISE DAPT Score to correctly predict bleeding, which was interpreted as good, as shown in Fig. 2 below.

Fig. 2

Receiver operating characteristic curve for the prediction of bleeding using the PRECISE DAPT score

Sensitivity: 0.846 (95% CI: 0.786--0.894) Specificity: 0.357 (95% CI: 0.128--0.649)

Positive Predictive Value (PPV): 0.946; Negative Predictive Value (NPV): 0.147

Calibration of the PRECISE-DAPT score revealed good performance. The model fit showed a non-significant Hosmer–Lemeshow p value of 0.588 (chi-square value of 6.53).

Discussion

To our knowledge, this is the first study to report the incidence of bleeding and performance of a predictive model in patients on DAPT in both private and public settings in East Africa.

According to our results, the incidence of overall bleeding was low (6.93%), and most bleeding events were minimal (5.4%) and did not require any interventions to stop. The incidence of major TIMI bleeding was 0%.

This is in contrast to other studies performed elsewhere. In the TRITON-TIMI trial, the rates of major non-CABG-related bleeding were 2.4% and 1.8% in patients receiving prasugrel and clopidogrel, respectively. This included life-threatening bleeding of 1.4% vs. 0.9% for the two groups, respectively [4]. In the landmark PLATO trial [5], the rates of major bleeding were 4.5% and 3.8% in patients receiving ticagrelor and clopidogrel, respectively.

Some of the possible explanations as to why our study revealed low incidences of major bleeding compared with other studies include the low number of patients on ticagrelor and prasugrel, both of which were associated with slightly higher incidences than clopidogrel in the TRITON-TIMI and PLATO trials. However, this may not explain the difference entirely because even in the clopidogrel versus aspirin trials, such as the CHARISMA trial [14], the rates of major bleeding were still higher than those in our study (1.7% vs. 1.3%, p = 0.09) in the clopidogrel vs. aspirin arms, respectively.

Another possible explanation is the greater use of glycoprotein IIb/IIIa receptor antagonists (54%) in the TRITON-TIMI trial [4] than in our trial (30.2%).

The populations studied also differed. The TRITON-TIMI trial was predominantly a Caucasian population (92%), whereas our study was predominantly a black African cohort (69.8%). However, in a study by Brittain et al. [15] comparing the risk of bleeding while on DAPT among US Black and White adults, there was no difference in post-discharge BARC 2–5 bleeding between the two groups of patients. Therefore, the racial difference in the studied population may not entirely explain the low incidence of bleeding in our study.

Although there was no difference between the incidence of bleeding and the type of DAPT used, possibly due to the small sample size, the use of ticagrelor/prasugrel or a GP IIb/IIIa inhibitor had an increasing effect on bleeding of 46% and 72%, respectively, compared with the use of clopidogrel and not using any GP IIb/IIIa inhibitor.

Although not widely used in clinical practice, international guidelines encourage the weighting of bleeding risk before the selection of the duration of DAPT. They suggest that patients with a high risk of bleeding should have a shorter period of DAPT [16]. Different tools have been validated for the assessment of bleeding risk at the time of initiation of DAPT to tailor the duration on the basis of bleeding risk. Although the PRECISE-DAPT Score has been validated by the PLATO trial validation cohort and the BernPCI registry validation cohort, it has not been studied in a low-resource setting.

In our study, the ability of the PRECISE-DAPT score to predict the one-year risk of bleeding was good, with an area under the curve of 0.699 (95% CI: 0.564–0.835) and a Hosmer–Lemeshow goodness-of-fit test, chi-square value of 6.53 and a p value of 0.588.

These findings are similar to those of other studies performed elsewhere. Fracesco Costa et al. [12] on derivation and validation of PRECISE-DAPT Score through a pooled analysis of patient’s datasets from clinical trials showed a c-index for out-of-hospital TIMI major and minor bleeding of 0.70 (95% CI 0.65–0.74) in the trial validation cohort and 0.66 (95% CI 0.61–0.71) in the Bern PCI registry validation cohort, similar to our study. Liang Dong, in a study evaluating the performance of the PRECISE-DAPT Score in predicting bleeding in Chinese elderly patients [17], reported that the C statistic of the PRECISE-DAPT model for the prediction of BARC > 2 bleeding in overall patients was 0.717 (95% CI, 0.656–0.777). The c-indexes for the PRECISE-DAPT Score in the GLOBAL LEADERS and GLASSY trials [18] were 0.67 (95% confidence interval [CI]: 0.63–0.71) vs. 0.63 (95% CI: 0.59–0.67) (p = 0.27) and 0.67 (95% CI: 0.61–0.73) vs. 0.66 (95% CI: 0.61–0.72), respectively. The SMART DATE Trial [19] analysed the clinical usefulness of the PRECISE-DAPT Score for predicting bleeding events in patients with ACS undergoing PCI. In patients with a non-high PRECISE-DAPT Score < 25, a 6-month DAPT was associated with a higher ischemic risk with similar bleeding risk, whereas in patients with a high PRECISE-DAPT Score > 25, a 6-month DAPT presented a similar ischemic risk with significantly increased major bleeding risk. Although we did not analyse ischemic risk, the utility of the PRECISE-DAPT Score in predicting bleeding risk in this study is similar to that in our study. Thus, the determination of the PRECISE-DAPT Score could improve clinical outcomes in patients with ACS undergoing PCI.

The discriminative ability of the PRECISE DAPT Score in our study was good (goodness-of-fit chi-square value of 6.53, p value of 0.588), similar to the findings of other studies, such as the study by Liang Dong et al. [16], which reported a chi-square value of 0.432 and a p value of 0.806 for the calibration of BARC > 2 bleeding in a 65-year-old or older Asian cohort of patients.

Limitations

The major limitation in our study is that out of the patients discharged on DAPT, 15% could not be contacted to establish where they had a bleeding episode and could therefore not be accounted for. It is worth noting that this high dropout rate could potentially impact the conclusions due to potentially missed bleeding events.

Although we assessed many possible risk factors associated with bleeding, other possible predisposing factors, such as malignancy and revascularization procedures, were not assessed. These are confounders to our study which can also potentially impact our conclusions.

Furthermore, we did not evaluate ischemic risk, which has risk factors similar to those associated with bleeding.

Conclusion

In conclusion, our study revealed that the incidence of major bleeding in patients on DAPT for one year was low, in fact 0%, in our cohort of patients. The performance of the PRECISE-DAPT Score in predicting the risk of bleeding was good and comparable to that of most studies performed in different parts of the world. In this context, we recommend that institutions should consider using this score when initiating DAPT to guide treatment decisions such as the duration of DAPT, among others. However, this should be done with caution due to the low specificity of this tool and high rates of loss to follow up. We also recommend a randomized trial with a larger sample size to support or refute our findings.