
==== Front
Eur Heart J
Eur Heart J
eurheartj
European Heart Journal
0195-668X
1522-9645
Oxford University Press UK

39016180
10.1093/eurheartj/ehae463
ehae463
Clinical Research
AcademicSubjects/MED00200
Eurheartj/15
Eurheartj/17
Eurheartj/16
Bleeding risk prediction after acute myocardial infarction-integrating cancer data: the updated PRECISE-DAPT cancer score
https://orcid.org/0000-0003-4625-2775
Dafaalla Mohamed Keele Cardiovascular Research Group, Centre for Prognosis Research, Keele University, Keele Rd, Stoke-on-Trent ST5 5BG, UK

https://orcid.org/0000-0002-3097-2834
Costa Francesco Department of Biomedical and Dental Sciences and Morphological and Functional Imaging, University of Messina, Messina 98100, Italy

https://orcid.org/0000-0001-6450-5815
Kontopantelis Evangelos National Institute for Health Research School for Primary Care Research, Division of Population Health, Health Services Research and Primary Care, School of Health Sciences, Faculty of Biology, Medicine and Health, Manchester Academic Health Science Centre, University of Manchester, Manchester, UK

Araya Mario Clinica Alemana, Hospital Militar de Santiago, Santiago, Chile

Kinnaird Tim Cardiology Department, University Hospital of Wales, Cardiff, UK

Micari Antonio Department of Biomedical and Dental Sciences and Morphological and Functional Imaging, University of Messina, A.O.U. Policlinic ‘G. Martino’, Messina 98100, Italy

Jia Haibo Department of Cardiology, The 2nd Affiliated Hospital of Harbin Medical University, Harbin, China
The Key Laboratory of Myocardial Ischemia, Chinese Ministry of Education, Harbin, China

Mintz Gary S Transcatheter Cardiovascular Therapeutics (TCT), Cardiovascular Research Foundation, New York, NY, USA

https://orcid.org/0000-0001-9241-8890
Mamas Mamas A Keele Cardiovascular Research Group, Centre for Prognosis Research, Keele University, Keele Rd, Stoke-on-Trent ST5 5BG, UK
National Institute for Health and Care Research (NIHR) Birmingham Biomedical Research Centre, UK

Corresponding author. Tel: 01782 679555, Email: mamasmamas1@yahoo.co.uk
Present address: Área del Corazón, Hospital Universitario Virgen de la Victoria, CIBERCV, IBIMA Plataforma BIONAND, Departamento de Medicina UMA, Spain, Centro de Investigación Biomédica en Red de Enfermedades Cardiovasculares (CIBERCV).
07 9 2024
17 7 2024
17 7 2024
45 34 Focus Issue on Cardio-Oncology, Genetics, Ischaemic Heart Disease 31383148
15 1 2024
29 3 2024
03 7 2024
© The Author(s) 2024. Published by Oxford University Press on behalf of the European Society of Cardiology.
2024
https://creativecommons.org/licenses/by-nc/4.0/ This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (https://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact reprints@oup.com for reprints and translation rights for reprints. All other permissions can be obtained through our RightsLink service via the Permissions link on the article page on our site—for further information please contact journals.permissions@oup.com.

Abstract

Background and Aims

This study assessed the impact of incorporating cancer as a predictor on performance of the PRECISE-DAPT score.

Methods

A nationally linked cohort of ST-elevation myocardial infarction patients between 1 January 2005 and 31 March 2019 was derived from the UK Myocardial Ischaemia National Audit Project and the UK Hospital Episode Statistics Admitted Patient Care registries. The primary outcome was major bleeding at 1 year. A new modified score was generated by adding cancer as a binary variable to the PRECISE-DAPT score using a Cox regression model and compared its performance to the original PRECISE-DAPT score.

Results

A total of 216 709 ST-elevation myocardial infarction patients were included, of which 4569 had cancer. The original score showed moderate accuracy (C-statistic .60), and the modified score showed modestly higher discrimination (C-statistics .64; hazard ratio 1.03, 95% confidence interval 1.03–1.04) even in patients without cancer (C-statistics .63; hazard ratio 1.03, 95% confidence interval 1.03–1.04). The net reclassification index was .07. The bleeding rates of the modified score risk categories (high, moderate, low, and very low bleeding risk) were 6.3%, 3.8%, 2.9%, and 2.2%, respectively. According to the original score, 65.5% of cancer patients were classified as high bleeding risk (HBR) and 21.6% were low or very low bleeding risk. According to the modified score, 94.0% of cancer patients were HBR, 6.0% were moderate bleeding risk, and no cancer patient was classified as low or very low bleeding risk.

Conclusions

Adding cancer to the PRECISE-DAPT score identifies the majority of patients with cancer as HBR and can improve its discrimination ability without undermining its performance in patients without cancer.

Structured Graphical Abstract

Structured Graphical Abstract Generation and validation of an updated PRECISE-DAPT cancer score. HBR, high bleeding risk; HES, Hospital Episode Statistics; MINAP, Myocardial Ischaemia National Audit Project; ONS, Office for National Statistics; STEMI, ST-elevation myocardial infarction.

Bleeding
PRECISE-DAPT
Cancer
Outcomes
National Institute for Health and Care Research 10.13039/501100000272 Birmingham Biomedical Research Centre 10.13039/501100018952
==== Body
pmc See the editorial comment for this article ‘Tailoring antithrombotic treatment in patients with acute myocardial infarction and cancer: virtue lies in balance’, by A. Jurado-Román and P. Agudo-Quílez, https://doi.org/10.1093/eurheartj/ehae445.

Introduction

Dual antiplatelet therapy (DAPT) has a key role in the prevention of ischaemic events in patients presenting with acute myocardial infarction (AMI) or undergoing percutaneous coronary intervention (PCI).1,2 While DAPT reduces the risk of such ischaemic events, it comes at the expense of bleeding complications, particularly in patients at high bleeding risk (HBR). High bleeding risk represents up to 50% of the population undergoing PCI, and individualization of treatment in relation to DAPT regimens and duration, balancing the risk of ischaemic events to major bleeding complications, underpins international guidelines.1,3 Standardized risk prediction tools have been instrumental in objectively identifying patients that would benefit from long/short antiplatelet duration based on the balance of ischaemic and bleeding risk.4,5 The PRECISE-DAPT score, which includes five clinical and laboratory features, has been demonstrated to predict bleeding in large external validation cohorts6,7 and is used to inform treatment decisions for DAPT duration.5,8 The PRECISE-DAPT score has been endorsed by clinical guidelines to guide duration of antiplatelet therapy in patients following PCI where patients deemed at HBR, defined as PRECISE-DAPT score ≥25, should be considered for shorter DAPT.5

One limitation of current tools, based on the set of predictors that were available during score validation, is the lack of relevant information that might enrich patient evaluation and risk prediction. Hence, updating these tools with additional features might allow more precise risk estimation and personalization of treatment among complex patient groups. Among these groups, cancer is a potential risk factor that has a significant prognostic impact. Patients with cancer represent a significant proportion of patients treated with PCI, with patients with current or a prior history of cancer representing up to 10% of all PCI procedures.9–11 Patients with cancer have an increased risk of both ischaemic and bleeding complications, owing primarily to the local impact of malignancy and the secondary risk of chemotherapy treatments and accompanying coagulopathies and thrombocytopaenia.12,13 Cancer is one of the Academic Research Consortium’s (ARC) major criteria for HBR. While it is considered a major risk factor by experts during a qualitative assessment, cancer has not been so far included in any of the standardized bleeding risk scores currently available for treatment individualization.14

The inclusion of cancer in standard risk assessment tools could be instrumental in improving risk prediction, thus driving personalized care in cancer patients. Therefore, we used multisource linked national prospective clinical registries including the UK Myocardial Ischaemia National Audit Project (MINAP) to study the impact of incorporating cancer diagnosis into the PRECISE-DAPT score.

Methods

Data source

We used the UK MINAP registry that captures data on the presentation profile and clinical care of patients hospitalized with a diagnosis of AMI in England, Wales, and Northern Ireland.15–17 The data collected have been used for research, to audit quality of care, and for public reporting of patients with AMI.16,18,19 The data routinely collected in the registry are patient demographics, admission time and method, cardiovascular comorbidities, clinical characteristics, relevant investigations, in-hospital pharmacological and interventional treatments, in-hospital outcomes, and discharge treatments.17,20–22 We included patients admitted with ST-elevation myocardial infarction (STEMI) and discharged on DAPT in England and Wales between 1 January 2005 and 31 March 2019. Data related to cancer diagnosis were obtained by linking data from the national MINAP registry with hospital admission data of the Hospital Episode Statistics (HES) registry.18,23 Hospital Episode Statistics Admitted Patient Care (HES APC) is a registry composed of data that are collected on all admissions to National Health Service hospitals in England.23 Active cancer was defined as patients who had a diagnosis of cancer at time of admission with STEMI. For each patient who was admitted with a primary diagnosis of STEMI, patients with a diagnosis of cancer at the time of admission were identified using the data from the HES APC database using the International Classification of Diseases (ICD), 10th edition, clinical modification codes (ICD-10-CM). The ICD-10 codes used to ascertain cancer diagnoses are listed in Supplementary data online, Table S1. The HES database has been used frequently in population-based studies from the national British registries as a reliable source of information about AMI and cancer diagnosis.24,25

We used data from the national death registry at the Office for National Statistics (ONS) to obtain data about date and cause of death.26 The ONS is the largest independent producer of official statistics in the UK. It is responsible for collecting and publishing statistics related to the economy, population, and society at national, regional, and local levels.26

Ethical approval

The study underwent formal ethical approval for the data linkages of MINAP, HES, and ONS registries. The ethical approval was provided by the Health Research Authority and the Health and Care Research Wales27 and the Confidentiality Advisory Group, which is an independent body that provides expert advice on the use of confidential patient information (REC reference 20/WA/0312).28

Outcomes

The primary clinical outcome was major bleeding within 1 year after STEMI admission, which was defined as a bleeding event that required hospital admission or led to death in hospital or following discharge. We used the ICD-10-CM codes to identify major bleeding events that required hospital admission from the HES APC registry (see Supplementary data online, Table S1). This definition reflects clinically actionable or life-threatening bleeding and is equivalent to the Bleeding ARC Types 2, 3, and 5.29

Validation cohorts

Overall, 216 709 STEMI patients were included, who were discharged on DAPT between 1 January 2005 and 31 March 2019. The study population was randomly divided into a training data set (75%, 162 476 cases) and a testing data set (25%, 54 233 cases). The training data set included 3452 patients with active cancer, and the testing data set included 1117 patients with active cancer.

The PRECISE-DAPT score

For the current study, we used the simplified PRECISE-DAPT score, a four-item score that was derived from pooled data set of eight randomized studies including 14 963 patients. The variables included in the score were age, prior haemorrhage, creatinine clearance, and haemoglobin levels. The difference with the full five-item version of the score was the exclusion of white cell count, which was the weakest predictor.5 The four-item version of the score has been extensively validated, confirming similar prediction performance compared to the original five-item score, and is used to inform DAPT duration decisions.5,8,30 A more detailed description of the generation of the PRECISE-DAPT and mini-PRECISE-DAPT scores is available in previous publications.5,8

Score generation

Binary cancer score

We generated a new modified score and compared its performance to the original four-item PRECISE-DAPT score. The modified score is a five-item modified PRECISE-DAPT score (binary cancer score) generated using a Cox regression model based on the following predictors: age, prior haemorrhage, creatinine clearance, haemoglobin levels, and cancer (binary variable). The regression coefficients were used to generate an overall score, which was then scaled from 0 to 100, with the higher values indicative of higher bleeding risk. Patients were then classified into four bleeding risk categories (very low, low, moderate, and high bleeding risk, based on the following cut-offs: ≤5, 6–14, 15–24, and ≥25, respectively.5,8

Ordinal cancer scores

We also assessed whether an ordinal cancer variable based on cancer type will improve score discrimination ability further. We tested the performance of two different ordinal scores: the three-category ordinal score and the four-category ordinal score.

The three-category ordinal score was developed using Cox regression model with cancer represented as an ordinal variable, and it was based on the following predictors: age, prior haemorrhage, creatinine clearance, haemoglobin levels, and cancer as ordinal variable [0 = no cancer, 1 = other cancers, 2 = colon and gastrointestinal tract (GIT) cancers]. The regression coefficients were then used to generate an overall score, which was then scaled from 0 to 100, with the higher values indicative of higher bleeding risk. Patients were then classified into four bleeding risk categories in similar fashion to the binary cancer score.

The four-category ordinal score was developed using Cox regression model with cancer represented as an ordinal variable, and it was based on the following predictors: age, prior haemorrhage, creatinine clearance, haemoglobin levels, and cancer as ordinal variable (0 = no cancer, 1 = other cancers, 2 = haematologic malignancies, 3 = colon and GIT cancers). The regression coefficients were then used to generate an overall score, which was then scaled from 0 to 100, with the higher values indicative of higher bleeding risk. Patients were then classified into four bleeding risk categories in similar fashion to the binary cancer score.

Statistical analysis

To make most of the available data, we used multiple imputations with chained equations to impute missing data. It was assumed that the missing data were missing at random, and we generated 10 imputed data. The imputation model closely matched the analytical model, including outcome.31 Analyses from these 10 data sets were combined using Rubin’s rules.32–34 We conducted sensitivity analysis on each imputed data set to ensure that the missing data did not impact the results (see Supplementary data online, Tables S3 and S4). We compared the score performance by reporting Harrell’s C-statistic, Somers’ D, the integrated discrimination improvement index (IDI), and the net reclassification index (NRI). The statistical analysis was done using Stata v16 software.

Results

A total of 216 709 STEMI patients who were discharged on DAPT were included, of which 4569 patients had active cancer. Supplementary data online, Table S2, shows the number of cancer patients in training and testing data sets with emphasis on common cancers in the UK.

Cancer patients were older [median age 73.7, interquartile range (IQR) 66.1–80.6 vs. 64.1, IQR 54.6–74.3] and less likely to receive coronary angiography (80% vs. 88%) and PCI (76% vs. 84%). The crude rate of major bleeding was higher in cancer patients at 30 days (1.8% vs. .7%), 90 days (3.3% vs. 1.5%), and 1 year (6.9% vs. 3.6%) post discharge. Table 1 shows the patient characteristics.

Table 1 Patients’ characteristics

	No cancer	Cancer	P-value	
n	212 140	4569		
Patients’ characteristics and clinical presentation		
 Age at admission, years, median (IQR)	64.1 (54.6–74.3)	73.7 (66.1–80.6)	<.001	
 Women	57 475 (27.1%)	1050 (23.0%)	<.001	
 Ethnicity			<.001	
  White	191 975 (90.5%)	4361 (95.4%)		
  BAME	20 165 (9.5%)	208 (4.6%)		
 BMI, kg/m2, median (IQR)	27.0 (24.2–30.3)	25.9 (23.1–29.1)	<.001	
 Cardiac arrest at presentation	18 111 (8.5%)	345 (7.6%)	.018	
 Killip class				
  I	106 721 (82.7%)	2298 (77.6%)	<.001	
  II	10 033 (7.8%)	310 (10.5%)		
  III	3788 (2.9%)	117 (4.0%)		
  IV	8461 (6.6%)	235 (7.9%)		
 Left ventricular ejection fraction			
  >50%	100 078 (47.2%)	1962 (42.9%)	<.001	
  35%–50%	91 261 (43.0%)	2047 (44.8%)		
  <35%	20 801 (9.8%)	560 (12.3%)		
 Comorbidities				
  Past medical history of angina	25 484 (12.0%)	733 (16.0%)	<.001	
  Previous MI	28 486 (13.4%)	785 (17.2%)	<.001	
  DM	34 351 (16.2%)	777 (17.0%)	.14	
  Hypertension	91 188 (43.0%)	2201 (48.2%)	<.001	
  Hypercholesterolaemia	66 641 (31.4%)	1391 (30.4%)	.16	
  Peripheral vascular disease	5886 (2.8%)	163 (3.6%)	.001	
  Stroke/TIA	10 091 (4.8%)	321 (7.0%)	<.001	
  Family history of CAD	72 667 (34.3%)	1087 (23.8%)	<.001	
 Smoking status				
  Never smoked	73 258 (34.5%)	1716 (37.6%)	<.001	
  Ex-smoker	56 695 (26.7%)	1827 (40.0%)		
  Current smoker	82 187 (38.7%)	1026 (22.5%)		
 Chronic kidney disease	5310 (2.5%)	241 (5.3%)	<.001	
 Asthma/chronic obstructive pulmonary disease	24 678 (11.6%)	691 (15.1%)	<.001	
 Previous PCI	19 444 (9.2%)	504 (11.0%)	<.001	
 Previous CABG	5734 (2.7%)	193 (4.2%)	<.001	
Inpatient management			
 Admitted by a cardiologist	187 653 (90.8%)	3714 (84.0%)	<.001	
 Glycoprotein IIb/IIIa inhibitors	33 522 (20.1%)	532 (14.5%)	<.001	
 Beta-blockers	39 894 (27.5%)	1094 (34.3%)	<.001	
 Warfarin	4745 (2.9%)	110 (3.1%)	.62	
 Loop diuretics	27 597 (17.1%)	814 (22.8%)	<.001	
 Aldosterone antagonists	24 782 (15.4%)	496 (13.9%)	.015	
 Coronary angiogram	171 806 (87.8%)	3212 (79.7%)	<.001	
 PCI	180 024 (84.9%)	3494 (76.5%)	<.001	
 CABG	1502 (.8%)	24 (.6%)	.20	
 Antiplatelet combination				
  Aspirin + clopidogrel	157 408(74.2%)	3605(78.9%)		
  Aspirin + P2Y12 inhibitor (ticagrelor/prasugrel)	54 732 (25.6%)	964 (21.1%)		
Clinical outcomes			
 Bleeding at 30 days	1541 (.7%)	84 (1.8%)	<.001	
 Bleeding at 90 days	3138 (1.5%)	153 (3.3%)	<.001	
 Bleeding at 1 year	7716 (3.6%)	313 (6.9%)	<.001	
Cancer type				
 Lung cancer		535(11.7%)		
 Prostate cancer	1523(33.3%)	
 Colon and GIT cancer	626(13.7%)	
 Breast cancer	190 (4.2%)	
 Renal/bladder cancer	379(8.3%)	
 Haematologic malignancy	728 (15.9%)	
 Other cancers	588 (12.9%)	
IQR, interquartile range; MI, myocardial infarction; HTN, hypertension; PVD, peripheral vascular disease; CKD, chronic kidney disease; TIA, transient ischaemic attack; CAD, coronary artery disease; PCI, percutaneous coronary intervention; CABG, coronary artery bypass graft; GIT, gastrointestinal tract; BAME, black, Asian and minority ethnic.

Modified PRECISE-DAPT score (binary cancer score)

We generated a modified PRECISE-DAPT score and assessed its performance. The binary cancer score was generated by modelling the variables included in the original PRECISE-DAPT score plus cancer as a binary variable (0 = no cancer, 1 = active cancer). The binary cancer score showed that patients with prior haemorrhage [hazard ratio (HR) 3.19, 95% confidence interval (CI) 2.91–3.50] and cancer (HR 1.82, 95% CI 1.60–2.07) have significantly higher odds of bleeding (Table 2). Figure 1 shows the binary cancer score nomogram and the corresponding rate of major bleeding at 1 year.

Figure 1 The modified score (binary cancer score) nomogram and the rate of major bleeding at 1 year

Table 2 Cox proportional hazard models of bleeding at 1 year for variables included in the binary cancer score

Variable	HR (95% CI)	P-value	
Age (per 1 year)	1.03 (1.03–1.03)	<.001	
Prior haemorrhage	3.19 (2.91–3.50)	<.001	
Haemoglobin	.88 (.84–.92)	<.001	
Creatinine clearance	.99 (.99–.99)	.001	
Cancer	1.82 (1.60–2.07)	<.001	

Performance of the binary cancer score vs. original PRECISE-DAPT score

We used the training data set to assess the ability of the binary cancer score to predict major bleeding events within 1 year post discharge. The original scaled score showed moderate accuracy (C-statistic .60; Table 3).

Table 3 Performance of the original PRECISE-DAPT score and the binary cancer score in the training and testing data sets

	Training data set	Testing data set	
	Scaled score	HBR vs. non-HBR	Scaled score	HBR vs. non-HBR	
Original score	
 HR	1.02	1.64	1.02	1.6	
 95% CI	1.02–1.02	1.55–1.72	1.02–1.02	1.47–1.75	
 Harrell’s C-statistic	.6				
 Somers’ D	.21				
 Sensitivity		62.94%		62.34%	
 Specificity		48.37%		48.54%	
 Positive predictive value		4.53%		4.51%	
 Negative predictive value		97.10%		97.07%	
 Accuracy		75.3%		75.8%	
Binary cancer score	
 HR	1.03	2.43	1.03	2.44	
 95% CI	1.03–1.04	2.32–2.56	1.03–1.04	2.32–2.56	
 Harrell’s C-statistic	.64				
 Somers’ D	.28				
 Sensitivity		42.41%		42.23%	
 Specificity		75.80%		75.92%	
 Positive predictive value		6.39%		6.40%	
 Negative predictive value		97.13%		97.12%	
 Accuracy		74.6%		74.5%	
 IDI		.04%		.01%	

The binary cancer score showed higher discrimination than the original PRECISE-DAPT score (C-statistics .64; HR 1.03, 95% CI 1.03–1.04) as it improved score specificity from 48.37% to 75.80% (Table 3). We compared the C-statistics of the original and binary cancer scores by performing a Wald test using the DeLong covariance estimates for the C-statistics estimate. It showed that the C-statistics values were significantly different (P < .001). Supplementary data online, Figure S1, shows the receiver operating characteristic curve of the original PRECISE-DAPT score and the binary cancer score. Figure 2 shows the calibration curve of the binary cancer, demonstrating that the modified score is a promising predictive tool as the calibration slope is 1 and the calibration in the large is 0. Supplementary data online, Figure S2, shows the observed (Kaplan–Meier) and predicted (Cox proportional hazard) bleeding probability for the binary cancer score where no significant differences between the observed and predicted values were noted.

Figure 2 Calibration curve of the binary cancer score

The overall NRI of the binary cancer score was 7% in both the training and testing data sets (Table 4). In cancer patients, 27% of patients who had major bleeding were reclassified as HBR (NRI event 27%) whereas 28% of patients who did not develop major bleeding were downgraded from HBR category. Table 5 shows the NRI of AMI patients with cancer.

Table 4 Net reclassification index of the binary cancer score in the training and testing data sets

	Training data set	Testing data set	
	Bleeding at 1 year	Bleeding at 1 year	
	No	Yes	Total	No	Yes	Total	
Risk category (HBR vs. not HBR)							
 Downgraded	45 156	1408	46 564	15 040	460	15 500	
 No change	107 644	4464	112 108	35 942	1509	37 451	
 Upgraded	3639	221	3860	1217	65	1282	
Total	156 439	6093	162 532	52 199	2034	54 233	
NRI			.07			.07	
IDI			.04%			.01%	

Table 5 Net reclassification index of acute myocardial infarction patients with cancer using the binary cancer score

	Bleeding at 1 year	
	No	Yes	Total	
Risk category (HBR vs. not HBR)				
 Downgraded	80	2	82	
 No change	2897	225	3122	
 Upgraded	1277	88	1365	
 Total	4254	315	4569	
NRI event	.27			
NRI no event	−.28			
Total NRI	−.01			

As most patients did not have active cancer at presentation, we performed a sensitivity analysis to confirm that adding cancer to the original score does not undermine score performance in this subgroup. The sensitivity analysis showed that the binary cancer score had better performance than the original score in patients without active cancer (see Supplementary data online, Table S5).

The modified PRECISE-DAPT score (binary cancer score) and bleeding risk

Figure 3 shows 1-year bleeding rates across the different risk categories of the binary cancer score in the whole study population. The bleeding rates of high, moderate, low, and very low bleeding risk were 6.3%, 3.8%, 2.9%, and 2.2%, respectively.

Figure 3 Annual bleeding rates across the different risk categories of the binary cancer score

Table 6 shows the bleeding risk categories in patients with and without cancer according to the original and binary cancer score in the whole study population. According to the original score, 65.5% of cancer patients were classified as HBR, and 21.6% were considered to have a low or very low bleeding risk. According to the binary cancer score, 94.0% of cancer patients were classified as HBR, 6.0% were classified as moderate bleeding risk, and no cancer patient was classified as low or very low bleeding risk.

Table 6 Bleeding risk categories in patients with and without cancer

	Original score	Binary cancer score	
	Frequency	%	Frequency	%	
No cancer					
 Very low risk	63 482	29.92	76 249	35.94	
 Low risk	17 360	8.18	41 915	19.76	
 Moderate risk	19 205	9.05	40 223	18.96	
 HBR	112 093	52.84	53 753	25.34	
Cancer group	
 Very low risk	604	13.22	0	0	
 Low risk	384	8.4	0	0	
 Moderate risk	587	12.85	292	6	
 HBR	2994	65.53	4277	94	

Ordinal cancer score based on cancer type

We compared the risk of bleeding of each cancer type to the risk of bleeding in other cancers to identify the cancer type with higher bleeding risk using Cox proportional hazard models. Supplementary data online, Table S6, shows that colon and GIT cancers have the highest risk of bleeding (HR 1.77, 95% CI 1.19–2.62). Therefore, a three-category ordinal score was developed where the cancer variable was represented as follows: 0 = no cancer, 1 = other cancers, and 2 = colon and GIT cancers.

The three-category ordinal cancer score (C-statistics .64; HR 1.04, 95% CI 1.05–1.05) did not show higher discrimination than the binary cancer score (C-statistics .64; HR 1.03, 95% CI 1.03–1.04). The proportion of cancer patients classified as HBR according to the binary score was 94.0%. The three-category ordinal cancer score classified all colon and GIT cancers as HBR (100%), but only 43.7% of the other cancers were classified as HBR. In patients without cancer, only 4.9% were classified as HBR according to the three-category ordinal score compared to 25.3% according to the binary cancer score.

As the three-category ordinal score did not show better discrimination than the binary score, we developed a four-category ordinal score and tested its performance to see whether adding additional categories based on cancer type will improve the score performance. The four-category ordinal score was developed, and the cancer variable was represented as follows: 0 = no cancer, 1 = other cancers, 2 = haematologic malignancies, and 3 = colon and GIT cancers. The four-category ordinal cancer score (C-statistics .64; HR 1.06, 95% CI 1.06–1.07) did not show higher discrimination than the binary cancer score (C-statistics .64; HR 1.03, 95% CI 1.03–1.04). Similar to the three-category ordinal score, the four-category ordinal cancer score classifies all colon and GIT cancers as HBR (100%), but classified 3.3% of haematologic malignancies and 19.2% of other cancers as HBR. Regarding patients without cancer, only .9% of patients without cancer were classified as HBR according to the four-category ordinal score compared to 25.3% according to the binary cancer score.

These findings shows that the three- and four-category ordinal scores do not provide better discrimination mainly because they reduce the PRECISE-DAPT ability to detect the HBR population in patients without cancer, which counterbalances any improvement in the prediction of bleeding risk in cancer subtypes.

Discussion

This is a population-based study that evaluated the impact of adding cancer diagnosis as a predictor on the performance of the PRECISE-DAPT score. In this national ACS population, which included patients with active cancer, the PRECISE-DAPT score showed moderate discriminative ability in AMI patients discharged on DAPT, and only 65.5% of cancer patients were classified HBR. A modified PRECISE-DAPT score generated by incorporating cancer as a binary covariate in the PRECISE-DAPT score improved score performance and ability to discriminate the HBR population in cancer patients without compromising the PRECISE-DAPT performance in patients without cancer. The modified score classified 94% of cancer patients as HBR, with the HBR group being associated with higher rates of major bleeding (Structured Graphical Abstract).

Patients presenting with AMI with concomitant cancer have a high risk of bleeding. Data from the nationwide Swedish quality registry showed that cancer is a strong predictor of severe bleeding in AMI patients.12 In fact, the European Society of Cardiology (ESC) included active malignancy as major criteria for HBR at time of PCI.35 Nevertheless, active malignancy is not part of the majority of bleeding risk assessment tools, probably because cancer patients are excluded from the clinical trials where these scores were derived from. Since national clinical registries like MINAP and HES provide information on cancer status, these were used as an opportunity to assess the feasibility and performance of integrating cancer diagnosis in the PRECISE-DAPT score.

Our study showed that a modified PRECISE-DAPT score that included cancer as a predictor outperformed the original PRECISE-DAPT score without undermining the PRECISE-DAPT performance in patients without cancer. Addition of cancer as a predictor improved the score performance and helped discriminate AMI patients with active cancer at HBR. Current guidelines recommend to classify all patients with active malignancy as HBR patients.35 According to the original PRECISE-DAPT score for the assessment of bleeding risk, only 65% of patients with cancer would have been considered HBR compared to 94% of patients if the modified score had been used. This indicates that around 29% of cancer patients who should be considered at HBR and receive shorter duration of DAPT would receive DAPT for longer periods if the original score was used, exposing them to increased risk of major bleeding.

Furthermore, the cut-offs adopted by the ARC for HBR (ARC-HBR) to define HBR (major bleeding rate >4% at the first year) are consistent with the cut-offs of the modified PRECISE-DAPT score.14 The annual incidence rate of bleeding events in the HBR group of the modified score is 6.3%, which is higher than the ARC-HBR cut-off, while the incidence rates for the moderate, low, and very low risk groups are <4%. It is also worth noting that around 6% of cancer patients were classified as moderate bleeding risk who should qualify for a standard duration of DAPT. These findings indicate that not all cancer patients should be classified as HBR, and further studies are required to generate more accurate risk assessment tools of bleeding events in cancer patients. Additionally, the binary cancer score has better performance than the original score as it improves score ability to correctly classify STEMI patients with cancer as HBR (i.e. higher specificity). These features make the modified PRECISE-DAPT score an efficient convenient tool that will help cardiologists make informed decisions about the use and duration of DAPT in AMI patients with and without cancer.

This study has important clinical and academic implications. To the best of our knowledge, this is the first study to show that integration of active cancer as a covariate improves the performance of the PRECISE-DAPT score. Further studies are recommended to externally validate the modified PRECISE-DAPT score. Our study also shows that the PRECISE-DAPT score is not only a convenient risk assessment tool with reasonable performance but also a flexible score that can be improved. However, there are few limitations that should be considered when interpreting the results of this study. The databases do not routinely capture indication for anticoagulation, use of direct oral anticoagulants, or the duration of antiplatelet use after discharge, which we could not adjust for, although this information is not a component of the PRECISE-DAPT score. They also lack data on cancer treatment at various disease stages, which limited our ability to evaluate the impact of these treatments on bleeding risk.

Conclusions

The modified PRECISE-DAPT score generated by adding cancer as a binary covariate had better performance and discrimination ability than the PRECISE-DAPT score. Consideration of cancer type does not improve performance and may compromise the performance of the score in non-cancer patients. The modified PRECISE-DAPT score will enable clinicians to make informed decisions about the bleeding risk particularly in cancer patients. Additional studies are recommended to assess the score performance in external populations.

Supplementary data

Supplementary data are available at European Heart Journal online.

Supplementary Material

ehae463_Supplementary_Data

Declarations

Disclosure of Interest

All authors declare no disclosure of interest for this contribution.

Data Availability

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

Funding

This study is funded by the National Institute for Health and Care Research (NIHR) Birmingham Biomedical Research Centre (BRC). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. Prof. Costa is funded by the European Union (ERC, ORACLE, ERC-2023-STG-101117469). Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.

Ethical Approval

The study underwent formal ethical approval for the data linkages of MINAP, HES, and ONS registries. The ethical approval was provided by the Health Research Authority and the Health and Care Research Wales27 and the Confidentiality Advisory Group, which is an independent body that provides expert advice on the use of confidential patient information.

Pre-registered Clinical Trial Number

None supplied.
==== Refs
References

1 Collet JP , ThieleH, BarbatoE, BarthélémyO, BauersachsJ, BhattDL, et al 2020 ESC guidelines for the management of acute coronary syndromes in patients presenting without persistent ST-segment elevation. Eur Heart J 2021;42 :1289–367. 10.1093/eurheartj/ehaa575 32860058
2 Costa F , MontaltoC, BrancaM, HongS-J, WatanabeH, FranzoneA, et al Dual antiplatelet therapy duration after percutaneous coronary intervention in high bleeding risk: a meta-analysis of randomized trials. Eur Heart J 2023;44 :954–68. 10.1093/eurheartj/ehac706 36477292
3 Valgimigli M , FrigoliE, HegD, TijssenJ, JüniP, VranckxP, et al Dual antiplatelet therapy after PCI in patients at high bleeding risk. N Engl J Med 2021;385 :1643–55. 10.1056/NEJMoa2108749 34449185
4 Baber U , MehranR, GiustinoG, CohenDJ, HenryTD, SartoriS, et al Coronary thrombosis and major bleeding after PCI with drug-eluting stents: risk scores from PARIS. J Am Coll Cardiol 2016;67 :2224–34. 10.1016/j.jacc.2016.02.064 27079334
5 Costa F , van KlaverenD, JamesS, HegD, RäberL, FeresF, et al Derivation and validation of the predicting bleeding complications in patients undergoing stent implantation and subsequent dual antiplatelet therapy (PRECISE-DAPT) score: a pooled analysis of individual-patient datasets from clinical trials. Lancet 2017;389 :1025–34. 10.1016/S0140-6736(17)30397-5 28290994
6 Castelijns MC , HagemanSHJ, TeraaM, van der MeerMG, WesterinkJ, CostaF, et al External validation of bleeding risk models for the prediction of long-term bleeding risk in patients with established cardiovascular disease. Am Heart J 2023;260 :72–81. 10.1016/j.ahj.2023.02.011 36841319
7 Munafò AR , MontaltoC, FranzinoM, PistelliL, Di BellaG, FerliniM, et al External validity of the PRECISE-DAPT score in patients undergoing PCI: a systematic review and meta-analysis. Eur Heart J Cardiovasc Pharmacother 2023;9 :709–21. 10.1093/ehjcvp/pvad063 37634083
8 Costa F , van KlaverenD, ColomboA, FeresF, RäberL, PilgrimT, et al A 4-item PRECISE-DAPT score for dual antiplatelet therapy duration decision-making. Am Heart J 2020;223 :44–7. 10.1016/j.ahj.2020.01.014 32151822
9 Bharadwaj AS , SwamyPM, MamasMA. Outcomes of percutaneous coronary interventions in cancer patients. Expert Rev Cardiovasc Ther 2020;18 :25–32. 10.1080/14779072.2020.1718493 31951772
10 Dafaalla M , Abdel-QadirH, GaleCP, SunL, López-FernándezT, MillerRJH, et al Outcomes of ST elevation myocardial infarction in patients with cancer: a nationwide study. Eur Heart J Qual Care Clin Outcomes 2023;9 :806–17. 10.1093/ehjqcco/qcad012 36921979
11 Potts JE , IliescuCA, Lopez MatteiJC, MartinezSC, HolmvangL, LudmanP, et al Percutaneous coronary intervention in cancer patients: a report of the prevalence and outcomes in the United States. Eur Heart J 2019;40 :1790–800. 10.1093/eurheartj/ehy769 30500952
12 Velders MA , HagströmE, JamesSK. Temporal trends in the prevalence of cancer and its impact on outcome in patients with first myocardial infarction: a nationwide study. J Am Heart Assoc 2020;9 :e014383. 10.1161/JAHA.119.014383 32067596
13 Giri J , NathanAS. The pros and cons of percutaneous coronary intervention in patients with cancer. JACC CardioOncol 2019;1 :156–8. 10.1016/j.jaccao.2019.11.002 34396176
14 Urban P , MehranR, ColleranR, AngiolilloDJ, ByrneRA, CapodannoD, et al Defining high bleeding risk in patients undergoing percutaneous coronary intervention: a consensus document from the academic research consortium for high bleeding risk. Eur Heart J 2019;40 :2632–53. 10.1093/eurheartj/ehz372 31116395
15 Birkhead JS , WestonCFM, ChenR. Determinants and outcomes of coronary angiography after non-ST-segment elevation myocardial infarction. A cohort study of the Myocardial Ischaemia National Audit Project (MINAP). Heart 2009;95 :1593–9. 10.1136/hrt.2008.164426 19508971
16 Rashid M , CurzenN, KinnairdT, LawsonCA, MyintPK, KontopantelisE, et al Baseline risk, timing of invasive strategy and guideline compliance in NSTEMI: nationwide analysis from MINAP. Int J Cardiol 2020;301 :7–13. 10.1016/j.ijcard.2019.11.146 31810815
17 Rashid M , KontopantelisE, KinnairdT, CurzenN, GaleCP, MohamedMO, et al Association between hospital cardiac catheter laboratory status, use of an invasive strategy, and outcomes after NSTEMI. Can J Cardiol 2020;36 :868–77. 10.1016/j.cjca.2019.10.010 32146069
18 Herrett E , SmeethL, WalkerL, WestonC; MINAP Academic Group. The Myocardial Ischaemia National Audit Project (MINAP). Heart 2010;96 :1264–7. 10.1136/hrt.2009.192328 20659944
19 Moledina SM , ShoaibA, WestonC, AktaaS, Van SpallG, KassamH, et al Ethnic disparities in care and outcomes of non-ST-segment elevation myocardial infarction: a nationwide cohort study. Eur Heart J Qual Care Clin Outcomes 2022;8 :518–28. 10.1093/ehjqcco/qcab030 33892502
20 Wu J , MamasM, RashidM, WestonC, HainsJ, LuescherT, et al Patient response, treatments, and mortality for acute myocardial infarction during the COVID-19 pandemic. Eur Heart J Qual Care Clin Outcomes 2021;7 :238–46. 10.1093/ehjqcco/qcaa062 32730620
21 Mamas MA , KwokCS, KontopantelisE, FryerAA, BuchanI, BachmannMO, et al Relationship between anemia and mortality outcomes in a national acute coronary syndrome cohort: insights from the UK Myocardial Ischemia National Audit Project registry. J Am Heart Assoc 2016;5 :e003348. 10.1161/JAHA.116.003348 27866164
22 Rashid M , WuJ, TimmisA, CurzenN, ClarkeS, ZamanA, et al Outcomes of COVID-19-positive acute coronary syndrome patients: a multisource electronic healthcare records study from England. J Intern Med 2021;290 :88–100. 10.1111/joim.13246 33462815
23 Herbert A , WijlaarsL, ZylbersztejnA, CromwellD, HardelidP. Data resource profile: Hospital Episode Statistics Admitted Patient Care (HES APC). Int J Epidemiol 2017;46 :1093–1093i. 10.1093/ije/dyx015 28338941
24 Thomas DS , Gentry-MaharajA, RyanA, FourkalaE-O, ApostolidouS, BurnellM, et al Colorectal cancer ascertainment through cancer registries, hospital episode statistics, and self-reporting compared to confirmation by clinician: a cohort study nested within the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS). Cancer Epidemiol 2019;58 :167–74. 10.1016/j.canep.2018.11.011 30616086
25 Coles B , TeeceL, WestonC, de BelderMA, Oliver-WilliamsC, WelchCA, et al Case-ascertainment of acute myocardial infarction hospitalizations in cancer patients: a cohort study using English linked electronic health data. Eur Heart J Qual Care Clin Outcomes 2022;8 :86–95. 10.1093/ehjqcco/qcab045 34156470
26 Deaths . Office for National Statistics. https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths#publications (20 October 2022, date last accessed)
27 Research Ethics Service and Research Ethics Committees . Health Research Authority. https://www.hra.nhs.uk/about-us/committees-and-services/res-and-recs/(2 November 2022, date last accessed)
28 Confidentiality Advisory Group . Health Research Authority. https://www.hra.nhs.uk/about-us/committees-and-services/confidentiality-advisory-group/(2 November 2022, date last accessed)
29 Mehran R , RaoSV, BhattDL, GibsonCM, CaixetaA, EikelboomJ, et al Standardized bleeding definitions for cardiovascular clinical trials: a consensus report from the Bleeding Academic Research Consortium. Circulation 2011;123 :2736–47. 10.1161/CIRCULATIONAHA.110.009449 21670242
30 Wester A , MohammadMA, OlivecronaG, HolmqvistJ, YndigegnT, KoulS. Validation of the 4-item PRECISE-DAPT score: a SWEDEHEART study. J Am Heart Assoc 2021;10 :e020974. 10.1161/JAHA.121.020974 34612051
31 Kontopantelis E , WhiteIR, SperrinM, BuchanI. Outcome-sensitive multiple imputation: a simulation study. BMC Med Res Methodol 2017;17 :2. 10.1186/s12874-016-0281-5 28068910
32 Rashid Hons M , Gale HonsCP, Curzen HonsN, Ludman HonsP, De Belder HonsM, Timmis HonsA, et al Impact of coronavirus disease 2019 pandemic on the incidence and management of out-of-hospital cardiac arrest in patients presenting with acute myocardial infarction in England. J Am Heart Assoc 2020;9 :e018379. 10.1161/JAHA.120.018379 33023348
33 Dafaalla M , RashidM, SunL, QuinnT, TimmisA, WijeysunderaH, et al Impact of availability of catheter laboratory facilities on management and outcomes of acute myocardial infarction presenting with out of hospital cardiac arrest. Resuscitation 2022;170 :327–34. 10.1016/j.resuscitation.2021.10.031 34718080
34 Dafaalla M , RashidM, BondRM, SmithT, ParwaniP, ThammanR, et al Racial disparities in management and outcomes of out-of-hospital cardiac arrest complicating myocardial infarction: a national study from England and Wales. CJC Open 2021;3 :S81–8. 10.1016/j.cjco.2021.09.026 34993437
35 Byrne RA , RosselloX, CoughlanJJ, BarbatoE, BerryC, ChieffoAet al 2023 ESC guidelines for the management of acute coronary syndromes. Eur Heart J 2023;44 :3720–826. doi:10.1093/eurheartj/ehad191 37622654
