The database concerning consecutive adult patients with AMI who were ≥ 18 years of age and were admitted to the cardiac intensive care unit (CICU) of the Heart Center at the Beijing Chaoyang Hospital affiliated to the Capital Medical University between December 1, 2010, and June 31, 2019, was analyzed. Patients who died or were discharged before a definite assessment of their cardiac function and GIB were excluded, as were patients with GIB before the onset of AMI and those with esophageal variceal bleeding. In total, 6,458 patients were enrolled and 36 were excluded. Of the analyzed cases, 2,663 patients (41.2%) had a non-ST-elevation myocardial infarction (NSTEMI) and 3,795 (58.7%) had an ST-elevation myocardial infarction (STEMI).

The diagnostic criteria for AMI were as follows: the STEMI diagnosis was based on the European Heart Association 2017 STEMI diagnostic criteria, including chest pain with a new ST-segment elevation in more than two adjacent electrocardiogram leads or a new left bundle branch block (6), and the NSTEMI diagnosis was based on the 2015 diagnostic criteria, including chest pain and new onset of an ST-segment depression in more than two adjacent electrocardiogram leads, with an increase in serum troponin I after admission to more than three times the upper limit of the normal value (7).

Overt GI bleeding was defined as hematemesis (the vomiting of bright red blood or dark brown granular material that resembles coffee grounds) and hematochezia or melena). Clinically important GI bleeding was defined as overt bleeding complicated by one of the following within 24 h after the onset of bleeding (in the absence of other causes): a spontaneous decrease of more than 20 mmHg in the systolic blood pressure; an increase of 20 beats per min in the heart rate or a decrease of more than 10 mmHg in the systolic blood pressure measured on sitting up; a decrease in hemoglobin of at least 2 g/dl (1.24 mmol/L); or a subsequent transfusion of two or more units of red blood cells during the bleeding episode (5).

This retrospective clinical trial was conducted at Beijing Chaoyang Hospital affiliated with the Capital Medical University, and the ethics committee of the hospital approved the study (approval number: 2019-scientific-5-1). Due to the retrospective nature of the study, the requirement for informed consent was waived.

All data were obtained from the Beijing Chaoyang Hospital, which is one of the most influential hospitals in China for the treatment of coronary heart disease. The Heart Center of the hospital established one of the first primary percutaneous coronary intervention teams in China in 1993 and is well-known for its quick response in performing reperfusion on eligible patients with AMI. The patients in the study are representative of the general Beijing population.

The electronic medical records were reviewed by the survey doctors (LY, FDJ, ZDP, ZZY, LN, and FYF), who collected the basic demographic information, previous medical history, admission diagnosis, therapeutic information, and discharge prescription(s) of all the patients. Diagnoses of digestive tract tumor history, peptic ulcer history, stroke, anxiety and depression status, and renal insufficiency were based on the International Classification of Diseases Tenth Revision (8) diagnostic information, which was provided by the Information Center of the Beijing Chaoyang Hospital. The codes of K25, K26, K27, and K92 were screened for. All GIB events were confirmed by independent gastroenterology consultant (HJY) after the onset of AMI and during hospitalization, and patients with preexisting GIB were excluded.

Diagnoses of hyperlipidemia, diabetes, and anemia were based on medical history combined with laboratory tests during admission. The smoking status and alcohol use were obtained from the electronic medical records. Non-steroidal anti-inflammatory drug (NSAID) history referred to the use of NSAIDs before admission, and a history of NSAID use was recorded if the patient had not stopped using them at least 3 months before admission. Antiplatelet drug history referred to the use of aspirin and/or a P2Y12 receptor antagonist (RA) within 3 months of admission. Aspirin and clopidogrel were administered in an oral dosage form of 100 and 75 mg, respectively, in all cases.

The cardiac function during the acute phase of AMI objectively reflects the severity of the AMI. The Killip classification method is simple and accurate and has been widely adopted by clinicians over the past five decades to assess the severity of an AMI. In this study, the Killip classification was used to determine the severity of the AMI and analyzed the relationship between the different cardiac function levels and GIB. The Killip classification was determined based on the previous literature as follows: class 1 (K1), no signs of heart failure; class 2 (K2), signs indicating mild to moderate heart failure (i.e., third heart sound gallop, rales halfway up the lung fields, or elevated jugular venous pressure); class 3 (K3), pulmonary edema; and class 4 (K4), cardiogenic shock or refractory hypotension (9).

Echocardiographic ejection fraction is a more quantitative and accurate method to evaluate the cardiac function than Killip classification. The latter is easier to implement without the help of medical equipment. In some cases, echocardiographic ejection fraction was collected, because not all patients were examined during CICU, or the data were lost. The two non-invasive methods to evaluate the cardiac function were compared in AMI episode.

Information about each patient regarding the use of antithrombotic drugs and gastric mucosal protective agents during hospitalization was provided by the Pharmaceutical Center of the Beijing Chaoyang Hospital. The available gastric mucosal protective agents included: (1) proton pump inhibitors (PPIs), such as omeprazole, pantoprazole, rabeprazole, and esomeprazole; (2) Histamine-2 receptor antagonists (H2RAs), such as cimetidine, ranitidine, and famotidine; and (3) oral gefarnate. An intravenous PPI was the first choice for the prophylactic treatment of peptic ulcers. Intravenous H2RAs were given when patients were allergic to PPIs or had liver and kidney damage. The PPIs or H2RAs were used for 3–5 days during a CICU stay. Oral gefarnate (one to two tablets, three times a day) was administered once the intravenous prophylaxis was withdrawn, and the dose was maintained for 3 months.

In line with previously established criteria, anemia was diagnosed when the baseline hematocrit value was < 39% for men and < 36% for women (10). Likewise, the chronic renal insufficiency was defined as a creatinine clearance of < 60 ml/min (11), and thrombocytopenia was defined as a nadir in-hospital platelet count of < 100 × 10⁹/L (12).

The endpoints were overt GIB and clinically important GIB. GIB event was performed by an independent gastroenterology consultant.

All statistical analyses were performed with SPSS Statistics 20.0 (IBM Company). The subjects were divided into four groups (Killip classes 1–4). Continuous variables with a normal distribution were expressed as the mean ± standard deviation (age and hospital stay); non-normally distributed variables were expressed as quartile intervals [median (Q1, Q3)], and countable data were expressed as frequency (n,%). According to the data type, either a one-way analysis of variance or a chi-squared test was used to compare the baseline characteristics between the groups. Univariate and multivariate conditional logistic regression models were constructed. The univariate logistic regression model was used to analyze the relationship between Killip classes 1–4 and GIB, and the multivariate regression model was used to adjust for common factors influencing GIB. All the variables were compared using a bilateral test, and p < 0.05 was statistically significant.

With respect to the 6,458 patients, the occurrence of GIB in Killip classes 1–4 was as follows: 42 out of 3,292 (1.27%) in the Killip class 1 group; 41 out of 2,372 (1.73%) in the Killip class 2 group; 26 out of 513 (5.07%) in the Killip class 3 group; and 22 out of 281 (7.83%) in the Killip class 4 group. The time of GIB from AMI onset (days) was shorter in the Killip class 1 group than in any other Killip class group, and the times of the other groups were comparable (Table 2).

There was a trend that the higher Killip classes had lower echocardiographic ejection fraction; and a significant difference between the ejection fraction of groups K2 and K3 (58.66 ± 11.70 vs. 48.14 ± 11.03, p < 0.001). With respect to groups K1 and K2, the ejection fraction was comparable (61.06 ± 10.94 vs. 58.66 ± 11.70, p = 0.054), as it was also comparable for groups K3 and K4 (48.14 ± 11.03 vs. 46.94 ± 12.37, p = 0.052) (see Table 2 and Figure 1).

The logistic regression analysis was performed using GIB as the dependent variable and the Killip classification as the independent variable. There was no significant difference between the class 2 and class 1 groups in the risk of GIB [odds ratio (OR) = 1.36, 95% confidence interval (CI): 0.88–2.09, and p = 0.164]. The risk in the class 3 group was 4.13 times higher than in the class 1 group (OR = 4.13, 95% CI: 2.51–6.79, and p < 0.001), and the risk in the class 4 group was 6.57 times higher than it was in the class 1 group (OR = 6.57, 95% CI: 3.86–11.18, and p < 0.001) (see Table 2 and Figure 2). After adjustment, the results show that the influence of the different Killip class levels of cardiac function on the occurrence of GIB remains significant (see Table 3).

The gastric (and sometimes the esophageal or duodenal) mucosal barrier is disrupted by the acute illness, and this disruption may present in the form of erosive gastritis ranging from asymptomatic superficial lesions and occult GI bleeding to overt clinically significant GIB (6, 13). However, although critically ill patients are often at risk of SUs, it has been discovered that not all patients with AMI develop them (5). The previous clinical studies have shown that the incidence of SUs is related to the severity of the primary disease (5).

Acute myocardial infarction is considered to be one of the common etiologies for the development of stress ulcers (such as gastritis and gastropathy), with less being intensively stressful than burns and cranial trauma (7, 9). On the other hand, the use of anticoagulant and dual antiplatelet therapy (DAPT) were recommended when treating patients with AMI (6). For these reasons, patients with AMI treated in CICU have an increased risk of GIB.

The 2017 ESC STEMI guideline recommended that a PPI in combination with DAPT is recommended in patients at high risk of GIB (6). It means prophylactic use of PPI is reasonable for high-risk bleeding patients and might not be necessary for those at low risk without pre-evaluation. Severe cardiac insufficiency in AMI episode increases the intensity of stress and gastrointestinal congestion. When considering the risk of GIB, the patient’s cardiac function status cannot be ignored.

The CRUSADE and DAPT score systems are recommended to identify the high-risk bleeding patients (14, 15). Both scoring systems identified the heart failure as a high risk factor for bleeding. Yet the degree of cardiac dysfunction during AMI scenario varies from asymptomatic to cardiogenic shock. Therefore, in the present study, it was hypothesized that different degrees of cardiac dysfunction have different contribution on GIB.

The cardiac dysfunction was already thought to be linked to GIB (16, 17). The present study found that the occurrence of GIB is significantly correlated, respectively, with each Killip class in patients with AMI.

Echocardiography was performed in 67.7, 79.8, 84.1, and 85.9% of the patients in the K1, K2, K3, and K4 groups, respectively. Meanwhile, Killip class was obtained for all cases. Since our study has demonstrated the relationship that the higher Killip classes had lower the echocardiographic EF values, the consistent assessment of cardiac function by two methods makes it credible to use Killip classification to evaluate the cardiac function.

The logistic regression model, after adjusting for age, sex, comorbidities, the use of protective agents for the gastric mucosa, and the use of antithrombotic drugs, showed that each Killip class has a different effect on GIB, and, thus, the classification can help to determine the risk of GIB in patients with AMI.

It is known that, in the acute phase of AMI, the location and duration of the occlusion of the culprit vessel, the infarct size, and the baseline characteristics lead to varying complications and result in different grades of hemodynamic disorders. Mild cases of AMI may be in a stable state throughout the acute phase, with fewer complications, and result in a shorter stay in the CICU. However, advanced cases may experience heart failure, hypotension, cardiopulmonary resuscitation, cardiogenic shock, multiorgan dysfunction, with more in-hospital complications, and higher mortality (18, 19).

The putative mechanisms include impaired cardiac function complicating low cardiac output leading to hypotension and subsequent gastrointestinal wall hypoxia, reducing gastric blood flow, and mucosal ischemia. The right-side cardiac dysfunction leads to the gastrointestinal wall congestion and hypoxia. Severe cardiac dysfunction aggravates the process and elevates the stress. The above pathological processes will lead to GIB.

Endoscopic studies have shown that gastric erosions are present in up to 90% of patients in the intensive care unit (20, 21). In most patients, these lesions are superficial and asymptomatic. However, clinically important bleeding occurs in approximately 1% to 4% of critically ill patients (22). The previous studies have suggested that the endoscopy after AMI increases complications, such as hypotension, arrhythmia, and repeated acute coronary syndrome (23–26). This seems to be confirmed by the concerns of medical staff regarding this invasive examination during the acute phase of AMI and could explain why only two of the patients in this study, who had an urgent need, underwent endoscopy. Nevertheless, it is believed that this study is meaningful, as it provides a basis for clinicians to judge the risk of GIB according to the level of heart function of patients with AMI.

This study found that the stress ulcer prophylaxis (SUP) had been given to all the AMI patients at admission and during hospitalization to prevent GIB. In 2008, the American College of Cardiology and the American Heart Association, in association with the American College of Gastroenterology recommends gastroprotection with a PPI for all patients receiving DAPT (27). The use of SUP has also been suggested for the prevention and treatment of SUs in critically ill patients, with cardio-cerebral accident listed as a potential stressor (28, 29). Yet in a recent meta-analysis, Wang et al. (30) suggested that SUP reduced clinically important GIB when compared with a placebo or no prophylaxis in all-cause critical patients with a high risk of developing GIB, but the reduction in bleeding might be unimportant for low-risk patients. Thus, it is reasonable to use SUP according to the risk stratification of GIB in patients with acute myocardial infarction.

The results of the current study provide another consideration for clinicians and medical practitioners to decide the risk of GIB in patients with AMI episode. Patients with higher Killip classes are more likely to develop GIB when other bleeding risk factors are the same.

This study does not consider death as an endpoint because it is focused on the association between the degree of cardiac dysfunction and GIB, which, it is believed, has not been investigated before.

This study has some limitations that should be addressed. Being a retrospective observational study, this study may have inherent shortcomings. It is less convincing than a perspective one and we hypothesized that randomized control clinical trials may be needed to test if prophylactic use of PPI is unnecessary in AMI patients with normal cardiac function and without high-risk bleeding conditions. Second, none of the patient in our study underwent GI endoscopy, thus the exact site of bleeding remained unknown and the endoscopic incidence of SUs in patients in different Killip classes was not investigated.