New model predicts in-hospital complications in myocardial infarction

INTRODUCTION AND OBJECTIVES: Ischemic cardiopathy constitutes the leading cause of death worldwide. Our aim was to evaluate the prognostic capacity of the leukoglycemic index as well as to create a predictive model of in-hospital complications in patients with ST elevation myocardial infarction. MATERIALS AND METHODS: This was a multicentral and cohort study, which included patients inserted in the Cuban Registry of acute myocardial infarction. The study investigated 900 patients with a validation population represented by 233 external subjects. In order to define the performance of the leukoglycemic index were evaluated the discrimination with the statistical C and the calibration by Hosmer – Lemeshow test. A model of logistic binary regression was employed in order to define the predictive factors. RESULTS: Optimal cut point of the leukoglycemic index to predict in-hospital complications was 1188 (sensibility 60%; specificity 61.6%; area under the curve 0.623; p < 0.001). In-hospital complications were significantly higher in the group with the leukoglycemic index ≥ 1188; a higher value was significantly associated with a higher risk to develop an in-hospital complication [RR (IC 95%) = 2.4 (1.804–3.080); p<0.001]. The predictive model proposed is composed by the following factors: age ≥ 66 years, leukoglycemic index ≥ 1188, Killip-Kimball classification ≥ II and medical history of hypertension. This scale had a good discrimination in both, the training and the validation population. CONCLUSION: The leukoglycemic index possesses a low performance when used to assess the risks for in hospital complications in patients with ST elevation myocardial infarction. The new predictive model has a good performance, which can be applied to estimate risk of in-hospital complications. This model would be able to contribute to the health systems of developing countries without additional cost; it enables prediction of the patients having a higher risk of complications and a negative outcome during the hospitable admission.


INTRODUCTION
Ischemic cardiopathy constitutes the leading cause of death in the world. According to the World Health Organization, from the year 2000 the deceases for this disease have progressively increased. In 2016, the disease claimed over nine million lives 1 . Approximately seven million people die every year as a consequence of this disease in worldwide, that corresponds to 12.8% of all deaths 2 .
In Cuba, heart ischemic diseases reached a death rate in the year 2020 of 165.8 deceases for each 100 000 inhabitants. Within them, the myocardial infarction occupies a preponderant place, with a dead rate of 69.7 for 100 000 inhabitants, higher than the numbers of the preceding year 3 .
The incidence of in-hospital complications has diminished strongly after ST segment elevation myocardial infarction (STEMI) because of percutaneous coronary intervention (PCI) and medical treatment. However, this appears to obscure the patient's prognosis.
Recently, research on the use of new cardiovascular markers has increased, with the objective to help to identify and stratify patients with greater risk and worse prognosis after suffering STEMI. However, inaccessibility to most of these markers in Cuba, limit its usefulness in habitual clinical practice.
In several studies, it has been shown that the leukocyte and blood glucose count are related to the genesis and progression of atherosclerotic disease as well as with the extension and complications of STEMI [4][5][6][7][8][9] . Quiroga and al. were the first investigators in propose the leukoglycemic index (LGI) as a prognostic marker of death and inhospital complications in patients with STEMI 10 .
We hypothesized that the LGI, measured in the first 24 hours of hospital admission, is a valid predictor of in-hospital complications in patients with STEMI. The purpose of the study was to evaluate the LGI's predictive capability and to create a predictive model.

Type of study and participants
An analytical, observational, multicenter, longitudinal and prospective cohort study was carried out, with data stored in the Cuban Registry of Acute Myocardial Infarction (RECUIMA) 11,12 . In this cohort 900 patients diagnosed with STEMI from January 2018 to the January 2020 were included with a consecutive form. The criteria of inclusion were: 1) diagnosis of STEMI, according to the 4th Universal Definition of Myocardial Infarction 13 ; 2) venous blood samples were obtained in the first 24 hours after admitted. The sample was used to estimate white blood count (WBC) and fast plasma glucose (FPG); and 3) absence of inflammatory systemic, infectious or hematological known disease. As part of validation population 142 patients were recruited from January to July 2021. These patients satisfied the same criteria of inclusion previously exposed.
Several clinical and epidemiological variables were extracted from the database: age, sex, medical history (diabetes mellitus, hypertension, dyslipidemia, smoking, prior myocardial infarction and prior chronic kidney disease), and the Killip Kimball functional class at admission. Other variables: reperfusion therapy (fibrinolysis or PCI), hospital stay, in-hospital complications and discharge state (alive or dead), as well as several laboratory exams (creatinine, triglycerides, FPG and WBC) were registered. The investigation was in accordance with the ethical principles set out in the Declaration of Helsinki and was approved by the Ethical Committee in Enrique Cabrera Teaching General Hospital.

Statistical analysis
The information was processed by the statistical packet IBM SPSS version 19.0.0. Continuous variables were presented as mean ± standard deviation and differences between the two groups were examined by independent-sample t-test. Categorical variables were described as counts (percentages) and compared by Pearson chi-square test (Pearson χ 2 test) or Fisher's exact test appropriately. A p value<0.05 with the 95% confidence interval was regarded as statistically significant. The performance of LGI was determined by the discrimination and the calibration: the discrimination was evaluated by the statistical C, also known as area under curve ROC; the method utilized for calibration was the Hosmer-Lemeshow test. The Youden index was utilized to determine the optimal cutoff point value of LGI for predicting primary endpoint. An analysis of logistic multivariate regression was accomplished to determine if the ILG was an independent predictor of complications, as well as identify others. Initial variables with good correlation in univariate analysis with the final outcome, were included in the multivariate model. After selecting the model of logistic binary regression, a predictive scale was created according to coefficients of every predictor. The new predictive model created in the universe of study was tried in a population of validation.

Baseline characteristic of study population
A total of 900 patients were included in the universe of study, with a mean (SD) age of 64 (12) years. Clinical characteristics and laboratory results of the total population and groups stratified by the occurrence of primary endpoint event were presented in Table 1.
LGI was significantly higher in patients with primary endpoint compared those without. The complicated patients showed higher age, higher prevalence of hypertension, diabetes mellitus, CKD, and a Killip-Kimbal functional classification ≥ II at admission, but lower prevalence of smoking. In terms of laboratory indicators, participants with the endpoint event had higher levels of creatinine, FPG and WBC. It was observed significant differences among both groups of patients when taking into account mortality and hospital stay. ROC curved analysis of LGI as predictor of inhospital complications showed a poor calibration according to Hosmer-Lemeshow test (χ 2 =21.208; p=0.006); its discriminating capability also was

Risk factors for in-hospital complications
Univariate and multivariate analysis results and predictors for in-hospital complications are presented in Table 3. Univariate analysis revealed that LGI≥1188, age, hypertension, diabetes mellitus, smoking, Killip-Kimball ≥ II and creatinine were risk factors of in-hospital complications in patients with STEMI. After adjusting potentials confounding factors, the multivariate analysis found that the LGI≥1188, age, Killip-Kimball functional classification ≥ II and hypertension were independent predictors of primary event. This model had a good calibration (χ 2 =8.964; p=0.35) and its AUC for predicting the occurrence of complications was 0.808 (95% CI 0.778-0.837; p<0.001), which is considered very accurate (Figure 1).

Development of a new predictive model of inhospital complications
The coefficients of logistic regression obtained were transformed to create a new predictive scale, called "Predictive Scale of Infarction Complications", abbreviated with the acronym PSIC (Table 4). This scale is integrated by four components: age≥66 years, LGI≥1188, Killip-Kimball classification ≥ II and a previous history of hypertension. The sum of the punctuation of this predictive scale in each patient varies from 0 to 6 points. With the new predictive scale, categories of risk were prepared according to the sum of these variables. Categories of risk were divided into lower risk (0-2 points), moderate risk (3-4 points) and higher risk (5-6 points) of in-hospital complications. The relation of these categories with the occurrence of complications in the universe of study is showed in the Table 5; it proved to be statistically significant (χ 2 =261.882; p<0.001).

Validation of the model "Predictive Scale of Infarction Complications"
The lethality in the validation population was 1.3%, and 63 patients had in-hospital complication (27%). The performance of the PSIC new model had a good calibration (Hosmer Lemeshow test; χ 2 =12.67; p=0.12) and a good discrimination (AUC 0.70; 95% CI 0.619-0.768; p<0.001).
Once more the validation population was divided according to the categories of risk predetermined for the PSIC model, and it was related with the occurrence of complications. This association proved to be again significant (χ 2 =28.19; p<0.001), the same result observed in the study population (Table 6).

DISCUSSION
In the present investigation, values of LGI were statically higher in the patients with in-hospital complications compared with those without. However, when performance was calculated, its discriminative capability was hindered.
Quiroga et al. were the first investigators who proposed the LGI as a scoreboard of fatal prognosis and in-hospital complications in patients with STEMI 10 . In the study, they correlated LGI with fatal development, heart failure and angina (p<0.001); the value obtained by AUC was 1600. There were several elements that do not permit    17 . The authors evaluated LGI as prognosis marker in 128 patients that suffered STEMI and were admitted between January 2009 and October 2010 in Celestino Hernández Robau Hospital. Major adverse cardiovascular events (MACE) were defined as Killip-Kimball ≥ II, supraventricular and ventricular arrhythmias, conduction disturbance and reinfarction or postinfarction angina. AUC obtained was 0.682 (95% CI 0.590-0.775; p<0.001), with an optimal cutoff t of 1158; this area is considered a poor discrimination, same it was described in our investigation, although we obtained a lightly higher cutoff point. When the authors performed a logistic regression multivariate analysis, they demonstrated that age≥75 years, systolic blood pressure<100 mm Hg and LGI≥1158 were independent fatal predictors of in-hospital complications or death, or both.
The present investigation aims to improve the prognosis of patients who suffer STEMI. The objective evaluation of the gravity of the disease through the punctuation system of the scale PSIC permits to identify patients with a high risk, who can benefit by invasive coronary therapeutic.

CONCLUSIONS
The LGI has a low predictive capacity for evaluating the risk of in-hospital complications of acute myocardial infarction. The new model "Predictive Scale of Infarction Complications" has a good performance which allows its application for the estimation of risk. The new model would be able to contribute the systems of health of developing countries of a clinical instrument without cost additional; it will permit predicting which ones the patients have bigger risk of an adverse outcome.