Introduction

As the global population continues to age, the over-65s and over-80s will be the fastest-growing segments of the population1. As a result, the elderly population will continue to grow and there will be a further increase in the number of elderly patients with acute coronary syndromes. Acute coronary syndrome (ACS) is an acute ischemic syndrome caused by rupture of unstable atherosclerotic plaques or fresh thrombosis secondary to erosion in the coronary artery2. ACS is an important cause of disability and death in patients3. With the rapid development of percutaneous coronary intervention (PCI), PCI has become the mainstay of treatment for patients with ACS4,5.Despite medical advances such as PCI, ACS still has high mortality, and MACE rates6,7. Along with the increasing number of elderly ACS patients treated with PCI, the regression after PCI has also become a widespread concern. Recent evidence suggests that malnutrition is an important factor in the prognosis of cardiovascular (CV) disease8. However, the nutritional status of patients is often overlooked, despite its association with poor outcomes in patients with cardiovascular disease9,10. Nutritional status affects patient regression after surgery, so early identification of patient nutritional status facilitates clinicians to make early clinical decisions and interventions to optimize clinical management to improve patient prognosis37. One type of modeling is prognostic modeling, which centers on assessing the likelihood of potential outcomes such as disease recurrence, mortality, incapacity, or complications manifesting at some point in the future based on the patient's current state of health38. Clinical prediction models can evaluate and classify the risk of patients based on fundamental clinical features and tests and examinations, thus aiding the identification of medium- and high-risk individuals early on. This can help clinicians create sensible management strategies and measures for controlling risk factors for patients outside of the hospital setting. In addition, the study of interaction prediction in various fields of computational biology provides valuable research directions. The study by Sun et al.39 provided a new deep learning algorithm called Graph Convolutional Networks with Graph Attention Networks (GCNAT), which promises to be a useful biomedical research tool for predicting potential metabolite disease associations in the future. It may be more convenient to analyse the nutritional status of ACS patients and allow us to better predict the risk of MACE in elderly ACS patients. Wang et al.40 developed a novel deep learning prediction model called DMFGAM that could become a powerful tool for predicting hERG channel blockers in the early stages of drug discovery and development. Predictive modelling will be widely used in the future to help clinicians make better diagnoses and treatments.

We constructed a column-line diagram of the clinical prediction model for the occurrence of MACE within 1 year after PCI in elderly ACS patients with each of the four dietary assessment tools. The risk of develo** MACE can be suggested more intuitively. According to the calibration curves, the three prediction models constructed from PNI, GNRI, and CONUT had good calibration ability; the decision analysis curves (DCA) suggested that the models constructed from PNI and GNRI had high clinical validity, while the prediction model constructed from BMI had poor clinical validity. The Clinical Impact Curve (CIC) suggests that the PNI and GNRI constructs have a lower rate of misdiagnosis than the CONUT and BMI constructs. The PNI can better predict the risk of develo** MACE, as reflected more intuitively by the predictive model we constructed.

There are many screening tools for malnutrition, but there is no consensus on which screening tool to use in patients with ACS. Based on our results, we suggest using the PNI score, which uses only 2 laboratory values and is very easy to calculate even without a specific automated calculator. Screening for malnutrition in elderly ACS patients undergoing PCI may identify patients at high risk of adverse cardiovascular outcomes who may benefit from targeted secondary prevention programs with supplementation to improve their prognosis.

Limitations of the study

The present study is a single-centre retrospective study with a relatively small number of patients and therefore has some drawbacks. There is no information in this study about patients' economic status, education, adherence, etc., which might help us to understand the causal factors of malnutrition. We did not compare the prognostic value of a nutritional screening tool with a more sophisticated comprehensive nutritional assessment. This is because malnutrition is a complex problem, especially in the elderly, with diverse etiologies and a wide range of determinants. The validity of assessing nutritional status through simple screening tools (PNI, GNRI, CONUT, and BMI) alone remains uncertain because of the lack of comparison with comprehensive nutritional assessments, such as subjective holistic assessments and mini-nutritional assessments. We only assessed nutritional status on admission and did not examine the relationship between changes in nutritional status over time and the incidence of MACE after PCI in elderly ACS patients. The results still need further validation with large samples and multicenter data. We welcome additions and improvements to this study from other researchers and medical centres in different countries.