Maintaining hemodialysis patient death risk prediction model based on plasma amino acid spectrum

Abstract

The invention relates to a maintenance hemodialysis patient death risk prediction model based on a plasma amino acid spectrum and construction and application of the model, and belongs to the technical field of biological medicine. The construction method comprises the following steps: detecting a baseline plasma sample of a maintenance hemodialysis patient through a Bruker nuclear magnetic resonance spectrometer; a cox proportional risk model is adopted to perform correlation analysis on each amino acid, and after important clinical covariables are corrected, it is determined that eight amino acid parameters are significantly related to the total cause death rate; obtaining a p value lt through a cox proportional risk model; the method comprises the following steps of: taking a parameter of 0.05 as a candidate parameter, screening potential biomarkers by adopting a support vector machine algorithm in combination with conventional clinical parameters which are proved to be related to hemodialysis patient prognosis, and constructing and obtaining a prediction model by adopting a logistic regression model according to the screened parameters; through the model provided by the invention, the patient death risk prediction capability is remarkably improved.

Description

Technical Field

[0001] This invention relates to a mortality risk prediction model for maintenance hemodialysis patients based on plasma amino acid profiles, as well as the construction and application of the model, belonging to the field of biomedical technology. Background Technology

[0002] Patients with end-stage renal disease (ESRD) often experience multiple complications and high mortality rates, relying on long-term dialysis and kidney transplantation for survival. Hemodialysis is the primary treatment for ESRD patients in my country, but the long-term survival rate of maintenance hemodialysis patients is significantly lower than that of the general population. Hemodialysis involves the clearance of various small-molecule metabolites, including multiple nutrients and amino acids. Studies have found that plasma branched-chain amino acids and other essential amino acids in maintenance hemodialysis patients are lower than in healthy controls, and further decrease after each hemodialysis session. However, the correlation between various plasma amino acids and the long-term survival rate of maintenance hemodialysis patients has not been reported, and the specific amino acids associated with prognosis remain unclear. Therefore, this research aims to comprehensively assess the plasma levels of various amino acids in maintenance hemodialysis patients using a magnetic resonance metabolomics platform, identify amino acid subclasses and specific amino acids significantly associated with all-cause mortality risk, and establish a predictive model for the long-term survival of maintenance hemodialysis patients. This will provide targeted follow-up treatment plans and improve patient survival rates, possessing significant social and economic value. Summary of the Invention

[0003] One of the technical problems that this invention aims to solve is how to establish a predictive model for the long-term survival of maintenance hemodialysis patients.

[0004] To address the aforementioned problems, the present invention provides a method for constructing a model for predicting mortality risk in maintenance hemodialysis patients based on plasma amino acid profiles. The method includes: detecting baseline plasma samples from maintenance hemodialysis patients using a Bruker nuclear magnetic resonance spectrometer; performing association analysis on each amino acid using a Cox proportional hazards model; identifying eight amino acid parameters significantly associated with all-cause mortality after adjusting for important clinical covariates; using parameters with p-values ​​<0.05 obtained through the Cox proportional hazards model as candidate parameters; combining these with previously confirmed conventional clinical parameters associated with hemodialysis patient prognosis; employing a support vector machine (SVM) algorithm to screen for potential biomarkers; and constructing and obtaining a patient mortality risk prediction model using a logistic regression model based on the screened parameters.

[0005] Preferably, the prediction model formula is:

[0006] y = 0.07×age + 0.41×sex + 0.41×CVD – 0.03×albumin + 0.06×BUN +0.002×creatinine + 0.65×diabetes + 1.09×lactate–29.59×leucine–31.64×valine–14.81×tyrosine + 23.20×BCAA – 2.89

[0007] Where y represents mortality risk, age represents age, sex represents sex, CVD represents baseline history of cardiovascular disease (CVD), albumin represents serum albumin, BUN represents blood urea nitrogen, creatinine represents serum creatinine, diabetes represents baseline history of diabetes, lactate represents blood lactate, leucine represents leucine, valine represents valine, tyrosine represents tyrosine, and BCAA represents branched-chain amino acids.

[0008] This invention provides an application of a model for predicting the mortality risk of maintenance hemodialysis patients based on plasma amino acid profiles. The prediction model formula is as follows:

[0009] y = 0.07×age + 0.41×sex + 0.41×CVD - 0.03×albumin + 0.06×BUN +0.002×creatinine + 0.65×diabetes + 1.09×lactate–29.59×leucine–31.64×valine–14.81×tyrosine + 23.20×BCAA – 2.89; where y is the risk of death, age is age, sex is sex, CVD is the baseline history of CVD, albumin is serum albumin, BUN is blood urea nitrogen, creatinine is serum creatinine, diabetes is the baseline history of diabetes, lactate is blood lactate, leucine is leucine, valine is valine, tyrosine is tyrosine, and BCAA is branched-chain amino acids.

[0010] Preferably, the application includes its use in the preparation of diagnostic kits for predicting the risk of death in maintenance hemodialysis patients after 5 years.

[0011] This invention provides a prognostic system for predicting the risk of death in maintenance hemodialysis patients after 5 years. The system includes a data processing module and a conclusion output module. The data processing module applies a predictive model formula, which is:

[0012] y = 0.07×age + 0.41×sex + 0.41×CVD - 0.03×albumin + 0.06×BUN +0.002×creatinine + 0.65×diabetes + 1.09×lactate–29.59×leucine–31.64×valine–14.81×tyrosine + 23.20×BCAA – 2.89; where y is the mortality risk, age is age, sex is sex, CVD is baseline CVD history, albumin is serum albumin, BUN is blood urea nitrogen, creatinine is serum creatinine, diabetes is baseline diabetes history, lactate is blood lactate, leucine is leucine, valine is valine, tyrosine is tyrosine, and BCAA is branched-chain amino acids; the conclusion output module receives the results obtained by the data processing module and outputs the prognostic result of the mortality risk of the tested person 5 years later.

[0013] Compared with the prior art, the present invention has the following beneficial effects:

[0014] This invention provides a model for assessing the long-term prognosis and mortality risk of maintenance hemodialysis patients, significantly improving the ability to predict patient mortality risk. It has important social and economic value for providing more targeted treatment options for patients and improving their survival rate. Attached Figure Description

[0015] Figure 1 This is a comparison chart of the prediction results of the prediction model of this invention in Chinese patients and British patients;

[0016] Figure 2 The optimal threshold for the prediction model of this invention and its sensitivity and specificity in Chinese and British patients are plotted. Detailed Implementation

[0017] Patients with end-stage renal disease who underwent maintenance hemodialysis at the Hemodialysis Center of Zhongshan Hospital affiliated with Fudan University in December 2015 were selected and followed up until June 2021. Patients under 18 years of age, those who had received maintenance hemodialysis for less than 3 months, those with a history of malignant tumors, or those with confirmed cardiovascular disease were excluded, resulting in a total of 368 patients included in the study. Blood samples were collected at baseline. The cause and date of death of patients were recorded during the follow-up period. Baseline EDTA plasma samples were analyzed using a Bruker nuclear magnetic resonance spectrometer to obtain quantitative data of 11 amino acids and 3 derived indicators, as shown in Table 1.

[0018] Table 1. Types of Amino Acids in Plasma Detected by Bruker Nuclear Magnetic Resonance Spectrometer

[0019] English abbreviations Full English name Chinese name Ala Alanine alanine Gly Glycine glycine His Histidine Histidine Ile Isoleucine Isoleucine Leu Leucine Leucine Phe Phenylalanine Phenylalanine Thr Threonine threonine Tyr Tyrosine Tyrosine Val Valine Valine Glu Glutamate glutamic acid Gln Glutamine glutamine BCAA Branched chain amino acids (Leu+Val+Ile) Branched-chain amino acids AAA Aromatic amino acids (Phe+Tyr+Trp) Aromatic amino acids FisherR Fischer ratio (Leu+Val+Ile) / (Phe+Tyr+Trp) F value

[0020] Note: Tryptophan did not reach the detection limit in this study and is therefore not listed. Derivative parameters related to tryptophan, such as AAA and FisherR, do not include tryptophan.

[0021] The relationship between each amino acid and all-cause mortality risk was assessed using a Cox proportional hazards model, after adjusting for important covariates (including age, sex, BMI, medical history, and routine laboratory indicators related to hemodialysis patient survival). Estimated hazard ratios (HRs) and 95% confidence intervals (CIs) for each standard deviation (per 1-SD) increase in amino acid levels are provided. A p-value < 0.05, combined with an FDR-corrected p-value < 0.2, was considered statistically significant. A total of eight plasma amino acid parameters significantly associated with mortality risk in hemodialysis patients were obtained, as shown in Table 2.

[0022] Table 2 Plasma amino acid parameters significantly associated with all-cause mortality risk in hemodialysis patients

[0023] English abbreviations Full English name Chinese name HR (95% CI) p value p FDR Leu Leucine Leucine 0.78 (0.66-0.94) 0.008 0.065 Phe Phenylalanine Phenylalanine 0.79 (0.66-0.95) 0.011 0.065 Tyr Tyrosine Tyrosine 0.80 (0.67-0.97) 0.019 0.082 Val Valine Valine 0.81 (0.68-0.97) 0.020 0.082 Glu Glutamate glutamic acid 1.34 (1.01-1.76) 0.041 0.162 Gln Glutamine glutamine 1.53 (1.10-2.14) 0.012 0.065 BCAA Branched chain amino acids (Leu+Val+Ile) Branched-chain amino acids 0.82 (0.69-0.98) 0.030 0.105 AAA Aromatic amino acids (Phe+Tyr+Trp) Aromatic amino acids 0.79 (0.66-0.94) 0.009 0.065

[0024] Parameters with p-values ​​<0.05 obtained from the Cox proportional hazards model were used as candidate parameters, and a support vector machine algorithm was used for further feature screening to obtain potential biomarkers. Based on the screened parameters, a logistic regression model was used to construct and obtain a mortality risk prediction model, and the area under the recipient operating characteristic curve (AUC) was calculated. All statistical analyses were performed using R version 4.3.0 and R studio (V1.2.5001).

[0025] Furthermore, a model for predicting the mortality risk of maintenance hemodialysis patients based on plasma amino acid profiles was obtained. This model is an all-cause mortality prediction model. Based on 14 quantifiable amino acid parameters, the four amino acids and eight clinical parameters required to construct the prediction model were obtained by combining the Cox proportional hazards model and support vector machine. The all-cause mortality prediction model was constructed through a logistic regression model, and the model formula is as follows:

[0026] y = 0.07×age + 0.41×sex + 0.41×CVD – 0.03×albumin + 0.06×BUN +0.002×creatinine + 0.65×diabetes + 1.09×lactate–29.59×leucine–31.64×valine–14.81×tyrosine + 23.20×BCAA – 2.89

[0027] Where y represents mortality risk, age represents age, sex represents sex, CVD represents baseline CVD history, albumin represents serum albumin, BUN represents blood urea nitrogen, creatinine represents serum creatinine, diabetes represents baseline diabetes history, lactate represents blood lactate, leucine represents leucine, valine represents valine, tyrosine represents tyrosine, and BCAA represents branched-chain amino acids.

[0028] Amino acid parameters combined with clinical indicators can effectively predict the risk of death in maintenance hemodialysis patients: the AUC of the all-cause mortality prediction model was 0.813. The AUC of the all-cause mortality prediction model including only clinical indicators was 0.787. The inclusion of amino acid parameters significantly improved the predictive ability of the mortality risk prediction model (P=0.034). Figure 1 (Figure A)

[0029] External validation of the model:

[0030] UK Biobank cohorts (e.g.) Figure 1 (Figure B) The corresponding amino acid concentrations were obtained using the high-throughput NMR metabolomics analysis service (Nightingale Health Ltd., Helsinki, Finland). A cohort of 128 end-stage renal disease patients undergoing maintenance hemodialysis was selected from the UK Biobank cohort as an external validation cohort. After data quality control, data on 12 amino acids and their related parameters were found to be common to both the Chinese and UK hemodialysis cohorts.

[0031] The AUC of the above mortality risk prediction model in the UK cohort is 0.798. (e.g.) Figure 1 (Figure B)

[0032] In the Chinese cohort, the mortality risk score had a sensitivity of 84.7% for ruling out mortality risk and a specificity of 91.5% for establishing mortality risk; when the mortality risk score was ≥0.614, the positive predictive value was 79.12% (e.g., Figure 2 (See Figure A in the original text). When this mortality risk score and its threshold were applied to the UK cohort, the sensitivity for excluding mortality risk was 89.5%, the specificity for establishing mortality risk was 88.7%, and the positive predictive value was 77.78%. (As shown in Figure A in the original text). Figure 2 (Figure A)

[0033] During a 5.5-year follow-up, the inventors observed 144 all-cause deaths. First, they performed a correlation analysis on 14 amino acid parameters using a Cox proportional hazards model. After adjusting for important clinical covariates, they identified 8 amino acid parameters significantly associated with all-cause mortality. Second, by using parameters with p-values ​​<0.05 obtained from the Cox proportional hazards model as candidate parameters, and combining them with previously confirmed conventional clinical parameters associated with hemodialysis patient prognosis, they employed a support vector machine algorithm to screen for potential biomarkers. This led to the construction of a predictive model. The model performance was evaluated by calculating the area under the recipient operating characteristic curve (AUC), and the advantages and disadvantages of the amino acid parameter-based model and the conventional clinical indicator model were compared.

[0034] To address the current situation where amino acid metabolism disorders and essential amino acid deficiencies in hemodialysis patients, but whose relationship with patient prognosis remains unclear, this invention, in a long-term follow-up cohort of hemodialysis patients from a single center, used plasma amino acid profiles provided by a Bruker nuclear magnetic resonance spectrometer to identify eight amino acid parameters significantly associated with all-cause mortality. The inclusion of these amino acid parameters significantly improved mortality risk prediction: the AUC of the all-cause mortality prediction model increased from 0.787 to 0.813 (P=0.034).

[0035] This invention utilizes patient plasma amino acid profile data obtained through the Bruker nuclear magnetic resonance spectroscopy metabolomics platform. By incorporating prognostic-related amino acid parameters into an all-cause mortality prediction model, it can predict the patient's mortality risk 5 years later.

[0036] In the attached diagram, Figure 1 The figures show a comparison of the prediction results of the prediction model of the present invention in Chinese patients and British patients; Figure A shows the ROC curve of the prediction model of the present invention in Chinese patients; Figure B shows the ROC curve of the prediction model of the present invention in British patients; Figure C shows the integrated ROC curve of the prediction model of the present invention in Chinese and British patients; Figure D shows a comparison of the statistical results of different models.

[0037] Figure 2 Figure A shows the sensitivity and specificity plots and survival curves of Chinese and British patients corresponding to the optimal threshold of the prediction model of this invention; Figure B shows the survival curves of Chinese patients grouped by the optimal threshold; Figure C shows the survival curves of British patients grouped by the optimal threshold; and Figure D shows the integrated survival curves of Chinese and British patients.

[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any form or substance. It should be noted that those skilled in the art can make various improvements and additions without departing from the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention. Any modifications, alterations, and equivalent changes made by those skilled in the art based on the above-disclosed technical content without departing from the spirit and scope of the present invention are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, and evolutions made to the above embodiments based on the essential technology of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for constructing a model for predicting mortality risk in maintenance hemodialysis patients based on plasma amino acid profiles, characterized in that, The method includes detecting baseline plasma samples from maintenance hemodialysis patients using a Bruker nuclear magnetic resonance spectrometer; using a Cox proportional hazards model to perform association analysis on each amino acid, and after adjusting for important clinical covariates, identifying 8 amino acid parameters that are significantly associated with all-cause mortality; Parameters with p-values ​​<0.05 were obtained through the Cox proportional hazards model as candidate parameters. Combined with previously confirmed routine clinical parameters that are related to the prognosis of hemodialysis patients, a support vector machine algorithm was used to screen potential biomarkers. Based on the screened parameters, a logistic regression model was used to construct and obtain a patient mortality risk prediction model.

2. The construction method according to claim 1, characterized in that, The prediction model is calculated as follows: y = 0.07×age + 0.41×sex + 0.41×CVD - 0.03×albumin + 0.06×BUN + 0.002×creatinine + 0.65×diabetes + 1.09×lactate–29.59×leucine–31.64×valine–14.81×tyrosine + 23.20×BCAA – 2.89; where y represents mortality risk, age represents age, sex represents sex, CVD represents baseline CVD history, albumin represents serum albumin, BUN represents blood urea nitrogen, creatinine represents serum creatinine, diabetes represents baseline diabetes history, lactate represents blood lactate, leucine represents leucine, valine represents valine, tyrosine represents tyrosine, and BCAA represents branched-chain amino acids.

3. An application of a model for predicting mortality risk in maintenance hemodialysis patients based on plasma amino acid profiles, characterized in that, The model formula is as follows: y = 0.07×age + 0.41×sex + 0.41×CVD - 0.03×albumin + 0.06×BUN + 0.002×creatinine + 0.65×diabetes + 1.09×lactate–29.59×leucine–31.64×valine–14.81×tyrosine + 23.20×BCAA – 2.89; where y is the risk of death, age is age, sex is sex, CVD is the baseline history of CVD, albumin is serum albumin, BUN is blood urea nitrogen, creatinine is serum creatinine, diabetes is the baseline history of diabetes, lactate is blood lactate, leucine is leucine, valine is valine, tyrosine is tyrosine, and BCAA is branched-chain amino acids.

4. The application according to claim 3, characterized in that, The applications include the development of diagnostic kits for predicting the risk of death in maintenance hemodialysis patients after 5 years.

5. A prognostic system for predicting the risk of death in maintenance hemodialysis patients after 5 years, characterized in that, The system includes a data processing module and a conclusion output module. The data processing module applies a prediction model formula, which is: y = 0.07×age + 0.41×sex + 0.41×CVD - 0.03×albumin + 0.06×BUN + 0.002×creatinine + 0.65×diabetes + 1.09×lactate–29.59×leucine–31.64×valine–14.81×tyrosine + 23.20×BCAA – 2.89; where y is the mortality risk, age is age, sex is sex, CVD is baseline CVD history, albumin is serum albumin, BUN is blood urea nitrogen, creatinine is serum creatinine, diabetes is baseline diabetes history, lactate is blood lactate, leucine is leucine, valine is valine, tyrosine is tyrosine, and BCAA is branched-chain amino acids; the conclusion output module receives the results obtained by the data processing module and outputs the prognostic result of the mortality risk of the tested person 5 years later.

Images