Marker, model for diagnosing primary aldosteronism and application thereof

Abstract

The application provides a marker for diagnosing primary aldosteronism, a model and application thereof, and combines a univariate analysis and a logistic regression analysis method, analyzes primary aldosteronism samples and primary hypertension samples in a national multicenter cohort, screens out a marker that can be used for predicting whether an individual has primary aldosteronism, and further utilizes a decision tree algorithm to construct a prediction model, realizes intelligent, rapid and simple prediction and diagnosis of primary aldosteronism based on LC-MS / MS technology and the prediction model, and meets clinical requirements.

Description

Technical Field

[0001] This invention belongs to the field of diagnostic technology for primary aldosteronism, specifically relating to biomarkers, models, and applications for diagnosing primary aldosteronism. Background Technology

[0002] Primary aldosteronism (PA) is a common cause of secondary hypertension, and its diagnosis plays a crucial role in hypertension management. Currently, the diagnosis of PA mainly relies on the aldosterone-to-renin activity ratio (ARR) as an initial screening indicator, followed by more complex confirmatory tests, such as a saline loading test or an oral captopril test, to verify the screening results. While this diagnostic process can provide a basis for the definitive diagnosis of PA to some extent, its overall diagnostic efficiency and accuracy still need improvement.

[0003] Structurally, existing technologies rely on multiple tests and complex procedures, which not only increase the economic and physical burden on patients but may also prolong the diagnostic cycle. The cumbersome diagnostic process further burdens the healthcare system. Current complex diagnostic methods and procedures are not suitable for primary aldosteronism, which appears to be becoming increasingly prevalent and continues to expand its disease spectrum.

[0004] Functionally, the diagnostic accuracy of ARR screening methods is affected by various factors, such as sample collection time, choice of testing method, and individual differences, which may lead to missed or misdiagnosed cases in some patients with primary aldosteronism. Furthermore, the criteria for judging whether an ARR is too high are not uniform, because the manifestations of this disease are a broad continuum: they can range from very mild preclinical stages to very severe conditions. Simply put, the severity of primary aldosteronism varies greatly, making it difficult to establish a unified diagnostic standard. The article "Performance of the Aldosterone to Renin Ratio as a Screening Test for Primary Aldosteronism," published in the *Journal of Clinical Endocrinology & Metabolism*, further confirms the diagnostic efficacy and limitations of ARR screening methods through a systematic review and meta-analysis. The sensitivity (ability to identify the disease) and specificity (ability to exclude the disease) of ARR vary greatly among different studies, ranging from 10% to 100% sensitivity and 70% to 100% specificity. Notably, three out of ten studies reported ARR sensitivity of <50%, indicating that ARR has limited ability to adequately identify patients with primary aldosteronism. Furthermore, meta-regression analysis revealed that the accuracy of ARR may also be affected by age and sex.

[0005] In summary, existing technologies have significant shortcomings in PA diagnosis, and a more accurate and convenient diagnostic method is urgently needed to address these issues. Summary of the Invention

[0006] To address the problems existing in current technologies, this invention provides biomarkers for predicting whether an individual has primary aldosteronism (PA) and their applications. Utilizing liquid chromatography-tandem mass spectrometry (LC-MS / MS) combined with a decision tree algorithm, biomarkers suitable for predicting PA are screened, and a multi-biomarker joint detection model is further constructed to achieve rapid and accurate diagnosis of PA. Employing this biomarker combination and the constructed detection model simplifies the diagnostic process, improves diagnostic efficiency and accuracy, and provides strong support for the early identification and effective treatment of patients with primary PA.

[0007] To achieve the above objectives, the present invention employs the following technical solution:

[0008] On the one hand, the present invention provides a model for predicting primary aldosteronism, the model being based on aldosterone concentration, renin activity, ARR value, age, sex, and K. + Constructed by combination of any one or more of the following markers: concentration, uric acid level, 18-hydroxycorticosteroid (18-OHF), and 18-hydroxycorticosterone (18-OHB).

[0009] The aldosterone (ALD) to renin activity (PRA) ratio (ARR) is currently the most commonly used screening indicator for acute pancreatitis (PA). However, because ALD and PRA are affected by various factors, relying solely on ARR values ​​for PA screening leads to high false positive and false negative results. For example, PRA and ALD levels are significantly affected by body position; renin activity is higher when lying down, while aldosterone levels are higher when standing. Sampling in different positions may result in inaccurate ARR calculations. Furthermore, some patients may have physiological variations in renin levels, causing ARR to fail to reach the diagnostic threshold, thus leading to missed diagnoses of PA. Therefore, to address the shortcomings of ARR screening technology, this invention proposes a combination of multiple biomarkers, including ARR values, which can accurately distinguish PA patients from healthy individuals, improving diagnostic performance and reducing the risk of misdiagnosis and missed diagnoses.

[0010] Further, the biomarkers include a combination of aldosterone concentration, renin activity, ARR value, and uric acid level, or a combination of aldosterone concentration, renin activity, ARR value, and sex, or a combination of aldosterone concentration, renin activity, ARR value, and K+. +A combination of concentrations, or a combination including aldosterone concentration, renin activity, ARR value, sex, and age, or a combination including aldosterone concentration, renin activity, ARR value, sex, and K. + A combination of concentrations, or including aldosterone concentration, renin activity, ARR value, age, sex, K + A combination of concentration and uric acid levels, or including aldosterone concentration, renin activity, ARR value, age, sex, K... + Concentration, a combination of 18-hydroxycorticosteroids and 18-hydroxycorticosterone.

[0011] Furthermore, the biomarkers include aldosterone concentration, renin activity, ARR value, sex, and K. + The combination of concentrations, the prediction of primary aldosteronism, is to determine whether a patient with primary hypertension has primary aldosteronism.

[0012] According to literature reports, the renin-angiotensin-aldosterone system (RAAS hormones) exhibits significant differences across genders and age groups. In women, RAAS hormone expression patterns are regulated by the menstrual cycle; however, when an individual is over 65 years old, renin levels decrease more significantly than aldosterone, a change that directly leads to an increase in the anorexic response rate (ARR). Furthermore, aldosterone plays multiple physiological roles. Firstly, it regulates vascular smooth muscle tone, affects the vascular response to vasoconstriction signals and arterial wall structure, and enhances vascular sensitivity to norepinephrine, thereby causing an increase in blood pressure. In addition, aldosterone also promotes the release of sodium... + The reabsorption of sodium and water leads to water and sodium retention and accelerates the absorption of potassium. + The excretion of potassium can lower a patient's blood potassium level, potentially leading to hypokalemia in severe cases. Therefore, this invention takes all of the above variables into consideration.

[0013] First, univariate analysis was used to initially screen for gender, age, and K in PA patients and non-PA individuals. + Several biomarkers showed significant differences, including concentration, uric acid level, ARR, PRA, ALD, 18-hydroxycorticosteroids (18-OHF), and 18-hydroxycorticosterone (18-OHB). However, since univariate analysis can only reveal the differences between groups for each indicator and cannot directly assess the comprehensive value of these indicators in predicting PA, this invention further employs logistic regression analysis to determine the diagnostic value of each variable through model coefficients and significance tests. Ultimately, ALD, PRA, ARR values, gender, and K were selected as biomarkers. +Five indicators with concentrations significantly correlated with PA diagnosis were analyzed. Data from multicenter samples showed that the model constructed using these five biomarkers achieved an AUC of 0.929, a sensitivity of 0.92, a specificity of 0.94, and an accuracy of 0.93. This represents a 114.67% improvement in sensitivity and a 13.73% improvement in accuracy compared to existing PA diagnostic tests. Furthermore, the testing process is simpler and more efficient, making it more promising for clinical application.

[0014] In some embodiments, the present invention combines the selected biomarkers in different ways and constructs prediction models using decision tree algorithms. The diagnostic performance of different decision tree models is evaluated. Results show that using a single biomarker to discriminate PA has lower diagnostic performance than models combining multiple biomarkers. Furthermore, models with more biomarkers do not necessarily have better diagnostic performance. The diagnostic performance of models constructed using 7, 8, or all candidate biomarkers is correlated with ALD, PRA, ARR values, gender, and K. + The combined model of the five biomarkers showed similarities, but the former was generally less effective than the latter. Therefore, this invention preferentially uses five biomarkers: aldosterone concentration (ALD), renin activity (PRA), ARR, sex, and K. + Concentration-based models can provide accurate diagnostic results in the initial screening stage, simplifying the diagnostic process, reducing the need for confirmatory tests, lowering medical costs, improving patient compliance, and enhancing the efficiency of clinical decision-making. They significantly improve the accuracy and efficiency of PA diagnosis, providing a powerful auxiliary tool for clinical practice.

[0015] In some methods, LC-MS / MS technology is used to accurately measure the levels of biomarkers in patient plasma. LC-MS / MS technology is characterized by high sensitivity, high specificity, and high accuracy. Compared with traditional detection methods such as chemiluminescence, it can more accurately measure the levels of hormones related to the RAAS system, reduce detection errors, and improve diagnostic accuracy.

[0016] Furthermore, the model is constructed using machine learning algorithms, including any one or more of random forest, decision tree, support vector machine, and logistic regression.

[0017] In some embodiments, the present invention compares parameters based on aldosterone concentration, renin activity, ARR, sex, and K. + For the five biomarkers of concentration, four supervised learning algorithms—decision tree, random forest, support vector machine, and logistic regression—were used to construct diagnostic models. Experimental results showed that the diagnostic model constructed by the decision tree algorithm had the best performance score, with a sensitivity of 0.92, a specificity of 0.94, an accuracy of 0.93, and an AUC of 0.929.

[0018] Preferably, the machine learning algorithm is a decision tree algorithm.

[0019] Decision trees are supervised learning algorithms that make decisions based on a tree structure. They construct a series of decision rules by learning the relationship between features in a dataset and the target variable. It divides the data into different categories, with each node representing a judgment condition for a feature, each branch representing the output result of that condition, and the final leaf node representing the final classification result. This invention uses R language (version 4.3.1) as the development tool and utilizes the rpart function in the rpart package to construct the decision tree model. In the diagnosis of PA, the decision tree model provided by this invention is based on multi-center data from across the country on PA screening, including the five screening indicators mentioned above. Through the analysis of clinical data, decision rules are constructed, and through training and optimization, the accuracy and reliability of the diagnostic results are ensured. In subsequent applications, simply inputting the integrated data into the model is sufficient to determine whether a patient has PA. Traditional diagnostic processes require multiple steps, including screening, diagnosis, and subtyping. This invention, by introducing a decision tree algorithm and combining gender and K... + The concentration-based diagnostic model provides doctors with a powerful auxiliary diagnostic tool, enabling rapid and accurate identification of disease states and reducing the risk of misdiagnosis and missed diagnosis. Furthermore, PA can be diagnosed with a single-point sample, eliminating the need for subsequent complex confirmatory testing procedures, significantly shortening the diagnostic cycle and improving hospital efficiency.

[0020] In some implementations, the diagnostic model for primary aldosteronism constructed using the decision tree algorithm is trained using a 10-fold cross-validation method, and the optimal hyperparameters of the model are determined to be: minimum number of split nodes (minsplit) of 20, complexity parameter (cp) of 0.01, maximum number of competing splits (maxcompete) of 4, and maximum depth (maxdepth) of 30.

[0021] In some implementations, based on aldosterone concentration, renin activity, ARR, sex, and K... + The decision tree model constructed based on concentration has the following discrimination conditions when applied: Table 5.

[0022] On the other hand, the present invention provides the use of a combination of biomarkers for preparing a reagent to predict whether an individual has primary aldosteronism, the biomarker combination including aldosterone concentration (ALD), renin activity (PRA), ARR value, age, sex, and K. + concentration.

[0023] Furthermore, the reagent is used to detect aldosterone concentration, renin activity, and K+ in individual samples. + Concentration; the ARR is calculated based on aldosterone concentration and renin activity; gender is determined based on the individual.

[0024] The samples include biomarkers in blood, urine, saliva, or sweat samples; the reagents are used to detect the presence, relative abundance, or concentration of biomarkers in the samples.

[0025] In some embodiments, the reagent for predicting whether a person has primary aldosteronism is a detection reagent prepared with the above-mentioned biomarker as the detection target, such as sample pretreatment reagents, antigens or antibodies, and other biological reagents and kits suitable for the detection of the biomarker; it can also be developed into a standardized reagent or kit suitable for the detection of the biomarker by liquid chromatography-ultraviolet (LC-UV) or liquid chromatography-mass spectrometry (LC-MS).

[0026] In another aspect, the present invention provides a kit for predicting whether an individual has primary aldosteronism, the kit comprising a detection reagent for a biomarker used as described above.

[0027] Furthermore, this invention provides a combination of biomarkers for predicting whether an individual has primary aldosteronism, said biomarkers including aldosterone concentration, renin activity, ARR value, sex, and K+. + Combinations of concentrations.

[0028] Furthermore, this invention provides a system for predicting whether an individual has primary aldosteronism, the system including a data analysis module; the data analysis module is used to analyze the detection values ​​of biomarkers, including aldosterone concentration, renin activity, ARR value, age, sex, and K+. + Any one or more combinations of concentration and uric acid level.

[0029] Furthermore, the biomarkers include aldosterone concentration, renin activity, ARR value, sex, and K. + Combinations of concentrations.

[0030] Furthermore, the data analysis module constructs a decision tree model using the detection values ​​of biomarkers to calculate the probability value of whether an individual has primary aldosteronism.

[0031] The method for in vitro detection of essential aldosteronism using the biomarker combination provided by the present invention includes the following steps:

[0032] (1) Blood sample collection and pretreatment;

[0033] (2) The concentration of aldosterone and renin activity in the sample after pretreatment in step (1) were determined by LC-MS / MS, and the ARR value was calculated.

[0034] (3) Data integration and input: The detection results of step (2) are integrated with gender, K +Concentration input decision tree model (PA-SingleDiagnosis model, PA-SDx);

[0035] (4) Model analysis and output: The model analyzes the data and outputs the diagnostic results.

[0036] This invention proposes an innovative in vitro diagnostic method that integrates LC-MS / MS technology with a decision tree diagnostic model, specifically designed for the rapid and accurate diagnosis of primary aldosteronism (PA) in hypertensive patients. First, LC-MS / MS technology is used to accurately measure the levels of RAAS system hormones in the patient's body, including ALD, PRA, and the calculated ARR ratio. Simultaneously, it integrates two clinically common and easily obtainable indicators: gender and potassium ion concentration. Then, a decision tree algorithm is used to construct a predictive model, achieving intelligent, rapid, and simplified diagnosis of primary aldosteronism.

[0037] The implementation of this invention significantly improves the accuracy and efficiency of PA diagnosis. PA diagnosis is typically a multi-step process. While initial screening with ARR tests is easily performed in outpatient settings, subsequent diagnostic procedures are cumbersome, potentially harmful, and costly. Using the decision tree model constructed from the biomarker combination provided by this invention, the diagnostic accuracy is improved by 9.23% compared to current screening techniques; compared to previous screening processes involving one or more confirmatory tests, the diagnostic cycle is significantly shortened. Furthermore, compared to CCT confirmatory tests, there are approximately half fewer false negatives, and it is expected to reduce the need for follow-up visits by about 50%, further conserving subsequent clinical resources.

[0038] The present invention has the following beneficial effects:

[0039] 1. This invention screened five biomarkers that can predict primary aldosteronism: aldosterone concentration (ALD), renin activity (PRA), ARR, sex, and K. + Concentration, and developed a new combination of predictive biomarkers.

[0040] 2. In this invention, aldosterone concentration (ALD) and renin activity (PRA) are detected based on LC-MS / MS technology, and the level of hormones in the RAAS system is accurately measured. A decision tree model is constructed using five screened biomarkers, which enables rapid diagnosis of primary aldosteronism, improves the accuracy of diagnostic results, and significantly shortens the diagnostic cycle.

[0041] 3. The decision tree model constructed by the biomarker combination provided by this invention not only improves the accuracy and efficiency of predicting primary aldosteronism, but also significantly saves clinical resources, reduces unnecessary examinations and treatments, and lowers medical costs; at the same time, by reducing the number of false negative patients, it reduces the need for subsequent second visits, further saving clinical resources. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the technical solution of the present invention.

[0043] Figure 2 The results show a comparison of ARR levels between women and men.

[0044] Figure 3 This is a PA prediction model (PA-SDx) built based on the decision tree algorithm. Detailed Implementation

[0045] To make the above-mentioned objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0046] Unless otherwise specified, all raw materials used in the examples are commercially available.

[0047] Example 1: Screening of biomarkers for predicting primary aldosteronism

[0048] This embodiment employs a prospective, multicenter cohort study approach, collecting blood samples from patients with essential aldosteronism and healthy controls. Different samples were analyzed using liquid chromatography-tandem mass spectrometry (LC-MS / MS). Based on logistic regression analysis, indicators showing significant differences between patients with essential aldosteronism and healthy controls were first screened. Five parameters with a clear association with early-stage essential aldosteronism were ultimately selected. Furthermore, a decision tree algorithm was used to construct a multi-biomarker joint detection model based on these five biomarkers. All models demonstrated good risk prediction capabilities in essential aldosteronism. The diagnostic efficacy of different models was evaluated through sensitivity, specificity, accuracy, and AUC analysis. Ultimately, the model constructed using the five biomarkers of this invention showed the highest diagnostic efficacy and can be used for efficient differentiation and diagnosis of essential aldosteronism and healthy individuals.

[0049] The specific steps are as follows:

[0050] I. Sample Collection

[0051] This study established a nationwide multicenter cohort by continuously sampling at Fuwai Hospital of the Chinese Academy of Medical Sciences, the First Affiliated Hospital of Chongqing Medical University, the General Hospital of the Chinese People's Liberation Army, the General Hospital of Tianjin Medical University, Fuwai Hospital Shenzhen of the Chinese Academy of Medical Sciences, the First Affiliated Hospital of Xi'an Jiaotong University, and Zhejiang Provincial People's Hospital.

[0052] Plasma samples were collected from patients with acute aldosteronism (PA) and essential hypertension (standing / lying position). 117 samples were collected from patients with essential aldosteronism, and 354 samples from patients with essential hypertension. Sample collection requirement: ≥2 mL of EDTA-anticoagulated plasma.

[0053] Requirements for standing blood collection: After getting up in the morning, maintain a non-lying position (you can sit, stand or walk) for at least 2 hours, sit quietly for 5-15 minutes, and then collect blood at 8-10 am.

[0054] Requirements for blood collection in supine position: Subjects should lie in bed overnight or lie in a supine position for 1.5-2 hours before blood collection.

[0055] The collected whole blood was stored at 2-8°C and the plasma was separated by centrifugation within 8 hours. The plasma was then stored at -80°C for testing.

[0056] II. Hormone Determination and Data Acquisition by LC-MS / MS

[0057] The collected plasma samples were processed and metabolites were identified using LC-MS / MS. Peak areas were obtained by integrating the chromatographic peaks of the metabolites, and data normalization and missing value imputation were performed. The resulting data were used for subsequent analysis, including univariate analysis and logistic regression analysis, to screen and rank the most effective differentially expressed metabolites between primary aldosteronism and essential hypertension samples. The sample processing and detection methods referenced the magnetic bead extraction method and LC-MS / MS detection in the invention patent CN114895043B, filed by our research team on July 13, 2022, entitled "Primary Aldosteronism Screening Kit and Diagnostic and Subtyping System." The specific steps are as follows:

[0058] (1) Take 400 μL of the sample to be tested, add 100 μL of inhibitor in sequence, vortex mix well, and incubate in a 37℃ oven for 3 hours;

[0059] (2) After incubation, add 400 μL of sample diluent, vortex to mix and stop incubation;

[0060] (3) Once the magnetic bead liquid has been fully vortexed into a suspension state, subsequent operations can be carried out;

[0061] (4) Take a 96-well plate and add 200 μL of magnetic bead suspension, 400 μL of EHK-equilibration solution, 900 μL of sample solution after step (2) treatment, 400 μL of EHK-elution buffer 1, 400 μL of EHK-elution buffer 2 and 130 μL of EHK-elution buffer in sequence.

[0062] Extraction was performed using a magnetic bead extractor. After extraction, the pretreatment plate was removed and centrifuged at 4000 rpm for 5 min. 100 μL of the eluted supernatant was transferred to a 96-well plate and detected and collected using liquid chromatography-tandem mass spectrometry.

[0063] III. Screening of predictive biomarkers for primary aldosteronism

[0064] (1) Screening differentially expressed metabolites using univariate analysis

[0065] Literature reports that RAAS hormones (renin-angiotensin-aldosterone system hormones) differ between sexes. The expression pattern of RAAS hormones in women is heavily regulated by the menstrual cycle. Specifically, during the premenstrual and ovulatory phases, the activity of aldosterone synthase reductase in women shows a significantly higher trend compared to men of the same age. Therefore, this invention uses a difference test method to further verify this result. Figure 2 As shown, the overall ARR level in the female population was significantly higher than that in the male population (P<0.05), indicating a significant difference in RAAS hormone levels between genders. Gender may be a key variable influencing the diagnosis of primary aldosteronism (PA). Furthermore, according to existing literature, when an individual is over 65 years old, renin levels decrease more significantly than aldosterone levels, a change that directly leads to an increase in ARR. In addition, aldosterone also promotes Na+... + The reabsorption of sodium and water leads to water and sodium retention and accelerates the absorption of potassium. + . excretion.

[0066] Therefore, during the screening process, we will consider the above indicators: age, gender, and Na. + Concentration, K + Variables such as concentration, blood pressure level, and RAAS hormones (including ARR) were all considered. To screen for metabolites with significant differences from the candidate variables, we used a nonparametric statistical method—the Mann-Whitney U test. This method is suitable for data that do not meet the normality assumption and can effectively assess the differences between two groups of samples. The results are shown in Table 1, where median[Q1-Q3] represents the median and the tertiary interval of the data.

[0067] The hormone levels of ALD, 18-OHF, and 18-OHB were measured using the LC-MS / MS method described above. Other ion concentrations, such as K+, were also measured. + Na +The results, such as height and weight, were obtained through the electrode method and obtained through routine clinical testing methods. ARR = plasma aldosterone concentration (ALD) / plasma renin activity (PRA). The renin activity was calculated by detecting the angiotensin I concentration in the samples before and after incubation, according to the formula: Angiotensin I production rate = (angiotensin I concentration after incubation - angiotensin I concentration before incubation) / incubation time. The angiotensin I concentration before incubation was very low and could be ignored.

[0068] Table 1. Results of candidate variable difference tests based on Mann-Whitney U test or chi-square test

[0069]

[0070] Based on the p-value results, gender, age, uric acid level, and K... + The p-values ​​for concentration, PRA, ALD, ARR, 18-OHF, and 18-OHB were less than 0.05, therefore these indicators were considered as potential candidate biomarkers. This process ensured that the selected variables had statistically significant differences between groups, thus providing a reliable basis for further multivariate analysis and model construction.

[0071] (2) Further screening of biomarkers with diagnostic value through machine learning algorithms.

[0072] Univariate analysis can only reveal the differences in each indicator between groups, but it cannot directly assess the comprehensive value of these indicators in predicting PA. To further identify key indicators with significant value in PA diagnosis, we adopted a machine learning algorithm—logistic regression analysis. Logistic regression is a statistical analysis method widely used in classification problems. It can effectively evaluate the joint effect of multiple independent variables on the dependent variable (i.e., the PA diagnosis result) and determine the diagnostic value of each variable through model coefficients and significance tests.

[0073] In the analysis, we included the PA diagnosis result as the dependent variable and the candidate biomarkers as independent variables in the model. The dependent variable, gender, showed a strong correlation with height, weight, and BMI. Multicollinearity may exist among these indicators, but no differences were shown in the U-test results. Considering that multicollinearity could interfere with the stability of the model and the interpretability of the results, we decided to remove height, weight, and BMI to avoid the impact of multicollinearity on the model. Finally, by fitting a logistic regression model, we evaluated the contribution of each independent variable to the primary aldosteronism diagnosis result and calculated the corresponding regression coefficients, standard errors, z-values, and p-values. We also analyzed the sensitivity, specificity, and AUC values ​​of each independent variable as a biomarker. The results are shown in Tables 2 and 3.

[0074] Table 2. Candidate variable detection results based on logistic regression analysis

[0075]

[0076] Table 3. Results of ROC analysis for single biomarkers

[0077]

[0078] The correlation between changes in the 17 biomarkers and the presence or absence of primary aldosteronism can be distinguished by the P-value in Table 2, or by the AUC value, specificity, and sensitivity in Table 3. Among these, the P-value and AUC value are the most intuitive and obvious. The smaller the P-value, the greater the influence of the indicator on the individual with primary aldosteronism compared to the individual with primary hypertension; the higher the AUC value, the more accurately the biomarker can distinguish between individuals with primary aldosteronism and those with primary hypertension.

[0079] Based on the results in Table 2, gender, K + Concentration, ALD, PRA, and ARR were confirmed as key factors in identifying primary aldehydes. The p-values ​​for all these indicators in the model were less than the pre-specified significance level (P<0.05), indicating that they have significant statistical value and diagnostic value in primary aldehyde diagnosis. Among them, PRA, ARR, and ALD had the smallest p-values, all <0.001.

[0080] As shown in Table 3, based on the presence or quantitative results of a biomarker used to distinguish between PA and non-PA populations, ARR had the highest AUC value (0.82), followed by PRA, ALD, and K. + The concentrations showed AUC values ​​of 0.737, 0.731, and 0.683. While other indicators such as 18-OHB and UA also had high AUC values, the results in Table 2 indicate that these indicators had no significant impact on the combined use of other biomarkers to identify primary aldehydes; therefore, they were not included as candidate biomarkers. In conclusion, gender and K are the preferred biomarkers. + Five indicators—concentration, ALD, PRA, and ARR—were selected as biomarkers for primary aldehyde diagnosis. The selection process for these five biomarkers was based not only on rigorous statistical analysis but also on relevant knowledge of physiology and clinical medicine, ensuring the scientific validity and practicality of the selected indicators.

[0081] Example 2: Model Construction and Evaluation

[0082] While a single biomarker can distinguish between blood samples with primary aldosteronism and those without, or to predict primary aldosteronism, generally combining multiple biomarkers to construct a predictive model results in higher accuracy in distinguishing or predicting the condition.

[0083] However, a single biomarker that is more accurate in predicting primary aldosteronism does not necessarily play a greater role in the combination with one or more other biomarkers. Also, the more biomarkers there are, the higher the predictive accuracy (AUC value) of the combination is not necessarily true. Therefore, a large number of validation experiments are still needed.

[0084] Therefore, based on the screening results of Example 1, this embodiment selected gender and K. + Five biomarkers significantly associated with primary aldosteronism were analyzed: concentration, ALD, PRA, and ARR. Additionally, although age, uric acid levels, 18-OHF, and 18-OHB were not significantly affected, the results in Table 1 showed that the p-values ​​for age, uric acid, 18-OHF, and 18-OHB were less than 0.05, indicating they could also be used to construct a predictive model. Therefore, in this embodiment, age, uric acid levels, 18-OHF, and 18-OHB were also included as biomarkers for model construction. The plasma samples used in this embodiment included 69 cases of primary aldosteronism and 216 cases of primary hypertension. The collection and analysis methods were the same as in Example 1, and predictive models were constructed and detected according to the following biomarker combinations:

[0085] Combination 1: ALD, PRA, ARR, and uric acid levels;

[0086] Combination 2: ALD, PRA, ARR, gender;

[0087] Combination 3: ALD, PRA, ARR, K + concentration;

[0088] Combination 4: ALD, PRA, ARR, gender, age;

[0089] Combination 5: ALD, PRA, ARR, gender, and K + concentration;

[0090] Combination 6: ALD, PRA, ARR, Age, Gender, K + Concentration and uric acid levels;

[0091] Combination 7: ALD, PRA, ARR, Age, Gender, K + Concentration and 18-OHB;

[0092] Combination 8: ALD, PRA, ARR, Age, Gender, K + Concentration, 18-OHB and 18-OHF;

[0093] Combination 9: ALD, PRA, ARR, age, sex, diastolic blood pressure, systolic blood pressure, blood urea nitrogen, creatinine, uric acid, potassium + Na + Cl- Adrenocorticotropic hormone, cortisol, 18-OHF, 18-OHB (i.e., the 17 markers shown in Tables 2 and 3).

[0094] R (version 4.3.1) was used as the development tool, and the rpart functions from the rpart package were used to construct decision tree models. The decision tree model (PA-SDx) constructed by combination 5 is shown below. Figure 3 As shown, the model can effectively classify and predict the target variable using the classification tree method of the rpart function (method = "class").

[0095] The model performance was evaluated using AUC, sensitivity, specificity, and accuracy. The results are shown in Table 4.

[0096] Table 4. Performance comparison of decision tree models constructed with different combinations of biomarkers

[0097]

[0098] As shown in Table 4, diagnostic models built using a single biomarker have lower diagnostic performance than models combining multiple biomarkers, and decision tree models built using a combination of four biomarkers have lower diagnostic performance than models combining five biomarkers. When using the five biomarkers in combination 5 to build the decision tree model, the model achieves the highest diagnostic accuracy, reaching 0.93, with an AUC of 0.929. When using seven biomarkers (combinations 6 and 7), eight biomarkers (combination 8), or seventeen biomarkers (combination 9) to build the diagnostic model, the diagnostic accuracy is very close to that of the decision tree model using combination 5, but the sensitivity and specificity are lower. Therefore, the combination of the five biomarkers in combination 5 is preferred, namely aldosterone concentration (ALD), renin activity (PRA), ARR value, sex, and K. + Concentration was used to construct a decision tree model for the diagnosis of primary aldosteronism.

[0099] Example 3: Comparison and selection of supervised classification algorithm models for constructing the optimal diagnostic model

[0100] In this embodiment, in order to select the best supervised classification algorithm for building the prediction model, the concentration matrix of the five best biomarkers selected in Example 2 is used as the original training dataset, and models under different supervised classification algorithms (including random forest, decision tree, logistic regression and support vector machine) are constructed according to the following methods. By comparing the performance of the different models constructed, the best supervised classification algorithm is selected.

[0101] Four diagnostic models, built based on random forest, decision tree, logistic regression, and support vector machine respectively, were optimized for hyperparameters (including learning rate, L2 regularization strength, maximum number of iterations, number of model layers, and number of neurons) using the Optuna framework and cross-validation on the training set. Ten-fold cross-validation was performed on the test set during training. The decision rules for each classification algorithm are as follows:

[0102] Decision Tree Classifier Algorithm: A decision tree is a tree-structured supervised learning model that recursively partitions features to form a decision path from the root node to the leaf nodes, used for classification or regression tasks. Its core idea is to decompose complex problems into multi-layered logical judgments through a series of decision rules. The hyperparameters of the decision tree model are: minimum number of split nodes (minsplit) = 20, complexity parameter (cp) = 0.01, maximum number of competitive splits (maxcompete) = 4, maximum depth (maxdepth) = 30, and the decision rules are as follows:

[0103] Table 5 Decision tree model selection criteria

[0104]

[0105] Logistic Regression Algorithm: Logistic regression is a linear probabilistic model used for classification tasks. Essentially, it predicts the probability of a sample belonging to a certain class by fitting a linear function. Its input parameters are: distribution family (binomial), link function (logit), maximum number of iterations (maxit) is 25, and convergence threshold (epsilon) is 1e-8. The predicted probabilities are as follows:

[0106] Z=2.937205+0.148269*ALD+0.009773*ARR-1.462620*K-0.784840*PRA+0.653381*Sex

[0107]

[0108] Note: e is the base of the natural logarithm; p represents the predictive value for whether an individual has primary aldosteronism.

[0109] Support Vector Machine (SVM): SVM is a model based on geometric margin and kernel function, used to solve binary classification problems. The core idea of ​​SVM is to find an optimal hyperplane that separates data points of different classes while maximizing the margin between the two classes. Its decision boundary is defined by the support vectors and kernel function parameters. Its input parameters are: model type (C-classification), kernel type (radial), probability estimate (TRUE), and generation formula:

[0110]

[0111] Where X represents the feature vector of the input sample (including K) + (PRA, ALD, ARR and Sex), the support vector set sv[i] is the i-th support vector, coefs[i] is the coefficient of the i-th support vector, γ is 0.1666667, and the bias term b is -0.06903083.

[0112] Random Forest Classifier: Random Forest is an ensemble learning algorithm that improves the accuracy and stability of classification or regression by constructing multiple decision trees and combining their predictions. Because its structure consists of a large number of decision trees, the splitting rules within each tree are complex, making the overall discriminative logic of the model difficult for humans to directly understand; it is considered a "black box model." Its input parameters are: model type (classification), number of trees (500), feature subset size (mtry) (2), and sampling with replacement (Replace) (TRUE).

[0113] In this embodiment, the training set and the test set use two completely different batches of samples. The training set consists of known samples, from which the inventors screen for biomarkers. The test set includes 39 cases of essential aldosteronism and 140 cases of essential hypertension, and is used solely to verify the diagnostic efficacy of the biomarker combination of this invention. The specific verification results are shown in Table 6.

[0114] Table 6. Performance comparison of models built using different supervised algorithms

[0115]

[0116] As shown in Table 6, the results were based on aldosterone concentration, renin activity, ARR value, sex, and K. +Among the five biomarkers (concentration, etc.), the diagnostic model constructed using the decision tree algorithm achieved the best performance score, with a sensitivity of 0.92, a specificity of 0.91, an accuracy of 0.91, and an AUC of 0.93. Therefore, the decision tree model is the preferred choice for the diagnosis of primary aldosteronism.

[0117] Example 4: Performance Validation of a Decision Tree Model Constructed with 5 Markers

[0118] To demonstrate the technical effectiveness of this invention, this embodiment further collected plasma samples from 46 patients with primary aldosteronism and 150 patients with primary hypertension. The predictive model (PA-SDx) was compared with the closest existing technology (ARR>30 for positive samples), and also with the current PA diagnostic test—the Captopril Challenge Test (CCT). The CCT method involves the patient taking 50 mg of captopril orally one hour after administration (sitting or standing). Plasma renin activity, aldosterone, and cortisol were measured before administration and at 1 and 2 hours after administration. The patient must remain seated throughout the test. In normal individuals, the captopril inhibition test resulted in a decrease in aldosterone concentration greater than 30%, while aldosterone levels were not suppressed in patients with primary aldosteronism. The diagnostic performance of the current screening technology and PA-SDx was then compared, and the results are shown in Table 7. The Positive Predictive Value (PPV) is the probability of actually having the disease when the test result is positive. PPV = number of true positives / (number of true positives + number of false positives). The Negative Predictive Value (NPV) is the probability of actually not having the disease when the test result is negative. NPV = number of true negatives / (number of true negatives + number of false negatives).

[0119] Table 7 Performance comparison of PA-SDx with currently commonly used clinical screening methods

[0120]

[0121] Note: Performance improvement calculation formula: (PA - SDx - ARR > 30) / ARR > 30

[0122] As shown in Table 7, based on multicenter data, when using the ARR=30 recommended by current guidelines as the cutoff point, the diagnostic sensitivity is 0.59, specificity is 0.85, accuracy is 0.79, and AUC is 0.72. However, using a decision tree model constructed based on five indicators, the sensitivity is 0.92, specificity is 0.94, accuracy is 0.93, and AUC is 0.93. Compared with existing technologies, this invention improves diagnostic performance from different perspectives, and overall, increases the accuracy by 18.05%.

[0123] Table 8 Performance comparison of PA-SDx with current clinical CCT confirmatory trials

[0124]

[0125] Note: Performance improvement calculation formula: (PA - SDx - CCT) / CCT

[0126] As shown in Table 8, in a multicenter cohort, compared with the CCT diagnostic test, when both are at the same specificity, the model's sensitivity is improved by 114.67%, demonstrating significant clinical benefits. In summary, compared with existing PA diagnostic techniques, the diagnostic model constructed using five biomarkers in this invention significantly improves predictive performance and accuracy, exhibiting optimal diagnostic efficacy.

[0127] Furthermore, compared to the primary aldosteronism screening and diagnostic classification system provided in the invention patent CN114895043B applied for by our research team on July 13, 2022, entitled "Primary Aldosteronism Screening Kit and Diagnostic and Subtyping System", the decision tree diagnostic model provided by this invention can directly replace the diagnostic test. It eliminates the need for a second diagnostic test after a positive screening result, thereby improving diagnostic accuracy, simplifying the diagnostic process, and increasing work efficiency.

[0128] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Anyone skilled in the art can make various modifications and alterations without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be determined by the claims.

Claims

1. A model for predicting primary aldosteronism, characterized in that, The model is constructed by training a decision tree algorithm on data from a combination of biomarkers, including aldosterone concentration, renin activity, ARR value, sex, and K. + Combinations of concentrations.

2. A model for predicting primary aldosteronism, characterized in that, The model is constructed by training a decision tree algorithm using data from a combination of biomarkers, including aldosterone concentration, renin activity, ARR value, age, sex, and K. + A combination of concentration and uric acid level.

3. A model for predicting primary aldosteronism, characterized in that, The model is constructed by training a decision tree algorithm using data from a combination of biomarkers, including aldosterone concentration, renin activity, ARR value, age, sex, and K. + Concentration, a combination of 18-hydroxycorticosteroids and 18-hydroxycorticosterone.

4. The model as described in claim 1, characterized in that, The biomarker combination includes aldosterone concentration, renin activity, ARR value, sex, and K. + Combinations of concentrations.

5. The model as described in claim 4, characterized in that, The prediction of primary aldosteronism is used to determine whether a patient with primary hypertension has primary aldosteronism.

6. The use of a reagent for detecting a combination of biomarkers in the preparation of a reagent for predicting whether an individual has primary aldosteronism, characterized in that, The biomarker combination includes aldosterone concentration, renin activity, ARR value, sex, and K. + Combinations of concentrations.

7. A kit for predicting whether an individual has primary aldosteronism, characterized in that, The kit includes reagents for detecting a combination of biomarkers, namely aldosterone concentration, renin activity, ARR value, sex, and K+. + Combinations of concentrations.

8. A system for predicting whether an individual has primary aldosteronism, characterized in that, The system includes a data analysis module; the data analysis module is used to analyze the detection values ​​of biomarkers, including aldosterone concentration, renin activity, ARR value, sex, and K+. + Combinations of concentrations.

9. The system as described in claim 8, characterized in that, The data analysis module uses the detection values ​​of biomarkers to construct a decision tree model and calculates the probability value of whether an individual has primary aldosteronism.

Images