Identifying KCNJ5 Mutation in Aldosterone-Producing Adenoma Patients With Baseline Characteristics Using Machine Learning Technology

Background Primary aldosteronism is characterized by inappropriate aldosterone production, and unilateral aldosterone-producing adenoma (uPA) is a common type of PA. KCNJ5 mutation is a protective factor in uPA; however, there is no preoperative approach to detect KCNJ5 mutation in patients with uPA. Objectives This study aimed to provide a personalized surgical recommendation that enables more confidence in advising patients to pursue surgical treatment. Methods We enrolled 328 patients with uPA harboring KCNJ5 mutations (n = 158) or not (n = 170) who had undergone adrenalectomy. Eighty-seven features were collected, including demographics, various blood and urine test results, and clinical comorbidities. We designed 2 versions of the prediction model: one for institutes with complete blood tests (full version), and the other for institutes that may not be equipped with comprehensive testing facilities (condensed version). Results The results show that in the full version, the Light Gradient Boosting Machine outperformed other classifiers, achieving area under the curve and accuracy values of 0.905 and 0.864, respectively. The Light Gradient Boosting Machine also showed excellent performance in the condensed version, achieving area under the curve and accuracy values of 0.867 and 0.803, respectively. Conclusions We simplified the preoperative diagnosis of KCNJ5 mutations successfully using machine learning. The proposed lightweight tool that requires only baseline characteristics and blood/urine test results can be widely applied and can aid personalized prediction during preoperative counseling for patients with uPA.

P rimary aldosteronism (PA, or Conn's syndrome) is characterized by the inappropriate production of aldosterone, and its prevalence is as high as 11.2% in newly diagnosed hypertensive patients, depending on the screening stringency and population. 1,2 Unilateral aldosterone-producing adenoma (uPA) is a common type of PA 3 and a curable form of hypertension.
Recent studies have shown that somatic mutations play a crucial role in the pathogenesis of uPA. 4 The KCNJ5 somatic mutation is associated with an almost 6-fold increase in the chance of complete clinical success after adrenalectomy in patients with uPA.
Furthermore, the presence of the KCNJ5 somatic mutation was an independent hypertension remission predictor after unilateral adrenalectomy in patients with uPA. 5 Presently, the identification of KCNJ5 mutations requires adrenalectomy and sampling of adrenal tumor tissues for Sanger sequencing.
This procedure is invasive, highly complex, and enables accurate diagnosis after adrenalectomy. However, the comprehensive data required to ensure surgical indication also necessitates access to highly sophisticated medical equipment, making this method impractical for general practitioners.
Regarding the development of a prediction method using clinical values commonly used in PA diagnostic steps, studies using machine learning for uPA are beginning to be seen recently. [6][7][8] However, there are currently no associated results on the detection of KCNJ5-mutated uPA. 9 Therefore, it is important to find a general approach to predict KCNJ5 mutations in patients with uPA. We aimed to develop a machine learning model to forecast the occurrence of KCNJ5 mutations in patients with uPA, using baseline demographic characteristics and laboratory data. This emphasizes the clinical utility of personalized therapy during preoperative counseling for blood pressure responses.

METHODS
STUDY DESIGN AND PARTICIPANTS. The inception cohort was based on the Taiwan Primary Aldosteronism Investigation (TAIPAI) database and the tissue bank and associated information is provided in the Supplemental Appendix. 10 A patient recruitment flowchart is shown in Figure 1 and all enrolled patients need to fulfill all criteria. In the general hospital, patients diagnosed with hypertension and suspected PA underwent screening for suspected PA, illustrated as decision point (DP) I in Figure 1. Before PA confirmation screening, antihypertensive medications were discontinued for at least 21 days before conducting confirmatory tests. 11 Patients with an initial aldosterone-renin ratio of >35 were confirmed to have PA through saline infusion or captopril tests (DP II). Subtypes of PAs (uPA and non-uPA) were identified by further examination, including advanced imaging studies or other invasive examinations if required (DP III). Identifying uPA requires the following criteria to be satisfied 11 (DP IV): 1) PA diagnosis was confirmed; 2) imaging evidence was available for a unilateral adrenal adenoma or hyperplasia; 3) lateralization of aldosterone hypersecretion with adrenal vein sampling or during dexamethasone suppression NP-59 single-photon emission computed tomography 12 to the abovementioned imaging findings; or 4) uPA was further confirmed after adrenalectomy with a pathologically proven CYP11B2-positive stained adenoma or through immunohistochemical evidence for (multiple) aldosterone-producing micronodule(s) after adrenalectomy. 13     accuracy, sensitivity, specificity, and F1) were averaged, and an average score was calculated to compare the approaches. The model training process is illustrated in Figure 2, and the detailed development information is described in the Supplemental Appendix.

RESULTS
In total, 328 patients were enrolled; there was a mild predominance of women (53%) among the participants, with an average age of 51.3 AE 11.6 years ( Table 1). The 87 parameters collected in this study, including routine blood and urine tests, clinical demographics, and comorbidities, are presented in Table 1 and Supplemental Table 1. The parameters that showed statistically significant differences between the 2 groups of patients with uPA are listed in    Table 2 and Figures 3B and 3C show the corresponding receiver operating characteristic curves for each feature selection method.
Supplemental Table 4 lists the full features and 2 selected features, and Figure 5 shows a visualized plot of the feature importance ranking.  Table 6 to 8).

DISCUSSION
The      However, the detailed mechanism remains unclear and further research is required to address these  Physicians are unlikely to diagnose a patient with a KCNJ5 mutation without advanced genomic sequencing of resected adrenal tissues, which requires invasive adrenalectomies or blind biopsies. In clinical practice, the patents received blood test, confirmation test, subtypes differentiation, and advancing image scan. Then, we will suggest treatment strategies based on previous finding of examination. After adrenalectomy, we will perform gene mutation analysis; however, most patients hesitated to receive adrenalectomy. Our predictive model can potentially aid in providing better surgical options for patients with unilateral aldosterone-producing adenoma (uPA) who could benefit from ipsilateral adrenalectomies and guide postoperative strategy.
proposed by Riley et al 24 and Skov et al. 25 This indicates that the credibility of our study falls at a confidence level of 95%. Therefore, we considered the sample size of our study to be sufficient. The recruited patients came from different cities in Taiwan, including 2 medical centers, 3 affiliated hospitals, and 2 regional hospitals and was further validated by an international cohort. We did not conduct a randomized trial to recruit patients; instead, the patients were identified and approached during regular visits during the study period. The population represents the occurrence of cases observed in real practice, which is considered representative of the East Asian population.

CONCLUSIONS
In this study, we developed machine learning models to identify the occurrence of KCNJ5 mutations in adrenal tissues of patients with uPA using baseline characteristics and routine blood/urine test results.
We found that the full and condensed versions could accurately identify KCNJ5 mutations preoperatively.
Our predictive model can potentially aid in providing better surgical options for patients with uPA who could benefit from ipsilateral adrenalectomies and guide postoperative following strategy.