Development and validation of a prediction score for subtype diagnosis of primary aldosteronism

To the Editor: Primary aldosteronism (PA) is a common cause of secondary hypertension, accounting for 5% to 20% of hypertensive patients.[1] The two main subtypes are aldosterone-producing adenoma (APA) and idiopathic hyperaldosteronism (IHA). Differentiating these subtypes is crucial for selecting appropriate treatment.[2] However, the current gold standard for subtype diagnosis, adrenal venous sampling (AVS), is invasive, costly, and not universally available.[3] This study aimed to develop a noninvasive prediction score using clinical and imaging parameters to distinguish APA from IHA. This study was approved by the Ethics Committee of the First Affiliated Hospital of Zhengzhou University (No. 2019-KY-366) and performed in accordance with institutional guidelines. Given the retrospective nature of the study, the Ethics Committee granted a waiver of informed consent. All data used in this study were fully anonymized prior to analysis to ensure patient privacy and confidentiality, and no direct patient contact or intervention was involved. In this study, a retrospective analysis was conducted, including 285 patients diagnosed with PA who underwent successful AVS at the Department of Endocrinology and Metabolism, First Affiliated Hospital of Zhengzhou University, between July 2021 and January 2024. These patients were randomly divided into a development cohort (n = 201) and a validation cohort (n = 84). The adrenal gland volumes (AGVs) of patients were adjusted based on the AGVs of 2295 healthy individuals who underwent routine physical examinations at the same hospital during the same period, with adjustments made for sex and age. The study also included 92 patients from the Department of Endocrinology at the First Affiliated Hospital of Chongqing Medical University, all of whom underwent successful AVS and had bilateral AGVs measured. An external validation cohort comprising 52 patients with APA and 40 patients with IHA was used to further verify the model's performance. AGVs were measured via computed tomography (CT), and the adjusted adrenal gland volume ratio (adjusted AGV ratio) was calculated. Clinical data, including plasma aldosterone concentration (PAC), plasma renin activity (PRA), aldosterone-to-renin ratio (ARR), and serum potassium levels, were collected. Logistic regression analysis identified independent predictors of APA, leading to the creation of a simplified clinical prediction score. Compared to patients with IHA, those with APA exhibited significantly lower serum potassium levels (2.87 [2.57–3.17] mmol/L vs. 3.27 [2.98–3.62] mmol/L, P <0.001), higher PAC (383.00 [256.60–564.40] pg/mL vs. 200.50 [154.98–254.78] pg/mL, P <0.001), and higher ARR (298.40 [177.70–524.00] vs. 154.73 [87.06–231.62], P <0.001). No significant differences were observed in other laboratory parameters or in left adrenal gland volume (LAGV), right adrenal gland volume (RAGV), adjusted LAGV, or adjusted RAGV. However, a significant difference was found in the adjusted AGV ratio between the two groups (P <0.001). Univariable logistic regression revealed that the adjusted AGV ratio, PAC, ARR, and lowest serum potassium level were significantly associated with APA diagnosis. In multivariable analysis, the adjusted AGV ratio (odds ratio [OR] = 3.406, 95% confidence interval [CI], 1.244–9.326; P = 0.017), ARR (OR = 1.005, 95% CI, 1.003–1.008; P <0.001), and lowest serum potassium level (OR = 0.281, 95% CI, 0.127–0.618; P = 0.002) remained significant predictors, while PAC was excluded due to collinearity with ARR. Based on these findings, a prediction score was developed incorporating three key predictors [Table 1]. An adjusted AGV ratio greater than 1.4646 was assigned 2 points, an ARR greater than 363.48 was assigned 7 points, and a serum potassium level of ≤ 3.06 mmol/L was assigned 2 points. The total score ranged from 0 to 11, with higher scores indicating a greater likelihood of APA. In the development cohort, the score achieved an area under the curve (AUC) of 0.781 (95% CI, 0.718–0.836). At a cutoff score of 9 points, the specificity reached 100% while maintaining a sensitivity of 35.65%, effectively ruling out IHA. The validation cohort demonstrated similar performance, with an AUC of 0.760 (95% CI, 0.655–0.847), confirming the model's robustness. Table 1 - Included factors and final points used for the prediction of APA. Risk factor Reference value (Wij) βi βi (Wij – WiREF) Pointsij = βi (Wij – WiREF)/B Adjusted AGV ratio 1.225 ≤1.46 0.97 = WiREF 0 0 >1.46 2.43 1.7885 2 ARR 0.005 ≤107.27 62.61 = WiREF 0 0 107.28–363.47 235.37 0.8638 1 >363.47 1418.09 6.7774 7 Lowest serum potassium –1.271 ≤3.06 mmol/L 2.50 1.7921 2 3.07–3.38 mmol/L 3.22 0.8770 1 >3.38 mmol/L 3.91 = WiREF 0 0 The identified three independent risk factors associated with APA were simplified, weighted according to their regression β coefficients, and used to develop a prediction score. The points for each factor were calculated based on the following calculations: (1) the distance D between each of the risk factors and the underlying risk group according to the estimated coefficient βi and the reference value Wij of each risk factor; (2) D = (Wij – WiREF) × βi; (3) constant B = 1; (4) the score related to each risk factor, points = D/B = (Wij – WiREF) × βi/B. Points were determined for each factor. AGV adrenal gland volume; ARR: Aldosterone-to-renin-ratio; Wij: Reference value assigned to a specific category of a risk factor; WiREF: Baseline risk reference value. Using Youden's index, the optimal cutoff point was determined to be 4. In the development cohort, this cutoff correctly identified APA in 61 of 115 patients (sensitivity 53.04%) and IHA in 78 of 86 patients (specificity 90.60%), with an overall accuracy of 69.15%. The validation cohort demonstrated similar accuracy (65.48%), with a sensitivity of 45.24% and specificity of 85.71%. When the cutoff point was raised to 9 to achieve 100% specificity, the prediction score correctly classified all 86 IHA patients in the development cohort, with a sensitivity of 35.65% and an accuracy of 63.18%. The validation cohort showed consistent results, confirming the score's reliability across different patient groups. In the total cohort, the score achieved an accuracy of 63.86%, sensitivity of 34.39%, and specificity of 100.00%. External validation was conducted using data from 92 patients at another tertiary hospital, including 52 APA and 40 IHA cases. The prediction score's performance remained consistent, with an AUC of 0.772 (95% CI, 0.671–0.857). At the cutoff score of 9 points, the specificity was maintained at 100%, with a sensitivity of 30.77%, highlighting the model's reliability across different populations and healthcare settings. The prediction score's practicality lies in its reliance on widely available clinical and imaging data. Unlike AVS, which requires specialized expertise and equipment, the variables used in the score can be easily obtained through routine laboratory tests and CT scans. This makes the score an attractive option for centers that lack the resources to perform AVS. Compared to existing models, this score simplifies the diagnostic process while maintaining high specificity. Previous models often required complex calculations or additional tests, limiting their utility in clinical practice.[4] By focusing on three key variables, the new score offers a straightforward approach to identifying patients who may benefit from unilateral adrenalectomy. The prediction score achieved high specificity, but its sensitivity was relatively low, highlighting the need to balance sensitivity and specificity in clinical decision-making. In practice, a score of 9 or higher can reliably exclude IHA and strongly sugggests APA, though AVS remains necessary to confirm APA diagnosis due to the limited sensitivity. Notably, since scores below 9 fail to exclude APA, patients with low scores but high clinical suspicion still require AVS for definitive diagnosis. Overall, this approach helps clinicians to avoid unnecessary AVS in definite IHA cases. The adjusted AGV ratio emerged as a critical variable for differentiating APA from IHA, reflecting asymmetrical adrenal gland enlargement in APA patients. This finding underscores the potential of imaging-based parameters in improving PA diagnosis. Incorporating ARR and serum potassium levels, which are routinely measured in clinical practice, further enhances the model's applicability. One of the main strengths of this study is the inclusion of external validation using deep learning to measure adrenal volume,[5] which confirmed the robustness of the prediction score across different patient populations. The use of a simplified scoring system ensures that the tool is accessible and easy to implement in clinical practice. However, it is essential to consider some limitations. First, the study was conducted in a limited number of centers, which may affect the generalizability of the results. In addition, the score does not account for postoperative outcomes, which could provide further insights into its clinical utility. Future research should focus on refining the score by incorporating additional variables that may improve its sensitivity without compromising specificity. Prospective, multicenter studies are also needed to validate the score's performance across different healthcare settings and patient populations. The clinical implications of this prediction score are significant. By providing a noninvasive method to provisionally exclude IHA, the score can help reduce the reliance on AVS, which is often a bottleneck in the diagnostic pathway for PA. This can lead to more timely and accurate diagnoses, ultimately improving patient outcomes. In conclusion, this novel prediction score offers a reliable, noninvasive method to identify APA in PA patients. Integrating clinical and imaging data provides a practical tool for guiding treatment decisions, particularly in centers without access to AVS. The score's high specificity makes it a valuable resource for clinicians, potentially reducing the reliance on invasive diagnostic procedures and improving patient outcomes.