PA is a common but severely underdiagnosed form of secondary hypertension [11, 12]. Elevated aldosterone levels are strongly associated with cardiovascular events and metabolic syndrome, and patients with PA are often prone to severe organ damage [13, 14]. Therefore, early detection and treatment of PA are necessary to minimize cardiovascular complications. Screening for PA is usually a multi-step process that uses ARR measurement as the initial screening condition [15]. In 2008, the European Society of Endocrinology published guidelines for the diagnosis and treatment of PA [16], recommending the use of ARR as a screening indicator; international and domestic guidelines and expert consensuses currently also recommend ARR as a screening index [2]. Additionally, drugs affecting the renin–angiotensin–aldosterone system should be discontinued before ARR measurement to improve the accuracy of PA screening [17]. In patients with hypertension, non-dihydropyridine CCBs and α-adrenergic blockers should be administered to control blood pressure. In this study, the missed diagnosis rate due to pre-washout ARR was nearly 25.0%. This highlights the importance of performing a repeat analysis of ARR after medication washout. However, these recommendations are not easy to implement universally, especially in patients with severe hypertension, where the hypotensive effects of drug washout are limited and patients cannot tolerate an entire 2–4 week washout. Additionally, some patients cannot tolerate non-dihydropyridine CCBs or α-adrenergic blockers owing to bradycardia or severe postural hypotension, making drug washout for PA screening difficult. Therefore, a complete washout before PA screening is difficult. Herein, 55.9% of patients (612/1095) were treated with two or more antihypertensive drugs to control blood pressure, further complicating drug washout. Consequently, simplifying the process of PA screening is an urgent clinical need to ensure patient safety and reduce blood pressure fluctuations while also improving the sensitivity and specificity of diagnosis. Thus, we included a retrospective analysis of ARR in 1095 patients who underwent hypertension screening and measured their PAC and DRC values before and after washout. The ARR values were analyzed to construct a screening model suitable for pre-washout PA screening.

Many clinical factors affect the ARR and unifying them is difficult. In this study, patients aged > 65 years or < 18 years and those with renal insufficiency were excluded, as were patients with factors such as renal hypertension, renal artery stenosis, and pregnancy, which interfere with ARR. Patients were instructed to maintain normal sodium intake throughout the screening process, and efforts were made to correct blood potassium levels to within the normal range; this was done to minimize the impact of blood potassium and sodium intake on ARR and prevent the impact of factors unrelated to medication on ARR.

Among the included patients, 45% had PA and 55% had EH. There was no significant difference in the sex ratio, duration of disease, BMI, or blood creatinine level between the two groups. However, the two groups had significant differences in age, hypertension grade, proportion of patients with refractory hypertension, number of antihypertensive drugs used, and blood sodium and potassium concentrations. Patients with PA were younger, had higher blood pressure, were poorly controlled with medication, and had lower blood potassium and higher blood sodium levels than patients with EH. Although recent studies showed that PA could present as hypertension with normal blood potassium, accounting for approximately 39–50% of all patients with PA [18, 19], hypokalemia cannot be used as a diagnostic criterion for PA [20] nor can normal blood potassium be used as an exclusion criterion. However, some studies have shown that the prevalence of PA is as high as 88.5% in patients with spontaneous hypokalemia and blood potassium concentrations < 2.5 mmol/L [21]. The prevalence of hypokalemia among Chinese patients with PA has been reported to be 74.8% [22]. Herein, the mean blood potassium concentration of the patients with PA was 3.40 ± 0.50 mmol/L, which was lower than normal, and the proportion of those with hypokalemia was 69.6%. This indicates that over half of the patients with PA had hypokalemia as the primary manifestation in addition to elevated blood pressure, suggesting that blood potassium levels can be considered adjunct markers in pre-washout screening. Hypertensive patients with a history of hypokalemia are more likely to be of clinical concern, and the retrospective design of the study (including only patients screened for PA based on clinical suspicion) may carry an element of selection bias, leading to an overestimation of the performance of the diagnostic model for subsequent screening [23].

Before washout, patients with PA still exhibited higher PAC, lower DRC, and higher ARR values than patients with EH (Fig. 3), indicating that these clinical manifestations of PA remained prominent despite medication acting as a confounding factor. Moreover, these indicators can be included as alternative predictors in the development of the screening model. To validate the effects of medication on the PAC, DRC, and ARR values during PA screening, we compared these values in patients with PA before and after washout. We found that the PAC levels did not change significantly before and after washout; however, the DRC levels were higher before washout, leading to decreased ARR before washout (Fig. 3). This finding indicates that drug interference primarily affects the DRC and produces a false-negative ARR result. Next, patients with PA were divided into false-negative, false-positive, and mixed drug groups based on the classes of drugs they used before washout. No significant differences between the pre-and post-washout PAC values were observed in all three groups (Table 3). In both the false-negative and mixed drug groups, DRC values decreased after washout (p < 0.05); however, the only significant difference was the elevation in post-washout ARR in the false-negative drug group (Table 3). Considering that the pre-washout drugs taken by the patients here were ACEIs/ARBs and CCBs and 10% of the patients were treated with diuretics, the drug-associated interfering factors were primarily due to false-negative drugs. Mixed drugs were also common, whereas using false-positive drugs alone was rare; this resulted in the drugs having both increasing and decreasing effects on the PAC values of patients with PA, leading to no significant differences. The primary effect of medication was a significant increase in the DRC value. The ARR was affected by these changes in PAC and DRC, resulting in either increased or decreased values, and there was a significant difference in the false-negative drug group. Additionally, although the pre-washout DRC values were elevated in the false-negative drug group, the absolute value of DRC did not increase significantly, possibly because DRC is affected by PA itself in addition to the effects of medication, and the overlap of these two factors attenuated the effects of medication. The sample size of the false-positive drug group was small, and there were no significant differences in the PAC, DRC, or ARR values before and after washout (Table 3). This result suggests that the sample size should be expanded in future studies to confirm the present results. Additionally, we conducted an in-depth analysis of false-negative drugs and made several inferences: ① ACEIs/ARBs mainly affect DRC, causing a decrease in ARR, with no significant impact on PAC. If ACEIs/ARBs are used and the pre-washout ARR is positive, it is speculated that it will remain positive after washout; ② CCBs have no significant impact on PAC and DRC, and there is no significant change in ARR before and after washout. For patients who find it challenging to complete washout, considering the use of CCB drugs may be an option; ③ Subgroup analysis for the above-mentioned false-negative drugs suggests that they have no significant impact on PAC. Therefore, for patients using these drugs, regardless of whether their ARR results are positive or not, attention should be paid to abnormal elevation of PAC to reduce missed diagnoses; ④ In patients treated with only ACEIs/ARBs here, the mechanisms of action of the drugs on ARR and PAC were the same as those in the CCT, in which PAC decreased and DRC increased. Thus, using this class of drugs could potentially lead to an increased probability of false-negative ARR values [24, 25]. A diagnosis of PA can be considered in the case of positive ARR and PAC results upon treatment with ACEIs/ARBs alone.

There were differences between the patients with PA and EH in terms of the classes of medication used. Because this was a retrospective study, the classes, duration, and doses of medication were not standardized during consultation and treatment; hence, achieving uniformity and standardization during follow-up consultation and treatment was difficult. The pre-washout treatment status was determined without uniform standardized drug treatment; therefore, we concluded that these indicators were subjectively influenced and unsuitable for inclusion as alternative predictors. Additionally, although there was a significant difference in the age between patients with PA and EH, the absolute age difference between the two groups was not significant nor was age considered an alternative predictor, possibly owing to the sample size. Taking these factors together, six indicators were selected: blood potassium; blood sodium; hypertension grade; pre-washout PAC, DRC, and ARR, all exhibiting significant differences between PA and EH. These values were clinically easy to obtain and yielded relatively stable test results.

The determination of cutoff values for these predictors using ROC curve analysis, subsequent conversion to binary variables, and logistic regression analysis were used to construct the PAPS3 screening model. The factors included in this model are also consistent with the predictors for PA screening currently used in most institutions [4, 26, 27]. Although most studies have not set an additional threshold for PAC when using ARR for PA screening, some have set PAC thresholds [4, 8, 26, 28], including PAC ≥ 9 ng/dL [26] and PAC > 16 ng/dL [28]. The PAPS3 model had a maximum score of 10, a cutoff value of 5.5, a sensitivity of 85.6%, and a specificity of 92.3% for screening PA. Calibration of the PAPS3 model was further performed using the Hosmer–Lemeshow test, which indicated a good fit with the original data and good predictive power. The stability of the PAPS3 model was further evaluated using internal bootstrapping validation, and the results confirmed that the model was stable.

The PAPS3 model showed high sensitivity, specificity, and accuracy, with an area under the curve > 0.9. There are many possible reasons for this. First, of the five indicators selected for the construction of the model, three (hypertension grade, blood potassium, and blood sodium) were not restricted to pre- or post-washout or affected by drugs before or after washout. Additionally, there was no significant difference in the PAC of patients with PA before and after washout. Subgroup analysis also indicated no significant difference in the PAC values between the drug treatment groups. This suggests that pre- and post-washout PAC was unaffected by large fluctuations caused by drug interference. As an indicator in the primary screening of PA, ARR plays a key role in the diagnostic process. Using ARR > 2.60 (ng/dL)/(mU/L) as the cutoff value, the present study found a consistency of 80.8% between pre- and post-washout ARR, indicating a high degree of overlap. Therefore, based on these factors, the PAPS3 model had high accuracy.

Despite the model’s accuracy, we also considered potential drawbacks during its development. First, the model was based on a retrospective analysis; all included patients had undergone drug washout. No patients who had difficulty tolerating drug washout were included, thereby introducing selection bias; however, there is currently no clinical solution to this problem. Additionally, patients with concomitant hypertension and hypokalemia, as well as those with pre-washout abnormalities in PAC and ARR, are more likely to draw the attention of clinicians and undergo proper PA screening; this bias can significantly inflate the measures on diagnostic tests [23]. Furthermore, a growing body of evidence shows that PA is present in some cases of mild hypertension and even in populations with normal blood pressure. These patients often do not receive further PA screening and are more likely to undergo follow-up only in primary care settings such as community clinics [29,30,31], causing some patients with PA to be missed. Additionally, there is great variability between study designs that is difficult to reconcile; variables include the population of interest, ARR threshold value, methods of renin and aldosterone measurement, and type and diagnostic thresholds of the confirmatory tests used, all of which introduce variations in the sensitivities and specificities of PA screening reported in different studies. Finally, a “gray area” in diagnosing PA makes a definitive diagnosis of EH or PA difficult.

In the past, researchers constructed similar models for the classification and diagnosis of PA. The Küpers score [32] is a typical predictive model for the staging of PA that includes typical adenoma imaging with thresholds of blood potassium < 3.5 mmol/L and glomerular filtration rate > 100 mL/(min. 1.73 m2). At Küpers score ≥ 5, the sensitivity and specificity for diagnosing unilateral PA are 53% and 100%, respectively. However, a Chinese study found that the Küpers score had low sensitivity and specificity for the Chinese population (62% and 53%, respectively), which did not apply to the elderly population. After modifying the model, it was found that urinary aldosterone levels, history of hypokalemia, and typical unilateral adenoma diameter > 1 cm had a diagnostic specificity of 90.5%. In China, He et al. used imaging histology techniques and clinical characteristics to establish a nomogram model that includes adrenal computed tomography imaging histology score, age, sex, blood potassium, and ARR to predict aldosterone-producing adenoma [33]. Although the PAPS3 model cannot distinguish the specific stage of PA, its advantages include the use of simple and easily obtainable predictors, convenience for clinical application, and high sensitivity and specificity.

ARR is recommended for the diagnosis of PA but is a highly variable test. In clinical practice, the factor most difficult to control but frequently encountered is that of a patient with hypertension undergoing screening for PA on medications that interfere with the measurement of the ARR. Robust detection of PA mandates that factors known to alter the ARR are controlled before sampling. Washout of interfering antihypertensive medications in non-hospitalized patients is not without risk; it is safe only in mildly hypertensive and regularly monitored patients. The key objective of the present study was to develop a diagnostic model for pre-washout screening of PA that is convenient for clinical application. Our analysis yielded the PAPS3 model, which has potential application in clinical practice and good predictive ability. However, PA should not be automatically diagnosed in patients who meet the PAPS3 scoring criteria; the recommended guidelines for PA screening, including an in-depth patient evaluation for possible PA, should still be followed. For patients able to undergo washout, the processes recommended by the guidelines for screening and confirmation should be followed. However, for patients who have difficulty tolerating washout of interfering drugs, the PAPS3 model can be recommended for the preliminary diagnosis of PA to reduce the risks to the patient during washout.