Welcome to JOURNAL OF SICHUAN UNIVERSITY (MEDICAL SCIENCE EDITION)
Volume 51 Issue 3
May  2020
Article Contents
Turn off MathJax

Citation:

Column:  

Diagnostic Efficiency of Different Screening Indexes for Primary Aldosteronism

  • Objective To investigate the the feasibility and diagnostic efficiencyvalue of different screening indexesmethods for screening primary aldosteronism (PA). Methods The clinical data of 499 patients with PA and 479 patients with essential hypertension diagnosed from Jan. 2009 to Dec. 2018 were retrospectively analyzed. The diagnostic performance of different screening indexs was compared by plotting receiver operating characteristic curves (ROC). Results The area under the ROC curve (AUC) of the plasma aldosterone concentration (PAC) to plasma renin activity (PRA) ratio (ARR) was greater than that of the ratio of the upright PAC to the angiotensin Ⅱ (AT-Ⅱ) (AA2R), upright PRA, upright PAC, supine ARR, and lowest blood potassium (P<0.05). The AUC of logistic regression model was greater than that of upright ARR (96.3% vs. 94.6%, P<0.05). There was no significant difference in AUC between decision tree model and upright ARR (94.1% vs. 94.6%, P>0.05). In the test set, the AUC difference between the logistic regression model and the decision tree model was not statistically significant (96.3% vs. 94.1%, P > 0.05). The calibration curve of the logistic regression model is closer to the 45 ° line, and the consistency between the prediction probability and the actual probability of the logistic regression model is better than that of the decision tree model. Conclusion For the screening of PA, upright ARR seems to be the best single screening index, while AA2R (radioimmunoassay) is not recommended. The diagnostic efficacy of logistic regression model including upright PAC, PRA and lowest blood potassium is better than that of single upright ARR.
  • 加载中
  • [1] FUNDER J W, CAREY R M, MANTERO F, et al. The management of primary aldosteronism: case detection, diagnosis, and treatment: an endocrine society clinical practice guideline. J Clin Endocrinol Metab,2016,101(5): 1889–1916. doi: 10.1210/jc.2015-4061
    [2] MULATERO P, STOWASSER M, LOH K C, et al. Increased diagnosis of primary aldosteronism, including surgically correctable forms, in centers from five continents. J Clin Endocrinol Metab,2004,89(3): 1045–1050. doi: 10.1210/jc.2003-031337
    [3] ROSSI G P, BERNINI G, CALIUMI C, et al. A prospective study of the prevalence of primary aldosteronism in 1,125 hypertensive patients. J Am Coll Cardiol,2006,48(11): 2293–2300. doi: 10.1016/j.jacc.2006.07.059
    [4] EIDE I K, TORJESEN P A, DROLSUM A, et al. Low-renin status in therapy-resistant hypertension: a clue to efficient treatment. J Hypertens,2004,22(11): 2217–2226.
    [5] WANG Z, CHEN Z, ZHANG L, et al. Status of hypertension in china: results from the china hypertension survey, 2012-2015. Circulation,2018,137(22): 2344–2356. doi: 10.1161/CIRCULATIONAHA.117.032380
    [6] MONTICONE S, D'ASCENZO F, MORETTI C, et al. Cardiovascular events and target organ damage in primary aldosteronism compared with essential hypertension: a systematic review and meta-analysis. Lancet Diabetes Endocrinol,2018,6(1): 41–50. doi: 10.1016/S2213-8587(17)30319-4
    [7] STOWASSER M, AHMED A H, PIMENTA E, et al. Factors affecting the aldosterone/renin ratio. Horm Metab Res,2012,44(3): 170–176. doi: 10.1055/s-0031-1295460
    [8] 李芳, 谢小英, 赵铁耘. 醛固酮_肾素比值联合低肾素水平诊断原发性醛固酮增多症的准确性. 四川大学学报(医学版),2010,41(3): 501–504.
    [9] YOUNG W F. Primary aldosteronism: renaissance of a syndrome. Clin Endocrinol (Oxf),2007,66(5): 607–618. doi: 10.1111/j.1365-2265.2007.02775.x
    [10] 朱杰, 赵玲, 王琳, 等. 尿醛固酮与不同体位血浆肾素活性比值对原发性醛固酮增多症的筛查价值. 解放军医学杂志,2018,43(7): 569–573. doi: 10.11855/j.issn.0577-7402.2018.07.05
    [11] 梁霞, 王小静, 余文珮, 等. 原发性醛固酮增多症诊断切点的研究及临床应用. 第三军医大学学报,2015,37(11): 1146–1149.
    [12] 李川, 周嘉, 梁杏欢, 等. 醛固酮与肾素的比值在原发性醛固酮增多症与原发性高血压鉴别中的应用. 中华高血压杂志,2015,23(6): 582–585.
    [13] BAUDRAND R, GUARDA F J, FARDELLA C, et al. Continuum of renin-independent aldosteronism in normotension. Hypertension,2017,69(5): 950–956. doi: 10.1161/HYPERTENSIONAHA.116.08952
    [14] YOUNG W F, Jr. Diagnosis and treatment of primary aldosteronism: practical clinical perspectives. J Intern Med,2019,285(2): 126–148. doi: 10.1111/joim.12831
    [15] VAIDYA A, MULATERO P, BAUDRAND R, et al. The expanding spectrum of primary aldosteronism: implications for diagnosis, pathogenesis, and treatment. Endocr Rev,2018,39(6): 1057–1088. doi: 10.1210/er.2018-00139
    [16] WOLLEY M J, STOWASSER M. New advances in the diagnostic workup of primary aldosteronism. J Endocr Soc,2017,1(3): 149–161.
    [17] POGLITSCH M. The AA2-ratio: improved screening for primary aldosteronism in hypertension. J Hypertens,2016,34: e270[2020-03-20].doi: 10.1097/01.hjh.0000492126.92947.fb.
    [18] POGLITSCH M, AHMED A H, RESL M, et al. The aa2-ratio: towards improved screening for primary aldosteronism in hypertension. J Hypertension,2016,34: e400[2020-03-20].doi: 10.1097/01.hjh.0000501024.99460.c9.
    [19] BURRELLO J, BUFFOLO F, DOMENIG O, et al. Renin-angiotensin-aldosterone system Triple-A analysis for the screening of primary aldosteronism. Hypertension,2020,75(1): 163–172. doi: 10.1161/HYPERTENSIONAHA.119.13772
    [20] POULSEN K. Measurements of renin-angiotensin-aldosterone. Pharmacol Rev,1973,25(2): 249–257.
    [21] 胡枫湫, 黄娟, 黄慧. 原发性醛固酮增多症诊断性试验在醛固酮瘤诊断中的临床应用与评价. 四川大学学报(医学版),2018,49(3): 469–473.
    [22] KONG A P S, CHOI C H, NG C M, et al. The use of aldosterone-renin ratio as a diagnostic test for primary hyperaldosteronism and its test characteristics under different conditions of blood sampling. J Clin Endocrinol Metab,2005,90(1): 72–78.
    [23] 张炯, 邹恒昀, 徐顺清, 等. 高血压患者不同体位醛固酮/血浆肾素活性值比较. 中国公共卫生,2010,26(3): 262–263.
    [24] BARIGOU M, AH-KANG F, ORLOFF E, et al. Effect of postural changes on aldosterone to plasma renin ratio in patients with suspected secondary hypertension. Ann Cardiol Angeiol (Paris),2015,64(3): 169–174. doi: 10.1016/j.ancard.2015.04.017
    [25] KOBAYASHI H, ABE M, SOMA M, et al. Development and validation of subtype prediction scores for the workup of primary aldosteronism. J Hypertens,2018,36(11): 2269–2276. doi: 10.1097/HJH.0000000000001855
    [26] HASHIMURA H, SHEN J, FULLER P J, et al. Saline suppression test parameters may predict bilateral subtypes of primary aldosteronism. Clin Endocrinol (Oxf),2018,89(3): 308–313. doi: 10.1111/cen.13757
    [27] KOBAYASHI H, HAKETA A, UENO T, et al. Scoring system for the diagnosis of bilateral primary aldosteronism in the outpatient setting before adrenal venous sampling. Clin Endocrinol (Oxf),2017,86(4): 467–472. doi: 10.1111/cen.13278
    [28] KOCJAN T, JANEZ A, STANKOVIC M, et al. A new clinical prediction criterion accurately determines a subset of patients with bilateral primary aldosteronism before adrenal venous sampling. Endocr Pract,2016,22(5): 587–594. doi: 10.4158/EP15982.OR
    [29] VENOS E S, SO B, DIAS V C, et al. A clinical prediction score for diagnosing unilateral primary aldosteronism may not be generalizable. BMC Endocr Disord,2014,14: 94[2020-03-20]. http://doi.org/10.1186/1472-6823-14-94. doi: 10.1186/1472-6823-14-94
    [30] SZE W C, SOH L M, LAU J H, et al. Diagnosing unilateral primary aldosteronism—comparison of a clinical prediction score, computed tomography and adrenal venous sampling. Clin Endocrinol (Oxf),2014,81(1): 25–30. doi: 10.1111/cen.12374
    [31] ZHANG Y, NIU W, ZHENG F, et al. Identifying unilateral disease in chinese patients with primary aldosteronism by using a modified prediction score. J Hypertens,2017,35(12): 2486–2492. doi: 10.1097/HJH.0000000000001488
    [32] NANBA K, TSUIKI M, NAKAO K, et al. A subtype prediction score for primary aldosteronism. J Hum Hypertens,2014,28(12): 716–720. doi: 10.1038/jhh.2014.20
    [33] RIESTER A, FISCHER E, DEGENHART C, et al. Age below 40 or a recently proposed clinical prediction score cannot bypass adrenal venous sampling in primary aldosteronism. J Clin Endocrinol Metab,2014,99(6): E1035–E1039. doi: 10.1210/jc.2013-3789
    [34] UMAKOSHI H, TSUIKI M, TAKEDA Y, et al. Significance of computed tomography and serum potassium in predicting subtype diagnosis of primary aldosteronism. J Clin Endocrinol Metab,2018,103(3): 900–908. doi: 10.1210/jc.2017-01774
    [35] XIAO L, JIANG Y, ZHANG C, et al. A novel clinical nomogram to predict bilateral hyperaldosteronism in chinese patients with primary aldosteronism. Clin Endocrinol (Oxf),2019,90(6): 781–788. doi: 10.1111/cen.13962
    [36] VELEMA M S, LINSSEN E J M, HERMUS A R M M, et al. A prediction model for primary aldosteronism when the salt loading test is inconclusive. Endocr Connect,2018,7(12): 1308–1314.
    [37] YAMASHITA T, SHIMIZU S, KOYAMA M, et al. Screening of primary aldosteronism by clinical features and daily laboratory tests: combination of urine pH, sex, and serum K+. J Hypertens,2018,36(2): 326–334. doi: 10.1097/HJH.0000000000001511
  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(7) / Tables(5)

Article views(3017) PDF downloads(22) Cited by()

Related
Proportional views

Column:  

Diagnostic Efficiency of Different Screening Indexes for Primary Aldosteronism

    Corresponding author: REN Yan, renyan@scu.edu.cn
  • 1. Adrenal Center, Department of Endocrinology and Metabolism, West China Hospital, Sichuan University, Chengdu 610041, China
  • 2. Department of Endocrine and Metabolic Diseases, Suining Central Hospital, Suining 629000, China
  • 3. Evidence-Based Medicine Center, West China Hospital, Sichuan University, Chengdu 610041, China
  • 4. Department of Nuclear Medicine, West China Hospital, Sichuan University, Chengdu 610041, China
  • 5. Health Management Center, West China Hospital, Sichuan University, Chengdu 610041, China

doi: 10.12182/20200360504

Abstract:  Objective To investigate the the feasibility and diagnostic efficiencyvalue of different screening indexesmethods for screening primary aldosteronism (PA). Methods The clinical data of 499 patients with PA and 479 patients with essential hypertension diagnosed from Jan. 2009 to Dec. 2018 were retrospectively analyzed. The diagnostic performance of different screening indexs was compared by plotting receiver operating characteristic curves (ROC). Results The area under the ROC curve (AUC) of the plasma aldosterone concentration (PAC) to plasma renin activity (PRA) ratio (ARR) was greater than that of the ratio of the upright PAC to the angiotensin Ⅱ (AT-Ⅱ) (AA2R), upright PRA, upright PAC, supine ARR, and lowest blood potassium (P<0.05). The AUC of logistic regression model was greater than that of upright ARR (96.3% vs. 94.6%, P<0.05). There was no significant difference in AUC between decision tree model and upright ARR (94.1% vs. 94.6%, P>0.05). In the test set, the AUC difference between the logistic regression model and the decision tree model was not statistically significant (96.3% vs. 94.1%, P > 0.05). The calibration curve of the logistic regression model is closer to the 45 ° line, and the consistency between the prediction probability and the actual probability of the logistic regression model is better than that of the decision tree model. Conclusion For the screening of PA, upright ARR seems to be the best single screening index, while AA2R (radioimmunoassay) is not recommended. The diagnostic efficacy of logistic regression model including upright PAC, PRA and lowest blood potassium is better than that of single upright ARR.

  • 原发性醛固酮增多症(primary aldosteronism,PA)是一组醛固酮分泌异常增多导致的疾病,其典型临床特点为高血压伴或不伴低血钾,主要分为肾上腺增生和醛固酮瘤(aldosterone-producing adenoma,APA)两种类型[1]。PA在高血压患者中占5%~10%[2-3],在难治性高血压中占比则高达20%[4]。根据最新的高血压流调数据估算,我国约有1 225万~2 450万PA患者[5]。PA患者发生卒中、冠心病、房颤、心衰、糖尿病、代谢综合征、左室肥厚的风险均显著高于原发性高血压(essential hypertension,EH)患者[6],因此,准确筛查出PA患者并给予正确的治疗对减少其心血管事件及靶器官损害的发生非常重要。

    目前美国内分泌学会指南及中国专家共识均推荐立位血浆醛固酮浓度(plasma aldosterone concentration,PAC)与血浆肾素活性(plasma renin activity,PRA)之比(aldosterone-to-renin ratio,ARR)作为PA筛查指标,但是由于缺乏统一的诊断方案和检测方法,ARR的诊断切点值在不同国家和地区变化很大,推荐的立位ARR切点值范围为20~40 (ng/dL)/(ng/mL·h)[1]。此外,与其他所有的生化检测一样,立位ARR也存在假阳性和假阴性结果[7]。因此,目前关于PA筛查指标的探索是PA临床研究的热点之一。

    本研究根据我们前期在四川大学华西医院2009年01月至2018年12月的住院电子病历系统基础上建立的PA回顾性数据库,拟评估立位ARR、立位PAC与血管紧张素-Ⅱ(AT-Ⅱ)之比(AA2R)、立位PRA、立位PAC、卧位ARR、最低血钾的筛查价值并确定其最佳切点值,并通过构建logistic回归和分类决策树模型探索多个联合指标的筛查诊断价值。

1.   对象和方法
  • 2009年01月01日至2018年12月31日经四川大学华西医院内分泌代谢科确诊的高血压患者978例,其中PA 499例,EH 479例。PA纳入标准(符合以下任意一条):①初筛试验阳性:立位ARR>30 (ng/dL)/(ng/mL·h),或立位ARR>20 (ng/dL)/(ng/mL·h)且PRA<1 ng/(mL·h),或立位ARR>20 (ng/dL)/(ng/mL·h)且PAC>15 ng/dL;且确诊试验(卡托普利试验和/或卧位盐水负荷试验)阳性;②低钾合并PRA低于可检测下限且PAC>20 ng/dL。EH纳入标准:①在未使用降压药物的情况下, 诊室收缩压 (SBP) ≥140 mmHg和 (或) 舒张压 (DBP)≥90 mmHg(1 mmHg=0.133 kPa);②PA初筛试验阴性或PA初筛试验阳性但确诊试验阴性。排除标准:①其他类型继发性高血压,如肾动脉狭窄、先天性肾上腺皮质增生症、嗜铬细胞瘤、库欣综合征、甲状腺功能亢进、甲状腺功能减退、肾素瘤;②肾素-血管紧张素-醛固酮系统检测结果缺失或未检测;③已发生明显慢性肾脏病变的患者。最终,纳入符合标准的PA患者499例,EH患者479例。本研究为回顾性研究,经四川大学华西医院伦理委员会批准(伦理号:2019年审229号)。

  • 收集患者临床资料,包括性别、出生年月、入院时间、出院时间、出院诊断、身高、体质量、平均血压、既往最低血钾水平、本院最低血钾水平、估计肾小球滤过率(estimated glomerular filtration rate,eGFR)、血肌酐、血尿酸、血电解质、立卧位PAC、立卧位PRA、立卧位AT-Ⅱ、肾上腺CT和/或MRI。

  • 立位肾素-血管紧张素-醛固酮系统(RAAS)检测为患者清晨起床后保持非卧位状态(可以坐位、站立或者行走)至少2 h,静坐5~15 min后抽取静脉血。卧位RAAS检测为患者夜间卧床休息7~8 h,清晨卧位状态抽取静脉血。所有血标本均采血后半小时内送检。PRA、AT-Ⅱ和PAC均采用放射免疫法检测。

  • 呈正态分布的计量资料采用$\bar x$±s表示,呈非正态分布的计量资料采用中位数及四分位间距表示。呈正态分布的计量资料两组间的比较采用独立样本t检验,呈非正态分布的计量资料两组间的比较采用Kruskal-Wallis检验。计数资料采用百分比表示,计数资料的比较采用卡方检验或Fisher检验。采用受试者工作特性(receiver operating characteristic,ROC)曲线评价立位ARR、立位AA2R、卧位ARR、最低血钾、立位PRA、立位PAC的诊断效能,并根据约登指数(Youden index,YI)确定最佳阳性切点值。预测模型构建:将数据集随机抽样分组为训练集(80%)和测试集(20%)两部分,利用训练集进行变量筛选,并构建logistic回归模型和分类决策树模型,利用测试集验证模型的准确性及泛化能力;通过绘制ROC曲线和校正曲线比较两种预测模型的区分度和一致性。使用R软件进行数据统计及模型的构建与评估。P<0.05为差异有统计学意义。

2.   结果
  • 本研究共纳入978例患者,男性473例,女性505例,PA 499例,EH 479例。PA组平均年龄为(48.7±11.9)岁,EH组平均年龄为(44.6±14.6)岁,两组差异有统计学意义(P<0.001);PA组女性占比高于EH组女性占比(P<0.05);PA组低钾比例高于EH组低钾比例(P<0.001);PA组肾上腺影像学异常率高于EH组肾上腺影像学异常率(P<0.001)。PA组的收缩压、立位PAC、立位ARR、立位AA2R、卧位PAC、卧位ARR、卧位AA2R、血钠、CO2结合力均高于EH组,差异有统计学意义(P均<0.05),而立位PRA、立位AT-Ⅱ、卧位PRA、卧位AT-Ⅱ、eGFR、血尿酸、最低血钾、血氯、血钙、血镁、血磷低于EH组,差异有统计学意义(P均<0.05)。两组间的舒张压、血肌酐、尿素氮差异无统计学意义(P>0.05)。见表1

    ItemPA group (n=499)EH group (n=479)P
    Age/yr.48.7±11.944.6±14.6<0.001
    Female/case (%)281 (56.30)224 (46.80)0.003
    Body mass index/(kg/m2)24.1±3.3224.8±3.460.001
    Systolic blood pressure/mmHg153±20.5150±19.50.029
    Diastolic blood pressure/mmHg94.8±13.994.4±14.90.639
    Adrenal-imaging abnormal/%86.2051.50<0.001
    Hypokalemia ratio/%86.2036.50<0.001
    Upright PAC*/(ng/dL)26.3 (20.6, 35.5)19.8 (15.2, 25.5)<0.001
    Upright PRA*/(ng/mL·h)0.32 (0.10, 0.75)3.67 (1.88, 7.00)<0.001
    Upright AT-Ⅱ*/(ng/L)56.8 (38.9, 66.0)68.3 (51.9, 84.1)<0.001
    Upright ARR*/((ng/dL)/(ng/mL·h))82.9 (35.4, 254.0)5.48 (3.10, 9.73)<0.001
    Upright AA2R*/((ng/dL)/ (ng/L))0.51 (0.36, 0.75)0.30 (0.22, 0.42)<0.001
    Supine PAC*/(ng/dL)25.7 (18.0, 35.5)15.2 (11.9, 19.9)<0.001
    Supine PRA*/(ng/mL·h)0.10 (0.06, 0.22)1.06 (0.53, 2.49)<0.001
    Supine AT-Ⅱ*/(ng/L)55.4 (37.0, 65.0)57.6 (40.4, 68.5)0.006
    Supine ARR*/((ng/dL)/(ng/mL·h))225.0 (99.6, 490.0)14.4 (6.7, 25.7)<0.001
    Supine AA2R*/((ng/dL)/ (ng/L))0.50 (0.32, 0.79)0.29 (0.22, 0.38)<0.001
    eGFR/(mL/(min·1.73m2))99.2±20.3104.0±19.00.002
    Serum creatinine/(mmol/L)70.5±20.171.2±17.40.539
    Serum uric acid/(mmol/L)324±96.9355±104.0<0.001
    Serum urea nitrogen/(mmol/L)4.85±1.524.89±1.480.649
    Serum sodium/(mmol/L)144±2.64142±2.14<0.001
    Lowest serum potassium/(mmol/L)2.74±0.643.57±0.55<0.001
    Serum chlorine/(mmol/L)103±3.11104±2.650.001
    Serum calcium/(mmol/L)2.20±0.122.26±0.11<0.001
    Serum magnesium/(mmol/L)0.88±0.100.90±0.09<0.001
    Serum phosphorus/(mmol/L)1.02±0.211.11±0.20<0.001
    CO2 combining power26.8±3.7024.6±2.85<0.001
     *Media (P25, P75); PA: Primary aldosteronism; EH: Essential hypertension; PAC: Plasma aldosterone concentration; PRA: Plasma renin activity; AT-Ⅱ : Angiotensin Ⅱ; ARR: The ratio of PAC to PRA; AA2R: The ratio of PAC to AT-Ⅱ; eGFR: Estimated glomerular filtration rate

    Table 1.  Comparison of clinical indicators between PA group and EH group

  • 分别绘制立位ARR、立位AA2R、立位PRA、立位PAC、卧位ARR、最低血钾的ROC曲线(图1),其诊断效能见表2

    Figure 1.  ROC curves of the upright ARR, upright AA2R, upright PRA, upright PAC, upright ARR, and lowest serum potassium

    VariableAUCCut-offSensitivitySpecificityPositive predictive valueNegative predictive value
    Upright ARR95.1%20 (ng/dL)/(ng/mL·h)
    86.2% 96.9% 96.4%87.4%
    Upright AA2R77.5%0.5 (ng/dL)/(ng/L)55.5%84.8%77.5% 66.5%
    Upright PRA93.2%1.2 ng/(mL·h)87.0%87.3% 88.1%87.1%
    Upright PAC71.9%22.6 ng/dL67.7%67.0%68.4% 66.9%
    Supine ARR93.2%41 (ng/dL)/(ng/mL·h)91.6%86.4% 88.2% 90.5%
    Lowest serum potassium84.2%3.2 mmol/L74.7%80.4%80.9%75.6%

    Table 2.  Diagnostic efficacy of upright ARR, upright AA2R, upright PRA, upright PAC, supine ARR, and lowest potassium

    AUC从大到小依次为:立位ARR(95.1%)、立位PRA(93.2%)和卧位ARR(93.2%)、最低血钾(84.2%)、立位AA2R(77.5%)和立位PAC(71.9%),除立位PRA和卧位ARR之间差异无统计学意义外,其余各指标的AUC差异均有统计学意义(P<0.05)。

  • 本研究构建了两种PA的筛查预测模型,分别是logistic回归模型和分类决策树模型。将978例患者随机分组为训练集(80%,784例)和测试集(20%,194例)(表3)。训练集中,PA 400例(51%),EH 384例(49%)。测试集中,PA 99例(51%),EH 95例(49%)。训练集和测试集的13个指标(无缺失值)的差异无统计学意义(P>0.05)。

    ItemTraining set (n =784)Test set (n =194)P
    PA51.00%51.00%1.000
    Age/yr.46.9±13.346.1±14.30.483
    Female50.90%54.60%0.393
    Lowest serum potassium/(mmol/L)3.15±0.713.12±0.780.610
    Upright PAC*/(ng/dL)22.8 (17.7, 29.6)22.8 (16.4, 30.7)0.775
    Upright PRA*/(ng/mL·h)1.17 (0.29, 3.75)1.31 (0.26, 4.11)0.857
    Upright AT-Ⅱ*/(ng/L)61.1 (46.4, 73.9)62.1 (47.4, 77.3)0.333
    Upright ARR*/((ng/dL)/(ng/mL·h))16.2 (5.47, 89.8)16.3 (4.91, 101)0.814
    Upright AA2R*/((ng/dL)/(ng/L))0.39 (0.27, 0.58)0.37 (0.25, 0.59)0.294
    Supine PAC*/(ng/dL)19.6 (14.0, 27.5)18.7 (13.3, 28.4)0.584
    Supine PRA*/(ng/mL·h)0.36 (0.10, 1.10)0.34 (0.10, 1.11)0.369
    Supine AT-Ⅱ*/(ng/L)56.3 (39.3, 66.1)56.6 (37.2, 66.9)0.874
    Supine ARR*/((ng/dL)/(ng/mL·h))56.1 (14.0, 247.0)47.3 (11.9, 204.0)0.336
    Supine AA2R*/((ng/dL)/ (ng/L))0.36 (0.25, 0.58)0.34 (0.24, 0.57)0.555
     *Median (P25, P75); PA: Primary aldosteronism; PAC: Plasma aldosterone concentration; PRA: Plasma renin activity; AT-Ⅱ : Angiotensin ; ARR: The ratio of PAC to PRA; AA2R: The ratio of PAC to AT-Ⅱ

    Table 3.  Comparison of training and test data

  • 该模型拟探索非比值指标对PA的筛查价值,故未将ARR和AA2R纳入模型,而将9个预测变量纳入(性别、年龄、最低血钾、立位PAC、立位PRA、立位AT-Ⅱ、卧位PAC、卧位PRA、卧位AT-Ⅱ)模型,得到变量重要性排序(图2)和变量回归系数(表4),将变量按照重要性排序,前4个依次为:立位PRA、最低血钾水平、卧位PAC、立位PAC(P<0.001)。

    Figure 2.  Logical regression model variable importance ranking diagram

    ItemCoefficientStandard errorZP
    Intercept3.035 81.162.6170.008 9
    Gender∶male−0.160 40.275 6−0.581 90.560 6
    Age0.0050.0110.4540.649 8
    Lowest serum potassium−1.699 70.253 6−6.702 6<0.001
    Upright PAC0.079 30.023 93.311 8<0.001
    Upright PRA−1.046 40.134 7−7.768 5<0.001
    Upright AT-Ⅱ0.004 40.008 70.501 30.616 2
    Supine PAC0.1120.026 64.217 8<0.001
    Supine PRA−0.538 10.230 6−2.333 60.019 6
    Supine AT-Ⅱ0.006 60.010.664 40.506 4

    Table 4.  Regression coefficients for the 9 predictors included in the logistic regression model

    结合变量重要性排序及临床实际操作简便性(立位PAC的检测较卧位PAC方便,可门诊检测),最终纳入最低血钾水平、立位PRA和立位PAC 3个预测变量。这3个变量的回归系数分别为−1.86、−1.3、0.15,P均<0.001(表5)。

    ItemCoefficientStandard errorZP
    Intercept4.883 20.850 35.742 8<0.001
    Lowest serum potassium−1.855 60.240 9−7.702 8<0.001
    Upright PRA−1.300 70.118 3−10.993 5<0.001
    Upright PAC0.151 50.019 87.649 4<0.001

    Table 5.  Regression coefficients for the 3 predictors included in the logistic regression model

    通过交叉验证和绘制ROC曲线,纳入上述3个变量的logistic回归模型在训练集的准确性为91.86%(95%CI : 91.06%~92.61%),敏感性93.28%、特异性90.38%、阳性预测值90.99%,阴性预测值92.81%,ROC曲线面积96.33%(95%CI: 95.81%~96.85%);在测试集的准确性为93.81%(95%CI : 89.44%~96.76%),敏感性89.90%、特异性97.89%、阳性预测值97.80%,阴性预测值90.29%,ROC曲线面积96.30%(95%CI: 93.33%~99.27%)。logistic回归模型在训练集和测试集的AUC差异无统计学意义(P=0.99)。最后根据所构建的logistic回归模型绘制列线图(图3)。

    Figure 3.  Nomogram of logistic regression model for PA screening

  • 该模型欲探索比值指标和非比值指标对PA的筛查价值,故将13个预测变量(性别、年龄、最低血钾、立位PAC、立位PRA、立位AT-Ⅱ、立位ARR、立位AA2R、卧位PAC、卧位PRA、卧位AT-Ⅱ、卧位ARR、卧位AA2R)纳入模型,得到变量重要性分数排序(图4),其中分数最高的前五个变量依次为立位ARR、卧位ARR、立位PRA、卧位PRA、最低血钾水平。通过训练集内部交叉验证得到交叉验证误差最小的模型(图5),该模型最终纳入的预测变量为立位ARR、卧位ARR和最低血钾,决策规则为:立位ARR≥20 (ng/dL)/(ng/mL·h),PA的概率为96%;立位ARR<20 (ng/dL)/(ng/mL·h),卧位ARR≥41(ng/dL)/(ng/mL·h),最低血钾<3.4 mmol/L,PA的概率为68%;立位ARR<20 (ng/dL)/(ng/mL·h),卧位ARR≥41 (ng/dL)/(ng/mL·h),最低血钾≥3.4 mmol/L,PA的概率为12%;立位ARR<20 (ng/dL)/(ng/mL·h),卧位ARR<41 (ng/dL)/(ng/mL·h),PA的概率为5%。

    Figure 4.  Decision tree model variable importance ranking diagram

    Figure 5.  Classification decision tree model

    通过交叉验证并绘制ROC曲线,决策树模型在训练集的准确性为86.78%(95% CI : 86.22%~87.32%),敏感性90.45%,特异性82.94%,阳性预测值84.67%,阴性预测值89.29%,ROC曲线下面积91.64%(95% CI: 91.14%~92.13%);在测试集的准确性92.78%(95% CI:88.19%~96.00%),敏感性88.89%,特异性96.84%、阳性预测值94.70%,阴性预测值89.32%,ROC曲线下面积94.14%(95% CI:90.77%~97.51%)。决策树模型在训练集和测试集的AUC差异无统计学意义(P>0.05)。

  • 在测试集中,logistic回归模型和决策树模型的AUC差异无统计学意义(96.3% vs. 94.1%,P>0.05)(图6)。此外,在测试集中,logistic回归模型的AUC大于立位ARR(96.3% vs. 94.6%,P<0.01),决策树模型和立位ARR的AUC差异无统计学意义(94.1% vs. 94.6%,P>0.05)。

    Figure 6.  Comparison of AUC between logistic regression model and classification decision tree model in ROC curves

  • 图7为两种模型的校准曲线,x轴为模型预测的PA发生概率,y轴为PA实际发生概率;理想曲线为45°参考线。logistic回归模型的校准曲线更接近45°线,表明logistic回归模型的预测概率与实际发生概率的一致性比决策树模型更好。

    Figure 7.  Comparison of logistic regression model and classification decision tree model calibration curves

3.   讨论
  • PA的诊断一直是临床上的难点,其中筛查是非常重要且关键的一步,寻找准确、灵敏、简易的筛查指标一直是研究热点之一。本研究通过对华西医院10年经内分泌代谢科住院确诊的PA和EH患者回顾性数据库的研究,探索了多种筛查指标和两种预测模型(logistic回归模型和分类决策树模型)对PA的筛查价值。

    我们首先通过绘制ROC曲线比较了立位ARR、立位AA2R、立位PRA、立位PAC、卧位ARR、最低血钾对PA的筛查价值并根据约登指数确定了每个指标对应的切点值。其中诊断效能最高的仍然是立位ARR(AUC 95.15%),最佳切点值为20 (ng/dL)/(ng/mL·h),敏感性和特异性分别达到86.17%和96.87%,这一切点值为2016年美国内分泌学会指南推荐的立位ARR的最低切点值 ,低于应用最广泛的切点值30 (ng/dL)/(ng/mL·h)[1]。国内李芳等[8]的研究结果显示立位ARR的最佳切点值为36.61 (ng/dL)/(ng/mL·h),但该研究仅纳入了APA患者,而未纳入肾上腺增生患者,因此不适用于肾上腺增生的PA患者筛查;而PA患者中60%以上是由增生引起的[9]。国内其他几项研究则提示我国人群立位ARR最佳切点值在13.01~27.2 (ng/dL)/(ng/mL·h)[10-12]之间,与本研究的切点值接近,提示我国人群PA筛查的ARR阳性切点值偏低,其最佳切点值应根据不同地区和人群特点确定。

    BAUDRAND等[13]对210例血压正常且PRA<1.0 ng/(mL·h)(并未考虑ARR值)的受试者进行了口服钠盐负荷试验,其中29例(14%)患者确诊试验阳性,而如果以初筛ARR>20 (ng/dL)/(ng/mL·h)作为阳性标准,然后再进行确诊试验,这些血压正常的患者中只有21%(6/29)(或整个210研究人群的3%)可按传统标准诊断为PA。该研究表明以PRA单独作为筛查指标可提高筛查率。但我们的研究结果发现把立位PRA作为单独的筛查指标,其AUC达到了93.2%,仅略低于立位ARR(95.15%)(P<0.05),而把立位PAC作为单独筛查指标则显著低于立位ARR(71.9% vs. 95.15%,P<0.05),甚至低于最低血钾(AUC 84.2%),提示立位PRA有较好的筛查价值,而立位PAC对于PA的筛查价值较小(AUC 71.9%,敏感性67.7%,特异性67.0%)。

    ARR为PAC与PRA之比,其大小主要受分母PRA的影响,在PRA极低的情况下(例如0.1 ng/mL·h),即使PAC低(例如4 ng/dL),ARR也可能升高,但几乎可以肯定这不符合PA诊断。我们的研究也发现立位PRA的诊断效能显著高于立位PAC,表明立位PRA在筛查中起主要作用,但仅以立位PRA或ARR单独作为筛查指标会导致假阳性偏高,而为了避免肾素过低引起的ARR假阳性,有学者提出以ARR联合PAC>15 ng/dL作为筛查指标[9]。2019年梅奥诊所YOUNG教授提出PA筛查阳性的标准为PAC≥10 ng/dL且PRA<1.0 ng/(mL·h)或肾素浓度(PRC)低于参考范围下限[14]。VAIDYA等[15]也提出对于有高血压和/或低血钾的患者,PRA被抑制且PAC>10 ng/dL,则认为初筛阳性,若PAC在5~10 ng/dL之间,则为潜在初筛阳性。但这种PAC和PRA组合(而非两者比值)的筛查方法的准确性还有待进一步的验证。

    近年研究致力于发现更好的PA筛查指标。AT-Ⅱ直接刺激醛固酮的合成,与肾素相比,AT-Ⅱ水平的下降和醛固酮与AT-Ⅱ之比(AA2R)的升高可能更能反映醛固酮的自主分泌,因此AA2R也可能是比较理想的筛查指标[16]。POGLITSCH等[17- 18]在2016年第26届国际高血压学会科学会议上发表的会议摘要表明质谱检测的AA2R可能优于ARR。2020年BURRELLO等[19]的研究表明质谱法检测的AA2R不劣于ARR。我们的研究发现立位AA2R的AUC明显低于其他所有筛查指标,说明其作为PA筛查指标的诊断效能较低。虽然AT-Ⅱ直接刺激醛固酮的分泌,但由于AT-Ⅱ为整个RAAS的重要中间环节,其在血浆中的浓度除了受醛固酮水平的负反馈调节外,还受血管紧张素转换酶和氨基肽酶的调控,因而AT-Ⅱ并不能较好反映醛固酮的自主分泌。此外,目前我院RAAS的检测均采用放射免疫法,检测抗体可能与血管紧张素Ⅰ和AT-Ⅱ的分解产物发生交叉反应,进而影响其检测准确性[20]。这些都可能是本研究中AA2R的准确性较低的原因。因此采用放射免疫法测定AT-Ⅱ计算的AA2R可能并不适合作为PA的初筛检查项目。

    有研究显示以卧位ARR>30 (ng/dL)/(ng/mL·h)且卧位PAC>17.8 ng/dL为初筛阳性条件时,筛查醛固酮瘤的敏感性96.8%,特异性90.5%[21],提示卧位ARR和卧位PAC对醛固酮瘤的筛查有一定价值,同样由于仅纳入了醛固酮瘤患者,该筛查标准并不适合增生患者的筛查。KONG等的研究发现卧位ARR的AUC大于立位ARR,卧位ARR切点值范围波动于20.5~112.06 (ng/dL)/(ng/mL·h),且卧位ARR最佳切点值均高于其对应原始研究中的立位ARR切点值[8, 10, 12, 22- 23]。本研究中卧位ARR的最佳切点值为41 (ng/dL)/(ng/mL·h),明显高于立位ARR的最佳切点值20 (ng/dL)/(ng/mL·h);卧位ARR的诊断效能(AUC 93.2%)与立位PRA(AUC 93.2%)差异无统计学意义(P>0.05),但均稍低于立位ARR(AUC 95.15%,P<0.05)。由于ARR的水平受体位和采血时间等因素的影响[22],导致不同体位下ARR的诊断切点值不同。卧位时肾血流量增加,抑制PRA的分泌,但ALD的分泌受体位影响较小,进而引起卧位ARR较立位ARR高[12, 24]。虽然立位ARR检查方便,门诊即可完成,且诊断效能更高,但立位ARR检查前,患者需保持站位或坐位2 h,采血前再静坐5~15 min,有部分患者可能由于身体原因无法完成上述过程,因此探索卧位ARR最佳切点值,有助于此类患者的筛查;同时对立位ARR阴性的住院疑诊患者,也可以作为立位ARR的补充。

    近年来国内外关于PA分型诊断临床预测模型的研究逐渐增多,包括Kupers评分、改良Kupers评分、Nanba评分、Kamemura评分、Kobayashi评分及中国瑞金医院列线图等[25-35],这些预测模型可在一定程度上辅助临床医师进行PA分型,减少不必要的AVS。此外VELEMA等开发了一种在盐负荷试验结果不确定时的PA确诊预测模型,该预测模型的敏感性为84.4%,特异性为94.3%,阳性和阴性预测值分别为90.5%和90.4%,此预测模型的诊断结果与专家小组的诊断结果高度一致[36]。而目前关于PA筛查的预测模型却鲜有报道,2018年YAMASHITA等通过logistic回归分析建立了PA的PFK筛查评分系统,纳入的指标包括尿PH、女性性别、低钾血症,但该评分系统的AUC仅为0.73[37],临床应用较少。为开发更好的筛查预测模型,本研究利用回顾性的PA数据库,通过R软件的caret包将数据随机分为训练集和测试集,利用训练集筛选预测变量并构建了logistic回归模型和分类决策树模型,然后通过交叉验证评估模型在训练集的准确性,最后通过测试集验证所构建模型对陌生数据预测的准确性和一致性。两种模型在训练集和测试集的诊断效能无显著差异,表明两种模型对陌生数据的预测能力稳定。在测试集中,两种模型的区分度相当(AUC 96.3% vs. 94.1%,P>0.05),但logistic回归模型的一致性优于分类决策树模型。

    Logistic回归模型最终纳入的指标为立位PAC、立位PRA、最低血钾,模型中的三个指标均可在门诊检测,且该模型的诊断效能优于目前广泛应用的立位ARR(96.3% vs. 94.6%,P<0.05),有利于提高PA的正确检出率。通过logistic回归模型绘制的列线图可直接得到患PA的概率,概率越高说明患PA的可能性越大,因此,该列线图不仅可以用于筛查,还可用于确诊,但还需前瞻性的研究进一步验证并确定概率切点值。

    分类决策树模型最终纳入的预测指标为立位ARR、卧位ARR和最低血钾,虽然该模型的诊断效能与立位ARR相当(94.1% vs. 94.6%,P>0.05)且卧位ARR需住院才能检测,而对于立位ARR阴性但怀疑PA的患者,可作为有效的补充筛查指标,以降低漏诊率。

    本研究探索了多种筛查指标和两种预测模型对PA的筛查诊断价值。研究结果显示立位ARR的诊断效能优于立位AA2R、卧位ARR、最低血钾、立位PRA和立位PAC。采用放射免疫法测得的立位AA2R的诊断效能明显低于其他大多数指标,故不推荐作为PA筛查指标。分类决策树模型的诊断效能与立位ARR相当,可作为立位ARR阴性住院患者的补充诊断方法。

Reference (37)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return