Within-Patient Reproducibility of the Aldosterone:Renin Ratio in Primary Aldosteronism
The plasma aldosterone concentration:renin ratio (ARR) is widely used for the screening of primary aldosteronism, but its reproducibility is unknown. We, therefore, investigated the within-patient reproducibility of the ARR in a prospective multicenter study of consecutive hypertensive patients referred to specialized centers for hypertension in Italy. After the patients were carefully prepared from the pharmacological standpoint, the ARR was determined at baseline in 1136 patients and repeated after, on average, 4 weeks in the patients who had initially an ARR ≥40 and in 1 of every 4 of those with an ARR <40. The reproducibility of the ARR was assessed with Passing and Bablok and Deming regression, coefficient of reproducibility, and Bland-Altman and Mountain plots. Within-patient ARR comparison was available in 268 patients, of whom 49 had an aldosterone-producing adenoma, on the basis of the “4-corner criteria.” The ARR showed a highly significant within-patient correlation (r=0.69; P<0.0001) and reproducibility. Bland-Altman plot showed no proportional, magnitude-related, or absolute systematic error between the ARR; moreover, only 7% of the values, for example, slightly more than what could be expected by chance, fell out of the 95% CI for the between-test difference. The accuracy of each ARR for pinpointing aldosterone-producing adenoma patients was ≈80%. Thus, although it was performed under different conditions in a multicenter study, the ARR showed a good within-patient reproducibility. Hence, contrary to previously claimed poor reproducibility of the ARR, these data support its use for the screening of primary aldosteronism.
Recent compelling evidence indicates that primary aldosteronism (PA) is associated with prominent cardiovascular and renal damage, adverse metabolic consequences, and an excess rate of cardiovascular events.1–3 In the Primary Aldosteronism Prevalence in hYpertension (PAPY) Study, PA was diagnosed in 11.2% of 1125 newly diagnosed hypertensive patients referred to hypertension centers; moreover, in 4.8% it was attributed to an aldosterone-producing adenoma (APA).4 Therefore, it is likely that PA represents the most common form of endocrine arterial hypertension that is surgically curable in a substantial proportion of the cases.
Currently the plasma aldosterone concentration (PAC):renin ratio (ARR) is the most popular screening test for identifying PA,4,5 and although it has obvious limitations,6 its widespread adoption can enhance the identification of PA.7 Nonetheless, the within-patient reproducibility of this test remains unknown.8
We report here the results of the study of within-patient comparison of the ARR that was determined at the baseline screening and repeated on average after a month.
Subjects and Methods
All of the patients originally recruited in the PAPY Study and 11 additional patients who were investigated with an identical protocol at the internal medicine department of the General Hospital of Legnano were included in this study. The PAPY Study protocol followed the Statement for Reporting Studies of Diagnostic Accuracy recommendations9 and the requirements of the Declaration of Helsinki.4 Briefly, consecutive newly diagnosed hypertensive patients, referred to specialized hypertension centers nationwide in Italy, were recruited after informed consent was obtained.4 A previous diagnosis of a secondary form of hypertension and the patient’s refusal to participate in the study were the only exclusion criteria. The general flowchart of the study protocol4,10 for the purpose of this study is summarized in Figure S1 (please see the online Data Supplement at http://hyper.ahajournals.org). Briefly, the patients were prepared from the pharmacological standpoint as follows4: drugs interfering with the renin-angiotensin-aldosterone system were withdrawn for ≥2 weeks (for angiotensin-converting enzyme inhibitors, angiotensin II type 1 receptor blockers, β-blockers, and diuretics) or 6 weeks (mineralocorticoid receptor antagonists) before performing the screening test and until the saline infusion test, if required by protocol. A long-acting calcium channel blocker (CCB) and/or doxazosin were prescribed whenever necessary to minimize the risks of uncontrolled hypertension and continued up to the saline infusion test.
At baseline screening, all of the patients underwent measurement of the 24-hour Na+ urine excretion and, the next morning, while they remained in the sitting position, of plasma renin activity (PRA), PAC, and plasma cortisol concentration (PCC) before and after a captopril challenge with 50 mg of oral captopril.10 In those with an ARR ≥40 and in 1 of every 4 of those with an ARR <40, the PRA, PAC, and PCC measurements were repeated before the saline infusion test, after, on average, 4 weeks, during which the treatment remained unchanged. For calculation of the ARR, the lowest value of the denominator, PRA, was set to 0.2 ng · mL−1 · h−1 to avoid overinflating the ARR.
The further diagnostic workup was based on the results of ARR baseline and/or after the captopril and/or the logistic discriminant function test.11 The patients positive at such tests were submitted to an imaging test for identification of adrenocortical nodules4 and also to adrenal vein sampling (AVS)12 or to dexamethasone-suppressed adrenocortical scintigraphy to identify a lateralized aldosterone excess production. AVS was used to diagnose lateralization of hyperaldosteronism only if bilaterally selective4; ACTH stimulation was not systematically used during AVS because, although it improves assessment of selectivity of catheterization, it does not enhance the diagnostic accuracy.13,14
Serum creatinine, serum and urine Na+ and K+ levels, PRA, PAC, PCC, and estimated glomerular filtration rate were measured as described.4 Normal ranges, intra-assay and interassay coefficients of variation, and antibody cross-reactivity for the hormonal measurements have already been reported.4
Identification of APA required all of the following criteria: (1) evidence of PA at the screening test as defined above; (2) lateralization of aldosterone secretion at AVS or at NP59 adrenocortical scintigraphy; (3) evidence of adenoma at pathology; and (4) demonstration of normokalemia and cure or improvement of hypertension at follow-up after adrenalectomy by reported criteria.4 Patients with biochemical evidence for PA but without conclusive evidence for a lateralized aldosterone excess were presumed to have idiopathic hyperaldosteronism (IHA).
Skewed variables as PRA, PAC, and PCC were analyzed after achievement of a normal distribution by log transformation. One-way ANOVA followed by the Bonferroni posthoc test was used to compare quantitative variables across groups. Distribution of categorical variables was investigated by χ2 analysis. To assess the reproducibility of the ARR, several techniques were used, including correlation and regression analyses and Bland-Altman and Mountain plots.15,16 For regression, the Passing and Bablok procedure was preferred to the ordinary linear regression method, because it makes no assumptions regarding the distribution of the samples and the measurement errors.17 In this analysis, a test for linearity is used to evaluate how well a linear model fits the data; moreover, the result does not depend on the assignment of the variables to the x and y axes. To confirm results, Deming regression was also used, because it takes measurement errors for both methods into account.18 Other than furnishing summary statistics for each ARR, both of these regression techniques use a test for linearity to evaluate how well a linear model fits the data and furnish a scatter plot with regression line and estimate intercept A and slope B along with their 95% CI. The CI is used to challenge the null hypothesis that A=0 and B=1. For the intercept A, the null hypothesis is accepted if the CI for A contains the value 0; rejection of this hypothesis leads to the conclusion that the 2 methods differ at least by a constant amount. For the slope B, the null hypothesis is accepted if the CI for B contains 1; rejection of this hypothesis implies that there is a proportional difference between the methods. To identify proportional error, magnitude-dependent bias, and systematic error, the Bland-Altman and Mountain plots were used.
Finally, to assess and compare the accuracy between ARR for identifying APA we also used receiver operator characteristics (ROC) curves10 with the MedCalc software (version 184.108.40.206, MedCalc Software). For all of the other analyses, the SPSS for Windows (version 17.0, SPSS Inc) was used; significance was set at P<0.05.
There were 268 patients with complete ARR data on the 2 occasions and with a conclusive diagnosis available for this study. The baseline features of the patients, divided by diagnosis (Table 1), did not differ for demographic, clinical, and hormonal data from the total PAPY Study cohort (all P values not significant).
No differences in age, body mass index, diastolic blood pressure (BP), serum creatinine, estimated glomerular filtration rate, and urine Na+ and K+ excretions were found between PA and primary hypertension (PH) patients. The APA patients had higher systolic BP and lower serum K+ levels than the PH patients. The 74 patients presumed to have IHA did not differ from the APA patients for baseline diastolic BP but had higher serum K+ than the APA patients.
At the time of the determinations of either ARR, 40.1% of the patients were untreated, 34.7% were on a CCB, and 25.2% were on a CCB and doxazosine. Significantly more APA than IHA or PH patients required the combination treatment to achieve BP control.
Hormonal Data at the First and Second ARR
On the occasion of both the first and the second hormonal measurements, both the APA and the IHA patients had elevated PAC, suppressed PRA, raised ARR, and lower serum K+ than the PH patients. Of note, from the first to the second assay, there was a decrease of PAC and the ARR, which reached significance in the IHA and the PH groups but not the APA group. This decrease was not accompanied by significant changes of PRA in any of the PA groups, although it was associated with a significant, albeit clinically negligible (0.06 ng · L−1 · h−1), increase in the PH group. In all of the groups, the decrease of PAC and ARR was paralleled by a significant fall of PCC.
Reproducibility of the ARR
When examining the relationship between first (ARR-1) and second measured (ARR-2) ARRs, we found a high within-patient correlation (R=0.69 [95% CI: 0.63 to 0.74]; P<0.0001) and reproducibility (coefficient of determination: 0.47). As both measurement values showed a skewed distribution, we applied the Passing and Bablok regression method. Figure 1 shows the scatter plot and regression line with 95% CI (Figure 1A). The 95% CI for the intercept A (−3.75 to 0.34) and the slope B (0.81 to 1.01) of the calculated regression equation comprised of 0 and 1 values, respectively. Thus, the ARR-1 and the ARR-2 did not differ by either a constant amount or a proportional difference. Identical results were obtained with the Deming regression, which considers the variance of the errors for both ARR-1 and ARR-2 (data not shown).
The Bland-Altman plot showed no evidence for proportional, magnitude-related, or absolute systematic errors between the 2 ARRs (Figure 1B; for interpretation please also refer to Figure S2). A Bland-Altman plot of the ARR-1:ARR-2 ratio, instead of their log-transformed differences on the y axis, showed that the mean value was close to 1, and most values were within the agreement interval, with only 7% of the values corresponding to a proportion close to that expected by chance, falling out of the 95% CI (Figure 1C). A scrutiny of the individual data of these PH patients (empty circles) revealed that they were attributed to PRA values above the normal range at the screening test, which had normalized on the occasion of the second test after 4 weeks. The Mountain plot was centered over 0 and showed no long tails, thus confirming that the 2 assays were unbiased (Figure 1D).
Diagnostic Accuracy of the 2 ARR
The ROC curves for ARR-1 and ARR-2 (Figure 2) were remarkably similar and their areas under the ROC curve (AUCs), which were both (P<0.001) >0.50, did not differ significantly, indicating that the accuracy of the 2 ARRs was similar. Consistent with this finding, no single APA patient who was pinpointed with the first ARR was missed with the second and vice versa.
We may be facing an epidemic of PA, and several studies suggest that we should screen all hypertensive patients for PA to prevent cardiovascular and metabolic complications.6,8,19 To this end, the availability of a screening test capable of providing an accurate identification of PA patients in the huge population of hypertensive patients would be essential. This test should be endowed with a high sensitivity, in order not to miss cases of APA, coupled with a decent specificity, to minimize false-positive results. Moreover, it should ideally provide reproducible results to give reliable information for confirming or excluding the diagnosis. Currently, the ARR is the most popular screening test for PA, although it represents in essence a “crude” bivariate analysis, which can furnish practically identical results starting from very different absolute PAC and PRA values.11,20–22 Although some empirical solutions to this potential pitfall have been proposed,6 the crucial issue of the reproducibility of the ARR has never been systematically addressed.
In the PAPY Study, the ARR had to be measured twice in a substantial proportion of the patients: those who tested positive by predefined criteria at the first ARR, and 1 of every 4 consecutive patients who tested negative. This design allowed a within-patient comparison of the 2 ARRs with exclusion of any bias associated with interindividual variability. Moreover, as antihypertensive drugs profoundly affect the ARR,23 all of the patients had to be carefully prepared pharmacologically to eliminate untoward influences of the antihypertensive treatment on the first measurement of the hormones and were thereafter maintained on the same treatment up to the second ARR.
Reproducibility of the ARR
The tight direct relationship between the 2 ARRs despite the fact that they were performed 4 weeks apart suggests that the ARR furnishes consistent results (Figure 1A) when performed under standardized conditions in the same patient.
The between-test, within-patient pairwise comparison showed a fall of PAC that translated into a decrease of the ARR in the IHA and PH but not in the APA group (Table 2). Because this decrease occurred despite no change in drug treatment and unchanged PRA, but was closely mirrored by a decrease of PCC in all of the groups, it was most likely related to the attenuation of the stress reaction to blood testing occurring from the first to the second blood drawing.
Importantly, the lack of PAC and the ARR decrease in the APA patients not only accords well with the view that aldosterone secretion is autonomous in this condition, but also suggest that a raised ARR on 2 separate occasions is a hallmark of APA. Hence, it would indicate the need to proceed to performing AVS without undertaking an exclusion test.24
To detect potential biases in the ARR measurement, as a proportional error, dependency of one ARR on the magnitude of the measure of the other and/or an absolute systematic error, Bland-Altman plots were used (for explanation, please refer to Figure S2). This showed no evidence for such types of errors (Figure 1B). The Mountain plot further confirmed that the 2 ARRs were unbiased, inasmuch as the mountain was centered on the 0 value.
At Bland-Altman plots, by far the majority of the values fell within the agreement interval and only very few, practically, those expected by chance, fell outside of this interval. A scrutiny of these individual data showed that the PRA was above the normal range in these outliers at the screening test but normalized when measured in the second one. Thus, in a tiny percentage of the patients, notwithstanding a strict adherence to the study protocol, a “carryover” effect of the previous antihypertensive treatment might occur. Therefore, >2 weeks of withdrawal of drugs stimulating renin secretion could be necessary in some patients to achieve unbiased ARR determination. Hence, these results overall provide evidence that the ARR furnishes reproducible results when performed under standardized conditions in adequately prepared patients.
Diagnostic Accuracy of the ARR and Clinical Implications
When comparing biochemical tests, the crucial point that separates clinical meaning from “pure” biochemistry25 is to establish whether the degree of agreement between the ARR is acceptable from the clinical standpoint. The AUC provides an estimate of test accuracy.10 Hence, to assess the diagnostic performance of the first and the second ARRs, we measured the AUC of both ROC curves, using the diagnosis of APA as a reference, because a conclusive diagnosis is feasible only for this subgroup of PA patients. The findings that both of the AUCs were close to 0.80 indicate that the 2 ARRs have an 80% accuracy.4,26 Moreover, the lack of a significant difference between the AUCs of the ARRs indicates a similar accuracy for identifying APA. In fact, all of the APA patients pinpointed with the first ARR were identified with the second and vice versa.
Treatment Effects and Accuracy of the ARR
It has been described that the long-acting CCB amlodipine would decrease PAC and, therefore, the ARR in PA patients,23 which is difficult to reconcile with the view that aldosterone secretion is mainly regulated via T-type Ca channels, as discussed later. At the first ARR of our patients on a CCB, 95% received nifedipine gastrointestinal therapeutic system or amlodipine. We found no treatment effect on hormone values in the majority of the patients with the aforementioned caveats. The changes occurring from the first to the second ARR were likely attributable, in our view, not to CCB treatment but rather to the waning of either a drug carryover effect and/or of a stress reaction. Because they occurred also in the untreated patients (data not shown), the possibility of a regression toward the mean cannot, however, be dismissed. Thus, these results further support the conclusion that a long-acting CCB and/or doxazosine do not markedly influence the PAC and PRA values, and, thus, the ARR, at least during the 4-week period that elapsed between the first and the second ARR. From the practical standpoint, this implies that these agents should be used before (and during re-)assessment of the ARR to avoid the risks of uncontrolled hypertension.
Limitations of the Study
The patients of the PAPY Study were referred to specialized hypertension centers where PAC and PRA could be measured after strict quality control criteria. Furthermore, the ARR was performed only after the withdrawal of interfering drugs. Thus, it remains to be determined whether our findings could be directly extrapolated to the general population of hypertensives. Data obtained in the Bussolengo Study, which was performed in the setting of general practice, would, however, support the feasibility and usefulness of measuring the ARR at this setting,27 provided that the patients are adequately prepared from the pharmacological standpoint.
In this study, the ARR was found to provide reproducible results and a reasonably accurate identification of the surgically curable forms of PA, like APA, when performed under carefully controlled conditions in patients adequately prepared from the pharmacological standpoint. This test is no more expensive than a lipid profile, but its result can have a much greater impact (than lipid profile) on the lifetime cardiovascular risk of those hypertensives who have a surgically curable subtype of PA.28 Hence, it should be offered, in our view, to all of the hypertensives who are candidates for adrenalectomy.6
CCBs, including benidipine, azelnidipine, and efonidipine, which block both L- and T-type Ca channels, were reported recently to inhibit angiotensin II- and KCl-induced expression of steroidogenic enzymes, including CYP11B2 (aldosterone synthase), in vitro.29 They also suppressed steroid biosynthesis in H295R cells, whereas nifedipine, an L-type CCB, was ineffective.30 Replacement of amlodipine, another L-type CCB, with efonidipine was associated with a significant decrease of both PAC31 and left ventricular mass index in essential hypertensive patients,32 indicating that the T-type Ca channels mediate aldosterone production and ensuing left ventricular hypertrophy. It was also reported recently that the dihydropriridine CCBs could have a mineralocorticoid receptor antagonistic effect,33 suggesting that they can be particularly effective in PA patients. It can, therefore, be anticipated that CCBs will be widely used in hypertensive patients during the screening for PA; hence, future work should be aimed at assessing the effect on the ARR of CCBs with and without T-type Ca channels and with and without mineralocorticoid receptor blocking properties.
The list of participating centers and PAPY study investigators are as follows: Padova, Italy, DMCS Internal Medicine 4, G.P.R., Andrea Semplicini, and A.C.P.; Padova, Italy, Endocrinology, F.M., Decio Armanini, Giuseppe Opocher, and Mee-Yung Mattarello; Ancona, Italy, Endocrinology, G.G., Vanessa Ronconi, and Marco Boscaro; Reggio Emilia, Italy, Azienda Ospedaliera ASMN di Reggio Emilia Internal Medicine, E.R.; Pisa, Italy, Internal Medicine, G.B. and Angelica Moretti; L'Aquila, Italy, Department of Internal Medicine and Public Health, C.F. and G.D.; Palermo, Italy, Internal Medicine, Giuseppe Andronico and Giovanni Cerasola; Brescia, Italy, Internal Medicine, D.R., Enzo Porteri, and Enrico Agabiti-Rosei; Legnano, Italy, Internal Medicine, G.P., Carlo Costantini, and Maria Teresa Lavazza; Roma, Italy, Internal Medicine, C.L. and Chiara Caliumi; Trieste, Italy, Internal Medicine, B.F.; Firenze, Italy, Endocrinology, M.Man. and Gabriele Parenti; Torino, Italy, Endocrinology, M.Mac. and Ezio Ghigo; Reggio Calabria, Italy, Nephrology, F.M., G.C., and Carmine Zoccali; Bari, Italy, Internal Medicine, A.B.
Sources of Funding
This study was supported by research grants from the Foundation for Advanced Research in Hypertension and Cardiovascular Diseases, the Società Italiana dell'Ipertensione Arteriosa, and Progetti di Ricerca di Interesse Nazionale 2007PRELCC_002. Moreover, the study was supported by the research grants 60% to A.C.P., G.P.R., T.M.S. and Progetto di Ateneo to T.M.S.
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
- Received July 25, 2009.
- Revision received August 20, 2009.
- Accepted October 21, 2009.
Rossi GP, Bernini G, Caliumi C, Desideri G, Fabris B, Ferri C, Ganzaroli C, Giacchetti G, Letizia C, Maccario M, Mallamaci F, Mannelli M, Mattarello MJ, Moretti A, Palumbo G, Parenti G, Porteri E, Semplicini A, Rizzoni D, Rossi E, Boscaro M, Pessina AC, Mantero F; for the PAPY Study Investigators. A prospective study of the prevalence of primary aldosteronism in 1,125 hypertensive patients. J Am Coll Cardiol. 2006; 48: 2293–2300.
Funder JW, Carey RM, Fardella C, Gomez-Sanchez CE, Mantero F, Stowasser M, Young WF Jr, Montori VM; for the Endocrine Society. Case detection, diagnosis, and treatment of patients with primary aldosteronism: an endocrine society clinical practice guideline. J Clin Endocrinol Metab. 2008; 93: 3266–3281.
Kaplan NM. Is there an unrecognized epidemic of primary aldosteronism? Con. Hypertension. 2007; 50: 454–458.
Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig LM, Moher D, Rennie D, de Vet HC, Lijmer JG; for the Standards for Reporting of Diagnostic Accuracy. The STARD statement for reporting studies of diagnostic accuracy: explanation and elaboration. Clin Chem. 2003; 49: 7–18.
Rossi GP, Belfiore A, Bernini G, Desideri G, Fabris B, Ferri C, Giacchetti G, Letizia C, Maccario M, Mallamaci F, Mannelli M, Palumbo G, Rizzoni D, Rossi E, Agabiti-Rosei E, Pessina AC, Mantero F; for the Primary Aldosteronism Prevalence in Italy Study Investigators. Comparison of the captopril and the saline infusion test for excluding aldosterone-producing adenoma. Hypertension. 2007; 50: 424–431.
Rossi GP, Sacchetto A, Chiesura-Corona M, De Toni R, Gallina M, Feltrin GP, Pessina AC. Identification of the etiology of primary aldosteronism with adrenal vein sampling in patients with equivocal computed tomography and magnetic resonance findings: results in 104 consecutive cases. J Clin Endocrinol Metab. 2001; 86: 1083–1090.
Seccia TM, Miotto D, De Toni R, Pitter G, Mantero F, Pessina AC, Rossi GP. Adrenocorticotropic hormone stimulation during adrenal vein sampling for identifying surgically curable subtypes of primary aldosteronism: comparison of 3 different protocols. Hypertension. 2009; 53: 761–766.
Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res. 1999; 8: 135–160.
Linnet K. Evaluation of regression procedures for methods comparison studies. Clin Chem. 1993; 39: 424–432.
Rossi GP, Pessina AC, Mantero F; for the Primary Aldosteronism Prevalence in Italy Study Investigators. More fuel to the debate on the “epidemics of primary aldosteronism.” Hypertension. 2008; 51: e1–e2.
Schwartz GL, Turner ST. Screening for primary aldosteronism in essential hypertension: diagnostic accuracy of the ratio of plasma aldosterone concentration to plasma renin activity. Clin Chem. 2005; 51: 386–394.
Seiler L, Rump LC, Schulte-Monting J, Slawik M, Borm K, Pavenstadt H, Beuschlein F, Reincke M. Diagnosis of primary aldosteronism: value of different screening parameters and influence of antihypertensive medication. Eur J Endocrinol. 2004; 150: 329–337.
Mulatero P, Rabbia F, Milan A, Paglieri C, Morello F, Chiandussi L, Veglio F. Drug effects on aldosterone/plasma renin activity ratio in primary aldosteronism. Hypertension. 2002; 40: 897–902.
Schirpenbach C, Seiler L, Maser-Gluth C, Rudiger F, Nickel C, Beuschlein F, Reincke M. Confirmatory testing in normokalaemic primary aldosteronism: the value of the saline infusion test and urinary aldosterone metabolites. Eur J Endocrinol. 2006; 154: 865–873.
Twomey PJ. How to use difference plots in quantitative method comparison studies. Ann Clin Biochem. 2006; 43: 124–129.
Akizuki O, Inayoshi A, Kitayama T, Yao K, Shirakura S, Sasaki K, Kusaka H, Matsubara M. Blockade of T-type voltage-dependent Ca2+ channels by benidipine, a dihydropyridine calcium channel blocker, inhibits aldosterone production in human adrenocortical cell line NCI-H295R. Eur J Pharmacol. 2008; 584: 424–434.
Dietz JD, Du S, Bolten CW, Payne MA, Xia C, Blinn JR, Funder JW, Hu X. A number of marketed dihydropyridine calcium channel blockers have mineralocorticoid receptor antagonist activity. Hypertension. 2008; 51: 742–748.