Discordance Between Imaging and Adrenal Vein Sampling in Primary Aldosteronism Irrespective of Interpretation Criteria Free

Abstract

Background

Subtyping of primary aldosteronism (PA) using imaging and adrenal vein sampling (AVS) may yield discordant results, causing confusion in management. Interpretation criteria for AVS lateralization may affect discordance rates.

Methods

We identified consecutive patients with PA who underwent AVS at a quaternary care center between January 2006 and May 2018. Patient demographics, laboratory results, diagnostic imaging, and AVS results were retrieved. Adrenal cross-sectional imaging was compared with AVS findings. The presence of lateralization was defined using varying thresholds for the lateralization index (LI) from >2:1 to >5:1. Discordance was defined by a unilateral lesion on imaging with contralateral or nonlateralization on AVS.

Results

A total of 342 patients were included; 68.7% had hypokalemia. With cross-sectional imaging, 191 (55.6%) patients had unilateral lesions, 47 (13.7%) had bilateral lesions, and 104 (30.4%) had normal imaging. Overall discordance rates were high, ranging from 22% to 28% for LI thresholds of >2:1 and >5:1, respectively. Discordance between imaging and AVS was positively correlated with LI threshold stringency (P < 0.001). Patients with normal or bilateral lesions on imaging frequently lateralized on AVS. Lateralization, when present, was approximately equal between left and right sides, irrespective of the LI threshold.

Conclusions

Discrepancies between imaging and AVS were common, even among patients with nonspecific imaging. Discordance was greatest with the strictest AVS interpretation criteria. Even under the most lenient thresholds, apparent discordance between imaging and AVS exceeded 20% and may limit the ability to make surgical decisions. Reliance on imaging alone for detecting lateralization may be misleading.

Primary aldosteronism (PA) is a potentially curable cause of hypertension and is present in up to 20% of individuals evaluated in hypertension clinics (1). Accurate subtyping of PA informs the use of highly effective disease-targeted treatments. Unilateral PA may be amenable to surgery, whereas bilateral forms of disease are treated with mineralocorticoid receptor antagonists. Adrenalectomy in patients with unilateral disease leads to cure or improvement of hypertension in most cases (2), improves quality of life (3), and is cost-effective (4), underscoring the importance of accurate subtyping.

Although adrenal vein sampling (AVS) is widely considered to be the definitive preoperative test for subtype diagnosis (5–8), its use may be limited because of technical challenges and lack of availability. As such, adrenal cross-sectional imaging with CT or MRI is often recommended as the first step for subtype classification owing to its relative accessibility and ease of performance (9, 10). Moreover, despite the paucity of supporting evidence, some experts argue that AVS can be avoided in certain situations where a solitary adrenal nodule is seen on CT/MRI in association with biochemical features of aldosterone excess, with recommendations to proceed directly to adrenalectomy in selected cases (7, 9, 11).

These recommendations are, in part, due to inconsistent reports comparing the performance of CT/MRI with AVS in identifying unilateral aldosterone hypersecretion. Even when a discrete unilateral lesion is detected on CT/MRI, aldosterone hypersecretion may arise from the contralateral side or bilaterally; furthermore, when apparent bilateral disease is seen on imaging, hypersecretion may be demonstrated to be unilateral with AVS. Previous studies examining the rates of discordance between subtyping modalities have been limited by nonstandardized imaging protocols and large differences in AVS interpretation criteria (8). The potential impact of different interpretation criteria on rates of discordance has not been studied. We hypothesized that stricter thresholds for determining lateralization on AVS would result in greater rates of discordance. In other words, application of more stringent lateralization criteria (a condition that is more likely to capture true cases of unilateral aldosterone hypersecretion) would result in greater apparent discrepancies between cross-sectional imaging and AVS and predispose toward a higher likelihood of misclassification by imaging. To address this, we conducted a cohort study of all patients undergoing subtyping for PA through a centralized regional center with a standardized practice and protocol and compared the results of CT/MRI vs AVS.

Patients and Methods

Study population and data sources

We identified 366 consecutive patients who underwent AVS for the workup of PA in Calgary, Alberta, Canada, between January 2006 and May 2018 through the regional AVS database. The study population represented all cases of PA undergoing AVS in Southern Alberta and in the interior of British Columbia, Canada. Cases were excluded if there were confounding adrenal disorders, such as Cushing syndrome or pheochromocytoma, or if cross-sectional imaging was not performed. The study was approved by the regional institutional ethics board.

All referred cases had a clinical diagnosis of PA using published criteria (12, 13). These were based on an elevated aldosterone/renin ratio (ARR) in conjunction with high-probability features of PA (i.e., resistant hypertension, spontaneous or diuretic-induced hypokalemia, and/or an adrenal nodule) (14). For further subtyping, all patients underwent adrenal cross-sectional imaging followed by AVS.

Measures

Patient demographics, laboratory results, diagnostic imaging reports, medications, and AVS results were systematically retrieved.

Renin was determined using plasma renin activity (PRA) from January 2006 to October 2014 and subsequently with direct renin concentration (DRC) from November 2014 to May 2018 (15). PRA was measured by radioimmunoassay with the GammaCoat Plasma Renin Activity I¹²⁵ assay (Diasorin, Stillwater, MN). To avoid overinflation of the ARR, the lowest reporting limit for PRA was set to 0.1 ng/mL/h (14). DRC was measured using the Diasorin Liaison XL platform using a chemiluminescent immunoassay calibrated to the WHO International Standard 68/356. The lowest reporting limit of 1.0 mIU/L was used. Serum aldosterone was measured either by solid-phase radioimmunoassay or chemiluminescent immunoassay. Both assays exhibited imprecision levels <10% across the measuring range and performed similarly. An ARR was considered elevated if >550 pmol/L/ng/mL/h when calculated using PRA or if >60 pmol/L/mIU/L when using DRC (16). Estimated glomerular filtration rate was calculated using the abbreviated Modification of Diet in Renal Disease Equation (17).

Subtyping

All study participants underwent cross-sectional imaging of the abdomen with either CT or MRI. Diagnostic imaging reports were reviewed. Studies were considered abnormal if any discrete nodule, bulkiness, or hyperplasia was noted in the radiology report. For example, if a patient was noted to have a left-sided adrenal nodule and bulkiness in the right adrenal gland, then that patient was considered to have bilateral lesions in the main analysis.

Washout of interfering drugs was attempted prior to ARR measurement and AVS. Most patients were treated with α-blockers, calcium-channel blockers, or direct vasodilators for blood pressure lowering when required. Mineralocorticoid receptor antagonists were discontinued for a minimum of 6 weeks prior to testing. During AVS, samples were collected for cortisol and aldosterone from the left and right adrenal veins as well as from the periphery at baseline and after a 250-µg cosyntropin bolus and a 6.25-μg infusion over 15 minutes. AVS interpretation was based on established criteria (18). First, the selectivity index (SI) was determined by comparing cortisol levels of adrenal veins vs the periphery. Successful cannulation of the adrenal veins was based on a SI ≥2:1 at baseline and/or ≥3:1 after cosyntropin stimulation (18, 19). Second, the lateralization index (LI) was calculated using the aldosterone/cortisol ratio collected at baseline. We applied varying thresholds for the LI, ranging from >2:1 (most lenient) to >5:1 (strictest), comparing the dominant vs nondominant sides to identify the presence of unilateral aldosterone hypersecretion. This range was selected as representative of the most commonly used thresholds among major centers (8, 12, 20). For patients who underwent multiple AVS procedures, only the first successful attempt was considered for the analysis.

Discordance and concordance

Adrenal cross-sectional imaging was compared with AVS findings. Discordance was defined by the presence of a unilateral lesion on imaging with either contralateral or nonlateralization on AVS. Concordance was present when there was a unilateral radiographic lesion with corresponding lateralization on AVS, bilateral lesions with nonlateralization, or normal adrenal imaging with nonlateralization. Finally, when imaging was normal or when bilateral adrenal lesions were seen and lateralization was identified with AVS, the results were considered nonspecific. These categories were selected because of the limitations in cross-sectional imaging to identify microadenomas and unilateral adrenal hyperplasia and its inability to distinguish between aldosterone-secreting adenomas and nonfunctioning adrenal incidentalomas (7, 21, 22). Therefore, our definition of discordance focused on the most clinically relevant cases where imaging and AVS results were completely contradictory and not explained by plausibly undetected adrenal lesions on imaging.

Analysis

Baseline demographic characteristics were reported for the cohort. Discrete data were presented as frequencies. Continuous variables, if normally distributed, were presented as means (with SD) but were otherwise presented as medians (with interquartile ranges). Rates of discordance and concordance between cross-sectional imaging and AVS were determined using varying thresholds for the lateralization index (ranging from >2:1 to >5:1). To account for the possibility of misclassification with imaging, we performed a sensitivity analysis by only including adrenal nodules of ≥1.0 cm in diameter (7, 23), with the contralateral gland appearing normal when defining the presence of a unilateral adrenal nodule and in determining discordance rates. Finally, discordance rates at an LI threshold of >3:1 were assessed across ARR bands (<2 times greater than the laboratory-specific upper limit of normal [ULN], 2 to 2.9 times ULN, 3 to 3.9 times ULN, 4 to 4.9 times ULN, and ≥5 times ULN) and aldosterone bands (<200 pmol/L, 200 to 399 pmol/L, 400 to 599 pmol/L, 600 to 799 pmol/L, 800 to 999 pmol/L, and ≥1000 pmol/L). Cochran Q test was used to compare lateralization rates and discordance and concordance rates for cases across LI threshold categories. A P value <0.05 was considered significant. Analyses were performed using Excel version 1808 (Microsoft, Redmond, WA).

Results

During the study period, 366 patients underwent AVS. Of these, 12 patients were excluded from further analyses for the following reasons: no imaging performed (n = 4), inability to complete AVS or missing AVS data (n = 3), and confounding adrenal disorders (n = 5). Of those not excluded, 342 (96.6%) had technically successful AVS procedures defined by the SI. The final study cohort therefore included 342 patients.

Characteristics of the cohort are shown in Table 1. The mean age of the study participants was 52.1 years, and 58.8% were male. At baseline, patients used a median of two antihypertensive drugs. The median screening ARR was more than 4 times higher than the cut-off used for case detection. Hypokalemia was common, occurring in two-thirds of individuals. When examined with cross-sectional imaging, 191 (55.8%) patients had unilateral lesions, 47 (13.7%) had bilateral lesions, and 104 (30.4%) had normal imaging. When present, unilateral abnormalities were more common on the left (65.4%) than on the right (34.6%). CT was performed in 318 (93.0%) patients, and MRI was performed in the remaining 24 (7.0%).

Successful AVS results are shown in Table 2. Lateralization, when present, was approximately equal between left and right sides across LI thresholds (47.3%:52.7% for >2:1; 46.8%:53.2% for >3:1; 44.9%:55.1% for >4:1; and 45.5%:54.5% for >5:1). As more strict LI criteria were applied, rates of apparent bilateral disease increased (Q₃ = 170.7, P < 0.001).

A comparison of discordance and concordance rates between imaging and AVS results with varying LI threshold is shown in Table 3. Discordance rates between imaging and AVS were positively correlated with LI threshold stringency (Q₃ = 45.2, P < 0.001). Using lenient thresholds of >3:1 and >2:1, the discordance rates were 23.7% and 21.6%, respectively. Using strict thresholds of >5:1 and >4:1, the discordance rates were highest at 28.1% and 26.3%, respectively.

Among patients with discordance, left-sided lesions on imaging were threefold more common compared with right-sided lesions across all LI thresholds. At the most lenient LI threshold of >2:1, over one-third of discordant cases with left-sided lesions on imaging lateralized to the right on AVS. In contrast, among discordant cases, <10% of those with right-sided lesions on imaging lateralized to the left on AVS.

Among individuals with nonspecific imaging results, lateralization was common, occurring in 42 (40.0%) patients with normal imaging and in 36 (76.6%) with bilateral lesions using an LI threshold of >2:1. At an LI threshold of >3:1, lateralization on AVS occurred in nearly one-third (32; 30.5%) of those with normal imaging and in over half of those with bilateral radiographic lesions (24; 51.1%). In those with normal imaging, AVS lateralization to the right was more common than to the left at all LI thresholds.

In a sensitivity analysis, we restricted the definition of a unilateral adrenal nodule to a lesion of ≥1.0 cm in its largest diameter with a normal contralateral gland. Discordance rates were broadly similar to those in the main analysis, and positive correlation with LI threshold stringency was maintained (Q₃ = 47.8, P < 0.001). At an LI threshold of >3:1, discordance rates were not significantly different across ARR and aldosterone bands.

Discussion

Accurate subtyping of PA is crucial in determining which patients may benefit most from adrenalectomy. Direct measurement of aldosterone secretion using AVS is widely considered to be the preoperative reference method to determine lateralization (5–8). Even so, some centers continue to rely on imaging to guide decision-making for individuals who may potentially benefit from curative adrenalectomy but for whom AVS is not readily available to confirm lateralization (7, 9, 11). In the current study, we found that discordance between CT/MRI and AVS was common irrespective of LI threshold. Overall discrepancy rates were high (range, 22% to 28%), suggesting that reliance on cross-sectional imaging alone may frequently result in misclassification and inappropriate surgery. Discordance was often associated with unilateral aldosterone hypersecretion by AVS and, when present, was most commonly lateralized to the right. Similarly, among cases with nonspecific imaging findings, lateralization on AVS was common and was typically to the right. When more stringent lateralization criteria were applied, there was a significant increase in apparent discordance rates.

Our study is consistent with and extends the findings of previous reports. No existing studies have compared differing lateralization criteria with CT/MRI findings, downstream treatment decisions, or clinical outcomes. Of the few studies that have examined the impact of different lateralization criteria on the classification rates of unilateral vs bilateral subtypes of PA, stricter lateralization criteria have consistently been associated with lower rates of apparent lateralization (24, 25). Furthermore, several studies have reported high rates of discordance between CT/MRI imaging and AVS (5–8), but our study shows that a meaningful proportion of such apparent discordance may be an artifact of excessively stringent AVS interpretation criteria. Although there have been no uniform standards for interpreting AVS across studies, reported discrepancy rates, especially when adjusted to our definition, have generally been similar to those observed here (6, 8, 26–28). The SPARTACUS study reported that left-sided lesions were more common on imaging, accounting for around 75% of unilateral lesions (29), which is similar to our observation. We also found that, with normal or bilateral imaging, AVS more commonly lateralized to the right. It is possible that the preponderance of left-sided lesions seen on CT/MRI and a higher rate of lateralization to the right with nonspecific imaging may be related to anatomical differences that potentially obscured the detection of right-sided adrenal adenomas (30). Furthermore, discrepancies between CT/MRI and AVS are likely related to the inability of imaging to distinguish between aldosterone-secreting adenomas and nonfunctioning adrenal incidentalomas (7, 21, 22) as well as the possibility of confounding from cortisol co-secretion (31).

Some experts and major centers have suggested that cross-sectional imaging alone can be used to guide treatment decisions for cases of marked aldosterone excess (e.g., with spontaneous hypokalemia) associated with a unilateral adrenal lesion (7, 9, 11). However, even in this cohort of subjects with high-probability features of aldosteronism, reliance upon CT/MRI alone can result in missed opportunities for potentially curative adrenalectomy for those with nonspecific imaging or the possibility of referral for unnecessary surgery in those with unilateral lesions but biochemically nonlateralizing disease. Accordingly, clinical prediction rules, which have relied heavily upon imaging findings (32), perform poorly in predicting unilateral disease (33–35). Even using the least stringent LI thresholds, discordance was present in up to one-quarter of cases. This is likely because CT/MRI findings do not correlate well with AVS, even when restricted to cases believed to be “typical” for unilateral PA with nodules ≥1.0 cm. Currently, limited alternatives to AVS exist to determine lateralization. ¹¹C-Metomidate positron emission tomography–CT has shown promise as an alternative modality, although its use is limited by cost and access to the necessary equipment and expertise (36). Although AVS may not always be the ideal test (owing to limited accessibility, technical challenges, and differences in interpretation criteria), it remains the best available option for PA subtyping.

There are many strengths of this study. Unlike many prior studies, in examining the agreement between CT/MRI and AVS according to varying interpretation criteria, we obtained data from one of the largest cohorts of patients with PA undergoing subtyping (8) and not from patients selected for unilateral disease. We had complete capture of all radiological procedures and laboratory tests for the region (15). AVS was performed using a standardized and published protocol, with a high rate of technical success (14, 37–39). We used a clinically relevant definition of discordance that affects management decisions.

There were also limitations of this study. First, this was a retrospective study. Therefore, we could not determine whether discordance between cross-sectional imaging and AVS resulted in differences in clinical decisions or outcomes because these data were not available. However, retrospective analysis was necessary because it was the most feasible method to examine varying classification criteria (40). The focus of the current study was to examine preoperative data as it pertains to real-life decision-making in subtyping PA. In a recent analysis conducted at our center, we found that a large proportion of cases of unilateral PA (confirmed by surgery) were missed using strict LI thresholds of >4:1 or >5:1; in contrast, a threshold of >3:1 was associated with very low rates (∼1%) of false lateralization (40). Therefore, at our institution, cases of apparent lateralization using a LI threshold of >3:1 are considered a surrogate (“silver standard”) for unilateral disease. AVS false-negative rates can never be known with certainty because independent verification would require confirmatory pathology, but it would be unethical to refer patients with apparent bilateral disease for surgery. Second, at our center, AVS was routinely offered to all patients desiring potential adrenalectomy (14, 15). Consequently, this cohort of patients may have been more likely to have severe or unilateral forms of PA. Even so, our selection of individuals with high-probability features of PA with clinical indications for AVS can be justified because these represent the types of patients who should receive AVS as per clinical practice guidelines (9); therefore, assessing factors that could affect this decision is relevant. Third, imaging modalities and protocols were not standardized for all patients. Although most subjects underwent CT scanning, 7.0% of patients had MRI. Differences in radiological outcomes were likely minimal because these modalities perform similarly in the evaluation of PA (41). Fourth, we defined the presence of PA solely by an elevated ARR. At our institution, we have long used high-probability features of PA (e.g., ARR >60 pmol/L/mIU/L and hypokalemia) for case detection, even without further conventional confirmatory testing with few (<3%) false-positive results (14, 15), with similar detection rates as other centers (42, 43). Finally, we relied on CT/MRI reports, which were not standardized. Recognition of subtle findings, such as “bulkiness” or “hyperplasia,” may have been different between radiologists. However, accurate identification of discrete nodules of ≥1 cm was less likely to be affected (7, 23), as we also confirmed in our sensitivity analysis. Further study is needed to examine if anatomical differences systematically affected the ability to recognize adrenal nodules on the left compared with the right.

In summary, we observed high rates of apparent discordance between imaging and AVS. This mismatch was more pronounced with stricter LI interpretation criteria. Even under the most lenient thresholds, apparent discrepancies between imaging and AVS exceeded 20% and may confound treatment decisions. Lateralization on AVS was common among those with nonspecific imaging results. These results suggest that cross-sectional imaging cannot be relied upon alone in subtyping PA. If clinical outcomes among surgically treated imaging-negative and imaging-positive patients are similar, then the absence of adrenal abnormality on imaging should not preclude the use of AVS, especially among those with high-probability features, such as hypokalemia or marked biochemical aldosteronism.

Abbreviations:

Abbreviations:
 
• ARR

  aldosterone–renin ratio

 
• AVS

  adrenal vein sampling

 
• DRC

  direct renin concentration

 
• LI

  lateralization index

 
• PA

  primary aldosteronism

 
• PRA

  plasma renin activity

 
• SI

  selectivity index

 
• ULN

  upper limit of normal

Acknowledgments

Disclosure Summary: A.A.L. is supported by the Hypertension Canada New Investigator Award. The remaining authors have nothing to disclose.

References