11C-metomidate PET CT versus Adrenal Vein Sampling for diagnosing surgically curable primary aldosteronism: prospective test validation, and impact of somatic genotype and ethnicity on outcomes


 Primary aldosteronism (PA) due to a unilateral aldosterone-producing adenoma (APA) is a common, curable cause of hypertension, but invasive methods of diagnosis and treatment contribute to <1% of patients being offered the chance of cure. The primary objective of our prospective within-patient study in 143 patients with PA was to compare accuracy of 11C-metomidate (MTO) PET-CT scanning with adrenal vein sampling (AVS) in predicting biochemical cure from PA and resolution of hypertension. Secondary outcomes addressed heterogeneity of underlying pathogenesis and prediction of patients most likely to achieve complete cure of hypertension. 128 patients reached 6-9 month follow-up. 77/78 surgical patients achieved one or more of the four hierarchically analysed Primary Aldosteronism Surgical Outcome (PASO) criteria for biochemical and clinical success. MTO was not superior to AVS but all four differences in accuracy favored MTO, with 95% CIs >-17%, the pre-specified margin of non-inferiority. The best univariate predictors of complete clinical cure were home systolic blood pressure (SBP) <135 mmHg after one month of spironolactone 100 mg daily (odds ratio 13.0 (3.72, 45.24) p<0.001) and KCNJ5 genotype of the APA (odds ratio 10.37 (2.50, 42.99) p=0.001). The latter remained significant in logistic regression on age, gender, ethnicity, and was itself predicted by elevated urine 18-hydroxycortisol:cortisol ratio. Our findings validate 11C-metomidate PET-CT for accurate, non-invasive detection of patients with unilateral PA, and identify patients most likely to benefit from adrenalectomy.


Introduction
Primary aldosteronism (PA) is a high morbidity cause of hypertension, resulting from excessive, renin-independent aldosterone secretion from one or both adrenal glands.
Atrial fibrillation (AF), stroke, heart failure and renal insufficiency are approximately twice as likely in PA compared to age-and sex-matched individuals with essential hypertension. [1][2][3][4] PA can often be surgically cured. Traditionally, patients are stratified into those in whom excess aldosterone is consequent upon a unilateral aldosteroneproducing adrenal adenoma (APA) and those with bilateral production (often termed idiopathic hyperaldosteronism, IHA). The binary distinction underpins clinical decision making despite the more recent appreciation that a continuum of disease is more likely; 5,6 current international consensus is that patients with APAs should be offered a laparoscopic adrenalectomy and patients with IHA should be treated with aldosterone antagonist drugs. 7 Despite the prevalence and penalty of PA, currently fewer than 1% of patients are identified and fully investigated. 8 Many reasons exist for this shortfall in clinical care, including low levels of clinical suspicion; time-consuming and imperfect confirmatory biochemical tests which require manipulation of many commonly prescribed antihypertensive medications; difficulties with patient selection for surgical intervention; and uncertainty about outcomes in patients hoping for surgical cure. The paradoxical consequence is patient and physician reluctance to embark on a diagnostic and therapeutic process with a high likelihood of prolonging life while reducing need for medication.
The 'MATCH' ('Is Metomidate superior to Adrenal vein sampling in predicting ouTCome from adrenalectomy in primary Hyperaldosteronism') study, addresses selection of patients for surgery, and the reasons for variable outcomes from intervention. The investigation and treatment of apparent APAs have not substantially changed in the past 30 years. 9,10 The two invasive interventions, with limited availability, namely AVS and surgery, might have been appropriate when PA was considered rare. PA is now known to cause 5-10% of unselected cases of hypertension, 11 which necessitates alternative strategies. Molecular imaging has the potential both to increase the number of PA patients who can be fully investigated and to elucidate which patients might benefit most from surgical intervention. 11 Cmetomidate (MTO) was developed as a tracer in the 1990s, after the anesthetic agent etomidate, the ethyl analogue of MTO, was discovered to inhibit cortisol production. 12,13 MTO was readily C 11 H3-labelled and in pre-clinical and clinical studies, the tracer concentrated in the adrenal. [14][15][16] However, ex vivo it bound, with similar potency, to the enzymes CYP11B1 (11beta-hydroxylase) and CYP11B2 (aldosterone synthase). 17 This discouraged evaluation in PA, where selective binding to CYP11B2 is required, in order to detect a higher signal from an APA than from either the normal adrenal, or an incidental non-aldosterone secreting adenoma. In 2012 we reported that in vivo selectivity could be achieved, if patients were pre-treated with dexamethasone for 3 days, which suppresses adrenal CYP11B1 (but not CYP11B2) protein expression. In 35 patients with PA, in whom a diagnosis of unilateral or bilateral PA had been reached by AVS, dexamethasone-suppressed MTO had 73% sensitivity and 87% specificity; assuming AVS was correct, and using a ratio of SUVmax between adenoma and contralateral adrenal of >1.25:1. 18 MTO has been adopted by some international centers, but its evaluation without dexamethasone suppression may have limited its applicability. 19 In order for MTO to enter routine clinical practice, as an alternative to AVS, a definitive, adequately-powered study was required in which the primary objective was to prospectively compare the accuracy of AVS and MTO PET-CT scanning in predicting the outcome of adrenalectomy, in patients with PA.
Guidelines to clinical practice in PA are based largely on expert opinion and retrospective analyses. The one randomised control trial to date, comparing AVS with CT, has not substantially changed practice, possibly because neither study group was deemed to follow standard guidance. 20 The MATCH study protocol mandated that every patient be investigated identically, according to international guidelines, with a pre-specified follow up schedule. Patients were clear of interfering medications at baseline and primary outcome visits, so that all outcomes of intervention could be accurately quantified, e.g. fall in aldosterone, and N-terminal pro-B-type natriuretic peptide (BNP) as a measure of cardiac function. In addition to standard investigations, every patient had a dexamethasone-suppressed MTO PET-CT, and a nested 'studywithin-study' prospective evaluation of spironolactone response as predictor of the clinical outcome from surgery.
The higher likelihood of biochemical than clinical cure after adrenalectomy 21 enabled the adoption of hierarchical primary endpoints, as used in previous clinical trials. [22][23][24] Biochemical outcome is a purer measure of accuracy in diagnosing unilateral disease, whilst cure of hypertension is of more concern to patients. Most patients with a unilateral APA will not be able to discontinue all antihypertensive medication, the mean reduction in pill burden being 50% at six months post-operatively. We explored whether, in addition to secondary, irreversible vascular changes, 6 undiagnosed contralateral disease may contribute to incomplete cure. 25,26 Clinical conservatism in recent decades has contrasted with advances in the molecular analyses of APAs.
MATCH, therefore, included pre-specified evaluations of whether variability in outcomes are related to genotype or transcript variations among APAs, and whether informative variations can be predicted in advance of surgery.

Methods
The study design is represented graphically in Fig. 1a  After the second investigation participants attended a 'baseline' visit for measurements of all clinical and biochemical parameters, whose response to treatment would assess the accuracy of MTO and AVS. Spironolactone therapy (or, if previously not tolerated, eplerenone) was initiated at 50 mg daily, and up-titrated to 100 mg daily after 2 weeks if tolerated. The home blood pressure response to spironolactone was measured 4 weeks after initiation of therapy. Participants who underwent adrenalectomy were followed up at 1, 3, 6, 12 and 24 months post-surgery while those managed medically were followed up at 9-12 and 24 months after initiation of spironolactone. The primary endpoint for the surgical group was assessed at 6 months post-surgery, and at 9-12 months after initiation of spironolactone for the medical group.
The primary objective of MATCH was to compare the accuracy of MTO and AVS in diagnosing unilateral PA. Accuracy was defined as each investigation's ability to successfully predict biochemical and clinical cure, as defined by the PASO consensus (Supplementary Table 1a), 21 at 6 months post-surgery. Since the rank order of partial and complete cures of PA and hypertension could be assumed (with more patients being biochemically than clinically cured), prospective studies in PA lend themselves to hierarchical analysis of each PASO outcome, in which all four are primary outcomes without needing to adjust for multiple statistical comparisons (Fig. 1b). Details of the study design, statistical plan, primary and secondary analyses are described in the online supplementary methods.

Primary Outcomes
In the full analysis set, MTO and AVS were graded 'high' probability in, respectively, 74/141 (52%) and 63/141 (45%) patients (Fig. 2a). Unilateral PA was diagnosed by one or both investigations in 86/134 patients (64%) (Fig. 2b). MTO and AVS were concordant in 38/89 (43%) of these unilateral cases, and agreed also in 20/45 (44%) of the bilateral cases. It is apparent that the higher than usual proportion of patients being diagnosed with unilateral PA is due to the large number in whom MTO is positive, over and above those diagnosed by AVS. Thus, 27/89 (30%) and 21/89 (24%), respectively, of the unilateral cases were diagnosed by MTO or AVS alone (Fig. 2b).

Outcomes from adrenalectomy
78 patients had evaluable data 6 months after surgery. Only one patient failed to achieve at least one of the primary outcomes. 69/78 (88%) achieved complete biochemical cure of PA and 24/78 (31%) achieved complete clinical cure (Fig. 2c). For all four PASO measures of cure, these were as likely to be achieved whether patients were diagnosed by one or both of the investigations. This is apparent from comparing the proportion of patients predicted to have unilateral disease, in Fig. 2c, with the proportion of patients achieving cure, in Fig. 2d. For the primary analysis of the accuracy of each investigation, a point was assigned to one or both investigations if they correctly predicted cure, and to the other or neither investigation if they wrongly predicted cure (Supplementary Table. 3). The Forest plot for the four sequentially analysed primary outcomes shows that MTO correctly predicted nominally more biochemical and clinical cures than AVS (Fig. 2e., Supplementary Fig.1). The difference in accuracy between tests was lower than the hypothesized 25%, and the upper bounds of the confidence intervals around the differences exceeded zero, meaning that MTO was not significantly superior to AVS. The lower bounds of the confidence intervals around the differences (between accuracy of MTO and AVS in predicting the four outcomes) all exceeded -17%, meaning that MTO meets the pre-specified limit for recognizing non-inferiority.
A supplementary analysis found a small impact of AVS 'failures' on the relative accuracy of MTO and AVS in predicting partial or complete biochemical success, but not on the other three PASO outcomes (Supplementary Table 4).

Secondary Outcomes
The quantitative data, at baseline and primary follow-up, used to calculate the binary success outcomes in Figure 2 are shown in Table 2, together with the comparable data for patients in the medically-treated group. As per study design and clinical requirements, the comparison of groups is a non-randomized annotation of outcomes which may respond differently to removal of aldosterone excess and mineralocorticoid receptor blockade.
Biochemically there were large and sustained changes in aldosterone, renin and their ratio (ARR) (Fig. 3a-c). Plasma aldosterone fell in all patients after surgery, by a mean of 83%, but rose in the medical group (Table 2). While renin was de-suppressed in both groups, the rise was greater in the surgical patients. Because renin can be estimated as either mass or activity, and the routine assay differed among centers, we took the opportunity, in the largest center, to measure both mass and activity at baseline and primary outcome time points. While generally well correlated, there were some outliers ( Supplementary Fig. 2). We therefore adopted a commonly used conversion factor in order to present all results for renin and ARR in Fig. 3, but also show these separately in Supplementary Fig. 3.
Blood pressures in the surgical group were higher at baseline and lower at follow-up than in the medical group (Table 2 and Fig. 3d). These analyses do not include the initial blood pressure responses to spironolactone 100 mg daily, which are reported below.

Cardiovascular outcomes -BNP and cardiovascular magnetic resonance imaging (CMR) sub-study
Plasma troponin did not change following adrenalectomy. By contrast, plasma BNP fell by a median of 31% ( Supplementary Fig. 4a-c). This finding is consistent with the CMR sub-study, in which end diastolic volume fell by almost 29% following surgery, but only by 9% on medical therapy ( Supplementary Fig. 4d). By contrast, left ventricular mass fell in both interventions.

Predictors of outcome from surgery
Univariate analysis confirmed that young age, female gender and lower starting blood pressure predicted patients who achieved complete clinical cure (Fig. 4a-c). 21 However, for SBP, we observed that starting pressure was similar in those with complete or absent cure, but higher in patients with partial cure (Fig. 4c). This paradox is probably due to the PASO definition of complete cure being an absolute BP (<135 mmHg), which is easier to reach from a lower starting point, whereas partial cure is defined by fall in BP (>20/10 mmHg), which is easier to achieve from a higher starting pressure. The majority, 7 of 13 (54%), of absent cures (as defined by PASO) were in Black patients, treated with a calcium channel blocker ( Fig. 4c-d).

Spironolactone test
Two analyses were planned. The first quantifed the correlation between fall in BP on aldosterone blockade with the fall in BP after removal of the source of aldosterone ( Fig. 4e). Because changes in a variable correlate with the baseline value, we performed Oldham corrections (see Fig. 4e legend) of the two sets of blood pressure changes, which removes the intrinsic correlation between two changes from a common starting value. 27 The two datasets clearly correlate, but for most patients would not accurately predict change in blood pressure following adrenalectomy. Of greater clinical utility is the binary analysis, Fig. 4f, in which patients whose systolic BP fell to <135 mmHg on spironolactone were always cured by surgery, partially or complete, with 12/18 (67%) able to discontinue all treatment. The BP achieved on spironolactone reflects both the pre-spironolactone BP, previously reported as a prediction of cure, and the fall during spironolactone therapy.

Sub-types of APAphenotype : genotype correlations, and immunohistochemistry (IHC)
RNA sequencing (RNA seq) detected, in most samples, the probable culprit somatic mutation, absent from the paired sample of adjacent adrenal (Supplementary table   5). The remainder were detected either by whole exome sequencing (WES), or by targeted sequencing of known gene hotspots. 61/70 (87%) samples had a mutation in the known hotspots of KCNJ5, CACNA1D, ATP1A1, ATP2B3, CTNNB1, GNAQ or CLCN2.
Some of the ATP1A1 deletions spanned different bases from those previously described, and the single mutation of CLCN2 is a deletion in four residues prior to three deleted in one of the germline families. 28,29 In two APAs without a known functional mutation, RNA seq detected somatic mutations predicted to delete function (SIFT = 0.0) of, respectively, VAPA and DPYSL2 (CRMP2), and showed both genes to be abundantly expressed in all samples of adjacent adrenal cortex. Both mutations were confirmed on genomic DNA analysis.
There were similar numbers of KCNJ5 and CACNA1D mutations, but with striking ethnic differences ( Fig. 5a[i]). CACNA1D mutations were much commoner than KCNJ5 in the APAs of Black participants, as previously reported, though a difference in our cohort was that the rarity of KCNJ5 mutations was true of both male and female Black participants ( Fig. 5a[ii]-[iii]). Almost all participants with KCNJ5 mutations, and two with our recently reported double-mutation of GNAQ and CTNNB1 30 were women, whereas the reverse was true of all the other mutations.
As is now well-recognized, the expression of CYP11B2 is slightly lower, and CYP11B1 substantially higher, in KCNJ5-mutant than other APAs. [31][32][33][34][35] This was apparent at RNA level in 40 adenomas from 39 adrenals (Fig. 5b). Immunohistochemistry (IHC) showed a similar difference, quantified as H-scores, in a subset of APAs stained for both CYP11B1 and CYP11B2 (Fig. 5c). In patients where MTO PET-CT predicted a mixture of functional and non-functional adenomas, with respectively high and low MTO uptake, IHC confirmed large differences in CYP11B2 expression between adjacent adenomas. This is illustrated for a pair of patients for whom CYP11B1 IHC was also available ( The first part of the heatmap, the CYP11B2 node, includes several genes previously noted to be upregulated in APAs, e.g. VSNL1, CALN1, 32,36-39 but of interest are genes which are significantly upregulated in aldosterone producing cell clusters (APCCs) compared to the zona glomerulosa (ZG). 40 The second part illustrates genes which we find to be selectively upregulated by somatic genotype. Most of the genes previously noted to differentiate KCNJ5-mutant from other, ZG-like APAs, again vary many-fold in their expression between the two types. In particular, nephronectin (NPNT), a gene which is selectively expressed in normal ZG compared to zona fasciculata (ZF), 31,34,41 was highly discriminant between KCNJ5-mutant and other APAs, and in addition was twice as abundant in CACNA1D-than the ATP1A1-or ATP2B3-mutant APAs.

Multivariate prediction, including genotype and urine steroid profile data
All 18 participants with KCNJ5 mutations achieved clinical cure, which was complete in 14/18 (78%) participants, an odds ratio of 10.4 compared to other patients (Fig. 6a).
Biochemical cure was complete in 17 patients, the 18 th having the rare p.Glu145Gln mutation. By contrast, the odds ratio for complete clinical cure in participants with CACNA1D mutations, which occurred in only 3/19 (16%), was 0.38, while they contributed 7/10 (70%) absences of clinical cures, and 3/5 (60%) patients with only partial biochemical cure.
These results posed the question whether the driver of these opposing outcomes for the two commonest genotypes was a difference in the syndrome of PA caused by the genotype, or the differences in age, gender and ethnicity associated with the genotype. In the planned (binary) logistic regression of complete versus partial or absent clinical cure on genotype, age, gender, and ethnic group, KCNJ5 was the major dependent variable, with an odds ratio of 11.12 (Fig. 6b). Conversely, CACNA1D genotype and ethnicity were difficult to separate as primary determinants of absence of cure. In a proportional odds model, in which all three clinical outcomes (complete, partial, absent) were considered against genotype, age, sex and ethnic group, only age was weakly significant, but the highest odds ratio, 7.84 (95% CI [1.24, 55.96] was for the post-hoc comparison of KCNJ5 vs CACNA1D genotype (Supplementary Table 6).
Previous studies have shown that KCNJ5-mutant APAs are associated with high serum levels of hybrid steroids, which require a combination of enzymatic steps typically restricted to the ZG (CYP11B2) and ZF (CYP11B1). [42][43][44] In our study urine multi-steroid profiling identified significantly higher ratios of the hybrid steroid 18-hydroxycortisol to cortisol in patients harboring KCNJ5 mutations, allowing for an almost complete separation from other PA cases (Fig. 6c). The only borderline-high result among the latter was from the solitary woman with an ATP1A1-mutant APA, who was completely cured. Conversely, the lowest 18-hydroxycortisol/cortisol ratio in the KCNJ5-mutant group was in the woman with a p.Glu145Lys mutation, who was not completely cured.

Discussion
MTO PET-CT, following pre-treatment with dexamethasone, enables non-invasive detection of unilateral APAs. Its non-inferiority to AVS, with nominal superiority in hierarchical analysis of all four of the PASO outcomes, has the potential to increase substantially the number of patients whose unilateral APAs can be discovered and removed (Fig. 2a,b). The cure of PA (i.e. partial or complete biochemical cure) in 74/78 surgical patients (out of 128 patients with completed outcomes), and >75% mean reduction in aldosterone, suggest that usual measures under-estimate the proportion of unilateral disease, and the value of its detection and treatment (Fig. 2c,d). The reduction in BNP ( Supplementary Fig. 4) is further evidence of benefit, and emphasizes the potential importance of linking suppression of aldosterone to reversal of longterm CV risk, rather than to cure of hypertension.
MTO has been available for more than two decades. 15,16 However, it has not entered routine practice because, in both early and recent studies which did not use dexamethasone, its in vivo selectivity for CYP11B2 was not apparent. 17,19 In MATCH, in which we compared the in vivo uptake of MTO into APA and adjacent adrenal, or into non-aldosterone producing adenomas, with the subsequent measurements, by qPCR and RNA seq, of CYP11B1 and CYP11B2, it is clear that uptake of MTO is much lower in CYP11B1 than CYP11B2-rich regions of the adrenal (e.g. Fig. 5d). We did not formally quantify patient preference between the two techniques, but even in the implausible scenario of more patients preferring the invasive to non-invasive investigation, the limited availability of routinely successful AVS contrasts with the potential for PET CT to become available in all PA centers. The main limitation of MTO is its short half-life which requires synthesis by an on-site cyclotron. 18F-ligands are currently under evaluation (ClinicalTrials.gov Identifier: NCT04529018). 45 According to guidance from the Endocrine Society, the primary goal of adrenalectomy is removal of autonomous aldosterone production, to reduce long-term cardiovascular risk. 7 Patients with PA lie on a spectrum between those who will do anything to reduce their pill burden, and those who would avoid surgery at any cost.
Probably, patients referred to the tertiary clinics which recruited for MATCH incline towards those willing to follow an interventional pathway without guarantee of complete cure. Because the proportion of referrals would rise if cure were more certain, and because the demand for intervention will outstrip supply as the proportion of PA who are diagnosed rises from <1% to 10% and beyond, MATCH had major secondary objectives of finding improved, prospectively evaluated predictive measures.
In designing MATCH, a key aim was to improve on the predictive power of univariate variables (young age, female gender, lower baseline blood pressure) [46][47][48] and various algorithms incorporating these and other variables. A further aim was to harness improved understanding of the molecular pathogenesis of APAs and their phenotypes into clinical decision-making. These objectives led to the incorporation of spironolactone testing as a study within a study, and to the genotyping of almost all of the APAs, and measurement of particular urine steroid metabolites as predictor of genotype. The 13-fold greater likelihood of complete cure following surgery, if SBP on spironolactone was <135 mmHg (Fig. 4f), probably reflects both the lower pretreatment pressure in these patients, and the contribution of aldosterone excess to this pressure. Since there was also a higher likelihood of SBP falling to <135 mmHg in patients with unilateral than bilateral PA, an interesting possibility would be to reduce radiation exposure by starting patients on spironolactone at an early stage, and dispense with the diagnostic CT, prior to PET CT, in those with high prior probability of unilateral disease.
Our secondary, pre-specified analyses build on our and others' demonstration that the molecular divisions among APAs, by genotype and gene expression, are becoming more distinct, at the same time as the conventional anatomical 'subtype' division into unilateral and bilateral is becoming more blurred. 6,[31][32][33][34][35] We are not the first to associate KCNJ5-mutation with better clinical outcome, or with hybrid steroid excess. [49][50][51] However, the observation has not been uniform; and in studies which are retrospective and/or the control group is wild-type APAs, it is harder to eliminate confounders, such as NFAs 'masquerading' as APAs, or to collect data of equal weight from most patients and adenomas. The 78% complete clinical cure rate for KCNJ5mutant APAs, compared to 31% overall (Fig. 6a), and only 15% once KCNJ5-and GNAQ/CTNNB1-mutant APAs are excluded, is dramatic. The complete cures included males and females, older and younger patients. The paucity of young Black women with KCNJ5-mutant APAs in MATCH makes it unlikely that good outcome is due to earlier detection in younger women consulting doctors for obstetric or contraceptive reasons.
CACNA1D mutation has not previously been associated with poorer outcome, but has been strongly linked both to a high proportion of somatic mutations in aldosterone producing cell clusters (APCCs), including those which may underpin bilateral 'hyperplasia', and to a higher prevalence than KCNJ5 mutation in Black male patients. [52][53][54] It is tempting therefore to postulate that the higher likelihood of absence of cure in Black patients reflects a point in the anatomical spectrum that is asymmetric rather than unilateral. 6 Some evidence in favor of this is that 4/5 of the partial biochemical cures are Black patients, whose renin remained suppressed despite their aldosterone falling by >75%, to <300 pmol/L. Conversely, KCNJ5 mutations and the GNAQ/CTNNB1 double-mutations arise typically in solitary adenomas, and are rarely seen in APCCs. To date, 16/16 double-mutant APAs have been completely cured. 30,31 In most endocrine conditions, it is normalization of the culprit hormone that is the main goal of intervention, and in MATCH this was achieved irrespective of genotype.
Furthermore, the large fall in plasma BNP adds weight to the recommendation for intervention as a means of cardiovascular protection ( Supplementary Fig. 4a and Table   2). Because the allocation to surgical and medical management was not randomized, and blood pressure control slightly less effective in the latter group, no definite conclusions can be drawn from the failure of BNP to fall in the medical patients. It could be that aldosterone suppression is more effective than mineralocorticoid receptor (MR) blockade, with aldosterone able to stimulate non-MR mediated effects. 55,56 But this hypothesis should wait for formal randomized control trials comparing some of the newer drugs in the aldosterone synthase inhibitor and mineralocorticoid antagonist classes.
In conclusion, our study validates dexamethasone-suppressed MTO PET-CT as a CYP11B2-selective investigation for lateralizing PA. MATCH shows that it will now be possible to diagnose unilateral APAs when AVS is either unavailable, unsuccessful, or not desired by the patient. Unlike AVS, MTO PET-CT is non-invasive and is unlikely to be operator-dependent. These advantages may attract more clinicians and patients to seek a diagnosis, especially if the 18F-analogue becomes available, and transportable to most hospitals with PET imaging facilities. 45 The longer-term impact of MATCH may be a reassessment of who undergoes lateralization and surgery, and a catalysis of decision-making by molecular, rather than anatomical, diagnosis. In this view, either surgery itselfa procedure usually undertaken for cure rather than preventionor the counselling of patients to whom it is offered, will emphasize its success rate in patients with elevated hybrid steroid secretion, and normal blood pressure on spironolactone. Conversely, patients advised of a <1/10 likelihood of complete clinical cure based on demographics, steroid secretion, and response to spironolactone, may now prefer optimized medical treatment, and/or to wait for nodular ablation, if/when this is shown to reduce BP, medication, and plasma BNP as much as surgery. When that time arrives, molecular imaging offers the potential advantage over AVS of localization as well as

Participants
Participants were recruited through tertiary endocrinology clinics at SBH and CUH, or complex hypertension clinics at SBH and GSTT. The criteria for inclusion consisted of patients aged 18 years or older with a probable or definite adenoma on CT or MRI scans of the adrenals, and meeting the Endocrinology Society criteria for diagnosis of PA (ARR > 750 pmol/mU or equivalent local value measured off interfering medications AND either (1) a plasma aldosterone >190 pmol/L after a 2L 0.9% saline infusion test, or (2) failure to suppress plasma aldosterone by 30% + persistent plasma renin suppression a 25 mg captopril challenge test, or (3) spontaneous hypokalemia + plasma renin below detection levels + plasma aldosterone > 550 pmol/L). Patients with elevated ARR could be put forward for consideration by the MDT for inclusion in the study as exceptional cases in whom confirmatory test was not performed if plasma aldosterone > 450 pmol/L AND (1) plasma renin <0.5 pmol/ml/hr (<9 mU/L) if measured on treatment with an ace-inhibitor (Lisinopril >20 mg or equivalent), or (2) an angiotensin receptor blocker (Losartan 100 mg or equivalent); OR if patient was aged <40 years and a definite adrenal adenoma was seen on CT or MRI.
Exclusion criteria included individuals who were (1) unlikely to proceed with surgery if recommended, (2) contraindicated for spironolactone/eplerenone therapy use, (3) unable to stop beta-blockers or direct renin blockers, (4) pregnant women or unable/unwilling to take secure contraceptive precautions whilst undergoing investigations, (5) unable/unwilling to take the dexamethasone required to prepare for a MTO scan (6) unwilling/unable to have both MTO and AVS, or (7) had a condition or drug regimen that was considered a contraindication by the principal investigator.

Adrenal vein sampling (AVS) and scoring
All AVS procedures were performed by one of three experienced interventional radiologists. 50μg/h of intravenous synacthen was started 1 hour before the procedure. In line with a consensus statement cannulation was considered successful if the cortisol level in each adrenal vein was ≥3x greater than that in the iliac and/or infrarenal IVC. 57 'High' probability of unilateral PA was diagnosed if the aldosterone/cortisol (A/C) ratio in one adrenal vein was ≥4 times that in the contralateral adrenal vein and if the contralateral vein had an A/C ratio less than that of the infrarenal IVC/iliac vein. 57,58 'Low' probability of unilateral PA was diagnosed with an A/C ratio <3, which was indicative of bilateral aldosterone secretion. An A/C ratio of 3-4 was considered 'intermediate'. In the event of one or both adrenal veins not being successfully cannulated, the outcome was declared 'failed'. Although, in clinical practice, it has been suggested that a focal right adrenal lesion in association with an A/C ratio in a unilaterally cannulated left adrenal vein <0.5 of the infrarenal IVC/iliac vein can provide justification for a right adrenalectomy, 59 this was not adopted in MATCH.

Minimization
The order that each participant underwent AVS or MTO PET-CT was determined by a minimization program within the study data center (Robertson Centre for Biostatistics, University of Glasgow), designed to maintain balance with respect to study site, gender, and age (<55 or ≥55 years). Both AVS and MTO scanning were performed with the patient taking non-interfering medications such as doxazosin, verapamil and hydralazine, according to Endocrine Society consensus guidelines. 7

Clinical decision-making
Each study patient was discussed in a MDT meeting attended by PIs, radiologists and surgeon. The MTO scan was presented and interpreted by the reporting radiologist prior to the AVS result being revealed. For each investigation, the question posed was: 'in the absence of any other information, would this result provide sufficient evidence for a referral for surgery?' The MDT decision was not bound unconditionally to the criteria for diagnosis of unilateral PA as outlined above. The probability of unilateral disease could, by consensus, be reduced if data from the contralateral adrenal suggested bilateral disease (e.g. contralateral adenoma with SUVmax higher than adjacent adrenal, or LI>4 but contralateral A/C not suppressed compared to IVC).

Baseline visit and initiation of spironolactone therapy
After the second investigation, patients attended for measurements of all clinical and biochemical parameters whose response to treatment would assess the accuracy of the MTO and AVS. Obtaining the baseline measurements at this point in the study allowed a maximum interval between discontinuation of any interfering medicines being taken at screening, and the opportunity for the blood pressure response to initiation of spironolactone to be assessed as a predictor of response to surgery.
After four days of home blood pressure recordings, 3 readings twice daily (24 in total), blood was drawn for electrolytes, renin activity and/or renin mass, aldosterone, BNP and troponin. In the rare occasion the minimum number of home blood pressure recordings were not available, office blood pressure recordings (average of 3 sequential readings) were used instead. Participants also performed a 24h urine collection for urine steroid profiles. Participants then commenced spironolactone 50 mg once daily for two weeks, increasing to 100 mg once daily depending on side effects and any concern about renal function or potassium homeostasis. In participants referred for surgery, therapy was continued, as tolerated, until the day of operation. Participants unable to tolerate spironolactone were treated with eplerenone 25 mg twice daily for two weeks, increasing to 50 mg twice daily. Further treatment titrations in the medical group were at the discretion of the study or referring physician.

Surgery
Laparoscopic adrenalectomy, using standard operative techniques, was performed by a single endocrine surgeon at two of the participating sites, unless the participant or referring physician preferred local referral. The serum cortisol value at the conclusion of the 72 hours preparatory dexamethasone before the MTO scan was used to guide the clinical concern about potential postoperative adrenal insufficiency. Where appropriate, supplementary hydrocortisone was provided and withdrawn after subsequent clinical and biochemical reassessment. The outcomes from surgery were assessed biochemically and clinically, at 3 and 6 months after surgery, and at 12 and 24 months in patients recruited in earlier years of the study. At each visit, the same measurements were performed as at the baseline visit, except that BNP, troponin, and 24h urine measurements were repeated only once, at the 6-month post-surgery visit, or 9-12 months in medically treated participants.

Outcomes
The primary objective of MATCH was to compare the accuracy of MTO PET-CT and AVS in diagnosing unilateral PA, by measuring normalisation of biochemical and clinical parameters following adrenalectomy. The criteria for normalisation were the PASO criteria, agreed by international consensus in 2017 (Supplementary table 1), for each of partial or complete cure of PA and hypertension. 21 Since the rank order of partial and complete cures of PA and hypertension could be assumed (with more patients being biochemically than clinically cured), prospective studies in PA lend themselves to hierarchical analysis of each PASO outcome, in which all four are primary outcomes without needing to adjust for multiple statistical comparisons (Fig.   1b). It is important to note that, whereas partial may seem softer versions of complete cure (either clinical or biochemical), they measure different parameters. Partial responses are changes in biochemistry, blood pressure, and defined daily dosage (DDD) of medications, whilst complete responses are numbers below an absolute target. Of note, the first outcome in the hierarchy, partial biochemical cure, is the sole outcome driven by reduction in the culprit hormone, aldosterone, to which excess cardiovascular risk in PA is attributed.
Secondary objectives of MATCH included serial measurements, up to two years postintervention, of the criteria for assessing biochemical and clinical success, analysed separately in the surgical and medically treated participants; and analyses of parameters which may, in the future, enable more reliable prediction of which patients can achieve complete clinical cure adrenalectomy. The main clinical test was a within-trial measurement of participants' response to 4 weeks treatment with spironolactone. Molecular analyses included RNA sequencing of APAs for genotyping and transcriptomics. We also report the left ventricular mass and end-diastolic volume measurements from the cardiovascular magnetic resonance imaging (CMR) sub-study of 50 patients, performed at baseline and 1 year after surgical or medical management.

Statistics and sample size
The primary analysis was a comparison of the accuracy of MTO PET-CT and AVS. Each test was deemed accurate if it recommended surgery and surgery resulted in cure (according to each PASO criterion), or if the test did not indicate surgery and surgery did not result in cure. Since surgery was almost exclusively undertaken when one or both tests indicated surgery was required, for those patients where neither test indicated surgery, both tests were treated in the analysis as having equal accuracy.
Only in those patients for whom the two tests gave conflicting recommendations was it possible to determine which of the two test was accurate, and which was not.
The difference in accuracy between the two tests was estimated amongst those who underwent surgery, using the Newcombe-Wilson score method for paired binary data, and tested using an exact McNemar test.
The sample size for the study was estimated from a table of all permutations of outcomes from each investigation and surgery, in which the frequency of each outcome reflected experience at participating centres (Supplementary Table 7). This estimated a need for 128 completed patients, with half undergoing surgery, and half of those (32) doing so as the result of one of MTO or AVS, but not both. This was based on having 90% power at a 5% significance level to detect a difference in accuracy between the two tests, assuming that on those occasions when only one of the tests was accurate, it would be MTO that was accurate 80% of the time. In addition, if the two tests were truly equally accurate, this sample size would have 90% power to demonstrate non-inferiority of MTO versus AVS within a non-inferiority margin of -17%.
The full analysis set was defined as all participants who had both investigations.
Because of pandemic-associated interruptions to the availability of laparoscopic adrenalectomy, the study was censored when more than 75 patients had undergone surgery, and at least one evaluable outcome visit. The analysis is reported for the 128 patients who had reached a primary outcome visit, either 6 months post unilateral adrenalectomy (including 5 for whom data from their most recent, 3-month visit was carried forward), or 9-12 months after their baseline visit in the medically-treated patients.

Secondary analyses
Secondary (and Primary) outcomes were compared between participants who underwent surgery and those who did not, and for those who underwent surgery, between those for whom surgery was indicated by AVS only, MTO only, or both tests.
Groups were initially compared using Fisher's Exact or Kruskal-Wallis tests, as appropriate. Linear, binary logistic, or proportional odds logistic regression models were then used, adjusted for age and gender. These models were extended to investigate other predictors of outcome after surgery, in the subset of participants who underwent surgery. There was particular interest in blood pressure at baseline and 4 weeks, age, gender, ethnicity, the number of classes/defined daily dose (DDD) of antihypertensive medications at baseline, tumour genotypes and immunohistochemical classification, contralateral suppression, numbers of nodules, blood biomarkers, SUVmax, and 24 hour urine steroid profile.
For each of the primary outcomes, a sensitivity analysis was carried out to assess the impact of including patients for whom AVS was classified as 'failed'. A bootstrap procedure was applied, sampling with replacement from the 78 patients who underwent surgery. Sampling was stratified by whether AVS was successful, or failed.
For each sample, the estimated difference in accuracy between MTO PET-CT and AVS was calculated for the combined sample of 78 patients, and for the 67 patients where AVS had not failed, and the difference between these estimated differences was taken as a measure of the sensitivity of the overall result to the inclusion of AVS failures.
One million replicates were drawn, and a 95% confidence interval was derived by the percentile method. If the 95% CI excluded zero, then the overall result was adjudged to be sensitive to the inclusion of AVS failures. If the 95% CI included zero, then the results were judged to be insensitive to the AVS failures.

Cardiovascular Magnetic Resonance Imaging (CMR) Sub-study
A CMR substudy was performed with imaging acquired at baseline prior to treatment allocation in 76 patients, and at follow up at 12.0+/-3.1 months post in 51 patients. All

Adrenal tissue collection
Adrenal tissue was collected directly from surgery. Once the adrenal was removed this was dissected into 5mm serial slices by the surgeon or histopathologist. All adrenal adenomas were identified and 10-200mg of tissue was removed from each APA and adjacent normal adrenal gland (AAG) for storage in separate tubes in RNALater © solution (AM7020, Invitrogen). Slices were sent to pathology for routine clinical hematoxylin and eosin staining and immunohistochemistry.

Immunohistochemistry (IHC)
IHC was performed on 3-μm sections cut from paraffin blocks using a fully automated systems. The CYP11B1 primary antibody clone RAT-87 (MABS502, Merck) was used at a dilution of 1:100 after heat-induced epitope retrieval at pH 9.0 (BOND Epitope Retrieval Solution 2, AR9640, Leica) for 20 minutes; and CYP11B2 primary antibody clone EPR10494 (ab168388, Abcam) was used at a dilution of 1:200 after proteolyticinduced epitope retrieval with Proteinase K (BOND Enzyme Pre-treatment kit, AR9551, Leica) for 10 minutes. The primary antibody binding to tissue sections was visualized using BOND Polymer Refine Detection system (DS9800, Leica).

Histopathological Analysis
For each case, APAs, aldosterone-producing nodules (APNs), aldosterone-producing micronodules (AP micronodules) formerlyl known as aldosterone-producing cell cluster (APCC), functional and non-functional hyperplasia were characterised based on the international histopathology consensus for unilateral primary aldosterone (HISTALDO). 61 AP micronodule numbers were counted within the adrenal adjacent to the main nodule. The cell types within the main nodules (lipid rich, lipid poor or mixed) were noted. The intensity and proportion (as a percentage) of the staining of tumours with CYP11B1 and CYP11B2 were evaluated semi-quantitatively using a scoring system as follows: 1= weak, 2= intermediate, and 3= strong. From this data, a H-score was assigned using the following formula: 3 x (% cell 3+) + 2 x (% cells 2+) + 1 x (%cells 1+)].
A consensus of either classical or non-classical histopathological findings was made in line with HISTALDO criteria, where "classical" histopathology associated with unilateral PA included a solitary APA or APN; "nonclassical" histopathology included multiple APMs or multiple APNs (or multiple APMs and multiple APNs together) or aldosterone-producing diffuse hyperplasia.

RNA and DNA extraction
RNA and DNA was extracted from tissue samples stored in RNALater © for RNA, whole exome or Sanger sequencing. Tissue homogenisation was performed in Trizol (Life Technology 15596026) using a FastPrep-24 5G Sample Preparation System (MP-Biomedicals) as per protocol. Genomic DNA (gDNA) was extracted from approximately 25-30mg of normal adrenal or APA tissue using QIAamp DNA Mini kit (51304, Qiagen) according to manufacturer's instructions. If AAG was not available to be used as controls, gDNA was extracted from blood using the salt extraction method. Total RNA was extracted from approximately 50-100mg of AAG or APA tissue using the Trizol method (Life Technology 15596026) followed by on-column DNaseI treatment (Invitorgen) and purification using Invitrogen's PureLink RNA minikit (#12183018A).

Quantitative polymerase chain reaction (qPCR)
qPCR was performed on all adenomas to assess for CYP11B2 and CYP11B1 mRNA expression. RNA was reverse transcribed to cDNA using Applied Biosystem's High Capacity RNA-to-cDNA kit. Taqman assay probes were used to quantify the gene of interest. The CT method was used to quantify gene expression levels. 62 Supplementary

Data Analysis
Sequence reads were trimmed to remove possible adapter sequences and nucleotides

Targeted Sanger Sequencing
cDNA from the APA and AAG was sequenced to confirm presence of KCNJ5 mutations.
PCR was performed with the following forward (caacttgctcgtcttcacca) and reverse (gagggtctccgctctcttct) primers, using AmpliTaq Gold™ Fast PCR Master Mix (ThermoFisher, 4390939) as per manufacturer's instructions. 5 l of PCR products were then purified using 2 l of ExoSAP-IT™ PCR Product Cleanup Reagent (Applied Biosystems, 78201.1.ml) prior to Sanger Sequencing. Sequencing of the PCR products was performed using LIGHTRUN Tube sequencing services from Eurofin (Germany).

Urinary Steroid Profiling by Liquid Chromatography-Tandem Mass Spectrometry
Steroids were extracted from 200 µL of a 24-hour urine collection after addition of an internal standard mixture, as described previously. 63,64 In brief steroids were deconjugated from their sulphate and/or glucuronide conjugates through enzymatic hydrolysis with helix pomatia (Sigma Aldrich, Gillingham, UK, 60°C for 3 hours). The steroids were then extracted via solid phase extraction (C18 100mg, Biotage, Hengoed, UK). The methanol eluent was collected, evaporated and the extract reconstituted and run by liquid chromatography tandem mass spectrometry (LC-MS/MS). The mass spectrometer was a Waters Xevo-XS with an Acquity ultra-high pressure liquid chromatography (uPLC) system, with an electrospray source in positive ionisation mode with a methanol and water gradient system (both with 0.1% formic acid). Gradient started at 30% methanol, linearly increasing to 39% over 6 minutes.
Next the column was washed at 98% methanol and re-equilibrated prior to the next injection. Separation was achieved on a Waters HSS T3 1.2 x 50mm 1.8µM column.
With each batch of samples, a calibration series, (0-1500 ng/mL), three spiked QC samples and seven pooled biological control urines were extracted. Steroids were quantified relative to their deuterated internal standard Cotisol-d4 or 18hydroxycortisol-d4, against this calibration series. In addition to previously published methods 18-hydroxycortisol (18OHF) was included in this method. Bias (% deviation), calculated as the accuracy at three spiked concentrations 20, 300 and 200ng/mL was <5% for cortisol and <10% for 18OHF at all concentrations. Imprecision (% RSD) measured from multiple extractions of the same urine sample (intra-assay (n=7), inter-assay (n=14) was 4.9% and 3.85% for cortisol, and 5.8% and 6.21% for 18OHF respectively.