Cardiovascular outcomes and all-cause mortality in primary aldosteronism after adrenalectomy or mineralocorticoid receptor antagonist treatment: a meta-analysis

Search strategy and selection criteria

The systematic review and meta-analysis was conducted according to the Preferred Reporting Items of Systematic Reviews and Meta-Analyses (PRISMA) statement and Cochrane methods. We searched PubMed, Cochrane library, Embase, and Cnki.net using terms associated with ‘hyperaldosteronism’, ‘adrenalectomy’, ‘mineralocorticoid receptor antagonist’, ‘cardiovascular’, ‘cerebrovascular’, ‘outcome’, ‘mortality’, ‘atrial fibrillation’, and ‘stroke’ (Supplemental Methods). The systematic review protocol was registered in PROSPERO (CRD42021256596).

Two investigators (SYC and WCH) searched the published studies in PubMed, Cochrane library, Embase, and Cnki.net (Supplementary Fig. 1, see section on supplementary materials given at the end of this article) from their inception to 18 December 2021, with language limited to either English or Chinese. We focused on publications of human subjects and case-control or cohort studies. A third investigator (VCW) resolved the disagreements between the two investigators. If potentially relevant, the full article would be retrieved and evaluated according to the following selection criteria.

Studies were included if they investigated the patients with PA who underwent MRA or adrenalectomy treatment, and mortality, cardiovascular or cerebrovascular events after treatment were reported. Exclusion criteria were non-human studies, duplicate records (study with the most relevant data was selected), without relevant outcomes, and absence of a control group (study presented the outcomes of either MRA or adrenalectomy treatment only). A manual search of reference lists from related studies, review articles, and meta-analyses was performed to identify additional reports not found in the computerized databases.

Data extraction

Two investigators (SYC and WCH) independently extracted relevant data from eligible studies according to a standardized spreadsheet and discrepancies were resolved by consensus or discussion with a third investigator (VCW). Extracted data included outcomes of interest, study characteristics (author, year of publication, country, study design, patient enrollment, number of participants, number of events, and duration of follow-up), and the baseline condition of patients (age at diagnosis, sex ratio, BMI, duration of hypertension, systolic BP, diastolic BP, percentage of comorbidities, such as diabetes mellitus (DM) and chronic kidney disease). When available, the mean and s.d. of each statistic were included. Full-text papers were reviewed for quality assessment and data synthesis. We further contacted the authors of enrolled articles to acquire additional details.

Quality assessment

The quality of included studies was evaluated by the Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) (28) by the Cochrane Bias Methods Group. ROBINS-I scored each study based on the following seven domains: ‘confounding’, ‘participants selection’, ‘classification of interventions’, ‘deviations from intended interventions’, ‘missing data’, ‘outcome measurement’, and ‘reported result selection’. Each domain was scored as ‘no information’ (0), ‘low risk of bias’ (1), ‘moderate risk of bias’ (2), ‘serious risk of bias’ (3), and ‘critical risk of bias’ (4). The overall risk of bias was concluded for each study.

Outcomes

Our pre-specified composite primary outcomes were long-term all-cause mortality and/or major adverse cardiovascular events (MACE), including CAD (myocardial infarction (MI) and revascularization with coronary artery bypass grafting (CABG) or coronary angioplasty (percutaneous coronary intervention (PCI))), stroke, arrhythmia, and congestive heart failure. We also analyzed the above-mentioned individual components as secondary outcomes to assess the incidence of each adverse event after treatment.

Statistical analysis

ORs and 95% CIs were calculated using the random-effects model with the Review Manager (RevMan) (Computer program), Version 5.4, The Cochrane Collaboration, 2020. We assessed the heterogeneity among studies by the Cochrane Q test and the null hypothesis of homogeneity would be rejected if P-value was less than 0.05. The extent of heterogeneity was categorized into mild (I² < 30%), moderate (30% ≤ I² < 50%), and substantial (I² ≥ 50%). Funnel plots were applied to identify potential publication bias. We also used STATA (Version 16, Statacorp. 2019) for prespecified subgroup analyses for categorical variables and continuous variables and meta-regression for continuous variables to assess the interaction between variables and the targeted treatments on composite primary outcomes.

We applied trial sequential analysis (TSA) (Copenhagen Trial Unit, Centre for Clinical Intervention Research, Denmark, software 0.9.5.10 Beta software) to the composite primary outcomes to assess the temporal and cumulative effect of each article on the present study. Taking into consideration the heterogeneity of patient selection, study design, quality of methodology, and duration of follow-up among studies, we adopted the random-effects model. We set the conventional non-superiority boundaries at significance levels of 0.05 and a power of 90% and calculated the α-spending boundaries using the O’Brien–Fleming procedure (29).

Search results and study characteristics

Our initial electronic search yielded a total of 355 articles. After duplicate removal, 293 studies were screened at the levels of title and abstract. Irrelevant and inappropriate records were removed, leaving 15 articles for full-text review for eligibility. Among them, six articles lacked a comparison between the MRA and adrenalectomy treatment. Eventually, 9 cohort studies (16, 17, 19, 22, 23, 24, 25, 26, 27) were eligible for our meta-analysis, resulting in a total of 8473 patients with PA who underwent either adrenalectomy (31.3%) or MRA (68.7%) treatment. The descriptive summary for included studies is reiterated in Table 1. Most studies reported similar baseline characteristics, including age, sex, BMI, DM, systolic BP, and diastolic BP, between surgically and medically treated patients. However, there was still some imbalance between treatment groups in some studies, which was reflected in our risk of bias assessment.

Quality assessment

The results of ROBINS-I tool assessment showed most of the included studies were within the range of low-to-moderate risk (Supplementary Table 1). Overall, only one study (23) had a moderate-to-serious risk and one (19) had a serious risk of bias. However, due to the limitation of a relatively small sample size and the fact that most studies were designed for comparison between PA and EH, most included studies could not properly adjust the baseline confounding variables (D1) of PA patients treated surgically or medically, yielding moderate-to-serious risk of bias. One study (23) reported 13.9% (12/86) of patients in the surgical group failed medical treatment prior to adrenalectomy, and therefore, it had a serious-to-critical risk of bias due to deviation from the intended intervention (D4). Authors of two studies (19, 25) failed to report the management of missing data (D5), causing a serious-to-critical risk of bias. Most included studies reached a moderate level in risk of bias due to the selection of reported results (D7).

Composite primary outcomes

Among nine included studies, three studies (19, 22, 27) reported all-cause mortality and MACE while the remaining (16, 17, 23, 24, 25, 26) only reported MACE as the primary outcome. Definitions of MACE differed mildly among studies. MACE of most studies comprised MI, CABG, and PCI. One study (24) only reported atrial fibrillation, another (25) only ischemic and hemorrhagic stroke, and the other (26) only CAD.

Composite primary outcomes were reached by 486 (18.4%) of 2648 patients in the adrenalectomy group and 1692 (29.0%) of 5825 patients in the MRA group. Superior effect of adrenalectomy over MRA was reported in four studies (19, 22, 25, 27), whereas three studies disclosed similar outcomes between two treatment modalities (16, 17, 23). The remaining two studies did not report either adrenalectomy or MRA treatment delivered better outcomes (24, 26).

Patients who underwent adrenalectomy had better composite primary outcomes (OR: 0.46 (95% CI, 0.38–0.56), P < 0.001, heterogeneity I² = 30%), including lower all-cause mortality and/or MACE rates, in the random-effects model (Fig. 1). Fixed-effects model results are shown in Supplementary Fig. 2. Funnel plot showed symmetrical distribution, indicating low risk of publication bias (Supplementary Fig. 2).

View Full Size

Forest plot (random-effects model) showing the risk of MACE and/or all-cause mortality between adrenalectomy and MRA treatment. Adx, adrenalectomy; MACE, major adverse cardiovascular events; MRA, mineralocorticoid receptor antagonist.

Citation: European Journal of Endocrinology 187, 6; 10.1530/EJE-22-0375

• Download Figure
• Download figure as PowerPoint slide

We separately analyzed all-cause mortality, one component of composite primary outcomes. The pooled result demonstrated lower all-cause mortality in the adrenalectomy group with the random-effects model (OR: 0.33 (95% CI: 0.15–0.73), P = 0.006, heterogeneity I² = 73%) (Fig. 2A). Fixed-effects model results are shown in Supplementary Fig. 3. A visual inspection of the funnel plot suggested no obvious publication bias (Supplementary Fig. 3). The comparison of MACE, the other component of composite primary outcomes, disclosed better outcome in the surgical group with the random-effects model (OR: 0.55, (95% CI: 0.40–0.74), P = 0.0001, heterogeneity I² = 68%) (Fig. 2B). Fixed-effects model results are shown in Supplementary Fig. 4. The funnel plot of MACE comparison suggested no obvious publication bias (Supplementary Fig. 4).

View Full Size

Forest plots (random-effects model) illuminating the risk of all-cause mortality (A) and MACE (B) between adrenalectomy and MRA treatment. Adx, adrenalectomy; MACE, major adverse cardiovascular events; MRA, mineralocorticoid receptor antagonist.

Citation: European Journal of Endocrinology 187, 6; 10.1530/EJE-22-0375

• Download Figure
• Download figure as PowerPoint slide

Secondary outcomes

In random-effects models, surgery was associated with a lower incidence of CAD (OR: 0.33 (95% CI: 0.15–0.75), P = 0.008, heterogeneity I² = 0%) (Fig. 3A), stroke (OR: 0.79 (95% CI: 0.34–1.83), P = 0.58, heterogeneity I² = 47%) (Fig. 3B), arrhythmia (OR: 0.46 (95% CI: 0.27–0.81), P = 0.007, heterogeneity I² = 36%) (Fig. 4A), and congestive heart failure (OR: 0.52 (95% CI: 0.33–0.81), P = 0.004, heterogeneity I² = 1%) (Fig. 4B). Stroke, however, did not reach statistical significance. The fixed-effects model forest plots of secondary outcomes are presented in Supplementary Figs. 5, 6, 7, and 8. Funnel plots of all four secondary outcomes confirmed zero to low risk of publication bias (Supplementary Figs. 5, 6, 7, 8).

View Full Size

Forest plots demonstrating the risk of (random-effects model) CAD (A) and stroke (B) between adrenalectomy and MRA treatment. Adx, adrenalectomy; MACE, major adverse cardiovascular events; MRA, mineralocorticoid receptor antagonist.

Citation: European Journal of Endocrinology 187, 6; 10.1530/EJE-22-0375

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Forest plots (random-effects model) illustrating the risk of arrhythmias (A) and congestive heart failure (B) between adrenalectomy and MRA treatment. Adx, adrenalectomy; MACE, major adverse cardiovascular events; MRA, mineralocorticoid receptor antagonist.

Citation: European Journal of Endocrinology 187, 6; 10.1530/EJE-22-0375

• Download Figure
• Download figure as PowerPoint slide

Subgroup analyses and meta-regression

Based on baseline conditions, we identified ten pre-specified subgroups and analyzed how different study factors influenced the composite primary outcomes (Supplementary Fig. 9A). Six studies were included for subgroup analyses of baseline plasma renin activity and aldosterone levels (Supplementary Fig. 9B and C). Our results showed that both subgroups in each analysis have the same direction of effect, indicating that baseline plasma renin activity and aldosterone levels have minimal to limited effect on the composite primary outcomes.

When the serum potassium was lower than 3.5 mmol/L (hypokalemia), indicating more advanced disease, the benefits of adrenalectomy over MRA on the composite primary outcomes could be insignificant (n  = 208, OR: 0.68 (95% CI: 0.33–1.40)). On the contrary, when the serum potassium was higher or equal to 3.5 mmol/L (normokalemia), implying less severe disease, the benefits of adrenalectomy over MRA on the composite primary outcomes became statistically significant (n  = 1,736, OR: 0.44 (95% CI: 0.35–0.57)).

We also conducted random-effects meta-regression to assess the interaction between each variable and the composite primary outcomes. Patient’s age modified the effect of adrenalectomy on composite primary outcomes (coefficient: 1.084 (95% CI: 1.005–1.169), P = 0.036), implying an 8.4% increase in OR per year increase of chronologic age (95% CI: 0.5–16.9%, P = 0.036). In other words, the older the patients are, the less advantage they may receive from adrenalectomy (Supplementary Fig. 9D).

We further performed subgroup analyses to compare the treatment effects of adrenalectomy and MRA in patients with unilateral PA. Two studies that enrolled 1012 patients were included (23, 27). In concordance with our main finding, lower risk of all-cause mortality and/or MACE (OR: 0.45 (95% CI: 0.34–0.59), P < 0.001, heterogeneity I² = 0%) and MACE (OR: 0.50 (95% CI: 0.38–0.66), P < 0.001, heterogeneity I² = 0%) in the adrenalectomy group (Supplementary Fig. 10).

Trial sequential analysis

For the composite primary outcomes, we applied TSA to estimate the statistical reliability of included data and overcome the limitation of relatively restricted sample sizes.

Calculation of the required information size (RIS) was based on assumptions and goals. We assumed a 19.5% event rate in the control arm (MRA), which was roughly the median of included studies. A 25% reduction in relative risk, equivalent to a 4.9% reduction in absolute risk, as the effect of the intervention (adrenalectomy) was considered clinically meaningful. A type 1 error of 5% and a power of 90% were adopted. According to the factors, the heterogeneity-adjusted RIS was calculated to be 5670 patients. The cumulative Z-curve surpassed the conventional boundary for statistical significance and the trial sequential monitoring boundary for benefits. Moreover, the accumulated case number of included studies was larger than the RIS, indicating that the current evidence reached a conclusion supporting the superior performance of adrenalectomy over MRA treatment in composite primary outcomes (Supplementary Fig. 11A).

Additionally, Supplementary Fig. 11B showcases the cumulative influence of each included trial temporally. From 2008 to 2021, as more trials were reported, we observed a consistently superior performance of surgical treatment over medical treatment. The cumulative Z-score has been constantly growing along with the total accrued sample size, reflecting the unwavering lowering of P-value of the meta-analysis.

Main findings

In this meta-analysis of 9 studies with 8473 patients with PA, we comprehensively reviewed the long-term composite primary outcomes and individual secondary outcomes after adrenalectomy vs MRA treatment. To our knowledge, this is the most updated quantitative review that evaluates the efficacy of medical vs surgical treatment upon the all-cause mortality and cardiovascular events in patients with PA. Our integrated results showed a significantly lower all-cause mortality rate, incidence of MACE, CAD (events of MI, CABG, and PCI), arrhythmia, and congestive heart failure after adrenalectomy than that after MRA therapy.

Our study is different from previous meta-analyses (30, 31), which reported a neutral result in cardiovascular events between surgical vs medical treatment and a potentially lower all-cause mortality after surgical treatment than after medical treatment. However, these older studies had a relatively high heterogeneity, and their size of included patient population was significantly smaller than ours. Collectively, these findings suggest that adrenalectomy not only normalized BP and aldosterone levels but also extenuated the unwanted cardiovascular events effectively. With the knowledge of the difference between the two main subtypes of PA, we could conclude that it is critical to diagnose PA among hypertensive patients, and further identification of lateralized PA is imperative to provide proper and better treatment (32).

The targeted treatments and clinical outcomes

One possible factor influencing cardiovascular outcomes was the higher incidence of metabolic syndrome, insulin resistance, and DM in patients with PA (6). Excessive aldosterone in patients with PA impairs insulin secretion in pancreatic beta-cells (33) and induces insulin resistance in peripheral tissues (34). Hypokalemia can also impair insulin secretion (35). The higher incidence of metabolic syndrome was possibly attributable to the concomitant glucocorticoid excess associated with PA. The co-secretion of glucocorticoid and mineralocorticoid is a frequent phenomenon observed in patients with PA, contributing to metabolic risks (36). Arlt et al. proposed that unilateral adrenalectomy effectively resolves the excess of glucocorticoid and mineralocorticoid simultaneously, while MRA alone is not sufficient to counteract the adverse metabolic risks and may require additional glucocorticoid-opposing agents to alleviate the problems (36).

Another possible reason for the inferior cardiovascular outcomes of MRA therapy observed in the present study could be the potentially inadequate MRA dose, making the plasma renin activity consistently suppressed (<1 µg/L/h) despite treatment, which was proven to be associated with a higher risk of cardiovascular events, diabetes, and mortality (37). Suppressed renin activity is considered a warning sign of excessive MR activation(5), implying insufficient antagonism of renal MR and potentially extra-renal ones. The over-activation of MR is deleterious far beyond hypertension; it has also been associated with ventricular hypertrophy (38), renal dysfunction (39), and mortality (8). Contrarily, removing APA solves the root cause of the mineralocorticoid excess, without the need to monitor plasma renin activity.

A German study found no difference in all-cause mortality between patients with PA treated with surgery or MRA vs matched EH controls (8). Intriguingly, they also observed a trend of increased mortality in all patients with PA compared to that of EH controls after 10 years of follow-up (8). A similar trend was identified in another study (37), which found higher cardiovascular events, mortality, diabetes, and atrial fibrillation in patients with PA treated with MRA compared with matched EH controls after long-term follow-up. Our meta-analysis demonstrated significant differences in the incidence of all-cause mortality and MACE between the two treatment modalities and confirmed the superiority of adrenalectomy.

Patients with PA after adrenalectomy are also associated with regression of left ventricular hypertrophy (38), improved metabolic outcomes, and increased insulin sensitivity (40). Researchers further found an improvement in myocardial fibrosis, carotid intima-media thickness, and arterial stiffness after adrenalectomy but not MRA therapy (41, 42). Some evidence further demonstrated that the endothelial dysfunction could be mitigated with surgical treatment in patients with PA (43), adding value to adrenalectomy. Thus, MRA therapy remains debatable regarding its ability in antagonizing metabolic risks and mortality among patients with PA.

Subgroup analyses and meta-regression

We found adrenalectomy delivers significantly better clinical outcomes, both composite primary outcomes and MACE, than MRA does in patients with unilateral PA. Because patients with unilateral PA could choose adrenalectomy or medical treatment, we raise an important issue that adrenalectomy could be a better choice in regard to clinical outcomes. From our metaregression, we found that patient’s age would influence the treatment outcomes, with an 8.4% increase in OR per year increase of chronologic age. In other words, the benefits of adrenalectomy might be flattened as patients age, making surgical advantages confounded by age. The reason might be that older patients with PA usually have a longer history of PA and hypertension, and the accumulated aldosterone- and hypertension-related damage could hardly be fully restored after surgery. Similarly, the international Primary Aldosteronism Surgical Outcome study also found that younger patients have a higher likelihood of complete clinical success (44). We hypothesize that conservative medical treatment could be considered in the older unilateral patients who were newly diagnosed. Further study is needed to define a proper age cut-off.

The benefit of adrenalectomy over MRA was constant among various subgroup characteristics of patients with PA. However, only when serum potassium level was not lower than 3.5 mmol/L (normokalemia) could adrenalectomy deliver more favorable composite primary outcomes. The results were likely attributable to different MRA dosages. However, limited enrolled cohorts and large s.d. made the interpretation difficult. Underpowered subgroups might also explain the non-significant finding.

In a recent multinational and multicenter study, Ohno et al. reported that adrenalectomy rates in unilateral PA were lower in Asia compared to those in Europe (45), which were often due to physician-derived factors such as good BP control, normokalemia, and adrenal venous sampling-related issues. Our results echoed their findings. Among the included studies, most patients with unilateral PA in Europe and North America underwent adrenalectomy as shown in Table 1. However, considerable proportions (44.2 and 36.5%,, respectively) of patients with unilateral PA in two recent cohort studies conducted in Asia (23, 27) were treated medically rather than surgically. The reasons were, however, not reported systemically. We further conducted a subgroup analysis of composite primary outcomes differentiating studies in which some patients with unilateral PA underwent adrenalectomy from studies in which most patients with unilateral PA underwent adrenalectomy (Supplementary Fig. 12). The result showed no significant difference between adrenalectomy rates and subsequent cardiovascular outcomes.

Reliability of the included data

With the aid of TSA, we perceived the sustained favorable outcomes of adrenalectomy over MRA, and most importantly, we are confident to emphasize that our findings and conclusions are robust since the patient number recruited far exceeded the RIS for interpretation of TSA.

Since the RIS of TSA has already been reached, the monitoring boundary crossed, and the consistent trend observed, we could conclude that the current evidence is sufficient to demonstrate the more advantageous effect of adrenalectomy for patients with PA in all-cause mortality and/or MACE. Nonetheless, because our conclusion was yielded from studies that consisted of various study designs and clinical scenarios, the Z curve crossed the neutrality line once in an earlier study (16).

Strengths and limitations

Guidelines published in the United States, Japan, and France (5, 46, 47) suggested that adrenalectomy of the affected side should be performed for patients with unilateral PA to normalize the BP and plasma aldosterone concentration. However, randomized controlled trials as well as high-quality prospective studies in support of the guidelines were lacking. Our result from this meta-analysis further supported and justified the current recommendation of multiple international guidelines.

Since cardiovascular events were frequently encountered in patients with PA, we collectively investigated MACE and all-cause mortality as the composite primary outcomes. The adoption of composite outcomes entitled us to encompass more diverse articles, overcoming the limitation of the scarcity of relevant materials. As a result, we could conclude robustly that adrenalectomy delivered better cardiovascular outcomes and lower all-cause mortality rates for patients with lateralized PA compared to MRA. To our knowledge, our work is the most updated meta-analysis to compare the all-cause mortality of patients with PA who underwent surgical or medical treatment and confirmed the advantage of the adrenalectomy.

Understanding the limitations of non-randomized controlled and retrospective studies, we introduced ROBINS-I tool for estimating the potential bias of included studies. Most enrolled studies were within low-to-moderate risk of bias, while few retrospective database studies possessed higher risks of bias.

Our meta-analysis comes with some limitations and should hence be interpreted carefully. First, while six out of the nine selected studies focused on the comparison of patients with PA vs EH patients, only the remaining three studies were designed for patients with PA alone. This led to the fact that only one study matched the patients with PA in surgical vs medical groups for their baseline characteristics. Since most of the included studies did not randomize the patients with PA in surgical vs medical groups, our results might be affected by these potential confounding factors as presented in the ROBINS-I risk assessment.

Second, although individual components of our composite primary outcomes, the all-cause mortality and MACE, have reached statistical significance to the advantage of adrenalectomy, they both harbored substantial heterogeneities. This might stem from factors such as different patient demographics, diverse baseline characteristics and comorbidities, lack of accumulated MRA dosage data, unmatched control groups, and discrepancy in the duration of follow-up among studies. Third, baseline characteristics between medically or surgically treated patients were imbalanced in some studies, meaning the selection of patients for adrenalectomy might be biased. For instance, in the study by Wu et al. (19) and Chang et al. (25), patients in the MRA group were significantly older and/or had diabetes more frequently than patients in the adrenalectomy group. In another study (17), medically treated patients had a higher BMI than surgically treated patients. This implies that there could be selection bias in these studies. However, according to our subgroup analysis (Supplementary Fig. 13), the balanced and imbalanced subgroups reported similar composite primary outcomes, suggesting that age and comorbidity imbalance in some studies have minimal influence on the overall meta-analysis. Fourth, the definitions of MACE varied among studies and the plasma renin levels during MRA treatment were not reported might also bias our results. However, while renin levels were demonstrated valuable in Hundemer’s studies (24), the information is still relatively new and therefore not available in older studies. Moreover, the concept has yet to be proven to be beneficial in prospective studies. Finally, APA and IHA were two distinct disease entities (48) with different presentations, suitable for adrenalectomy and MRA treatment, respectively. Nevertheless, our subgroup analysis showed the positive results of adrenalectomy were constant in patients with unilateral PA.

A more appropriate study would be a randomized trial comparing unilateral PA treated by adrenalectomy vs unilateral PA treated by MRA. However, this could delay the potentially curative surgery by many years, which is unethical for those participants randomized to the MRA arm. The unilateral PA subgroup analysis in our study, although with a relatively small sample size, sheds a light on this controversial question.