Proportion of Patients With Hypertension Resolution Following Adrenalectomy for Primary Aldosteronism: A Systematic Review and Meta‐Analysis

Abstract

Unilateral primary aldosteronism (PA) is often treated with adrenalectomy, but hypertension resolution rates are variable. A valid estimate of the postoperative normotension rate is necessary to inform the utility of PA testing and treatment. The authors searched MEDLINE In‐Process & Other Non‐Indexed Citations, Embase, and Cochrane Central Register of Controlled Trials. Prospective adult cohort studies with surgically treated PA that reported resolution of hypertension without the aid of medications were included. Among 2620 abstracts identified by the search, 25 studies in the systematic review with data on 1685 patients were investigated. The pooled proportion of normotension following adrenalectomy was 52% (95% confidence interval, 0.44–0.60). Meta‐regression demonstrated a significant negative association between length of follow‐up and proportion of normotension, with normotension dropping by 6.7% per year of follow‐up (coefficient −0.006; 95% confidence interval, −0.01 to 0.002). Overall, approximately half of the patients experienced hypertension resolution, although this outcome may not be durable in all patients.

Primary aldosteronism (PA) is defined as a form of drug‐resistant and potentially curable hypertension with evidence of excess aldosterone secretion, suppressed plasma renin activity, and often hypokalemia.1 Initially thought to be rare, PA is the leading cause of secondary hypertension. With increased screening and detection, the incidence of PA among hypertensive patients is currently reported at approximately 10%.2 This syndrome has multiple etiologies, including inherited gene mutations,3 development of autoantibodies,4 and ectopic aberrant functional adrenal receptors.5 The clinical management is guided by the definition of unilateral vs bilateral adrenal involvement. With rare exception, adrenalectomy is reserved for unilateral adrenal disease.

The proportion “cured” following adrenalectomy for patients with PA has been reported to range from 0% to 100%.6 However, “cure” is variably defined in the literature. These definitions include normal blood pressure (BP), normal BP without aid of antihypertensive medications, a reduced need for antihypertensive medications, and biochemical normalization, among others. Biochemical normalization of the aldosterone‐renin ratio is likely a necessary part of a definition of “cure.” At this time, however, there is no global consensus on the interpretation of postoperative aldosterone‐renin ratios with the existence of variable assays and cutoffs for even initial diagnosis. Therefore, from the patient's perspective and in the absence of patient‐level clinical end point data, attainment of medication‐free normotension may be the most uniform and recognizable health benefit of surgical PA treatment. Individually reported results of long‐term resolution of hypertension without antihypertensive medications are between 30% and 70% following adrenalectomy.7

The primary aim of this systematic review and meta‐analysis was to determine the proportion of hypertension resolution without the aid of antihypertensive agents following adrenalectomy in patients with PA. To our knowledge, no previous meta‐analysis has been performed. A secondary objective was to identify patient‐level factors reported to be associated with hypertension resolution in patients managed with adrenalectomy.

Methods

Protocol

This systematic review and meta‐analysis was conducted and reported in accordance with the Meta‐analysis of Observational Studies in Epidemiology and Preferred Reporting Items for Systematic Reviews and Meta‐Analyses statement (Study Protocol S1).8, 9

Data Sources and Search Strategy

We searched the following electronic databases without restrictions: MEDLINE In‐Process & Other Non‐indexed Citations (1946 through January 2016), Embase (1980 through January 2016), and Cochrane Central Register of Controlled Trials (1991 through January 2016). Our search strategy was developed in consultation with an experienced research librarian. Databases were searched using a combination of four search themes: (1) relevant diagnosis (PA), (2) relevant intervention (adrenalectomy, unilateral or bilateral), (3) relevant outcomes (proportion of hypertension resolution), and (4) study designs (observational, cohort studies). The initial search was completed using the Boolean operator “or” to combine and map the Medical Subject Heading (MeSH) for each theme. The search themes were combined using the Boolean “and” operator. The MeSH terms and keywords used can be viewed in Study Protocol S2 and the MEDLINE search strategy can be found in Study Protocol S3. We reviewed the reference lists of the search results and key papers to find other potential articles for inclusion.

Study Selection

Two reviewers (J.B. and M.E.) independently evaluated articles in a two‐stage process. We initially reviewed all abstracts retrieved in the electronic database search and identified those abstracts in which adrenalectomy appeared to be used for treatment of adrenal hypertension in adults 18 years or older for full text review. Any disagreement was resolved by consensus.

In the second stage, reviewers (J.B. and M.E.) completed a full text review of articles that met the above inclusion criteria. Full‐text articles were included in the systematic review if they included adults (defined as 18 years or older) with adrenal hypertension, evaluated hypertension resolution after adrenalectomy, and used a prospective cohort study design.

The primary outcome was the resolution of hypertension following adrenalectomy. For the purposes of our study, resolution of hypertension was defined as a normal BP in the absence of antihypertensive medications. Studies were included if they reported prevalence of hypertension resolution as a proportion using our definition.

Any disagreement between reviewers was resolved by consensus and recorded. The kappa (κ) statistic was used to assess agreement between reviewers regarding the inclusion of full‐text articles.10

Data Extraction and Quality Assessment

Both reviewers independently performed data extraction, and disagreement was resolved by consensus. The following data were extracted: number of patients, patient demographics, duration of hypertension, tumor pathology, definition of hypertension resolution, proportion of hypertension resolution, follow‐up rate, and follow‐up duration.

Each reviewer assessed each included publication separately for risk of bias. The evaluated study quality items included description of demographics, baseline differences, enrollment of consecutive patients, study design, follow‐up interval, and participant retention rate. These measures were scored by each reviewer using the Newcastle‐Ottawa Scale (NOS)11 for cohort studies, which is made up of eight dichotomous criteria that assess the selection of cohorts, cohort compatibility, and outcome. To evaluate the selection of the cohort, representation of the cohort, selection of the nonexposed cohort, ascertainment of exposure, and demonstration that the hypertension resolution was not present at the start of the study were evaluated. Comparability between studies was assessed based on whether demographics were described, baseline differences reported, and confounders discussed. The risk of bias for outcome was also assessed using assessment of outcome, follow‐up time interval, and adequacy of follow‐up.

Data Synthesis and Analysis

The proportion of cure was calculated by dividing the number of patients with normotension without antihypertensive medications following adrenalectomy by the total number of patients who underwent adrenalectomy for each study. If the authors reported the proportion of cure following adrenalectomy based on pathology, and therefore presented two separate cohorts, these cohorts were extracted separately. For overall proportion of hypertension resolution, the total sample size was combined.

The pooled proportion of cure from each study was combined across studies using Mantel‐Haenszel‐weighted DerSimonian and Laird random‐effects models.12 Calculating I ² and Q statistics as well as conducting tests of homogeneity quantified heterogeneity across studies. The I ² statistic quantifies the percentage of variability between studies due to factors other than sampling variability.13

In an effort to explore potential sources of heterogeneity and to enhance understanding of factors associated with different rates of cure, we conducted a series of stratified analyses to determine whether the pooled proportion of cure varied by type of pathology (aldosterone‐producing adenoma [APA] and bilateral idiopathic hyperaldosteronism [IHA]), mean follow‐up interval (3–12 months, 13–24 months, 25–36 months, and >36 months), and duration of hypertension (<6 years and >6 years). Studies were included in the pathology stratification if all patients in the cohort had the same pathology. Duration of hypertension was stratified using mean duration of hypertension greater than or less than 6 years, as reported in the literature as a predictive factor for resolution of hypertension following adrenalectomy.14

Meta‐regression was performed by duration of follow‐up and duration of hypertension using mean follow‐up (months) and duration of hypertension (months) as continuous variables.

Finally, small‐study effects potentially due to publication bias were assessed through visual inspection of funnel plots and Begg's funnel plot asymmetry test.15 All statistical analysis was performed using Stata, version 11.0 (Stata Corp., College Station, TX).

Results

Search Results

The results of our search strategy are summarized in Figure 1. Among 2620 citations identified for review, 248 met inclusion criteria for full‐text review. Of these 248, 25 articles were included in the systematic review and meta‐analysis, including data on 1685 patients.7, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39 There was excellent agreement between reviewers on inclusion of full‐text articles (k=0.93).

Figure 1.

Study flow.

Characteristics of Included Studies

The characteristics and study outcomes of the 25 included cohort studies are shown in the Table. Publication dates ranged from 1985 to 2015, while the number of study participants varied from 17 to 212. The mean age was reported in 17 of the 25 cohorts and ranged from 44.5 to 56.7 years. The duration of preoperative hypertension was reported in 10 of the included cohort studies and ranged from 63.6 to 162 months. The mean serum potassium level preoperatively was reported in 11 of the studies and ranged from 3.1 to 3.5 mmol/L. Of the 25 included prospective cohort studies, 11 reported the mean number of BP medications prior to adrenalectomy and ranged from 1.1 to 3.5. Adrenal venous sampling was reported as a confirmatory tool in establishing the diagnosis for PA patients in four of the included studies, with an additional four studies reporting between 25% and 75% undergoing the confirmatory diagnostic test.

Table 1.

Characteristics and Study Outcomes of the 25 Included Cohort Studies

Study	Year	Study Location	Participants, No.	Mean Age, y	Mean BMI	Duration of Hypertension, mo	Mean Pre‐Op Potassium. mmol/L	Mean Number of BP Medications Pre‐Op
Bernini et al7	2012	Italy	19	45	27.7	82.8	3.1	1.1
Catena et al16	2007	Italy	54	53	28.6	120	3.2	2.54
Chioncel et al17	2012	Romania	17	44.9	NR	NR	NR	3.2
Dobrucki et al18	2006	Poland	25	56.7	29.2	NR	3.21	NR
Giacchetti et al19	2009	Italy	42	50	26.9	99.6	3.35	2.1
Gockel et al20	2007	Germany	40	NR	NR	NR	NR	NR
Gordon et al21	1989	Australia	26	NR	NR	NR	NR	NR
Grady et al22	1996	United States	32	NR	NR	NR	NR	NR
Groth et al23	1985	Switzerland	50	NR	NR	NR	NR	NR
Holaj et al24	2015	Czech Republic	21	51.4	29.2	144	3.5	3.5
Horky et al25	1987	Czech Republic	38	NR	NR	NR	NR	NR
Lin et al26	2012	Taiwan	20	46	NR	82.8	3.3	1.9
Meria et al27	2003	France	212	48	NR	144	3.47	NR
Mourad et al28	2008	France	58	52	NR	137	NR	2.6
Omura et al29	2004	Japan	50	NR	NR	NR	NR	NR
Pang et al30	2007	Australia	62	50.2	NR	NR	NR	2.6
Rossi et al31	2011	Italy	19	NR	NR	NR	3.4	NR
Rossi et al32	2013	Italy	110	NR	25.9	NR	3.19	2.7
Rutherford et al33	1998	Australia	68	51	NR	NR	NR	NR
Shen et al34	2009	China	163	42	NR	NR	NR	NR
Sukor et al35	2010	Australia	22	50	NR	NR	3.1	2.7
Walz et al36	2008	Germany	183	49.6	NR	122.4	NR	NR
Worth et al37	2015	United States	58	52.6	31.5	162	3.5	3.3
Wu et al38	2015	China	148	56.3	NR	NR	NR	NR
Zhang et al39	2004	China	130	44.5	NR	63.6	NR	NR

Abbreviations: BMI, body mass index; BP, blood pressure; NR, not reported.

Risk of Bias Assessment

A summary of the methodological details of the cohort studies is presented in Table S1. As these were surgical cohorts, they were limited by not having a nonexposed (ie, medically managed) cohort. Of all other evaluated indicators of cohort selection, the included studies were strong, as the cohorts were made up of patients referred to tertiary centers for management of PA, they used surgical records, and all patients had hypertension at the time of adrenalectomy. All studies reported the primary outcome (proportion of hypertension resolution), as it was one of the inclusion criteria. The mean follow‐up time interval was not reported in several studies. The proportion of patients who underwent follow‐up was greater than or equal to 85% in all but one study.

Proportion of Hypertension Resolution

A total of 1685 patients were included in the 25 cohort studies reporting a pooled estimate of proportion of hypertension resolution of 52% (95% confidence interval [CI], 0.44–0.60; Figure 2). The data were heterogeneous (I ²=92.1%).

Figure 2.

Overall proportion of hypertension resolution following adrenalectomy.

Exploration for Sources of Heterogeneity

Stratification was conducted to determine sources of heterogeneity. For the cohorts reporting a mean duration of hypertension of less than 6 years, the proportion was 52% (95% CI, 0.25–0.79; I ²=90.8%; Figure S1) compared with 41% (95% CI, 0.29–0.48; I ²=88.5%; Figure S1) for cohorts with means ≥6 years. The meta‐regression that examined the association of duration of hypertension in months as a continuous variable with proportion cured was not significant (coefficient, −0.0007; standard error, 0.001; 95% CI, −0.003 to 0.002 [P=.54]; Figure S2).

From the included studies, 12 reported the proportion of hypertension resolution for cohorts with APA pathology exclusively and this was found to be 54% (95% CI, 0.41–0.67; I ²=91.1%; Figure S3). This is compared with two cohorts that reported rates for IHA pathology exclusively, and found that all patients with IHA remained hypertensive following adrenalectomy.

The results of this meta‐analysis were also stratified by follow‐up interval. For included studies with mean follow‐up intervals greater than 36 months, the prevalence of hypertension resolution was 42% (95% CI, 0.33–0.51; I ²=79.1%; Figure S4) compared with 57% for follow‐up intervals of 25 to 36 months (95% CI, 0.44–0.70; I ²=62.5%; Figure S4), 60% for follow‐up intervals of 13 to 24 months (95% CI, 0.20–1.01; I ²=92.8%; Figure S4), and 60% (95% CI, 0.20–1.01; I ²=92.8%; Figure S4) for follow‐up intervals less than 12 months. Meta‐regression examining the association between mean follow‐up interval in months for studies that followed their patients for a minimum of 12 months postoperatively and the proportion of hypertension resolution was significant (coefficient −0.0056, standard error 0.0020; 95% CI, −0.0093 to −0.0019 [P=.006]; Figure 3). Practically, this analysis shows that with each year of follow‐up, there is a 6.7% drop in the proportion of patients who remain normotensive without medications.

Figure 3.

Meta‐regression of proportion of hypertension resolution by follow‐up interval.

Small‐Study Effects Analyses

Analyses of the effects of small studies were completed. There was no evidence of publication bias when assessed through visual inspection of funnel plots (Figure S5) and Begg and Mazumdar (rank correlation) test for asymmetry.15

Discussion

This systematic review and meta‐analysis of 25 cohort studies suggests that approximately half of the patients who undergo adrenalectomy for the treatment of unilateral PA will experience normotension without the use of antihypertensive medications. However, our analysis suggests that this may not be a durable outcome in all patients over the course of prolonged follow‐up. This does not necessarily mean that the patients were not “cured” of PA or that they derived no lasting benefit from specific PA treatment; rather, as the mean follow‐up interval in these studies was as long as 77 months, it is possible that some patients developed either essential hypertension or another disorder associated with hypertension such as chronic kidney disease. PA is known to carry a high risk of causing renal impairment and such renal injury may not be fully reversible even upon resolution of the PA state.40 It is also conceivable that a prolonged delay in initial PA diagnosis could contribute to permanent renal injury with persistent hypertension even after definitive correction of the PA state. Biochemical recurrence of PA remains a theoretical possibility but is not widely recognized or reported in other literature sources. Contralateral suppression of the remaining adrenal gland leading to normotension in the postoperative period with a recurrence of hypertension once the remaining adrenal gland has recovered could also explain this finding. Long‐term biochemical monitoring of aldosterone levels may help further explain any underlying mechanisms of recurrent hypertension. The possibility of nonpersistent hypertension resolution should not be interpreted as an argument against adrenalectomy since a comparative group of the alternate strategy of primary medical therapy has not been studied in a similar fashion. Primary medical therapy has its own costs and potential complications and requires regular long‐term clinical follow‐up, all of which would need to be considered in a true comparison to surgical management.

Recent efforts have focused on identifying predictive factors that may determine which patients obtain hypertension resolution following adrenalectomy, and the presence or absence of such factors may account for some of the residual heterogeneity observed in this review. These proposed factors include age at time of surgery, sex, pathology, duration of hypertension, presence of hypokalemia at diagnosis, number of antihypertensive drugs required preoperatively, and positive response to mineralocorticoid receptor antagonists (MRAs) such as spironolactone.19, 41 Again, these studies are often limited by virtue of their derivation from small cohorts with relatively short‐term follow‐up.

A predictive model called the aldosteronoma resolution score was developed by Zarnegar and colleagues14 that features four predictive factors: number of antihypertensive medications preoperatively, duration of hypertension, body mass index, and sex. One short‐term study in Japanese patients has suggested that the score has reasonable accuracy to predict normotension.42 Our large meta‐analysis and stratified analyses do not support using the duration of hypertension as a predictive factor. Rather, our analysis suggests that there is no significant difference between the duration of hypertension preoperatively and the proportion of hypertension resolution. This is a novel finding that may be explained by the fact that prior studies of predictive factors are necessarily based on retrospectively analyzed clinical case data. The actual “onset” of hypertension in a given individual may be entirely unknown or inaccurate if dependent on patient recall. Thus, the duration of preexisting hypertension may still be a truly predictive factor for future hypertension resolution but its analysis through patient chart data abstraction may be inadequate to accurately determine the exact relationship. This is a key consideration when developing future guidelines and argues for the importance of a workup of all patients with secondary hypertension regardless of how long they have had a diagnosis of hypertension. The three other predictive factors included in this model could not be evaluated in our meta‐analysis without patient‐level data.

Our meta‐analysis supports pathology as being a possible predictive factor for a medication‐free BP outcome. From the studies that qualified for this meta‐analysis, two suggested a 0% hypertension resolution rate when unilateral adrenal hyperplasia was the final pathology. This result is highly preliminary as it was derived from only six patients, and therefore has low power, but it may suggest that the magnitude of benefit may be smaller for patients with unilateral IHA pathology.23, 25 Other papers not included in the meta‐analysis do appear to show clinical benefit to surgical intervention even with diffuse hyperplasia as the pathology.43 Recent work in functional histopathology has now shown that the traditional gross cellular differentiation between adenoma and hyperplasia may sometimes be functionally incorrect and therefore future studies that aim to determine pathology‐linked clinical outcomes will likely need to use a more modern and functional approach to the subtype definitions.44

Study Limitations

There are a number of limitations of this study. The studies included were all prospective, surgical cohorts but patient selection bias cannot be fully excluded. One of the major challenges in the management of PA is making the correct diagnosis. Patients often present for workup of PA on several antihypertensive medications, including MRAs, which can complicate the interpretation of the aldosterone‐renin ratio. Another important point is that diagnostic techniques have evolved over time including the use of adrenal vein sampling (AVS), which was likely underused in many of the older cohorts, in comparison to present recommendations. While this raises the possibility of inappropriate selection of patients for surgery, it may also have limited the inclusion of unilateral PA patients who would have benefitted from surgery but were not treated for want of a clear diagnosis. Future outcome‐oriented study cohorts should have the diagnosis supported by AVS and in doing so, will ensure that only truly unilateral surgical cohorts are included in analysis.

Second, we were unable to access patient‐level data and therefore were limited in the outcome predictors that could be analyzed. With patient‐level data, it would be possible to determine the predictive impact of factors such as younger age, degree of hypokalemia, treatment with MRAs, and number of antihypertensive medications, which have all been reported as useful in predicting outcome following adrenalectomy.45

Third, normotension was achieved in each of these cohorts, but biochemical cure was not reported in many, and long‐term clinical outcomes were not assessed. Several studies have looked at the impact of PA compared with essential hypertension and found that patients with PA have significant cardiac and renal complications compared with hypertension alone.46, 47, 48 This suggests that biochemical cure as an outcome may be of equal or greater significance for the overall health of these patients and that the maximal benefits of PA management might only be realized when the aldosterone excess is removed in addition to hypertension control. As PA registries are developed in the future, it will be useful for postoperative biochemical measures of renin and aldosterone to be included as an important variable for inclusion of reports of surgical treatment outcomes.

Fourth, these are surgical cohorts that do not include patients who either choose to be medically managed or are not offered surgery for some other reason, such as comorbid medical conditions. Therefore, our pooled estimate of cure is limited to only those who are selected to undergo surgery and we are unable to draw inferences about the probability of the resolution of hypertension in an unselected group of patients with PA.

The final limitation of this study is that we focused on a specific single primary outcome with unproven impact on hard patient outcomes. However, as it would be difficult to conduct a large, prospective, clinical event–driven, randomized controlled trial comparing surgical and medical interventions for the treatment of PA, we need to rely on the published historical data with meta‐analysis. The percentage of patients who experience clinical benefits, including improvement in BP management and reduction in the number of antihypertensive medications, is likely higher than the proportion of complete resolution reported in this meta‐analysis. Less than half of the included studies published postoperative biochemical data. It was not possible to analyze biochemical cure to confirm both the diagnosis and resolution of PA in these cohorts without standardization of the aldosterone‐renin ratio measurement and interpretation between centers.

Since not all patients who undergo surgery have hypertension resolution and many patients with PA have well‐controlled BP with MRA drugs,49 there is an ongoing debate as to the overall clinical importance of investing the time and effort needed to pursue a PA diagnosis to the point of determining surgical candidacy. In most centers, patients with PA will require anatomical imaging and adrenal vein sampling to define PA as either bilateral or unilateral adrenal disease, with patients with unilateral forms being offered surgery.50 Early data suggest that clinical outcomes may be superior in patients who undergo adrenalectomy but the steps required to define surgical candidates have both risks and costs.51 For a disease such as PA with seemingly high prevalence, interventional cost and risk‐benefit analyses will be highly dependent on accurate determination of the beneficial outcomes seen across multiple populations of affected patients.

The bulk of the global published experience to date with surgery for PA is that just over 50% of patients achieve medication‐free normotension. Larger, longer studies may help to better define predictors of outcomes, and prospective clinical trials are needed to confirm patient‐level benefits of treatments. Until then, cost‐effectiveness studies may use the present data for the estimates of benefit and individual patients may be counseled as to their expected BP outcomes with surgical intervention.

Conclusions

The costs associated with the workup and management of PA are not negligible.52 It is therefore important to have an accurate understanding of the clinical benefits for the patient, including control of hypertension and quality of life, as well as the impact on the healthcare system such as reduction in follow‐up visits and postoperative medications associated with performing adrenalectomy in these patients.35 The results of this meta‐analysis could be used to inform a future economic analysis. This pooled estimate of the magnitude of surgical benefit may be useful in patient counseling, guideline writing, and possibly future economic models of different PA treatment approaches.

Sources of Funding

DMR is supported by an Alberta Innovates Health Solutions Population Health Investigator Award.

Disclosures

The authors have nothing to disclose.

Supporting information

Study Protocol S1. Focused Question:

Study Protocol S2. Comprehensive Search Themes

Study Protocol S3. MedLine Search Strategy

Table S1. Methodological Details

Table S2. Stratification of Sources of Heterogeneity

Figure S1. Proportion of Hypertension Resolution Stratified by Duration of Hypertension Preoperatively

Figure S2. Duration of Hypertension Preoperatively as a Continuous Variable

Figure S3. Proportion of Hypertension Resolution for Aldosterone‐Producing Adenomas

Figure S4. Proportion of Hypertension Resolution Stratified by Follow‐Up Interval

Figure S5. Funnel Plot.

J Clin Hypertens (Greenwich). 2016;18:1205–1212. DOI: 10.1111/jch.12916. © 2016 Wiley Periodicals, Inc.