Effect of diuretics on plasma aldosterone and potassium in primary hypertension: A systematic review and meta‐analysis

By contrast with drugs inhibiting the renin‐angiotensin‐aldosterone system (RAAS), diuretics stimulate renin release by the kidneys. Although plasma aldosterone (PA) is thought to be mainly regulated by RAAS activity, serum potassium has been shown to be an important factor in animal models and humans. Here we perform a systematic review and meta‐analysis of randomised controlled trials (RCT) in hypertension investigating the effects of diuretic therapy on PA and the correlation of change in PA with that of potassium and blood pressure (BP).

Compared to RAAS inhibitors, diuretics have a more complex mechanism of action 3 which includes an initial reduction in plasma volume and a sustained decline in peripheral resistance, thereby improving an underlying haemodynamic defect of hypertension. [4][5][6][7] Under acute and chronic conditions, diuretics induce an increase in plasma renin activity (PRA) 8 but whether diuretics also increases PA has been debated. 9 Apart from the level of activation of the RAAS, potassium, [10][11][12] which is also affected by diuretic treatment, plays an important role in the regulation of PA production and some diuretics (such as mineralocorticoid receptor antagonists) are known to have direct inhibitory action on aldosterone formation.
The objective of the present study was to perform a systematic review and meta-analysis of randomised clinical trials (RCTs) where diuretics were used to treat hypertension and measurements of PA were available before and after diuretic treatment to determine if they lead to a sustained increased in PA and whether this differs according to class of diuretic. Secondary objectives were to establish if there is correlation between difference in PA and that of serum potassium and if the decrease in BP relates to the difference in PA.

| Search strategy
This systematic review and meta-analysis was carried out in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, 13 similarly to our previously published review. 14 16 September 2020). Studies which were included were trials of diuretics used either as mono-or multitherapy in hypertension, and which examined how they affected plasma aldosterone +/À renin, serum potassium and blood pressure. The keywords used included "thiazide", "thiazide-like", "potassium-sparing", and "diuretic". Medical subject headings (MeSH) and non-MeSH terms were used to search the databases for relevant publications. The full search strategy for MEDLINE is provided in the Supporting Information. The review protocol was not registered but the PROSPERO database was searched before starting the review to ensure that a similar review was not ongoing.

| Study selection and eligibility criteria
Papers were initially screened by title and abstract. Studies were eligible for inclusion if they were an RCT performed in hypertensive human subjects ≥18 years old, examining antihypertensive effects of a thiazide, thiazide-like, loop, mineralocorticoid receptor antagonist (MRA) or potassium-sparing diuretic with a duration of at least 1 week. Studies investigating novel diuretics or "diuretic-like" drugs not currently licensed for clinical use were excluded as were studies in which subjects had pulmonary arterial hypertension or heart failure. We included in the analysis all the diuretic compounds currently licenced in the UK for the treatment of arterial hypertension. We also included drugs which were licenced for hypertension at the time that the original study was conducted. All studies were required to have examined PA with results available before and during diuretic treatment or, in the case of placebo-controlled trials, during treatment with placebo. Studies were eligible if diuretic therapy was added to either no previous therapy or to stable background treatment. The search was limited to the English language only and review articles were disregarded. Titles and abstracts were screened by one author (R.J.M.), and the same author reviewed the full-text articles.

| Data collection process
Data were extracted independently by one author (R.J.M.) using a standard form. This included author, year of publication, class of diuretic(s) and dose used, protocol (including presence of background therapy and whether placebo controlled), sample size, average age, sex distribution, ethnicity (if available) and prevalence of diabetes (if available). For the outcome measurements, mean (± standard deviation/standard error) of values for BP and aldosterone before and during diuretic (and before and during placebo treatment in placebocontrolled studies) and the difference between values on and before treatment were extracted from the relevent articles. Where available, the differences in renin, serum potassium and systolic blood pressure (SBP) were extracted. If standard deviations were not reported these were calculated from standard errors, P values or confidence intervals.
The duration of diuretic treatment at the time of measurement was also recorded. Where only graphical reports of measurements were available, an estimation from the graph was taken if it was judged to be accurate to within 10%. Units of aldosterone were converted to pmol/L for analysis if other units were used.

| Quantitative data synthesis and statistical analysis
Meta-analysis was conducted using Comprehensive Meta-Analysis Software Version 3 (Biostat, Englewood, New Jersey, USA). 15 Net changes in PA, renin, potassium and SBP were obtained as the difference from baseline after treatment with either diuretic or placebo. If there was no standard error of the mean change stated, it was estimated from the P value, number of observations and size of the change. A random-effects model was used to compensate for between-study heterogeneity in terms of demographic inconsistencies and different diuretic doses 16 with calculation of the standardised mean difference in PA and its 95% confidence interval (CI). The standardized mean difference expresses the size of the intervention effect in each study relative to the standard deviation of the measurement (in this case PA). Raw mean differences were calculated for plasma potassium, and SBP and standardised differences for PRA.
Statistical heterogeneity was assessed using Cochran's Q test. 17 P < .05 was considered statistically significant and all tests were two-tailed.

| Meta-regression
Random-effects meta-regression was performed using the method of moments to evaluate the association between standardised difference in PA and change in SBP in the overall data and in subjects previously untreated with another antihypertensive. The same method was used to examine the association between change in PA and change in serum potassium.

| Publication bias
Potential publication bias was assessed by inspection of Begg's funnel plot asymmetry and Egger's asymmetry tests. 18 We did not perform any risk of bias within individual studies because we were not examining the primary outcome of the studies.

| Description of studies
The study selection process is detailed in a flow chart as per PRISMA guidelines ( Figure 1

| Primary outcome meta-analysis: Effect of placebo and diuretic on plasma aldosterone
Placebo had a negligible effect on PA in the present analysis ( Figure 2D). The standardized difference in mean PA after placebo was À0.11 (95% CI À0.36, 0.14). With diuretic therapy, all diuretic classes led to a significant increase in PA but there was no betweenclass heterogeneity ( Figure 2A)  Figure 2B). There was a similar increase in PA for those in whom diuretic was added to previous antihypertensives: 0.301 (0.040, 0.563; Figure 2C). After separating MRA from potassium-sparing diuretics, the same homogeneity between classes was found.  TIE, tienilic acid. *More than one result in the same study.

| Meta-regression: Relation of change in plasma aldosterone to that of serum potassium after diuretic
Random-effects meta-regression was also used to examine whether the change in PA was associated with that of serum potassium and we found there was no relationship (coefficient À0.09, 95% CI À0.32, 0.15, P = .466; Figure 4A) with similar findings in studies where the participants were on no background therapy (coefficient À0.14, 95% CI À0.36, 0.07, P = .186; Figure 4B).

| Meta-regression: Relation of change in SBP to change in plasma aldosterone
Random-effects meta-regression was performed to examine whether change in SBP was associated with change in PA. In the overall population, changes in SBP were independent of changes in PA (coefficient À0.005, 95% CI -0.019, 0.008, P = .44; Figure 5A). However, in studies where the participants were not on background treatment with another antihypertensive, there was a significant relationship between the change in SBP and change in PA (coefficient À0.02, 95% CI -0.034, À0.01, P < .001; Figure 5B).

| Publication bias
The funnel plot of standard error vs effect size was asymmetric and suggestive of potential publication bias. Presence of publication bias was also suggested by Egger's linear regression (P = < 0.001). After adjustment of effect size for potential publication bias using the "trim and fill" correction, 14 potentially missing studies on the left side of the funnel plot were imputed, leading to a corrected overall effect size that was slightly but not significantly less than the initial estimate (0.21 (95% CI 0.17-0.26; Figure 6). The main finding of the present meta-analysis is that diuretics lead to an increase in PA that does not differ significantly between classes of diuretic, but which is significantly associated with change in SBP in previously untreated subjects. In the studies where PRA was also measured before and after treatment, a raise in PRA occurred (a finding in line with a previously published systematic review 14 ), demonstrating activation of the RAAS, which is likely to be the driver of the increased PA.
Whether activation of the RAAS could be harmful has been debated. The RAAS is a complex system in which angiotensin II acts through two main receptor subtypes, the AT1 and AT2 receptors. All classic physiological effects of angiotensin II, such as vasoconstriction, aldosterone production and water retention, are largely mediated by the AT1 receptor which promotes hypertension, endothelial Whilst there is evidence that PRA is helpful in selecting patients who will benefit from diuretic therapy, 67 the potential use of change in PRA in guiding dose-titration and selecting class of diuretic remains speculative and was not confirmed by a recently published systematic review. 68 PA is another potential biomarker that could be used to guide treatment. In our analysis the change in SBP related to that of PA. Failure of PA to rise after initiation of a diuretic could therefore be used to assess adherence to diuretic therapy and to guide dose adjustment, although this is speculative and would require testing in prospective studies. It could be also speculated that the use of a concomitant medication might play a role in limiting the rise in PA, which in turn could have a beneficial effect per se since that aldosterone may cause cardiac remodelling without affecting arterial pressure. [69][70][71] Apart from the RAAS, the other major factor regulating aldosterone secretion is potassium. In humans and in experimental animals, alterations in potassium balance as well as acute increments in serum potassium can stimulate aldosterone production. For example, in F I G U R E 4 Meta-regression plot of the association between mean changes in PA with diuretic therapy and the change in serum potassium (mEq/L) in (A) overall and (B) previously untreated with another antihypertensive. Std diff, standardised difference F I G U R E 5 Meta-regression plot of the association between mean changes in PA with diuretic therapy and the change in SBP in (A) overall and (B) previously untreated with another antihypertensive. SBP, systolic blood pressure; Std diff, standardised difference F I G U R E 6 Funnel plot displaying publication bias in the studies reporting the impact of diuretic therapy on PA change. Open diamond represents observed effect size; closed diamond represents imputed effect size. Std diff, standardised difference normal subjects, infusion of 10 mEq of potassium produces a 25% increase in plasma aldosterone. 10 Changes in dietary potassium intake for as little as 24 hours can also substantially modify the secretion of aldosterone from the adrenal glands induced by acute potassium administration: high dietary potassium intake enhances aldosterone secretion, while low potassium intake reduces it. 11 Our results suggest that variation of serum concentration of potassium per se has a limited effect in regulating PA. However, there are suggestions that this mechanism could be relevant in specific populations. 72 Finally, the subanalysis investigating MRA showed similar effects of these agents on PA compared to other diuretic classes. It has been reported that similarly to other RAAS inhibitors, 73,74 after an initial suppression/blockade of aldosterone, the PA level often returns to normal or even rises above pre-treatment levels. 75,76 This review is subject to several limitations. We were unable to stratify results by ethnicity (and difference in RAAS activity between ethnic groups have been described) because the majority of studies were performed in Caucasians and in many studies ethnicity was not reported. Studies in specific ethnic groups will be required to determine if effects of diuretics on PA differ according to ethnicity.
The use of background therapy in some studies and a variable dose in others prevent a useful estimate of the effect size relating to a standard dose of diuretic. The duration of studies was relatively short and very few studies were performed with loop diuretics (which are not commonly used in hypertension). The MRA/potassium sparing group was mostly composed of spironolactone, which in many trials was used at high dose unrepresentative of its current use in primary hypertension. Finally, the present study is limited by the lack of availability of some full-text articles.
In conclusion, this systematic review and meta-analysis demonstrates that diuretic therapy in hypertension leads to an increase in PA which does not differ between classes of diuretics. The use of drugs that might antagonise the effects of this rise in PA or the combination of other diuretics with RAAS inhibitors is a rational means to prevent the adverse effects of this rise in PA, but whether such a treatment strategy leads to BP-independent effects remains to be tested.