A systematic review and meta-analysis of the impact of mineralocorticoid receptor antagonists on glucose homeostasis.

BACKGROUND: Spironolactone, a nonselective mineralocorticoid receptor antagonist (MRA), may have a deleterious effect on glycemia. The objective of this review was to assess current knowledge on MRAs' influence (spironolactone, eplerenone, and canrenone) on glucose homeostasis and the risk of diabetes.
METHOD: A systematic review was conducted using the Medline database on articles published from 1946 to January 2017 that studied the effects of MRAs on any glucose-related endpoints, without any restrictions regarding the participants' characteristics.Study design, patient population, dose and duration of intervention, and the quantitative results on glycemic markers were extracted, interpreted for result synthesis, and evaluated for sources of bias. From the articles included in the qualitative analysis, a select number were used in a meta-analysis on studies having measured glycated hemoglobin (HbA1c) or risk of diabetes.
RESULTS: Seventy-two articles were selected from the Medline database and references of articles. Results on spironolactone were heterogeneous, but seemed to be disease-specific. A potential negative effect on glucose regulation was mainly observed in heart failure and diabetes trials, while a neutral or positive effect was detected in diseases characterized by hyperandrogenism, and inconclusive for hypertension. Interpretation of data from heart failure trials was limited by the small number of studies. From a meta-analysis of 12 randomized controlled studies evaluating spironolactone's impact on HbA1c in diabetic patients, spironolactone had a nonsignificant effect in parallel-group studies (mean difference 0.03 [-0.20;0.26]), but significantly increased HbA1c in crossover studies (mean difference 0.24 [0.18;0.31]). Finally, eplerenone did not seem to influence glycemia, while limited data indicated that canrenone may exert a neutral or beneficial effect.The studies had important limitations regarding study design, sample size, duration of follow-up, and choice of glycemic markers.
CONCLUSION: Spironolactone may induce disease-specific and modest alterations on glycemia. It is uncertain whether these effects are transient or not. Data from the most extensively studied population, individuals with diabetes, do not support a long-term glycemic impact in these patients. Further prospective studies are necessary to establish spironolactone's true biological effects and their clinical implications.

Introduction

Increased activity of the renin-angiotensin-aldosterone system (RAAS) is present in many cardiovascular (CV) diseases, including hypertension and heart failure (HF).[ Aldosterone contributes to many of the negative processes related to RAAS in these pathologies, such as myocardial fibrosis, sodium retention, increased blood pressure, and inflammation.[ Mineralocorticoid receptor antagonists (MRAs), such as spironolactone (SPIRO) and eplerenone (EPLE), block the deleterious effects of aldosterone that are mediated by the mineralocorticoid receptor. Consequently, this pharmacological activity makes MRAs effective in treating hypertension, particularly resistant hypertension,[ and in reducing the risk of morbidity and mortality in HF patients.[ MRAs are also used for the treatment of primary aldosteronism[ and edema associated with liver cirrhosis or nephrotic syndrome.[

Despite its beneficial impact on CV events, SPIRO, a nonselective MRA, has “off-target” effects on progesterone, androgen, and glucocorticoid receptors. These effects include the displacement of androgen from the androgen receptor, inhibition of enzymes in the testosterone synthesis pathway (17α-hydroxylase and 17–20 desmolase), increases in the conversion of testosterone to estradiol,[ and inhibition of estrone sulfatase and 17β-HSD type 1 which leads to increases in estradiol pool.[ These mechanisms cause gynecomastia, breast tenderness, menstrual irregularities,[ and erectile dysfunction.[ Although these off-target effects are undesirable in most conditions, they are useful for the treatment of disorders related to hyperandrogenism.[ Thus, SPIRO is also a treatment for idiopathic hirsutism[ and polycystic ovary syndrome (PCOS), a disease characterized by excess androgens, oligoovulation or anovulation and/or polycystic ovaries.[

There is a growing amount of evidence suggesting that SPIRO's “off-target effects” could also include detrimental effects on glucose homeostasis.[ A potential cause of this negative effect is the fact that SPIRO increases cortisol levels through an off-target effect: the blockade of the glucocorticoid receptors.[ Cortisol, a glucocorticoid, increases glucose through lipolysis and gluconeogenesis. On the other hand, EPLE, a selective MRA, has a very low activity on other steroid receptors.[ As such, it is believed that it does not inhibit adrenal cell aldosterone or cortisol production and does not affect glucose metabolism.

Glucose intolerance and diabetes are already frequent comorbidities in some of the diseases that require treatment with an MRA, and are associated with an increased risk of CV events.[ Thus, it is critical to determine if MRAs modulate glycemia in any of the patient populations that use them. The objective of this article was to assess current knowledge on the subject in existing literature, in the context of growing use of MRAs in HF and in other diseases. Also, the information on potentially additional adverse effects of MRAs could be used by physicians to guide their treatment choices. We conducted a systematic review of randomized controlled trials (RCTs), prospective studies, and observational studies, evaluating the influence of the MRAs SPIRO, EPLE, and canrenone, regardless of comparator group, on any biomarkers of glucose homeostasis in a variety of populations. Healthy individuals, patients at risk of CV disease, HF patients, and patients with other non-CV diseases were included into this analysis. We then performed a meta-analysis with appropriate datasets.

Methods

Search strategy

A search was conducted on the Medline database on articles written from 1946 till January 2017. In addition, a manual search was performed on references of the retrieved articles from Medline, based on the eligibility criteria. The following search terms were used: glucose, or glucose metabolism disorders, or insulin, or glycosylated hemoglobin A; and steroid receptors, or aldosterone, or mineralocorticoid receptor antagonists, or spironolactone, or eplerenone; and humans, or double-blind method.

Eligibility criteria

Any prospective RCTs or prospective or retrospective cohort studies that contained measures of glucose metabolism, before and after treatment with an MRA, were reviewed. We did not put any constraints on the types of glycemic markers, because we wished to collect any information that was relevant to the effect on glucose control. MRAs were restricted to SPIRO, EPLE, and canrenone (an active metabolite of SPIRO). The MRA drospirenone was excluded because it is mainly used as a contraceptive. There were no limitations for the comparator or the absence of a comparator. However, studies in which an MRA was evaluated in combination with another drug but without any comparator group were excluded. For example, in the case where a combination of SPIRO with a thiazide diuretic was being used, the article was accepted only if the study design included a comparator group consisting of either 1 of these 2 drugs in monotherapy. A minimum treatment period with an MRA of 1 week was required for inclusion. As we were interested in comparing the effects of MRAs in various diseases, we included studies irrespective of study population (healthy, at risk of CV disease, HF, and other non-CV diseases), or whether the effect on glucose metabolism was part of the primary or secondary endpoints. We limited our language selection to English, French, and Russian.

Study selection, data extraction, and synthesis of results

Eligibility assessment and data collection was performed independently by the first and second authors. Any differences were resolved through discussions and consensus. Articles were selected after an evaluation of the title, abstract, or full article. The results on Medline were alphabetized by the first author's name in each study, to easily identify and eliminate duplicates. Data extraction was conducted using an MsExcel spreadsheet. The following characteristics were extracted: study design, sample size, disease of participants, study medication and dose, time of treatment and follow-up, the markers of glucose homeostasis, and effects of the study medication on the markers.

Although all markers of glucose homeostasis were collected for the systematic review, our primary endpoints were the change in glycated hemoglobin (HbA1c) and onset of diabetes in the context of RCTs, as these are markers of long-term glucose control. All available summary measures of glycemic markers were recorded: baseline and posttreatment means or medians, mean changes within treatment groups or treatment phases, mean differences between groups, and odds ratios or risk ratios.

Meta-analysis

Prospective RCTs that evaluated SPIRO's effect on HbA1c and that had a comparator group were also included into quantitative analyses. As HbA1c is an indicator of glucose control over a period of 3 months, we considered it to be the most reliable marker to include in the meta-analysis, as opposed to glucose or insulin that may vary greatly between blood tests. HbA1c data is reported as a mean difference (MD) and accompanying 95% confidence interval (CI) and was pooled using a Hartung-Knapp method random-effects model with the “meta” package in R version 3.1.3 (The R Project for Statistical Computing).[ Separate analyses were conducted for parallel-group versus crossover studies. We assessed presence of statistical heterogeneity using the Cochrane P value (P < .10 significant) and the degree of heterogeneity using the I2 statistic with a value >50% considered substantial.[

Quality and risk of bias

The first and second authors evaluated independently the quality and the risk of bias of each study considering the following criteria: study design (retrospective vs prospective; observational vs interventional), randomization, blinding (double-blind vs single-blind vs open-label), trial registration, choice of comparator, presence of a washout period, dose of study medication and regimen, duration of treatment and follow-up, sample size and statistical power, choice of glycemic markers, analytical methods, baseline characteristics/medication and between-group imbalances, quality of laboratory measurements, line of therapy for an MRA, and comprehensive description of methodology and results. However, we did not exclude studies based on this evaluation.

Ethical review

Ethical approval was not necessary for this study as it only included previously published summary data. It did not involve animal or human test subjects, and did not require access to any personal data.

Results

Figure 1 presents the selection process. From 1682 articles that were identified through the Medline database (excluding duplicates), 117 articles were excluded due to language barriers and 338 reviews were removed. Among the remaining articles, 873 were excluded from the title and 259 were excluded for the abstract. Finally, 35 articles were removed after reading the full text. An additional 12 articles were identified from the references of the articles that were found in the Medline search results. Thus, 72 articles were included into this literature review. Among these articles, 12 studies were chosen to be included in the meta-analysis according to our selection criteria, as they consisted of RCTs measuring effects on HbA1c. We did not have a sufficient number of studies on the risk of diabetes.

Figure 1

Flow of information.

Tables 1–8    present each study's characteristics. Studies were grouped according to different patient populations. A variety of markers of glycemia were evaluated. Synthesis of the findings according to each disease is presented in Table 9, grouped according to healthy individuals, patients at risk of CV complications, HF patients, and patients suffering from other illnesses unrelated to CV disease. Limits of each individual study can be accessed in the Supplemental Content.

Table 1

Results for healthy volunteers.

Table 6

Results for PCOS and hirsutism.

Table 6 (Continued)

Results for PCOS and hirsutism.

Qualitative review

Healthy volunteers

Only 2 small prospective studies were conducted on 18 and 13 healthy volunteers, and had a short follow-up period of 10 and 14 days, respectively (Table 1).[ The first study compared the use of high dose SPIRO (100 mg every 6 hours) in combination with adrenocorticotropic hormone (ACTH) or cortisol versus a glucocorticoid receptor antagonist (RU486).[ The second study included the use of 50 mg EPLE, without a comparator group.[ In both studies, the MRA exerted a neutral effect on glucose control, although HbA1c or the risk of diabetes was not evaluated.

Hypertension

We identified multiple studies (14 studies) that were performed on hypertensive patients. Eleven studies used SPIRO, and 3 used EPLE (Table 2 ). Studies with SPIRO included 7 RCTs,[ 1 prospective nonrandomized trial without controls,[ and 3 observational studies.[ Studies with EPLE consisted of 1 RCT[ and 2 prospective trials without control groups.[

Table 2

Results for hypertension.

Sample sizes varied from 15 to 1141 patients, and study duration varied between 2 months and 10 years. Doses ranged from 25 to 100 mg/d for SPIRO,[ with the exception of 2 studies, where doses went up to 200[ and 400 mg.[ EPLE was used at doses of 25 to 50 mg.[ Although a number of biomarkers were used to evaluate the effect of the MRAs on glycemia, only 3 studies measured the effect on HbA1c.[ The onset of diabetes was not assessed in any of the studies. The markers that were measured included: glucose, insulin, area under the curve (AUC) glucose, AUC insulin, homeostasis model assessment of insulin resistance (HOMA-IR), and quantitative insulin sensitivity check index (QUICKI).

Results for SPIRO were heterogeneous. In studies that compared SPIRO to placebo or that lacked a comparator group, SPIRO exerted a negative or slightly negative effect on some glycemic markers.[ In most studies comparing SPIRO to chlorthalidone, SPIRO had a more beneficial effect on glycemia than its comparator.[ Use of SPIRO in comparison to or in combination with hydrochlorothiazide, or in comparison to trichlormethiazide, did not yield any conclusive results.[ It is worth mentioning that thiazide diuretics are known to be associated with worsening glucose control.[ One study comparing SPIRO to perindopril or placebo did not find any significant differences between groups in terms of glucose.[ On the other hand, EPLE exerted a neutral effect in all reports.[

Obesity and metabolic syndrome

Two RCTs evaluating the effect of SPIRO were conducted on obese individuals (Table 3).[ Seven studies were done on patients with metabolic syndrome. Among these studies, 3 evaluated the effect of SPIRO.[ They consisted of 1 RCT[, 1 prospective nonrandomized trial,[ and 1 prospective trial without a control group.[ One randomized, double-blind, placebo-controlled, parallel-group study compared directly SPIRO to EPLE, as well as to a placebo group.[ Another RCT (crossover) compared EPLE to placebo.[ The last 2 studies were double-blind placebo-controlled trials on canrenone; however, the allocation was based on blood pressure characteristics rather than randomization (Table 3).[

Table 2 (Continued)

Results for hypertension.

The sample sizes ranged from 8 to 156 patients, with study duration lasting from 1 month (treatment period in crossover study) to 9 months. SPIRO was used at doses of 25 to 75 mg, EPLE doses ranged from 25 to 100 mg, and doses of canrenone were between 50 and 100 mg. A variety of biomarkers were also measured in these study populations, such as glucose, insulin, HOMA-IR, AUC glucose, AUC insulin, insulin sensitivity index, glucose effectiveness, and IV glucose tolerance. HbA1c and onset of diabetes were not measured in any of these studies.

These few small studies suggest that SPIRO does not exert a negative effect on glucose control in patients with obesity or metabolic syndrome, although their statistical power was limited. EPLE, a selective MRA, was not found to have a significant effect on glycemia. Studies with canrenone, which is more selective for the mineralocorticoid receptor than SPIRO,[ suggest that it may exert a beneficial effect in this population.

Diabetes

Multiple studies were conducted on patients with diabetes. Indeed, we identified 20 prospective studies that were performed on diabetic patients (Table 4 ).[ SPIRO was used in 16 of these studies, with 15 RCTs[ and 1 prospective study without controls.[ Three RCTs used EPLE,[ and 1 RCT was performed with canrenone.[

Table 3

Results for obesity or metabolic syndrome.

Sample sizes varied greatly between 16 and 268 patients. Study duration was between 8 weeks and 1 year, with the exception of 1 study that had a treatment period of 1 week. SPIRO was given at doses of 25 to 50 mg. EPLE doses ranged between 50 and 100 mg, and canrenone was given at a dose of 25 mg. Measured biomarkers included cortisol, glucose, insulin, HOMA-IR, HOMA-β, adiponectin, and fructosamine. In contrast with other diseases, almost all of the studies (18 studies) assessed the effect on HbA1c. This parameter increased between 0.16% and 0.6% with the use of SPIRO in studies that detected a significant association between SPIRO and changes in HbA1c.

Sixteen studies evaluated SPIRO. Among the 15 studies that measured HbA1c, 6 studies found that it significantly impaired glucose control,[ and 3 observed a nonsignificant, harmful trend with this drug.[ We must mention that 2 of these studies, conducted by the same group, had similar and overlapping populations.[ Thus, the second study[ was not included in the meta-analysis. Three studies that did not find a significant change in glucose metabolism with SPIRO used a placebo[ or no comparator.[ Another study that reported no significant change compared SPIRO with losartan,[ a drug from a pharmacological class that is known to decrease the risk of diabetes.[ The other 2 studies that did not find a significant change used hydrochlorothiazide as a comparator,[ which, as previously mentioned, is known to cause hyperglycemia.[ In the 3 EPLE studies, there was no significant impact on glycemia. Moreover, adiponectin, a protective adipocytokine, increased with EPLE in one of the studies.[ Similarly, canrenone did not influence glucose metabolism in this population.

Heart failure

We found a limited number of studies in HF. Among a total of 5 studies that were conducted on HF patients,[ 2 studies used SPIRO and consisted of 1 RCT[ and 1 retrospective cohort study.[ Two substudies of large RCTs used EPLE.[ Finally, 1 RCT compared directly the 2 drugs (Table 5).[

Table 4

Results for diabetes.

Sample sizes in this disease were quite large, ranging between 107 and 6497 patients, with the exception of 1 study that included 16 patients. The duration of the studies varied from 4 months to 2.8 years. SPIRO was used at doses of 25 mg, while EPLE was administered at doses of 25 to 50 mg. Many biomarkers were measured, including glucose, insulin, HOMA-IR, cortisol, and adiponectin. Most notably, HbA1c (in 1 study[) and the incidence of diabetes[ were measured for this disease.

SPIRO had a deleterious effect on glucose homeostasis in HF patients. HbA1c increased by 0.2%.[ Furthermore, this negative effect correlated with an increase in cortisol levels,[ suggesting that SPIRO exerts its negative effect through an increase in this hormone. On the other hand, EPLE did not have any effect on glucose homeostasis.

Polycystic ovary syndrome and idiopathic hirsutism

Fourteen studies evaluated the effect of SPIRO on patients with disorders related to hyperandrogenism (polycystic ovary syndrome [PCOS] or idiopathic hirsutism) (Table 6 ).[ Among this large number of studies, 7 were RCTs,[ 4 studies were prospective but without controls,[ 2 studies were prospective with medication assigned based on each patient's needs,[ and one study was of observational design.[

Table 4 (Continued)

Results for diabetes.

Sample sizes and duration of follow-up were somewhat limited. Almost all of the studies had sample sizes from 14 to 100 patients. Only 1 study included >100 patients (total sample of 198 patients).[ Study duration was between 2 weeks and 12 months. The doses of SPIRO ranged from 50 to 200 mg, and were usually higher than those used in other diseases. Although a number of biomarkers were collected, none of the studies measured levels of HbA1c or the incidence of diabetes. Rather, the studies evaluated the effect on glucose, insulin, HOMA-IR, fasting immunoreactive insulin (IRI), AUC insulin, AUC glucose, oral glucose tolerance test (OGTT), and insulin sensitivity indices.

Interestingly, SPIRO seemed to have a neutral or even a beneficial effect on glycemia in these patients, as opposed to other diseases where it had a tendency to exert a deleterious effect. In PCOS and hirsutism, it has been suggested that the favorable effect may be due to the decrease, and therefore the improvement, in the levels of testosterone, mediated by SPIRO.[

Hyperaldosteronism

Four articles were published on the effect of SPIRO in primary hyperaldosteronism (Table 7).[ Three of these studies were prospective, without randomization.[ Surgery or pharmacological treatment was chosen based on patients’ needs. Patients with adenomas underwent adrenalectomy, while patients with idiopathic hyperaldosteronism were treated with SPIRO. The fourth study was a noninterventional cross-sectional study in which only SPIRO was used as a treatment.[

Table 5

Results for heart failure.

Sample sizes were rather small, ranging from 9 to 47 patients, and the follow-up varied between 6 months and 5.7 years. The doses of SPIRO were between 25 and 300 mg. HbA1c was used as a biomarker in only 1 study,[ and incidence of diabetes was not assessed in any of them. The other biomarkers in these studies included insulin, C-peptide, glucose, HOMA-IR, homeostatic model assessment of β-cell function (HOMA-βF), glucose disposal rate, insulin sensitivity index, metabolic clearance rate of glucose, OGTT, fasting insulin to glucose ratio, hyperinsulinemic-euglycemic clamp, AUC insulin, and AUC glucose.

The results on hyperaldosteronism varied, as the effects were different depending on the different biomarkers. Given the limited number of investigations and patients, no definitive conclusion can be made regarding this disease.

Other conditions

Finally, 4 studies were published on other diverse patient populations (Table 8). Two studies were performed on patients with kidney disease, including 1 RCT[ and 1 sequential fixed-dose study.[ One retrospective cohort study evaluated patients with hypertension and hepatitis C.[ A fourth article presented preliminary results of an RCT on patients with nonalcoholic fatty liver disease.[

Sample sizes varied widely between 9 and 240 patients, and the duration of follow-up also differed from 8 weeks to 5.4 years. SPIRO was administered at doses of 25 to 50 mg. HbA1c was not evaluated; however, incidence of diabetes was a measured biomarker. Other markers included glucose, insulin, HOMA-IR, and QUICKI. The results were inconclusive.

In most pathologies, few RCTs (between one and 3) evaluated the effect of SPIRO specifically on HbA1c. There was a sufficient number of RCTs only in patients with diabetes (6 parallel-group trials and 6 crossover trials), where the majority of studies measured this specific marker. There were no RCTs that measured and reported the risk of diabetes. Consequently, overall, 2 meta-analyses were conducted on prospective RCTs with diabetic patients. The first quantitative analysis was performed on 6 parallel-group studies, and the second analysis included 6 crossover studies.

In the parallel-group studies (Fig. 2), the difference in mean of HbA1c between SPIRO and the comparator group was nonsignificant (mean difference 0.03 [95% CI: −0.20 to 0.26]). However, a significant difference in mean was observed in the crossover studies (mean difference 0.24 [95% CI: 0.18–0.31]; Fig. 3). There was no indication of heterogeneity in either one of the meta-analyses (I2 = 0%).

Figure 2

Meta-analysis of parallel-group studies.

Figure 3

Meta-analysis of crossover studies.

Discussion

Summary

Overall, the multiple studies conducted on SPIRO yielded heterogeneous results. These differences may be due in part to the small sample sizes in many of the studies, heterogeneous study designs and medical conditions, as well as the variability in the glycemic markers that were evaluated. However, certain trends are apparent when summarizing the impact of SPIRO in some distinct health conditions (Table 9 ), suggesting that SPIRO's effect may be disease-specific. On the other hand, our review confirms that EPLE does not have an impact on glucose homeostasis in any of the diseases that were studied. The very few investigations on canrenone suggest that it exerts a neutral or a beneficial effect.

According to our review, SPIRO may have an adverse effect in diabetes and HF. It does not seem to have a significant impact on glucose levels in the metabolic syndrome or hyperaldosteronism. On the other hand, it may either have a neutral or even a beneficial effect on glucose metabolism in diseases characterized by hyperandrogenism. Results from studies performed on healthy individuals, as well as in those on patients with hypertension, were inconclusive. These observations may also be related to the fact that HbA1c, a more sensitive biomarker of long-term glycemic control, was primarily measured in studies with HF and diabetic patients. This marker was used in very few studies on patients with other diseases. In investigations that found a negative effect on HbA1c, the average increase was, mostly, between 0.2% and 0.3%. The long-term effects of such increases in HbA1c remain largely unknown. However, such increases may have significant long-term clinical consequences as a 1% increase in HbA1c translates into a 15% increase in all-cause mortality and 25% increase in CV mortality in patients with diabetes.[ Overall, we may observe that SPIRO seems to exert a moderately negative effect on glucose regulation in patients who suffer from CV diseases or who have illnesses that increase the risk of developing heart disease, such as diabetes. On the other hand, SPIRO seems to exert a potentially favorable effect on non-CV hormonal diseases, such as PCOS.

Results from the meta-analyses with SPIRO are ambiguous, as they were nonsignificant in the parallel-group analysis, but significant in the crossover studies. The most distinguishable difference between these 2 sets of studies was the duration of treatment. The parallel-group studies had minimum treatment duration of 3 months, while the crossover studies had a maximum treatment phase of 2 months. We postulate that perhaps this contrast in the duration of follow-up may have contributed to these conflicting results. Indeed, diabetic patients that undergo a longer duration of treatment, such as the participants in the parallel-group trials, may be more likely to have their hypoglycemic agents adjusted if their glucose control worsens during the study. As such, if SPIRO did exert a significantly harmful effect on glycemia, it may have been masked by an adjustment of the patient's antidiabetic medication that is used to improve glucose metabolism. In the absence of large studies investigating the risk of diabetes, this explanation remains speculative. Additionally, there were fewer studies that used a placebo for the comparator group in the parallel-group studies. Some comparators, such as hydrochlorothiazide, are known to have a harmful impact on glycemia. However, these differences in the choice of comparator did not lead to any heterogeneity in study results. Therefore, this difference is probably not a significant limitation. Finally, SPIRO's effect on glucose homeostasis may simply be transient. Further research is required to explain these results, but the meta-analyses that were conducted confirm that if any deleterious effect exists, it would be modest.

A recent, systematic review and meta-analysis of randomized placebo-controlled trials, regarding SPIRO's glycemic effects, was conducted by Zhao et al.[ From 18 RCTs, 8 studies provided information on the change in HbA1c. We included 12 studies into our meta-analysis. The additional 4 studies in our analysis consisted of 2 RCTs with an active comparator rather than a placebo (exclusion criteria for Zhao et al),[ as well as 2 studies that could have potentially been included into their meta-analysis.[ In the meta-analysis of these studies, SPIRO was associated with a significant increase in HbA1c levels (mean difference 0.16; 95% confidence interval 0.02–0.30). SPIRO's impact on glucose, insulin, and HOMA-IR was nonsignificant. The value of HbA1c was slightly lower but in a similar range to the numeric value that we found in our crossover studies (0.24; 95% CI 0.18–0.31). However, the authors pooled parallel-group and crossover studies into a single analysis. In fact, in their meta-analysis, when crossover studies were excluded in their sensitivity analyses, the difference in HbA1c was much smaller, not statistically significant (0.05; 95% CI −0.14 to 0.25), and very similar to our results computed from pooled parallel-group studies (0.03; 95% CI −0.20 to 0.26). Additionally, when the authors pooled 3 studies on HbA1c that had a minimum duration of 3 months (all parallel-group studies), the estimate, once again, was small and nonsignificant (0.05; 95% CI −0.14 to 0.25). This observation is consistent with our own findings, where SPIRO did not have an effect on parallel-group studies with a longer duration of treatment. Zhao et al suggest that perhaps SPIRO's effect on glycemia is short-term, and does not persist on a long-term basis. This transient effect may also explain these results. The investigators also mention the possibility that SPIRO's anti-androgen effect may play a role in its impact on glucose control. The results that we obtained from the studies on patients with PCOS (not included in Zhao et al's paper) are in agreement with this hypothesis and potentially validate their assumptions. Overall, as we included more studies into our review, as well as into our meta-analysis of papers on HbA1c, our paper provides complementary and supportive information to the earlier report by Zhao et al.

Potential mechanisms of action

A number of mechanisms have been proposed to explain SPIRO's effects on glucose sensitivity. Given the positive correlation between the increase in HbA1c and the increase in cortisol, this glucocorticoid has been central to many hypotheses.[ SPIRO's off-target effect on glucocorticoid receptors could lead to a reflex increase in cortisol,[ a key player in glucose homeostasis through lipolysis and gluconeogenesis. Therefore, excess cortisol could potentially have a deleterious effect on glucose metabolism. Furthermore, cortisol has a similar affinity to the mineralocorticoid receptor as aldosterone.[ The 11β-hydroxysteroid dehydrogenase type II (11β-HSDII) enzyme regulates cortisol levels and its activity through a conversion of this steroid to its inactive form (cortisone).[ This transformation prevents cortisol from exerting additional effects and allows aldosterone to bind to its receptor. However, this enzyme is expressed at lower levels in skeletal muscle, liver, and adipose tissue.[ Consequently, these tissues may be more sensitive to high levels of this glucocorticoid.

Others have suggested that mineralocorticoid receptor blockade itself could lead to cortisol accumulation through a reduction in clearance[ or an inhibition of the negative feedback on the hypothalamo-pituatary axis.[ Another hypothesis is that the increase in HbA1c may be due to a compensatory increase in aldosterone, as the non-genomic mineralocorticoid receptors are not blocked.[ Nevertheless, such hypotheses are not consistent with the lack of impact that EPLE has on glucose homeostasis. Indeed, it would be difficult to understand why the selective antagonist, EPLE, that exerts its effect on the same mineralocorticoid receptor, would not have a negative impact on glucose control. On the whole, more research is needed to establish the exact mechanisms by which cortisol may exert these effects.

These mechanisms can be responsible for the fact that the effects differ according to different diseases. The increase in cortisol by SPIRO could have a detrimental effect on glucose tolerance in diseases that already have increased baseline levels of this hormone and are related to CV disease, such as metabolic syndrome,[ diabetes,[ hypertension,[ and HF.[ This hypothesis is supported by a high rate of diabetes in the Cushing syndrome, a disease characterized by cortisol excess.[

The off-target anti-androgen effect of SPIRO may also play a role in modulating glycemia, because testosterone levels affect glucose homeostasis.[ SPIRO's anti-androgenic effect may be either harmful or beneficial to glucose regulation, depending on the disease. In conditions that are characterized by hyperandrogenism, such as PCOS, the high baseline levels of testosterone may be linked to a risk of insulin resistance or even diabetes, and a decrease in this hormone during treatment with SPIRO may exert a beneficial effect on glucose tolerance.[ On the contrary, circulating levels of testosterone are decreased in disorders related to CV disease, such as HF[ and diabetes.[ It has been suggested that low levels of testosterone could also be associated with insulin resistance[; consequently, the decrease in this hormone, mediated by the use of SPIRO, may result in an unfavorable milieu for glucose homeostasis. Overall, the use of SPIRO could tip the scale from risk to benefit, and vice-versa, depending on the baseline testosterone levels in each disease.

In contrast to SPIRO, current knowledge suggests that EPLE's selectivity may explain its neutral effect on glycemia. Similarly, canrenone's neutral or even beneficial effect on glucose control is possibly due to its more selective nature than its parent molecule SPIRO. Indeed, it has a decreased affinity for the androgen receptor in comparison to SPIRO.[ However, it is not possible to draw any conclusions on canrenone from such a small number of studies.

Study limitations

Our review has important limitations. Regarding the limits of individual studies, many used designs prone to bias, such as retrospective or observational designs (see Table, Supplemental Content, illustrating study limits). For our review, one of the most important biases from an observational study would be confounding by indication. Indeed, the prescription of an MRA may depend on the severity of the disease. If MRA users were sicker than nonusers, the effect observed on glycemia may have been related to disease severity rather than exposure to an MRA. This bias could overestimate the potential association between MRA exposure and glucose metabolism. Second, retrospective observational studies may not always include all of the important clinical variables that could be measured in RCTs, leading to differential and non-differential bias. In addition, confounders that require detailed information on clinical parameters and lifestyle were not measured in many studies, causing residual confounding bias. Confusion bias may also exist when the variable is associated with the exposure and outcome.

Among prospective studies, certain methodological choices may have also predisposed the studies to bias. For instance, some of these studies were nonrandomized. Rather, the prescription of an MRA was based on the patient's personal needs, symptoms, or disease etiology. Such study designs could lead to a selection bias. Also, certain prospective studies were not blinded. In such cases, analyses could potentially be influenced by the knowledge of the treatment group. Furthermore, the lack of a washout period in some prospective trials may have generated a carryover effect.

Additionally, a number of articles had an incomplete description of the study design. This limited our capacity of assessing the quality of these studies. Moreover, the strength of evidence of studies was often weak because most studies had a short follow-up period, a small sample size, and/or markers that are not associated with long-term glucose metabolism (HbA1c or development of diabetes). Many studies used comparator drugs that are known to have a positive or negative effect on glycemia, leading to possible overestimation or underestimation of MRAs’ harmful glycemic effects, respectively. Nevertheless, this method did not induce heterogeneity, at least in our meta-analysis. Other studies did not have a control group. Also, some results were inconsistent within a study, as different glycemic markers had apparently opposite effects. Finally, in several articles, published results came from post-hoc analyses.

With respect to the limitations of the review process, the studies were quite different in terms of study design, study population, duration of treatment, doses, comparator medication, and types of glycemic markers. Few studies measured the effect on HbA1c in most diseases, with the exception of diabetes. This restricted the number of studies that we could include into the meta-analysis. In addition, there were a limited number of studies, and even fewer RCTs, in diseases such as metabolic syndrome, HF, and hyperaldosteronism, preventing us from drawing conclusions about the effects of MRAs in these patients. Also, as some studies were conducted by the same groups, there was some overlap between study populations.[ Moreover, the use of a single database may have slightly limited the number of selected articles. Although Medline is a comprehensive database of scientific publications, a second search engine may have provided additional relevant articles. Finally, only published articles were reviewed, leading to a potential publication bias. In general, studies that fail to reject the null hypothesis are less likely to be published. In our review, the absence of these studies may have resulted in an overestimation of MRAs’ glycemic effects. Furthermore, if effects on glucose control were not part of the primary or secondary endpoints, some authors may have failed to report the effect that was measured on glycemia in their papers, as glucose markers are routinely measured in RCTs or observational studies. This may create an outcome reporting bias. As such, it is possible that certain studies found a significant association between an MRA and glucose homeostasis, but were not published because this variable was not part of their primary endpoint and the effect on their main outcome of interest was not significant. Although less likely, this publication bias may induce an underestimation of MRAs’ glycemic effects.

Conclusion

The results of this systematic review indicate that different studies reported different effects of SPIRO on glucose homeostasis. Although these effects could be disease-specific, the inconsistencies between the studies and the limited quality of the study designs prevent us from drawing any definitive conclusions. Even within certain diseases, results were heterogeneous. Current evidence indicates that if spironolactone has any deleterious impact on glucose homeostasis, it is likely to be modest, and perhaps transient. On the other hand, EPLE, a selective MRA, does not appear to have an effect on glycemia in any of the diseases. Similarly, canrenone, a metabolite of SPIRO, seems to have a neutral or even positive effect. In the future, further investigations will be necessary to understand whether these potential pharmacological differences are clinically significant in terms of the long-term risk of diabetes or other clinically relevant outcomes.

Table 7

Results for hyperaldosteronism.

Table 8

Results for other conditions.

Table 9

Summary of results.

Table 9 (Continued)

Summary of results.