Major depressive disorder (MDD) is a common and costly mental disorder that is characterized by pervasive low mood and various other symptoms, such as cognitive and physical symptoms. Its probability of recurrence is high, with an average of four episodes during a patient’s life (Limosin et al., 2007). Nearly 30% of individuals with MDD always have symptoms present and develop a chronic condition that is reliant on the illness stage and other risk factors such as childhood trauma and personality (Angst et al., 2009; Boschloo et al., 2014). It substantially affects an individual’s psychosocial functioning and exhausts quality of life (Malhi and Mann, 2018). The Global Burden of disease Study 2019 states that (Diseases and Injuries, 2020) depressive disorders are among the top 10 causes of disability-adjusted life-years (DALYs) for the 10–49-years age group and are among the top three causes of DALYs for women. Barely any breakthrough in the optimization of MDD diagnosis and the improvement of MDD treatment outcomes have been made over the past several decades despite the considerable efforts exerted worldwide. Some novel antidepressants targeting the N-methyl-D-aspartate receptor or gamma-aminobutyric acid are still in the early stages for development or approval. The Sequenced Treatment Alternatives to Relieve Depression study (www.star-d.org) has shown that approximately half of patients with nonpsychotic MDD respond to level 1 antidepressant treatment (citalopram), and only nearly 30% patients achieve remission. Theoretically, the cumulative remission rate of four sequential treatments is 67%, and the likelihood of achieving remission is high in the first two medication trials and then decreases. Thus, probing into the pathophysiology of MDD and customizing effective therapeutic strategies for every individual suffering from MDD are essential.

Endocrine system abnormalities, including abnormalities of the adrenal, gonadal, and thyroid axes, have been observed in depression for many decades, and the alterations in mood and cognition after treatment with endocrine function-targeting medications further suggest that hormones play an important role in the pathophysiology of MDD (Dwyer et al., 2020). The hypothalamic-pituitary-adrenal (HPA) axis is a crucial neuroendocrine system that controls stress reactions and orchestrates emotions and many other bodily processes (Dwyer et al., 2020). In particular, arginine-vasopressin (AVP), in addition to corticotropin-releasing factor (CRF), enhances the release of adrenocorticotropic hormones (ACTH) and participates in acute stress response (Rene et al., 2000; Lolait et al., 2007). The up-regulation of its receptor may contribute to maintaining corticotrophic responsiveness to chronic stress or depression (Rotondo et al., 2016). Early studies have found that cortisol concentrations in plasma and cerebrospinal fluid were elevated in MDD (Holsboer and Barden, 1996; Nemeroff, 1996) and that the frequency of the failure to respond to the dexamethasone suppression test was increased in patients with depression, despite the low sensitivity (about 44%) that limited its use as a diagnostic tool (Arana et al., 1985). Subsequent studies further observed blunted cortisol circadian rhythms (Stetler and Miller, 2011), excessive HPA axis activity (Amsterdam et al., 1987; Ising et al., 2007; Menke et al., 2016), and impaired negative feedback in MDD (Arborelius et al., 1999; Dwyer et al., 2020). Keller and colleagues found that high plasma cortisol was associated with worsened cognitive performance in patients with MDD and healthy controls and that patients with psychosis had higher cortisol level than healthy subjects and depressed patients without psychosis (Keller et al., 2017). Interestingly, patients with bipolar disorder (some of whom were in depressive episodes) had increased cortisol response to the combined dexamethasone/corticotrophin-releasing hormone test (Watson et al., 2004). Notably, HPA axis dysregulation and normalization failure after treatment are associated with poor clinical prognosis, including low response rates to antidepressants and high relapse and chronicity (Nelson and Davis, 1997; Young et al., 2004; Vreeburg et al., 2013).

On the basis of the abovementioned insights, some researchers have attempted to explore medications that modify HPA axis function for depression treatment. These medications include glucocorticoid (GR)/mineralocorticoid receptor (MR) antagonists, vasopressin receptor antagonists, and steroidogenesis inhibitors. However, the results of these clinical trials were mixed. For example, Jahn and his team (Jahn et al., 2004) found that metyrapone (a cortisol synthesis inhibitor) was effective as an adjunctive treatment for MDD, whereas another study (McAllister-Williams et al., 2016) had negative results. Similarly, evidence for the glucocorticoid receptor antagonist mifepristone was ambiguous (Watson et al., 2012; Block et al., 2018). Some medications had initial promising results but were then discontinued (Serradeil-Le Gal et al., 2005; Sanofi-Aventis, 2008), and several compounds had been abandoned due to their severe side effects or lack of efficacy (Binneman et al., 2008; Zorrilla and Koob, 2010; Dwyer et al., 2020). A previous meta-analysis that explored antiglucocorticoid and related treatments for psychosis assessed depression symptoms as secondary outcomes and found limited evidence (Garner et al., 2016). None of these strategies have been successfully translated into clinical use, and the efficacy and side effects of these strategies for the treatment of MDD remain uncertain. We therefore systemically searched available studies and conducted a meta-analysis to determine the effects and safety of HPA axis-based medications for MDD.

By following the guidance of Preferred Reporting Items for Systematic Reviews and Meta-analysis, we prepared a study protocol with objectives, search strategy, participants, study type, outcome measurements, and data synthesis strategy for study organization and reporting (PROSPERO registration number: CRD42021247279).

As shown in Supplementary Figure S1 (see Appendices), 1907 studies were included for title and abstract screening after 551 duplications were excluded. Subsequently, 38 studies were fully reviewed. A total of 23 publications that included 2,972 subjects were finally included in our analysis.

This meta-analysis of 16 RCTs and seven open-label studies that included 2,972 subjects examined the efficacy and safety of HPA-targeting treatments for depression as a monotherapy or as an add-on treatment. The following medications were explored: metyrapone, mifepristone, ketoconazole, V_1B receptor antagonists (TS-121, SSR149415, and ABT-436), CFH receptor antagonists (CP-316,311, GSK561679, and R121919), dexamethasone, spironolactone, and fludrocortisone. Our results indicated the favorable but small effect of these medications as a whole in treating individuals with depression or MDD. Effectiveness was remained significant during the 2 weeks follow-up period. Sponsorship might affect the results to a small extent. Evidence for ketoconazole, dexamethasone and fludrocortisone was scant, whereas the efficacy of mifepristone and V_1B receptor antagonists was more convincing than that of other medications. In particular, the V_1B receptor antagonist revealed a medium effect size. Our findings yielded insufficient evidence to support the efficacy of HPA-modulating medications for cognitive impairment in patients with depression. Although the action of these medications as the sole treatment for patients with depression was promising, additional studies were needed to verify their efficacy as augmenting agents. Safety analysis favored placebos. The rate of gastrointestinal side effects and dizziness were higher in subjects using antiglucocorticoids and related medications than in subjects using placebos.

Mifepristone, a GR antagonist, has been one of the most tested HPA-targeting medications for depression or psychosis. A meta-analysis reported by Garner and her team concluded the possible efficacy of mifepristone for psychosis (Garner et al., 2016). In our analysis, almost all of the included mifepristone studies were performed with patients with psychotic depression. Previous studies suggested that hyperactivity of the HPA axis may lead to dopaminergic and noradrenergic dysregulation and dysfunction. For example, researchers have found that dexamethasone, adrenocorticotropin, and CRH boost dopamine metabolites (Rothschild et al., 1984; Posener et al., 1994; Posener et al., 1999). Duval and colleagues (Duval et al., 2006) found lower cortisol and growth hormone response to dopamine receptor agonists and alpha 2-adrenoreceptor agonists, respectively, in patients with psychotic major depression than in healthy controls and nonpsychotic patients. The function of somatodendritic and postsynaptic 5-hydroxytryptamine 1 A (5-HT_1A) receptors is also regulated by corticosteroids and decreases in response to chronic stress (Grino et al., 1987; Guillaume et al., 1987; Laaris et al., 1997; Fairchild et al., 2003; McAllister-Williams et al., 2007). This situation may be the underlying mechanism of the efficacy of mifepristone used alone or with selective serotonin reuptake inhibitors for psychotic depression. However, the whole picture of this mechanism is complex and remains unclear. Interestingly, some researchers have hypothesized that the effectiveness of mifepristone may be associated with mifepristone plasma levels. One of the included studies (Block et al., 2018) examined this hypothesis and found a greater response rate in patients with high mifepristone plasma levels than in those without. Thus, the lack of significant difference for depression between mifepristone and placebos reported by some studies may be attributed to subtherapeutic mifepristone plasma levels. However, predicting plasma levels in accordance with dosage is difficult due to the nonlinear kinetics of mifepristone when administered at doses exceeding 50 mg (Sitruk-Ware and Spitz, 2003). Further studies may benefit from examining not only the level of mifepristone but also the levels of its active metabolites which increase reliably with dosage (Heikinheimo et al., 1987; Sitruk-Ware and Spitz, 2003). Some researchers have suggested that the efficacy of antiglucocorticoid treatment may be associated with baseline cortisol levels. Previous studies indicated that HPA disturbances, including abnormal basal and post-dexamethasone cortisol levels, are more likely to manifest in patients with psychotic depression than in non-psychotic depressed subjects (Contreras et al., 2007). Lombardo and colleagues conducted a meta-analysis (Lombardo et al., 2019) to investigate this hypothesis and found that patients who responded to cortisol synthesis inhibitors rather than GR antagonists had higher baseline cortisol levels than non-responders. The different pharmacological actions of these medications may be a confounding factor. The efficacy of mifepristone may also be associated with inflammation levels instead of with baseline cortisol levels given the evidence provided by a previous animal study by Zhang and colleagues (Zhang et al., 2018). Other researchers focused on GR antagonism and hypothesized that the cerebrospinal fluid levels of cortisol and mifepristone may affect the therapeutic effects of these drugs (Golier and Yehuda, 2018). Studies have suggested that mifepristone has a potent p-glycoprotein pump antagonist role that can control the transport of cortisol across the blood-brain barrier (Schatzberg and Lindley, 2008). These hypotheses await testing in the future.

Cortisol synthesis inhibitors, including metyrapone and ketoconazole, were also broadly examined in patients with MDD. In general, the quality of most studies examining these two medications was relatively low, and our results showed no benefit from treatment. Similarly, our results did not reveal the beneficial effect of either CRH receptor antagonists or MR-modulating treatments due to scarce evidence. CRH system hyperactivation had been reported in patients with MDD, and the abnormality was restored after antidepressant treatment (Nemeroff et al., 1991). Some clinical trials had explored related compounds for depression treatment but few are ongoing due to side effects or lack of efficacy (Dwyer et al., 2020). New compounds with different targets of the CRH system and different pharmacokinetic profiles may be useful (Dwyer et al., 2020). MR is another essential nuclear receptor that binds to cortisol with high affinity. It is mainly distributed in the prefrontal-limbic circuit (Dwyer et al., 2020). MR dysfunction was also found in depressed patients with equivocal results. Increases and decreases in MR function have been reported (Lopez et al., 1998; Lopez et al., 2003; Young et al., 2003; Wang et al., 2008). Thus, Otte and colleagues (Otte et al., 2010) examined the efficacy of the use of a MR antagonist (spironolactone) and a MR agonist (fludrocortisone) as an adjunct to escitalopram in patients with MDD. Although they found no significant difference in altered HAMD scores between interventions and controls, they observed that responders in the fludrocortisone group responded earlier than the controls and patients in the spironolactone group. Response time is an important index and should be assessed in future studies. Early response is linked to reduced mental health care costs and personal suffering. The fludrocortisone, according to (Otte et al., 2010), may play an accelerator role in depression, instead improving overall psychopathology. Existing evidence indicated the potential antidepressant effect of MR-modulation. For example, MR upregulation was one of the earliest response to antidepressants in several animal studies (Barden et al., 1995; Yau et al., 2002), which was also observed in studies on GR antagonists (Bachmann et al., 2003). The continued response (at least 2 weeks) to GR antagonists and cortisol synthesis inhibitors seen in our results may be partly mediated by MR up regulation and MR/GR balance resetting (Schatzberg and Lindley, 2008). In addition, MR stimulation led to increased 5-HT_1A receptor expression in mice (Rozeboom et al., 2007). While we cannot deny the possibility that some observed effects of fludrocortisone are exerted by GR stimulation, despite that its affinity for GR is much lower than for MR. On the contrary, evidence from MR antagonist spironolactone weas more controversial. Previously, (Holsboer, 1999), demonstrated adverse effects of spironolactone for depression treatment, and similarly, (Otte et al., 2010), found no improvement for depressed patients treated with spironolactone as an add-on strategy. However, other studies, for example studies focusing on patients with premenstrual syndrome (O'Brien et al., 1979; Wang et al., 1995) and euthymic patients with bipolar disorder (Juruena et al., 2009), have reported beneficial results of spironolactone on mood or residual symptoms improvement. Thus, MR-targeting medications are promising approaches in the future.

Three studies examining V_1B receptor antagonists were graded as high quality, and our results showed a medium effect size for significant improvement in depression symptoms. Early animal studies had found that vasopressin-containing neurons were distributed in limbic areas and that their messenger RNA expression was increased by chronic stress (Rabadan-Diehl et al., 1995; Hernando et al., 2001; Stemmelin et al., 2005). AVP was elevated in approximately 25% of patients with MDD (van Londen et al., 1997) and normalized after antidepressant treatment (De Bellis et al., 1993). Although the preliminary results were positive, the potential antidepressant efficacy of the V_1B receptor antagonist needs further repetition.

Overall, a higher rate of adverse events was observed in patients treated with HPA-targeting medications than in controls, and the quality of evidence was relatively high. No serious adverse events were reported, and few patients dropped out due to side effects. Thus, these medications were well-tolerated. Notably, given the homology between glucocorticoid receptors and progesterone receptors and the potential risk of inducing abortion and amenorrhea in females, current GR antagonists may benefit from the further development of receptor selection (Schatzberg and Lindley, 2008).

It is important to mention that glucocorticoid secretion is characterized by a complex, circadian and ultradian pattern which is under the influence of genes, age, gender and environments (Kalafatakis et al., 2016). This pulsatility has substantial, multi-level effects not only on the peripheral tissues, but also on the central nervous system. Recently, many researchers have found significant neurobiological effects of glucocorticoid ultradian rhythm on human, including cognitive, emotional and behavioral processes (Kalafatakis et al., 2018; Kalafatakis et al., 2019; Kalafatakis et al., 2021). Different cortisol replacement regimens resulted in varying outcomes, and subjects with regimen that failed to mimic the ultradian profile had lower health-related quality of life (Bleicken et al., 2010; Tiemensma et al., 2014; De Bucy et al., 2017). It seems like few studies in this meta-analysis have considered this issue, and most of these studies had simplified drug administration strategies. No study to date has elucidated the effects of different treatment regimens on depression, and it remains unclear whether treatment outcomes improve if HPA-modulating drugs are administered according to cortisol ultradian rhythmicity. Given the complexity and variability of glucocorticoid pulsatility and its neurobiological significance, it is a great challenge for future studies to optimize the application and therapeutic evaluation of HPA axis-modulating therapies for MDD. In addition, it is also noteworthy that even though many preclinical and clinical studies suggest a strong association of HPA-axis in stress induced mental disorders like depression and posttraumatic stress disorder. There is also evidence that not all depressed patients do display alterations of the HPA axis. A HPA-axis dysregulation biomarker is not employed clinically in routine settings. Therefore, it is unclear whether such treatments can benefit all patients or not. This study suggests that patients with psychotic depression and treatment-resistant MDD may be benefit from HPA axis examination and HPA axis-modulating therapy. However, more clinical trials are warranted to this issue.

One main limitation of this research is the variable quality of the included studies, which may bias our results. Many of them have “unclear risk of bias” items due to the lack of necessary information, such as randomization, allocation, or blinding procedures. Additional well-designed RCTs with prolonged follow-up periods investigating HPA-modulating treatments in different depression subtypes are needed. In addition to the change in depression rating scores, other outcome measures, such as response rate, remission rate, time to respond, and functional improvement along with cognitive function must be investigated. These measures are more understandable and feasible for clinical use than other measures. Simultaneously, the measurement of the plasma level of medications, peripheral and central neuroendocrine indexes, and inflammation factors may be useful for establishing optimal doses and understanding underlying mechanisms. Given the remarkable sex difference in the HPA system (Kokras et al., 2019), patient stratification may be useful for diminishing confounding factors. In addition, head-to-head trials are also important to determine whether these treatments are a better choice for patients with MDD than standard treatments.

In conclusion, this is the first meta-analysis examining the efficacy and safety of HPA-targeting medications in patients with MDD. Our results suggest the favorable but small effect of these medications as a whole for MDD. Specifically, evidence for mifepristone and V_1B receptor antagonists are more convincing than that for other medications. Although adverse effect analysis favors placebos, these drugs are generally well-tolerated. HPA-based medications are a promising field for depression treatment, but additional high-quality RCTs, including head-to-head trials, are needed to verify findings.