Pineal Abnormalities in Psychosis and Mood Disorders: A Systematic Review

The pineal gland (PG) is a small interhemispheric brain structure that influences human physiology in many ways, most importantly via secretion of the hormone melatonin which is known to regulate sleep and wakefulness. Here, we systematically reviewed existing neuroimaging studies of PG structure, and/or melatonin release (MLT) in psychosis and mood disorders. Medline, PubMed, and Web of Science databases were searched (on 3 February 2023), yielding 36 studies (8 PG volume, 24 MLT). The findings showed smaller-than-normal PG volume in people with schizophrenia, regardless of symptom severity and illness stage; and smaller-than-normal PG volume in major depression, with some indication of this being present only in certain subgroups, or in those with high scores on the ‘loss of interest’ symptom. There was considerable evidence of lower-than-normal MLT as well as aberrant MLT secretion pattern in schizophrenia. A similar picture, though less consistent than that seen in schizophrenia, emerged in major depression and bipolar disorder, with some evidence of a transient lowering of MLT following the initiation of certain antidepressants in drug-withdrawn patients. Overall, PG and MLT aberrations appear to represent transdiagnostic biomarkers for psychosis and mood disorders, but further work is needed to establish their clinical correlates and treatment implications.


Introduction
For centuries, the pineal gland (PG) has been referred to as the 'third eye' or 'ajna chakra', 'pineal eye', and 'seat of the soul', and it was only in the late 19th century that we gained a clear understanding of its structure and various influences on the mammalian physiology, especially its role in the regulation of sleep cycle and wakefulness [1][2][3][4][5]. In humans, the PG is a small interhemispheric brain structure, resting proximally on the posterior aspect of the diencephalon. It is located within 1-2 mm of the midline and becomes visible between habenular and posterior commissures around 7 weeks of gestation [6]. Average PG dimensions in human adults are 5-9 mm in length, 1-5 mm in width, and 3-5 mm in height, and it weighs roughly between 100 and 180 mg depending on the age and sex [7,8]. The main function of the PG is to receive and transmit the light-dark signals from our surroundings and, accordingly, to produce and secrete the hormone melatonin [9]. Pineal melatonin (MLT) is known to modulate circadian rhythms and is involved in sleep regulation [10], reproductive physiology [8,11,12], and immunological regulation [13].
In recent decades, there has been much interest in examining PG and MLT production in people with psychosis and mood disorders, given that poor regulation and/or quality of sleep feature prominently in both of these disorders [14,15]. There are reports of smaller PG volumes, compared to healthy people, both in people with psychosis [16,17] and mood disorders [18] though not all studies have found this (e.g., [19]). It is possible that PG and MLT aberrations represent transdiagnostic biomarkers across these disorders or, 3 two reviewers were compared, and the degree of agreement was assessed. Any di ancies, as well as any studies with a score of 1, were discussed with a third reviewe to reach a consensus. The reasons for excluding studies at all stages were documente a flowchart of the study selection process, see Figure 1).

Data Collection Process
Data extraction for all selected studies was conducted by the first author (SC). dition, data for a random selection of 18 studies were independently extracted by th ond author (AB) to verify the extraction process.

Data Items and Analysis
For each of the selected study, the following data were extracted: author and year, study population and sample characteristics (sample size, mean age and sex bution for the patient and healthy comparison groups; and, in addition, for patient gr

Data Collection Process
Data extraction for all selected studies was conducted by the first author (SC). In addition, data for a random selection of 18 studies were independently extracted by the second author (AB) to verify the extraction process.

Data Items and Analysis
For each of the selected study, the following data were extracted: author and study year, study population and sample characteristics (sample size, mean age and sex distribution for the patient and healthy comparison groups; and, in addition, for patient groups, diagnosis, age at illness onset, duration of illness, symptoms, current medication, and treatment history), design of the study, imaging modality or methods for assessing PG structure and/or function, intervention characteristics (where relevant), key study outcomes (i.e., diagnosis effects in PG structure and/or function; any association with patient characteristics), and funding sources. For all key outcome variables, group averages (mean, SD) and the size of the correlations (where reported between PG volume/MLT and relevant sample characteristics (e.g., symptoms or medication dose) were also extracted. All extracted data were compiled into a Microsoft Excel spreadsheet and analysed descriptively, considering the statistical significance of the findings as reported by the study authors for group comparisons (patients vs. healthy controls; or involving different patient groups) and/or correlations between the PG structure/function and patient characteristics. We made no specific assumptions where the study authors had assessed relevant sample characteristics (for example, symptoms) but not examined them in relation to the PG structure or function; we simply noted them as 'not reported' in the extracted data.

Quality Appraisal
The quality of selected studies was assessed using Newcastle-Ottawa Quality Assessment Scale [26] for non-randomised, cross-sectional studies, and Critical Appraisal Skills Programme [27] scores for randomised controlled trials (see Supplementary Tables S1 and S2 for quality rating of the selected studies).

Results
Overall, the search yielded 36 studies conducted in 16 different countries (UK, Germany, France, Sweden, Italy, Belgium, Poland, Canada, Brazil, USA, Australia, Turkey, Iran, China, Japan, and Taiwan). There was significant variability in sample sizes ranging between 5 and 87 participants for the healthy group, between 7 and 162 participants for the psychosis group (excluding bipolar disorder as this patient group was included mostly within the context of mood disorders, as bipolar versus unipolar depression), and between 6 and 50 participants for the mood disorders group (major depression or bipolar disorders). Eight of 36 reviewed studies investigated PG structure using MRI (4 studies in psychosis or schizophrenia, 3 studies in mood disorders, and 1 study involved schizophrenia as well as mood disorders), and 24 studies examined MLT in blood, urine and/or cerebrospinal fluid (8 in psychosis, 15 in mood disorders, 1 study involved both schizophrenia and mood disorders). Most of these 32 studies used a cross-sectional study design comparing the patient (psychosis and/or mood disorders) and healthy control groups, and some studies also examined possible associations between PG structure and/or MLT and symptom ratings or medication within the patient samples. Lastly, four studies examined medication effects on MLT in major depression and also included comparison with healthy controls. No study met our eligibility criteria for examining antipsychotic effects on MLT.

PG Volumes: Effects of Diagnosis and Possible Association with Clinical Characteristics
The details and key findings extracted from the reviewed studies in relation to PG volumes in psychosis and mood disorders are presented in Table 1.
PG volume was found to be significantly smaller in people with schizophrenia or psychosis, compared to healthy controls, in four of the five studies that investigated this [16,17,28,29]. An early study with a modest sample size [19] reported no significant difference between the patient and control groups although the mean PG volume was still numerically lower in the patient group (see Table 1a,b). In the two most recent studies with relatively large sample sizes [17,29], PG volume was found to be smaller-than-normal not only in the first-episode and chronic schizophrenia groups, but also in those at a high risk of developing psychosis. Furthermore, no relationship between the PG volume and age or any clinical characteristics of the patient sample (symptoms, age of illness onset, duration of illness and/or treatment) was detected in any of the four studies that found significantly smaller-than-normal PG volume in psychosis or schizophrenia samples [16,17,28,29].
PG volume was also found to be significantly smaller in people with major depressive disorders, compared to healthy controls, in two of the three studies that examined this [16,18]; however, there was no association between PG volumes and depressive symptom ratings in either of these studies (see Table 1b,c). One study [30] that did not observe a significant difference between the PG volumes of currently depressed or remitted groups of mood disorder patients and healthy controls, however, reported significantly smaller PG volumes in non-melancholic patients, compared to melancholic depressed patients; it also reported a negative association between PG volumes, and the 'loss of interest' symptom ratings but no association with any other symptoms (for example, the Beck Depression Inventory-II scores) when examined across the patient sample. Furthermore, this study [30] showed no significant difference between the PG volumes of bipolar disorder patients and healthy controls, and no significant association between total PG volume and any clinical characteristics, namely, the number of episodes, duration of illness or the medication dose, in this group of patients. A further study of bipolar disorder patients [31] also found them to not differ from healthy controls in PG volumes, although a group of BD patients were found to have smaller PG volumes than the controls in a later study [16].
There was only one study [16] directly comparing PG volumes in schizophrenia and mood disorders, and it showed (i) significantly lower PG volume in the schizophrenia group, compared to patient groups with unipolar depression or bipolar disorder, (ii) no significant difference between unipolar depression and bipolar disorder groups; (iii) and significantly lower PG volumes, compared to controls, in both unipolar depression and bipolar disorder groups (as already noted, both unipolar depression and bipolar disorder groups had larger volumes compared to those in people with schizophrenia).

PG Function (MLT): Effects of Diagnosis and Association with Clinical Characteristics
There were eight studies examining MLT level or MLT secretion patterns in people with schizophrenia (see Table 2). All these studies, except one [44], observed (overall) lowerthan-normal MLT level [45][46][47][48] or an aberrant pattern of MLT secretion in this group of patients [48][49][50]. Specifically, there was a different pattern in patients and controls, with MLT increasing at 8 p.m. and peaking at 2 a.m. in controls but no significant peaks in patients [49]; earlier peaks in drug-free patients, relative to controls [49]; and delayed rhythm and atypical sleep-wake cycles in 50% of the patients, relative to controls [50]. In one study [48] that compared acutely unwell and chronic patient groups, acutely unwell patients had higher MLT, relative to chronic patients, but both groups still had lower MLT relative to healthy controls.
There was no relationship between MLT aberrations and symptoms in most of the studies that examined this in people with schizophrenia [45,47,49], and in one study [50] that did observe lower MLT in patients with positive symptoms, relative to those without positive symptoms, this effect was explained by age ( Table 2). Antipsychotic medication seemed to be associated with a change towards normalisation of the MLT secretion pattern in that medicated patients did not differ significantly from controls [51]. In the only study [44] that did not find lower-than-normal MLT levels in schizophrenia patients, MLT levels were found to be positively associated with antipsychotic medication dose as well greater symptom severity, possibly reflecting higher antipsychotic doses prescribed to those with greater symptom severity.
There were 15 studies examining MLT level or MLT secretion in people with major depression and/or bipolar disorders ( Table 3). Most of these studies showed significantly lower MLT, especially nocturnal MLT, in depressed patients, compared to healthy controls [52][53][54][55][56][57][58], or atypical MLT secretion patterns such as delayed peak [59] or delayed onset [60]. Relatively fewer studies showed a non-significantly lower MLT in patients [61], no significant difference between the patient and control groups [62,63], or lower MLT level in patients with psychotic depression, but higher in those with non-psychotic depression, both compared to controls [64]. One of these studies [56] had also examined and observed lower MLT in panic disorder patients, compared to healthy controls. Interestingly, in one study [65] that had assessed MLT in both blood and cerebrospinal fluid (CSF), blood but not CSF MLT was lower in depressed patients, and CSF but not blood MLT was lower in bipolar disorder patients, both relative to healthy controls. However, an earlier study [66] had shown lower serum MLT in bipolar disorder patients (at 1 a.m. and 5 a.m.; but only at 1 a.m. in euthymic subgroup) relative to the healthy group.
Most studies in mood disorders had only small-to-modest and mixed sex samples (with marked inter-study variation in average depression severity, see symptom ratings for individual studies, Table 3), and did not examine relationship with symptom profiles or medication. A few studies which did examine such associations presented a mixed picture, with no relationships between basal MLT or secretion patterns and medication [52] or symptom severity [55], or a negative relationship with depression severity [53]. Some studies also showed MLT to correlate positively with body mass index (BMI), and to be influenced by sex of the participant while others did not (Table 3). Lastly, there was only one study [67] comparing MLT in people with schizophrenia (n = 12) with that in people with depression (n = 60] and, unexpectedly, its findings revealed higher MLT in schizophrenia, relative to depressed patients (Table 4).
There were four studies (all involving 20 or fewer patients) investigating the influence of antidepressant medications on MLT in depressed patients ( Table 5). Two of these studies, with only or predominantly females, showed a transient lowering of MLT following certain antidepressants [desipramine [68]; fluoxetine [69]], while the remaining two mixed-sex studies showed no pre-to post-change in MLT [clomipramine [70]; fluoxetine [71]]. Further complicating the picture regarding MLT patterns in depressed patients, three of these four studies [68,69,71] had reported no significant differences in MLT between the patient and control groups at baseline, while one study [70] reported higher MLT in patients, relative to healthy controls (see Table 5).       Interestingly, no significant difference between the patient and control groups in daytime or nocturnal urinary MLT. Two weeks of drug-naïve or drug-free prior to the study.
MLT (between 6 p.m. and 7 a.m.) measured via blood Between-groups Significantly higher nocturnal MLT levels in patients (0.18 ± 0.14 nmol/L) compared to controls (0.15 ± 0.10 nmol/L). When the MDD group was split into non-psychotic (n = 15) and psychotic (n = 7) sub-groups, lower MLT in the psychotic subgroup (0.11 ± 0.11 nmol/L), but higher MLT in the non-psychotic subgroup (0.19 ± 0.15 nmol/L) relative to controls. Two weeks of drug-naïve or drug-free prior to the study.
MLT positively correlated with height (r = 0.09) and weight (r = 0.17) but relationship with symptom ratings, and other variables not reported.   MLT (between 8 a.m. and 10 a.m.) measured via CSF and blood.

Discussion
This systematic review evaluated currently existing evidence to identify possible aberrations in pineal gland (PG) volume and melatonin production in people with psychosis or mood disorders and their possible associations with patient characteristics, especially symptom severity.

PG Volume in Psychosis and Mood Disorders
Our review of existing findings in psychosis and mood disorders yielded consistent evidence of smaller PG volumes in first-episode and chronic schizophrenia patients as well as in people at a high risk of developing psychosis, compared to healthy controls, with no significant influence of symptom severity or medication status [16,17,28,29], and also some evidence of smaller PG volumes in schizophrenia patients, relative to mood disorder patients [16]. The same direction of effects (i.e., smaller PG volume in patients than controls) was present in people with bipolar disorders [16] but not significantly so in all studies [30,31]. Smaller-than-normal PG volume was also present in people with major depression, independent of symptom severity [16,18], although there was also evidence of this being applicable to only certain sub-types, or to those experiencing 'loss of interest' as a symptom [30]. Taken together, these findings suggest that smaller-than-normal PG volume may represent a transdiagnostic biomarker for schizophrenia and mood disorders, deserving of further study in relation to sex differences, functional and clinical outcomes, including treatment responsiveness.

MLT Production and Secretion Patterns in Psychosis and Mood Disorders
The majority of the existing studies conducted in schizophrenia observed lower-thannormal MLT in acutely unwell as well in chronic patient groups [45][46][47][48] or an atypical pattern of MLT secretion in this clinical population [49][50][51]. The studies, however, also showed some differences between acutely ill and chronic patients [48] and there was cross-sectional evidence of possible normalisation of MLT secretion pattern in medicated patients [51], suggesting that MLT production might be influenced by both illness and medication-related factors. As yet, there are no published longitudinal investigations of such possible influences in MLT production of schizophrenia and related populations.
Our review of studies in people with major depression revealed considerable evidence of lower-than-normal MLT, especially nocturnal MLT [49,[52][53][54][55][56][57], or atypical MLT secretion patterns, including delayed peak [59] or delayed MLT onset [60]. However, this was not found in some studies [62,63] and appeared in others to be dependent on presence of psychotic symptoms [64] or methods for assessing MLT [66]. Some data showed an association between lower MLT and greater depression severity [53]. The findings in bipolar disorder patients were also mixed, again suggesting that MLT in major depression and bipolar disorders may be sensitive to multiple influences. Lastly, there was some evidence of a transient lowering of MLT following the initiation of certain antidepressants in drugfree patients [desipramine [68]; fluoxetine [69]] suggesting that medication may have been a confounding factor and contributed to the mixed pattern of MLT findings in mood disorders. Interestingly, all four studies that investigated medication effects on MLT showed either no MLT difference between the patient and control groups (3 studies), or higher MLT in depressed patients relative to controls (one study), at baseline. This is hard to explain and may relate to some non-specific patient selection bias in such studies (e.g., selection of only those patients who were considered safe enough to withdraw medication for some time in order to meet study eligibility criteria).)

Limitations of the Reviewed Evidence and the Review Processes
The findings of this systematic review should be considered taking a number of limitations into account. First, most of the reviewed studies had small-to-moderate sample sizes (with n< 50 per group), providing limited power to compare PG structure and function between the patient and healthy control groups and to investigate sex differences in diagnosis-related effects; therefore, they are likely to have been underpowered to meaningfully examine any associations between the PG structure/function and patient characteristics of potential relevance, especially symptom profiles. Second, a number of studies either did not provide sufficient information on patient characteristics or did not examine them in relation to the PG structure or function, which would have allowed greater insight into this topic and a clearer interpretation of negative results (for example, 31). Third, the methods for assessing MLT varied greatly between the reviewed studies with MLT assessed via different routes (in blood, urine, CSF) and the samples collected at different times (e.g., one-off, in the morning or evening, over 8-24 h). Fourth, there were only four studies examining antidepressant effects on MLT in mood disorders, and all of these were of fair quality at best (see Supplementary Table S2); and there were no studies assessing antipsychotic effects on the PG structure or function in people with psychosis. Making solid conclusions about diagnosis or symptom related differences without a clear understanding of medication influences in the PG structure or function is difficult. Lastly, we deliberately focused on psychosis and mood disorders but it is possible that the PG structure and function are also affected in other disorders, and did not conduct a meta-analysis to formally assess the effect of the moderator variables (due to the limited number of studies with required power and information available for review), restricting confidence in some of our findings until they are examined and supported by future research.

Conclusions and Future Directions
Our review provided consistent evidence that smaller-than-normal PG volume represents a transdiagnostic biomarker across psychosis and mood disorders, and possibly other disorders linked with sleep dysfunction. We also found lower-than-normal MLT as well as aberrant MLT secretion pattern (flattened or shifted rhythm) in psychosis and mood disorder patients, relative to healthy controls, but the findings on these measures appeared fragile, possibly due to various influencing factors, such as severity of certain symptoms, specific symptom profiles and medications. With these findings, we make a number of recommendations for the future scientific enquiry in this area.
First, we suggest that future clinical studies should employ a transdiagnostic approach involving multiple patient groups and assessment on multiple symptom dimensions, including the quality of sleep and sleep disturbances. A similar approach can be taken to clarify the association between specific symptom dimensions and PG volume and function in non-clinical samples, especially to rule out medication-related confounders and potentially identify underlying latent factors.
Second, we emphasise that future studies should have large enough sample sizes to detect diagnosis related effects of small-to-medium size. The studies should also aim to include sufficient number of males and females to allow meaningful investigation of sex differences, given their influence in many brain structures and function [81] as well as in prevalence rate, prognosis, symptom profile and presentation, and treatment response in many disorders, including schizophrenia [82,83] and affective disorders [84].
Third, we recommend future studies to consider the chronotype of study participants as well as take any regional and seasonal variations into account for both the control and patient groups, given their relevance for mental disorders as well as MLT production [85].
Fourth, we highlight the need for multimodal investigations involving assessment of both PG structure and function in the same samples to clearly understand how PG volume aberrations, with or without covarying for whole-brain volume, may influence MLT production in the short and long-term in interaction with age, BMI, sex, and possible illness-or symptom-related influences in clinical and non-clinical samples.
Finally, we encourage future studies to examine how currently used antipsychotics for treating psychosis might influence the PG volume and MLT in parallel to changes in sleep parameters, and also consider the use of melatonin as potential treatment avenues for reducing side effects of antipsychotics [86][87][88], ameliorating various cognitive and information processing disturbances typically present in people with schizophrenia [89,90], and treating sleep disturbances in a range of psychiatric disorders, including but not limited to psychosis and mood disorder [91,92].