Introduction
Over 60% of people with bipolar disorder (BD) are overweight or obese (Goldstein et al., Reference Goldstein, Liu, Zivkovic, Schaffer, Chien and Blanco2011). Obese patients have a less favorable illness course than normal-weight patients, including more frequent and severe depressions, poorer responses to pharmacologic and psychotherapeutic interventions, and greater inter-episode cognitive impairment (Depp et al., Reference Depp, Strassnig, Mausbach, Bowie, Wolyniec, Thornquist and Harvey2014; Fagiolini et al., Reference Fagiolini, Frank, Houck, Mallinger, Swartz, Buysse and Kupfer2002; Fagiolini, Frank, Scott, Turkin, & Kupfer, Reference Fagiolini, Frank, Scott, Turkin and Kupfer2005; Fagiolini, Kupfer, Houck, Novick, & Frank, Reference Fagiolini, Kupfer, Houck, Novick and Frank2003; McElroy et al., Reference McElroy, Kemp, Friedman, Reilly-Harrington, Sylvia, Calabrese and Shelton2015; McIntyre, Mandel, & Pappadopulos, Reference McIntyre, Mandel and Pappadopulos2011; Peters et al., Reference Peters, Shesler, Sylvia, da Silva Magalhaes, Miklowitz, Otto and Deckersbach2015). Magnetic resonance imaging (MRI) studies suggest that the link between higher body mass index (BMI) and poor clinical outcomes has a neurobiological basis. We and others have found that patients with higher BMIs have greater limbic brain volume reductions, lower white matter (WM) integrity, and more pronounced neurochemical changes (Bond et al., Reference Bond, Ha, Lang, Su, Torres, Honer and Yatham2014, Reference Bond, da Silveira, MacMillan, Torres, Lang, Su and Yatham2016, Reference Bond, Silveira, MacMillan, Torres, Lam and Yatham2017; Islam, Metcalfe, MacIntosh, Korczak, & Goldstein, Reference Islam, Metcalfe, MacIntosh, Korczak and Goldstein2018; Kuswanto et al., Reference Kuswanto, Sum, Yang, Nowinski, McIntyre and Sim2014). Similar findings have been reported in major depressive disorder, anxiety disorders, and schizophrenia, suggesting that higher BMI is a risk factor for greater brain illness severity across many psychiatric illnesses (Cole et al., Reference Cole, Boyle, Simmons, Cohen-Woods, Rivera, McGuffin and Fu2013; Coplan et al., Reference Coplan, Fathy, Abdallah, Ragab, Kral, Mao and Mathew2014; Li et al., Reference Li, Deng, He, Li, Huang, Li and Deng2013; Opel et al., Reference Opel, Redlich, Grotegerd, Dohm, Heindel, Kugel and Dannlowski2015).
Excessive weight gain begins early in BD and accrues over time (Hu et al., Reference Hu, Torres, Qian, Wong, Halli, Dhanoa and Yatham2016). However, little is known about how longitudinal weight changes impact the progression of brain abnormalities, or the time course over which it happens, due to a lack of prospective studies. Studying weight gain in first episode mania patients provides a unique opportunity to investigate the relationship between weight gain and brain illness progression, or neuroprogression, from the time of first diagnosis of BD. We previously reported that BD patients with clinically significant weight gain [CSWG, defined by the Food and Drug Administration as gaining ⩾7% of baseline (BL) weight (Sachs & Guille, Reference Sachs and Guille1999)] in the 12 months after their first manic episode experienced more mood episodes and worse psychosocial functioning than patients without CSWG, showing a relationship between weight gain and clinical illness progression (Bond et al., Reference Bond, Kunz, Torres, Lam and Yatham2010; Hu et al., Reference Hu, Torres, Qian, Wong, Halli, Dhanoa and Yatham2016). In the only prospective MRI study in BD, we found that first episode mania patients with CSWG also experienced greater volume loss in the left orbitofrontal cortex (OFC), left anterior cingulate cortex (ACC), and left middle temporal gyrus, all limbic brain areas important to BD (Bond et al., Reference Bond, Su, Honer, Dhanoa, Batres-y-Carr, Lee and Yatham2019). This suggests that CSWG is a risk factor for more rapid neuroprogression in BD and that CSWG-related neuroprogression is present beginning at the first diagnosis of BD.
We conducted the current analyses to investigate whether CSWG is a risk factor for the progression of neurochemical abnormalities in first episode mania BD patients. We focused on hippocampal neurochemical abnormalities because the hippocampus is important to the pathophysiology of BD and hippocampal neurochemical abnormalities are well documented (Otten & Meeter, Reference Otten and Meeter2015; Yildiz-Yesiloglu & Ankerst, Reference Yildiz-Yesiloglu and Ankerst2006). We measured two specific metabolites, N-acetylaspartate (NAA) and glutamate + glutamine (Glx), because lower NAA and greater Glx are the most consistently reported neurochemical abnormalities in BD (Gigante et al., Reference Gigante, Bond, Lafer, Lam, Young and Yatham2012; Maddock & Buonocore, Reference Maddock and Buonocore2012) and we previously found in cross-sectional studies that higher BMI predicted lower hippocampal NAA and greater Glx at recovery from the first manic episode (Bond et al., Reference Bond, da Silveira, MacMillan, Torres, Lang, Su and Yatham2016, Reference Bond, Silveira, MacMillan, Torres, Lam and Yatham2017). We hypothesized that patients with CSWG over 12 months would have greater 12-month decreases in NAA and greater increases in Glx than patients without CSWG.
Methods
Participants
First episode mania BD patients and a control group of healthy comparator subjects (HS) from the University of British Columbia (UBC) Systematic Treatment Optimization Program for Early Mania (STOP-EM) were recruited between July 2004 and February 2016 (Yatham, Kauer-Sant'anna, Bond, Lam, & Tam, Reference Yatham, Kauer-Sant'anna, Bond, Lam and Tam2009). The UBC Clinical Research Ethics Board approved the STOP-EM project. Written informed consent was obtained from participants before any study procedures took place.
STOP-EM enrolled BD patients aged 15–35 years who experienced their first manic/mixed episode in the preceding 3 months. The diagnoses of BD and first mania were made according to DSM-IV-TR criteria and were determined by a psychiatric interview with a research psychiatrist, a review of medical records, and administration of the Mini International Neuropsychiatric Interview (MINI). Patients with comorbid psychiatric and substance use disorders were eligible for enrollment so long as their primary diagnosis was BD. Patients were excluded if their manic episode was not due to BD (e.g. they had a substance-induced manic episode or a manic episode secondary to another medical condition), or if they had a neurological illness or head trauma with loss of consciousness >30 min.
STOP-EM enrolled HS aged 15–35 years who had no personal history of psychiatric illness and no family history in first-degree relatives, confirmed by clinical interview and the MINI. HS were recruited from the greater Vancouver area via flyers, media advertisements, and posts in online forums such as Craigslist.
Clinical assessments
At BL, psychiatric diagnoses were confirmed in patients and excluded in HS. Sociodemographic data and information about illness course in patients, including prior depressive and hypomanic episodes and current medication treatments, were collected. Patients were prescribed maintenance treatment for BD according to Canadian clinical practice guidelines (Yatham et al., Reference Yatham, Kennedy, O'Donovan, Parikh, MacQueen, McIntyre and Gorman2005, Reference Yatham, Kennedy, O'Donovan, Parikh, MacQueen, McIntyre and Beaulieu2006, Reference Yatham, Kauer-Sant'anna, Bond, Lam and Tam2009, Reference Yatham, Kennedy, Parikh, Schaffer, Beaulieu, Alda and Berk2013, Reference Yatham, Kennedy, Parikh, Schaffer, Bond, Frey and Berk2018; Yatham, Kauer-Sant'anna, et al., Reference Yatham, Kauer-Sant'anna, Bond, Lam and Tam2009).
At each study visit [BL, 6 months (M6), and 12 months (M12)], clinical status was assessed in patients using the Young Mania Rating Scale (YMRS), the Montgomery-Asberg Depression Rating Scale (MADRS), and the Positive and Negative Syndrome Scale (PANSS) positive subscale. Medications were recorded. Participants' height was measured with a stadiometer and they were weighed in a non-fasting state in light clothing with footwear removed. Standard definitions of BMI [weight (kg)/height (m)2] and BMI categories (underweight: BMI <18.50, normal weight: BMI = 18.50–24.99, overweight: BMI = 25.00–29.99, and obesity: BMI ⩾30.00) were used. CSWG was defined as gaining ⩾7% of BL weight (Sachs & Guille, Reference Sachs and Guille1999).
Magnetic resonance spectroscopy (MRS) and MRI protocols
MRS data were collected at BL and M12 using a Philips Achieva 3-Tesla scanner (Best, the Netherlands). T2-weighted coronal, sagittal, and axial images for anatomical parameters were obtained. A point resolved spectroscopy sequence [echo time (TE) = 35 ms, repetition time (TR) = 2000 ms] was used to acquire data from 30 mm × 15 mm × 15 mm voxels in both hippocampi. The sagittal image was used to align the long axis of the voxel with the long axis of the hippocampus. Its position in the superior/inferior and medial/lateral directions was then adjusted based on the coronal and axial images to include the maximum amount of hippocampus and minimize the presence of other brain structures and cerebrospinal fluid (CSF). Screen captures of voxel slice positioning and anatomical landmarks were referenced to ensure consistency of voxel placement across participants and timepoints. A screen capture showing voxel placement is in online Supplementary Fig. S1a. A chemical shift selective pulse and crusher gradient were employed to suppress the water signal. Water-unsuppressed signals were also obtained to reference metabolite signals and for eddy current correction. A total of 128 water-suppressed and 16 non-water-suppressed averages were acquired.
At BL and M12, T1-weighted structural MR images were acquired to quantify hippocampal volumes and the proportion of the MRS voxels occupied by gray matter (GM) and WM. The T1-weighted images were obtained using a 3D axial MPRAGE (T1 Turbo Field Echo) sequence [field of view = 25.6 cm, matrix = 256 × 256, isotropic voxels (1 × 1 × 1 mm3 acquired and reconstructed), TR/TE = 7.6/3.5 ms, T/R head coil, flip angle = 8°, shot interval = 1800 ms, TI = 794 ms, SENSE = 0], generating 180 1-mm-thick contiguous slices of the whole brain.
MRS and MRI data extraction
LCModel 6.3 was used to extract concentrations of NAA and Glx normalized to the unsuppressed water spectrum (Provencher, Reference Provencher1993). Sample 1H-MRS spectra from a patient and a HS are shown in online Supplementary Fig. S1b. Normalized metabolite concentrations were converted to institutional absolute concentrations [millimolar (mm) units] by correcting for voxel GM, WM, and CSF water concentrations and water and metabolite signal decay during the TE (Supplementary Materials and Methods). Values for water and metabolite T1 and T2 relaxation times were taken from the literature (Kosior, Lauzon, Federico, & Frayne, Reference Kosior, Lauzon, Federico and Frayne2011; Posse et al., Reference Posse, Otazo, Caprihan, Bustillo, Chen, Henry and Alger2007).
Poor-quality data for individual metabolite values were excluded by (1) multiplying the %s.d. error from LCModel by the corresponding absolute metabolite concentration to obtain the absolute error estimate and (2) rejecting data when the absolute error estimate was >30% of the median metabolite concentration across all measurements. Compared to the conventional threshold of rejecting data with %s.d. error estimates >20%, this method ensures that there is less bias towards rejecting low-concentration metabolite values (Kreis, Reference Kreis2016).
Freesurfer 6.0 subcortical segmentation was used to extract hippocampal structural data from the T1-weighted MRI scans. The volumes were normalized by estimated total intracranial volume generated by FreeSurfer to reduce image distortion from head motion, voxel-scaling, and other artifacts, to control for between-participant (e.g. sex-related) differences in head size, and in keeping with previous studies of weight and brain volumes (Gunstad et al., Reference Gunstad, Paul, Cohen, Tate, Spitznagel, Grieve and Gordon2008; Mathalon, Sullivan, Rawles, and Pfefferbaum, Reference Mathalon, Sullivan, Rawles and Pfefferbaum1993; Whitwell, Crum, Watt, & Fox, Reference Whitwell, Crum, Watt and Fox2001). The proportion of the MRS voxels occupied by GM and WM was determined by FSL 4.1.9 FAST (Smith et al., Reference Smith, Jenkinson, Woolrich, Beckmann, Behrens, Johansen-Berg and Matthews2004; Woolrich et al., Reference Woolrich, Jbabdi, Patenaude, Chappell, Makni, Behrens and Smith2009).
Data analyses
Statistical analyses were carried out using IBM SPSS Statistics for Windows 25.0 (SPSS Inc, Chicago, IL) at a two-tailed significance level of α = 0.05, except where corrected for multiple comparisons as noted below.
Demographic and clinical variables were examined with independent-sample t tests, χ2 tests, and Fisher's exact test. Patient-HS differences in delta-BMI (ΔBMI) were investigated using a general linear model for repeated measures with age and gender as covariates. (All Δvalues here and below are defined as M12 value − BL value so that a positive Δvalue represents an increase and a negative value a decrease.) Patient-HS differences in CSWG were examined with χ2 tests.
The impacts of diagnosis and CSWG on left and right Δhippocampal volumes were investigated with general linear models for repeated measures with diagnosis and CSWG as factors. Within-subject changes from BL to M12 in the proportions of the MRS voxels occupied by GM and WM were examined used paired t tests.
Of a possible 736 datapoints for NAA and Glx (92 participants × 2 metabolites × 2 hippocampi × 2 timepoints), 101 (13.7%) were missing due to poor data quality (n = 9) or because the participant was lost to follow-up or had not reached the M12 follow-up visit (n = 92). Missing MRS and clinical data were imputed using the expectation maximization method, which uses an iterative procedure to fit unbiased expected values to missing data (Dempster, Laird, & Rubin, Reference Dempster, Laird and Rubin1977). Missingness was random (Little's MCAR χ2(5) = 3.579, p = 0.609).
Primary analysis
For our primary analysis, we investigated whether patients with CSWG had greater 12-month decreases in left and right hippocampal NAA and greater increases in left and right Glx than patients without CSWG when adjusting for clinical and treatment variables. We created four general linear models for repeated measures, one for each Δmetabolite. To correct for multiple comparisons across four metabolites, a p value of 0.013 (0.05/4) was considered significant. In each model, the BL and M12 concentrations of the appropriate metabolite were the dependent variables. CSWG was the predictor of interest. To adjust for clinical and treatment variables that could impact CSWG and/or Δmetabolites and confound the relationship between them, the following were included as covariates: the number of pre-manic mood episodes (depressive + hypomanic), mean YMRS, MADRS, and PANSS-positive scores (BL score + M6 score + M12 score / 3), and the total 12-month doses of lithium, valproate, and second generation antipsychotics (SGAs). Total medication doses were defined as the sum of the BL + M6 + M12 doses, with SGA doses converted to chlorpromazine equivalents. To ensure that findings were independent of BL-M12 changes in hippocampal volumes or the GM/WM composition of the MRS voxels, these were also included as covariates.
Secondary analyses
We conducted two secondary analyses to explore additional aspects of the relationship between weight gain and Δmetabolites. (1) We investigated whether the continuous variable of ΔBMI predicted Δmetabolites in patients by repeating the primary models with ΔBMI substituted for CSWG. Because we previously reported that CSWG, but not ΔBMI, predicted worse clinical outcomes in our patients, suggesting the possibility of a floor effect for the impact of weight gain (Bond, Kunz, Torres, Lam, & Yatham, Reference Bond, Kunz, Torres, Lam and Yatham2010; Hu et al., Reference Hu, Torres, Qian, Wong, Halli, Dhanoa and Yatham2016), we hypothesized that CSWG would be a better predictor of Δmetabolites. (2) We investigated whether the relationship between CSWG and Δmetabolites was unique to BD by repeating the models in HS, with CSWG as the predictor and hippocampal volumes and the GM/WM composition of the MRS voxels as covariates; and in the combined sample of patients + HS with diagnosis, CSWG and a diagnosis × CSWG interaction as predictors, and hippocampal volumes and the GM/WM composition of the voxels as covariates.
Results
Participant characteristics, CSWG, and hippocampal volumes
Ninety-two participants (58 patients, 34 HS) were included in our analyses. Sociodemographic characteristics of patients and HS are shown in Table 1, and clinical characteristics of patients in Table 2. Mood rating scale scores in patients were low at BL and M12. Ninety-three percent of patients were treated with mood stabilizers and/or SGAs at BL and 84.5% were treated at M12.
Table 1. Sociodemographic characteristics of first episode mania patients and healthy subjects
BMI, body mass index; ΔBMI, 12-month change in body mass index; CSWG, clinically significant weight gain.
A blank cell indicates the variable was only measured at one timepoint.
Table 2. Clinical and treatment characteristics of first episode mania patients
MADRS, Montgomery-Asberg Depression Rating Scale; PANSS, Positive and Negative Syndrome Scale; YMRS, Young Mania Rating Scale.
A blank cell indicates the variable was only measured at one timepoint.
a Less than the total of (pre-manic hypomanic episodes + pre-manic depressive episodes) because eight patients reported experiencing both hypomanic and depressive episodes.
b Reported for only patients taking each medication, not for the entire sample.
Patients and HS had similar mean BL BMIs and proportions with normal weight, overweight, and obesity (Table 1). M12 ΔBMI was 1.4 kg/m2 in patients v. 0.4 kg/m2 in HS (F (1) = 6.415, p = 0.013). Thirty-three percent of patients v. 17.6% of HS experienced CSWG (χ2(1) = 2.473, p = 0.116).
Neither diagnosis nor CSWG significantly predicted mean Δhippocampal volumes and there were no diagnosis × CSWG interactions (Δleft hippocampus – diagnosis: F (1) = 0.830, p = 0.365; CSWG: F (1) = 0.921, p = 0.340; diagnosis × CSWG: F (1) = 0.342; p = 0.560; Δright hippocampus – diagnosis: F (1) = 2.127, p = 0.148; CSWG: F (1) = 0.000 p = 1.000; diagnosis × CSWG: F (1) = 0.661; p = 0.418). (BL and M12 hippocampal volumes are listed in online Supplementary Table S1.)
Primary analysis
Primary and secondary models passed diagnostic tests including normality of residuals, homoscedasticity, and non-collinearity of independent variables. There were no significant within-subject changes from BL to M12 in the proportions of the MRS voxels occupied by GM or WM (left GM: t = −0.038, p = 0.969; left WM: t = 0.222, p = 0.825; right GM: t = −1.180, p = 0.241; right WM: t = 1.304, p = 0.196).
After correction for multiple comparisons, CSWG in patients predicted a greater decrease in left hippocampal NAA (−0.406 mm in patients with CSWG v. +0.324 mm in patients without CSWG; F (1) = 8.764; p = 0.005) (Fig. 1). The effect size (ES; Cohen's d) was moderate at −0.52. The finding remained significant when we repeated the analysis excluding one outlying ΔNAA of 7.16 mm (F (1) = 7.059; p = 0.011), and when we repeated the analysis including only patients with complete (non-imputed) MRS and clinical data (F (1) = 7.975; p = 0.010). CSWG did not predict a greater decrease in right NAA (−0.332 mm v. −0.566 mm; F (1) = 0.937; p = 0.338) or greater increases in left Glx (+2.300 mm v. +2.963 mm; F (1) = 1.832; p = 0.183) or right Glx (−0.074 mm v. +0.228 mm; F (1) = 0.012; p = 0.912).
Fig. 1. Relationship between CSWG and change in mean left hippocampal NAA in first episode mania patients and healthy subjects.
Secondary analyses
ΔBMI as a predictor of Δmetabolites in patients
ΔBMI predicted a greater decrease in left NAA (F (1) = 18.499; p < 0.001). The finding persisted when the analysis was repeated with the outlying ΔNAA value removed (F (1) = 13.544; p = 0.001) and when it was repeated in patients with complete MRS and clinical data (F (1) = 21.837; p < 0.001). ΔBMI did not predict a greater decrease in right NAA or a greater increase in left or right Glx (p values = 0.344–0.722).
CSWG as a predictor of Δmetabolites in HS and patients + HS
A non-significant trend suggested that CSWG predicted a greater decrease in left NAA in HS (−0.464 mm in HS with CSWG v. +0.461 mm in HS without CSWG; F (1) = 3.652; p = 0.067) (Fig. 1). The ES was moderate at −0.53. The number of HS with CSWG was small (N = 6 of 34 HS) and the 95% confidence interval thus wide, making this result difficult to interpret. However in the model including patients + HS, there was a main effect of CSWG on Δleft NAA (F (1) = 7.602; p = 0.007), but no main effect of diagnosis (F (1) = 0.016; p = 0.900) and no diagnosis × CSWG interaction (F (1) = 0.001; p = 0.977), confirming that CSWG had similar effects in patients and HS.
The remaining models in HS found that CSWG did not predict a greater decrease in right NAA or a greater increase in left or right Glx (p values = 0.392–0.994). Similarly, the remaining models in patients + HS did not find main effects of diagnosis (p values = 0.073–0.331), CSWG (p values = 0.397–0.668), or diagnosis × CSWG interactions (p values = 0.288–0.975) on Δright NAA or Δleft/Δright Glx.
Discussion
This was the first prospective study on the relationship between weight gain and neurochemical abnormalities in BD. Our main finding was that first episode mania patients with 12-month CSWG had a greater 12-month decrease in left hippocampal NAA than patients without CSWG. CSWG was thus a risk factor for more rapid progression of one of the most consistently reported brain abnormalities in BD. The relationship was robust, occurring early in the illness and in patients with modest weight gain relative to that seen over the course of BD. In contrast, our hypothesis that patients with CSWG would experience a greater 12-month increase in Glx was not supported.
The relationship between CSWG and decreasing NAA was not unique to patients. There was also a trend-level significant decrease in left hippocampal NAA in HS with CSWG with an ES similar in magnitude to that observed in BD patients with CSWG. This raises provocative questions about the true cause of neurochemical differences between BD patients and non-BD comparators. Weight gain accrues over the course of BD, and in patients with longstanding illnesses obesity rates are over 60% greater than in the general population (Goldstein et al., Reference Goldstein, Liu, Zivkovic, Schaffer, Chien and Blanco2011). If BD patients differ systematically from non-BD comparators in obesity, then either BD diagnosis or the difference in obesity rates could account for neurochemical differences between the two groups. In our sample, weight was a stronger predictor than diagnosis. Additional studies in patients with longer illnesses are needed to confirm this.
Interpreted alongside our previous prospective volumetric MRI report (Bond et al., Reference Bond, Su, Honer, Dhanoa, Batres-y-Carr, Lee and Yatham2019), the current study constitutes proof-in-principle that weight gain is a risk factor for neuroprogression from the time of first diagnosis of BD. Our previous MRI study, conducted in the same cohort as the current one, found that CSWG predicted greater 12-month volume loss in limbic brain regions important to BD, including the left OFC, left ACC, and left middle temporal gyrus (Bond et al., Reference Bond, Su, Honer, Dhanoa, Batres-y-Carr, Lee and Yatham2019). The left hemisphere predominance of CSWG-related volumetric and neurochemical brain changes is noteworthy in light of the fact that left-sided brain lesions are linked to depression (Satzer & Bond, Reference Satzer and Bond2016). It thus suggests a possible neurobiological explanation for the link between obesity and more frequent and severe depressive episodes in BD (Fagiolini et al., Reference Fagiolini, Kupfer, Houck, Novick and Frank2003; Goldstein, Liu, Schaffer, Sala, & Blanco, Reference Goldstein, Liu, Schaffer, Sala and Blanco2013). Obesity is also a risk factor for depression in the general population and the CSWG-related reduction in NAA in our HS also suggests a possible mechanism underlying this relationship.
The relationship between CSWG and decreasing hippocampal NAA was independent of hippocampal volume changes, which were not related to CSWG. The lack of association between weight and hippocampal volumes in BD, which was also previously reported in a small cross-sectional study (Viana-Sulzbach, Pedrini, Bücker, Brietzke, & Gama, Reference Viana-Sulzbach, Pedrini, Bücker, Brietzke and Gama2016), has implications for understanding the relationships between weight gain and neuroprogression in BD. First, it suggests that CSWG predicts primary changes in hippocampal NAA rather than changes that are secondary to hippocampal structural abnormalities. Second, together with our previous volumetric report, it suggests that some limbic brain areas are more prone to CSWG-related volume loss than others, at least early in the course of BD.
The hippocampus and NAA play important roles in the pathophysiology of BD. The hippocampus, particularly its rostral division, is a key limbic brain structure with reciprocal connections to brain areas involved in generating and modulating emotional responses including the prefrontal cortex, amygdala, and ventral striatum (Small, Schobel, Buxton, Witter, & Barnes, Reference Small, Schobel, Buxton, Witter and Barnes2011). Neuroimaging studies demonstrate smaller hippocampal volumes in BD patients (Hallahan et al., Reference Hallahan, Newell, Soares, Brambilla, Strakowski, Fleck and McDonald2011; Hartberg et al., Reference Hartberg, Jorgensen, Haukvik, Westlye, Melle, Andreassen and Agartz2015; Haukvik et al., Reference Haukvik, Westlye, Morch-Johnsen, Jorgensen, Lange, Dale and Agartz2015; Otten & Meeter, Reference Otten and Meeter2015). NAA is an osmolyte, a storage depot for glutamate (as N-acetylaspartylglutamate), and the second most abundant amino acid in the brain after glutamate. It is synthesized in mitochondria in neurons and oligodendrocytes and is considered to be a marker of neuron and myelin function and mitochondrial health (Baslow, Reference Baslow2003; Maddock & Buonocore, Reference Maddock and Buonocore2012; Nordengen, Heuser, Rinholm, Matalon, & Gundersen, Reference Nordengen, Heuser, Rinholm, Matalon and Gundersen2015). Reduced NAA is one of the most robust neurochemical findings in BD (Maddock & Buonocore, Reference Maddock and Buonocore2012).
Despite the prospective design of this study, we cannot be certain of the causal direction of the relationship between CSWG and decreasing left hippocampal NAA since CSWG and ΔNAA occurred over the same 12-month period. One possibility is that CSWG causes decreasing NAA, possibly by altering biomarkers that impact the brain such as inflammatory cytokines, adipokines, and others. In support of this explanation, previous studies have shown that the hippocampus is vulnerable to obesity-related damage. In animal models, diet-induced obesity led to hippocampal inflammation, decreased neurogenesis, and impaired hippocampus-related cognitive functioning (Grayson et al., Reference Grayson, Fitzgerald, Hakala-Finch, Ferris, Begg, Tong and Benoit2014; Kanoski & Davidson, Reference Kanoski and Davidson2011; Kanoski, Zhang, Zheng, & Davidson, Reference Kanoski, Zhang, Zheng and Davidson2010; Stangl & Thuret, Reference Stangl and Thuret2009). In human studies, higher BMI in elderly subjects predicted greater prospectively measured hippocampal atrophy (Cherbuin, Sargent-Cox, Fraser, Sachdev, & Anstey, Reference Cherbuin, Sargent-Cox, Fraser, Sachdev and Anstey2015).
However, alternate explanations need to be considered. One is reverse causality – that hippocampal abnormalities cause weight gain. Supporting this, the hippocampus plays a role in regulating food intake and body weight. It receives projections from brain areas involved in appetitive and ingestive behaviors, including the visual, olfactory, and prefrontal cortices, the hypothalamus, and the ventral striatum (Martin & Davidson, Reference Martin and Davidson2014). It integrates external food stimuli, internal interoceptive and reward cues, and learned memories to influence where, when, and the degree to which food is consumed (Kanoski & Davidson, Reference Kanoski and Davidson2011). In animal models, experimentally induced impairment in hippocampal functioning caused significant weight gain (Davidson et al., Reference Davidson, Hargrave, Swithers, Sample, Fu, Kinzig and Zheng2013), while in human studies subjects with bilateral hippocampal damage had markedly excessive food intake (Rozin, Dow, Moscovitch, & Rajaram, Reference Rozin, Dow, Moscovitch and Rajaram1998). A second alternate explanation is that a third variable, such as genetic polymorphisms, causes both increased weight and reduced hippocampal functioning. In support of this hypothesis, polymorphisms in the fat mass and obesity (FTO) gene are associated with increased BMI and reduced brain volumes across the lifespan (Ho et al., Reference Ho, Stein, Hua, Lee, Hibar, Leow and Thompson2010; Melka et al., Reference Melka, Gillis, Bernard, Abrahamowicz, Chakravarty, Leonard and Pausova2013).
Strengths of the current study include its prospective design and state-of-the-art measurement of hippocampal neurochemicals. Examining a first episode mania BD sample allowed us to minimize confounding factors present in people with longer illnesses, such as variable illness durations, high comorbidity rates, and multiple medication trials. We rigorously adjusted our models for potentially confounding clinical and treatment variables and for diagnosis-, CSWG-, and time-dependent differences in the MRS voxels that could have impacted our results. Including HS in our analyses allowed us to investigate whether our findings were unique to BD. Limitations included our relatively small sample, particularly of HS, and the fact that we did not collect data on diet and physical activity, which could modulate the relationships between weight gain and the brain. We also did not collect data on pre-manic use of mood stabilizers or SGAs, although it stands to reason that most patients were naive to these medications prior to their first mania. We were unable to distinguish between different causes of CSWG, such as medications and genetic factors, which could conceivably have different biological effects. We also did not gather data on biomarkers altered by weight gain, such as inflammatory cytokines, adipokines, and others, and so we could not investigate whether these mediated the relationship between CSWG and hippocampal neurochemical changes.
In conclusion, we found that BD patients with CSWG in the 12 months after their first manic episode and HS with 12-month CSWG had greater decreases in left hippocampal NAA than comparators without CSWG. Together with our previous prospective volumetric MRI study, this suggests that CSWG is an important risk factor for neuroprogression in BD and that this relationship is present from the time of first diagnosis of BD. Our findings further suggest that neurobiological abnormalities that are typically considered to be BD diagnosis related may in fact be due to patient-control differences in BMI. Further studies are needed to determine how the relationship between weight gain and brain changes in BD patients evolves over longer illness durations and whether it plays a role in other psychiatric illnesses with high obesity rates, such as schizophrenia and major depressive disorder.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721000544.
Acknowledgements
The authors thank the BD patients and HS who enrolled in STOP-EM. A modified version of this report was presented at the 16th Annual Meeting of the International Society for Bipolar Disorders, July 13–16, 2016, Amsterdam, the Netherlands.
Financial support
The data for this manuscript were generated from the Systematic Treatment Optimization Program for Early Mania (STOP-EM), which was supported by an unrestricted grant to LNY from AstraZeneca Canada. The sponsor had no input into the design or conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.
Conflict of interest
Dr Bond has received consulting fees and/or research grants from: Alkermes PLC, Myriad Genetics, Pfizer, the Canadian Institutes of Health Research (CIHR), NIDA, and the University of Minnesota Foundation. Dr Silveira reports no conflicts of interest. Dr Torres has received speaking/consulting fees from Lundbeck, Sumitomo Dainippon, and Community Living British Columbia. Dr Lam has received speaking/consulting fees and/or research grants from Akili, Allergan, Asia-Pacific Economic Cooperation, BC Leading Edge Foundation, Brain Canada, Canadian Institutes of Health Research, Canadian Network for Mood and Anxiety Treatments (CANMAT), Canadian Psychiatric Association, CME Institute, Hansoh, Janssen, Lundbeck, Lundbeck Institute, Medscape, Mind Mental Health Technologies, Otsuka, Pfizer, St. Jude Medical, University Health Network Foundation, and Vancouver General Hospital Foundation. Dr Yatham has received speaking/consulting fees and/or research grants from Alkermes, Allergan, AstraZeneca, Bristol Myers Squibb, CANMAT, CIHR, Dainippon Sumitomo Pharma, Janssen, Lundbeck, Otsuka, Sunovion, and Teva.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
