Indirect Calorimetry to Measure Metabolic Rate and Energy Expenditure in Psychiatric Populations: A Systematic Review

Psychiatric and metabolic disorders are highly comorbid and the relationship between these disorders is bidirectional. The mechanisms underlying the association between psychiatric and metabolic disorders are presently unclear, which warrants investigation into the dynamics of the interplay between metabolism, substrate utilization, and energy expenditure in psychiatric populations, and how these constructs compare to those in healthy controls. Indirect calorimetry (IC) methods are a reliable, minimally invasive means for assessing metabolic rate and substrate utilization in humans. This review synthesizes the extant literature on the use of IC on resting metabolism in psychiatric populations to investigate the interaction between psychiatric and metabolic functioning. Consistently, resting energy expenditures and/or substrate utilization values were significantly different between psychiatric and healthy populations in the studies contained in this review. Furthermore, resting energy expenditure values were systematically overestimated when derived from predictive equations, compared to when measured by IC, in psychiatric populations. High heterogeneity between study populations (e.g., differing diagnoses and drug regimens) and methodologies (e.g., differing posture, time of day, and fasting status at measurement) impeded the synthesis of results. Standardized IC protocols would benefit this line of research by enabling meta-analyses, revealing trends within and between different psychiatric disorders.


Introduction
Psychiatric and metabolic disorders constitute two of the greatest burdens to the health of individuals globally. The disability-adjusted life years (DALYs) incurred by psychiatric and metabolic disorders have been continually rising across most of the world [1,2]. Elevated systolic blood pressure, fasting glucose, and body-mass index (BMI) were identified as the 1st, 3rd, and 4th leading risk factors for DALYs, respectively [1,2]. Psychiatric disorders also rank among the top 10 leading causes of global disease burden and show no signs of decreasing incidence [2].
Moreover, mounting evidence indicates a bidirectional, longitudinal relationship between metabolic and psychological wellbeing [3]. Evidence suggests that the years of life lost due to psychiatric disorders are often underestimated, and do not truly reflect the true morbidity and high premature mortality rates observed in populations with psychiatric disorders [1]. Furthermore, greater than half of the excess mortality observed in psychiatric populations is attributable to cardiovascular diseases or other metabolic conditions (e.g., diabetes mellitus, non-alcoholic fatty liver disease, obesity, cholesterol disorders, etc.), which are not adequately treated, and sometimes exacerbated, by currently available psychiatric therapies like some antidepressants and antipsychotics [4,5]. Conversely, preliminary evidence suggests that insulin resistance in the brain contributes to the pathophysiology of psychiatric disorders, and that drugs which increase insulin sensitivity could be effective in treating some psychiatric patients [6]. The foregoing indicates that metabolic comorbidities are overly represented in patients with psychiatric disorders, and that these comorbidities might be more modifiable and higher-priority treatment targets to optimize quality of life and patient well-being [4,5].
For the purposes of this review, metabolism refers to the biochemical processes wherein organisms convert chemical energy from food into the cellular energy and building blocks (i.e., proteins, lipids, carbohydrates, and nucleic acids) required to sustain life. Adenosine triphosphate (ATP) is the primary chemical energy currency in eukaryotic cells. At rest, cellular energy homeostasis is maintained primarily by the oxidation of intracellular carbohydrate, lipid, and protein stores to facilitate the biochemical reactions necessary to synthesize ATP for cellular energy [7]. Oxidation of these substrates yields carbon dioxide (CO 2 ), water (H 2 O), and heat as metabolic by-products [8]. Additionally, comparatively small amounts of ATP are produced through anaerobic energy pathways involving anaerobic glycolysis and the phosphagen system [7,9]. The total energy expenditure (TEE)defined as the total heat energy released by the human body per day [10]-of an individual can be divided into three components: (1) resting energy expenditure (REE), which refers to the energy requirements necessary to establish energy homeostasis at rest; (2) activity energy expenditure (AEE), which refers to the energy requirements necessary to establish energy homeostasis during physical movement; and (3) the thermogenic effect of food, which refers to the energy requirements necessary to digest, absorb, and store nutrients from food.
As a large percentage of REE is derived from oxidative reactions, the associated metabolic rates can be validly and reliably estimated based on the rate of oxygen (O 2 ) consumption and CO 2 , H 2 O, and heat production [11]. In laboratory settings, measurement of human energy expenditure (EE) can be achieved via direct or indirect calorimetry (IC) methods.
Direct calorimetry is considered the "gold standard" for assessing EE in humans and involves determining EE within a thermally isolated chamber that precisely measures changes in temperature that can only be attributed to metabolic heat production. Due to the relative unavailability of human direct calorimeters, IC methodologies that are more practical for clinical research purposes have been developed to reliably estimate human EE. The IC technique considered the most valid and reliable for estimating average EE in free-living humans over a period of days is the doubly labelled water method (DLW), which involves the oral ingestion of the stable isotopes deuterium ( 2 H) and 18 O and subsequent measurement of their dilution and elimination rates [12][13][14]. The valid implementation of this technique involves well controlled, complex, and expensive analytical techniques, which make the DLW method one that has not been broadly adopted by researchers [12]. In contrast, respiratory gas exchange systems are a widely used, relatively non-invasive IC methodology for the real-time assessment of human EE.
The respiratory gas exchange method for determining metabolic rate is based on the measurement of inspired and expired respiratory gases for the calculation of the volume and rate of oxygen consumption ( . VO 2 ) and carbon dioxide production ( . VCO 2 ). The ratio between . VCO 2 and . VO 2 , known as the respiratory exchange ratio (RER) in organisms, can be used to quantify the substrate(s) being oxidized to fuel energy metabolism by comparing measured RER values to the known RERs associated with the complete oxidation of different macronutrients [8]. When at a metabolic steady state, RER typically ranges between 0.7 and 1.0, with the former indicating that 100% of oxidative metabolism is fueled by lipids and the latter indicating that 100% of metabolism is fueled by carbohydrates. A steady-state RER of~0.85 indicates that oxidative metabolism is fueled by a mixture of carbohydrate and fat oxidation with relatively minimal contributions from protein metabolism in healthy, non-fasting individuals [15]. Respiratory gas exchange is well established as a reliable and valid method for calculating EE in resting and exercising animals, including humans, for periods ranging from a few minutes to several hours, and has been widely used with human research participants for nearly a century [16][17][18].
While the complex bidirectional relationship between metabolic and psychiatric health continues to be elucidated, a paucity of studies have directly investigated energy expenditure in psychiatric populations. Notwithstanding, evidence has begun to emerge suggesting metabolic perturbations in populations with psychiatric disorders due to lifestyle (e.g., increased sedentary behaviour in depression), iatrogenic (e.g., antipsychotics or antidepressants associated with increased fat mass and low-density lipoprotein cholesterol), and other unknown factors [19]. Thus, as interest in the metabolic bases of psychiatric conditions grows, it is necessary to synthesize the extant literature on EE testing in psychiatric populations to identify/create best practices and inform the methodologies of future studies. The aim of this systematic review is to summarize the methodologies and observations from studies which have used IC to measure metabolic rate/energy expenditure in adult psychiatric populations.

Search Strategy
An open search was conducted on the EMBASE, PsycInfo, and PubMed databases, as well as ClinicalTrials.gov and the first 10 pages of Google Scholar, through 26  ). This search identified 2862 additional records that were incorporated into the review. In total, 14 of the reports included in the current review were identified in the first search, while 5 more relevant reports were included from the second search. The raw search results were exported to covidence.org and two reviewers (JD and LO) performed the title and abstract review, full-text review, and data extraction, with discrepancies resolved by discussion with input from a third party (RM) until consensus was reached. A PRISMA flow diagram ( Figure 1) was used to organize the search results and Mendeley Desktop 1.19.8 was used to manage references and deduplication. Due to the high heterogeneity and small number of available studies (mainly consisting of non-randomized or observational/cross-sectional studies), statistical synthesis was not considered viable.

Study Selection and Eligibility Criteria
Original interventional and observational studies including clinical trials, retrospective studies, and open-label studies were eligible for inclusion in the first search if they were conducted in adult humans and published in English. Studies that used IC to assess resting metabolic rate in populations with psychiatric disorders were included in the current review. Review papers, conference, and poster presentations were excluded to avoid capturing the same study sample twice. We also excluded studies in which methods other than IC were used to assess metabolic rate (e.g., equation-derived estimation using anthropometrics), studies involving participants with eating disorders or neurodegenerative diseases (to avoid confounding effects on metabolism due to closely linked pathogenesis), and studies involving individuals <18 years of age.

Study Selection and Eligibility Criteria
Original interventional and observational studies including clinical trials, retrospective studies, and open-label studies were eligible for inclusion in the first search if they were conducted in adult humans and published in English. Studies that used IC to assess resting metabolic rate in populations with psychiatric disorders were included in the current review. Review papers, conference, and poster presentations were excluded to avoid capturing the same study sample twice. We also excluded studies in which methods other than IC were used to assess metabolic rate (e.g., equation-derived estimation using anthropometrics), studies involving participants with eating disorders or neurodegenerative diseases (to avoid confounding effects on metabolism due to closely linked pathogenesis), and studies involving individuals <18 years of age.

Quality Appraisal Strategy
Risk of bias was measured using the Newcastle-Ottawa Scale for non-randomized case-control and cohort studies [20], the ROBINS-I tool for non-randomized interventional studies [21], the ROB2 tool for randomized trials [22], and the Joanna Briggs Institute (JBI) critical appraisal checklist for analytical cross-sectional studies [23].

Quality Appraisal Results
Results of the quality appraisal are shown in Table 1. Overall, the studies by Caliyurt et al. [24], Hassapidou et al. [30], Gaist et al. [27], and Virkkunen et al. [41] were deemed to have a moderate risk of bias due to confounding factors for which reasonable attempts

Quality Appraisal Strategy
Risk of bias was measured using the Newcastle-Ottawa Scale for non-randomized case-control and cohort studies [20], the ROBINS-I tool for non-randomized interventional studies [21], the ROB2 tool for randomized trials [22], and the Joanna Briggs Institute (JBI) critical appraisal checklist for analytical cross-sectional studies [23].

Quality Appraisal Results
Results of the quality appraisal are shown in Table 1. Overall, the studies by Caliyurt et al. [24], Hassapidou et al. [30], Gaist et al. [27], and Virkkunen et al. [41] were deemed to have a moderate risk of bias due to confounding factors for which reasonable attempts were made by the authors to mitigate their impacts, whereas Gewirtz et al. [29] was deemed to have a serious risk of bias due to confounding factors with no attempt to assess and/or mitigate their potential impacts. The rest of the studies were deemed to be of high methodological quality.

Characteristics and Findings of Included Studies
Of the 19 studies included, 8 were cross-sectional studies, 6 were case-control studies, 4 were non-randomized interventional studies, and 1 was a randomized controlled trial. There were 6 studies conducted in the United States, 3 in Australia, 2 each in Italy, Greece, and Finland, and 1 each in Japan, South Korea, Sweden, and Turkey. The sample sizes ranged from 8 to 989 participants. In aggregate, 1875 participants were enrolled across the 19 included studies as follows: 198 healthy controls (HCs); 92 bipolar-I disorder (BD-I); 287 schizophrenia/schizophreniform disorder; 10 seasonal affective disorder (SAD); 15 treatment-resistant depression (TRD); 1117 severe mental illness (SMI); 11 major depressive disorder (MDD); 145 habitually violent offenders with antisocial personality disorder (APD). Of the 19 included studies, 18 used only respiratory gas exchange measurements as the method of indirect calorimetry and metabolic rate calculations; whereas one study used a respiratory gas exchange method to measure REE and used the DLW method to measure TEE over 10 days. Moreover, 15 studies were observational, while 4 were interventional, investigating the effects of a prescribed Mediterranean diet, bright light therapy, or various drugs on body composition and metabolism.
There was a notable degree of heterogeneity in the IC protocols reported by each study with respect to the sampling duration, time of day, digestive state, and postural positioning, as well as how each study defined "steady state" and the participants being "at rest", all of which may affect IC measurements. While in summary the results herein indicate that REE is consistently altered in psychiatric patients compared to healthy controls, the directionality (i.e., whether/when REE is increased or decreased) and whether these alterations represent a cause or rather a consequence of the psychiatric condition, remain to be illuminated. The IC methodologies and results of each study are described in Table 2.

Resting Energy Expenditure in Psychiatric versus Control Participants
In studies where absolute REE values were similar between psychiatric patients and controls, differences in substrate utilization could be observed as patients tended to exhibit reduced lipid oxidation compared to healthy controls [26,37]. Some studies reported higher REE values in patients with Bipolar I Disorder [24] and Seasonal Affective Disorder in off-light conditions during the winter [27], whereas other reports found consistently lower REE and/or non-oxidative glucose metabolism in psychiatric patients [33,41,42]. One study found no difference in measured REE between BD-I patients and healthy controls [39], and another study found that schizophrenia patients had higher RER than controls, but REE did not differ between groups when corrected for FFM [37].

Predictive Equations versus Indirect Calorimetry to Measure Energy Expenditure
Predictive equations for metabolic rate/EE were not as accurate in psychiatric patients as they were in healthy controls [32,36,[38][39][40]. Four studies demonstrated commonly used predictive REE equations to systematically overestimate REE in psychiatric patients [36,[38][39][40]. Between studies, there were discordant results in the suitability of specific predictive equations at estimating REE. For example, three studies found the Harris-Benedict equation to overestimate REE [36,38,39], whereas Sugawara et al. [40] did not find significant predictive bias in the Harris-Benedict equation in patients with schizophrenia or schizoaffective disorders. Furthermore, Miniati et al. [32] found that REE values measured by IC differed significantly from those estimated using three commonly used predictive equations, thus rendering the equations clinically inappropriate for estimating REE in female patients with BD-I, compared to IC.  BD-I patients and healthy controls had comparable insulin resistances (mean ± SEM HOMA-IR = 2.7 ± 0.7 vs. 2.5 ± 0.7, for patients and controls, respectively; p = 0.79). BD-I patients had 13% lower fat oxidation at rest than HCs, but resting metabolic rates were comparable.
Activity monitors revealed that neither mean total daily energy expenditure, nor energy expenditure during PA was different between patients and controls.
There were also no differences in mean time spent in sedentary, moderate, or vigorous activity.  There was 6 out of the 15 patients with evidence of occult hypothyroidism, all of whom responded to thyroid hormone medication and achieved a normal metabolic rate and/or thyroid hormones, with a reduction in depression, except for one who was lost to follow up. The remaining 9 participants were euthyroid, and/or eumetabolic or hypermetabolic and thus thyroid hormone interventions were not initiated.  Progressive statistically significant reductions in mean weight, fat mass, waist circumference, and BMI throughout the duration of monitoring (p < 0.001). The mean final weight loss was 9.7 kg and BMI decreased to 30.7 kg/m 2 (p < 0.001). The mean final fat mass loss was 8.0 kg, and the mean final waist circumference reduction was 10.3 cm (p < 0.001) compared to baseline. Significant and continual reductions were observed in fasting plasma glucose, total cholesterol, and triglycerides concentrations throughout the study (p < 0.001). REE decreased significantly in completers at months 3 and 6 compared to baseline (p < 0.001).         Habitually violent, incarcerated offenders with APD had significantly lower non-oxidative glucose metabolism, basal glucagon, and free fatty acids when compared with normal controls, but glucose oxidation and CSF 5-HIAA did not differ markedly between these groups. The effect sizes for lower non-oxidative glucose metabolism among incarcerated and non-incarcerated APD subjects were 0.73 and 0.51, respectively, when compared with controls, indicating that this finding was not explained by incarceration. Habitually violent offenders with APD have markedly lower glucagon and non-oxidative glucose metabolism when compared with healthy controls, and these findings were more strongly associated with habitual violent offending than low CSF 5-HIAA levels.

Discussion
The present systematic review sought to summarize the methodological and observational data made available in the English literature reporting metabolic rate and/or energy expenditure data measured by IC methodologies in human psychiatric populations. In general, the studies included herein reported mixed findings, mainly due to the high degree of heterogeneity between study objectives and methodologies. As measured EE values are highly dependent on the IC protocol employed (i.e., time spent resting, body positioning, time of day, digestive status), consequently, inter-study comparability is limited. Nevertheless, in the available case-control studies, most investigations observed healthy controls and psychiatric patients to have differences in metabolism, reflected by the measures of REE, TEE, or RER [24,33,37,41,42].
Such disparate between-study results could be due to differences in many factors such as measurement methodologies, pharmacologically-induced alterations in metabolism, or metabolic profiles among different psychiatric illnesses. However, it is noted that the limited evidence does suggest disparate metabolic profiles between those with or without psychiatric illness, and potentially between those with different psychiatric illnesses ( Table 2). Characterizations of these metabolic differences, and whether their relationship to the psychiatric illness are correlative or causative, are still to be determined. Moreover, the duration/chronicity and severity of psychiatric disease could affect REE values directly or indirectly through changes in food intake, intake, eating behaviour, and the accompanying changes in body composition. For example, reduced appetite and low energy are common symptoms of depressive episodes [43]. Over time, reduced caloric intake coupled with low levels of physical activity can cause changes in body composition and reductions in REE. The stability of the disease state in this case could mediate the change in REE, as a more chronic depressive phenotype would confer greater metabolic aberrancies relative to healthy controls or individuals with less chronic depressive phenotypes.
Another general finding from this review was that the correlations between measured and predicted EE values were largely weak, especially in patients. This is perhaps due to underlying metabolic aberrations or medications often present in the patient populations [3,44]. Nevertheless, this highlights the importance of diligent IC methodologies and the use of gas exchange IC devices when feasible, rather than predictive equations, for determining EE in psychiatric populations. Taken together, this evidence suggests that the commonly used predictive REE equations are likely unreliable and invalid for estimating REE in psychiatric patients, regardless of whether they are accurate for healthy controls. Further research efforts are necessary to develop valid and reliable predictive equations for use in psychiatric populations.
Drug therapy can aggravate or alleviate metabolic aberrations that may be present in psychiatric patients. Many antipsychotic drugs act by influencing monoaminergic neurotransmission in distinct brain regions known to influence eating behaviour. For example, olanzapine and clozapine have high affinity for the 5-HT2C receptor, which serve important functions in the regulation of motor behaviour and appetite [45]. As such, consumption of these antipsychotics often precipitates increases in food intake and body weight, whereas other antipsychotics such as aripiprazole are less problematic from a weight-gain perspective, due in part to a lower affinity for 5-HT2C receptors [46]. Heterogenous methodologies and patient populations, as well as the small sample sizes of the studies included in this review, precluded direct comparisons of the effects of different drug classes on metabolic responses. Randomized, controlled studies comparing multiple classes of drugs within a homogenous patient population, as well as studies comparing the same drugs in different patient populations, will help elucidate these effects. A preponderance of studies focused on investigating the effects of antipsychotics or other psychiatric drugs on EE, without a control group of medication-free psychiatric participants. Although clinically relevant and representative of naturalistic samples, this leaves a knowledge gap in the overall characterization of EE in psychiatric patients. Pharmacological intervention involving antipsychotic or mood-modulating drugs may also have indirect effects on EE by changing voluntary movement patterns due to altered motivational states. This may lead to systematic over-or under-estimation of EE in psychiatric populations that is a function of the drug, rather than the illness per se. For example, ref. [35] assessed TEE and AEE in male schizophrenia patients undergoing pharmacological intervention with clozapine for symptom management. The authors found that TEE was significantly lower than World Health Organization recommendations, due largely to sedentarism in people with schizophrenia who take clozapine. It is unclear what to make of these results, however, as there were no direct comparisons made to healthy age-matched controls or patients not taking clozapine. Additionally, the use of pharmacological interventions in the available research makes it difficult to quantify if differences in metabolic activity in psychiatric patients are attributed to psychiatric illness or a consequence of the drug response. Without further dedicated efforts to classify the unique metabolic profiles of psychiatric patients, it is difficult to elucidate the effects of psychiatric illness on metabolism.
Overall, 18 studies with low or moderate risk of bias fit the inclusion criteria for review, while one included study had a high risk of bias. Due to the paucity of available studies and high heterogeneity among investigations, no meta-analysis could be conducted as valid mathematical analyses were not viable in this review. Despite this limitation, the trends observed and discussed herein merit reporting and future consideration. This systematic review reinforces the importance of methodological consistency and rigor for obtaining valid and reliable measurements of EE via IC in psychiatric patients to enable comparisons between subjects as well as between groups. For example, it is important that patients are conclusively in a rested and relaxed state, as even slight fidgeting or anxiousness during IC measurements can influence REE outcomes and can often go unrecognized by the examiner. For this reason, it is recommended that IC measurements should last at least 20-30 min to allow sufficient time for individuals to reach a true resting equilibrium. All included studies in this review which reported the timing of IC assessments measured participants for at least 20 min, however some groups [32,41,42] did not provide details on the timing of the IC assessments. Thus, methodological variability can increase uncertainty and introduce undue heterogeneity in the interpretation and translation of the results to a broader synthesis of evidence. Best practices for respiratory gas exchange IC have not yet been established or agreed upon by the scientific community, so researchers must develop and validate their own protocols, which contributes to the high methodological heterogeneity. The creation of a consortium of scientists, clinicians, and researchers with the goal of establishing an international, gold-standard methodology for respiratory gas exchange IC would be an important step toward reducing the issue of methodological heterogeneity.

Conclusions
At present, a scarcity of research has investigated IC in psychiatric populations. Based on the available evidence, it appears that the metabolic profiles of psychiatric patients may be discordant to those of healthy controls. Furthermore, the results herein indicate that predictive equations used to estimate EE in healthy individuals are likely inadequate for estimating EE in psychiatric patients for reasons that remain unclear, warranting further investigation. Lastly, studies examining the effects of pharmacological intervention on metabolic rate confound the interpretation of the unique effects of psychiatric illness on bioenergetics. Further research efforts are necessary to elucidate the complex relationships among energy metabolism, the brain, and behaviour, thus enabling future meta-analyses.

Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.