Mood versus energy/activity symptoms in bipolar disorder: which cluster of Hamilton Depression Rating Scale better distinguishes between mania, depression, and euthymia?

Introduction: Although bipolar disorder (BD) is traditionally included among mood disorders, some authors believe that changes in energy and motor activity, rather than mood changes, represent the true cardinal symptoms in mania and depression. The aim of the current study was to identify which cluster of the Hamilton Depression Rating Scale (HAM-D) better distinguishes between mania, depression and euthymia. Method: A group of 106 patients with BD were followed for 13 years and repeatedly assessed with the HAM-D as well as with other clinical scales. To perform a comparison, HAM-D items were classified according to clinical criteria into three clusters: energy/ activity symptoms, mood symptoms, and other symptoms. Item response theory (IRT) analyses were performed to provide a test information curve for those three clusters. We measured the prevalence of one cluster of symptoms over the other two throughout the latent trait. Results: Considering HAM-D items individually, the IRT analysis revealed that there was a mixture of mood and energy/ activity symptoms among the most discriminative items, both in depression and in euthymia. However, in mania, only energy/activity symptoms – i.e., general somatic symptoms and retardation – were among the most informative items. Considering the classification of items, both in depression as in mania, the energy/activity cluster was more informative than the mood cluster according to the IRT analysis. Conclusion: Our data reinforce the view of hyperactivity and motor retardation as cardinal changes of mania and depression, respectively.


Introduction
Although traditionally included among mood disorders, bipolar disorder (BD) is especially characterized by changes in energy and motor activity.
Actigraphy studies have revealed that mania 1-3 and depression 4,5 are associated respectively with increased and decreased motor activity. Moreover, several factor analysis studies of manic symptoms led to the conclusion that hyperactivity was the core feature of mania. [6][7][8] Comparing to the Diagnostic and Statistical Manual of Mental Disorders, 4th edition, Text Revision (DSM-IV-TR), 9 the DSM-5 10 brought an important change regarding the diagnostic criteria for manic episodes. In DSM-IV-TR, overactivity was included among the items for the diagnosis of mania, but this change could be absent. Now, in addition to elevated mood or irritability, increased energy or activity is mandatory. In contrast, motor or energy disturbance is not a cardinal symptom in neither version of the psychiatric classification manual for the diagnosis of major depressive episodes.
Item 2 of the Young Mania Rating Scale (YMRS) 11 -the most used scale for manic symptoms -assesses increased motor activity/energy. However, in the Hamilton Depression Rating Scale (HAM-D) 12 -the gold standard scale for depressive symptoms -energy and activity are not associated with a single item. In HAM-D, three items are linked to activity (8 -retardation, 9 -agitation, and 7 -work and activities), 13 and at least one item is linked to energy (13 -general somatic symptoms).
Our objective was to identify which item cluster of the HAM-D is most informative about the occurrence and severity of BD clinical states. We intended to compare items related to mood symptoms to items related to energy/activity, in order to establish which item cluster is the most relevant one for the assessment of mania, depression, and euthymia.

Subjects
The sample was selected from an outpatient research unit at Instituto de Psiquiatria, Universidade Not all the patients participated in the study at the same time, and they were not necessarily assessed during the entire period of 13 years. The only exclusion criterion was refusal to participate in the study. All patients were 18 years old or older and signed an informed consent form. The local ethics committee approved the project.

Assessment
Patients were administered the Structured Clinical Interview for DSM-IV (SCID) 14 for the evaluation of the psychiatric diagnosis. Each patient was repeatedly assessed with the YMRS, 11 the HAM-D, 12

and the Clinical Global
Impressions Scale for use in bipolar illness (CGI-BP). 15 Only three evaluations were considered for data analysis: the one regarding the most severe manic episode, the one regarding the most severe depressive episode, and the one regarding the least symptomatic period of euthymia.
We start from the premise that the most serious episodes, i.e., the ones with the highest scores, would be the most representative of mania and depression. This premise followed a criterion of convenience. The three evaluations could have been performed any time during the 13year follow-up period. DSM-IV-TR criteria were applied for the diagnosis of manic and depressive episodes and euthymia. CGI-BP scores were used as the criterion of severity. In cases of draw, YMRS and HAM-D total scores were considered, i.e., if two manic episodes had the same CGI-BP score, the episode with the highest YMRS score was chosen. Similarly, if two depressive episodes had the same CGI-BP score, the episode with the highest HAM-D score was chosen.
To perform a comparison between energy/activity and mood symptoms assessed by the HAM-D, the items of the scale were classified into three clusters, according to personal clinical criteria defined by the first author: mood, energy/activity and other symptoms. The mood symptom cluster included changes related not only to sadness but also to anxiety: depressed mood, feelings of guilt, suicide, psychic anxiety, somatic anxiety, and hypochondriasis. Hypochondriasis was included in this symptom cluster because it is especially related to worry and anxiety. Work and activities, retardation, agitation, general somatic symptoms, and genital symptoms were included in the energy/activity symptom cluster.
General somatic symptoms and genital symptoms were considered energy/activity symptoms because "loss of energy and fatigability" and "loss of libido" are rated in these items, respectively. Finally, the remaining items were classified as "other symptoms," namely, insomnia (three items), gastro-intestinal somatic symptoms, loss of weight and insight.

Statistical analyses
In descriptive statistics, the stage of disease in each patient was considered as reference: mania, euthymia, and depression. Mean age and standard deviation (SD), and total scores of the scales were calculated for each stage. Item response theory (IRT) analyses were conducted to understand the levels of discrimination and information of symptoms in relation to levels of severity in each stage. For that, depressive, euthymic, and manic stages were equated together through the one-step method 16  IRT is a set of statistical models that assume logistic distribution of samples through a latent trait.
In psychiatric research, IRT models are quite useful because they help researchers understand symptom profiles (item error and information/sample distribution) throughout levels of severity (latent trait) of a specific psychopathology. 19 In this study, the generalized partial credit model (GPCM), 20 a two-parameter model for polytomous responses with different numbers of categories was adopted to analyze items and test discrimination, difficulty, and information.
GPCM discrimination index (a) reveals how much separation an item or cluster of items is able to establish between individuals supposed to show different results. i.e., that one cluster of symptoms is prevalent when compared with the other two, whereas a non-significant chi-square indicated no prevalence of a specific set. All analyses were conducted using R software with the following packages: LTM, mIRT and psych.

Results
Among the 243 patients attending the BD outpatient clinic, 106 were selected, because they had experienced at least one manic episode, one depressive episode and one euthymic period over the 13 years covered by the study. They were followed on average for 5.5 years (SD=3.1). Among the 106 participants, 74 were women (69.8%). In their last evaluation, the mean age (X) of the participants was 52.5 years (SD=11.7). Regarding  The invariance test conducted through MGCFA showed that metric invariance was the best solution to explain the covariance matrix of the one-factor structure in HAM-D, i.e., HAM-D was reliable to assess depression through different episodes. Table 1 reveals fit and error indexes found for the unidimensional model as well as invariance levels.
The GPCM could be performed because the unidimensional solution was acceptable, also showing good invariance across clusters. Table 2 depicts GPCM estimates discrimination (a) and difficulty (θ) for each one of the 18 items of HAM-D.    However, during euthymic states, the curves depicting energy/activity symptoms seem to fade, and mood symptoms become most informative. In less severe depressive states, mood symptoms inform more than energy/activity symptoms. This means that when a patient has less severe depression cycling to or from euthymia, he/she tends to have more mood symptoms than energy/activity symptoms or other. During euthymic states, those mood symptoms are prevalent across the whole level of severity. In higher levels of manic episodes, mood symptoms also appear to be the most discriminative cluster of symptoms.
The chi-square test revealed prevalence of energy symptoms in both depression and mania episodes (χ²=3.29, p<0.05, and χ²=4.20, p<0.05, respectively). However, in mania, only energy/activity symptomsi.e., general somatic symptoms and retardation -were among the most informative items.
Considering the classification of items, in both depression and mania, the energy/activity cluster was more informative than the mood cluster according to the IRT analysis.
In a previous study, 21 we had already found that energy/activity symptoms are more relevant than mood symptoms in BD. A sample of 118 hospitalized patients in mania were assessed with a six-item mania subscale of Schedule for Affective Disorders and Schizophrenia-Changed version (SADS-C). 22 In a confirmatory factor analysis, increased energy was the item with the highest factor loadings. Moreover, an IRT analysis showed that symptoms related to energy were more informative about mania severity than those related to mood.
According to several factor analysis studies, hyperactivity is the core symptom of mania. For example, Bauer et al. 8

Conclusion
Our data reinforce the view of hyperactivity and motor retardation as cardinal changes of mania and depression, respectively.

Disclosure
No conflicts of interest declared concerning the publication of this article.