Journal of Affective Disorders 101 (2007) 35–41www.elsevier.com/locate/jad

Research report

Towards a reconceptualization of mixed states, based on anemotional-reactivity dimensional model

Chantal Henry⁎, Katia M'Baïlara, Alain Desage, Sébastien Gard,David Misdrahi, Eduard Vieta

Hôpital Charles Perrens, Bâtiment Lescure, 121 rue de la Béchade, 33076 Bordeaux Cedex, FranceLaboratoire de Psychologie Clinique et Psychopathologie JE 2358, 3ter place de la Victoire,

Université Victor Segalen, Bordeaux 2, Bordeaux, FranceBipolar Disorders Program, Hospital Clinic, IDIBAPS, Clinical Institute of Neuroscience,

University of Barcelona, Villarroel 170/Rossello 140, 8036 Barcelona, Spain

Received 29 October 2006; accepted 30 October 2006Available online 22 January 2007

Abstract

Background: DSM-IV criteria for mixed states may be too restrictive and may actually exclude patients who do not meet the fullcriteria for a manic and depressive state. Using this DSM-IV definition, many patients who are considered depressed may havemixed features, which can explain why some bipolar depressive states can worsen with antidepressants and can be improved bymood stabilizers or atypical antipsychotics. A dimensional approach not exclusively focused on the tonality of affect would help todefine a broader entity of mixed states. The aim of this study was to apply a dimensional model to bipolar episodes and to assessthe overlap between the groups defined using this model and using categorical diagnosis.Method: We assessed 139 DSM-IV acutely ill bipolar I patients with MAThyS (Multidimensional Assessment of Thymic States byHenry et al. in press), a scale that assesses five quantitative dimensions exploring excitatory and inhibition processes, and that is notfocused on tonality of mood but on emotional reactivity. We studied the relationship between clusters defined by statistical analysesand DSM-IV bipolar mood states.Results: This study showed the existence of three clusters. Cluster 1 was characterized by an inhibition in all dimensions andcorresponded to the depressive cluster (more than 90% of patients met the criteria for DSM-IV Major Depressive Episode (MDE)).Cluster 2 showed a general excitation and was mainly DSM-IV manic or hypomanic patients (90%). Cluster 3 (Mixed) was morecomplex and the diagnosis included MDE (56%) in most of the cases associated with manic or hypomanic symptoms, mixed states(18%) defined by DSM-IV criteria, and manic or hypomanic states (25%). Emotional reactivity was relevant to distinguish Cluster1 (Depressive), exhibiting emotional hypo-reactivity, from Cluster 2 (Manic) and 3 (Mixed), characterized by emotional hyper-reactivity. Sadness was reported equally in all three clusters.Conclusion: A dimensional approach using the concept of emotional reactivity seems appropriate to define a broad mixed stateentity in patients who would be diagnosed with MDE according to DSM-IV. Further studies are needed to test the relevance of thismodel in therapeutic strategies.© 2006 Elsevier B.V. All rights reserved.

Keywords: Bipolar disorders; Mixed states; Major depressive episode; Emotional reactivity; Treatment

⁎ Corresponding author. Tel.: +33 5 56 56 34 50; fax: +33 5 56 56 17 14.E-mail address: chenry@perrens.aquisante.fr (C. Henry).

0165-0327/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.jad.2006.10.027

36 C. Henry et al. / Journal of Affective Disorders 101 (2007) 35–41

1. Introduction

Manic and depressive states are not mutually exclusiveand their combination defines the mixed states. Mixedstates are common, difficult to treat, potentially severe,and may carry a high risk of suicide (Dilsaver et al., 1994;Vieta, 2005a). However, mixed states are very heteroge-neous entities and this broad diversity is illustrated by thereported rates among manic patients, from 14 to 67%(Akiskal et al., 1998). For Kraepelin (1913), the conceptof mixed state was broader than that defined in the currentclassifications (for the evolution of the mixed statesconcept see Marneros, 2001). Kraepelin described 6mixed states, frommanic states with dysphoric features todepressive states with some agitation. Currently, someauthors use the concept of “mixed depressive state” or“agitated depression” to describe “soft” mixed statesdefined by a depressive state associated with manic orhypomanic symptoms (Koukopoulos and Koukopoulos,1999; Benazzi, 2002; Akiskal et al., 2005). However,there is no definition in the current classifications thatdistinguish this “soft” mixed state. Suppes et al. (2005)have recently identified mixed hypomania as a commontype of episode within the bipolar spectrum. Akiskal andBenazzi (2005) also showed the existence of dysphorichypomania and discussed the interest of categorial anddimensional conceptualizations of a broad spectrum ofmixed states. Therefore, to meet the criteria for a mixedstate using the DSM-IV, a patient should have a full manicstate associatedwith a depressive episode.More precisely,to fulfill the criteria for a manic episode the moodmust beexalted and this mood state must be associated with threesymptoms, or four if the mood is irritable. If a patient metthe criteria for a depressive episode in combination withan irritable mood but if this were associated with onlythree symptoms of mania, he would not meet the criteriafor a mixed state. He would only meet the criteria for adepressive episode. Thus, a single symptom is sufficientto shift from one affective state to another, along with thecorresponding change in treatment strategy. A categorialapproach does not allow for the classification of all thesestates on a continuum. As suggested by some authors(McElroy et al., 1992, 1995; Bauer et al., 1994), adimensional model could be helpful to better characterizemood states in bipolar disorders and more particularly,those with an admixture of manic and depressive features.In 1997, Perugi and colleagues operationalized newmixed state criteria in part based on the concept ofemotional instability and they showed that mixed stateswere more frequent than expected using the DSM-IV.

Based on the Kraepelinian concept of mood states,we built a tool called MAThyS (Multidimensional

Assessment of Thymic States) to define mood statesas the function of a dimensional approach (Henry et al.,in press). Kraepelin defined mood states as starting fromexcitement or inhibition of the three domains of psychiclife: cognitive processes (train of thought rather than itscontents), mood, and volition (expressed in psychomo-tor activity). We extended this by changing the tonalityof mood (sadness vs. euphoria) for emotional reactivity(hypo- vs. hyper-reactivity), which is closer to theconcept of dimensions. An emotion is characterized byits tonality (pleasant/unpleasant) and also by its intensity(the arousal) (Lang, 1995).

The dimensions assess inhibitory or excitatory pro-cesses; therefore they can be applied to manic or depres-sive states and to states presenting with an admixture ofboth. Emotional reactivity may be considered as a newcomponent in the comparison of current mood scales.The goal of this dimension is to discriminate mood statesas a function of emotional reactivity rather than basedexclusively on the tonality of mood (euphoria/depressivemood).

The aims of this study were: 1) to conduct a clusteranalysis using a dimensional approach in order to ex-plore the clinical heterogeneity of bipolar states, 2) toassess the relationship between the different clusters andthe diagnoses using the DSM-IV criteria, 3) to describethe dimensional profile of each cluster and moreparticularly to test the relevance of emotional reactivityto define mood states, and particularly mixed states.

Our hypothesis was that mixed states correspond to abroader entity than those defined in DSM-IV and that adefinition based on the tonality of mood may fail tocapture this heterogeneity, whereas a definition based onemotional reactivity may be more appropriate. Thevalidity of a broader definition of mixed states may havetherapeutic implications because episodes currentlyconsidered as depressive states may actually qualify tobe mixed states, and may respond well to primarily anti-manic treatment, and poorly to antidepressants.

2. Methods

2.1. Patients

Bipolar patients (n=139) were recruited fromconsecutive inpatient admissions in a general psychiatryunit, and from outpatients in a specific consultation forbipolar disorders (Charles Perrens Hospital, Bordeaux).

Patients were interviewed by a trained psychologistusing the French version of the Diagnosis Interview forGenetic Studies (Nurnberger et al., 1994) for DSM-IVdiagnosis (APA, 1994). We focused on the part assessing

37C. Henry et al. / Journal of Affective Disorders 101 (2007) 35–41

diagnosis of mood disorders. All patients enrolled in thisstudy met the criteria for bipolar disorder. Patients wererated before starting any pharmacological treatment forthe current episode.

2.2. Assessments

The severity of the mood episode was quantified withboth theMontgomery andAsbergDepressionRating Scale(MADRS) (Montgomery and Asberg, 1979) and the Bechand Rafaelsen manic scale (MAS) (Bech et al., 1978).

The dimensional scale MAThyS (MultidimensionalAssessment of Thymic States) (Henry et al., in press) is avisual analogical scale to explore mood episodes usingquantitative dimensions. It assesses five a priori dimen-sions that can fluctuate from inhibition to excitation. It isclose to the model proposed by Kraepelin, who definedmood states as starting from the excitement or inhibitionof the three domains of psychic life. In this model, asopposed to the two other dimensions, mood was notdefined by a quantitative dimension that could be de-scribed by a continuum.

In MAThyS, the five dimensions are: emotionalreactivity, cognition speed, psychomotor activation,motivation and sensory perception. Emotions are notdefined as a function of their “tonality” (e.g., euphoria,irritability, sadness, anxiety), but as quantitative dimen-sions defined by the intensity of emotional reactivity,thus linking environmental stimuli with emotionalreactivity. Manic and mixed states may be better definedby emotional hyper-reactivity, meaning that patients feelemotions with a greater intensity and depending on theenvironmental context, rather than by the tonality ofaffect, which tends to fluctuate (Henry et al., 2003). Allthe other dimensions (motricity, motivation, cognitiveprocesses, sensory perception) are more common andmay also fluctuate from inhibition to excitation.

The scale is composed of 20 items (3 to 5 for eachdimension) and also assesses tonality of affect during theprevious week with a Likert Scale (5 frequencies: neverto constantly). It gives a total score (0 to 200), and 5scores for each dimension. The higher the score, thegreater the excitation and vice versa.

The validation of the scale was done on anindependent sample of 196 subjects, including controlsand bipolar patients during euthymic, manic/hypomanicand depressive states. The confirmatory and exploratoryanalysis showed: good face validity and acceptability(the participation rate was 96.4% in the depressive groupand 86.3% in the manic group and all respondentscompleted all items on the scale (no missing data), whichsupports the acceptability of the instrument, and suggests

that it was well understood and easy to complete), goodconvergent validity, good internal consistency betweensubscales (Cronbach alpha coefficients N0.70), lowinter-item redundancy, and moderate linkage betweenMAThyS and the MADRS (r=−0.45) and the MAS(r=0.56) (Henry et al., in press).

2.3. Statistical analysis

Statistical analyses were done with SPSSV.12.0, usinghierarchical cluster analysis (HCA), which is an explor-atory tool designed to reveal natural groupings (orclusters) that would otherwise not be apparent within adata set. This method is recommended for clustering asmall number (less than a few hundred) of objects.Hierarchical cluster analysis begins by separating eachobject into a cluster by itself. At each stage of the analysis,the criterion by which objects are separated is relaxed inorder to link the two most similar clusters until all of theobjects are joined in a complete classification tree. Thebasic criterion for any clustering is distance. Objects thatare near each other should belong to the same cluster, andobjects that are far from each other should belong todifferent clusters. This method gives a dendrogram, inwhich the horizontal axis shows the distance betweenclusters when they are joined and the vertical axis showsthe objects. In our study, the objects were the bipolarpatients and the criteriawere the answers to the 20 items ofthe visual analogic scale. Other analyses were descriptiveor the means were compared using Chi2 or ANOVA.

3. Results

3.1. Sample characteristics

The sample of 139 bipolar patients was composed of38 (27.3%) men and 101 (72.7%) women, with a meanage at interview of 40.05 (±11.76) years old; (range 14–65 years old).

Forty-two (30.2%) patients had always been single,64 (46%) patients were married, and 33 (23.7%) weredivorced or widowed. Most patients presented type Ibipolar disorder 78 (56.1%).

3.2. Cluster analysis

A cluster analysis using MAThyS items identifiedthree clusters among all groups of bipolar patients.Clusters 2 and 3 had more in common with each otherthan with Cluster 1.

The main socio-demographic and clinical character-istics of the three clusters are represented in Table 1.

Table 1The main socio-demographic and clinical characteristics of the threeclusters

Cluster 1 Cluster 2 Cluster 3 Difference

Number (%) 42(30.22%)

21(15.11%)

76(54.68%)

Age (mean)(±SD)

40.95(±11.68)

38.76(±14.06)

39.91(±11.23)

F(2,138)=0.252;p=0.777

Gender(male) n (%)

13 (31%) 6(28.6%)

19 (25%) Chi2=0.5, ddl=2,p=0.778

BP I n (%) 21 (50%) 19(90.5%)

38 (50%) Chi2=9.71,ddl=2, p=0.008

Suicide attempt 5 (11.9%) 0 (0%) 21(27.6%)

Chi2=10.09,ddl=2, p=0.006

Fig. 2. A general representation of the correspondence between thethree clusters and DSM-IV diagnosis.

38 C. Henry et al. / Journal of Affective Disorders 101 (2007) 35–41

There was no difference in gender (Chi2 =0.5, ddl=2,p=0.778) and mean age (F(2,138)=0.252; p=0.777)between the three clusters. There was more BP I (90%)in Cluster 2 than in the two other clusters (Chi2 =9.71,ddl=2, p=0.008). The incidence of suicide attemptsduring the current episode was different among the threeclusters with a higher rate in Cluster 3 (27.6%)(Chi2 =10.09, ddl=2, p=0.006).

The score on the MADRS was statistically differentin each cluster (F(2,136)=52.29; pb0.0001) (Fig. 1).Depressive symptomatology was more severe in Cluster1 (26.52±9.8) than in the two other clusters (Cluster 2:7.5±4.2) (Cluster 3: 13.36±7.5). The score on themanic scale was statistically different in each cluster (F(2,136)=55.31; pb0.0001). Manic symptomatologywas the highest in Cluster 2 (18.75±8.3) (Cluster 3:9.24±5.5) (Cluster 1: 2.29±5.0).

3.3. Relationship between the three clusters andDSM-IV diagnosis

Fig. 2 shows a general representation of thecorrespondence between the three clusters and DSM-IV diagnoses. In Cluster 1, the majority of the patientsmet the criteria for a Major Depressive Episode; in

Fig. 1. Comparison between the three clusters on the depressive scale(MADRS) and the manic scale (MAS).

Cluster 2, more than 90% suffered from manic orhypomanic episodes and the rest was represented bymixed episodes. However, Cluster 3 was more complex.

Table 2 reports the proportion of DSM-IV moodepisodes within the three clusters. Cluster 1 was clearlyassociated with the diagnosis of major depressiveepisode. Cluster 2 was represented essentially bymanic, hypomanic and some mixed episodes. However,what we consider the mixed cluster is more complexthan expected. Only 18% of patients met the criteria formixed episodes as defined by DSM-IV. However, 56%of patients met the criteria for a Major DepressiveEpisode, and among them the majority also had manic

Table 2DSM-IV categories within the three clusters

DSM-IV diagnosis Cluster 1(N=42)

Cluster 2(N=21)

Cluster 3(N=76)

MDE 92.8% 0% 56.6%- pure MDE −83.3% −0% −9.2%- MDE (with manicsymptoms)

−9.5% −0% −47.4%

Mixed episode 7.2% 9.5% 18.4%(Hypo)manic episode 0% 90.5% 25%

Fig. 4. Tonality of emotions reported by patients during the previousweek.

39C. Henry et al. / Journal of Affective Disorders 101 (2007) 35–41

symptoms associated with the depressive episode.Twenty percent had a manic or hypomanic state.

3.4. Characteristics of the three clusters using MAThyS

The results revealed a significant difference betweenthe three groups in the total MATHyS scores [Cluster 1:mean=51 (±21.22) range: 3.5–94 ; Cluster 2: mean=165.79 (±16.41) range: 134.5–193; Cluster 3: mean=118.19 (±20.02) range: 72.5–159.5] (F(2,138)=269.12;pb0.0001) (Fig. 3), and in each dimension: emotionalreactivity (F(2,138)=243.99; pb0.0001), cognition (F(2,138)=91.97; pb0.0001), psychomotor activation (F(2,138)=71.91; pb0.0001), motivation (F(2,138)=122.47; pb0.0001) and sensory perception (F(2,138)=66.77; pb0.0001).

The first cluster clearly exhibited an inhibition in alldimensions as shown by the comparison with the meanobtained in control subjects. Conversely, Cluster 2 showedan excitation in all dimensions. In the third cluster,emotional hyperreactivity was the most salient dimensionin comparison to themean (t (15.4; ddl=163; pb0.0001)).

3.5. Tonality of emotions assessed using MAThyS

Fig. 4 shows the comparison in the tonality ofemotions reported during the previous week between thethree clusters. The frequency of all emotions wasstatistically different between the three clusters with ahigher frequency in Cluster 2 (Manic) and Cluster 3(Mixed) (Cluster 1: 7.5 (±3.6); Cluster 2: 13.67 (±4.7);Cluster 3: 11.83 (±4.8); F(2;138)=17.88; pb0.0001).

Sadness is the single emotion reported in the sameway in the three clusters (Cluster 1: 2.29 (±1.3); Cluster2: 2.05 (±1.3); Cluster 3: 2.30 (±1.2); F(2;138)=0.37;p=0.69). Other tonalities of affect were not so similarin the three clusters: joy (Cluster 1: 0.50 (±0.6); Cluster2: 2.00 (±1.14); Cluster 3: 1.12 (±0.7); F(2;138)=30.03;

Fig. 3. Characteristics of the 5 dimensions using MAThyS in the threeclusters.

pb0.0001), irritability (Cluster 1: 0.93 (±1.0); Cluster 2:2.24 (±1.3); Cluster 3: 1.92 (±1.2); F(2;138)=12.66;pb0.0001); panic (Cluster 1: 1.04 (±1.1); Cluster2: 1.19 (±1.3); Cluster 3: 1.62 (±1.3); F(2;138)=3.19;p=0.044); anxiety (Cluster 1: 1.88 (±1.1); Cluster2: 2.14 (±1.2); Cluster 3: 2.5 (±1.1); F(2;138)=4.18;p=0.017); anger (Cluster 1: 0.67 (±0.9); Cluster 2: 1.81(±1.2); Cluster 3: 1.34 (±1.2); F(2;138) = 8.84;pb0.0001); and exaltation (Cluster 1: 0.19 (±0.5); Clus-ter 2: 2.24 (±1.1); Cluster 3: 1.03 (±1.1); F(2;138)=31.07; pb0.0001).

4. Discussion

Our study suggested that mood states in bipolardisorders were distributed into three clusters when de-fined by a dimensional approach based on excitatory andinhibition processes. Cluster 1 was characterized by aglobal inhibition and may correspond to pure depressiveepisodes, also traditionally described as endogenous de-pression with a general retardation. Cluster 2 representedthe manic pole and exhibited a general excitation. Clus-ter 3 ranged from depressive episodes to manic episodesusing the DSM-IV criteria, and could represent mixedstates in a broader definition than DSM-IV.

The first conclusion is that the DSM-IV definition formixed states seemed too restrictive. Requiring thecombination of a full depressive and manic state, theDSM-IV excluded patients exhibiting either a manicepisode associated with depressive symptoms or a majordepressive episode associated with manic and hypo-manic symptoms. Our results showed that mixedepisodes were not unique conditions and were closerto the concept of mixed states as defined by Kraepelin,who described six mixed states.

The advantage of a dimensional approach based onexcitatory and inhibition processes is that all affective

40 C. Henry et al. / Journal of Affective Disorders 101 (2007) 35–41

states can be represented on a continuum. This model iscomplementary to the categorial approach as suggestedby Dayer et al. (2000). These authors also proposed adimensional approach to define mixed states, based onthree dimensions: depressive, manic and dysphoricdimensions. However, this approach still seems tooclose to categorial definitions because they are mainlybased on the tonality of mood.

Using factor analysis of rating scales, a few studieshave proposed new representations of mixed states(Dilsaver et al., 1999; Cassidy et al., 1998; Sato et al.,2002). Generally, these studies have shown a bimodaldistribution of depressive factors, supporting the ideathat mixed states are not a unique condition. However,the factor analyses were done using tools based oncategorial definitions and thus on the tonality of mood asthe main criteria. A recent study illustrating the greatdifficulty in defining episodes with mixed features is thatby Suppes and collaborators (2005) on mixed hypoma-nia. This study highlighted the frequency of depressivesymptoms co-occurring with hypomanic symptoms,with the co-occurrence defining a mixed hypomania.To define this new syndrome (DSM-IV does not includemixed hypomania), two scales were used: a manic scaleand an inventory of depressive symptomatology. How-ever, it is generally necessary to use diagnostic criteria todefine a syndrome and rating scales are only indicators ofthe severity of the episodes. This apparent weakness ispossibly due to the impossibility of patients havingdepressive features to respond to the main criteria for ahypomanic episode: an exalted, expansive or irritablemood. It means that in the case of mixed hypomania, thetonality of mood is not the best choice to define this state,or any mixed state. The dimensional model may addressthis problem. Therefore, the concept of emotionalreactivity is useful to describe mixed states. Emotionalhyper-reactivity refers to the fact that all emotions can befelt with an unusual intensity. As a function of the en-vironmental context, or linked to temperamental fea-tures, the tonality of mood may be associated with a highlability or with a main euphoric or dysphoric coloratura.Conversely, emotional hypo-reactivity may describe aspecific type of depression characterized by a generalinhibition in all dimensions. The distinction betweenhypo-reactive depression and a type of depression as-sociated with emotional hyper-reactivity may be usefulto understand responses to treatment. Hence, depressivemixed states defined by the co-occurrence of a MajorDepressive Episode and at least three hypomanic symp-toms (Benazzi, 2002), and mixed hypomania may beconsidered as the same kind of episode: “a soft mixedstate”, and both depressive mixed states and mixed hy-

pomania may require the same pharmacological strategy.“Soft”, however, may not necessarily be the most appro-priate adjective because in Cluster 3 (Mixed) the inci-dence of suicide attempts was particularly high (morethan twice that in Cluster 1 (Depressive)). Emotionalhyper-reactivity associated with depressive moodseemed to be associated with a higher risk of suicideattempts.

The emotional reactivity sub-score on MAThyS waslinked to the frequency of all emotions reported by thepatients. Regarding tonality of mood, sadness was theonly affect reported in the same manner in all threeclusters. However, as expected, joy and exaltation weremore frequent in Cluster 2 (Manic), while irritability andanger were highly reported in Cluster 2 (Manic) andCluster 3 (Mixed). Anxiety and panic were morefrequent in Cluster 3 (Mixed) than in Cluster 1(Depressed). The higher incidence of panic and anxietyin the mixed cluster is consistent with reports showing agreater load of comorbidity and suicidality in patientswith mixed features (Vieta et al., 2001).

How it is possible to explain that sadness wasreported with the same frequency in all three clusters?The MAThyS scale was not created to discriminatemood episodes as a function of the tonality of mood.Thus, at the end of the dimension assessment, patientswere asked to report the frequency of each emotion inthe previous week. Therefore, patients could reportsadness and joy very frequently or never, but one wasnot exclusive for the other and we did not ask thepatients to determine which one was the most frequent.

Our results suggested that a dimensional approachcould be useful to improve the definition of mood states.Emotional reactivity, a dimension usually neglected,seems to be a fundamental component to discriminatemood states. The tonality of mood is only an aspect ofemotion, but emotions are better defined by their valence(pleasant/unpleasant) and their intensity (Lang, 1995).Therefore, it seems judicious to add this dimension todefine affective disorders.

Using this dimensional model, major depressiveepisodes with an emotional hyper-reactivity couldrepresent a mixed state where antidepressants may notbe indicated. Whether or not these states are classifiedamong depressive states, they can be considered asresistant to antidepressant treatment. Antipsychotics thathad been reported to improve bipolar depression, such asolanzapine (Tohen et al., 2003) and quetiapine (Calabr-ese et al., 2005), may work not only because of theirputative antidepressant action (Yatham et al., 2005), butalso because of their action on manic and mixed featuresarising from a major depressive episode (Vieta, 2005b).

41C. Henry et al. / Journal of Affective Disorders 101 (2007) 35–41

Further studies are needed to explore the response totreatment as a function of this dimensional model. Thismodel could be particularly appropriate to assess ad-olescent depressions, which often belong to mixed states(Dilsaver et al., 2005) but are unrecognized and treatedinappropriately by antidepressants.

References

Akiskal, H.S., Benazzi, F., 2005. Toward a clinical delineation ofdysphoric hypomania — operational and conceptual dilemmas.Bipolar Disord. 7 (5), 456–464.

Akiskal, H.S., Hantouche, E.G., Bourgeois, M.L., Azorin, J.M.,Sechter, D., Allilaire, J.F., Lancrenon, S., Fraud, J.P., Chatenet-Duchene, L., 1998. Gender, temperament, and the clinical picturein dysphoric mixed mania, findings from a French national study(EPIMAN). J. Affect. Disord. 50, 175–186.

Akiskal, H.S., Benazzi, F., Perugi, G., Rihmer, Z., 2005. “Unipolar”depression re-conceptualized as a depressive mixed state:implications for the antidepressant-suicide controversy. J. Affect.Disord 85, 245–258.

Bauer, M.S., Whybrow, P.C., Gyulai, L., Gonnel, J., Yeh, H.S., 1994.Testing definitions of dysphoric mania and hypomania: preva-lence, clinical characteristics and inter-episode stability. J. Affect.Disord. 32 (3), 201–211.

Bech, P., Rafaelsen, O.J., Kramp, P., Bolwig, T.G., 1978. The maniarating scale: scale construction and inter-observer agreement.Neuropharmacology 17, 430–431.

Benazzi, F., 2002. Depressive mixed state frequency: age/gendereffects. Psychiatry Clin. Neurosci. 56 (5), 537–543.

Calabrese, J.R., Keck Jr., P.E., Macfadden, W., et al., 2005. Arandomized, double-blind, placebo-controlled trial of quetiapine inthe treatment of bipolar I or II depression. Am. J. Psychiatry 162(7), 1351–1360.

Cassidy, F., Forest, K., Murry, E., Carroll, B.J., 1998. A factor analysisof the signs and symptoms of mania. Arch. Gen. Psychiatry 55,27–32.

Dayer, A., Aubry, J.M., Roth, L., Ducrey, S., Bertschy, G., 2000. Atheoretical reappraisal of mixed states: dysphoria as a thirddimension. Bipolar Disord. 2 (4), 316–324.

Dilsaver, S.C., Chen, Y.W., Swan, A.C., Shoaib, A.M., Krajewski, K.J.,1994. Suicidality in patients with pure and depressivemania. Am. J.Psychiatry 151, 1312–1315.

Dilsaver, S.C., Chen, Y.W., Shoaib, A.M., Swann, A.C., 1999.Phenomenology of mania: evidence for distinct depressed,dysphoric, and euphoric presentations. Am. J. Psychiatry 156,426–430.

Dilsaver, S.C., Benazzi, F., Rihmer, Z., Akiskal, K.K., Akiskal, H.S.,2005. Gender, suicidality and bipolar mixed states in adolescents.J. Affect. Disord. 87, 11–16.

Henry, C., Swendsen, J., Van Den Bulke, D., Sorbara, F., Leboyer, M.,2003. Emotional hyperreactivity as the fundamental mood charac-teristic of manic and mixed states. Eur. Psychiatr. 18, 124–128.

Henry, C., M'baïlara, K., Poinsot, R., Falissard, B. Construction etvalidation d’une échelle dimensionnelle de l’humeur : La MAThyS(Multidimensional Assessment of Thymic States). In press inl’Encéphale.

Koukopoulos, A., Koukopoulos, A.E., 1999. Agitated depression as amixed state and the problem of melancholia. Bipolarity: beyondclassic mania. Psychiatr. Clin. North Am. 22 (3), 564–574.

Kraepelin, E., 1913. La folie maniaco-depressive, french translation byGeraud M, Bordeaux, Mollat Eds.

Lang, P.J., 1995. The emotion probe: studies of motivation andattention. Am. Psychol. 50, 372–385.

Marneros, A., 2001. Origin and development of concepts of bipolarmixed states. J. Affect. Disord. 67, 229–240.

McElroy, S.L., Keck, P.E., Pope, H.G., Hudson, J.F., Faedda, G.L.,Swann, A.C., 1992. Clinical and research implications of thediagnosis of dysphoric or mixed mania or hypomania. Am. J.Psychiatry 149 (12), 1633–1644.

McElroy, S.L., Strakowski, S.M., Keck Jr., P.E., Tugrul, K.L., West,S.A., Lonczak, H.S., 1995. Differences and similarities in mixedand pure mania. Compr. Psychiatry 36 (3), 187–194.

Montgomery, S., Asberg, A., 1979. A new depression scale designedto be sensitive to change. Br. J. Psychiatry 134, 382–389.

Nurnberger Jr., J.I., Blehar, M.C., Kaufmann, C.A., et al., 1994.Diagnostic interview for genetics studies. Rationale, uniquefeatures, and training. NIMH Genetics Initiative. Arch. Gen.Psychiatry 51 (11), 849–859.

Perugi, G., Akiskal, H.S., Micheli, C., Musetti, L., Paiano, A., Quilici,C., Rossi, L., Cassano, G.B., 1997. Clinical subtypes of bipolarmixed states: validating a broader European definition in 143cases. J. Affect. Disord. 43, 169–180.

Sato, T., Bottlender, R., Kleindienst, N., Möller, H.J., 2002.Syndromes and phenomenological subtypes underlying acutemania: a factor analytic study of 576 manic patients. Am. J.Psychiatry 159, 968–974.

Suppes, T., Mintz, J., McElroy, S.L., et al., 2005. Mixed hypomania in908 patients with bipolar disorder evaluated prospectively in theStanley Foundation Bipolar Treatment Network: a sex-specificphenomenon. Arch. Gen. Psychiatry 62 (10), 1089–1096.

Tohen, M., Vieta, E., Calabrese, J., Ketter, T.A., Sachs, G., Bowden,C., Mitchell, P.B., Centorrino, F., Risser, R., Baker, R.W., Evans,A.R., Beymer, K., Dube, S., Tollefson, G.D., Breier, A., 2003.Efficacy of olanzapine and olanzapine-fluoxetine combination inthe treatment of bipolar I depression. Arch. Gen. Psychiatry 60(11), 1079–1088.

Vieta, E., 2005a. The treatment of mixed states and the risk ofswitching to depression. Eur. Psychiatr. 20, 96–100.

Vieta, E., 2005b. Bipolar mixed states and their treatment. Expert Rev.Neurotherapeutics 5 (1), 1–6.

Vieta, E., Colom, F., Corbella, B., Martinez-Aran, A., Reinares, M.,Benabarre, A., Gasto, C., 2001. Clinical correlates of psychiatriccomorbidity in bipolar I patients. Bipolar Disord. 3, 253–258.

Yatham, L.N., Goldstein, J.M., Vieta, E., Bowden, C.L., Grunze, H.,Post, R.M., Suppes, T., Calabrese, J.R., 2005. Atypical anti-psychotics in bipolar depression: potential mechanisms of action.J. Clin. Psychiatry 66 (supp 5), 40–48.