Group-based trajectory modeling A novel approach to ...

Journal of Affective Disorders ] (]]]]) ]]]–]]]

Contents lists available at SciVerse ScienceDirect

Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad

Research report

Group-based trajectory modeling: A novel approach to examining symptom trajectories in acute bipolar episodes Katia M’Bailara a,b,i,n, Olivier Cosnefroy c, Eduard Vieta d, Jan Scott e,g,i, Chantal Henry f,g,h,i a

University Bordeaux, Psychologie, Sante´ et Qualite´ de vie, EA4139, F-33000 Bordeaux, France Department of Psychiatry Adulte, CHS Charles Perrens, Bordeaux, France c University Pierre Mende s France, Grenoble, France d Bipolar Disorders Program, Hospital Clinic, University of Barcelona, IDIBAPS, CIBERSAM, Barcelona, Spain e Academic Psychiatry, Institute of Neuroscience, Newcastle University, NE1 4LP, UK f INSERM, Unite´ U955, Cre´teil F-94010, France g University Paris-Est, Faculte´ de Me´decine, UMR-S 955, Cre´teil F-94010, France h AP-HP, Groupe Henri-Mondor Albert-Chenevier, Pˆ ole de psychiatrie, Cre´teil F-94010, France i Fondation Fondamental, Cre´teil F-94010, France b

a r t i c l e i n f o

abstract

Article history: Received 7 June 2012 Received in revised form 1 July 2012 Accepted 2 July 2012

Background: Pattern analysis can aid understanding of trajectories of symptom evolution. However, most studies focus on relatively homogeneous disorders with a restricted range of outcomes, prescribed a limited number of classes of medication. We explored the utility of pattern analysis in defining shortterm outcomes in a heterogeneous clinical sample with acute bipolar disorders. Method: In a naturalistic observational study, we used Group-based trajectory modeling (GBTM) to define trajectories of symptom change in 118 bipolar cases recruited during an acute DSM IV episode: major depression (56%), (hypo)mania (26%), and mixed states (18%). Symptoms were assessed weekly for a month using the MATHYS, which measures symptoms independent of episode polarity. Results: Four trajectories of symptom change were identified: Persistent Inhibition, Transient Inhibition, Transient Activation and Over-activation. However, counter to traditional predictions, we observed that bipolar depression shows a heterogeneous response pattern with cases being distributed approximately equally across trajectories that commenced with inhibition and activation. Limitations: The observational period focuses on acute outcomes and so we cannot use the findings to predict whether the trajectories lead to stable improvement or whether the clinical course for some clusters is cyclical. As in all GBTM, the terms used for each trajectory are subjective, also the modeling programme we used assumes dropouts are random, which is clearly not always the case. Conclusion: This paper highlights the potential importance of techniques such as GBTM in distinguishing the different response trajectories for acutely ill bipolar cases. The use of the MATHYS provides further critical insights, demonstrating that clustering of cases with similar response patterns may be independent of episodes defined by mood state. & 2012 Elsevier B.V. All rights reserved.

Keywords: Bipolar disorder Mood Trajectory Inhibition Activation Acute episode-MATHYS

1. Introduction The selection of optimum treatments for the management of bipolar disorders (BD) is often undermined by the heterogeneity of clinical presentations. A further level of complexity is the variability in intra-episode BD symptoms and the range and nature of sub-syndromal manifestations of the disorder. For example, depressive episodes may present with retarded, melancholic, agitated or atypical features and manic patients may n Corresponding author at: Laboratoire de psychologie, Sante´ et Qualite´ de vie EA413, Universite´ de Bordeaux, 3ter place de la Victoire, 33076 Bordeaux, France. Tel.: þ0033 5 57 57 30 27; fax: þ 0033 5 57 57 19 77. E-mail address: [email protected] (K. M’Bailara).

present as elated, dysphoric, labile, paranoid, etc. (Cassidy et al., 1998). Furthermore, whilst mixed states are defined be the presence of syndromal levels of depressive and manic symptoms, patients not meeting full ‘mixed state’ criteria frequently present concurrently with sub-syndromal symptoms of one pole of BD alongside syndromal symptoms of the opposite pole (i.e. depressive symptoms during hypomanic episodes and vice versa) (Benazzi, 2007; Henry et al., 2010). In order to better reflect this heterogeneity of BD, DSM V proposes a ‘specifier’ of mixed symptoms for manic, hypomanic and depressive episodes (http://www.dsm5.org/ProposedRevisions/). Other researchers have suggested that activation rather than mood state may be a more appropriate means of defining different BD presentations (Angst, 2011, Angst et al., 2010).

0165-0327/$ - see front matter & 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jad.2012.07.007

Please cite this article as: M’Bailara, K., et al., Group-based trajectory modeling: A novel approach to examining symptom trajectories in acute bipolar episodes. Journal of Affective Disorders (2012), http://dx.doi.org/10.1016/j.jad.2012.07.007

2

K. M’Bailara et al. / Journal of Affective Disorders ] (]]]]) ]]]–]]]

Although research has helped to better understand the many and various acute BD presentations, less is known about whether episodes evolve differently according only to episode polarity (depression, hypomania, etc.) or as a function of specific behavioural or vegetative symptom profiles (e.g. level of activation; agitated or retarded, etc.). Until recently, research in this area was hampered by the fact that few rating instruments adequately measured the full range of key symptom dimensions across all polarities. Most assessment scales were developed when more traditional views of depression and mania (as polar opposites) prevailed and so the assessments make assumptions about which symptoms should be measured based on the mood state defining the presenting polarity of the episode, rather than simultaneously measuring the range of mood, cognition and behavioural manifestations of BD regardless of the predominant mood (Johnson et al., 2010). The above issue is critically important in BD as it invariably confounds the naturalistic assessment of treatment outcomes, whilst studies of comparative effectiveness in BD are more complex than for other disorders. Clinical heterogeneity, as seen e.g. in the range of presentations of acute bipolar episodes, introduces a lack of clarity about which specific classes of drugs are best used for different acute BD symptom profiles. Contemporary research indicates that some treatments, e.g. atypical antipsychotics, are now viewed as efficacious acute treatments of both manic and depressive episodes. However, as randomized controlled trials select homogeneous subgroups who meet traditional episode criteria and trials report aggregate outcomes using conventional ratings scales for assessing change in either manic or depressive symptoms, it is not established whether the benefits of some treatments result from their action on a dimensional characteristic shared by a clusters of patients (e.g. increased activation levels) rather than whether the index BD episode is manic or depressive. Furthermore, we do not know if rate of improvement of symptoms in a depressive episode that is achieved with a specific treatment can be predicted from the rate of improvement of symptoms in a manic episode achieved with the same class of medication. To truly understand the role of medications that can be effective across polarities we need to find novel ways to measure change in symptoms and also new methods to examine response patterns that extend beyond ‘good versus bad’ outcome categories defined by cutoffs on a mania or a depression rating scale. In unipolar depression research, pattern analysis validly differentiates initial rate of change in symptoms in response to treatment and can predict (to a certain extent) the outcomes achievable in the continuation and maintenance phases (Quitkin et al., 1984). More recently, Marques et al. (2010) have applied this approach to treatment outcomes in schizophrenia, concluding that trajectory models of response, rather than the simple responder/non-responder dichotomy, provide a better statistical account of how antipsychotics may work. However, these studies largely address samples with relatively homogeneous presentations (e.g. acute psychosis) and observe the response pattern attained with a limited number of classes of medications (e.g. atypical antipsychotics). Before applying response pattern analysis models in BD, we need to establish the nature of symptom change in routine clinical settings and examine whether the trajectories generated lead to the identification of meaningful clusters of individuals with similar response characteristics. Studies need to include not only an appropriate statistical approach to pattern analysis but also a symptom measure that can potentially differentiate if changes in symptoms represent a beneficial shift towards euthymia or ‘overshoots’ euthymia and marks a ‘switch’ into (hypo)mania or depression. This paper describes a ‘exploratory study’ study aimed a) To clarify trajectories of change in acute BD episodes over time. To do this, we used group-based trajectory modeling

(GBTM), which is a statistical method designed to explore heterogeneity in clinical groups by identifying distinct trajectories of change (Nagin, 2005). b) To assess more subtle changes in symptoms and reduce the risk of false positive classifications of acute outcome (e.g. an individual who meets good outcome criteria for improvement in depression using standard rating scales, but has actually developed hypomanic symptoms). To do this, we employed the MATHYS (Multidimensional Assessment of Thymic States) which is an assessment tool that measures BD symptom dimensions and severeity irrespective of the polarity of the acute BD episode (Henry et al., 2008).

2. Experimental procedures 2.1. Sample With ethical approval, we recruited a convenience sample of acutely ill BD inpatients and outpatients who were willing and able to give written informed consent to participate in the study. Those exhibiting comorbidity, suicidality or psychotic symptoms were included (unless consent was an issue). As this was an observational study, treatment remained under the control of the responsible clinical team and any changes were made independently from the investigators. 2.2. Clinical assessment All participants were assessed using the mood section of the French version of the DIGS, a structured clinical interview incorporating DSM-IV diagnostic criteria (Nurnberger et al.,1994) during an acute episode. The severity of the mood episode was quantified with both the (Montgomery and Asberg, 1979) Montgomery and Asberg Depression Rating Scale (MADRS) and the Bech and Rafaelsen Manic scale (MAS) (Bech et al., 1978) and then patients were asked to complete the MATHYS, rating how they felt during the preceding week (week 1); the MATHYS was repeated on three further occasions (end of weeks 2, 3, 4). There are few clinical tools to assess bipolar episodes independently of polarity, but this was of critical importance in trying to make a more sophisticated interpretation of trajectories of change over time. We therefore employed the MATHYS (Multidimensional Assessment of Thymic States), a self-report scale that can be used especially to assess activation levels and emotional reactivity regardless of current BD episode status (Henry et al., 2008). This scale, designed a priori, includes five relevant quantitative dimensions (an English version can be accessed at: http://www.biomedcentral.com/content/supplemen tary/1471-244X-8–82-S1.doc). Thus, classic features, such as cognition, motivation, psychomotor agitation or retardation and sensory perception, are assessed quantitatively (i.e., racing thoughts or subjectively feeling that their thoughts are slower, physical agitation or retardation, and increase or decrease in sense perception). Examples of items include: ‘My brain never stops’; ‘My brain seems to be functioning in slow motion’. Similar concepts are applied to the evaluation of emotion (i.e. focusing on whether the patient felt emotion with normal intensity, greater intensity, or less intensity). Examples of these items include: ‘My emotions are very intense’; ‘My emotions are not very strong’. Analysis of the psychometric properties of the scale reveal that is has good validity and internal consistency (Cronbach’s alpha coefficient¼0.95), and that scores are moderately correlated with both the Montgomery Asberg Depression Rating Scale (depression score; r ¼ 0.45) and the Bech-Rafaelson Mania Scale (mania



score; r ¼0.56). There are five MATHYS subscales representing the quantitative dimensions (defined a priori) of emotional reactivity, cognitive speed, motivation, psychomotor activation and sensory perception. The total scale comprises 20 items, each rated on a visual analogue scale (VAS) with scores ranging from inhibition (score ¼1) to activation (score ¼10), and a rating of 5 indicating the normal state for that individual. A total score of 100 therefore represents euthymia, whilst scores o100 represent states of inhibition and 4100 represents activation or excitation. In another study, change in MATHYS score over time was a satisfactory measure of categorical outcome, with an Effect Size¼0.3 at weeks 6 and 24, and showed similar changes to the most commonly used depression and mania rating scales (HRSD; YRMS) (article under Henry et al., submitted for publication). This study also uses the total score to asses change in symptom dimensions over time. 2.3. Statistical analysis Trajectories of activation/inhibition level were modeled using the four consecutive MATHYS scores. Semi-parametric mixture models were estimated using the PROC TRAJ procedure (Jones et al., 2001) from SASs version 9.1 (2006). This method is particularly well suited as it allows identification of both the number of subgroups and number of subjects per sub-group as well as estimating trajectory shapes (Jones and Nagin, 2007; Nagin and Tremblay, 2001; Nagin, 1999). Given the nature of the MATHYS scale, we selected the CNORM model estimation. The Bayesian Information Criterion (BIC) was used to select the optimal number of groups and the shape of the trajectories of each group. We employed four criteria for identifying the number of subgroups: Jeffreys’ criterion was used for choice of model (Jeffrey, 1961) as it makes it possible to examine the relative credibility of different models according to the data. We then iteratively compared the value of BIC for a model of n groups with the previous (n 1 groups) and next (nþ 1 group) options. In parallel, we analyzed the average probability of belonging to each group (average posterior probability) using a minimum threshold for consideration of 70 as suggested by Nagin’s ‘rule of thumb’ (Nagin, 2005, p.88). Finally, the number of subjects in each group and the relevance of the theoretical solutions were also taken into account. The selection of trajectory shapes was based on the polynomial model best fitting the observations. By comparing the value of BIC and applying the recommendations of Jones (2005), we first selected a cubic trajectory for all groups and then reduced

1

3

the order of the polynomial models of each group until the parameters were significant. Models with 3–5 groups were analyzed. Comparing the BIC values, we found that the 4-group model best fitted the data. Similarly, in terms of average posterior probabilities, the 4-group solution was deemed suitable as the average estimated for each group exceeded.85. All parameters (constant, linear, quadratic) were significant.

3. Results The sample comprised 118 patients who met DSM-IV criteria for an acute BD episode: 56% (n ¼66) had major depression, 26% had either mania (n¼19; 15%) or hypomania (n ¼12; 10%), and 18% (n¼21) a mixed state. The mean age was 39.7 years (SD 11.3); 75% (n ¼89) were female and the majority of cases had BD I (60.5%; n¼ 71), with 39.5% (n ¼47) meeting criteria for BD II. At baseline assessment, the sample mean scores were: MADRS 15.84 (SD 9.62), MAS 9.49 (SD 8.98) and MATHYS 106.54 (SD 45.25). This was a naturalistic observational study, so treatment remained entirely the responsibility of the clinical team (but following the current guidelines) with no interventions allowed by the researchers. None of the patients was participating in any clinical drug trials, and so most individuals received classic combinations of medication selected according to the episode polarity. The GBTM analysis identified four distinct trajectories of symptom change using the MATHYS (see Fig. 1). These were – Trajectory 1: Persistent Inhibition, defined as MATHYS o100 and no change during follow-up). About 10.2% of the sample (n ¼12) demonstrated this trajectory, all of whom had major depression. The group mean MATHYS score at baseline was 52.29 (min 42.42 to max 62.16) and remained unchanged at the endpoint. – Trajectory 2: Transient Inhibition, this term is used to describe a trajectory of inhibition (a similar initial level as seen in Trajectory 1) followed by initial rapid progress towards euthymia, with a slowing of the rate of change such that there was little difference between the penultimate and final MATHYS scores. However, the mean final MATHYS score was shifted beyond euthymia into the activation range i.e. MATHYS 4100. Twenty five cases (21.2%) were included in this subgroup. The baseline score of the MATHYS was 57.64 (min 46.15 to max 69.14) and reached 106.84 (min 96.85 to max 116.82) at the endpoint.

Persistent Inhibition (10.2%)

Transient Activation (47,4%)

Transient Inhibition (21.2%)

Overactivation (21.2%)

2

3

4

Week Fig. 1. Trajectories as defined by GBTM analysis of mean total MATHYS scores over time.


4


80% 70%

76%

Persistent Inhibition Transient Inhibition

65%

Transient Activation 60%

Overactivation

50%

44%

40%

35%

33%

30% 20% 10% 0%

18% 14% 4%

10% 0% 0%

0%

Major depression

Mixed States

(hypo)Mania

Fig. 2. Distribution of cases by diagnosis across four groups defined by trajectory of change.

– Trajectory 3: Transient Activation, this term defined a cluster with moderate activation (MATHYS 120) at baseline, that showed linear improvement in symptoms achieving euthymia by the final follow-up (MATHYS 100). This trajectory incorporated the largest proportion of cases (n¼56; 47.4%). The baseline score of the MATHYS was 118.92 (min 112.83 to max125.01) and reached 101.26 (range 95.71–106.80) at the endpoint. The shape parameter is linear ( 0.84). – Trajectory 4: Over-activation, this cluster comprised those patients who were highly over-activated at baseline (MATHYS 160), with linear change over the follow-up period. The rate of symptom improvement was faster for this trajectory than the others, but the final MATHYS score of 120 (i.e. moderately activated) indicates most cases had not achieved euthymia. There were 25 cases (21.2%) in this subgroup. The baseline score of the MATHYS was 161.77 (min 151.17 to max 172.37) and reached 119.53 (min 110.92 to max 128.14) at the endpoint. The shape parameter is linear ( 2.03). Fig. 2, demonstrates the case composition of the clusters for each trajectory according to polarity of the acute BD episode. This demonstrates that about half the major depression cases are found within Trajectories 1 and 2 (baseline MATHYS score in the inhibition range) but half are within Trajectories 3 and 4 (baseline MATHYS score in the activation range). The majority of mixed cases (76%) are found in the Transient Activation group (Trajectory 3). Interestingly, most depressive or mixed cases follow Transient Activation or Inhibition (Trajectories 2 and 3), suggesting they progress towards euthymia (or slightly exceed that rating) over the one month follow-up period. The majority (65%), but not all (hypo)manic cases are in the Over-activation group (Trajectory 4), which shows a rapid linear rate of change, but this cluster has achieved only partial recovery by the end of one month.

4. Discussion This study explored the utility of GBTM in defining patterns of symptom change in acute bipolar episodes. GBTM was originally introduced to examine patterns that evolved over long periods of time (Lacourse et al., 2003; Mustillo et al., 2003) including the development of nicotine dependence in adolescents (Hu et al., 2008) and the emergence of depressive symptoms from early childhood to late adolescence (Dekker et al., 2007). It is only more

recently that it has been used to assess heterogeneity in shortterm outcomes such as treatment responses in clinical treatment trials or interventions (Brown et al., 2008; Peer and Spaulding 2007; Nagin and Odgers, 2010). Interest in this approach is evolving in psychiatry and in schizophrenia, Levine and colleagues have demonstrated the utility of trajectory analysis for understanding clinical trial data (Levine and Leucht, 2010; Levine et al., 2012) and analyzing data from epidemiological cohorts (Levine et al., 2011). The primary reason for using the statistical model and symptom assessment scale described was that, if it proved possible to describe different trajectories of symptom change or recovery in a general clinical sample of acute cases, it would suggest that it may well be feasible to apply this approach in the future to define clinical sub-groups in BD and predictors of treatment response, regardless of episode polarity. When symptom dimensions were assessed weekly using the MATHYS, four distinct trajectories were identified. As anticipated, patients with initial MATHYS scores 4100 (Trajectories ‘‘transient activation’’ and ‘‘overactivation’’), exhibited a progressive decrease in activation during the follow-up. However, there were two types of trajectory depending on the degree of activation at baseline. As is frequently observed in studies where initial symptom levels are high, the rate of improvement was quicker in trajectory ‘‘overactivation’’, which had the highest level of activation at baseline (although the final MATHYS score was only equivalent to the initial score of trajectory ‘‘transient activation’’). It is noteworthy that Trajectory transient activation included a heterogeneous case mix with the majority of mixed states allocated to this cluster, which are known to be difficult to treat ¨ (Muzina, 2009; Kruger et al., 2005; Gonza´lez-Pinto et al., 2010). For cases showing global behavioral inhibition (MATHYS score o100) at baseline, there were two distinct trajectories. Trajectory ‘‘transient inhibition’’ cases appeared to improve gradually, with some slowing in the rate of change in the final week of follow-up, but with their final MATHYS scores in the euthymic range (although slightly exceeding euthymia). Conversely, Trajectory ‘‘persistent inhibition’’ indicated no improvement and the cluster comprised cases with persistent severe inhibition. Patients with a major depressive episode were distributed across all the trajectories, with 51% depression cases allocated to Trajectories ‘‘persistent inhibition’’ and ‘‘transient inhibition’’ (with initial inhibition scores) and the remainder to Trajectory ‘‘transient activation’’ and ‘‘overactivation’’ (with moderate to high activation scores). These findings suggest that: i) clinical categories defined by the level of activation identify at least two



clusters of bipolar depression, ii) since trajectories of improvement differ, the modeling provides preliminary evidence of heterogeneity in the acute treatment response that may need to be taken into account in future treatment trials. The finding that bipolar depression cases do not constitute a homogeneous group, is concordant with recent publications such as the STEP-BD and studies by our own group (Goldberg et al., 2009; Frye et al., 2009; Henry et al., 2007a,b; Vieta et al., 2010; Henry et al., 2010). However, the use of GBTM elucidates that even short-term BD depression outcomes show great heterogeneity. Although a small but significant cluster of depression cases made up Trajectory ‘‘persistent inhibition’’, the distribution of depression cases across multiple trajectories highlights that future randomized treatment trials and comparative effectiveness studies may benefit from the use of novel statistical approaches that refine our ability to classify treatment responses. Whilst simple categorical outcomes may meet the needs of regulatory bodies that license medications, they fail to inform clinicians about how to differentially prescribe medications for the heterogenous mix of BD cases classified together according only to polarity. Obviously, this preliminary study has a number of limitations. First, the PROC TRAJ used in this study assumes missing data is missing completely at random. This assumption is open to dispute as many missing values are probably not random and so this may lead to bias in the estimation of our parameters. Moreover, there are a small numbers of cases in some trajectory clusters however GBTM can be applied on a total sample of more than 118 patients and 4 repeated measures. Second, the observational period focuses on acute outcomes and so we cannot use the findings to predict whether the trajectories lead to stable improvement or whether the clinical course for some clusters is cyclical, (or indeed if any cases experience ‘switching’ at a later stage). To further develop this approach in BD, studies that include more patients and provide assessments over a longer period of time are needed. Third, this was an observational study and it is not possible to comment on how prescribed treatments (combinations used, adequacy of doses, adherence, timing of introduction of treatment for the acute episode, etc) influenced trajectories of change in symptoms across or within clusters. However, we emphasize that our goal was to observe trajectories for cases representative of those seen in day to day clinical practice as we wanted to assess generic patterns. Exploring response patterns in homogeneous sub-samples of patients treated under controlled research conditions is the next step in the process. In conclusion, this study demonstrated that pattern analysis, especially if combined with a dimensional symptom assessment such as the MATHYS, can be used to explore trajectories of change in heterogeneous populations of individuals with an acute BD episode. The approach may be particularly relevant for BD depression. There is a need for ‘modeling’ differential responses to mood stabilizers, antipsychotics and/or antidepressants, in order to better reflect clinical reality and take symptom dimensions such as activation into account as well as (or even instead of) episode polarity. The MATHYS appears to offer additional insights into the different clusters of symptom dimensions observable across BD cases meeting episode criteria. In the future, larger studies will establish if the trajectories identified in this study (persistent inhibition, transient inhibition, transient activation and over-activation) are stable models or if more possible trajectories exist. The GBTM approach will hopefully moves us closer to the goal of establishing a more sophisticated understanding of symptom evolution in BD in the short-term, to develop a clearer rationale for specific therapeutic strategies over the long-term, and also to investigate any links between symptom dimensions, biomarkers of BD and trajectory groups.

5

Role of funding source This study was integrated in a hospital protocol of clinical search financed by the Ministry of Research in France.

Conflict of interest There is no conflict of interest.

Acknowledgements This paper was written in collaboration with members of ENBREC (European Network of Bipolar Research Experts Centres): Jan Scott, Eduard Vieta and Chantal Henry.

References Angst, J., 2011. Problems in the current concepts and definitions of bipolar disorders. World Psychiatry 10 (3), 191–192. Angst, J., Cui, L., Swendsen, J., Rothen, S., Cravchik, A., Kessler, R.C., Merikangas, K.R., 2010. Major depressive disorder with subthreshold bipolarity in the National Comorbidity Survey Replication. American Journal of Psychiatry 167 (10), 1194–1201. Bech, P., Rafaelsen, O.J., Kramp, P., Bolwig, T.G., 1978. The mania rating scale: scale construction and inter-observer agreement. Neuropharmacology 17, 430–431. Benazzi, F., 2007. Bipolar disorder-focus on bipolar II disorder and mixed depression. Lancet 17, 935–945. Brown, C.H., Wang, W., Kellam, S.G., Mutheń, B.O., Petras, H., Toyinbo, P., Poduska, J., Ialongo, N., Wyman, P.A., Chamberlain, P., Sloboda, Z., MacKinnon, D.P., Windham, A., 2008. Methods for testing theory and evaluating impact in randomized field trials: intent-to-treat analyses for integrating the perspectives of person, place, and time drug alcohol depend. Prevention Science and Methodology Group Suppl. 1, S74–S104. Cassidy, F., Forest, K., Murry, E., Carroll, B.J., 1998. A factor analysis of the signs and symptoms of mania. Archives of General Psychiatry 55, 27–32. Dekker, M.C., Ferdinand, R.F., van Lang, N.D.J., Bongers, I.L., Van Der J.D., Ende, Verhulst, F.C., 2007. Developmental trajectories of depressive symptoms from early childhood to late adolescence: gender differences and adult outcome. Journal of Child Psychology Psychiatry, and Allied Disciplines 48, 657–666. Frye, M.A., Helleman, G., McElroy, S.L., Altshuler, L.L., Black, D.O., Keck Jr, P.E., Nolen, W.A., Kupka, R., Leverich, G.S., Grunze, H., Mintz, J., Post, R.M., Suppes, T., 2009. Correlates of treatment-emergent mania associated with antidepressant treatment in bipolar depression. American Journal of Psychiatry 166, 164–172. Goldberg, J.F., Perlis, R.H., Bowden, C.L., Thase, M.E., Miklowitz, D.J., Marangell, L.B., Calabrese, J.R., Nierenberg, A.A., Sachs, G.S., 2009. Manic symptoms during depressive episodes in 1380 patients with bipolar disorder: findings from the STEP-BD. American Journal of Psychiatry 166, 173–181. Gonza´lez-Pinto, A., Barbeito, S., Alonso, M., Alberich, S., Haidar, M.K., Vieta, E., Tabare´s-Seisdedos, R., Zorrilla, I., Gonza´lez-Pinto, M.A., Lo´pez, P., 2010. Poor long-term prognosis in mixed bipolar patients: 10-year outcomes in the vitoria prospective naturalistic study in Spain. Journal of Clinical Psychiatry 7, 671–676. Henry, C., Luquiens, A., Lancon, C., Sapin, H., Astruc, B., Perrin, E., Falissard, B., Lukasiewicz, M.. Inhibition/activation in bipolar disorder: a further validation of the mutlidimensional assessment of thymic states scale (MAThyS) in an open label six months clinical trial. Submitted for publication. Henry, C., M’Baı¨lara, K., Mathieu, F., Poinsot, R., Falissard, B., 2008. Construction and validation of a dimensional scale exploring mood disorders: MAThyS (Multidimensional assessment of thymic states). BMC Psychiatry 19 (8), 82. Henry, C., M’Baı¨lara, K., Desage, A., Gard, S., Misdrahi, D., Vieta, E., 2007a. Towards a reconceptualization of mixed states, based on an emotional-reactivity dimensional model. Journal of Affective Disorders 101, 35–41. Henry, C., M’bailara, K., Le´pine, J.P., Lajnef, M., Leboyer, M., 2010. Defining bipolar mood states with quantitative measurement of inhibition/activation and emotional reactivity. Journal of Affective Disorders 127, 300–304. Henry, C., M’Baı¨lara, K., Poinsot, R., Casteret, A.A., Sorbara, F., Leboyer, M., Vieta, E., 2007b. Evidence for two types of bipolar depression using a dimensional approach. Psychotherapy Psychosomatics 76, 325–331. Hu, M.C., Muthen, B., Schaffran, C., Griesler, P.C., Kandel, D.B., 2008. Developmental trajectories of criteria of nicotine dependence in adolescence. Drug and Alcohol Dependence 98, 94–104. Jeffreys, H., 1961. Theory of Probability (3e). Oxford University Press Ed., New York. Johnson, S.L., Morriss, R., Scott, J., Paykel, E., Kinderman, P., Kolamunnage-Dona, R., Bentall, R.P., 2010. Depressive and manic symptoms are not opposite poles in bipolar disorder. Acta Psychiatrica Scandinavica 123, 206–210. Jones, B.L., Nagin, D.S., 2007. Advances in Group-Based Trajectory Modeling and an SAS Procedure for Estimating Them. Sociological Methods and Research 35, 542–571. Jones, B.L., Nagin, D.S., Roeder, K., 2001. A SAS procedure based on mixture models for estimating development trajectories. Sociological Methods and Research 29, 374–393.


6


Jones, B.L., 2005. PROC TRAJ Retrieved 06/11/08. Available from /http://www. andrew.cmu.edu/user/bjones/strtxmpl1.htmS. ¨ ¨ Kruger, S., Trevor Young, L., Braunig, P., 2005. Pharmacotherapy of bipolar mixed states. Bipolar Disorder 7, 205–215. Lacourse, E., Nagin, D., Tremblay, R.E., Vitaro, F., Claes, M., 2003. Developmental trajectories of boys’ delinquent group membership and facilitation of violent behaviors during adolescence. Development and Psychopathology 15, 183–197. Levine, S.Z., Leucht, S., 2010. Elaboration on the early-onset hypothesis of antipsychotic drug action: treatment response trajectories. Biological Psychiatry 68, 86–92. Levine, S.Z., Rabinowitz, J., Faries, D., Lawson, A.H., Ascher-Svanum, H., 2012. Treatment response trajectories and antipsychotic medications: examination of up to 18 months of treatment in the CATIE chronic schizophrenia trial. Schizophrenia Research 137 (1–3), 141–146. Levine, S.Z., Lurie, I., Kohn, R., Levav, I., 2011. Trajectories of the course of schizophrenia: From progressive deterioration to amelioration over three decades. Schizophrenia Research 126, 184–191. Marques, T., Arenovich, T., Agid, O., Sajeev, G., Muthen, B., Chen, L., Kinon, B., Kapur, S., 2010. The different trajectories of antipsychotic response: antipsychotics versus placebo. Psychological Medicine 20, 1–8. Montgomery, S., Asberg, A., 1979. A new depression scale designed to be sensitive to change. The British Journal of Psychiatry 134, 382–389. Mustillo, S., Worthman, C., Erkanli, A., Keeler, G., Angold, A., Costello, E.J., 2003. Obesity and psychiatric disorder: developmental trajectories. Pediatrics 111, 851–859.

Muzina, D.J., 2009. Pharmacologic treatment of rapid cycling and mixed states in bipolar disorder: an argument for the use of lithium. Bipolar Disorder 11, 84–91. Nagin, D.S., Tremblay, R.E., 2001. Analyzing developmental trajectories of distinct but related behaviors: a group-based method. Psychological Methods 6, 18–34. Nagin, D.S., Odgers, C.L., 2010. Group-based trajectory modeling in clinical research. Annual Review of Clinical Psychology 27, 109–138. Nagin, D.S., 1999. Analyzing developmental trajectories: a semiparametric, groupbased approach. Psychological Methods 4, 139–157. Nagin, D.S., 2005. Group-based Modeling of Development. Harvard University Press, Cambridge, MA. Nurnberger Jr., J.I., Blehar, M.C., Kaufmann, C.A., et al., 1994. Diagnostic interview for genetics studies. Rationale, unique features, and training. NIMH genetics initiative. Archives of General Psychiatry 51, 849–859. Peer, J.E., Spaulding, W.D., 2007. Heterogeneity in recovery of psychosocial functioning during psychiatric rehabilitation: an exploratory study using latent growth mixture modeling. Schizophrenia Research 93, 186–193. Quitkin, F., Rabkin, J., Ross, D., Stewart, J.W., 1984. Identification of true drug response to antidepressants: use of pattern analysis. Archives of General Psychiatry 41, 782–786. Vieta, E., de Arce, R., Jimeńez-Arriero, M.A., Rodriguez, A., Balanza´, V., Cobaleda, S., SIN-DEPRES Group, 2010. Detection of subclinical depression in bipolar disorder: a cross-sectional, 4-month prospective follow-up study at community mental health services (SIN-DEPRES). Journal of Clinical Psychiatry 71, 1465–1474.
