|Archived: Predictability in Bipolar Disorder
A major clinical challenge in caring for patients with severe psychiatric disorders is to formulate a confident prognosis and to anticipate needs for treatment and support over prolonged periods of risk. This challenge is particularly compelling in potentially disabling or fatal major mood disorders, which follow notoriously complex courses with marked individual differences [2,4,11,25]. Bipolar I disorder is an attractive condition in which to study prognostication and prediction since it appears longitudinally to be the most diagnostically stable of major psychiatric disorders [16,17]. It has also become fashionable to consider its prevalence about as large as for unipolar major depressive disorder (MDD) [1,11]. Our interest in this topic was stimulated several years ago by discovering a long-neglected study in a German journal from 1938 by Eliot Slater of the Maudsley Institute in London . It indicated that the proposal by Emil Kraepelin a century ago that the course of manic-depressive illness (MDI, now bipolar disorder [BD] and recurrent major depression [MDD]) is toward acceleration (shorter cycles) with more recurrences. Slater pointed out, and we verified, that this impression arises from a sampling error based on assessing cycle-length in pooled samples of individuals with relatively slower or more rapid natural cycling (low and high recurrence counts with correspondingly low and high cycle-lengths). Kraepelin’s findings disappear when cycle-length is considered with samples matched for episode-counts . The literature on this topic includes 38 studies over the past century, with no agreement as to whether cycle-acceleration occurs in MDI or BD [5,11]. Moreover, data from the McLean First Episode Project found, among over 300 BD subjects, that within-subject slope-functions (cycle-length vs. episode-count) followed a random distribution, with evidence of slowing as well as of acceleration and a central node at little or no change in intervals with more recurrences . These findings are inconsistent with the “kindling” hypothesis that suggests worsening in BD with more recurrent episodes, possibly associated with cerebral damage . Although acceleration can be questioned, the clinical impact of multiple recurrences in BDs is real and can be devastating.
This experience led us to wonder to what extent the course of BD is a random or chaotic phenomenon both within and between individuals, starting from the widely accepted view that BD tends to present at younger ages than MDD. In international samples involving several thousand subjects, onset-age ranked: BD-I < BD-II < MDD. Younger onset-age increased the probability of diagnosis of BD as well as predicting greater long-term disability. Moreover, familial mood disorders were more likely with younger onset, possibly with at least two subpopulations: a young-onset, familial group and an older-onset, sporadic group, suggesting that earlier onset may be associated with biological, possibly genetic, contributions in BD . Another feature of early onset is that an inevitable delay in diagnosis and appropriate treatment of BD averages 8–10 years from first symptoms, but can exceed 16 years with juvenile onset . Such delays tend to be longer in BD-II than BD-I cases, especially when initial illnesses are depressive or not severe, and are somewhat longer among women . Delay of diagnosis and of appropriate treatment for BD (including judicious use of antidepressants ) can lead to greater morbidity, disability and costs , but also increases risk of fatal outcomes since at least half of long-term risk of suicidal behaviors in mood disorder patients occurs within the initial 2–5 years of illness [21,22].
We next tested the hypothesis that the course of BD might be predicted by the nature of initial episodes of illness [6,7]. In a sample of 550 US & EU BD-I subjects, manic or hypomanic first-episodes were followed by an excess of later manialike recurrences, that depression was followed by a 7-fold excess of later depressions, that initial mixed manic-depressive states were followed almost entirely by either depressions or more mixed-states, and that there was a 2-fold excess of future depression over mania overall in BD-I patients. Recently, we found in 1081 European BD-I patients followed for 15.7 years that onset-types ranked: depression (59%) > mania (13%) > psychosis (8.0%) > anxiety (7.6%) > hypomania (6.7%) > mixed-states (5.5%) . Each onset-type was predictive of marked differences in later percent-of-time in depression or manialike states: initial anxiety and depression were strongly predictive of later depression and initial mixed-states predicted more later depression than mania, whereas initial hypomania, psychosis or mania strongly predicted later manialike morbidity. Overall these BD patients spent 2.1-fold more time in depression than mania-like states .
Predominant polarity, defined as =2.0-fold overall excess of depression or of manialike episodes, was proposed by colleagues in Barcelona, and is found in half of BD subjects. Identification of predominant polarity in BD is not specifically predictive. Nevertheless, it reveals associations with predictive significance, and itself may be predicted by the polarity of initial episodes. In data concerning 928 BD-I disorder subjects from 5 collaborating international sites, we found that initial episode-polarity strongly predicted later predominant types of illness episodes . That is, initial mania or psychosis was followed by a 6-fold excess of manialike episodes, whereas initial depression or anxiety was followed by a 7-fold excess of depressive recurrences; initial mixed-states were followed by a striking 47-fold excess of later predominance of depression , supporting the proposal that mixed-states are more closely related to bipolar depression than to mania . Separation of BD patients into those in whom mania or depression predominates appears to be a useful distinction in other ways, too. For example, those with predominant depression were more likely: to receive ECT, have a longer latency from onset of initial illness to diagnosis, present initially in depression or a mixed-state, be psychotic at some time, meet diagnostic criteria for an Axis-II personality disorder, attempt suicide, experience mixed-states, have rapid-cycling (=4 recurrences within a 12-month period), to be married, female, and unemployed at some time. In contrast, predominant mania was associated with initial mania or psychosis, substance abuse, more education, and more familial mood disorders .
An important aspect of prognosis in mood disorders is to deal with prevalent initial presentations in depression among BD-I and BD-II, their early differentiation from MDD patients, and timely prediction of those later diagnosed with BD. This is a matter of particular concern in juveniles presenting in depression, a high proportion of whom are later considered to have BD, are at particularly increased risk of suicidal behaviors if bipolar, and in whom excessive antidepressant treatment may have adverse effects that include unanticipated and potentially dangerous mood-switching and possibly risk of suicide [2,3,23,24]. We first compiled information from reports of characteristics that effectively differentiated BD from MDD. In descending order of statistical significance, they included: mood-switching during antidepressant treatment, lability or prominent anger or diagnostic consideration of borderline disorder, suicidal acts, family history of probable BD, substance abuse, depression with agitation or mild hypomanialike features, psychotic symptoms, =4 depressive recurrences before mania/hypomania, cyclothymic temperament, “atypical” depressive features (hypersomnia, hyperphagia, psychomotor slowing), attention deficits, and earliest affective illness before age 25 [1,20,23,24]. Since much of the preceding work involved adults who presented in depression and were later diagnosed with BD, with Research Fellow Giulia Serra, we also considered juvenile antecedent clinical features that would predict future BD vs. MDD . In 334 Roman patients, features that significantly predicted later BD, in descending order of statistical significance included: adolescent mood swings, substance abuse, eating disorders, restlessness or hyperactivity in childhood, depressive symptoms, familial BD, male sex, cyclothymic or hyperthymic temperament, lack of anxiety symptoms, relatively early age at first antecedents, and having =2 antecedents. These analyses are limited by retrospective reliance on detailed psychiatric records of adults with firm diagnoses of BD or MDD. With colleague Paola Salvatore, we also evaluated early features of adults with secure diagnoses of BD that might identify the type of first-lifetime major episodes. This approach supported anticipation of first major episode-types in 263 McLean First Episode, psychotic BD-I patients . Among subjects with non-manic presentations (mixed-states or depression) vs. mania, in descending order of Odds Ratio, predictors included: early, often juvenile, attention disturbances, depression, evidence of perceptual perplexity, substance abuse leading to detoxification, unstable or mixed affective states, antidepressant treatment, impulsivity, and states of anhedonia or diminished vitality closely preceding first acute episodes of BD illness .
Another characteristic of BD is the distinction of long-term course-type of those who present initially in depression, followed by mania or hypomania and then a euthymic interval (DMI pattern), or its opposite (MDI), as proposed by our late collaborator, Anathasios Koukopoulos. In a recent analysis of 855 patients in Roman & Sardinian mood disorder centers founded by Dr. Koukopoulos and collaborator Leonardo Tondo, we found an approximately equal likelihood of the DMI and MDI course-patterns, although MDI was more associated with BD-I diagnoses and DMI with BD-II syndrome marked by recurrent depression and some hypomania . The DMI course-type was associated with female sex, depression or anxiety as a first-lifetime episode, and an excess of depression in long-term follow-up, and importantly, an overall clinical response to long-term mood-stabilizing treatments that was 71% less favorable than with MDI patients. Associated with the MDI course were a BD-I diagnosis, male sex, earlier onset-age, manic or hypomanic or psychotic first episodes, substance abuse, and long-term treatment with a mood-stabilizer, as well as a superior response to such treatment .
Finally, with Research Fellow Alberto Forte, we recently analyzed both McLean First Episode findings from the start of BD-I, pooled with findings in over 40 reported studies (N=6119 subjects) that evaluated long-term morbidity patterns in clinically treated BD and MDD patients. In BD-I, BD-II & unipolar MDD cases, the proportion of weeks unwell averaged between 40% and 50%. In MDD, unsurprisingly, virtually all morbidity was depressive, but strikingly in BDs, depression accounted for more than three-quarters (77%) of total time ill, and depression was somewhat more prominent in BD-II than BD-I subjects (83% vs. 71%). Prognostically, DMI was associated with initial depression, and MDI with first-episode mania. These findings underscore the current state of therapeutics for BD, in which available treatments appear to be relatively effective long-term for manic and psychotic manifestations, but far from satisfactory for bipolar depression. Moreover, the morbidity found in MDD patients raises a broader question of limited effectiveness of long-term treatments for depressive illness in general. Such findings might be colored by the clinical impression that depression tends to be more clearly recalled and complained about by many mood-disorder patients, at least compared to hypomania (“me at my best”), although full manic episodes are likely to be memorable by changes in treatment, hospitalization, or legal or financial troubles. Instead, we suspect that a major contributor to such outcomes is inconsistent and time-limited adherence to recommended long-term treatments (mood-stabilizers in BD or antidepressants in MDD, and psychotherapy in both) rather than, or in addition to, limited prophylactic effectiveness of available treatments. Regardless of its basis, depressive morbidity in all mood disorders remains a major clinical & public health challenge as depression has become the leading medical basis of long-term disability, according to the WHO, and because depression is strongly associated with impaired social and vocational functioning, substance abuse, accidents, self-harm, and suicide in young adult years as well as with excess mortality from medical disorders in later life. A theme running through all of the studies summarized here is that depression in BD is predictable, at least from the type of first major episode (depressive, mixed, or anxious), and possibly by factors that can be identified as early as in childhood. It is to be hoped that such information can contribute to improving long-term outcomes in all mood disorders, and guide further development of more effective and better-tolerated and accepted mood-altering treatments.