. 63
( 68 .)


Fig. 17.2 Course over 15-year follow-up. Bipolar I (BPI), n = 135; bipolar II (BPII), n = 71.
Adapted from Judd et al. (2002).

Robb et al. (1997) employed a self-report measure, the Illness Intrusiveness
Rating Scale, and found that euthymic BP-II patients suffered greater impairment
than BP-I subjects (Robb et al., 1997).
Judd et al. (2002) compared the course of illness for BP-I (n ¼ 135) and
BP-II (n ¼ 71) followed in the National Institute of Mental Health (NIMH)
Collaborative Depression Study over 15 years of follow-up. Figure 17.2 shows
outcomes from the Longitudinal Interval Follow-up Evaluation (LIFE: Keller et al.,
1987), one of the few measures that harvests systematic data over time periods
sufficiently long enough to describe course of illness. The impressive differences in
percentage of weeks depressed (BP-I ¼ 31%; BP-II ¼ 52%) and the ratio of weeks
depressed to weeks with hypomaniaþmania (BP-I ¼ 3:1; BP-II ¼ 37:1) demon-
strates the possibility for capturing long-term outcomes.

Difficulties in conducting clinical trials for mixed episodes
The DSM-IV definition of mixed episodes requires a period of at least 1 week
during which a patient meets full criteria for mania and major depression. Few
studies reporting results for mixed episodes, however, use the DSM-IV definition.
Leaving aside the confusing array of ˜˜mixed states™™ described as fulfilling one or
more criteria for ˜˜mixiety™™ (Akiskal and Pinto, 1999), many studies of mixed states
are problematic. Within the limits of brevity permissible here, it is possible to
highlight only two commonly underappreciated issues related to interpretation of
studies reporting results for mixed states.
Some studies operationalize the term ˜˜mixed state™™ by using cut-off scores on
formal rating scales, which are not validated as diagnostic instruments. For
373 Investigational strategies

instance, it is not possible to state with certainty that a patient meeting criteria for
mania and also having a Hamilton Depression Rating Scale score > 20 meets the
DSM criteria for a mixed episode.
Most treatment results published for mixed episodes are derived from clinical
trials conducted primarily to gain regulatory approval for treatment of acute
mania. Studies designed for this objective are subject to potentially serious distor-
tions when outcomes are reported. In such studies response rates are operationally
defined based on improvement on the mania rating scale without considering the
persistence or worsening of depressive severity. Thus patients with moderate
mania at study entry who become severely depressed during the study can meet
the outcome criteria for treatment response. Furthermore, accession bias can have
dramatic consequences. At the baseline study visit, clinically depressed bipolar
patients with just one or two moderate manic symptoms, such as irritability,
agitation, insomnia, and racing thoughts, may well meet rating scale criteria for
a mixed episode. In light of longitudinal data indicating that mixed episodes are
very likely to be chronic (Kupfer et al., 2001), it is surprisingly common to observe
that a substantial percentage of subjects in clinical trials with mixed episodes at
baseline meet remission criteria at the week 1 follow-up assessment.
Addressing factors related to the specificity of diagnosis and treatment outcome
has great potential to improve the prospects for research on mixed episodes.
Standardizing criteria for accession of more uniform samples or at least clearly
defining the condition is a critical need. Himmelhoch et al. (1976) recognized the
importance of secondary factors such as comorbid conditions and psychoactive
substance use in the phenomenology of mixed episodes. Kraepelin (1921) noted
the occurrence of mixed states and distinguished transient mixed states which
might arise in the course of cycling from a manic episode to a depression, from
persistent states in which the symptoms of mania and depression co-occur
chronically. Entry criteria for studies focusing in mixed states could promote
more homogeneous samples by excluding subjects with known secondary factors
and requiring a duration longer than 4 weeks to improve the homogeneity of their
sample. Studies reporting outcomes for mixed episode should include composite
outcome measures that employ concurrent assessment of mania and depression
rather than reporting outcome measure related exclusively to mania (e.g., 50%
improvement from baseline mania rating scale score).

Difficulties in conducting clinical trials for rapid cycling
What is rapid cycling? The term ˜˜rapid cycling™™ was coined by Dunner et al. (1977)
based on the high frequency of four episodes or more per year among those
patients with poor response to lithium. Since then, treatment for rapid cycling
374 G. Sachs and M. Graves

has become recognized as an important area of unmet clinical need. The DSM-IV
classifies rapid cycling as a course specifier rather than a subtype of bipolar
disorder. Although rapid cycling is associated with relatively poor response to
treatment and persistence of higher rates of cycling than non-rapid cycling (Bauer
et al., 1994; Baldessarini et al., 2000), bipolar illness, indices such as family history,
and age of onset do not separate rapid-cycling from non-rapid-cycling patients
(Bauer et al., 1994). Furthermore, prospective follow-up reveals rapid cycling is
seldom persistent. Among subjects diagnosed as rapid cycling on entry into the
NIMH Collaborative Depression study, Coryell et al. (1992) found only a third
demonstrated four or more episodes through the first year of prospective follow-
up and in only 3% did rapid cycling persist through 3 years.
The DSM-IV concept of rapid cycling retains Dunner and colleagues™ definition
of rapid cycling as four or more episodes in 1 year (Dunner et al., 1977).
Importantly, the DSM-IV concept of rapid cycling requires counting episodes;
either as four episodes separated by periods of remission or a switch from an
episode of one polarity to an episode of opposite polarity. Strict application of the
DSM definitions can provide upper as well as lower boundaries for annual episode
frequency consistent with the rapid-cycling concept. Notably, the DSM requires
separate episodes be bounded by a period of full or partial remission lasting at least
8 weeks and the definition of mania requires the presence of symptoms for at least
1 week. Therefore, in the course of a year, patients following this 9-week pattern
could have no more than six episodes. A pattern of continuous cycling in which
a patient abruptly switches from a 1-week period meeting criteria for mania to a
2-week period of symptoms meeting criteria for depression could produce a higher
cycle frequency. Even repetition of this 3-week pattern throughout a year could
produce an annual cycle frequency only as high as 17 per year. Many patients and
practitioners, however, report cycle frequencies greatly exceeding 17 per year. In
fact, it has become common to hear descriptions of patients with multiple cycles
within a single day. These various forms of so-called ˜ultrarapid cycling™ are
characterized by ˜˜truncated episodes™™ (Bauer et al., 1994). The concept of trunc-
ated episodes allows a phase to count toward the diagnosis of rapid cycling even
when that phase is too short to qualify as a DSM-defined episode. The advantage of
the truncated episode concept is obvious: it allows the rapid-cycling designation to
be applied to patients such as those described with 48-h periods of depression
alternating with 48-h periods of mood elevation. The problem, however, is that
when we suspend the requirement to meet the definition for an episode we lack
criteria to distinguish meaningfully mixed episodes from rapid cycling or even
reliably distinguish a phase of illness from an emotion. Using the concept of
truncated episodes in clinical trials, therefore, requires researchers to use great
caution and standardized counting procedures.
375 Investigational strategies

What are the lessons from clinical trials for rapid cycling?
Perhaps the most striking lesson from trials that have focused on rapid cycling is
how difficult it is to study this condition. To date the only published parallel-
group double-blind controlled trial reporting results for rapid cycling found a
significant advantage for lamotrigine over placebo. This trial is remarkable not
only because it represents the best available evidence pertaining to the treatment of
rapid cycling, but also because the results document a clearly consistent pattern of
differential response across multiple outcome measures in which BP-II rapid
cycling appears robustly responsive to lamotrigine and BP-I patients appear
minimally, if at all, responsive to lamotrigine (Calabrese et al., 2000).
Published reports of lithium treatment for rapid cycling suffer from the limita-
tions of mirror design, but do suggest lithium can be a beneficial treatment for
some rapid-cycling patients (Dunner et al., 1977; Maj et al., 1989). One small
single-blind trial described benefit for thyroid hormone at hypermetabolic doses
(Bauer and Whybrow, 1990). There are, however, no published studies that
examine the use of valproate, carbamazepine, gabapentin, oxcarbazepine, conven-
tional antipsychotics, or atypical antipsychotics under double-blind conditions.
Some recent clinical trials may create confusion when reporting results for rapid-
cycling patients for outcomes other than rapid cycling. For example, post hoc analyses
have been conducted to examine results for rapid-cycling patients enrolled in double-
blind trials testing the efficacy of potential treatments for acute episodes of mania or
depression. Such results, with appropriate caveats, can be perfectly acceptable, but
confusion may arise for several reasons. First, the reported outcome measure for
acute episodes is easily misinterpreted as an outcome for rapid cycling per se. Change
in rating scale scores over a brief (usually 3“4-week) phase of blinded treatment does
not speak to the issue most pertinent for rapid cycling “ change in cycle frequency
over time. Second, even interpretation of acute response rates should take into
consideration whether the process of randomization balances treatment groups for
the presence of rapid cycling. Third, published reports often fail to clarify whether
results presented for rapid-cycling patients refer to current or lifetime rapid cycling,
and fail to make clear that the diagnosis of rapid cycling relied on a single question
asked during the baseline assessment of an acutely ill patient. Consistent with the
DSM-IV classification, rapid cycling concurrent with study entry might represent a
valid state designation. Bipolar individuals who are not currently cycling rapidly, but
who have experienced a period of rapid cycling in the remote past, might define a
clinically important trait. Bipolar patients shown to be prone to periods of cycle
acceleration or affective switch may represent a distinct subgroup, perhaps at high
risk of becoming manic during treatment with standard antidepressant medications.
Consistent use of operationalized definitions and validated procedures for assessment
376 G. Sachs and M. Graves

are needed to examine the state-versus-trait question. (See Definitions for character-
izing rapid cycling, below.)
Post hoc analysis of results for patients with rapid cycling have been presented
for the acute mania trials (Bowden and McElroy, 1995; Tohen et al., 1999) and
illustrate common problems related to assessment and interpretation. The acute
mania trials conducted by Bowden and McElroy, and Tohen et al. relied on
retrospective diagnosis of rapid cycling based on simply asking acutely manic
patients how many episodes they had had. Post hoc analysis of the Bowden trial is
sometimes cited as evidence of the superior efficacy of valproate over lithium. This
interpretation cannot be correct. The randomization process in that acute mania
study was not intended to balance the treatment groups for rapid cycling and for
purely random reasons did not result in the assignment of any rapid-cycling
patient to the lithium treatment group. Tohen et al. found a higher response
rate in rapid-cycling patients treated with olanzapine than non-rapid-cycling
patients. This outcome indicates that manic episodes in subjects with rapid cycling
responded to olanzapine sooner than manic episodes in non-rapid-cycling bipolar
patients, but does not address rapid cycling.
The problem of counting episodes is a serious obstacle to research on rapid
cycling. Figure 17.3 depicts the course of a single hospitalized patient over a period
of about 90 days. The graphed daily ratings, made by nursing staff using independ-
ent scales to rate separately the severity of mania and depression, reveal extreme
variability of mood state. While this patient was described clinically as simply
having a manic episode, an observer counting the peaks and valleys generated by
the daily ratings could easily describe this patient™s course as ultrarapid cycling,
and a rater applying DSM-IV criteria would likely diagnose a mixed episode. This
discrepancy, which is a serious impediment to research on rapid cycling, requires

t Depression

0 20 40 60 80

Fig. 17.3 Daily independent mania and depression ratings. Depression is the most common chief
complaint during mania. Adapted from Goodwin and Jamison (1990).
377 Investigational strategies

Table 17.1 Counting phase shifts

Previous assigned mood state

Current mood state Depression Hypomania Mania Mixed

Depression No phase shift Yes Yes No phase shift
Hypomania Yes No phase shift No phase shift No phase shift
Mania Yes No phase shift No phase shift No phase shift
Mixed Yes Yes Yes No phase shift

investigators to develop clear operational procedures that can be applied consist-
ently by raters at multiple sites.
Standardized assessment tools for prospective longitudinal follow-up may offer
advantages over retrospective assessments like LIFE. The advantages of prospective
rating may be particularly important when attempting to track multiple brief
fluctuations in mood state.
Defining a phase of illness is a key research need. Kramlinger and Post (1996)
defined an alternative to the DSM episode definitions of ultrarapid cycling to
account for phase shifts that occur within a 24-h period. The 15-point Bunney-
Hamburg Rating Scale, administered twice daily by nurses, was used to identify a
phase shift. Subjects also marked 100-mm visual analogue lines, where the left side
was anchored by ˜˜best ever™™ (manic) and the right side by ˜˜worst ever™™ (depressed).
A new depressive episode was defined as a sudden increase of three points in the
assessment score or at least 7 days with a depression rating greater than 7. The end of
the depressive episode was determined by a sudden decrease of 3“6 points, a
sustained period of at least 7 days with a score less than five, or a switch in polarity
from depression to mania.
There are some obvious problems with this technique: a patient who was already
in a major depressive episode could experience several brief depressions due to
transient three-point fluctuations in the course of a single day. Perhaps more
important than the technical issues is the fact that patients are typically not
observed in inpatient units while undergoing treatment for rapid cycling. In
order for a trial to target rapid cycling effectively, it would need to take place
over a period of 6 months or more. Outcome measures that rely on trained
observer ratings are impractical, because subjects are not likely to remain hospi-
talized for the long periods required to assess multiple episodes.
Methods developed in our clinic for counting phases can be more readily
applied to outpatient studies. Phase changes are defined as the appearance of
a new mood state with duration of 48 h (Table 17.1). If the new mood state is
not sustained for 48 h, it must meet the DSM-IV criteria for an episode. This
378 G. Sachs and M. Graves

Month 1 Mood chart
Psychotic Sxs
Daily mood ratings indicate highest and

lowest mood state for each date.


. 63
( 68 .)