<<

. 25
( 68 .)



>>

in more correct diagnoses than strict structured interviewing (Dunner and Tay,
1993; Brugha et al., 2001). All patients were systematically SCID-CV-interviewed
for a history of manic/hypomanic episodes, and for DSM-IV hypomanic symp-
toms during the index MDE. The SCID-CV-structured question on racing
thoughts was supplemented by the Koukopoulos and Koukopoulos™ definition
(1999) of crowded thoughts (head continuously full of ideas that the patient is
unable to stop). The SCID-CV skip-out instruction of the stem question about
past hypomanic mood was not followed, as a negative answer would not allow
assessment of the other hypomanic symptoms. It was shown (Benazzi and
Akiskal, 2003a) that systematic assessment of all past hypomanic symptoms
increased the frequency of BP-II diagnoses, as overactivity was easier to remem-
ber (by patient and family members/close friends) than hypomanic mood (diag-
nosis of hypomania always required hypomanic mood, which was easier to
remember after having remembered overactivity). A history of mania/hypomania
was always investigated soon after having made the diagnosis of MDE, before the
assessment of study variables, to avoid a possible bias related to knowledge of
indicators of bipolarity. BP (I þ II) family history was investigated with the Family
History Screen (Weissman et al., 2000), a structured interview for psychiatric
history of first-degree relatives.
Depressive mixed state (DMX) was defined as an MDE plus more than two
concurrent hypomanic symptoms, following Benazzi and Akiskal (2001). The
more clinically useful definition of DMX was found to be one based on a
minimum number of hypomanic symptoms (three) during MDE versus one
based on the combination of specific hypomanic symptoms, by multivariate
analyses (Benazzi, 2002e, 2003b). Hypomanic symptoms during the MDE lasted
at least 1 week, appeared during the MDE, and were present at the time of the
interview.
The DSM-IV 4-days™ minimum duration of hypomania for BP-II diagnosis
(a cut-off not based on data: Dunner, 1998) was not followed. Instead, at least 2
143 Atypical depression and bipolar spectrum



Table 6.1 Comparisons between atypical depression (AD) and non-atypical depression
(N-AD) (n ¼ 433)

AD N-AD
(n ¼ 182) (n ¼ 251)

Variable: mean (SD), % T/Z df P

Unipolar 24.1 51.3 “5.7 0.0000
Bipolar II 75.8 48.6 5.7 0.0000
Female gender 74.1 58.9 3.2 0.0010
Age at index MDE (years) 40.2 (13.5) 46.4 (15.3) “4.3 431 0.0000
Age at onset of first MDE (years) 22.2 (10.6) 29.7 (14.0) “6.0 431 0.0000
> 4 MDEs 76.3 69.3 1.6 0.1084
MDE symptoms > 2 years 44.5 37.0 1.5 0.1161
Axis I comorbidity 58.2 45.8 2.5 0.0108
MDE severity by GAF 51.3 (7.9) 50.0 (9.8) 1.4 431 0.1409
Psychotic features 2.7 11.9 “3.4 0.0005
> 2 hypomanic symptoms 63.1 35.8 5.6 0.0000
during index MDE
Bipolar (I þ II) family history 50.8 32.1 3.9 0.0001
Irritability 60.4 47.4 2.6 0.0075
Psychomotor agitation 31.8 21.5 2.4 0.0156
Distractibility 76.3 62.9 2.9 0.0030
More talkativeness 22.5 14.7 2.0 0.0369
Racing/crowded thoughts 75.2 58.9 3.5 0.0004
Mood reactivity 100 74.5
Weight gain 36.8 1.9 9.6 0.0000
Overeating 46.7 3.1 10.9 0.0000
Hypersomnia 63.7 7.1 12.5 0.0000
Leaden paralysis 69.2 17.1 10.9 0.0000
Interpersonal rejection sensitivity 85.1 40.6 9.3 0.0000

MDE, major depressive episode; GAF, Global Assessment of Functioning.




days of hypomania were required for BP-II diagnosis, following previous reports
(Akiskal et al., 1977; Cassano et al., 1992; Coryell et al., 1995; Akiskal, 1996; Angst,
1998; Akiskal et al., 2000; Benazzi, 2001c). Most present-study BP-II patients had
had more than one episode of hypomania. Often, family members or close friends
supplemented the clinical information during the interview. Study variables are
reported in Table 6.1.
144 F. Benazzi



Statistics
Means were compared with the t-test, proportions were compared by the two-
sample test of proportion. Univariate and multivariate logistic regression was
used to study associations. Two-way ANOVA was used to study interaction.
Maximum-likelihood logit estimation was used for discriminant analysis.
STATA Statistical Software, Release 7, was used (Stata Corporation, College
Station, TX, USA). P-values were two-tailed, and the probability level was
P < 0.05.


Results
Frequency of AD was 182/433 (42.0%). AD versus non-AD (Table 6.1) had
significantly more BP-II, females, lower age, lower age of onset, more axis I
comorbidity, fewer psychotic features, more DMX, more BP family history, and
more hypomanic and atypical symptoms.
To test if AD associations with female gender, early onset, DMX, and BP family
history were specific features of AD, or were due to its association with BP-II
(which could be a confounding factor, as it was associated with all these variables),
multivariate logistic regression controlled for BP-II was used. Logistic regression
of AD versus females gender found OR ¼ 1.9, z ¼ 3.2, P ¼ 0.001, and when con-
trolled for BP-II found OR ¼ 1.9, z ¼ 2.9, P ¼ 0.003. Logistic regression of AD
versus onset found OR ¼ 0.9, z ¼ À5.4, P ¼ 0.000, and when controlled for BP-II
found OR ¼ 0.9, z ¼ À4.2, P ¼ 0.000. Logistic regression of AD versus DMX found
OR ¼ 3.0, z ¼ 5.5, P ¼ 0.000, and when controlled for BP-II found OR ¼ 2.4,
z ¼ 4.2, P ¼ 0.000. Logistic regression of AD versus BP family history found
OR ¼ 2.1, z ¼ 3.0, P ¼ 0.002, and when controlled for BP-II found OR ¼ 1.6,
z ¼ 1.8, P ¼ 0.062. To test if the association between AD and axis I comorbidity
was specific for AD, or was instead related to a common early age of onset,
multivariate logistic regression was used. Logistic regression of AD versus axis I
comorbidity found OR ¼ 1.6, z ¼ 2.5, P ¼ 0.011, and when controlled for onset,
found OR ¼ 1.4, z ¼ 1.6, P ¼ 0.093.
To test if the association between AD and BP-II was related to early onset (both
had an early onset), multivariate logistic regression was used. Logistic regression of
AD versus BP-II found OR ¼ 3.3, z ¼ 5.5, P ¼ 0.000, and when controlled for
onset, found OR ¼ 2.5, z ¼ 4.0, P ¼ 0.000. Two-way ANOVA was also used to
find if the lower age of onset in AD versus non-AD was related to an interaction
between AD and BP-II. The result was that AD had F ¼ 23.2, P ¼ 0.000, BP-II had
F ¼ 26.9, P ¼ 0.0000, and there was an interaction between AD and BP-II (F ¼ 4.7,
P ¼ 0.0306). Discriminant analysis of BP-II, female gender, early onset, DMX, and
145 Atypical depression and bipolar spectrum



Table 6.2 Discriminant analysis of bipolar II (BP-II), female gender, early onset, major
depressive episode with more than two concurrent hypomanic symptoms (depressive
mixed state; DMX), and BP family history for predicting atypical depression (n ¼ 433)

Variable Coefficient Z P

BP-II 0.4 1.3 0.188
Female 0.5 2.0 0.046
Onset “0.0 “3.4 0.001
DMX 0.6 2.1 0.031
BP family history 0.2 0.9 0.329


BP family history for predicting AD (n ¼ 433 sample) is presented in Table 6.2. Results
found that AD was significantly associated only with female gender, early onset, and
DMX. BP-II AD versus UP AD (Table 6.3) had significantly lower age of onset, more
recurrences, more depression chronicity, more depressive mixed state, more BP family
history, and more irritability. AD symptom frequency was not significantly different.
Sensitivity, specificity, correctly classified, and receiver-operating characteristics
(ROC) area of AD, BP family history, early onset of first MDE, many MDE
recurrences (> 4), and DMX for predicting BP-II diagnosis, by logistic regression,
are presented in Table 6.4. Results showed that AD had a high specificity for
predicting BP-II, second only to BP family history. Discriminant analysis of AD,
BP family history, early onset of first MDE, many MDE recurrences, and DMX for
predicting BP-II diagnosis are presented in Table 6.5. Results showed that AD was
a near-significant predictor of BP-II compared to the other variables.
Associations among AD symptoms in the whole sample are presented in
Table 6.6. Mood reactivity was significantly associated with all AD symptoms,
apart from leaden paralysis (4/5). All the other AD symptoms were significantly
associated with each other. Associations among AD symptoms in the BP-II sub-
sample are presented in Table 6.7. Mood reactivity was significantly associated
with 3/5 AD symptoms. The other AD symptoms were often, but not always,
significantly associated with each other (7/10). Associations among AD symptoms
in the UP subsample are presented in Table 6.8. Mood reactivity was not signifi-
cantly associated with any AD symptom. The other AD symptoms were often, but
not always, significantly associated with each other (8/10).
A comparison of AD symptoms between mood-reactive (MR) and non-mood-
reactive (N-MR) MDE patients is presented in Table 6.9. Findings showed that AD
symptoms were significantly more common in MR MDE.
Given the strong significant association between BP-II and AD, the association
between BP-II and mood reactivity was tested. Logistic regression of BP-II versus
mood reactivity found OR ¼ 2.0, z ¼ 2.6, P ¼ 0.009.
146 F. Benazzi



Table 6.3 Comparisons between bipolar II atypical depression (BP-II AD) and unipolar
(UP) AD

BP-II UP
(n ¼ 138) (n ¼ 44)

Variable: mean (SD), % T/Z df P

Female gender 73.9 75.0 “0.1 0.8846
Age at index MDE (years) 39.9 (13.0) 41.1 (14.9) “0.5 180 0.6077
Age at onset of first MDE (years) 21.3 (9.8) 25.2 (12.4) “2.1 180 0.0329
> 4 MDEs 81.8 59.0 3.0 0.0020
MDE symptoms > 2 years 50.0 27.2 2.6 0.0080
Axis I comorbidity 59.4 54.5 0.5 0.5661
MDE severity by GAF 50.9 (7.9) 52.7 (7.8) “1.3 180 0.1885
Psychotic features 3.6 0.0 1.2 0.2019
> 2 hypomanic symptoms 68.8 45.4 2.8 0.0051
during index MDE
Bipolar (I þ II) family history 59.7 27.2 3.7 0.0002
Irritability 66.6 40.9 3.0 0.0024
Psychomotor agitation 33.3 27.2 0.7 0.4494
Distractibility 79.7 65.9 1.8 0.0606
More talkativeness 25.3 13.6 1.6 0.1054
Racing/crowded thoughts 78.2 65.9 1.6 0.0998
Mood reactivity 100 100
Weight gain 39.8 27.2 1.5 0.1313
Overeating 47.8 43.1 0.5 0.5863
Hypersomnia 62.3 68.1 “0.6 0.4860
Leaden paralysis 71.7 61.3 1.3 0.1933
Interpersonal rejection sensitivity 84.0 88.6 “0.7 0.4554
n atypical symptoms 4.0 (1.0) 3.8 (0.9) 1.1 180 0.2386

MDE, major depressive episode; GAF, Global Assessment of Functioning.


Conclusions
Frequency of AD was 42.0%, in line with the findings of Nierenberg et al. (1998) in
a large UP sample. AD versus non-AD had significantly more BP-II, more females,
lower age of onset, more axis I comorbidity, fewer psychotic features, more DMX,
more BP family history, more hypomanic and atypical symptoms, in line with
previous reports (Kendler et al., 1996; Rabkin et al., 1996; Levitan et al., 1997;
Sullivan et al., 1998; Sotsky and Simmens, 1999; McGrath et al., 2000; Williamson
et al., 2000; Benazzi and Akiskal, 2001). More BP-II in AD was reported in some
recent studies (Angst, 1998; Perugi et al., 1998; Agosti and Stewart, 2001; Angst
147 Atypical depression and bipolar spectrum



Table 6.4 Sensitivity (SE), specificity (SP), correctly classified (CC), and receiver operating
characteristics (ROC) area of atypical depression (AD), positive family history of bipolar (I þ II)
disorders, early-onset first major depressive episode (MDE), many MDE recurrences (> 4),
MDE with more than two concurrent hypomanic symptoms (depressive mixed state: DMX)
for predicting BP-II diagnosis, by logistic regression (n ¼ 405)

Variable, % SE SP CC ROC

Bipolar family history 56.6 79.8 66.3 0.68
AD 53.5 76.2 62.7 0.64
DMX 59.3 73.1 64.9 0.66
> 4 MDEs 80.9 40.2 64.4 0.60
Early onset 86.2 39.0 67.0 0.70




Table 6.5 Discriminant analysis of atypical depression (AD), positive family history of bipolar
(I þ II) disorders, early-onset first major depressive episode (MDE), many MDE recurrences
(> 4), MDE with more than two concurrent hypomanic symptoms (depressive mixed state:
DMX) for predicting bipolar II diagnosis (n ¼ 405)

Variable Coefficient Z P

Bipolar family history 1.2 3.5 0.000
AD 0.6 1.9 0.054
DMX 0.6 1.9 0.049
> 4 MDEs 0.7 1.8 0.062
Early onset “0.0 “3.9 0.000




Table 6.6 Associations among atypical depression symptoms in the whole sample (n ¼ 405),
by univariate logistic regression (odds ratio, *P < 0.05; **P < 0.01)

Mood Weight Leaden
reactivity gain Overeating Hypersomnia paralysis

Weight gain 3.1*
Overeating 9.3** 83.7**
Hypersomnia 4.6** 8.8** 9.4**
Leaden paralysis 1.5 3.0** 3.5** 2.6**
Interpersonal rejection
sensitivity 2.2** 2.0* 1.7* 2.4** 2.4**
148 F. Benazzi



Table 6.7 Associations among atypical depression symptoms in the bipolar II subsample
(n ¼ 241), by univariate logistic regression (odds ratio, *P < 0.05; **P < 0.01)

<<

. 25
( 68 .)



>>