A prediction rule for diagnosing hypomania

Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 317–322

Contents lists available at ScienceDirect

Progress in Neuro-Psychopharmacology & Biological Psychiatry j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / p n p b p

A prediction rule for diagnosing hypomania Franco Benazzi ⁎ Hecker Psychiatry Research Center, Forli (IT), a University of California at San Diego Collaborating Center, San Diego, USA Department of Psychiatry, University of Szeged, Szeged, Hungary Department of Psychiatry, National Health Service, Forli, Italy

a r t i c l e

i n f o

Article history: Received 12 October 2008 Received in revised form 7 December 2008 Accepted 15 December 2008 Available online 24 December 2008 Keywords: Bipolar II disorder Diagnosis Hypomania Overactivity Prediction rule Symptoms

a b s t r a c t Background: Missing the diagnosis of past hypomania, and thus of bipolar II disorder, is common. Study aim was to find a ‘prediction rule’ for facilitating the diagnosis of past hypomania. Methods: In an outpatient psychiatry private practice (non-tertiary care), a consecutive sample of 275 bipolar II disorder (BP-II) remitted patients, and consecutive, independent, sample of 138 major depressive disorder (MDD) remitted patients, had been interviewed for different study goals during follow-up visits by a senior bipolar-trained psychiatrist. Using the Structured Clinical Interview for DSM-IV, modified and validated by Benazzi and Akiskal [Benazzi F (2007). Lancet 369: 935–945] to improve the probing for past hypomania, patients had been questioned about the most common symptoms and duration of recent threshold and subthreshold hypomanic episodes. The sample was retrospective in nature. A prediction rule was tested. This is a score resulting from the sum of the weighted scores of each hypomanic symptom which was an independent predictor of hypomania. Its cutoff score for discriminating hypomania was based on the highest figure of correctly classified hypomanias and on the most balanced combination of sensitivity and specificity. A second, independent sample of 138 BP-II and 71 MDD remitted outpatients was tested to replicate the findings. Results: By univariate logistic regression, hypomanic symptoms distinguishing BP-II and MDD included ‘increase in goal-directed activity’ (overactivity) (OR = 28.3), ‘elevated mood’ (OR = 14.9), ‘increased talkativeness’ (OR = 9.2), ‘inflated self-esteem’, ‘decreased need for sleep’, ‘excessive risky activities’, and ‘irritable mood’. By multivariable logistic regression, the independent predictors of hypomania resulted ‘increase in goal-directed activity’ (OR = 14.9, weighted score = 15), ‘elevated mood’ (OR = 7.5, weighted score = 7), ‘increased talkativeness’ (OR = 3.6, weighted score = 4); ‘irritable mood’, ‘inflated self-esteem’, ‘decreased need for sleep’, and ‘excessive risky activities’ had ORs between 2.04 and 2.39, with a weighted score = 2. The prediction rule showed that the cutpoint score ≥ 21 had the highest figure of correctly classified hypomanias (88%, ROC area = 0.94), showing the most balanced combination of sensitivity (87%) and specificity (89%). This prediction rule, tested on the second sample, found that the same cutoff score ≥ 21 correctly classified the highest figure of hypomanias (94%, ROC area = 0.97), showing the most balanced combination of sensitivity (93%) and specificity (95%). To cross this cutoff score, overactivity was always required (as the sum of the scores of elevated mood and of the other symptoms did not reach this cutoff). However, scores 10 to 20 correctly classified only slightly lower figures of hypomanias. Conclusions: A prediction rule for hypomania was tested. The scores of overactivity plus at least some hypomanic symptom (among elevated mood, irritability, inflated self-esteem, less sleep, talkativeness, excessive risky activities) correctly classified 88% of hypomanias. Instead, elevated mood without overactivity, plus even all the other symptoms, did not reach the best figure of correctly classified. However, lower cutoff scores, up to 10, classified slightly lower figures of hypomanias, but with less balanced combinations of sensitivity and specificity. These findings may have diagnostic utility, because BP-II versus MDD is likely to be a more severe disorder. This prediction rule, if replicated and fine-tuned in different settings, may help clinicians better probing past hypomania, thus reducing the common misdiagnosis of BP-II as MDD. © 2008 Elsevier Inc. All rights reserved.

Abbreviations: BP-II, bipolar II disorder; DSM-IV-TR, diagnostic and statistical manual of mental disorders, forth edition, text revision; GAF, Global Assessment of Functioning scale; ICD-10, the ICD-10 classification of mental and behavioural disorders; MDD, major depressive disorder; MDE, major depressive episode; OR, odds ratio; 95%CI, 95% confidence intervals; ROC, Receiver-Operating-Characteristics analysis; SCID-CV, Structured Clinical Interview for DSM-IV Axis I Disorders-Clinician Version; SD, standard deviation. ⁎ Via Pozzetto 17, 48015 Castiglione Cervia RA, Italy. Tel.: +39 335 6191 852; fax: +39 054 330 069. E-mail address: [email protected]. 0278-5846/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.pnpbp.2008.12.007

318

F. Benazzi / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 317–322

1. Introduction Bipolar II disorder (BP-II) is defined, according to DSM-IV-TR (American Psychiatric Association, 2000) by recurrent hypomanic and major depressive episodes (MDE). The DSM-IV-TR ‘prototypical’ symptom (its criterion A, always to be present) is elevated mood, and/or irritable mood (mood change). Instead, ICD-10 (the WHO Classification of Mental and Behavioral Disorders) criteria for hypomania require both elevated mood (but not irritable mood) and increased activity (overactivity) (World Health Organization, 1992). In DSM-IV-TR instead, overactivity is one of the criterion B symptoms, optional and of similar weight. DSM-IV-TR major depressive disorder (MDD) (i.e. single or recurrent MDE) is distinguished by no history of mania or hypomania. Hypomania is thus the core feature of BP-II. ‘Classic’ diagnostic validators (e.g. Robins and Guze, 1970) such as family history (reviewed by Goodwin and Jamison, 2007; Benazzi, 2007a,b; Vieta and Suppes, 2008) and diagnostic stability (Coryell et al., 1995; Akiskal et al., 1995; Suppes et al., 2005; Vieta and Suppes, 2008) support the diagnostic validity of the DSM-IV-TR category of BP-II (ICD10 has no diagnostic criteria for BP-II). Several different clinical features of BP-II versus MDD also support the diagnostic utility (Kendell and Jablensky, 2003) of DSM-IV-TR BP-II. BP-II, versus MDD is likely to be a more severe disorder, e.g. likely to have a more unstable course, more recurrences, a longer duration of illness, more suicide attempts, more substance abuse and other axis I disorders, more complex treatments (Akiskal et al., 1995; Judd et al., 1998, 2003; Perugi et al., 2003; Suppes et al., 2005; Akiskal et al., 2006a; Rihmer, 2007; Benazzi, 2007c; Judd et al., 2008; Vieta and Suppes, 2008). The main clinical manifestation of BP-II is syndromal and subsyndromal depression, while hypomania (often lasting only days to weeks) occupies only a small proportion of its symptomatic periods (which are 40 to 50% of the follow-up weeks) (Judd et al., 2003). Thus, as BP-II patients usually present for treatment of depression and not for hypomania (which is often a pleasant period of overfunctioning) (Hecker, 1898, English translation by Koukopoulos, 2003), misdiagnosing BP-II as MDD is common (Goodwin and Jamison, 2007; Benazzi, 2008a). Diagnosing past hypomania needs a skilful probing. Semi-structured interviewing by trained clinicians has been shown to outperform structured interviewing by correctly classifying significantly more BP-II (Dunner and Tay, 1993; Simpson et al., 2002; Regeer et al., 2002; Akiskal and Benazzi, 2005; Kessler et al., 2006; Angst, 2007; Kessler, 2007). The DSM-IV-TR stem criterion A for hypomania, i.e. mood change (elevated and/or irritable mood), has been shown to facilitate the misdiagnosis of BP-II as MDD by its stem question skip-out instruction (i.e. no history of mood change, then no bipolar disorder) (Akiskal et al., 1977; Benazzi, 2003a; Benazzi, 2008a; Angst, 2008). The priority of mood change has not been validated. Kraepelin did not set any priority among its basic domains of ‘manic-depressive insanity’, i.e. elevated mood, thinking, and activity. A growing body of empirical evidence suggests that overactivity (‘increase in goal-directed activity’) may have at least the same priority of mood change as stem criterion for hypomania. As overactivity is an observable behavior easier to remember than emotions, it can be more useful than mood change for better probing past hypomania (Angst et al., 2003, 2005; Akiskal et al., 2006b; Benazzi, 2007a; Angst, 2008; Cassidy et al., 2008). Furthermore, the ‘core’ symptoms of hypomania have not been validated (Goodwin and Guze (1996) suggested ‘euphoria, hyperactivity, logorrhea’). Several DSM-IV-TR and ICD-10 ‘hypomanic’ symptoms show little specificity, being found in many other psychiatric disorders, such as irritability, psychomotor agitation, risky activities, and distractibility. Study aim was to find a simple ‘prediction rule’ to improve the probing for past hypomania, and thus reducing the misdiagnosis of BP-II as MDD. 2. Methods Detailed study methods can be found in previous reports (Benazzi and Akiskal, 2006; Benazzi, 2006; Benazzi, 2008b). Setting: An outpatient psychiatry private practice (a non-tertiary care), first or

second line of treatment of depression in the wealthy Italian EmiliaRomagna region. Most people can afford a fee-for-service private visit in this region, limiting a possible selection bias. Interviewer: The author (FB), a clinical (26 years in practice) and mood disorder research psychiatrist. Population: A consecutive remitted sample of 275 BP-II, and a consecutive, independent, remitted sample of 138 MDD outpatients, previously recorded for different study goals. Exclusion criteria: comorbid substance-related disorders and borderline personality disorder (anyway uncommon in the study setting [Benazzi, 2000]), because confounding the diagnosis of hypomania (Akiskal et al., 2006b); cognitive disorders (assessed by clinical evaluation using a semi-structured interview based on DSM-IV-TR criteria); clinically significant (i.e. impairing functioning) general medical illnesses (assessed by clinical evaluation). The interviews had been conducted in a state of remission during follow-up visits (which, at the beginning of the remission, included visits every other week), in order to overcome possible biases related to a current episode. Remission was assessed clinically and by the GAF (Global Assessment of Functioning scale), requiring a score N80 for at least one month (as functional remission may not follow symptom remission). Patients had been previously interviewed about history of threshold and subthreshold hypomanic episodes (defined below). Often (in around 70% of cases, no exact figure recorded), family members or close friends supplemented clinical information during the diagnostic interviews. A second, independent sample of 138 BP-II and 71 MDD remitted outpatients, previously described (Benazzi, 2006), was tested to replicate the findings on the ‘prediction rule’. This sample had been assessed by the same interviewer following the same methods, in the same setting, at a different time-point, and for different purposes. Data were recorded between 2002 and 2006. All patients meeting study criteria were included, combining data from four independent samples (i.e. all recorded patients assessed during remission for a study goal, assessed with the same methodology). There was therefore no selection bias. Among these large independent samples, three were chosen and combined at random to become the first sample, and the remaining sample (after the random grouping of the first three samples) was chosen as the second independent sample. Data were recorded during follow-up interviews. At random, according to the number of patients in the waiting room of the clinic, patients had been interviewed (i.e. when there was enough time to carry out correctly the interview). All interviewed patients with a mood disorder (BP-II or MDD) were included in the samples. The study was conducted according to the Helsinki declaration, all procedures were carried out with the adequate understanding and written consent of the subjects, formal approval to conduct past and future studies by the present study methods, and by recording clinical variables representing an in-depth clinical evaluation, had been obtained from local ethics review board. 2.1. Diagnostic interviewing The Structured Clinical Interview for DSM-IV Axis I DisordersClinician Version was used (First et al., 1997), modified and validated by Benazzi and Akiskal to improve the probing for past hypomania (Benazzi and Akiskal, 2003; Akiskal and Benazzi, 2005) (FB inter-rater reliability k for diagnosing past hypomania by the SCID-CV was 0.73 [Benazzi, 2003b]). Remission was assessed by the GAF (in the SCIDCV). As BP-II patients rarely present for treatment of a hypomanic episode, interview was based on recall of symptoms of past hypomanic episodes, as in most studies on hypomania (American Psychiatric Association, 2000; Goodwin and Jamison, 2007). During follow-up visits, patients had been re-interviewed and rediagnosed by the SCID-CV. Relatively few variables were recorded because of the limited time of a follow-up visit in this busy private practice. The SCID-CV is partly semi-structured and is based on clinical

F. Benazzi / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 317–322 Table 1 Comparison of bipolar II disorder (BP-II) versus major depressive disorder (MDD) on past episodes of hypomanic symptoms, by univariate logistic regression Variables: mean (SD), %

BP-II, n = 275 MDD, n = 138 OR (95%CI)

Age, years Females

40.4 (10.9) 67.2%

39.9 (11.4) 70.8%

Hypomanic symptoms: Elevated mood Irritable mood Inflated self-esteem Decreased need for sleep Increased talkativeness Racing thoughts Distractibility Increase in goal-directed activity Excessive risky activities

81.4% 65.9% 71.0% 52.9% 66.4% 66.4% 61.3% 92.0% 57.6%

22.7% 53.2% 22.0% 14.8% 17.6% 62.6% 56.7% 28.8% 28.6%

1.00 (0.98–1.02) 0.84 (0.54–1.32)

14.87 (8.99–24.60)** 1.69 (1.11–2.57)* 8.67 (5.35–14.05)** 6.46 (3.80–10.98)** 9.23 (5.55–15.33)** 1.17 (0.64–2.16) 1.20 (0.66–2.18) 28.30 (15.96–50.20)** 3.38 (2.17–5.27)**

(OR = odds ratio; 95%CI = 95% confidence intervals; * = p b 0.05; ** = p b 0.01).

evaluation (differently from the epidemiological structured interviews, using ‘fixed’ questions administered by lay interviewers). Wording of the questions can be changed to improve and to check the understanding by the interviewee. This is an important advantage versus fully structured interviews, because this interview method has been shown to reduce the rate of false negatives BP-II (Dunner and Tay, 1993; Simpson et al., 2002; Benazzi, 2003a). The skip-out instruction of the stem question on history of mood change (elevated/irritable mood) was not followed, in order to assess all past hypomanic symptoms, following Dunner and Tay (1993). A minimum duration of threshold and subthreshold hypomania of 2 days was required, instead of the DSM-IV-TR minimum duration of 4 days, on the basis of data supporting this cutoff, i.e. no differences on diagnostic validators between BP-II whose hypomania lasted 2 to 3 days and DSM-IV-TR BP-II (Benazzi, 2001; Angst et al., 2003; Judd et al., 2003; Benazzi and Akiskal, 2006; Angst, 2008; Vieta and Suppes, 2008); indeed, the DSM-IV-TR cutoff of 4 days is not data-based (Dunner, 2003). Around 30% of the BP-II samples met the 2-day minimum duration, i.e. around 70% met full DSM-IV-TR criteria for hypomania. 2.2. Definition of hypomanic episode Questioning was about the most common symptoms and duration of the most recent threshold and subthreshold hypomanic episodes. Threshold hypomanic episodes met DSM-IV-TR criteria (apart from duration); subthreshold hypomanic episodes required at least 2 symptoms, following Judd and Akiskal (2003). All the symptoms of the threshold and subthreshold hypomanic episodes were assessed and recorded in the BP-II and MDD samples. 2.3. Testing study aim ‘Prediction rules’ are used in many fields of medicine (Chunilal et al., 2003). On the basis of the sum of the scores given to specific risk factors for an outcome, the risk of the outcome is quantified (Chunilal et al., 2003). The prediction rule, in this case, was a score resulting from the sum of the scores of each hypomanic symptom which had been shown to be an independent predictor of hypomania. A cutoff score was then chosen on the basis of the highest figure of correctly classified hypomanias together with the most balanced combination of sensitivity and specificity. The prediction rule was then tested on an independent sample, as a first step toward its validation. 2.4. Statistics Univariate and multivariable logistic regression were used to study associations and to control for confounding. A prediction rule for hypomania was created (Chunilal et al., 2003). By multivariable logistic

319

regression, hypomanic symptoms (the independent variables) found significantly different between BP-II and MDD were regressed versus hypomania (the dependent variable), and independent predictors of hypomania were found. A weight was given to each hypomanic symptom which was an independent predictor of hypomania. For rounding the ORs, a weight was given to each hypomanic symptom 1) by rounding its odds ratio (OR) to the nearest number with a single digit, and 2) also by rounding the OR to the nearest even integer number (Chunilal et al., 2003; Rothman, 2002; Armitage et al., 2002). According to Rothman (2002, p 51) ‘when rounding a number ending in 5, it is customary rounding upward, but it is preferable to use an unbiased strategy, such as rounding to the nearest even number’. As the results obtained by the two methods of rounding the OR were almost identical, the simple method was reported, i.e. rounding the OR to the nearest even integer number. The sum of the scores of each independent predictor symptom was then included in a ROC (Receiver-OperatingCharacteristics) analysis to find the best cutoff (cutpoint) score discriminating hypomania. This cutoff score was chosen on the basis of the highest percentage of correctly classified hypomanias together with the most balanced combination of sensitivity and specificity. STATA Statistical Software, Release 10.1, was used (StataCorp, College Station, TX, USA, 2008). P values were two-tailed, alpha level was set at 0.05. 3. Results Comparisons between BP-II and MDD by univariate logistic regression (Table 1) showed that the hypomanic symptoms with significantly larger ORs in BP-II included, in decreasing order of strength (i.e. highest ORs), ‘increase in goal-directed activity’ (overactivity) (OR = 28.3), ‘elevated mood’ (OR = 14.9), ‘increased talkativeness’ (OR= 9.2), ‘inflated self-esteem’, ‘decreased need for sleep’, ‘excessive risky activities’, and ‘irritable mood’. ‘Racing thoughts’ and ‘distractibility’ did not result significantly different. The core symptoms of hypomania were then investigated by multivariable logistic regression, including in the model the hypomanic symptoms significantly different between BP-II and MDD (Table 2). It resulted that independent predictors of hypomania were, in decreasing order of strength, ‘increase in goal-directed activity’ (OR= 14.9, weighted score = 15), ‘elevated mood’ (OR= 7.5, weighted score = 7), ‘increased talkativeness’ (OR = 3.6, weighted score = 4); ‘irritable mood’, ‘inflated self-esteem’, ‘decreased need for sleep’, and ‘excessive risky activities’ had ORs between 2.04 and 2.39, with a weighted score = 2 (Rothman, 2002). Table 3 shows the testing of the prediction rule for hypomania. It shows the percentage of correctly classified hypomanias according to several cutpoints of the score resulting from the sum of the scores of the independent predictor hypomanic symptoms. The score range was 0–34, its median was 25.5. The highest figure of correctly classified (88%), and the most balanced combination of sensitivity (87%) and specificity (89%), was shown by a cutpoint score ≥ 21. Scores between 10 and 21 had slightly lower figures of correctly classified, but lower specificity. Scores above 21 had a progressive decrease of correctly classified, while increasing specificity. The validity of this prediction

Table 2 Multivariable logistic regression of hypomania (dependent variable) versus hypomanic symptoms (predictor variables), weighted by rounding odds ratios (OR) Predictor variables of hypomania

OR (95%CI)

Weighted score

Elevated mood Irritable mood Inflated self-esteem Decreased need for sleep Increased talkativeness Increase in goal-directed activity Excessive risky activities

7.53 (3.66–15.48)⁎* 2.04 (1.02–4.06)* 2.33 (1.16–4.71)* 2.39 (1.09–5.26)* 3.60 (1.72–7.52)** 14.95 (7.06–31.65)** 2.35 (1.15–4.79)*

7 2 2 2 4 15 2

Log likelihood = −110.25, Likelihood ratio chi-squared test = 288.43, p b 0.0001. (OR = odds ratio; 95%CI = 95% confidence intervals; * = p b 0.05; ** = p b 0.01).

320

F. Benazzi / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 317–322

Table 3 Prediction rule for classifying past hypomania: cutpoints of the sum of the weighted scores of hypomanic symptoms (independent predictors of hypomania) in the entire first sample, correctly classifying hypomania Cutpoint

SE%

SP%

CC%

≥0 ≥2 ≥4 ≥6 ≥7 ≥8 ≥9 ≥10 ≥11 ≥13 ≥15 ≥17 ≥19 ≥21 ≥22 ≥23 ≥24 ≥25 ≥26 ≥27 ≥28 ≥30 ≥32 ≥34

100.0 99.6 98.8 98.5 98.5 98.1 97.7 97.4 96.6 95.9 94.0 93.3 90.7 87.4 81.1 80.3 75.5 75.1 72.5 61.8 60.7 51.1 32.2 13.3

0.0 10.6 36.3 50.0 53.7 55.3 57.5 63.6 63.6 67.4 69.7 75.0 84.8 88.6 91.6 92.4 93.9 96.2 96.2 96.9 97.7 98.4 99.2 99.2

67.1 70.4 78.3 82.5 83.8 84.0 84.5 86.3 85.8 86.5 86.0 87.3 88.8 87.8 84.5 84.3 83.5 82.0 80.3 73.3 72.8 66.6 54.2 41.4

(CC = correctly classified; SE = sensitivity; SP = specificity).

rule was supported by a high area under the ROC curve of 0.94 (95% CI = 0.91–0.96) (the ROC curve range is 0.5–1, the closer it is to 1 the highest it is the discriminative power of a test). As a first step toward the validation of this prediction rule, the prediction rule was tested on an independent sample, in the same setting, previously recorded for different goals. The sample had 138 patients with BP-II and 71 patients with MDD. BP-II had a mean (SD) age of 39.0 (9.8) years, females were 67.3%. MDD had a mean (SD) age of 39.2 (10.6) years, females were 77.1%. It was found that a cutoff score≥21 identified the highest figure of correctly classified hypomanias (94%), and the most balanced combination of sensitivity (93%) and specificity (95%). The area under the ROC curve was 0.97 (95%CI=0.95–0.99). Scores between 10 and 21 had similar figures of correctly classified, but lower specificity. Scores above 21 had a progressive decrease of correctly classified, while increasing specificity. 4. Discussion 4.1. The core symptoms of hypomania It should be noted that subthreshold hypomania was not uncommon in the study samples of MDD, as previously reported (Cassano et al., 2004; Angst, personal communication, 2008; Benazzi, 2009), a finding which, by itself, questions the diagnostic validity of the current criteria for hypomania. This study result replicates, in a clinical setting, recent epidemiological findings about the frequency of subthreshold hypomania, which was higher than that of threshold hypomania (Angst et al., 2003; Judd and Akiskal, 2003; Merikangas et al., 2007). By univariate analysis, the most distinguishing hypomanic symptoms (i.e. those having the highest ORs) were overactivity (‘increase in goaldirected activity’, OR = 28.3), ‘elevated mood’ (OR = 14.8), and ‘increased talkativeness’ (OR= 9.2); inflated self-esteem, decreased need for sleep, excessive risky activities, and irritable mood were other, less strongly associated distinguishing symptoms. By multivariate analysis it resulted that the ‘core’ symptoms of hypomania (i.e. its independent predictors) were overactivity (OR = 14.9), ‘elevated mood’ (OR = 7.5), ‘increased talkativeness’ (OR = 3.6), ‘excessive risky activities’ (OR= 2.3), ‘decreased need for sleep’ (OR= 2.3), ‘inflated self-esteem’ (OR= 2.3), and irritability

(OR= 2.0). This cluster of symptoms fits the three basic domains of hypomania described by Kraepelin, i.e. excited activity (DSM-IV-TR ‘increase in goal-directed and risky activities’, ‘increased talkativeness’, ‘decreased need for sleep’), excited mood (DSM-IV-TR ‘elevated and irritable mood’), and excited thinking (DSM-IV-TR ‘increased talkativeness’, the outward manifestation of mental overactivity). About this last symptom, when mental overactivity was represented by a question asking about ‘racing thoughts’, racing thoughts did not show discriminative power, apparently negating the interpretation of increased talkativeness as a sign of mental overactivity. However, talkativeness is an unquestionable observable behavior which reflects accelerated thinking, while, from a patient's view, ‘racing thoughts’ may represent a wide range of mental overactivities not necessarily leading to more talkativeness (e.g. the ruminations of depression). Study results found that the core symptoms of DSM-IV-TR hypomania followed historical descriptions focusing on excited mood, thinking, and activity (Falret, 1854; Kahlbaum, 1882; Hecker, 1898; Kraepelin, 1913) (English translations by Sedler, 1983; Baethge et al., 2003; Koukopoulos, 2003; Barclay, 1921; respectively). This clinical picture of hypomania has diagnostic utility on the basis of its discrimination of a more severe disorder dominated by recurrent depression (i.e. BP-II) from a less severe, often recurrent depression (i.e. MDD) (see Introduction). 4.2. A prediction rule for hypomania A prediction rule was created on the basis of the above findings. The prediction rule for hypomania tested in Table 3 shows that the highest figure of correctly classified hypomanias (88%), and the most balanced combination of sensitivity (87%) and specificity (89%), was shown by a cutpoint score ≥ 21. For clinical practice, a person presenting for depression and probed for past hypomania, if showing a cutpoint score ≥ 21 of the sum of the scores of the core hypomanic symptoms, could have nearly 90% probability of having BP-II. This cutoff score could be crossed by the sum of the scores of overactivity plus some symptom of hypomania, but not by the sum of the scores of elevated mood without overactivity, plus all the other symptoms. However, lower scores, up to 10, correctly classified slightly lower figures of hypomanias, but with less balanced combinations of sensitivity and specificity. This finding runs against DSM-IV-TR diagnostic criteria for hypomania. While overactivity could be a stem criterion for hypomania according to recent studies (see Introduction), it is instead a criterion B in DSM-IV-TR, i.e. it does not necessarily has to be present like the criterion A (mood change). Because the sum of the scores of overactivity (which had the highest score) and just some other core hypomanic symptoms could cross the prediction rule cutoff score of 21, diagnosing BP-II could be much simplified by a probing more focused on overactivity and elevated mood (the sum of these two symptoms alone crossed the cutoff of 21). Furthermore, by following the prediction rule, overactivity may be present without elevated mood, but elevated mood needs the presence of overactivity to cross the 21 score cutoff. The prediction rule cutoff required, to be crossed, the presence of overactivity, or the presence of elevated mood and overactivity, suggesting that both symptoms could be a stem criterion A for hypomania (as suggested by the ICD-10, World Health Organization,1992), and also that overactivity may have an edge over elevated mood. However, Table 3 shows that lower scores, up to 10, correctly classified slightly lower figures of hypomanias, but with less balanced combinations of sensitivity and specificity, leaving still open the priority of overactivity on mood as stem criterion. 4.3. Toward a validation of a prediction rule for hypomania As a first step toward the validation of this prediction rule, the prediction rule was tested on an independent sample (Benazzi, 2006) which had been assessed by the same interviewer, following the same methods in the same setting, at a different time-point, and for

F. Benazzi / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 317–322

different purposes. The highest figure of correctly classified hypomanias (94%), and the most balanced combination of sensitivity (93%) and specificity (95%) was found by a cutoff score ≥ 21, replicating the findings in the first sample. Also, lower scores, up to 10, correctly classified slightly lower figures of hypomanias, but with less balanced combinations of sensitivity and specificity. Study findings show that a prediction rule could be a useful tool for a better (and quicker) probing for past hypomania. Its main focus on observable behaviors (i.e. overactivity, increased talkativeness, decreased need for sleep, excessive risky activities) increases its reliability (Simpson et al., 2002) and its chance of being remembered compared to past emotions (Benazzi and Akiskal, 2003). Replications by different groups in different settings are essential for fine-tuning a prediction rule of general utility for clinicians. That the present prediction rule was created in a nontertiary care setting may make it of more general utility, pending replications. Only apparently the present study has a similarity to the Bipolarity Index (Sachs, 2004), as this Index results in scores (yet to be validated, apart from its face validity) built on variables describing several classic features of bipolar disorder (e.g. mania/hypomania, family history, age at onset, high recurrence), and not on symptoms of mania or hypomania as in the present study. 5. Limitations The single interviewer (FB) was experienced in diagnosing BP-II (Benazzi, 2007a). Study samples were part of databases previously recorded for different study goals, making unlikely an interviewer's bias. The description of past hypomania may be limited by recall bias, but it is not possible to examine hypomania directly, in a large enough sample, because BP-II individuals rarely present to the clinician for hypomania (American Psychiatric Association, 2000; Goodwin and Jamison, 2007). Only long-term follow-ups could have the chance to observe hypomanic episodes (e.g. Judd et al., 2003). The present author does not have the resources to go beyond cross-sectional and retrospective studies. Assessment during remission, focus on recent episodes, and often (in around 70% of cases, no exact figure recorded) concurrent interviewing of key informants, should have reduced the impact of recall bias. A methodologic advantage of this paper is that BP-II and MDD patients were studied in a state of remission. This provides assurance that the phenomenology of the hypomania was not contaminated by depression, in particular its cognitive bias for remembering the ‘positive’ experiences of hypomania (Peselow et al., 1995). The systematic collection of data by validated instruments should also have reduced any unintended biases. Recent evidence (reviewed by Angst, 2008) has shown that probing for history of mood change as skip-out, stem criterion (i.e. no history of mood change, no bipolar disorder) can increase the false negatives BP-II. Assessing all past hypomanic symptoms (the present study method), following Dunner and Tay (1993), has been shown to increase the true positives BP-II. Compared to mood change, overactivity is an observable behavior (as well as several other core symptoms, i.e. ‘increased talkativeness’, ‘decreased need for sleep’, ‘excessive risky activities’) which is easier to remember by patients and key informants, while mood change could be seen by patients as normal mood fluctuations or, if presented as causing more or less severe problems, it could be seen as a sign of a severe mental illness which may be denied by patients (Akiskal and Benazzi, 2005; Angst, 2008). 6. Conclusions A prediction rule for hypomania was created. Overactivity plus some core hypomanic symptoms (especially elevated mood, talkativeness, decreased need for sleep, excessive risky activities) correctly classified around 90% of hypomanias, while elevated mood would require the presence of overactivity. However, lower scores correctly classified slightly lower figures, but with less balanced combinations

321

of sensitivity and specificity. The study findings were replicated in an independent sample. These findings may have diagnostic utility. BP-II is likely to be a more severe and a more difficult to treat disorder compared to MDD. This prediction rule, if replicated and fine-tuned in different settings, may help clinicians better probing past hypomania, thus reducing the current high misdiagnosis of BP-II as MDD (Benazzi, 2008a). References Akiskal HS, Djenderedjian AM, Rosenthal RH, Khani MK. Cyclothymic disorder: validating criteria for inclusion in the bipolar affective group. Am J Psychiatry 1977;134:1227–33. Akiskal HS, Maser JD, Zeller PJ, Endicott J, Coryell W, Keller M, et al. Switching from ‘unipolar’ to bipolar II. An 11-year prospective study of clinical and temperamental predictors in 559 patients. Arch Gen Psychiatry 1995;52:114–23. Akiskal HS, Benazzi F. Optimizing the detection of bipolar II disorder in outpatient private practice: toward a systematization of clinical diagnostic wisdom. J Clin Psychiatry 2005;66:914–21. Akiskal HS, Akiskal KK, Lancrenon S, Hantouche EG, Fraud FP, Gury C, et al. Validating the bipolar spectrum in the French National EPIDEP Study: overview of the phenomenology and relative prevalence of its clinical prototypes. J Affect Disord 2006a;96:197–205. Akiskal HS, Akiskal KK, Lancrenon S, Hantouche E. Validating the soft bipolar spectrum in the French National EPIDEP Study: the prominence of BP-II 1/2. J Affect Disord 2006b;96:207–13. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed. Washington, DC: American Psychiatric Association; 2000. text revision (DSM-IV-TR). Angst J. The bipolar spectrum. Br J Psychiatry 2007;190:189–91. Angst J. Bipolar disorder-methodological problems and future perspectives. Dialogues Clin Neurosci 2008;2:129–39. Angst J, Gamma A, Benazzi F, Ajdacic V, Eich D, Rossler W. Toward a re-definition of subthreshold bipolarity: epidemiology and proposed criteria for bipolar-II, minor bipolar disorders and hypomania. J Affect Disord 2003;73:133–46. Angst J, Adolfsson R, Benazzi F, Gamma A, Hantouche E, Meyer TD, et al. The HCL-32: towards a self-assessment tool for hypomanic symptoms in outpatients. J Affect Disord 2005;88:217–33. Armitage P, Berry G, Matthews JNS. Statistical methods in medical research. 4th edi. Oxford: Blackwell Science Ltd; 2002. Baethge C, Salvatore P, Baldessarini RJ. “On cyclic insanity” by Karl Ludwig Kahlbaum, MD: a translation and commentary. Harv Rev Psychiatry 2003;11:78–90. Benazzi F. Borderline personality disorder and bipolar II disorder in private practice depressed outpatients. Compr Psychiatry 2000;41:106–10. Benazzi F. Is 4 days the minimum duration of hypomania in bipolar II disorder? Eur Arch Psychiatry Clin Neurosci 2001;251:32–4. Benazzi F. Diagnosis of bipolar II disorder: a comparison of structured versus semistructured interviews. Prog Neuro-Psychopharmacol Biol Psychiatry 2003a;27:985–91. Benazzi F. Depression with racing thoughts. Psychiatry Res 2003b;120:273–82. Benazzi F. Does temperamental instability support a continuity between bipolar II disorder and major depressive disorder? Eur Psychiatry 2006;21:274–9. Benazzi F. Bipolar disorder – focus on bipolar II disorder and mixed depression. Lancet 2007a;369:935–45. Benazzi F. Is there a continuity between bipolar and depressive disorders? Psychother Psychosom 2007b;76:70–6. Benazzi F. Bipolar II disorder: epidemiology, diagnosis and management. CNS Drugs 2007c;21:727–40. Benazzi F. Misdiagnosis of bipolar II disorder as major depressive disorder. J Clin Psychiatry 2008a;69:501–2. Benazzi F. A relationship between bipolar II disorder and borderline personality disorder? Prog Neuropsychopharmacol Biol Psychiatry 2008b;32:1022–9. Benazzi F. Does hypomania distinguish bipolar II disorder from major depressive disorder? Psychother Psychosom 2009;78(1):55–8. Benazzi F, Akiskal HS. Refining the evaluation of bipolar II: beyond the strict SCID-CV guidelines for hypomania. J Affect Disord 2003;73:33–8. Benazzi F, Akiskal H. The duration of hypomania in bipolar-II disorder in private practice: methodology and validation. J Affect Disord 2006;96:189–96. Cassano GB, Rucci P, Frank E, Fagiolini A, Dell'Osso L, Shear MK, et al. The mood spectrum in unipolar and bipolar disorder: arguments for a unitary approach. Am J Psychiatry 2004;161:1264–9. Cassidy F, Yatham LN, Berk M, Grof P. Pure and mixed manic subtypes: a review of diagnostic classification and validation. Bipolar Disord 2008;10(1 Pt 2):131–43. Chunilal SD, Eikelboom JW, Attia J, Miniati M, Panju AA, Simel DL, et al. Does this patient have pulmonary embolism? JAMA 2003;290:2849–58. Coryell W, Endicott J, Maser JD, Keller MB, Leon AC, Akiskal HS. Long-term stability of polarity distinctions in the affective disorders. Am J Psychiatry 1995;152:385–90. Dunner DL, Tay KL. Diagnostic reliability of the history of hypomania in bipolar II patients and patients with major depression. Compr Psychiatry 1993;34:303–7. Dunner DL. Clinical consequences of under-recognized bipolar spectrum disorder. Bipolar Disord 2003;5:456–64. First MB, Spitzer RL, Gibbon M, Williams JBW. Structured clinical interview for DSM-IV axis I disorders-clinician version (SCID-CV). Washington, DC: American Psychiatric Press; 1997. Goodwin DW, Guze SB. Psychiatric diagnosis. 5th ed. New York: Oxford University Press; 1996.

322

F. Benazzi / Progress in Neuro-Psychopharmacology & Biological Psychiatry 33 (2009) 317–322

Goodwin FK, Jamison KR. Manic-depressive illness. 2nd ed. New York: Oxford University Press; 2007. Judd LL, Akiskal HS, Maser JD, Zeller PJ, Endicott J, Coryell W, et al. A prospective 12-year study of subsyndromal and syndromal depressive symptoms in unipolar major depressive disorders. Arch Gen Psychiatry 1998;55:694–700. Judd LL, Akiskal HS. The prevalence and disability of bipolar spectrum disorders in the U.S. population: re-analysis of the ECA database taking into account subthreshold cases. J Affect Disord 2003;73:123–31. Judd LL, Akiskal HS, Schettler PJ, Coryell W, Endicott J, Maser JD, et al. A prospective investigation of the natural history of the long-term weekly symptomatic status of bipolar II disorder. Arch Gen Psychiatry 2003;60:261–9. Judd LL, Schettler PJ, Akiskal HS, Coryell W, Leon AC, Maser JD, et al. Residual symptom recovery from major affective episodes in bipolar disorders and rapid episode relapse/recurrence. Arch Gen Psychiatry 2008;65:386–94. Kendell R, Jablensky A. Distinguishing between the validity and utility of psychiatric diagnoses. Am J Psychiatry 2003;160:4-12. Kessler RC, Akiskal HS, Angst J, Guyer M, Hirschfeld RM, Merikangas KR, et al. Validity of the assessment of bipolar spectrum disorders in the WHO CIDI 3.0. J Affect Disord 2006;96:259–69. Kessler RC. The global burden of anxiety and mood disorders: putting the European study of the epidemiology of mental disorders (ESEMeD) findings into perspective. J Clin Psychiatry 2007;68(suppl 2):10–9. Koukopoulos A. Ewald Hecker's description of cyclothymia as a cyclical mood disorder: its relevance to the modern concept of bipolar II. J Affect Disord 2003;73:199–205. Kraepelin E. Manic-depressive insanity and paranoia. Edinburgh: Livingstone E & S; 1921. Merikangas KR, Akiskal HS, Angst J, Greenberg PE, Hirschfeld RM, Petukhova M, et al. Lifetime and 12-month prevalence of bipolar spectrum disorder in the national comorbidity survey replication. Arch Gen Psychiatry 2007;64:543–52.

Peselow ED, Sanfilipo MP, Fieve RR. Relationship between hypomania and personality disorders before and after successful treatment. Am J Psychiatry 1995;152:232–8. Perugi G, Toni C, Travierso MC, Akiskal HS. The role of cyclothymia in atypical depression: toward a data-based reconceptualization of the borderline-bipolar II connection. J Affect Disord 2003;73:87–98. Regeer EJ, Rosso ML, ten Have M, Vollebergh W, Nolen WA. Prevalence of bipolar disorder: a further study in The Netherlands. Bipolar Disord 2002;4(Suppl 1):37–8. Rihmer Z. Suicide risk in mood disorders. Curr Opin Psychiatry 2007;20:17–22. Robins E, Guze SB. Establishment of diagnostic validity in psychiatric illness: its application to schizophrenia. Am J Psychiatry 1970;126:983–7. Rothman KJ. Epidemiology. Oxford: Oxford University press; 2002. Sachs G. Strategies for improving treatment of bipolar disorder: integration of measurement and management. Acta Psychiatr Scand Suppl 2004;422:7-17. Sedler MJ. Falret's discovery: the origin of the concept of bipolar affective illness. Am J Psychiatry 1983;140:1127–33. Simpson SG, McMahon FJ, McInnis MG, MacKinnon DF, Edwin D, Folstein SE, et al. Diagnostic reliability of bipolar II diagnosis. Arch Gen Psychiatry 2002;59:736–40. Suppes T, Mintz J, McElroy SL, Altshuler LL, Kupka RW, Frye MA, et al. Mixed hypomania in 908 patients with bipolar disorder evaluated prospectively in the Stanley Foundation Bipolar Treatment Network: a sex-specific phenomenon. Arch Gen Psychiatry 2005;62:1089–96. Vieta E, Suppes T. Bipolar II disorder: arguments for and against a distinct diagnostic entity. Bipolar Disord 2008;10(1 Pt 2):163–78. World Health Organization. The ICD-10 classification of mental and behavioural disorders. Geneva: World Health Organization; 1992.