Journal of Affective Disorders 170 (2015) 23–29

Contents lists available at ScienceDirect

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

Research report

Psychometric properties of the Hong Kong Chinese (Cantonese) TEMPS-A in medical students$ Chi Ming Leung a,n, Arthur D.P. Mak b, Yu Tao Xiang c, Sing Lee b, Connie T.Y. Yan a, Tony Leung b, Daniel Bessonov d, Kareen K. Akiskal d, Hagop S. Akiskal d a

Department of Psychiatry, Shatin Hospital, Hong Kong, China Department of Psychiatry, The Chinese University of Hong Kong, Hong Kong, China c Faculty of Health Sciences, University of Macau, Macao SAR, China d International Mood Center, University of California, San Diego, USA b

art ic l e i nf o

a b s t r a c t

Article history: Received 27 April 2014 Received in revised form 26 July 2014 Accepted 6 August 2014 Available online 27 August 2014

Background: The self-rated auto-questionnaire, the Temperament Scale of Memphis, Pisa, Paris and San Diego (TEMPS-A) is the latest development in the study of temperamental attributes. It has been used and validated in different cultures and countries. The current study aims at validating the Chinese (Cantonese) version of the TEMPS-A and comparing the psychometric properties of the long and short forms of the translated scale. Methods: The Chinese (Cantonese) version of TEMPS-A was prepared with the standard translation and back-translation method, and approved by the original authors (HSA & KKA). It was administered to medical students of the two local universities, and results were analyzed. Results: 613 valid questionnaires were returned. The Cronbach-Alpha coefficients for the depressive, cyclothymic, hyperthymic, irritable and anxious temperament subscales were 0.63, 0.82, 0.78, 0.80, and 0.84, respectively. The strongest correlation was observed between the cyclothymic and irritable temperaments (R ¼0.600). Factor analysis yielded one large composite (depressive and anxious) and four homogenous factors, cyclothymic, anxious, hyperthymic and irritable. A newly reconstituted 43item short form, based on methods suggested by the original authors yielded similar factor structure. Limitations: The narrow age range of subjects somewhat limits generalization of the results. However, external and concurrent validations against other validated scales have been demonstrated for the original English versions as well as against the most commonly used languages of the world; furthermore, such validation has also been demonstrated for Chinese (Mandarin). Conclusions: The Chinese (Cantonese) version of TEMPS-A and the reconstituted 43-item short form were found to have good internal consistency and factor structures comparable to those of other languages from diverse cultures across the planet. We propose that the Cantonese TEMPS-A is a useful tool for local use. & 2014 Elsevier B.V. All rights reserved.

Keywords: Temperaments TEMPS-A Chinese-Cantonese

1. Introduction The study of human temperaments in modern times was made a respectable scientific endeavor by Eysenck and Eysenck (1969). Many other instruments have been developed since then, of which the latest are those of Cloninger et al. (1994) and Akiskal and Akiskal (2005a). Temperaments refer to temporally stable behavioral traits with strong affective reactivity, has general appeal in empirically unraveling the kaleidoscopic profile of human behavioral liabilities and assets on the one hand (Akiskal and Akiskal, 2005b), which have ☆ Disclaimer statement: Authors have no sources of support or financial interest to report for this work. n Corresponding author. Tel.: þ 852 26367754; fax: þ 852 26475321. E-mail address: [email protected] (C.M. Leung).

http://dx.doi.org/10.1016/j.jad.2014.08.026 0165-0327/& 2014 Elsevier B.V. All rights reserved.

been buttressed by genetic data using different technologies (Kang et al., 2008; Gonda et al., 2011; Greenwood et al., 2012, 2013), and immediate clinical significance in their hypothesized relationship with affective disorders in a state-trait fashion on the other (Akiskal, 1981). Indeed, while circularity of bipolar disorder has intrigued generations of clinicians and researchers, the phenomenological continuity of cyclothymic temperament (Kretschmer, 1936; Akiskal et al., 1977, 1979; Akiskal and Mallya, 1987; Akiskal and Akiskal, 1992; Akiskal and Akiskal, 2005a) with bipolar disorders has vast implications in understanding bipolarity in terms of etiology and treatment formulation. Affective temperaments have since been operationalized (Akiskal and Mallya, 1987), the commonest instrument used being the self-rated auto-questionnaire, the Temperament Scale of Memphis, Pisa, Paris and San Diego (TEMPS-A). The TEMPS-A

24

C.M. Leung et al. / Journal of Affective Disorders 170 (2015) 23–29

was constructed based on both historical and clinical grounds (Kraepelin, 1921; Kretschmer, 1936; Akiskal et al., 1977,1979). The depressive, cyclothymic, hyperthymic and irritable temperaments in the original framework were further supplemented by the anxious axis (Akiskal, 1998; Akiskal et al., 2005a). Temperamental attributes thus measured have been correlated with genetic, psychophysiological and clinical measures (Gonda et al., 2009; Goto et al., 2011). With the original 110-item version being too long for population-based surveys a 39-item short version was later devised (Akiskal et al., 2005b), showing good psychometric properties in terms of concurrent validity with the Temperament and Character Inventory (TCI) (Cloninger et al., 1994), internal consistency, as well as preservation of the proposed 5-subscale structure of temperaments. Despite the apparent universality of affective temperaments in its original Euro-American conception, as borne out in validation studies in different ethnic groups, these studies also showed variations in factor structure and item composition across different cultures (Vahip et al., 2005; Akiyama et al., 2005; Gonda et al., 2011). TEMPS has translations in over 25 different languages (Karam et al., 2005; Blöink et al., 2005; Vahip et al., 2005; Borkowska et al., 2010; Pompili et al., 2008; Figueira et al., 2008; Rozsa et al., 2008; Krebs et al., 2006), including Japanese and Korean (Matsumoto et al., 2005; Kang et al., 2008). To date, TEMPS has not yet been validated in Chinese Cantonese communities. Research of temperaments in Chinese communities is of particular interest. Clinical studies (MAK, 2009) and population-based surveys (Lee et al., 2009) in China have found prevalence of bipolar spectrum disorders and sub-threshold hypomania comparable with that of the West. Description of temperamental profiles in the Chinese population would be important in etiological research for bipolar spectrum disorders, and in informing clinical practice in the management of the millions of Chinese people estimated to suffer from bipolar spectrum disorder (Lin et al., 2013). The current study aims at validating the Chinese (Cantonese) version of the TEMPS-A (cTEMPS-A) and exploring its psychometric properties and applicability. In view of the potential application in large-scale population based studies in China, we also attempted to derive a short version of the cTEMPS-A using two alternative methods: one is adopting the same 39 items of the original short version of the TEMPS-A, and the other is factor analyzing cTEMPS-A using the same methodology as the original authors (Akiskal et al., 2005b). We examined and compared psychometric properties of both versions.

2. Methods 2.1. Translation The original English version-110 item of the Temperament Scale of Memphis, Pisa, Paris and San Diego-Auto-questionnaire (TEMPS-A) was adopted and translated independently into Chinese (Cantonese) by three bilingual psychiatrists (C.M.L, A.M., Y.T. X.) who then compared the texts item by item until consensus was reached. Selection of wording and phrasing emphasized broad comprehensibility to both speakers of the official Mandarin language and the Cantonese dialect, which is more common in Southern China. To assure criterion equivalence with the original version, content, semantic, technical and conceptual equivalence were all emphasized in the translation process, with a slight preference for content equivalence over semantic equivalence where the two were in conflict (Flaherty et al., 1998). Informal pilot interviews with clinical and non-clinical subjects were conducted to help in this process, by checking the wordings of the Chinese translation, so that translated items are broadly

comprehensible and retain the meaning of the original items. The agreed-upon text was then back-translated to English by another psychiatrist (J.S.) who had no prior knowledge of the scale. The back-translated scale was scrutinized by the original author Akiskal who made suggestions for improvement. The panel of translators made further amendments to the Chinese text which was again back-translated into English. The original author endorsed the linguistic authenticity of the final draft of the cTEMPS-A, that is used in this study. 2.2. Sample The cTEMPS-A was distributed to five classes of medical students (year one to year five) of two local universities for completion in mid-2009. Written informed consent was obtained. Approval from the Clinic Research Ethics Committee of the Chinese University of Hong Kong had been sought before the commencement of the study. 2.3. Statistical analysis Data was analyzed using the Statistical Package for Social Science (SPSS) version 17. Internal consistency was measured using the Cronbach-Alpha coefficient. Factor loadings were calculated using principal component analysis (PCA) with varimax rotation with cutoff coefficient set at 0.3. Correlation between subscales was calculated using Pearson's bivariate correlation. Z-scores were calculated for different temperaments assuming normal distribution. To examine differences in scores, Chi-square test was used for categorical variables. For continuous variables, ANOVA was used for parametric variables. Significance level was set at po0.05. 2.4. Short versions We explored two alternative methods to derive a short version of the cTEMPS-A. 2.4.1. The original 39-item short version The corresponding 39 items of the original short version of TEMPS-A (Akiskal et al., 2005b) were used, and this version was analyzed for its internal consistency, factor structure and loading, subscale correlations, and Z-scores. 2.4.2. The reconstituted short version of the cTEMPS-A A reconstituted short version of the cTEMPS-A was derived based on the method used by the original authors in 2005 (Akiskal et al., 2005b). The first 84 items of the TEMPS-A, consisting of dysthymic, cyclothymic, hyperthymic and irritable temperaments, were factor analyzed with PCA and varimax rotation. Items loading onto one factor with value equal or more than 0.45 were added to the 26 items of anxious temperament. The combined items were further factor analyzed. The cut-off of 0.45 was selected instead of 0.35 in the original short version, to strike a balance between inclusiveness and length of the short version. The final items were selected to form a short version based on factor loadings.

3. Results 3.1. Subjects There were 680 year one to year five medical students and a total of 613 valid questionnaires were collected after 23 (3.61%) were discarded for incomplete entry. The response rate was 93.5%. There were 274 males (44.7%) and 339 females (55.3%) with a mean age of 20.8 (range 17–30, S.D.¼2.0). Nearly all (99.9%) were single.

C.M. Leung et al. / Journal of Affective Disorders 170 (2015) 23–29

Table 1 Pearson correlation among the five temperament subscales (110-item) cTEMPS-A.

25

Table 2 Gender difference in mean scores of five temperament scales (110-item) cTEMPS-A.

Scale

Cyclothymic

Hyperthymic

Irritable

Anxious

Scale

Males 7 S.D.

Females7 S.D.

pValue

Depressive Cyclothymic Hyperthymic Irritable

0.432nn

 0.314nn  0.030

0.389nn 0.600nn 0.059

0.558nn 0.559nn  0.233nn 0.558nn

Depressive Cyclothymic Hyperthymic Irritablea Anxious

8.2 7 3.3 6.4 7 4.6 9.5 7 4.2 4.3 7 3.6 6.4 7 4.7

8.6 7 3.1 7.17 4.3 8.7 7 4.1 4.2 7 3.5 7.3 7 5.1

0.07 0.06 0.03nn 0.70 0.03nn

nn

pValue o 0.01. a

Menstruation item (no. 84) omitted for comparison. po 0.05 ANOVA.

nn

3.2. Internal consistency The Cronbach-Alpha coefficients for the depressive, cyclothymic, hyperthymic, irritable and anxious temperament subscales were 0.63, 0.82, 0.78, 0.80, and 0.84, respectively. 3.3. Correlations between temperaments The correlations among the different temperament subscales are shown in Table 1. The strongest correlation was observed between the cyclothymic and irritable temperaments (R¼0.600). No correlation was observed between the cyclothymic and hyperthymic subscales nor between the irritable and hyperthymic subscales. Depressive temperament held a negative correlation with hyperthymic temperament.

in nature (e.g., A85 ‘I have been a worrier for as long as I can remember’, A86 ‘I am always worrying about one thing or another.’). The other factors were more homogenous. Factor II consisted mostly of items representing rapid shifts of mood symptoms between high and low phases, reflecting cyclothymic features (e.g., C23 ‘I get sudden shifts in mood and energy.’; C25 ‘My ability to think varies greatly from sharp to dull for no apparent reason.’). Factor III represented the anxious temperament, with a preponderance of somatic and cognitive features of anxiety (e.g., A 93 ‘When stressed, my hands often tremble.’; A100 ‘I am always thinking someone might break bad news to me about a family member.’. Factor IV represented hyperthymic (e.g., H47 ‘I have great confidence in myself.’) and factor V irritable temperament (e.g., I80 ‘I have been told that I become violent with just a few drinks.’).

3.4. Gender difference in temperament scores The gender differences in mean scores of different temperaments are summarized in Table 2. The females scored significantly lower with the hyperthymic and higher with the anxious subscales. The females also scored higher with the depressive and cyclothymic subscales, though the differences just fall short of statistical significance, p ¼0.07 and 0.06 respectively. 3.5. Distribution of Z-scores Assuming normal distribution, the percentage of respondents scoring between þ 1 and þ2S.D. and above in different subscales is summarized in Table 3. There was no sex difference in the percentage of respondents scoring 4 þ2S.D. in all five temperaments. Distribution of all temperament scores is positively skewed, with hyperthymic subscale least deviated from normality. 3.6. Age effect on temperaments There was no significant difference in the mean scores of any of the temperamental subscales between subjects aged below or above 20. 3.7. Factor analysis The results of factor analysis are summarized in Table 4. Ninety items were loaded onto the five factors with value 0.3. The first factor accounted for 13.23% of variance with eigenvalue of 14.57. The second accounted for 5.34% of variance with eigenvalue of 5.88. The third factor accounted for 3.52% of variance with eigenvalue of 3.87. The fourth factor accounted for 2.95% of variance with eigenvalue of 3.24. The fifth factor accounted for 2.78% of variance with eigenvalue of 3.06. Total variance explained by the five factors was 27.82%. Factor I was a composite factor dominated by items from both depressive (e.g., D1 ‘I am a sad, unhappy person.’, D6 ‘For as long as I can remember, I have felt like a failure.’) and anxious subscales on the original item, the latter being mainly psychic (over-worrying)

3.8. Short versions 3.8.1. The 39-item original short version 3.8.1.1. Factor analysis. The factor structure of the 39 items of the short version of cTEMPS-A is shown in Table 5. The factor structure and component items of each factor were identical to those of the original English short version of the TEMPS. Factor I consisted of 12 cyclothymic items identical to the original (C23, C24, C25, C26, C29, C30, C32, C34, C35, C37, C38, and C39), compatible with the original interpretation as a cyclothymic factor. It accounted for 15.15% of variance with eigenvalue of 5.91. Factor II also comprised of eight items identical to the original (D1, D2, D4, D5, D19, I65, I66, and I81), compatible with the original interpretation as a depressive factor. It accounted for 7.65% of variance with eigenvalue of 2.98. Factor III accounted for 5.40% of variance with eigenvalue of 2.11. It was defined by the same eight items as the original short form (H61, I71, I72, I73, I74, I77, I79, and I80), thus may retain the original definition as an irritable factor. Factor IV accounted for 4.20% of variance with eigenvalue of 1.64. It was defined by the same eight items as the original short form (H45, H48, H51, H52, H54, H55, H58, and H60), retaining the original interpretation as a hyperthymic factor. The fifth factor accounted for 4.01% of variance with eigenvalue of 1.56. It contained the same three anxious items (A 99, A98, and A100) and retains the original interpretation as an anxious factor. Total variance explained by the five factors amounted to 36.42%. Item 79 had a low factor loading value of 0.25. 3.8.1.2. Internal consistency. The Cronbach-Alpha coefficients for the depressive, cyclothymic, hyperthymic, irritable and anxious temperament subscales were 0.75, 0.78, 0.67, 0.67, and 0.60, respectively. 3.8.2. The reconstituted short version of the cTEMPS-A 3.8.2.1. Factor analysis. Using a cut-off of factor loading coefficient equal to or greater than 0.35, 54 items out of the first 84 items of the c-TEMPS A loaded onto either one factor of a four-factor

26

C.M. Leung et al. / Journal of Affective Disorders 170 (2015) 23–29

Table 3 Distribution of Z-scores on five temperament subscales (110-item) cTEMPS-A. Scale

Total subjects

Inside the distribution (%)

1–2 S.D. (%)

4 þ 2 S.D. (%)

Depressive Cyclothymic Hyperthymic Irritable Anxious

613 613 613 613 613

80.0 83.4 82.9 81.7 84.3

13.1 12.2 13.7 13.6 10.6

3.9 4.4 3.4 4.7 5.1

Male Scale Depressive Cyclothymic Hyperthymic Irritable Anxious

Total subjects 274 274 274 274 274

Skewness 0.422 0.687 0.120 1.218 1.047

Inside the distribution (%) 85.4 81.4 83.6 82.1 85.0

1–2 S.D. (%) 11.0 13.5 13.1 12.8 9.9

4 þ 2 S.D. (%) 3.6 5.1 3.3 5.1 5.1

Female Scale Depressive Cyclothymic Hyperthymic Irritable Anxious

Total subjects 339 339 339 339 339

Skewness 0.453 0.399 0.191 1.178 0.962

Inside the distribution (%) 81.1 85.0 82.3 81.4 83.8

1–2 S.D. (%) 14.8 11.2 14.2 14.2 11.2

4 þ 2 S.D. (%) 4.1 3.8 3.5 4.4 5.0

Table 4 Factor structure after varimax rotation of the (110-item) cTEMPS-A. Factor I D1 D2 D3 D4 D6 D7 D8 D13 D19 C22 C24 C33 C38 I65 I66 I69 I70 I78 I81 A85 A86 A87 A88 A89 A91 A92 A107 A108

0.79 0.72 0.57 0.58 0.61 0.53 0.48 0.38 0.49 0.36 0.42 0.67 0.37 0.62 0.59 0.71 0.71 0.39 0.48 0.81 0.77 0.71 0.70 0.69 0.69 0.79 0.54 0.38

Factor II

Factor III

C23 C25 C26 C27 C29 C30 C34 C35 C36 C39 C40 C41 C42 H56 H61 I64 I67 I71 I72 I73 I74 I77

A93 A94 A95 A96 A97 A98 A99 A100 A101 A102 A104 A105 A106 A109 A110 C32

0.65 0.47 0.51 0.71 0.53 0.49 0.48 0.52 0.45 0.45 0.42 0.56 0.37 0.45 0.48 0.47 0.50 0.53 0.49 0.63 0.46 0.52

0.54 0.48 0.35 0.52 0.35 0.39 0.42 0.64 0.62 0.49 0.79 0.52 0.54 0.52 0.48 0.45

Table 5 Factor structure after varimax rotation of the short (39-item) cTEMPS-A.

Factor IV

Factor V

Factor I

D14 D15 D16 D18 H47 H48 H49 H50 H51 H52 H53 H54 H57 H58 H59 H60 I82 A103

H45 H62 I75 I76 I79 I80 I83

C23 C24 C25 C26 C29 C30 C32 C34 C35 C37 C38 C39

0.52 0.31 0.65 0.52 0.56 0.44 0.40 0.57 0.38 0.55 0.53 0.50 0.56 0.44 0.32 0.40 0.43 0.35

0.32 0.54 0.48 0.45 0.58 0.61 0.59

Only items with factor loading coefficient ≧0.3 included. D¼ Depressive, C ¼ Cyclothymic, H ¼Hyperthymic, I ¼ Irritable, A ¼Anxious.

structure. With the remaining 26 items added for further principal component analysis and varimax rotation, 43 items converged onto five factors separately with a factor loading coefficient equal to or greater than 0.45. The results are shown in Table 6. These 43 items formed the reconstituted short version of cTEMPS-A further analyzed below. Factor I accounted for 20.47% of variance with eigenvalue of 9.0. It was defined by 14 items: three depressive (D1, D2, D6), seven anxious (A85, A86, A87, A88, A89, A91, and A92) items of psychic nature, three irritable (I68, I69, I70), and one cyclothymic item (C33). We interpret it as a combined anxious-depressive factor.

0.61 0.39 0.52 0.49 0.64 0.61 0.38 0.52 0.61 0.37 0.52 0.52

Factor II

Factor III

Factor IV

Factor V

D1 D2 D4 D5 D19 I65 I66 I81

H61 I71 I72 I73 I74 I77 I79 I80

H45 H48 H51 H52 H54 H55 H58 H60

A99 A98 A100

0.62 0.59 0.55 0.23 0.63 0.62 0.50 0.62

0.47 0.55 0.60 0.69 0.52 0.60 0.25 0.40

0.55 0.66 0.55 0.53 0.48 0.47 0.51 0.50

0.70 0.74 0.46

Only items with factor loading coefficient ≧0.2 included. D¼Depressive, C ¼Cyclothymic, H¼ Hyperthymic, I ¼Irritable, A ¼ Anxious.

Table 6 Factor structure of reconstituted short version (43-item) of cTemps-A. Factor I D1 D2 D6 C33 I68 I69 I70 A85 A86 A87 A88 A89 A91 A92

0.49 0.55 0.45 0.50 0.71 0.68 0.68 0.73 0.71 0.69 0.69 0.69 0.62 0.76

Factor II

Factor III

Factor IV

Factor V

C23 C25 C26 C29 C35 C39

H61 I64 I66 I67 I71 I72 I73 I75 I76

D18 H47 H48 H50 H52 H53 H54 H58 I82

A93 A100 A101 A104 A105

0.64 0.58 0.57 0.53 0.59 0.52

0.56 0.57 0.53 0.59 0.47 0.54 0.61 0.50 0.52

0.48 0.57 0.56 0.53 0.63 0.53 0.50 0.49 0.54

0.50 0.52 0.62 0.64 0.45

Only items with factor loading coefficient≧0.45 included. D¼Depressive, C ¼Cyclothymic, H¼ Hyperthymic, I ¼Irritable, A ¼ Anxious.

Factor II accounted for 7.13% of variance with eigenvalue of 3.14. It was defined by six cyclothymic items (C23, C24, C25, C28, C35, and C39), thus may be interpreted as a cyclothymic factor.

C.M. Leung et al. / Journal of Affective Disorders 170 (2015) 23–29

Factor III accounted for 4.82% of variance with eigenvalue of 2.12, defined by nine items, of which one was hyperthymic (H61), eight were irritable (I64, I66, I67, I71, I72, I73, I75, and I76), interpretable as an irritable factor. Factor IV accounted for 3.97% of variance with eigenvalue of 1.75. It was defined by nine items: with one depressive (D18), seven hyperthymic (H47, H48, H50, H52, H53, H54, and H58), and one irritable (I82), interpretable as a hyperthymic factor. Factor V accounted for 3.79% of variance with eigenvalue of 1.67. It was defined by five items depicting somatic aspects of anxiety (A93, A100, A101, A104, and A105), and interpretable as an anxious (somatic) factor. The five factors together explained a total of 40.18% of the variance. 3.8.2.2. Internal consistency. The Cronbach-Alpha coefficients for the anxious–depressive, cyclothymic, hyperthymic, irritable and anxious (somatic) temperament subscales were 0.91, 0.70, 0.71, 0.75, and 0.55, respectively.

4. Discussion 4.1. Internal validity We validated the cTEMPS-A in a group of non-clinical Chinese adults in Hong Kong. The cTEMPS-A showed satisfactory linguistic likeness to the original, retain the original five-factor structure, and showed good internal consistency of the subscales. Correlations between the temperament subscales were consistent with theoretical considerations and empirical evidence as set out by the original authors (Akiskal and Akiskal, 1992; Akiskal, 1998, 2002; Placidi et al., 1998). These serve to support the internal validity of the cTEMPS-A. 4.2. Factor structure Similar to the case for most other versions of the TEMPS-A (Karam et al., 2005; Pompili et al., 2008; Figueira et al., 2008; Borkowska et al., 2010), the original depressive and anxious overworrying items mix to form an “anxious–depressive” temperament in the cTEMPS-A, while the anxious items with physical symptoms converge to form a relatively homogeneous factor. The unitary concept or spectrum view of anxiety and depression is age old (Gersh and Fowles, 1979; Kendell, 1974; Stavrakaki and Vargo, 1986). In addition, depression is not a unitary construct (Clark et al., 1999) and the cognitive model considers a core cognitive dysfunction shared by depression and anxiety (Haaga et al., 1991). Thus, such over-lapping between the original depressive and anxious temperament factors is not surprising. The finding agrees well with the composite nature of the Chinese term for depression “you yu 忧郁”, which literally means worry and stagnation. The Chinese-Mandarin version of the TEMPS-A, translated and validated by another research team, reported a somewhat different factor structure (Lin et al., 2013). Sampling and linguistic difference (Mandarin vs. Cantonese), together with methodological distinction, may account for this disparity. Moreover, the present Cantonese version is based on a population of medical students, which may make it somewhat less generalizable to the general population. The heterogeneity of the original depressive factor was also reflected in the relatively low internal consistency of the depressive subscale (0.63) as compared with the other subscales (0.78– 0.84). On the other hand, the hyperthymic factor contained the most homogeneous distribution of the original hyperthymic items with little contamination from other axes.

27

Most of the temperaments were positively correlated with one another, with alpha values ranging from 0.45 to 0.59, in parallel with values reported in the literature (Karam et al., 2005; Pompili et al., 2008; Borkowska et al., 2010). The hyperthymic temperament, however, is distinctly different by having significant and substantially negative correlations with the depressive and anxious temperaments while having little correlation with other two. That hyperthymic temperament was not correlated with other temperaments was evident even in the original version of TEMPS-I (Akiskal et al., 1998). Such a negative relationship with the depressive temperament was echoed by other Continental European samples, such as German (Blöink et al., 2005) and Polish studies (Borkowska et al., 2010), but not by others, e.g., Lebanese (Karam et al., 2005). This may imply that correlation between depressive and hyperthymic temperament could be more characteristic of a bipolar diathesis in Mediterranean samples contrasted with Continental European populations. On the other hand, the cyclothymic temperament, while having no correlation with the hyperthymic temperament, correlated strongly and positively with both the irritable and depressive temperaments. Such a relationship is in line and parallels cyclothymic temperament with bipolar II tendencies, the course of which is predominantly depressive, intermixed with recurrent periods of euphoria and irritability (Coryell et al., 1995; Perugi and Akiskal, 2002). It would be of interest to examine this hypothetically predictive relationship with bipolar disorders in prospective cohort studies particularly in China, because the quality of hypomania appears different in at least one study, comparing Sardegna and China (Carta et al., 2014). Females were significantly different from males with hyperthymic (lower) and anxious (higher) temperaments. They also scored higher, though just falling short of statistical significance, from their counterparts with depressive and cyclothymic (both higher) temperaments. Most other studies reported between-gender differences in all temperaments (Karam et al., 2005; Pompili et al., 2008). Both the lower hyperthymic and higher anxious scores, or depressive and cyclothymic as well, in females could be attributed to their traditional social roles and status in most communities; whether there are constitutional differences in this respect cannot be answered by the methods of this study. The more narrow distribution of the Z-scores in the hyperthymic temperament, as compared to other domains, could reflect the more “modest” and even “inhibited” personality styles of Chinese people in general. Likely, the constitution of our medical student subjects, possibly less adventurous than their college counterparts, also contributes to the finding; these interpretations are in line with those of Figueira et al., 2008. The lack of differential effect by age for different temperaments can be attributed to a rather narrow age range in our student sample. Rooted in the Euro-American literature, the original temperament items unavoidably contained items that might be culturally sanctioned, or were loaded on the subscales different from the original versions. This probably reflected culturally disparate idioms of expressing emotional experience on the one hand, and culturally-specific normative behavioral styles that may be somewhat beyond the reach of semantic and content equivalence on the other. The Turkish version deleted 10 unsuitable items (Vahip et al., 2005), while the Japanese researchers chose only 40 items for inclusion in the Japanese temperament inventory (Akiyama et al., 2005). This is not surprising since the construction of the scale by the original authors was rooted in theoretical and EuroAmerican historical considerations. In our sample of university students in Hong Kong, item I82 (‘I could be a revolutionary.’) loads onto the hyperthymic domain. It would be of interest to example the same item in Chinese communities of different sociocultural backgrounds, such as Taiwan and Mainland China,

28

C.M. Leung et al. / Journal of Affective Disorders 170 (2015) 23–29

where ‘revolutionary’ may have other connotations. However, despite these relatively minor regional and cultural variations, the overall temperament constructs embodied in TEMPS-A emerge as remarkably universal. As mentioned earlier, anxiety items with “worrisome” contentA85 (‘I have been a worrier for as long as I can remember.’), A86 (‘I am always worrying about one thing or another.’), A87 (‘I keep on worrying about daily matters that others consider minor.’), A 88 (‘I cannot help worrying.’), A89 (‘Many people have told me not to worry so much.’), A91 (‘I am unable to relax.’), A 92 (‘I often feel jittery inside’.), A107 (‘I am an insecure person.’) and A108 (‘Even minor changes in routine, stress me highly.’) merge with the depressive items to form the anxious–depressive, or super “depressive” factor. Again, the dual psychic and biological components in the phenomenon of depression, as exemplified in the Chinese term for depression “you yu 忧郁”, meaning “worry” and “stagnation”, is well demonstrated. Cyclothymic items-C22 (‘I often feel tired for no reason.’), C33 (‘I am told that I often get pessimistic about things, and forget previous happy times.’) and C38 (‘The way I see things is sometimes vivid, but at other times lifeless.’) also load onto this super factor, probably reflecting occurrence of bipolar traits in depressive individuals, in support of a spectrum model of bipolarity (Akiskal et al., 1977, 2002). Several irritable items-I64 (‘I am a grouchy (irritable) person.’), I67 (‘I am highly critical of others.’), I72 (‘When crossed, I could get into a fight.’), I73 (‘People tell me I blow up out of nowhere.’), I74 (‘When angry, I snap at people.’), I77 (‘I can get so furious I could hurt someone.’) and hyperthymic items-H56 (‘People tell me that I often get my nose into others’ business.’) and H61 (‘When I disagree with someone, I can get into a heated argument.’) load onto the cyclothymic factor in our sample. This is understandable in view of the high association between the cyclothymic and irritable temperament. On the other hand, items H56 (‘People tell me that I often get my nose into others’ business.’) and H61 (‘When I disagree with someone, I can get into a heated argument.’) probably reflected more of irritability than of euphoric disinhibition that is usually associated with hyperthymic temperament. Depressive items-D14 (‘I am the kind of person you can always depend on.’), D15 (‘I put the needs of others above my own.’), D16 (‘I am a hard working person.’), and D18 (‘It is natural for me to be neat and organized.’) describing “positive” personal attributes, load onto the hyperthymic factor, probably reflecting either the neutrality of these items or their desirability. Indeed they seem to reflect obsessive-compulsive traits embodied in the concept of the melancholic temperament in the Geman literature (Mundt et al., 1997). Indeed, they did not find their place in the 39-item short version of TEMPS. In the Chinese context, the common cultural tendency to celebrate discipline and self-sacrifice may have further moved these items towards cultural norm rather than reflecting depressive tendencies. This could mean that the melancholic temperament, as conceived in the German literature, may have some universality, witness its resonance with what can be regarded as Chinese “virtues”. 4.3. Short versions The 39-item cTEMPS-A upheld the original five-factor structure, accommodating items identical to the original English short version in each temperament subscale. Such robust factor structure with its clean item contents has been borne out in its various translations, supporting its cross-national applicability. This also served as a testament to the degree of criterion equivalence achieved in the Chinese version. Low loading values may have been the trade-off for standardization of subscale items, with D5 (‘I give up easily.’) and I79 (‘I am known to swear a lot.’) in the 0.2–

0.3 range. The relatively low internal consistency compared with the long version also results from excluding other items that may overlap in content but lend cohesiveness between items in a subscale. With the 43-item cTEMPS-A, items that did not apply well to the local culture, translating into lower correlation coefficients, were either dropped or reallocated. Compared with the 39-item short version TEMPS, it has similar length, factor strength, and comparable internal consistency. The main difference between the two lies in the body of the anxious–depressive factor foremost, with the reconstituted 43-item cTEMPS-A bearing more resemblance to the cTEMPS-A and arguably the original TEMPS-A than did the original 39-item TEMPS-A. We thereby submit that the 43item cTEMPS could be a valid and reliable tool for population screening locally, as an “economical” alternative of the 110-item cTEMPS-A. This is a compromise from a feasibility point of view which would require further study. One must be cautious to adopt it on a large scale but from a pragmatic point of view, local screening can go ahead.

4.4. Limitations The current study was obviously limited by the narrow age range and non-clinical nature of the subjects, which may limit generalization to other samples. The recruitment of only medical students may also have meant pre-selection of individuals with higher dysthymic and obsessive-compulsive traits, as has been reported for physicians in a previous study on professionals’ temperamental characteristics (Akiskal et al., 2005c), or lower hyperthymic tendency, as has been reported for professional choice in Spanish university students (Figueira et al., 2008). However, factor structure of the cTEMPS-A was broadly similar to that of the original English version, and the 39-item short forms were identical across Chinese and English versions. On the other hand, factor structure of the original TEMPS-A, derived from clinical samples, was similar to that based on a clinically-well student sample (Maremmani et al., 2005), indicating the robustness of the temperament constructs across clinical and nonclinical samples. Although the original version of the TEMPS-A has undergone concurrent validation against other validated instruments of temperament scales (Maremmani et al., 2005), in the present Hong Kong Cantonese study, we did not include any instrument to serve as reference to assess external or predictive validity. Such work should take care to include cross-culturally valid external validators. The Chinese version of the TCI could be one candidate, with evidence for its criterion equivalence to the original English version (Parker et al., 2003). Of even greater interest, however, would be to tease out whether any cross-national variance is attributed to variation of the criterion used to study temperamental phenomena, or to a true variance in temperamental phenomena in a cross-ethnic sense. Much use would be made of the TEMPS-A (Akiskal and Akiskal, 2005a), in its different linguistic versions, on cross-ethnic samples, and possibly translational research (Kang et al., 2008). At our center, further field work comparing the 43- and 110item versions of cTEMPS in clinical subjects, of a more comprehensive age range, for their predictive validity is being planned. Coupling with established personality inventories will complement the study in external validity, before embarking on representative population sampling and translational research on temperaments. In the other direction, follow-up study of what medical specialties our sample subjects will be doing could shed further light on the relationship between temperamental traits and specialty choice that has been pioneered in Italian and

C.M. Leung et al. / Journal of Affective Disorders 170 (2015) 23–29

Portuguese studies by the developers of TEMPS-A (Akiskal et al., 2005c; Figueira et al., 2008). Role of funding source Disclaimer statement: All authors have no sources of support, funding or financial interest to report for this work.

Conflict of interest Disclaimer statement: All authors have no conflicts of interests to declare for this work.

Acknowledgment We would like to thank the medical students of the two medical schools in Hong Kong for completion of the questionnaires, Dr. J So (J.S.) for the backtranslation, and Mr. A Tsang for statistical advice.

References Akiskal, H.S., 1981. Subaffective disorders: dysthymic, cyclothymic and bipolar II disorders in the “borderline” realm. Psychiatr. Clin. N. Am. 4, 25–46. Akiskal, H.S., 1998. Toward a definition of generalized anxiety disorder as an anxious temperament type. Acta Psychiatr. Scand. 98, 66–73. Akiskal, H.S., 2002. The bipolar spectrum–the shaping of a new paradigm in psychiatry. Curr. Psychiatry Rep. 4, 1–3. Akiskal, H.S., Akiskal, K., 1992. Cyclothymic, hyperthymic and depressive temperaments as subaffective variants of mood disorders. In: Tasman, A., Riba, M.B. (Eds.), Annual Review of Psychiatry, Vol. 11. American Psychiatric Press, Washington, pp. 43–62. Akiskal, H.S., Mallya, G., 1987. Criteria for the “soft” bipolar spectrum: treatment implications. Psychopharmacol. Bull. 23, 68–73. Akiskal, H.S., Djenderedjian, A.M., Rosenthal, R.H., Khani, M.K., 1977. Cyclothymic disorder: validating criteria for inclusion in the bipolar affective group. Am. J. Psychiatry 134, 1227–1233. Akiskal, H.S., Khani, M.K., Scott-Strauss, A., 1979. Cyclothymic temperamental disorders. Psychiatr. Clin. N. Am. 2, 526–554. Akiskal, H.S., Placidi, G.F., Maremmani, I., Signoretta, S., Liguori, A., Gervasi, R., Mallya, G., Puzantian, V.R., 1998. TEMPS-I: delineating the most discriminant traits of the cyclothymic, depressive, hyperthymic and irritable temperaments in a nonpatient population. J. Affect. Disord. 51, 7–19. Akiskal, H.S., Akiskal, K.K., 2005a. TEMPS: Temperament Evaluation of Memphis, Pisa, Paris, and San Diego (Special issue). J. Affect. Disord. 85 (1–2), 1–242. Akiskal, H.S., Akiskal, K.K., Haykal, R.F., Manning, J.S., Connor, P.D., 2005a. TEMPS-A: progress towards validation of a self-rated clinical version of the Temperament Evaluation of the Memphis, Pisa, Paris, and San Diego autoquestionnaire. J. Affect. Disord. 85, 3–16. Akiskal, H.S., Mendlowicz, M.V., Jean-Louis, G., Rapaport, M.H., Kelsoe, J.R., Gillin, J. C., Smith, T.L., 2005b. TEMPS-A: validation of a short version of a self-rated instrument designed to measure variations in temperament. J. Affect. Disord. 85, 45–52. Akiskal, K.K., Akiskal, H.S., 2005b. The theoretical underpinnings of affective temperaments: implications for evolutionary foundations of bipolar disorder and human nature. J. Affect. Disord. 85, 231–239. Akiskal, K.K., Savino, M., Akiskal, H.S., 2005c. Temperament profiles in physicians, lawyers, managers, industrialists, architects, journalists, and artists: a study in psychiatric outpatients. J. Affect. Disord. 85, 201–206. Akiyama, T., Tsuda, H., Matsumoto, S., Miyake, Y., Kawamura, Y., Noda, T., Akiskal, K.K., Akiskal, H.S., 2005. The proposed factor structure of temperament and personality in Japan: combining traits from TEMPS-A and MPT. J. Affect. Disord. 85, 93–100. Blöink, R., Brieger, P., AkiskaL, H.S., Marneros, A., 2005. Factorial structure and internal consistency of the German TEMPS-A scale: validation against the NEOFFI questionnaire. J. Affect. Disord. 85, 77–83. Borkowska, A., Rybakowski, J.K., Drozdz, W., Bielinski, M., Kosmowska, M., Rajewska-Rager, A., Bucinski, A., Akiskal, K.K., Akiskal, H.S., 2010. Polish validation of the TEMPS-A: the profile of affective temperaments in a college student population. J. Affect. Disord. 123, 36–41. Carta, M.G., Massidda, D., Moro, M.F., Aguglia, E., Balestrieri, M., Caraci, F., Dell'Osso, L., Di Sciascio, G., Drago, F., Faravelli, C., Hardoy, M.C., Calò, S., Pollice, R., Fortezzo, A., Akiskal, H., 2014. Comparing factor structure of the Mood Disorder Questionnaire (MDQ): in Italy sexual behavior is euphoric but in Asia mysterious and forbidden. J. Affect. Disord., 155; , pp. 96–103. Clark, D.A., Beck, A.T., Alford, B.A., 1999. Scientific Foundations of Cognitive Theory and Therapy of Depression. John Wiley & Sons, New York. Cloninger, C.R., Przybeck, T.R., Svrakic, D.M., Wetzel, R.D., 1994. The Temperament and Character Inventory (TCI): A Guide to Its Development and Use. Center for Psychobiology of Personality, Washington University, St. Louis, MO. Coryell, W., Endicott, J., Maser, J.D., Keller, M.B., Leon, A.C., Akiskal, H.S., 1995. Longterm stability of polarity distinctions in the affective disorders. Am. J. Psychiatry 152, 385–390. Eysenck, H.J., Eysenck, S.B.G., 1969. Personality Structure and Measurement. Routledge, London.

29

Figueira, M.L., Caeiro, L., Ferro, A., Severino, L., Duarte, P.M., Abreu, M., Akiskal, H.S., Akiskal, K.K., 2008. Validation of the Temperament Evaluation of Memphis, Pisa, Paris and San Diego (TEMPS-A): Portuguese-Lisbon version. J. Affect. Disord. 111, 193–203. Flaherty, J.A., Gaviria, F.M., Pathak, D., Mitchell, T., Wintrob, R., Richman, J.A., Birz, S., 1998. Developing instruments for cross-cultural research. J. Nerv. Ment. Dis. 176, 257–263. Gersh, R.F., Fowles, D., 1979. Neurotic depression: the concept of anxious depression. In: Depue, R. (Ed.), The Psychobiology of Depressive Disorders: Implications for the Effects of Stress. Academic Press, San Diego, C.A.. Gonda, X., Fountoulakis, K.N., Rihmer, Z., Lazary, J., Laszik, A., Akiskal, K.K., Akiskal, H.S., Bagdy, G., 2009. Towards a genetically validated new affective temperament scale: a delineation of the temperament phenotype of 5-HTTLPR using the TEMPS-A. J. Affect. Disord. 112, 19–29. Gonda, X., Vazquez, G.H., Akiskal, K.K., Akiskal, H.S., 2011. From putative genes to temperament and culture: cultural characteristics of the distribution of dominant affective temperaments in national studies. J. Affect. Disord. 131, 45–51. Goto, S., Terao, T., Hoaki, N., Wang, Y., 2011. Cyclothymic and hyperthymic temperaments may predict bipolarity in major depressive disorder: a supportive evidence for bipolar II1/2 and IV. J. Affect. Disord. 129, 34–38. Greenwood, T.A., Akiskal, H.S., Akiskal, K.K., 2012. Genome-wide association study of temperament in bipolar disorder reveals significant associations with three novel Loci. Biol. Psychiatry 72, 303–310. Greenwood, T.A., Badner, J.A., Byerley, W., Keck, P.E., McElroy, S.L., Remick, R.A., Sadovnick, A.D., Akiskal, H.S., Kelsoe, J.R., 2013. Heritability and genome-wide SNP linkage analysis of temperament in bipolar disorder. J. Affect. Disord. 150, 1031–1040. Haaga, D.A., Dyck, M.J., Ernst, D., 1991. Empirical status of cognitive theory of depression. Psychol. Bull. 110, 215–236. Kang, J.I., Namkoong, K., Kim, S.J., 2008. The association of 5-HTTLPR and DRD4 VNTR polymorphisms with affective temperamental traits in healthy volunteers. J. Affect. Disord. 109, 157–163. Karam, E.G., Mneimneh, Z., Salamoun, M., Akiskal, K.K., Akiskal, H.S., 2005. Psychometric properties of the Lebanese-Arabic TEMPS-A: a national epidemiologic study. J. Affect. Disord. 87, 169–183. Kendell, R.E., 1974. The stability of psychiatric diagnosis. Br. J. Psychiatry 124, 352–356. Kraepelin, E., 1921. Manic Depressive Insanity and Paranoia. Livingstone, Edinburgh. Krebs, M.O., Kazes, M., Olie, J.P., Loo, H., Akiskal, K.K., Akiskal, H.S., 2006. The French version of the validated short TEMPS-A: the temperament evaluation of Memphis, Pisa, Paris and San Diego. J. Affect. Disord. 96, 271–273. Kretschmer, E., 1936. Physique and Character. Kegan, Paul, Trench, Trubner and Co. Ltd, London. Lee, S., Ng, K.L., Tsang, A., 2009. A community survey of the twelve-month prevalence and correlates of bipolar spectrum disorder in Hong Kong. J. Affect. Disord. 117, 79–86. Lin, K., Xu, G., Miao, G., Ning, Y., Ouyang, H., Chen, X., Hoang, N., Akiskal, K.K., Akiskal, H.S., 2013. Psychometric properties of the Chinese (Mandarin) TEMPS-A: a population study of 985 non-clinical subjects in China. J. Affect. Disord. 147, 29–33. MAK, A.D.P., 2009. Prevalence and correlates of bipolar II disorder in major depressive patients at a psychiatric outpatient clinic in Hong Kong. J. Affect. Disord. 112, 201–205. Maremmani, I., Akiskal, H.S., Signoretta, S., Liguori, A., Perugi, G., Cloninger, R., 2005. The relationship of Kraepelian affective temperaments (as measured by TEMPS-I) to the tridimensional personality questionnaire (TPQ). J. Affect. Disord. 85, 17–27. Matsumoto, S., Akiyama, T., Tsuda, H., Miyake, Y., Kawamura, Y., Noda, T., Akiskal, K.K., Akiskal, H.S., 2005. Reliability and validity of TEMPS-A in a Japanese non-clinical population: application to unipolar and bipolar depressives. J. Affect. Disord. 85, 85–92. Mundt, C., Backenstrass, M., Kronmiller, K.T., Fiedler, P., Kraus, A., Stanghellini, G., 1997. Personality and endogenous/major depression:an empirical approach to typus melancholicus. 2. Validationof typus melancholicus core-properties by personality inventory scales. Psychopathology 30, 130–139. Parker, G., Cheah, Y.C., Parker, K., 2003. Properties of the temperament and character inventory in a Chinese sample. Acta Psychiatr. Scand. 108, 367–373. Perugi, G., Akiskal, H.S., 2002. The soft bipolar spectrum redefined: focus on the cyclothymic, anxious-sensitive, impulse-dyscontrol, and binge-eating connection in bipolar II and related conditions. Psychiatr. Clin. N. Am. 25, 713–737. Placidi, G.F., Signoretta, S., Liguori, A., Gervasi, R., Maremmani, I., Akiskal, H.S., 1998. The semi-structured affective temperament interview (TEMPS-I). Reliability and psychometric properties in 1010 14-26-year old students. J. Affect. Disord. 47, 1–10. Pompili, M., Girardi, P., Tatarelli, R., Iliceto, P., De Pisa, E., Tondo, L., Akiskal, K.K., Akiskal, H.S., 2008. TEMPS-A (Rome): psychometric validation of affective temperaments in clinically well subjects in mid- and south Italy. J. Affect. Disord. 107, 63–75. Rozsa, S., RIhmer, Z., Gonda, X., Szili, I., Rihmer, A., Ko, N., Nemeth, A., Pestality, P., Bagdy, G., Alhassoon, O., Akiskal, K.K., Akiskal, H.S., 2008. A study of affective temperaments in Hungary: internal consistency and concurrent validity of the TEMPS-A against the TCI and NEO-PI-R. J. Affect. Disord. 106, 45–53. Stavrakaki, C., Vargo, B., 1986. The relationship of anxiety and depression: a review of the literature. Br. J. Psychiatry 149, 7–16. Vahip, S., Kesebir, S., Alkan, M., Yazici, O., Akiskal, K.K., Akiskal, H.S., 2005. Affective temperaments in clinically-well subjects in Turkey: initial psychometric data on the TEMPS-A. J. Affect. Disord. 85, 113–125.