TEMPS-A: progress towards validation of a self-rated clinical version of the Temperament Evaluation of the Memphis, Pisa, Paris, and San Diego Autoquestionnaire

Abstract

Background

Our aim was to validate the Temperament Evaluation of the Memphis, Pisa, Paris, and San Diego Autoquestionnaire (TEMPS-A) in a clinical population.

Methods

The study was conducted in two Memphis mood clinics involving 398 affectively ill patients with young to middle index age (42 years±13 S.D.), who were 95% white, 62% female, and 51% bipolar spectrum. A subset of 157 of the entire sample were retested in 6–12 months, and the entire sample was then subjected to factor analysis (PCA extraction method with varimax rotation).

Results

We obtained high test–retest reliability ranging from 0.58 for the irritable, to 0.68, 0.69 and 0.70, respectively, for the cyclothymic, dysthymic and hyperthymic. The hypothesized four-factor structure of the TEMPS-A was upheld, with the cyclothymic explaining 14% of the variance, followed by the irritable, hyperthymic, and dysthymic together accounting for another 14%. Internal consistency was excellent, with Chronbach alphas ranging from 0.76 for the dysthymic to 0.88 for the cyclothymic. Exploratory factor analysis revealed 2 super factors, Factor I loading on cyclothymic, irritable, and dysthymic temperaments, and Factor II loading heavily on the hyperthymic. The 50-item TEMPS-A-Clinical Version was constructed by using a cutoff of alpha ≥0.4 for traits loading exclusively on their original temperaments. We also proposed a longer 69-item version for future study, in which we permitted a greater number of traits based on clinical considerations (alpha cutoff 0.30).

Limitation

The sample was preponderantly white, and may not generalize to other U.S. ethnic groups. This earlier version of TEMPS-A did not include the anxious temperament.

Conclusions

We psychometrically validated the TEMPS-A in affectively ill outpatients, leading to an instrument suitable for use in psychiatric, especially affectively ill, populations. It is noteworthy that in this clinically ill population we succeeded in measuring traits which could make subjects vulnerable to affective episodes, as well as those of adaptive nature. For instance, the dysthymic emerged as bound to routine, self-blaming, shy-nonassertive, sensitive to criticism, yet self-denying, dependable, and preferring to work for someone else rather than be the boss. The hyperthymic had the highest number of “positive” traits: upbeat, fun-loving, outgoing, jocular, optimistic, confident, full of ideas, eloquent, on the go, short-sleeper, tireless, who likes to be the boss, but single-minded, risk-taker, and unlikely to admit to his/her meddlesome nature. The cyclothymic emerged as labile with rapid shifts in mood; unstable in energy, self-esteem and socialization; unevenly gifted and dilettante; yet keen in perception, intense in emotions, and romantic. The irritable emerged as skeptical and critical (which might be considered intellectual virtues), but otherwise having the “darkest” nature of all temperaments: grouchy, complaining, dissatisfied; anger- and violence-prone, and sexually jealous. The foregoing temperament attributes, observed in a moderately severe group of patients with affective disorders, nonetheless testify to the evolutionary context of these disorders—“submissive” behavior, territoriality, romantic charm, and last, but not least, sexually jealous with its associated specter of violence. We hypothesize that the putative social and limbic mechanisms underlying mood disorders appear to have archaic origins on an evolutionary scale. We finally submit that the traits underlying affective disorders are very much part of human nature.

Introduction

The Temperament Evaluation of Memphis, Pisa, Paris, and San Diego (TEMPS) has been in the making for at least a quarter of a century. Its early development is intimately linked with the Mood Clinic Data Questionnaire (MCDQ), a semi-structured clinical instrument that the first author (H.S.A.) had constructed in order to collect systematic diagnostic data on consecutive patients in our Memphis mood clinic (Akiskal et al., 1978). The criteria for diagnosis in the MCDQ were largely derived from the Washington University framework (Feighner et al., 1972), in which the only validated personality disorder was anti-social personality. The unreliability of other personality constructs can be seen in the fact that our clinic patients–previously evaluated by other clinicians in the greater Memphis metropolitan area–had often been tagged with different personality attributes, i.e., a manic–depressive patient would be diagnosed “hysterical” (as they were labeled in those days) at the excited phase of the illness and “passive-dependent” during the depressive phase. The state dependency of such diagnoses did not seem to bother the clinicians who had made them.

But it did bother us. Instead of using the then DSM-II schema (American Psychiatric Association, 1968) for personality disorders and traits, we resorted to German concepts: such as Kurt Schneider's (1958) classic description of “psychopathic” (i.e., abnormal) personalities. We operationalized, modified or otherwise enriched Schneider's descriptions with those of Kraepelin's (1899/1921) “personal dispositions”, taking into consideration our own clinical experience in the setting of the Mood Clinic (Akiskal et al., 1977, Akiskal et al., 1979). Our first set of operationalizations pertained to the “depressive”, “cyclothymic”, and “hyperthymic” types; an “irritable” type was added later (Akiskal and Mallya, 1987, Akiskal, 1992). Other Schneiderian personality constructs (e.g., the “status-seeking”, “attention-seeking”, and “fanatic” types) which we had also operationalized for the MCDQ, will not be considered in this paper.

It is relevant to point out that unlike Kraepelin (1899/1921), Schneider (1958) did not believe that his “psychopathic types” represent the underlying foundations of mood disorders. In line with Kraepelin, we preferred to consider them dispositions to mood states; indeed, we conceptualized them as “subaffective” or trait affective expressions of mood disorders (Akiskal, 1981, Sass et al., 1993). These traits were hypothesized to precede and follow episodes of these disorders in an affective trait-affective disorder continuum. In other words, we used the superb descriptions of Schneider, but adhered to the Kraepelinian conceptual framework. For us, it made better sense for the affectively ill to be described by traits which reflect an affective disposition (Akiskal, 1992), rather than the maladaptive interpersonal framework used in DSM-II, DSM-III, and DSM-IV (American Psychiatric Association, 1968, American Psychiatric Association, 1980, American Psychiatric Association, 1994). Hence our preference for the term “temperament” (emotional reactivity) rather than “personality disorder”. We also submit that temperament, as we define it, embraces both affective liabilities and assets, which makes it more attractive for theory, research, and practice (Akiskal and Akiskal, 1988, Akiskal and Akiskal, 1992, Akiskal, 1996).

While our early work on affective temperaments did influence other clinical researchers (Depue et al., 1981, Klein, 1990, Gunderson et al., 1994), it had marginal influence on the actual formulation of axis II personality types in such formal diagnostic systems as the DSM-III and DSM-IV (American Psychiatric Association, 1980, American Psychiatric Association, 1994) and ICD-10 (World Health Association, 1992). A “depressive personality disorder”, however, did appear in the DSM-IV appendix as a proposed type for further study. By contrast our work (Akiskal et al., 1977, Akiskal et al., 1980) had major impact in formulating the subaffective disorders, leading to the inclusion of cyclothymia and dysthymia, conceived as low grade affective disorders on axis I (DSM-III; American Psychiatric Association, 1980).

The foregoing inconsistent decisions by the architects of the DSM-III and DSM-IV were not due to lack of operationalization of the affective temperaments. Indeed, these manuals (1980) liberally borrowed–almost word for word–our criteria for cyclothymia and dysthymia for axis I use. Nor was it due to the absence of psychometric measures for these affective temperaments; indeed Depue et al. (1981) had developed one for cyclothymia, and eventually Gunderson et al. (1994) developed it for depressive personality. There even existed a scale for hypomanic personality inventory (Eckblad et al., 1986). The inconsistency with which DSM-III and DSM-IV dealt with the affective temperaments was due to the difficulty of resolving what is “subaffective” and what is “personality disorder”—a particularly troubling area of contention between axis I and II (Akiskal et al., 1979, Widiger, 1989, Akiskal and Akiskal, 1992).

With cyclothymic and dysthymic disorders conceptualized as subthreshold mood disorders in these manuals, the old stereotypes of describing mood disorders with the “dramatic” and “anxious” clusters of axis II have continued to dominate clinical and research thinking, despite evidence of their state-dependency (O'Connell et al., 1991). It should also be mentioned in this context that borderline personality disorder nowadays seems to engulf much of the personality disfunction in the affective, particularly bipolar, domain (Akiskal, 2004). It is a curious perversion of history that DSM-IV (2004) uses the Schneiderian perspective-personality disorders as orthogonal to mental disorders classified on axis I—without the benefit of the majestic descriptive terminology for his psychopathic types. For instance, the borderline type could have been termed “labile”, which is a more accurate portrayal of the emotionality of these patients.

Another dilemma that the first author (H.S.A.) faced in his mood clinic was the unwieldy nature of systems of characterizing temperament or personality in formal psychometric testing, requiring 4 to 5 gradations for each trait (see, for instance, the General Behavior Inventory of Depue et al., 1981). Such an approach was obviously impractical in the clinical setting of a service-based mood clinic. That's why, despite great conceptual affinity between our approach and that of Depue to cyclothymia, regrettably, no formal collaboration was forged between us. We also felt that self-report in bipolar patients would be notoriously unreliable. As a result, our MCDQ-based diagnostic approach to the affective temperaments relied heavily on the clinical skills of trainees and attendings working in our setting. This, in turn, explains why, until recently, the use of the temperament constructs under discussion was largely limited to the relatively few psychiatrists who were trained in, or otherwise exposed to, the Memphis (and subsequently to the San Diego) Mood Clinic.

We published the full version of the Memphis clinician interview form for temperaments in 1987 (Akiskal and Mallya, 1987). This paper did in the beginning attract attention primarily because the temperaments could now be used to define “soft bipolar disorder” (e.g., Cassano et al., 1989, Akiskal and Pinto, 1999). Subsequently it was formally adopted for research purposes in an Italian collaboration (e.g., Perugi et al., 1990). Nearly a decade later, Pisa psychiatrists also expressed interest in psychometrically validating the Akiskal–Mallya criteria in a large (n=1010), young (14–25 year-old) community sample evaluated by the interview method (Placidi et al., 1998, Akiskal et al., 1998). This became the TEMPS-I, the first psychometrically valid instrument measuring the affective temperaments in the classic German sense. Concurrent validity against the Temperament and Character Inventory of Cloninger et al. (1994) is just being published for the first time in this issue of the Journal (Maremmani et al., 2005).

In the meanwhile, in part guided by earlier Pisa work, the first two authors (H.S.A. and K.K.A.) reformulated the depressive and hyperthymic types in an expanded version that could be more easily administered by a clinician. For each trait it was necessary to formulate a set of specific questions that define it in language easily understood by the subject being interviewed, thereby making standardization easier. As an illustration, the original operationalized depressive temperament criteria are shown in Table 1, reformulated in a format suitable for full scale standardization in Table 2. We were skeptical whether the cyclothymic could be formatted in an analogous fashion: in particular, our doubts involved the feasibility of a self-rated version for the cyclothymic type. These doubts eventually dissipated, when as part of the NIMH Collaborative Study of Depression, it was found that self-rated trait “mood lability” was a highly specific predictor (86%) of those major depressives who during prospective follow up became bipolar II (Akiskal et al., 1995). The cyclothymic type, then, was expanded from these criteria, building upon our own interview-based operationalization (Akiskal et al., 1977, Akiskal et al., 1979).

While the Italian reformulation of the “Memphis temperament” concepts was in progress, French psychiatrists too expressed interest in the use of affective temperaments. In this dual collaboration termed EPIDEP and EPIMAN, centered in Paris but conducted in four regions of France, the version that served for French translation (Hantouche and Akiskal, 1997) and psychometric validation (Akiskal et al., 2005a, this issue) was intermediate between that of the Italian and San Diego versions. Both interview and auto-questionnaire versions were implemented in French.

The TEMPS-A (autoquestionnaire version) was further developed in San Diego, in the Clinical Research Center, and now bears the title of “Temperament Evaluation of Memphis, Pisa, Paris and San Diego-Autoquestionnaire (TEMPS-A)” (see Akiskal et al., 2005b, this issue).

Parenthetically, our French collaborators were also interested in the concept of anxious temperaments, which were originally part of the MCDQ in a rudimentary form. These preliminary criteria rather than being tested in psychiatric settings were first tested in a French general medical practice population. The latter did uphold a heterogeneous tripartite conceptualization of the anxious temperaments, in which the generalized anxious temperament nonetheless occupied a central position (Hantouche and Akiskal, 2005, this issue). However, the French study on anxious temperaments could not have examined the latest version of the generalized anxious temperament (GAT), because that version was published in 1998, subsequent to the initiation of the French National Study of EPIDEP in 1994. Indeed, the definitive version of the GAT, was ultimately published as part of a theoretical paper on the evolutionary functions of worrying conceived as an altruistic trait (Akiskal et al., 1998). Finally, as collection of data on the Memphis mood clinic sample was initiated before the development of the generalized anxious temperament, the present contribution deriving from that clinic does not include examination of the anxious temperament along with the other four.

The San Diego International Mood Center has tested the TEMPS-A in a research population (Akiskal et al., 2005b, this issue), has collaborated on a German Mûnster version (Erfurth et al., 2005, this issue) that tested it in a clinically well student population. The Japanese (Akiyama et al., 2005, this issue) and the Turkish (Vahip et al., 2005, this issue) versions, have their own unique cultural context. Many of the respective principal investigators of these different language versions spent short training periods in San Diego. Arabic, Danish, Greek, Hungarian, Polish, Portuguese, Spanish and Swedish versions are now in the process of being tested. All in all, counting the Italian and French versions, TEMPS now exists in 12 languages in addition to American English.

The present contribution derives from the mood clinic setting, and it is meant to be a contribution for clinical practice. It is noteworthy that all versions of TEMPS-A, whether slated for use in normal or affectively ill subjects, incorporate traits which we have formulated to reflect evolutionary functions, i.e., falling in and out of love, being the boss over a territory, being a follower, etc. This is based on work conducted earlier in Memphis and Paris on the adaptive and “positive” facets of temperaments (Akiskal et al., 1979, Akiskal and Akiskal, 1988, Akiskal and Akiskal, 1992).