Classification of adult patients with type 2 diabetes using the Temperament and Character Inventory


*Noriko Yoda, MD, Department of Psychiatry Graduate School of Medical Science, Kyoto Prefectural University of Medicine, 465 Kajii-cho Kawaramachi-Hirokoji Kamigyo-ku, Kyoto 602-8566, Japan. Email:


Aim:  Adult patients with type 2 diabetes were classified using the Temperament and Character Inventory (TCI). This classification was used to propose effective therapeutic approaches based on subtypes of psychological characteristics.

Methods:  The TCI and various psychological tests were administered to 89 patients (54 men and 35 women). Cluster analysis was performed using three temperament factors of the TCI as variables for subclassification: novelty seeking (NS), harm avoidance (HA) and reward dependence (RD).

Results:  The patients were divided into two clusters based on the TCI results. Cluster 1 had a low NS/high HA/low RD pattern, which indicated resistance to change and lack of cooperativeness, and results of other tests indicated that patients in cluster 1 were obsessional and had weak intention and personality disorder symptoms. In contrast, cluster 2 had a high NS/low HA/high RD pattern, indicating a socially active person who easily becomes dependent on others, and results of other tests indicated that these patients were histrionic and less anxious. In cluster 1 there was a significant negative correlation between hemoglobin A1c (HbA1c) and SD-1 (TCI self-directedness) ‘responsibility’ and a significant positive correlation between HbA1c and State anxiety of State–Trait Anxiety Inventory. In cluster 2 there was a significant positive correlation between HbA1c and HA-2 ‘fear of uncertainty’.

Conclusions:  Patients with type 2 diabetes were classified using the TCI into two subgroups. These two groups differed in psychological characteristics and had a different pattern in correlation with glycemic control.

THE THERAPEUTIC TARGET of diabetes treatment is maintenance of good glycemic control and prevention of complications. This requires patient self-management of activities such as exercise and eating habits. Because therapeutic effects vary depending on the degree of self-management, techniques to promote these skills are necessary1 and psychotherapy is reported to be effective.2 For effective support, an understanding of the psychological characteristics of the patient is necessary.

In previous studies, depression has been found in approximately 30% of all diabetic patients,3 40% of diabetic patients under treatment had anxiety symptoms,4 and diabetes doubled the risk of depression.5 Factors associated with poor glycemic control include depression,2,6 anxiety,7 alexithymic tendency,8 extreme personality features,9 and stress load,10 whereas problem-solving ability11 and diabetes locus of control12 are factors associated with good glycemic control. An empowerment approach with psychological support to increase the self-worth and self-control of the patient has been shown to increase the therapeutic effect.13 Therefore, psychological conditions have a major influence on glycemic control, and effective treatments have been designed based on these findings.

The current study was based on the hypothesis that there are plural personality patterns in diabetic patients. This hypothesis was established because contradictory findings regarding diabetic characteristics have been reported. One study found that fewer negative emotions and altruistic features are associated with poor glycemic control,14 whereas lack of relationship between personality and level of glycemic control has also been reported.15 We began to suspect that diabetic personality is not homogenous or stereotypical; that is, diabetic personality may be subclassifiable and this may lead to treatment approaches that correspond to particular characteristics and thus may be more effective. The study design was planned on this basis; we first divided the patients into subgroups, and then investigated differences in the relationship between glycemic control and psychological test results.

The Temperament and Character Inventory (TCI) was used for classification of the personality of patients with type 2 diabetes. The degree of glycemic control has not been investigated previously with regard to personal ‘temperament’ and ‘character’, which form the basis of personality independent from psychological conditions according to the Cloninger personality theory.16–19 This theory is based on Cloninger's original biosocial model and the TCI is reported to be effective for classification of patients based on personality differences.20–22



The subjects were adult patients with type 2 diabetes who were under treatment at the outpatient clinic of the Department of Endocrinology and Metabolism, University Hospital, Kyoto Prefectural University of Medicine, between 28 November 2005 and 12 September 2006, and who gave informed consent after receiving an explanation of the study. The subjects took home the psychological test questionnaire, filled it in, and sent it back by mail. The questionnaire was distributed to 146 patients and answers were returned by 89 patients (return rate: 61.0%). The study was approved by the Ethics Committee of the University on 19 October 2005.


Evaluation of glycemic control

The mean hemoglobin A1c (HbA1c) level for 1 year was used as an index of glycemic control. Patients with mean HbA1c levels <8.0% and ≥8.0% were placed in the good and poor glycemic control groups, respectively, according to the American Diabetes Association criterion.23

Evaluation of psychological characteristics

The patients completed the Japanese editions of the following self-reported questionnaire-based psychological tests, all of which have been verified for their reliability and validity: the TCI,24 the Self-Rating Depression Scale (SDS), the State–Trait Anxiety Inventory (STAI), the 20-item Toronto Alexithymia Scale (TAS-20),25 the Stress Coping Inventory (SCI), and the Parental Bonding Inventory (PBI). The details of these tests are as follows.

The TCI is a self-reported questionnaire established based on Cloninger's personality theory. The Japanese edition, version 1.0,24 which was translated from the original TCI version 9 (revised in 1992), was used in the study. The patients answered yes or no to 240 questions associated with ‘temperament’ and ‘character’: ‘temperament’ is subdivided into four factors (novelty seeking, NS; harm avoidance, HA; reward dependence, RD; and persistence, P), and ‘character’ is subdivided into three factors (self-directedness, SD; cooperativeness, C; and self-transcendence, ST). Each factor is further classified into 1–5 items.

PBI assesses parental rearing attitudes, based on the memory of subjects regarding their parents in their first 16 years.

The TAS-20 measures alexithymic tendency using 5-step answers to 20 questions. Questions consist of three factors: ‘difficulty in identifying feelings’, ‘difficulty in describing feelings’, and ‘externally oriented thinking’.

The SCI measures coping with stress, using 3-step answers to 64 questions. Coping methods are approximately divided into cognitive and emotional strategies.

The SDS measures depressive tendency, using 4-step answers to 20 questions; and the STAI measures anxiety, using 4-step answers to 20 questions each on state and trait anxieties.

Statistical analysis

For subclassification of the patients, a two-step cluster analysis was performed using the NS, HA, and RD scores in the Temperament category of the TCI as variables. The correlation between the mean HbA1c value for 1 year and each psychological test score (TAS-20, STAI, SDS, PBI, SCI, and TCI) was calculated and a t-test was performed for each cluster. Regression coefficients of psychological test items that correlated with HbA1c were calculated in each cluster to investigate the relationship between psychological features and glycemic control. SPSS Version 12.0J for Windows (SPSS Inc, Chicago, IL, USA) was used, and P < 0.05 was regarded as significant.


Clinical background

The clinical background of the patients is shown in Table 1. The 89 patients (54 men, 35 women) had a mean age of 63.34 ± 10.21 years and a mean HbA1c level of 7.39 ± 1.18%. The good and poor glycemic control groups included 73 and 16 patients, respectively. Six patients received only dietary and exercise therapy, 33 patients were treated with oral hypoglycemic agents, and 50 with insulin self-injection.

Table 1.  Subject characteristics
 All patientsCluster 1Cluster 2χ2 test or t-test
  • Comparison between cluster 1 and cluster 2.

  • Glycemic control: good control, HbA1c < 8.0%; poor control, HbA1c ≥ 8.0%.

  • HbA1c, hemoglobin A1c.

n (%)8943 (48.31)46 (51.69)
Gender (M/F)54/3527/1627/19n.s.
Age (years)63.34 ± 10.2165.00 ± 9.7161.78 ± 10.53n.s.
HbA1c (%)7.39 ± 1.187.57 ± 1.337.22 ± 1.00n.s.
Glycemic control (good/poor)73/1633/1040/6n.s.
Duration of illness (years)14.10 ± 10.6015.19 ± 10.8513.09 ± 10.36n.s.
Treatment period (years)10.12 ± 8.6411.42 ± 8.938.91 ± 8.27n.s.
Diabetic complication (+/–)67/2233/1034/12n.s.

Cluster analysis for subclassification

The results of cluster analysis for subclassification of patients are shown in Fig. 1. Using NS, HA and RD as variables in the subclassification, the patients could be divided into two clusters. A Bonferroni multiple comparison categorized clusters 1 and 2 as including patients with a low NS/high HA/low RD pattern and an inverse high NS/low HA/high RD pattern, respectively.

Figure 1.

Cluster analysis: T scores for novelty seeking (NA), harm avoidance (HA), and reward dependence (RD) on the Temperament and Character Inventory for (a) cluster 1 and (b) cluster 2. Data were analyzed using the Bonferroni method for all pairwise comparisons.

Clinical background in each cluster

The clinical background of patients in each cluster is shown in Table 1. Clusters 1 and 2 consisted of 43 (48.3%) and 46 patients (51.7%), respectively. There were 27 men and 16 women in cluster 1, and 27 men and 19 women in cluster 2, and the mean ages were 65.0 ± 9.7 and 61.8 ± 10.5 years, respectively. The mean HbA1c values were 7.6 ± 1.3% and 7.2 ± 1.0%, respectively. Glycemic control was good in 33 patients and poor in 10 in cluster 1, and good in 40 patients and poor in 6 in cluster 2. One, 12, and 30 patients received dietary and exercise therapy, oral hypoglycemic drugs, and insulin self-injection, respectively, in cluster 1; and five, 21, and 20 patients in cluster 2 received these respective treatments. There were no significant differences in gender, age, HbA1c level, percentage of patients with good glycemic control, duration of illness, treatment period, and diabetic complications between the clusters.

Psychological test scores in clusters

The means and standard deviations of the psychological test scores in each cluster are shown in Table 2. Significant differences between the clusters were indicated on t-test. Cluster 1 was characterized by a low NS/high HA/low RD pattern; SD and C were low; the TAS-20 II score was high, showing difficulty with describing feelings; and the STAI-1 and -2 and SDS scores were high, showing anxiety and depression. In cluster 2, in addition to the high NS/low HA/high RD pattern, SD and C were high; the TAS-20 II score was low, indicating no problems with awareness and verbalization of feelings; and the STAI-1 and -2 and SDS scores were low, showing that anxiety and depression were low. There were no significant differences between the PBI and SCI scores in each cluster.

Table 2.  Clusters 1 and 2: descriptive statistics (mean ± SD)
 Cluster 1 (n = 43)Cluster 2 (n = 46)
  • **

    P < 0.01.

  • PBI, Parental Bonding Inventory; SCI, Stress Coping Inventory; STAI, State–Trait Anxiety Inventory; TAS-20, 20-item Toronto Alexithymia Scale; TCI, Temperament and Character Inventory.

 Difficulty identifying feelings15.05 ± 5.7413.28 ± 4.51
 Difficulty describing feelings**14.28 ± 4.2111.70 ± 3.51
 Externally oriented thinking20.58 ± 3.4318.87 ± 4.65
 1 State anxiety**46.52 ± 12.5838.15 ± 10.99
 2 Trait anxiety**47.95 ± 12.1538.50 ± 10.07
 SDS**44.33 ± 8.2536.24 ± 8.56
 Paternal care21.53 ± 7.5721.78 ± 7.21
 Paternal overprotection11.60 ± 6.0510.88 ± 5.42
 Maternal care26.57 ± 5.1526.80 ± 6.26
 Maternal overprotection11.31 ± 5.689.64 ± 6.24
 Cognitive strategy26.88 ± 13.7127.78 ± 12.90
 Emotional strategy24.56 ± 10.7123.67 ± 10.39
 Novelty seeking**14.79 ± 4.2619.20 ± 4.71
 Harm avoidance**24.19 ± 4.0815.15 ± 4.79
 Reward dependence**12.30 ± 3.5915.63 ± 3.36
 Persistence3.98 ± 1.784.33 ± 1.63
 Self-directedness**25.07 ± 6.8730.50 ± 5.21
 Cooperativeness**24.26 ± 6.2528.67 ± 5.51
 Self-transcendence12.40 ± 5.0913.04 ± 6.39

Psychological test scores and HbA1c according to cluster

Significant correlation coefficients between HbA1c level and psychological test scores in each cluster are shown in Table 3. For cluster 1, partial correlation was used because age and HbA1c were significantly correlated. For cluster 2, Pearson's correlation was used because of the absence of a significant correlation with age. In cluster 1 there was a significant negative correlation between HbA1c and SD-1 ‘responsibility’ (r =−0.470, r2 = 0.112, P = 0.029) and a significant positive correlation between HbA1c and STAI-1. In cluster 2 there was a significant positive correlation between HbA1c and HA-2 ‘fear of uncertainty’ (r = 0.334, r2 = 0.111, P = 0.023). No significant correlation was found between HbA1c and scores on psychological tests other than the TCI.

Table 3.  Correlation coefficients (significant values) between test scores and HbA1c
 Cluster 1Cluster 2
  • *

    P < 0.05.

  • Data in cluster 1 are controlled for age; data in cluster 2 are not controlled for age.

  • HA, harm avoidance; HbA1c, hemoglobin A1c; SD, self-directedness; STAI, State–Trait Anxiety Inventory; TCI, Temperament and Character Inventory.

TCI Fear of uncertainty (HA2) 0.334*
TCI Responsibility (SD1)−0.470* 
STAI-1 State anxiety0.435* 


Personality characteristics in each cluster

Cloninger investigated the basic stimulus–response characteristics in traditional personality categories using a combination of three temperament factors (NS, HA and RD) and obtained the results shown in Table 4.16 Based on this analysis, patients in cluster 1 (low NS/high HA/low RD) were obsessional and those in cluster 2 (high NS/low HA/high RD) were histrionic. NS is related to triggering behavior and represents the extent to which novel stimulation is accepted, thereby serving as an index of eagerness for investigation or a preference for repetition, impulsiveness or calm and stoic behavior, and a tendency for spending or saving money. HA is related to inhibition of behavior and is associated with worrying and pessimism, thereby serving as an index of cautiousness and seeking assurance or optimism with overconfidence, a preference for shyness and quiet inactivity or active stimulation, and a need for a long or short time for adaptation. RD is related to maintenance of behavior and is associated with social friendly relations and sensitivity, thereby serving as an index of attaching greater importance to cooperation with companions or distancing oneself from others and hiding inner feelings, and sensitivity or indifference to social admiration, approval, and rejection.

Table 4.  Basic stimulus-response characteristics of traditional personality categories, as defined by Cloninger16
  1. HA, harm avoidance; RD, reward dependence; NS, novelty seeking.


Cluster 1 has a low NS/high HA/low RD pattern that indicates resistance to change in accustomed styles and a lack of flexibility, but tolerance of solitude and engagement in practical thinking unaffected by sentimental feelings. This pattern consists of weak objective-directed behavioral control indicated by the low SD score, low C, poor ability to describe feelings, high anxiety and depression, worry, and easily feeling fatigued. In contrast, cluster 2 consists of a high NS/low HA/high RD pattern that indicates an extrovert, social, emotional, and active personality that adapts to different environments, is easily dependent on others, and is easily affected by sentiments. Social solitude is painful, and persons of this type tend to force themselves to look for adaptation.26 They are able to behave with intention and are cooperative, and their alexithymic tendency, anxiety and depression levels are low.

Clusters and glycemic control

In cluster 1, anxiety increased and responsibility decreased as glycemic control worsened. The characteristics of the patients (low NS/high HA/low RD) suggest that they may have anxiety toward new treatment and may become resistant to it. Such patients may also have anxiety regarding a danger such as hypoglycemia, and may try to prevent it by taking excessive calories, leading to a tendency for hyperglycemia. Regarding their low responsibility, these patients are less aware of self-involvement in their own behavior and its outcomes, and may easily accuse others. Therefore, they may exhibit similar attitudes to treatment for diabetes and life management, such as diets and exercise, leading to poor glycemic control.

Cluster 1 was typed as obsessional, and the SD and C scores were low, suggesting personality disorder symptoms19 and showing weak intention and low C. Such patients are bound by order, perfectionism and unity, and flexibility, openness and efficiency are sacrificed. They demand that others obey their ways, and may strongly resist instructions that intervene in their lifestyles and demand changes in a one-sided manner, even when instructions are provided as therapy. But if these patients are convinced and accept the instructions, rather than being persuaded by a physician, they may maintain glycemic control methodically. Thus, responsibility may be the key to glycemic control in this patient class.

Cluster 2 was typed as histrionic, and glycemic control improved as fear of uncertainty decreased, suggesting the influence of secretion of stress hormones. This can lead to poor glycemic control in patients who are sensitive to uncertainty and easily feel anxiety, although these patients are included in cluster 2, in which depression and anxiety are relatively low. These patients are less likely to have a fear of new treatment and tend to respond bravely to such treatment, which may lead to good glycemic control. Cluster 2 patients are originally less anxious: the SD and C scores were high, showing intentional behavior and C, and they do not have difficulty in establishing a therapeutic relationship. But fear of uncertainty (i.e. cautiousness) may interfere with glycemic control. These correlations should be carefully considered because the study had a cross-sectional design and glycemic control may have affected the test scores.

For some cluster 1 patients the following psychological support may be necessary: drug therapy and/or cognitive therapy reducing depression and anxiety, elevation of awareness of self-responsibility and promotion of active involvement to increase self-worth, supporting autonomy to increase feelings of control,27 setting therapeutic goals with the support of medical staff rather than being given treatment by attending physicians, therapy attaching greater importance to self-directed individual styles, and narrative-based medicine, in which patients ‘spin the story’ themselves.28,29

In contrast, for cluster 2 patients, group treatment is appropriate because they are social and cooperative. The presence of other patients as models reduces the fear of uncertainty, which may lead to better glycemic control. A previous study has shown that patients treated in group consultations had better glycemic control than control subjects.30 Active introduction to groups, such as a diabetes group education class, is appropriate because a syntonic atmosphere is better than a focus on individuality. Cultivation of a sense of solidarity with the entire group, that is, heading for a common goal through teaching and encouragement of each other, may be more effective for patients in cluster 2.

Limitations of the study

There were four major limitations in the study: the relatively small sample size; lack of controls; cross-sectional research design that might have resulted in current glycemic control affecting the test scores; and the concern that diabetic patients do not typically have particularly pathological personalities, compared with personality deviations of anorexic or bulimic patients or patients with alcohol dependence. This may lead to a limitation in demonstrating a significant relationship between glycemic control and personality features in diabetic patients. These limitations need to be addressed in a future study with a greater number of patients and healthy controls with a longitudinal research design.


Patients with type 2 diabetes were classified into two subgroups using the TCI. These groups differ in psychological characteristics, which had different patterns in correlation with glycemic control. The results suggest that diabetes self-management support should be based on the psychological characteristics of subtypes of patients.