Eating style in seasonal affective disorder: Who will gain weight in winter?

Eating Style in Seasonal Affective Disorder: Who Will Gain Weight in Winter? Kurt Kr~iuchi, Simone Reich, and Anna Wirz-Justice Patients with seasonal affective disorder (SAD) selectively eat more carbohydrates (CHO), particularly sweets but also starch-rich foods, during their depression in winter. The Dutch Eating Behaviour Questionnaire (DEBQ) was administered to female SAD patients, healthy female controls, and female medical students to determine their eating style, together with the modified Seasonal Pattern Assessment Questionnaire (SPAQ+). SAD patients showed higher values for "emotional" (EMOT} eating than the students, and these in turn had higher values than the controls. In comparison to controls, SAD patients and students had high values for the factor "external" (EXT) eating, but there was no difference between the groups with

respect to "restraint" (REST) eating. This is in strong contrast to patients with bulimia and anorexia nervosa, who are high REST eaters, indicating that SAD patients do not have a similar eating disorder. Additional items showed that SAD patients selectively eat sweets under emotionally difficult conditions (when depressed, anxious, or lonely). Configural frequency analysis showed that seasonal body weight change (SBWC) is high in subjects with high EMOT and REST eating together with a high body mass index (BMI). This result is in accordance with the concept of disinhibition of dietary restraint in extreme emotional situations, e.g., the depressive state.

EASONAL AFFECTIVE DISORDER (SAD) is characterized by recurrent episodes of major depression in autumn and winter that remit in spring and summer. 1 Often, atypical symptoms are present and are highly correlated with each other: increased sleep need, fatigue, and appetite, particularly for carbohydrates (CHO), together with weight gain. 2 We have replicated the selective nutrient choice of SAD patients in a series of independent studies. 3-5 The highest CHO intake occurs in the afternoon and evening. 4 Successful light therapy selectively suppresses CHO intake. 4 Additionally, patients with the highest sweets intake in the second half of the day respond best to light treatment. 5 SAD patients not only have appetite and weight disturbances, but also dysfunctional eating attitudes; however, these are not as extreme as those of bulimics. 6 A significant correlation between seasonal body weight changes (SBWC) and body dissatisfaction was interpreted as discomfort with the body shape during winter when weight is increased. 6 However, the eating style of SAD patients has not been studied in detail. There are valid and reliable instruments for measuring human eating style, such as the Dutch Eating Behaviour Questionnaire (DEBQ). 7,8 The DEBQ has been developed to

distinguish three factors based on three concepts of human eating behavior. First, the factor of "external" (EXT) eating is derived from the cognitivephysiological theory of emotions. 9 It deals with the question of whether persons eat more than normal because they are more easily steered by external stimuli--the sight, smell, amount, and availability of food, the time-of-day signal, and the lack of clearly recognized internal signals of hunger and satiation. ~°-12 Second, the factor of "emotional" (EMOT) eating is based on a psychosomatic concept of obesity, 13 which refers to the tendency of some persons to deal with anxiety, insecurity, irritability, and depression with increased eating. 14,15 This behavior can be learned in early childhood as a way of solving emotional problems via food. 16,17Third, the factor of "restraint" (REST) eating is based on the tendency to restrict food intake to maintain body weight or to promote weight losslS--a strategy clearly characteristic of certain clinical eating disorders (e.g., bulimia and anorexia nervosa). 8 The DEBQ can be used to examine the relationship between REST eating (an inhibition factor) and both EMOT and EXT eating (disinhibition factors) in the light of restraint theory) Efforts to establish determinants of the direction and extent of weight change in depression have used two psychometric measures: dietary restraint, a measure that discriminates dieting, weightconscious individuals from nondieters, and disinhibition of such dietary restraint. Depressive patients who scored high on the restraint scale also reported that they gained weight when they were depressed.19,2° Scores on the disinhibition factor corre-

S

From the Psychiatric University Clinic, Basel, Switzerland. Address reprint requests to Kurt Kriiuchi, Psychiatric University Clinic, Wilhelm KIein-Strasse 27, CH-4025 Basel, Switzerland. Copyright © 1997 by WB. Saunders Company 0010-440X/97/3802-0001 $03.00/0 80

Copyright© 1997by W.B. Saunders Company

ComprehensivePsychiatry, Vol. 38, No. 2 (March/April), 1997: pp 80-87

EATING STYLE IN SAD

81

lated with measured weight change in patients during the course of severe, major depression. 2~ We therefore asked whether these concepts are applicable for seasonal body weight variation, which is a characteristic symptom of SAD. This survey study compared diagnosed SAD female patients with two nonclinical populations, healthy female controls in the same age range and a younger group of female medical students, to answer the following questions: (1) Do SAD patients have a specific eating style (EXT, REST, or EMOT eating)?; (2) Is a particular food choice linked with certain situations or mood states?; and (3) Is there a relationship between SBWC and a specific eating style? METHOD

Subjects and Recruitment

Diagnosed Patients With SAD In the summer of 1992, all 234 patients (183 females and 51 males), who had been previously interviewed during the course of earlier Swiss studies and diagnosed as SAD at that time (Rosenthal et al52 criteria or DSM-III-R2~), were sent a letter inviting them to participate in a questionnaire study during their nonsymptomatic time of year. SAD patients are predominantly women, and thus we restricted our analysis to women only. Eighty-four unmedicated female SAD patients (46% of total 183 female patients) returned a complete set of questionnaires. The demographic characteristics of this subsample did not differ statistically from those of the total group, indicating a representative sample of the Swiss SAD population.

Nonclinical Populations Control group 1 (similar age and body mass index as SAD patients). In the summer of 1992, in parallel to the study with SAD patients, healthy middle-aged women were recruited by word-of-mouth, and 38 completed the set of questionnaires. All participated in a study on seasonal differences in psychological and behavioral measures. Control group H (medical students). In the spring of 1994, third-year medical students were asked during a lecture to participate in the questionnaire study; 42 women returned a complete set of questionnaires. This additional nonclinical population represents a homogenous group with respect to socioeconomic status and demographic characteristics. Students were chosen as a comparison group because they are often in stressful situations that can lead to disturbed eating behavior. 8

Questionnaires The DEBQ The DEBQ is a robust, reliable, and valid instrument that identities three dimensions of eating style: EMOT, REST, and EXT. 7,8 We translated the English version7 into German and added six open questions to determine which foods were eaten in which situations (for the formulation of all questions, see the Results). The answers were classified into specific food/drink

categories (sweets: chocolate, cake, etc.; CHOstarch: pasta, rice, potatoes, bread, etc.; fruits: apples, oranges, etc.; protein: meat, sausage, fish, etc.; dairy products: milk, cheese, yogurt, etc.; caffeine: coffee, tea, and cola; alcohol: beer, wine, and spirits; and fat). 4 The factor structure of the German version of the questionnaire was replicated in our total sample and showed a good cross-validation in comparison to the English version7,8 (data not shown, see the Discussion).

The Seasonal Pattern Assessment Questionnaire The Seasonal Pattern Assessment Questionnaire (SPAQ) was initially developed as a screening for SAD. 2 The SPAQ is an 18-item retrospective self-report providing information concerning the presence and severity of seasonal variation in mood, sociability, energy, sleep, appetite, and body weight. It permits documentation of which items change with the seasons, in which month they change, and the amplitude of these changes. We have expanded the SPAQ with more detailed questions about seasonal sleeping and eating behavior (SPAQ+). 24 Nine specific food and two drink items were added to assess seasonal variation in food selection (e.g., "In which season do you eat most chocolate?", with multiple answers possible). The results of these questions will be punished elsewhere. A comparison between the original SPAQ and the SPAQ+ did not show significant differences and demonstrated a good test-retest reliability. 25

Statistics All statistics were determined using SYSTAT 5.2 for Macintosh (Systat Inc, Evanston, IL). One-way analysis of variance (ANOVA) was calculated for group comparisons. The chisquare statistic (contingency table) was used for frequency comparisons. Post hoc comparisons for the ×2 statistic and for ANOVA were performed with Bonferroni-corrected a values (P < .05). Interrelationships between the three eating factors (EMOT, EXT, and REST) were calculated by Pearson's productmoment correlation (r). To identify significant combinations (configurations) of eating behavior-related variables with SBWC, a hierarchical configural frequency analysis (CFA) was performed according to the method used by Lienert. 26 This nonparametric approach was selected to avoid any problems with nonnormal distribution of the different parameters. Significant differences in configurations between the groups were analyzed with a two-group CFA, which corresponds to a parametric discriminant analysis. 26 The following variables were preselected for CFA: SBWC, BMI, and EXT, EMOT, and REST eating factors. All variables were categorized by median division (median: SBWC, 2 kg; BMI, 21.73 kg/m2; EXT, 1.90; EMOT, 1.46; and REST, 1.65).

RESULTS Demographic Characteristics In comparison to control group I (CON-I) and SAD patients, control group II (CON-II) subjects were younger (mean _+ SD: SAD, 48.0 ___ 12.6 years; CON-I, 49.8 + 16.5; and CON-II, 24.2 _+ 2.9; one-way ANOVA: F(2,161) = 63.17, P < .0001; post hoc: CON-II > CON-I = SAD) and had a lower BMI (SAD, 23.2 _ 2.8 kg/m2;

82

KR,~UCHI, REICH, AND WIRZ-JUSTICE

CON-I, 23.1 _+ 4.1; and CON-II, 20.7 +_ 1.8; oneway ANOVA: F(2,161)= 8.44, P < .0005; post hoc: CON-II < CON-I = SAD). SAD patients reported slightly more life events than the two other groups (CON-I, 15.8%; CON-II, 25.6%; and SAD, 34.9%; X2(df=2) = 4.89, NS). All groups rated present health to be good (ratings: excellent, good, not especially good, and bad; X2(df=6) ~" 0.492, NS).

SPAQ Item and Score Analysis

*

t

3.0

8

2.5

2.0

EXTERNAL

RESTRAINT EATING FACTOR

EMOTIONAL

Fig 2, Eating typology with the DEBQ in SAD patients (n = 84) and 2 nonclinical populations ICON-I, similar age and BMI as SAD patients, n = 38; CON-II, medical students, n = 42), EXT eating factor: 1-way ANOVA, F(2,161) -- 15.5, P < .0001; post hoc, SAD = CON-II > CON-I. REST eating factor: 1-way ANOVA, F(2,161) = 2.50, NS. EMOT eating factor: 1-way A N O V A , F(2,161) = 24,5, P < .0001; post hoc, SAD > CON-II > CON-I. Results are the mean -+ SEM, all women, post hoc. * P < .05 (Bonferroni-adjusted cxvalues).

ria for SAD (SAD, 59.5%; CON-I, 13.2%; and CON-II, 16.7%) but did show, as expected, significantly more nonseasonals (SAD, 20.2%; CON-I, 73.7%; and CON-II, 61.9%; X2(df=4) ----- 44.22, P < .0001; post hoc: SAD < CON-I = CON-II). However, no significant differences were found with respect to the occurrence of S-SAD between the groups (SAD, 20.2%; CON-I, 13.2%; and CON-II, 21.4%). DEBQ Eating T y p o l o g y

14

~t

m 12

O

~ 6 < z 0 4 < ~

ml

1.5

SAD patients had a significantly higher seasonality score and a higher value for seasonal problems than CON-I and CON-II, who were not statistically different from one another (Fig 1). This was also true for all of the individual items except mood. SAD showed the highest seasonal mood item value, and CON-I the lowest (0 = no seasonal mood change and 4 = extremely marked seasonal mood change) (SAD, 2.64 + 1.06; CON-I, 1.29 _+ 1.11; CON-II, 1.83 + 1.08; one-way ANOVA: F(2,161) = 22.66, P < .0001; post hoc: SAD > CON-I = CON-II). If subjects showed seasonality, all showed a winter pattern as defined by Fornari et al.27 When individuals were defined according to questionnaire answers as SAD, subsyndromal SAD (S-SAD), or within the normal range (nonseasonal) with the criteria used by Kasper et al., 28 there were also significant differences between the groups. In comparison to SAD patients diagnosed as such 2 to 8 years previously, significantly fewer subjects from the nonclinical populations fulfilled the criteFigure 1:

3.5

2

0 Fig 1. SPAQ seasonality score and seasonal problems score in SAD patients (n = 84) and 2 nonclinical populations (CON-I, similar age and BMI as SAD patients, n = 38; CON-II, medical students, n = 42). SPAQ score: 1-way ANOVA, F(2, 161) -- 15.5, P < .0001; post hoc SAD > CON-I = CON-II. Seasonal problems: 1-way ANOVA, F(2,161) = 33.2, P < .0001; post hoc, SAD > CON-I = CON-II. Results are the mean +_ SEM, All women, post hoc. * P < .05 (Bonferroni-adjusted values).

The DEBQ was administered to CON-I and SAD patients in summer, when the latter were no longer depressed; the CON-II group was studied in spring. SAD patients and CON-II showed a higher EXT eating factor than CON-I. The groups did not differ with respect to REST eating. The EMOT eating factor dearly separated the three groups. SAD showed the highest EMOT eating factor value, and CON-I the lowest (Fig 2). The EMOT eating factor was significantly correlated with EXT eating (r = .527, N = 164, P < .0001) and with REST eating (r = .242, N = 164, P < .01), whereas the EXT eating factor was not significantly correlated with REST eating (r = .027, N = 164, NS). Similar correlation coefficients were found for each subgroup separately (SAD, CON-I, and CON-II; data not shown).

EATING STYLE IN SAD

Situative Eating/Drinking Behavior The open questions added to the DEBQ asked which foods/drinks were ingested under which situations or emotions. These were then analyzed with respect to specific macronutrient categories as previously defined 4 (see the Methods). The results are presented in Table 1. All groups restrained their intake of sweets and starches to an equal extent to control weight. They also did not differ with respect to their food choice when tempted by external factors. Only with respect to clearly labeled EMOT eating were significant differences found. In general, SAD patients ate and drank more often than the nonclinical populations when they were in an emotional state. Concerning food/drink selection during depressive mood, SAD patients chose sweets more often than CON-II, and CON-II more often than CON-I. In comparison to CON-I and SAD, CON-II drank fewer caffeine-containing drinks. All other food/ drink categories did not statistically differ with respect to the three groups (SAD patients showed a tendency to select more starch-rich foods than CON-I). The second question related to EMOT eating asked what foods were preferred when subjects were anxious or stressed. Again, SAD patients ate more sweets than CON-I and slightly (NS) more than CON-II. The third question revealed that when they were lonely, SAD patients consumed more sweets than CON-II, and CON-II more than CON-I. SAD patients drank alcohol more frequently than the nonclinical populations. Finally, there was no selective eating behavior when subjects were bored. All groups chose sweets most often under these diffuse emotional conditions. Although not statistically significant, it is noteworthy that SAD patients mentioned alcohol with respect to all four emotional questions. These final questions recapitulate what appears, from our prior studies using food logs, to be characteristic for SAD patients3-5: they prefer CHO, primarily as sweets, and eat these when depressed, anxious, or lonely.

Relationship Between SBWC and Eating Style Compared with SAD, CON-I and CON-II reported significantly lower SBWC (SAD, 3.0 --+ 2.0 kg, range 0 to 8; CON-I, 1.6 -+ 1.2 kg, range 0 to 4; and CON-II, 1.4 ___ 1.5 kg, range 0 to 7; one-way ANOVA: F(2,161)= 15.4, P < .000l; post hoc:

83

Table 1. Comparison of Situative Food/Drink Intake in SAD Patients and Two NonClinical Populations

Intake

DiagnosedSAD NonclinicalPopulations (n = 84) CON-I (n = 38) CON-II (n = 42)

Depressed mood: "When you are depressed or discouraged do you eat/drink certain foods or drinks? When yes, what?" Sweets 57.1"t 13.2t 35.7 CHOstarch 17.9 2.6 7.1 Fruits 2.4 0 0 Caffeine 21.4t 21.1t 2.4 Alcohol 14.3 2.6 2.4 Dairy products 11.9 5.3 0 Anxiety: "When you are anxious, worried or tense do you eat/drink certain foods or drinks? When yes, what?" Sweets 27.4* 5.3 16.7 CHOstarch 10.7 0 2.4 Fruits 3.6 0 0 Caffeine 8.3 10.5 2.4 Alcohol 16.7 10.5 2.4 Dairy products 9.5 2.6 0 Loneliness: "When you are lonely do you eat/drink certain foods or drinks? When yes, what?" Sweets 38.1"t 7.9t 16.7 CHOstarch 9.5 5.3 4.8 Fruits 4.8 2.6 0 Caffeine 10.7 7.9 2.4 Alcohol 14.3t 2.6 0 Dairy products 7.1 2.6 0 Boredom: "When you are bored do you eat/drink certain foods or drinks? When yes, what?" Sweets 20.2 15.8 23.8 CHOstarch 9.5 7.9 2.4 Fruits 7.1 10.5 4.8 Caffeine 10.7 13.2 2.4 Alcohol 6,0 2.6 0 Dairy products 8,3 2.6 0 EXT." "Do certain foods or drinks tempt you to eat/drink? When yes, what?" Sweets 22,6 15.8 19.0 CHOstarch 6.0 2.6 9.5 Crackers 2.4 0 0 Protein 1.2 2.6 2.4 Fruits 1.2 2.6 2.4 Caffeine 6.0 13.2 2.4 Alcohol 14.3 10.5 2.4 Dairy products 2.4 2.6 0 REST: "When you are watching your weight and try to eat/ drink less, do you avoid certain foods ro drinks? When yes, what?" Sweets 53.6 42,1 61.9 CHOstarch 33.3 15,8 28.6 Protein 7.1 10.5 2.4 Fat 19.0 23,7 21.4 Alcohol 6.0 10.5 0.0 NOTE. Values are the percentage of subjects per group who reported food/drink intake of the respective food/drink category. ×21d~21 statistic based on Bonferroni-adjusted values (e.g., critical c~ value for depressed mood questions, .05/6 = .0083; group comparisons only performed when previous test provded significant, critical value, .05/3 = .017. * P < .05 vCON-I. t P < .05 vCON-[I.

84

SAD > CON-I = CON-H); 58.3% of SAD patients, 18.4% of CON-I, and 23.4% of CON-II showed a relevant SBWC greater than 2 kg (X2(~f=2)=23.7, P < .0001; post hoc: SAD > CON-I = CON-II). To find predictors for SBWC, significant combinations (configurations) of eating behavior-related variables with SBWC were sifted out by a fivedimensional hierarchical CFA 26 of the five preselected variables for the pooled groups (N = 164). Two of a total of 32 configurations were significant (total X2(df=26) = 108.2, z = 6.67, P < .0001). One significant configuration included all variables below the median (post hoc t e s t : X 2 ( d f = l ) = 47.37, P < .0001) and the other contained all variables above the median (post hoc test: X2(af=l)= 27.95, P < .0001). Because some of the configurations showed the expected values less than five (for definition, see Lienert26), the most significant four dimensional CFA, which contained the variables EMOT, REST, BMI, and SBWC, was chosen for detailed analysis (Fig 3; total X2(df=ll)= 66.92, z = 5.95, P < .0001; all three- or two-dimensional CFAs exhibited lower significance values). Again, two significant configurations were found: one included all variables below the median (Fig 3; post hoc test, all negative: X2(df=l) = 30.86, P < .0001) and the other contained all variables above the median (Fig 3; post hoc test, all positive: X2(df=l) -~20.14, P < .0001). All expected values were above five, indicating a correct ×2 analysis.26 Similar results were found for SAD patients analyzed alone (Fig 3A; post hoc test, all negative: XZ(df=l) = 45.10, P < .0001; all positive: X2(af=1) = 9.22, P < .05). To define significant discriminative configurations between SAD patients and controls, multigroup CFAs were calculated. 26 Based on the results of the hierarchical CFA, four-dimensional multigroup CFAs were performed (dimensions: SBWC, EMOT, REST, and BMI). In a first step, no significant differences between CON-I and CON-II were f o u n d (X2(df=]5) = 14.98, NS), and these subjects were therefore combined into a single control group. A further two-group dimensional CFA showed significant discriminative configurations between SAD patients and the pooled controls (Fig 3; XZ~d/_~5)= 43.98, P < .0001). The best discriminative configuration between SAD patients and nonclinical subjects contained all variables above the median (post hoc test, all positive: X2(df=~)= 14.86, P < .0001). A higher frequency of SAD patients than controls showed this configuration

KR,~,UCHI, REICH, AND WIRZ-JUSTICE

"k

.25 SAD

(N=84)

.20

-k .15 oel .10

Z .05

O~ 0

:B

.25

F< ,.d

.20

,15

.10

.05

0 REST

. . . .

+

++

+

. . . .

+

+

+

+

EMOT

-

-

+

+

-

- +

+

-

-

+

+

-

-

+

+

BMI

-

+

-

+

-

+

+

- +

-

+

-

+

-

SBWC

.

.

.

.

++

++

.

.

.

.

+ ++

+

+

Fig 3. Summary of CFA of SBWC, BMI, and EMOT and REST eating factors. All variables were categorized by median division (+, above median; - below median). (A) Distribution of configurations (relative frequencies of n = 84} in SAD patients (X2td~_llj = 66.52, P < .0001). *Significant configurations (P < .05) within the SAD group (Bonferroni-adjusted value, .05/16 = .0031). The same configurations were significant in the pooled sample (N = 164). (B) Distribution of configurations (relative frequencies of n = 80) in the pooled control population (X2(~_111= 40.08, P < .0001; CON-I and CON-II did not statistically differ). §Significant discriminative configurations between SAD patients and nonclinical subjects (P 8,05; Bonferroni-adjusted ~ value, .05/16 = .0031 ). The two configurations with a statistical trend (§P < .1) became significant after agglutination (X2(~=21 = 16.76, P < .0002, indicated with a bracket).

(Fig 3). The two configurations with a statistical trend (P < .1) became significant after agglutination (post hoc test: X2(df=2)-~ 16.76, P < .0002). This agglutinated configuration contains REST eaters without an EMOT eating style and a small SBWC independent of BMI. In comparison to SAD patients, a higher frequency of nonclinical subjects exhibited this agglutinated configuration (Fig 3). DISCUSSION

According to the SPAQ, 26.4% of the nonclinical control population CON-I fulfilled the question-

EATING STYLE IN SAD

naire-based criteria for S-SAD or SAD. 28A comparable percentage (31%) has been found in a previous survey of a representative random sample in Basel, 29 indicating a high prevalence for seasonal problems in this part of Switzerland. Thus, the controls were an appropriate nonclinical group without preselection for absence of seasonality. The DEBQ provided new insight into the often reported phenomenon of seasonal changes in appetite and body weight in SAD. In both nonclinical populations, CON-I and CON-II, values for the three factors of the DEBQ were not statistically different from those previously published for healthy Dutch women 7 and a group of female British students, 8 respectively. Additionally, in all three groups of subjects, the intercorrelations between the three factors were similar to the original results, with no relationship between REST and EXT eating, a fairly strong relationship between EMOT and EXT eating, and a weak but significant correlation between EMOT and REST eating. 7 Comparable to previous studies of the DEBQ, the factor structure of our questionnaire data again showed a striking stability. 7,8 Eating style did not change with depressive state (SAD before and after light treatment, n = 13; data not shown) or according to season (winter v summer, n = 12 for SAD and n = 40 for control pairs; data not shown). Thus, this reliability suggests that the DEBQ as used in this German translation documents a very stable individual trait. The DEBQ clearly separated the two nonclinical populations from diagnosed SAD patients. In comparison to the nonclinical populations, SAD patients showed the highest EMOT eating factor, and CON-II scored higher on this factor than CON-I. SAD patients and CON-II exhibited a similar EXT eating factor, which was higher than in CON-I. No difference between the groups was seen in REST eating. Similar results were found with age- and BMI-adjusted values (data not shown), indicating a real group-specific finding independent of demographic differences (CON-I! were younger and slimmer than the two other groups). Based on these findings of high values of disinhibition factors, SAD patients in particular, but also CON-II (students), represent high-risk groups for eating disorders. 7,8 Indeed, the prevalence rate for bulimia in a student population has been estimated to be higher than in a community sample of older adults. 3° In a Swiss student population, a relatively high preva-

85

lence rate of about 4% has been found. 31 Recent reports have drawn attention to the similarity in certain clinical symptoms between bulimic and SAD patients. 6,32-34 Patients with bulimia nervosa may have winter worsening of binge eating, 6,32-34 and respond to light therapy 35 and serotonergic medication. 36,37 Thus, there seems to be some clinical justification for the original suggestion of a comparable eating disorder etiology in SAD and bulimia. 3s We found that SAD patients in particular, but also CON-II, were strongly motivated to eat by EMOT and EXT factors without showing REST eating more often than CON-I. This contrasts with the demonstrated high values for REST eating in bulimic patients. 8 The contrasts in eating style with respect to REST eating suggest that bulimics and SAD patients can be separated. It is important to note that these results were found in summer when SAD patients were not depressed, indicating that SAD patients have a disturbed eating style that is independent of clinical state. It may be that this trait of SAD patients leads to disinhibited eating during autumn and winter when they are depressed. CHO craving is one of the core symptoms associated with SAD, and this is reflected in a specific increase in CHO-rich food intake during the winter depressive phase. 2 The SPAQ+ showed that the amplitude of this seasonal food choice in SAD patients is particularly large for chocolate, cakes, and pastries, and the increase begins earlier than for controls (data not shown; see Kr~iuchi and Wirz-Justice29). The additional items of the DEBQ, which were open questions without any indication of the kind of answer, clearly demonstrated that SAD patients specifically chose CHO--particulady sweets--when they were depressed, anxious, or lonely. Again, CON-II was situated between CON-I and SAD. Therefore, during depressive states, increased CHO intake may represent a self-healing attempt. A comparable finding has been seen in another atypical form of depression, the hysteroid dysphoria state, which is also often accompanied by overeating or craving for sweets. 39 It is well known that the sensation of sweetness or bitterness initially serves as one of the first attractive or repellent sensory experiences ("early learning of pleasure and aversion"). J7 But adult humans also give the highest "hedonic ratings" to substances that are sweet, particularly if they are combined with a high fat content. 4° Thus, during the depressive state, behaviors are activated that

86

KRAUCHI, REICH,AND WlRZ-JUSTICE

were positively experienced in early childhood (regressive behavior). However, not only can developmental psychological parallels be drawn, but also physiological differences. In the allesthesia test, SAD patients perceive high sucrose concentrations as more pleasant when they are depressed than when they are euthymic after light therapy or in summer. 41 This change in hedonic ratings is reflected in changes in metabolic functions: in the glucose tolerance test, SAD patients have impaired insulin sensitivity only during the winter depressive phase. 41 Taken together, these results suggest that a real "metabolic demand" for glucose underlies the characteristic CHO craving in SAD patients during their winter depression. Another core symptom associated with SAD is the SBWC. We have previously reported that SAD patients who more often eat CHO, particularly starch-rich foods, during their depressive phase in autumn and winter were those with the largest seasonal body weight increase. 3,4 Weight change has long been recognized as an important symptom in affective disorders in general. 42 Although weight loss is more commonly reported, up to a third of patients report weight gain. 21,43A number of clinical features of depression have been examined as putative predictors of the direction and extent of weight change in individuals with major depression, but with inconclusive results. Dietary restraint 19,2° and disinhibition of dietary restraint 21,44 have been reported to distinguish weight gainers from weight losers. However, this relationship between disinhibition and weight change was no longer significant when BMI, age, and sex were c o n t r o l l e d for. 45 Moreover, the BMI itself is significantly correlated with weight change, suggesting that the heavy get heavier and the light get lighter when depressed. 45 However, all these published results have been found in patients on antidepressant medication. Therefore, eating behavior and

body weight regulation during the course of a depressive episode may be confounded by pharmacological treatment. For a positive or negative energy balance, not only is food intake important, but also energy expenditure (exercise, diet-induced thermogenesis, and resting metabolic rate). Therefore, a pharmacological intervention may interact in a complex manner (e.g., appetite and metabolic processes). The advantage of our study was that none of the SAD patients were on antidepressant medication and therefore were free of such confounding influences. The CFA showed that SAD patients with a large BMI who reported high REST eating in addition to high EMOT eating showed a large ( > 2 kg) body weight increase in winter (this configuration was more often found in SAD patients than in the nonclinical populations). Conversely, SAD patients with a small BMI who reported a low REST eating in addition to a low EMOT eating showed a small body weight increase in winter. Additionally, control subjects more often manifested a "successful restraint" eating style that led to no SBWC (independent of BMI). However, this successful restraint eating in controls was only seen when no disinhibiting EMOT eating factor was present. These findings are in accordance with the concept of disinhibition of dietary restraint in extreme emotional situations, e.g., during the depressive state. The determinants of seasonal changes in CHO intake, appetite, and body weight in connection with depressed mood are still unknown. Their elucidation would improve our understanding of the link between feeding behavior and depression. ACKNOWLEDGMENT

This report is based on the doctoral thesis of S.R. The SAD patients had all participated in prior studies supported by the Swiss National Science Foundation. We thank Professor J. Ktichenhoff and Dr H.-J. Haug for helpful comments on the manuscript.

REFERENCES

1. RosenthalNE, Sack DI, Gillin JC, LewyAJ, GoodwinFK, Davenport Y, et al. Seasonal affectivedisorder: a description of the syndromeand preliminary findings with light therapy. Arch Gen Psychiatry 1984;4l:72-80. 2. Rosenthal NE, Genhart M, Jacobsen FM, Skwerer RG, Wehr TA. Disturbances of appetite and weight regulation in seasonal affective disorder. Ann NY Acad Sci 1987;499:216230. 3. Kr~iuchiK, Wirz-JusticeA. The four seasons: food intake frequency in seasonal affectivedisorder in the course of a year. PsychiatryRes 1988;25:323-338.

4. Kr~iuchi K, Wirz-Justice A, Graw E The relationship of affective state to dietary preference: winter depression and light therapy as a model. J AffectDisord 1990;20:43-53. 5. Kr~iuchi K, Wirz-Justiee A, Graw E High sweets intake late in the day predicts a rapid and persistent response to light therapy in winter depression. PsychiatryRes 1993;46:107-117. 6. Berman K, Lam RW, Goldner EM. Eating attitudes in seasonal affectivedisorder and bulimia. J AffectDisord 1993;29: 219-225. 7. Van Strien T, Frijters JER, Bergers GPA, Defares PB. The Dutch Eating Behaviour Questionnaire (DEBQ) for assessment

EATING STYLE IN SAD

of restrained, emotional and external eating behaviour. Int J Eat Disord 1986;5:295-315. 8. Wardle J. Eating style: a validation study of the Dutch Eating Behaviour Questionnaire in normal subjects and women with eating disorders. J Psychosom Res 1987;31:161-169. 9. Schachter S. The interaction of cognitive and physiological determinants of emotional state. In: Berkowitz L (ed): Advances in Experimental Social Psychology. Vol. 1. New York, NY: Academic, 1964. 10. Nisbett RE. Taste, deprivation and weight determinants of eating behaviour. J Pers Soc Psychol 1968; 10:107-116. 11. Nisbett RE, Kanouse DE. Obesity, food deprivation and supermarket shopping behaviour. J Pers Soc Psychol 1969;12: 289-294. 12. Schachter S, Gross L. Manipulated time and eating behavior. J Pers Soc Psychol 1968; 10:98-106. 13. Kaplan HJ, Kaplan HS. The psychosomatic concept of obesity. J Nerv Ment Dis 1957;125:181-201. 14. Bruch H. Transformation of oral impulses in eating disorders: a conceptual approach. Psychiat Q 1961 ;35:458-481. 15. Bruch H. Psychological aspects in overeating and obesity. Psychosomatics 1961 ;5:269-274. 16. Bruch H. Hunger and instinct. J Nerv Ment Dis 1969; 149: 91-114. 17. Steiner JE. Human facial expressions in response to taste and smell stimulation. Adv Child Dev Behav 1979; 13:257-295. 18. Herman CP, Mack D. Restrained and unrestrained eating. J Pets 1975;43:647-660. 19. Polivy J, Herman CP. Clinical depression and weight change: a complex relation. J Abnorm Psychol 1976;85:338340. 20. Zielinski JJ. Depressive symptomatology: deviation from a personal norm. J Commun Psychol 1978;6:163-167, 21. Weissenbnrger J, Rush AJ, Gilers DE, Stunkard AJ. Weight change in depression. Psychiatry Res 1986; 17:275-283. 22. Rosenthal NE, Kasper S, Schulz PM, Wehr TA. New concepts and developments in seasonal affective disorder. In: Thompson C, Silverstone T (eds): Seasonal Affective Disorder. London, UK: CNS Clinical Neuroscience, 1989:97-132. 23. Diagnostic and Statistic Manual of Mental Disorders. Ed. 3. Rev. (DSM-III-R). Washington, DC: American Psychiatric Association, 1987. 24. Wirz-Justice A, Kr~iuchi K, Graw P. Epidemiologie der SAD. Ther Woche Neurol Psychiatr 1993;7:478-482. 25. Marriott PF, Greenwood KM, Armstrong SM. An assessment of SPAQ and SPAQ + reliability. Light Treat Biol Rhythms Bull 1993;5:33-34. 26. Lienert GA. Verteilungsfreie Methoden in der Biostatistik. Vol. 2. Meisenheim am Glam, Germany: Verlag Anton Hain, 1978. 27. Fornari VM, Braun DL, Sundy SR, Sandberg DE, Matthews M, Chert L, et al. Seasonal patterns in eating disorder subgroups. Compr Psychiatry 1994;35:450-456. 28. Kasper S, Wehr TA, Bartko JJ, Gaist PA, Rosenthal NE. Epidemiological findings of seasonal changes in mood and

87

behavior: a telephone survey of Montgomery County, Maryland. Arch Gen Psychiatry 1989;46:823-833. 29. Krauchi K, Wirz-Justice A. Seasonal patterns of nutrient intake in relation to mood. In: Anderson GH, Kennedy SH (eds): The Biology of Feast and Famine: Relevance to Eating Disorders. Orlando, FL: Academic, 1992:157-182. 30. Bushnell JA, Wells JE, Hornblow AR, Oakley-Browne MA, Joyce P. Prevalence of three bulimia syndromes in the general population. Psychol Med 1990;20:671-680. 31. Ferrero FP, Rouget PC. Bulimia and weight variations in a Swiss population. Int J Eat Disord 1991 ;10:563-569. 32. Blouin A, Blouin J, Aubin P, Carter J, Goldstein C, Boyer H, et al. Seasonal patterns of bulimia nervosa. Am J Psychiatry 1992;149:73-81. 33. Brewerton TD, Krahn DD, Hardin TA, Wehr TA, Rosenthal NE. Findings from the Seasonal Pattern Assessment Questionnaire in patients with eating disorders and control subjects: effects of diagnosis and location. Psychiatry Res 1992;52:71-84. 34. Lam RW, Solyom L, Tompkins A. Seasonal mood symptoms in bulimia nervosa and seasonal affective disorder. Compr Psychiatry 1991;32:552-558. 35. Lam RW, Goldner EM, Solyom L, Remick RA. A controlled study of light therapy for bulimia nervosa. Am J Psychiatry 1994; 151:744-750. 36. Freeman CPL, Hampson M. Fluoxetine as a treatment for bulimia nervosa. Int J Obes 1987;1 l(3Suppl):171-177. 37. Blouin AG, Blouin JH, Perez EL, Bushnik T, Zuro C, Mulder E. Treatment of bulimia with fenfluramine and desipramine. J Clin Psychopharmacol 1988;8:261-269. 38. Rosenthal NE, Heffernan MM. Bulimia, carbohydrate craving and depression: a central connection? In: Wurtman RJ, Wurtman JJ (eds): Nutrition and the Brain. Vol. 7. New York, NY: Raven, 1986:139-166. 39. Schuman M, Gitlin M J, Fairbanks L. Sweets, chocolate, and atypical depressive traits. J Nerv Ment Disord 1987;175:491495. 40. Drewnowski A, Greenwood MRC. Cream and sugar: human preferences for high-fat foods. Physiol Behav 1983;30: 629-633. 41. Kr~tuchi K, Keller U, Leonhardt G, Hang H-J, Brunner DP, Hetsch C, et al. Depressed SAD patients exhibit impaired insulin-glucose homeostasis and altered palatability of sucrose in winter. Society lbr Light Treatment and Biological Rhythms Abstr 1994;6:33. 42. Kraepelin E. Compendium der Psychiatrie. Vol. 11. Leipzig, Germany: Abel, 1883. 43. Paykel ES. Depression and appetite. J Psychosom Res 1977;21:401-407. 44. Ruderman AJ. Dysphoric mood and overeating: a test of restraint theory's disinhibition hypothesis. J Abnorm Psychol 1985;94:78-85. 45. Stunkard A J, Fernstrom MH, Price RA, Buss E, Frank E, Kupfer DJ. Weight change in depression: influence of "disinhibition" is mediated by body mass and other variables. Psychiatry Res 1991;38:197-200.