Psychiatry Research, 17, 275-283 Elsevier
Weight
Change
Jan Waissenburger,
275
in Depression A. John
Rush,
Donna
E. Giies,
and Albert
J. Stunkard
Received May 6, 1985: revised version received October 31. 198.5; accepted January 31, 1986.
Abstract. This report describes the weight changes of 109 outpatients during the course of a depressive illness and relates these changes to several potential predictors: age, gender, diagnosis, and scores on the Hamilton Rating Scale for Depression (HRSD), the Beck Depression Inventory, and the three factors on the Eating Questionnaire. Weight changes ranged from -33 to +50 pounds, with 40% of the patients reporting weight gain, 30% weight loss, and 30% no change in weight. Weight loss occurred more rapidly than did weight gain. The disinhibition factor of the Eating Questionnaire was significantly correlated with weight change during depression and, on a stepwise discriminant function analysis, differentiated weightgaining from weight-losing patients at a high level of statistical significance. Severity of depression also differentiated weight-gaining from weight-losing patients in the discriminant function analysis, but only on the HRSD and at a level of more modest statistical significance. Key Words. Depression, Weight
loss has been
body weight, affective
recognized
as a feature
disorder,
of depression
restraint, since
disinhibition. Kraepelin’s
classic
description (Kraepelin, 1904; Beck et al., 1961; American Psychiatric Association, 1968; Zung et al., 1974), while weight gain has been recognized more recently (Stunkard, 1957; Feighner et al., 1972; Bruch, 1974; Stunkard and Rush, 1974; Spitzer et al., 1978; American Psychiatric Association, 1980; Davidson et al., 1982). Factors influencing the directon of weight change are unclear. Loss of weight has been related to two factors-severity of depression (Hamilton, 1960; Beck et al., 1961; Zung et al., 1974) and the presence of endogenous depression (Poll&, 1965; Rosenthal and Klerman, 1966; Spitzer et al., 1978). Gain in weight has been related to the converse-mildness of depression and the presence of nonendogenous, neurotic, or atypical depression (Paykel, 1977; Davidson et al., 1982; Liebowitz et al., 1984). The nature of the relationship between these variables and weight change, however, is uncertain. In addition, some research has shown weight loss to be a poor predictor of endogenous depression (Kendell and Gourlay, 1970; Nelson and Charney, 1981). More recently, it has been suggested that psychological traits, notably “restrained eating,” or the tendency to restrict food intake in order to control body weight, may influence weight changes during depression (Herman and Polivy, 1980). Thus, Polivy
Jan Weissenburger, M.A., is Research Scientist, A. John Rush, M.D., is Betty Jo Hay Professor; and Donna E. Giles, Ph.D., is Research Assistant Professor, Affective Disorders Unit, Department of Psychiatry, The University of Texas Health Science Center at Dallas. Albert J. Stunkard, M.D., is Professor, Department of Psychiatry, University of Pennsylvania, Philadelphia, PA. (Reprint requests to Dr. A.J. Rush, Affective Disorders Unit, Dept. of Psychiatry, University of Texas Health Science Center, 5323 Harry Hines Blvd., Dallas, TX 75235, USA.) 0165-1781/86/$03.50
@ 1986 Elsevier Science Publishers
B.V.
276 and Herman (1976) reported that six patients who scored high on their Restraint Scale said that they tended to gain weight during depressive moods, while six patients who scored low on the scale said that they lost weight during such moods. In a related study, patients who scored high on the Restraint Scale also reported depressionrelated weight gains (Zielinski, 1978), while in another study subjects who scored high on the Restraint Scale overate when a depressive mood was experimentally induced (Frost et al., 1982). The promising construct of restrained eating has been further developed in studies that show it is not a unitary concept but comprises three distinctly different factors: “restraint” itself, “disinhibition,” and “hunger” (Stunkard and Messick, 1985). The Eating Questionnaire, which was designed to measure these factors, was administered to 109 depressed patients in the present study. In addition, four other potential predictors of weight change during depression were assessed: age, gender, diagnosis, and severity of depression as assessed by the Hamilton Rating Scale for Depression (Hamilton, 1960) and the Beck Depression Inventory (Beck et al., 1961).
Methods Subjects. Subjects were outpatients evaluated in the Affective Disorders Unit, University of Texas Health Science Center at Dallas, from March 1982 through May 1983, who met Research Diagnostic Criteria (RDC) (Spitzer et al., 1978) for unipolar major depression (n q 93), or bipolar I (n q 7) or II (n = 9) disorder, depressed phase. The majority of patients had no prior treatment during the presenting episode of depression. A small group had received some form of psychological treatment, but patients who were taking antidepressants upon entry into the clinic or who had been on antidepressants at any time during the current episode of depression were not included in the study. To minimize the confounding effects of weight loss, the RDC for endogenous depression were slightly modified to omit weight loss as a criterion. Twenty-seven unipolar patients were thus classified as endogenous, 66 as nonendogenous. Procedure. The Eating Questionnaire (Stunkard and Messick, 1985) was used to elucidate the relationship between eating behavior traits and weight change during clinical depression. The Eating Questionnaire measures three primary dimensions of eating behavior: (1) cognitive restraint of food intake, (2) disinhibition, and (3) hunger (Stunkard and Messick, 1985). Cognitive restraint measures concern about, and conscious control of, eating behavior. Disinhibition measures the tendency toward periodic loss of control of eating behavior and the tendency to eat to relieve emotional states. The hunger factor measures the subjective sensation of hunger. All subjects completed the Sl-item Eating Questionnaire (Stunkard and Messick, 1985) the 21-item Beck Depression Inventory (BDI) (Beck et al., 1961), and weight history and demographic questionnaires. Experienced clinical evaluators, without knowledge of the experimental hypotheses, completed the I7-item Hamilton Rating Scale for Depression (HRSD) (Hamilton, 1960) on all patients. Diagnoses were made according to RDC (Spitzer et al., 1978) based on structured interview, following the Schedule for Affective Disorders and Schizophrenia-Lifetime Version (SADS-L) (Spitzer and Endicott, 1978). Evaluators who conducted the Hamilton ratings, SADS-L, and RDC diagnoses had no knowledge of the Eating Questionnaire data. All patients scored > 10 on the HRSD. Analyses of the HRSD and BDI ratings were based on total score, minus the weight and appetite loss items. Patients were asked whether they had experienced a weight change during the current episode of depression. If so, they indicated the total number of pounds gained or lost and the number of
277 weeks over which the weight change occurred. Weekly weight change during the depressive episode (pounds of weight change divided by the number of weeks of depression) was computed for each patient. Patients who reported weight gain, regardless of total pounds gained or the number of weeks during which the gain occurred, were classified as weight gainers. Those who reported weight loss during the current episode of depression were classified as weight losers. The total score on each of the Eating Questionnaire factors was computed for each subject: cognitive restraint (range: O-21), disinhibition (range: O-15), and hunger (range: O-15). Betweengroup comparisons were conducted with x2 analyses, Student’s t tests, analyses of variance, and Newman-Keuls multiple comparisons. Relationships among variables were assessed with Pearson’s product-moment correlations, partial correlations, and linear regression analyses. The relative contribution of demographic, diagnostic, symptom severity, and Eating Questionnaire variables to weight change during the current episode was examined with stepwise discriminant analyses. Results Fig. 1 shows the distribution of total weight change for all patients (n = 109). Total weight change ranged from -33 to +50 pounds. For weight gainers (n =43), the average amount of weight gained was 17.3 pounds (SD 11.8). For weight losers (n = 33), the average loss was 11.3 pounds (SD = 7.9). Duration of the current episode of depression for the total sample ranged from 1 to 192 weeks, with a mean of 20.1 weeks (SD = 3 1.9). The mean duration was far greater for weight gainers (29.8 weeks; SD = 3 1.9) than for weight losers (7.5 weeks; SD 9.3). q
Fig. 1. Distribution
of total weight change for all subjects
r-l
0
-35 -30 -2, -20 -m 30 -5
cl
,
90 4, 7.0 2,
30 51 .o 4, 50
Weiqht Change (pounds)
Weight change expressed as an average weekly value (total pounds of weight change divided by weeks of depression) is presented in Fig. 2. The number of pounds of weight lost per week (mean = 2.1; SD = 1.0) was significantly greater than the number of
278 pounds gained (mean = 1.2; SD = 1.1) suggesting than weight gain (t 3.7; df 74; p < 0.001). Table 1 presents the proportion of patients in types of weight change: loss, no change, and gain. among the proportions of each diagnostic group (~2 = 2.87; df 2; p = 0.24). q
that weight loss occurs more rapidly
q
each diagnostic group with the three There were no significant differences in the three weight change categories
q
Fig. 2. Distribution
of weekly weight change for all subjects
35
I
30
25 * g
20
t? 6 2 E 1
15
10
1l-i
1
5
0 Weekly Weight Change (pounds)
Table 1. Numbers and proportion type
of each diagnostic
group by weight change
Weight change type No change
Loss Depressive subgroups Unipolar-endogenous
(n = 27)
Unipolar-nonendogenous Bipolar
I or II (n = 16)
(n = 66)
n
%
n
Oh
. Gain n
Oh 44.5
10
37.0
5
18.5
12
20
30.4
24
35.8
22
32.8
3
18.8
4
25.0
9
56.2
Table 2 shows the average weight gain or loss for each diagnostic group. A two-way (diagnosis x weight change) analysis of variance, with absolute weight change as the dependent variable, yielded a nonsignificant main effect of diagnosis (F = 2.8 1; df = 2, 70;~ = 0.07). Since the proportions of patients in each weight change category and the amount of reported weight change were not significantly different among diagnostic groups, subsequent analyses were done with all diagnostic groups combined.
279 Table 2. Total weight change (pounds) within each diagnostic group by weight change type Weight change Loss Depressive subgroups
n
Mean
Gain SD
Mean
n
SD
Unipolar-endogenous
10
-12.3
9.4
12
16.3
11.0
Unipolar-nonendogenous
20
- 9.4
6.0
22
15.8
11.4
3
-20.7
10.1
9
22.1
13.6
Bipolar
I or II
The correlations among variables for the sample of 109 patients are illustrated in Table 3. Note that the disinhibition factor of the Eating Questionnaire correlated 0.35 (p < 0.001) with weight change, while the hunger factor correlated 0.28 (p < 0.01). Neither the HRSD nor the BDI scores correlated with weight change. Table 3. Pearson correlations Restraint Restraint
among variables for the total sample (n = 109)
Disinhibition
Hunger
Age
Sex
Disinhibition
0.32
1.0
Hunger
0.241
0.713
Age
0.16
Sex
0.373
1 .o
-0.02
0.02
1 .o
0.03
0.05
0.18
1 .o
HRSD
-0.14
0.09
-0.02
-0.01
0.08
BDI
-0.03
0.221
0.05
-0.11
-0.01
0.353
0.282
0.02
0.03
Weight change (pounds)
HRSD BDI Change
1.0
0.241
1.0 0.473 -0.09
1.0 0.07
1 .o
HRSD = Hamilton Rating Scale for Depression BDI = Beck Depression Inventory. 1. p -= 0.05. 2. p < 0.01. 3. p < 0.001.
Factors Related to Direction of Weight Change. There was no difference in the proportion of men and women that gained or lost weight (~2 = 0.70, df = 1,p = 0.40). Table 4 presents group means and standard deviations for age, the three Eating Questionnaire factors, and HRSD and BDI scores for each type of weight change: loss, no change, and gain. HRSD and BDI scores were computed without appetite and weight loss items. Scores on the three Eating Questionnaire scales were significantly higher in the weight gain group than in the other two groups (p < 0.05). Hamilton scores for the weight loss group were significantly higher than scores for the other groups @ < 0.05), but there were no differences between the groups in BDI score or in age. The combination of variables that best discriminated weight gain from weight loss was examined with stepwise discriminant analysis. Prior probabilities were set equal to the proportion of patients in each group. Eating Questionnaire disinhibition (F= 27.2; df = 1, 74;~ < 0.001) and HRSD scores (Fz4.58; df= 1,73;p < 0.05) yielded a maximum Wilks’ lambda of 0.69 (F = 16.6, df = 2, 73; p < 0.01). Weight-gaining
280 patients had higher disinhibition and lower HRSD scores than weight-losing patients. The discriminant function correctly classified 76.7% (33/43) of the weight gainers and 69.7% (23/33) of the weight losers, using a jackknifed classification procedure. To examine the reliability of this finding, the sample was randomly divided into two subgroups, and a stepwise discriminant analysis was performed on each subgroup. In the analysis of the first subgroup, disinhibition was the only variable to enter the discriminant function, resulting in a maximum Wilks’ lambda of 0.76 (F 11.7; d’= 1,37; p < 0.05). In the analysis of the second subgroup, three variables entered the discriminant function: disinhibition, restraint, and the HRSD score. The final maximum Wilks lambda was 0.48 (F 12. I; df = 3, 34; p < 0.01). This partial replication, in relatively small subsamples, suggests that the influence of disinhibition is stable. q
q
Table 4. Summary statistics on selected variables for each weight change type Weight change Loss (n = 33) Variables
Age
Mean
39.5
No change (n = 33)
Gain (n = 43)
SD
Mean
SD
Mean
SD
12.1
36.4
9.4
37.6
12.3
P 0.6
P NS
Restraint
7.7
4.7
8.2
5.8
11 .o
5.2
4.5
0.01
Disinhibition
5.3
3.7
6.4
4.6
9.9
3.8
13.8
5.1
3.4
5.5
2.9
8.1
4.0
8.9
HRSD
Hunger
19.5
5.1
16.6
4.3
17.1
4.6
3.7
0.03
BDI
26.0
9.5
25.4
9.2
26.3
8.6
0.1
NS
HRSD = Hamilton Rating Scale for Depression. BDI = Beck Depression Inventory. 1. df = 2. 106.
Discussion This study helps to provide an answer to the longstanding puzzle as to why some persons gain weight during a depressive episode, while others lose weight. The results confirm earlier reports that psychological factors may determine the direction of weight changes during depression. Thus, the disinhibition factor to the Eating Questionnaire, which measures the tendency toward periodic loss of control of eating behavior, was highly correlated with weight change (rx0.35;~
281 eating behavior-restraint, disinhibition, and hunger. By contrast, the Restraint Scale, which was used in the two earlier studies, purported to measure only restraint. In fact, it probably did not measure restraint. Scores on the original Restraint Scale correlated with the restraint subscale of the Eating Questionnaire in an earlier study at a statistically insignificant level of 0.168. Correlation of the Restraint Scale with the disinhibition subscale was a highly significant 0.840 (Stunkard and Messick, unpublished data). It appears that the earlier studies, too, found disinhibition, and not restraint, to be related to weight gain during depression. These observational studies are supported by an experimental study by Frost et al. (1982), who assessed the influence of experimentally induced depressive mood on the eating behavior in the laboratory of normal weight subjects who had received the Restraint Scale. They, too, found that disinhibition or the so-called “weight lability” factor and not cognitive restraint, per se, was associated with increased food intake during a depressive mood. Disinhibition is thus related not only to weight changes during depression, but also to eating behavior in normal persons who are made sad. It is apparently triggered by sensitivity to mood changes across a broad spectrum-from mild, acute, and experimental to severe, chronic, and endogenously induced. The finding of the relationship between the HRSD score and weight change may be interpreted as confirming the clinical impression that more severe depression is associated with weight loss. It appears to extend Paykel’s (1977) finding of increased severity of depression among patients with decreases in appetite. However, the finding was not a robust one. It reached only a low level of statistical significance; severity related only to direction and not to extent of weight loss; and the relationship was found only when severity was measured by the HRSD and not by the BDI. It is possible that the specific items on the depression scales influenced the results. The BDI includes more items measuring cognitive symptoms and mood, whereas the HRSD is heavily weighted toward vegetative symptoms such as sleep, libido, and somatic concerns. Multivariate analyses have shown that these latter symptoms are more highly correlated with weight loss than are cognitive symptoms or mood (Rhoades and Overall, 1983). Viewed from this perspective, the present results suggest that depressed patients who lose weight have more vegetative symptoms, but are not necessarily more depressed. The most important aspect of the findings of the depression inventories may be to highlight the greater predictive value of the disinhibition factor in explaining weight changes in depression. A word on the use of self-report weights is in order. Clearly, measured weights are to be preferred. Nevertheless, a growing body of research indicates that self-reported weights are sufficiently valid to be used in a variety of research endeavors (Charney et al., 1976; Pirie et al., 1981; Schlichting et al., 1981; Stunkard and Albaum, 1981). In the absence of systematic bias, any errors in reporting would have the effect of obscuring positive findings and thus would be in a conservative direction. In summary, present findings confirm that eating behavior, particularly disinhibition, significantly relates to the reported direction of weight change during clinical depression. Independent replication of these findings is needed, as are more refined analyses of weight change within depressive subgroups, within groups by type of weight change, and longitudinally throughout the course of depression and during clinical recovery.
282
Acknowledgments. The research reported was supported in part by a grant from the National Institute of Mental Health (MH-28459) to A. John Rush, M.D. The authors express their appreciation to Randy Askins, Jo Ann Brown, B.S.N., Carol J. Fairchild, M.S.N., and Nancy Lee, Ph.D., for assistance in data collection and compilation; to David Savage for secretarial support; and to Kenneth Z. Altshuler, M.D., Professor and Chairman, for his administrative support.
References Psychiatric Association. DSM-II: Diagnostic and Statistical Manual of Mental 2nd ed. APA, Washington, DC (1968). American Psychiatric Association. DSM-III: Diagnostic and Statistical Manual of Mental Disorders. 3rd ed. APA, Washington, DC (1980). Beck, A.T., Ward, C.H., Mendelson, M., Mock, J.E., and Erbaugh, J.K. An inventory for measuring depression. Archives of General Psychiatry, 4, 561 (1961). Bruch, H. Eating Disorders: Obesity, Anorexia and the Person Within. Basic Books, New York (1974). Charney, E., Goodman, H.C., McBride, M., Lyon, B., and Pratt, R. Childhood antecedents of adult obesity: Do chubby infants become obese adults? New England Journal of Medicine,
American
Disorders.
295, 6 (1976).
Davidson, Archives
J.R.T.,
of General
Miller,
R.D.,
Psychiatry,
Turnbull,
39,527
CD.,
and Sullivan,
J.L. Atypical
depression.
(1982).
Feighner, J.P., Robins, E., Guze, S.P., Woodruff, R.A., Winokur, G., and Munoz, R. Diagnostic criteria for use in psychiatric research. Archives of General Psychiatry, 26,57 (1972). Frost, R.O., Goolkasian, G.A., Ely, R.J., and Blanchard, F.A. Depression, restraint, and eating behavior. Behavior Research and Therapy, 20, 113 (1982). Hamilton, M. A rating scale for depression. Journal of Neurology, Neurosurgery and Psychiatry,
23, 56 (1960).
Herman, C.P., and Polivy, J. Restrained eating. In: Stunkard, A.J., ed. Obesity. W.B. Saunders, Philadelphia (1980). Kendell, R.E., and Gourlay, J. The clinical distinction between psychotic and neurotic depression. British Journal of Psychiatry, 117,257 (1970). Kraepelin, E. Lectures on Clinical Psychiatry. Wood, New York (1904). Liebowitz, M.R., Quitkin, F.M., Stewart, J.W., McGrath, P.J., Harrison, W., Rabkin, J., Tricamo, E., Markowitz, J.S., and Klein, D.F. Phenelzine versus imipramine in atypical depression: A preliminary report. Archives of General Psychiatry, 41, 669 (1984). Nelson, J.C., and Charney, D.S. The symptoms of major depressive illness. American Journal
of Psychiatry,
138,
I (1981).
Paykel, E.S. Depression and appetite. Journal of Psychosomatic Research, 21,401 (1977). Pirie, P., Jacobs, D., Jeffrey, R.W., and Hannan, P. Distortion in self-reported height and weight data. Journal of American Dietetic Association, 78, 601 (198 1). Polivy, J., and Herman, C.P. Clinical depression and weight change: A complex relation. Journal
of Abnormal
Psychology,
85, 338 (1976).
J.D. Suggestions for a physiological classification of depression. British Journal of Psychiatry, 111,489(1965). Rhoades, H.M., and Overall, J.E. The Hamilton Depression Scale: Factor scoring and profile classification. Psychopharmacology Bulletin, 19,9 1 (1983). Rosenthal, S.H., and Klerman, G.L. Content and consistency in the endogenous depressive pattern. British Journal of Psychiatry, 112, 47 1 (1966). Schlichting, P., Hoilund-Carlsen, P.F., and Quaade, F. Comparison of self-reported height and weight with controlled height and weight in women and men. International Journal of Pollitt,
Obesity,
5,6
(198 1).
Spitzer, R.L., and Endicott, J. Schedule for Affective Disorders and SchizophreniaL$erime Version. 3rd ed. New York State Psychiatric Institute, 722 W. 168th St., New York, NY 10032 (1978).
283 Spitzer, R.L., Endicott, J., and Robins, E. Research Diagnostic Criteria: Rationale and reliability. Archives of General Psychiatry, 36, 773 (1978). Stunkard, A.J. The dieting depression: Incidence and clinical characteristics of untoward responses to weight reduction regimens. American Journal of Medicine, 23, 77 (1957). Stunkard, A.J., and Albaum, J. The accuracy of self-reported weights. American Journal of Clinical Nutrition, 34, 1593 (198 1). Stunkard, A.J., and Messick, S. The Three-Factor Eating Questionnaire to measure dietary restraint, disinhibition and hunger. Journal of Psychosomatic Research, 29,71 (1985). Stunkard, A.J., and Rush, A.J. Dieting and depression reexamined: Critical review ofreports of untoward responses during weight reduction for obesity. Annals of Internal Medicine, 81, 526 (1974). Zielinski, J.J. Depressive symptomatology: Deviation from a personal norm. Journal of Community Psychology, 6, 163 (1978). Zung, W.W.K., Coppedge, H.M., and Green, R.L. The evaluation of depressive symptomatology: A triadic approach. Psychotherapy and Psychosomatics, 24, 171 (1974).