Predictors Dan A. Oren, Frederick
of Response to Phototherapy Affective Disorder M. Jacobsen,
Thomas
A. Wehr, Christine
in Seasonal
L. Cameron,
and Norman
E. Rosenthal
We examined data from 44 women with seasonal affective disorder (SAD) to determine whether any demographic, diagnostic, or symptomatic characteristics would be predictive of a favorable response to phototherapy. Preexistent hypersomnia was particularly associated with lessening of depression after
phototherapy. In contrast to a report elsewhere, both “typical” and “atypical” depressive symptoms correlated with improvement after phototherapy. This is a US government work. There are no restrictions on its use.
S
more than one study, we used data only from the first study in which they took part. We performed an analysis of variance to assure that the five treatment paradigms used had similar degrees of efficacy. We recruited patients through referrals from therapists familiarwith our program or by a newspaper notice. Criteria for entry to this study were (1) diagnosis of winter seasonal affective disorder’; (2) female gender (to obtain a more homogeneous sample); (3) willingness to comply with the requirements of the experiments; and (4) HAM-D” score of at least 13. All patients provided informed consent after the protocols were fully explained to them. Light treatment was administered in both outpatient and inpatient settings using a box containing fluorescent lamps producing 2.3 x lOI photonslsicm’of full-spectrum white’ or green light.” The number of quanta delivered was the same as that commonly used in studies of broad-spectrum phototherapy at 2,500 lux. Patients were instructed to sit in front of the light fixture and glance briefly at the light once or twice per minute while engaged in reading or desk work. This treatment was administered for 1 week for 2 to 6 hours per day. depending on the flexibility of the individual’s schedule and the experimental design of the study in progress. To predict responses to phototherapy. we considered pretreatment individual HAM-D and ATY item scores, as well as age of onset of illness, age of participation in study, and number of years since onset of first winter depressive episode as independent variables (Table l), and change in HAM-D and ATY scores as dependent variables in a stepwise multiple logistic regression using the SAS package (SAS Institute. Cary, NC). Significance level for entry to the regression was .O.S;significance level for variables to stay in the regression at each step was .05. Finally. we determined correlations between pretreatment HAM-D, ATY, and
disorder (SAD) is an annual mood disorder characterized by fall/ winter depressions alternating with periods of euthymia or hypomania in the spring and summer.’ Many patients with the diagnosis show “atypical” vegetative symptoms (e.g., fatigue, hypersomnia, increased appetite, and weight gain), variability with respect to severity, and “reverse” diurnal variation with mood worsening in the late afternoon. Although depressed SAD patients have a robust antidepressant response to phototherapy,‘” only one published study has analyzed clinical factors that may predict a positive response to phototherapy. In that report, Stinson and Thompson concluded that clinical factors likely to predict a therapeutic response to phototherapy include unipolar compared with bipolar II history, high degree of “atypical” symptoms of depression (including hypersomnia), and a low level of “endogenous” depressive symptoms. In the present report, we have analyzed clinical response data from several phototherapy studies performed by our group over the past 8 years5~’ to determine whether any predictive factors can be identified. easonal
affective
METHOD We examined retrospectively data from the “active” condition of each controlled study comparing an “active” with an “inactive placebo” form of phototherapy for SAD performed at the National Institute of Mental Health (NIMH) between 1981 and 1989. Levels of depression were based on the 21-item Hamilton Depression Scale (HAMD),“’ as well as atypical symptoms (ATY) such as hypersomnia, carbohydrate craving, and weight gain.” In the more recent studies, these measures were obtained by the Structured Interview Guide for the Hamilton Seasonal Affective Disorder version (SIGH-SAD).l’ Depression scores were obtained weekly by trained raters who were blind to experimental conditions. Interrater reliability in each study examined was over 95%. When patients participated in
Comprehensive
HAM-D plus ATY scores. each of those scores.
and posttreatment
change
in
From the Clinical Psychobiology Branch, National Institute of Mental Health, Bethesda, MD. Address reprint requests to Dan A. Oren, M.D., Clinical Psychobiology Branch, National Institute of Mental Health, Bldg. 10, Room 4S239, 9000 Rockr-ille Pike, Bethesda, MD 20892. This is a US government work. There are no restrictions on its use. 0010-440X192/3302-0001$0.00/0
Psychiatry, Vol. 33, No. ‘2(March/April), 1992: pp 11I-l 14
111
112
OREN ET AL
Table 1. Potentially Predictive Factors Examined in Regression Analysis Pretreatment symptoms Depressed mood Work and activity level Social withdrawal Genital symptoms Gastrointestinal somatic symptoms Weight loss Weight gain Appetite increase Increased eating Carbohydrate craving Early-night insomnia Middle-night insomnia Late-night insomnia Hypersomnia Generalized somatic symptoms Fatigue Guilt feelings Suicidal@ Psychic anxiety Somatic anxiety Hypochondriasis
scores on individual depression subscale items with improvement in total depression scores. Hypersomnia consistently predicted improvement in mood scores following phototherapy. Table 3 shows the correlations between pretreatment severity scores on the 21-item HAM-D and ATY and total scores, and the improvement in score after light treatment. The magnitude of improvement in both subscales correlated with severity before treatment. High scores on the ATY scale also correlated positively with improvement in total score. Reexamination of these data for the group as a whole using 17-item HAM-D scores” did not change the significance of these correlations. Reexamination of these data using percentage improvement in scores as the dependent measure failed to sustain a correlation with pretreatment HAM-D scores, but did sustain a correlation between the ATY scores.
Insight into illness Psychomotor retardation
DISCUSSION
Agitation “Reverse” diurnal variation Severity of diurnal variation Depersonalization or derealization Paranoia Obsessional-compulsive
traits
Age of onset of illness Age of participation in study Number of years since onset of first winter depressive episode
To rule out possible confounding factors of patients who were treated with green light and inclusion of postmenopausal subjects, we repeated the statistical analysis for subgroups of the original population that did not include these subjects.
RESULTS
Data from 44 female patients from five studies were analyzed. There was no significant difference in efficacy of active treatments between the five different studies (F = 0.77, df = 4, P > .05), with the mean improvement of the typical HAM-D scores being 55%. Fifteen patients met Research Diagnostic Criteria for unipolar depression, 27 patients met criteria for bipolar II depression, and two patients met criteria for bipolar I depression.13 Table 2 lists the regression formulas linking
Pooling of patients from studies in which we administered different light therapy regimens creates a potential problem in the interpretation of the results and may account for the small number of predictors found. Nevertheless, the ability to pool these data is supported by our observing no difference in efficacy between the different regimens that were pooled. The study suggested that in depressed SAD patients an antidepressant response to phototherapy was particularly associated with hypersomnia. This may provide a rationale for the method used in some studies of SAD of obtaining a more homogeneous sample by entering only hypersomnic patients into investigative protocols. The greatest correlation between subscale scores that we found was between pretreatment ATY score and posttreatment improvement in ATY score, supporting the finding of Stinson and Thompson4 and supporting the value of measuring ATY scores in winter SAD. In contrast, our correlation of higher pretreatment HAM-D scores with greater magnitudes of improvement on that scale differs from that of Stinson and Thompson, who indicated that greater severity of typical depressive symptoms
PREDICTORS OF RESPONSE TO LIGHT IN SAD
113
Table 2. Multiple Regression Formulas Predicting Improvement
in Depression Scores After Phototherapy as Predicted by
Individual Subscale Items Group as a whole (N = 44) Improvement in
= 6.3+
2.7 Hypersomnia+
3.1 Suicidalityt
2.7 Carbohydrate
P = ,002
i- = .38
r = .4J
r=
= 9.0+
2.9 Suicidality
P = ,007
r = .40
= -3.7+
1.3 Appetite in-
1.7 Weight gain+
1.5 Social with-
HAM-D + ATY df = 3.40, F =
craving .56
6.08 Improvement in HAM-D df=1,42,F= 7.94 Improvement in ATY df = 5‘38, F =
crease+
drawal+ r = .62
P < .OOOl r = .48
1.2 Carbohydrate
1.l Hypersomnia
craving+
r = .JO
r = .74
1.1 Carbohydrate
1.2 Social with-
r = .J8
11.77 White light treatment only (N = 35) Improvement in
= 21-
0.42 Age of onset+
3.4 Hypersomnia
P = .003
r = .31
r=
= 14p
0.42 Age of onset+
2.8 Psychic anxiety
P = ,001
r = .23
r = .35
= -2.o+
1.9 Appetite in-
2.0 Weight gain+
HAM-D f ATY df = 2.32, F =
.49
7.14 Improvement in HAM-D df = 2,32, F = 8.68 improvement in ATY df = 4.30, F =
crease+
craving+
P < .OOOl r = .26
r = .46
r = .52
drawal+ I = .58
10.5 White light treatment only, age less than 46 (N = 27) Improvement in
= 9.5-
4.6 Hypersomnia
P = .007
r = .51
= 4.0+
2.3 Hypersomnia+
3.1 Suicidality
P= .Ol
r = .43
r=
=
1.7 Weight gain+
1.8 Appetite in-
HAM-D + ATY df=
1,25,F=
8.67 Improvement in HAM-D; df = 2.32. F =
.56
5.46 Improvement in ATY df = 3.23, F =
1.J Hypersomnia
crease+ P = .0005
r = .52
r = .65
r = .J3
a.59 NOTE. The number of points of improvement
on the HAM-D and ATY depression scales can be predicted by multiplying the
appropriate individual item scores by the cofactors listed above and adding them as shown in the above equations.
predicted a worse response to treatment. Stinson and Thompson based their conclusion on percentage improvement as the outcome measure, while we used absolute scores as the outcome measure, perhaps accounting for this discrepancy. Indeed, when we used percentage improvement in our analysis, we failed to identify a significant correlation that was present when absolute scores were used. This variation may be explained by patients with low pretreatment scores, such as two subjects whose pretreatment scores were 13 and posttreatment scores
were 0. Patients with such low pretreatment scores and essentially complete improvement will fall along a regression curve when evaluated by absolute means, but not when percentage improvements are examined. Presentation of SAD studies only in terms of percentage of depression score improvement may therefore miss important features of the responders to phototherapy. Although “atypical” depressive features played the largest role in predicting response to phototherapy, the “endogenous” depressive fea-
114
OREN ET AL
Table 3. Correlations of Pretreatment
Severity Scores on the
HAM-D and ATY Subscales and Total Score With Magnitude of Response to Light Treatment Pretreatment AN
HAM-D
Total
Group as a whole (N = 44) Magnitude of improvement ATY
0.78’
HAM-D
0.17
-0.16 0.40t
0.24 0.44*
Total
0.510
0.22
0.43t
White light treatment only (N = 35) Magnitude of improvement ATY
0.761
HAM-D
0.21
-0.17 0.3711
0.44**
0.22
Total
0.52tt
0.18
0.41**
White light treatment only, age ~46 (N = 27) Magnitude of improvement ATY
0.7211
HAM-D
0.12
-0.23 0.3881
0.14 0.41(1/
Total
0.44***
0.15
0.36
Significance for group as a whole: *P = .OOOl; tP = ,007; s/J = ,003; §P = .0004. Significance for white tight treatment only: llf = .OOOl: ilp = .03; l*P = ,009; ttP = .002; SSP = .Ol. Significance for white tight treatment only, age
of Response to Phototherapy Affective Disorder M. Jacobsen,
Thomas
A. Wehr, Christine
in Seasonal
L. Cameron,
and Norman
E. Rosenthal
We examined data from 44 women with seasonal affective disorder (SAD) to determine whether any demographic, diagnostic, or symptomatic characteristics would be predictive of a favorable response to phototherapy. Preexistent hypersomnia was particularly associated with lessening of depression after
phototherapy. In contrast to a report elsewhere, both “typical” and “atypical” depressive symptoms correlated with improvement after phototherapy. This is a US government work. There are no restrictions on its use.
S
more than one study, we used data only from the first study in which they took part. We performed an analysis of variance to assure that the five treatment paradigms used had similar degrees of efficacy. We recruited patients through referrals from therapists familiarwith our program or by a newspaper notice. Criteria for entry to this study were (1) diagnosis of winter seasonal affective disorder’; (2) female gender (to obtain a more homogeneous sample); (3) willingness to comply with the requirements of the experiments; and (4) HAM-D” score of at least 13. All patients provided informed consent after the protocols were fully explained to them. Light treatment was administered in both outpatient and inpatient settings using a box containing fluorescent lamps producing 2.3 x lOI photonslsicm’of full-spectrum white’ or green light.” The number of quanta delivered was the same as that commonly used in studies of broad-spectrum phototherapy at 2,500 lux. Patients were instructed to sit in front of the light fixture and glance briefly at the light once or twice per minute while engaged in reading or desk work. This treatment was administered for 1 week for 2 to 6 hours per day. depending on the flexibility of the individual’s schedule and the experimental design of the study in progress. To predict responses to phototherapy. we considered pretreatment individual HAM-D and ATY item scores, as well as age of onset of illness, age of participation in study, and number of years since onset of first winter depressive episode as independent variables (Table l), and change in HAM-D and ATY scores as dependent variables in a stepwise multiple logistic regression using the SAS package (SAS Institute. Cary, NC). Significance level for entry to the regression was .O.S;significance level for variables to stay in the regression at each step was .05. Finally. we determined correlations between pretreatment HAM-D, ATY, and
disorder (SAD) is an annual mood disorder characterized by fall/ winter depressions alternating with periods of euthymia or hypomania in the spring and summer.’ Many patients with the diagnosis show “atypical” vegetative symptoms (e.g., fatigue, hypersomnia, increased appetite, and weight gain), variability with respect to severity, and “reverse” diurnal variation with mood worsening in the late afternoon. Although depressed SAD patients have a robust antidepressant response to phototherapy,‘” only one published study has analyzed clinical factors that may predict a positive response to phototherapy. In that report, Stinson and Thompson concluded that clinical factors likely to predict a therapeutic response to phototherapy include unipolar compared with bipolar II history, high degree of “atypical” symptoms of depression (including hypersomnia), and a low level of “endogenous” depressive symptoms. In the present report, we have analyzed clinical response data from several phototherapy studies performed by our group over the past 8 years5~’ to determine whether any predictive factors can be identified. easonal
affective
METHOD We examined retrospectively data from the “active” condition of each controlled study comparing an “active” with an “inactive placebo” form of phototherapy for SAD performed at the National Institute of Mental Health (NIMH) between 1981 and 1989. Levels of depression were based on the 21-item Hamilton Depression Scale (HAMD),“’ as well as atypical symptoms (ATY) such as hypersomnia, carbohydrate craving, and weight gain.” In the more recent studies, these measures were obtained by the Structured Interview Guide for the Hamilton Seasonal Affective Disorder version (SIGH-SAD).l’ Depression scores were obtained weekly by trained raters who were blind to experimental conditions. Interrater reliability in each study examined was over 95%. When patients participated in
Comprehensive
HAM-D plus ATY scores. each of those scores.
and posttreatment
change
in
From the Clinical Psychobiology Branch, National Institute of Mental Health, Bethesda, MD. Address reprint requests to Dan A. Oren, M.D., Clinical Psychobiology Branch, National Institute of Mental Health, Bldg. 10, Room 4S239, 9000 Rockr-ille Pike, Bethesda, MD 20892. This is a US government work. There are no restrictions on its use. 0010-440X192/3302-0001$0.00/0
Psychiatry, Vol. 33, No. ‘2(March/April), 1992: pp 11I-l 14
111
112
OREN ET AL
Table 1. Potentially Predictive Factors Examined in Regression Analysis Pretreatment symptoms Depressed mood Work and activity level Social withdrawal Genital symptoms Gastrointestinal somatic symptoms Weight loss Weight gain Appetite increase Increased eating Carbohydrate craving Early-night insomnia Middle-night insomnia Late-night insomnia Hypersomnia Generalized somatic symptoms Fatigue Guilt feelings Suicidal@ Psychic anxiety Somatic anxiety Hypochondriasis
scores on individual depression subscale items with improvement in total depression scores. Hypersomnia consistently predicted improvement in mood scores following phototherapy. Table 3 shows the correlations between pretreatment severity scores on the 21-item HAM-D and ATY and total scores, and the improvement in score after light treatment. The magnitude of improvement in both subscales correlated with severity before treatment. High scores on the ATY scale also correlated positively with improvement in total score. Reexamination of these data for the group as a whole using 17-item HAM-D scores” did not change the significance of these correlations. Reexamination of these data using percentage improvement in scores as the dependent measure failed to sustain a correlation with pretreatment HAM-D scores, but did sustain a correlation between the ATY scores.
Insight into illness Psychomotor retardation
DISCUSSION
Agitation “Reverse” diurnal variation Severity of diurnal variation Depersonalization or derealization Paranoia Obsessional-compulsive
traits
Age of onset of illness Age of participation in study Number of years since onset of first winter depressive episode
To rule out possible confounding factors of patients who were treated with green light and inclusion of postmenopausal subjects, we repeated the statistical analysis for subgroups of the original population that did not include these subjects.
RESULTS
Data from 44 female patients from five studies were analyzed. There was no significant difference in efficacy of active treatments between the five different studies (F = 0.77, df = 4, P > .05), with the mean improvement of the typical HAM-D scores being 55%. Fifteen patients met Research Diagnostic Criteria for unipolar depression, 27 patients met criteria for bipolar II depression, and two patients met criteria for bipolar I depression.13 Table 2 lists the regression formulas linking
Pooling of patients from studies in which we administered different light therapy regimens creates a potential problem in the interpretation of the results and may account for the small number of predictors found. Nevertheless, the ability to pool these data is supported by our observing no difference in efficacy between the different regimens that were pooled. The study suggested that in depressed SAD patients an antidepressant response to phototherapy was particularly associated with hypersomnia. This may provide a rationale for the method used in some studies of SAD of obtaining a more homogeneous sample by entering only hypersomnic patients into investigative protocols. The greatest correlation between subscale scores that we found was between pretreatment ATY score and posttreatment improvement in ATY score, supporting the finding of Stinson and Thompson4 and supporting the value of measuring ATY scores in winter SAD. In contrast, our correlation of higher pretreatment HAM-D scores with greater magnitudes of improvement on that scale differs from that of Stinson and Thompson, who indicated that greater severity of typical depressive symptoms
PREDICTORS OF RESPONSE TO LIGHT IN SAD
113
Table 2. Multiple Regression Formulas Predicting Improvement
in Depression Scores After Phototherapy as Predicted by
Individual Subscale Items Group as a whole (N = 44) Improvement in
= 6.3+
2.7 Hypersomnia+
3.1 Suicidalityt
2.7 Carbohydrate
P = ,002
i- = .38
r = .4J
r=
= 9.0+
2.9 Suicidality
P = ,007
r = .40
= -3.7+
1.3 Appetite in-
1.7 Weight gain+
1.5 Social with-
HAM-D + ATY df = 3.40, F =
craving .56
6.08 Improvement in HAM-D df=1,42,F= 7.94 Improvement in ATY df = 5‘38, F =
crease+
drawal+ r = .62
P < .OOOl r = .48
1.2 Carbohydrate
1.l Hypersomnia
craving+
r = .JO
r = .74
1.1 Carbohydrate
1.2 Social with-
r = .J8
11.77 White light treatment only (N = 35) Improvement in
= 21-
0.42 Age of onset+
3.4 Hypersomnia
P = .003
r = .31
r=
= 14p
0.42 Age of onset+
2.8 Psychic anxiety
P = ,001
r = .23
r = .35
= -2.o+
1.9 Appetite in-
2.0 Weight gain+
HAM-D f ATY df = 2.32, F =
.49
7.14 Improvement in HAM-D df = 2,32, F = 8.68 improvement in ATY df = 4.30, F =
crease+
craving+
P < .OOOl r = .26
r = .46
r = .52
drawal+ I = .58
10.5 White light treatment only, age less than 46 (N = 27) Improvement in
= 9.5-
4.6 Hypersomnia
P = .007
r = .51
= 4.0+
2.3 Hypersomnia+
3.1 Suicidality
P= .Ol
r = .43
r=
=
1.7 Weight gain+
1.8 Appetite in-
HAM-D + ATY df=
1,25,F=
8.67 Improvement in HAM-D; df = 2.32. F =
.56
5.46 Improvement in ATY df = 3.23, F =
1.J Hypersomnia
crease+ P = .0005
r = .52
r = .65
r = .J3
a.59 NOTE. The number of points of improvement
on the HAM-D and ATY depression scales can be predicted by multiplying the
appropriate individual item scores by the cofactors listed above and adding them as shown in the above equations.
predicted a worse response to treatment. Stinson and Thompson based their conclusion on percentage improvement as the outcome measure, while we used absolute scores as the outcome measure, perhaps accounting for this discrepancy. Indeed, when we used percentage improvement in our analysis, we failed to identify a significant correlation that was present when absolute scores were used. This variation may be explained by patients with low pretreatment scores, such as two subjects whose pretreatment scores were 13 and posttreatment scores
were 0. Patients with such low pretreatment scores and essentially complete improvement will fall along a regression curve when evaluated by absolute means, but not when percentage improvements are examined. Presentation of SAD studies only in terms of percentage of depression score improvement may therefore miss important features of the responders to phototherapy. Although “atypical” depressive features played the largest role in predicting response to phototherapy, the “endogenous” depressive fea-
114
OREN ET AL
Table 3. Correlations of Pretreatment
Severity Scores on the
HAM-D and ATY Subscales and Total Score With Magnitude of Response to Light Treatment Pretreatment AN
HAM-D
Total
Group as a whole (N = 44) Magnitude of improvement ATY
0.78’
HAM-D
0.17
-0.16 0.40t
0.24 0.44*
Total
0.510
0.22
0.43t
White light treatment only (N = 35) Magnitude of improvement ATY
0.761
HAM-D
0.21
-0.17 0.3711
0.44**
0.22
Total
0.52tt
0.18
0.41**
White light treatment only, age ~46 (N = 27) Magnitude of improvement ATY
0.7211
HAM-D
0.12
-0.23 0.3881
0.14 0.41(1/
Total
0.44***
0.15
0.36
Significance for group as a whole: *P = .OOOl; tP = ,007; s/J = ,003; §P = .0004. Significance for white tight treatment only: llf = .OOOl: ilp = .03; l*P = ,009; ttP = .002; SSP = .Ol. Significance for white tight treatment only, age
