German Journal of Psychiatry

ISSN 1433-1055


Light Therapy of Affective Disorders with and without Seasonality: Biological Markers

Holger Hill, Christiane Batt, Sabine Maier, Matthias Weisbrod, and Hermes Kick

From the Department of Psychiatry, University of Heidelberg. Corresponding author: Dr. Holger Hill, Psychiatrische Universitätsklinik, Voss-Str. 4, D-69115 Heidelberg, Germany, tel. ++49-6221-565469, fax: ++49-6221-565998, e-mail: Holger_Hill@ukl.uni-heidelberg.de


Clinical symptoms and response to light therapy treatment are important factors for the definition of seasonal affective disorder (SAD) as a subgroup of major depression. As part of a study on 39 depressive outpatients who underwent a two-week course of light therapy, we attempted to strengthen the definition of SAD with biological markers such as auditory evoked potentials (AEP).

According to the changes in their Hamilton depression scores over the course of therapy, patients were classified into three response groups: (1) The nonresponder group with no or marginal improvement, (2) the intermediate group with only a temporary improvement and (3) the responder group with an overall improvement. Clinically, patients with SAD symptoms showed a significantly better response to light therapy. Electrophysiologically, the responder group showed a smaller P300 amplitude before and during treatment compared with both other groups, as did the SAD compared with the non-SAD patients. Additionally, patients under antidepressant medication had a reduced P300 amplitude in contrast to the drug-free patients.

Our conclusion is that the amplitude of P300 as a biological trait marker supports the classification of SAD as a subgroup of depression and suggests that the pathophysiological basis varies among the subtypes of affective disorder.

(German J Psychiatry 1998;1:41-52)

Key Words: light therapy, phototherapy, seasonal affective disorder, depression, P300, event-related potentials


As first described by Rosenthal et al. (1984), 'the diagnosis of seasonal affective disorder (SAD) - a recurrent major depression in autumn and winter remitting in spring and summer - has been operationalised and incorporated into the Diagnostic and Statistical Manual of Mental Disorders, DSM-III-R (American Psychiatric Association 1987)' (Wirz-Justice 1993). Light therapy as a natural treatment was developed on the basis of circadian rhythm research. The antidepressant effect of this therapy has been well established (e.g. Terman et al. 1989, Tam et al. 1995), despite the extent of a placebo component as discussed by Eastman et al. (1990, 1992). However, hypotheses about the physiological basis of SAD and the different assessments of light therapies (light intensities, duration and time of day of treatment) are the subject of controversial discussion.

SAD can be defined as a subgroup of major depression from the diagnostic and clinical points of view. The finding of biological parameters such as electroencephalographic (EEG) differences would strengthen this definition. P300, a positive event-related potential (ERP) component of the EEG, reflects attentional and decision processes. It peaks around 300 ms after the occurrence of a rare stimulus and is well established in psychiatric research. Several studies have shown a reduced amplitude of the auditory evoked P300 in depressive patients compared with healthy controls, e.g. Roth et al. (1981), Blackwood et al. (1987) and Gangadhar et al. (1993). Pfefferbaum et al. (1984) found reduced amplitudes of auditory and visual evoked P300 only in drug-free patients. Maurer and Dierks (1988) and Sara et al. (1994) observed only nonsignificantly reduced P300 amplitudes in depressive patients compared with controls. As a possible explanation for their results, Sara et al. mention the heterogeneity of groups of depressive patients classified using DSM-III criteria and that P300 abnormalities may be related only to a specific aspect of the depressive syndrome.

This interpretation is supported by studies on subgroups of depressive patients. Pholien et al. (1987) found reduced amplitudes in melancholic compared with nonmelancholic depressive patients (DSM III classification) and intermediate amplitudes in a dysthymic group, although all groups of patients showed reduced amplitudes compared with controls. Santosh et al. (1994) investigated patients with a DSM III-R melancholic depression and found a reduced P300 in those subjects who had additional psychotic features. Hansenne et al. (1994) found reduced P300 amplitudes in depressive patients who had attempted suicide compared with patients who had not attempted suicide.

In SAD patients, Murphy et al. (1993) found no differences in visual evoked potentials (N1, P2, N2, P300) compared with controls. Duncan et al. (1990) described an increase of visual P300 amplitude in SAD patients during the assessment of a light therapy course with a correlation between P300 changes and the improvement in psychopathology. Auditory P300 was not affected by light therapy treatment in that study.

In the present study we measured the auditory P300 before and during a course of light therapy on depressive patients. The finding of group differences or changes during therapy will give more emphasis for the definition of SAD as a subgroup of major depression and suggest a different pathophysiological base of SAD. Additionally, differences in ERP will be of clinical importance as predictive variables for the assessment of different therapies.

Because the therapeutic efficacy of the light intensity used is still a matter of dispute (e.g. Wirz-Justice 1993), we chose a cross-over design with two different light intensities to confirm intensity-dependent responses in light therapy.

Material and Methods

The study included 39 depressive out-patients (30 female, 9 male) who fulfilled the DSM III-R criteria for major depressive disorder (296.32: n = 38; 296.31: n = 1) and had a minimum score of 13 on the 21-item Hamilton Depression Rating Scale (Hamilton 1967) before therapy. Thirty patients showed a seasonal pattern of depression. Twenty-two of them fulfilled all 4 items, 8 fulfilled 2 or 3 items of the DSM III-R criteria for an SAD diagnosis. Nine patients had no seasonal symptoms. The mean age was 48 years (SD 10.6). Twenty-three patients were drug free, 16 patients were treated with antidepressant medication before and during therapy (nine of them with antidepressants only, two with antidepressants and neuroleptics, three with antidepressants and lithium, one with lithium only, one with a tranquilliser only). The last changes in medication were made at least two weeks before the start of light therapy.

Light therapy was performed for two hours daily in the early evening for a period of two weeks between November and February. To test therapeutic efficacy in relation to light intensity, a cross-over design was devised. Patients were divided into two groups: Group 1 received full-spectrum fluorescent light (Theralux ATL-6) with an intensity of 2500 lux for the first week and 300 lux for the second week, group 2 received the inverse light-intensity pattern. The patients sat at a distance of 1 m from the light source (2500 lux treatment) and were instructed to look periodically into it. For the 300 lux treatment, the lights were dimmed and the distance increased to 3 m.

Before therapy (day 1), after week one (day 8) and after week two (day 15), Hamilton depression scores were measured in the morning and evening by the same qualified investigator.

Therapy response was defined as the relative change in the Hamilton depression score during the two-week light therapy course, based on the mean of morning and evening scores. Response was calculated as follows: Response [%] of day d = 100 * (Hamilton day 1 - Hamilton day d) / Hamilton day 1 (where d = 8 or 15 respectively).

The electrophysiological recordings were performed in the late morning. Nineteen gold electrodes were placed according to the international 10-20 system, all referred to linked mastoids. Vertical and horizontal eye movements were recorded with two electrodes supra-/infraorbitally and two electrodes at the outer canthal positions. Impedances were below 5 kOhm. P300 was generated with randomised stimuli: A frequent 1000 Hz tone with an occurrence of 80 % and a 2000 Hz tone (target) with a 20 % occurrence. Duration of the tones was 50 ms including 10 ms rise and fall. The interstimulus interval was 1.5 s, digitation rate was 256 Hz. Data were filtered online with a 0.1 Hz to 70 Hz bandpass. The task was to count the target tone silently. To assure that the subjects performed the task correctly, the number of targets was changed between sessions from 30 to 41.

Data processing: Eye movements and eye blinks were corrected using the regression method of Anderer et al. (1989a). A prestimulus interval of 200 ms was used for baseline correction. Data were controlled manually, only artifact-free trials were included in further analysis. The averaged waveforms were lowpass filtered with 16 Hz. The largest positive deflection in the time range from 250 to 500 ms was defined as the P300 peak. Some data sets showed no clear P300 pattern at peripheral electrode sites. Therefore, only the nine electrodes with the best P300 pattern were included in further analysis (F4, Fz, F3, C4, Cz, C3, P4, Pz, P3). Statistical tests were performed with MANOVAs, for data of nominal scale level the Chi2-test was used.


According to the different variables, we subdivided clinical and ERP results into light-intensity pattern, response group definition, SAD symptoms and antidepressant medication.

Clinical data

To examine the influence of light intensity on therapy response, a MANOVA with the two independent factors light-intensity pattern and SAD symptoms and Hamilton depression score as a dependent factor (repeated measuring for day 1, 8 and 15) was computed. The hypothesised higher efficacy of the 2500 lux light intensity should have shown different Hamilton depression scores at day 8 and probably at day 15 in both treatment groups. However, there were no significant interactions which would confirm our hypothesis:

Light intensity pattern * day: F(2, 70) = 1.85, p = 0.16 (Table 1).

Light intensity pattern * SAD symptoms * day: F(2, 70) = 0.83, p = 0.44.

Table 1: Means and standard deviations of Hamilton depression scores for the two light-intensity pattern groups (first week/second week).

day 1
day 8
day 15
300/2500 lux (n=23)
19.2 (4.3)
11.8 (6.3)
9.5 (6.9)
2500/300 lux (n=16)
19.4 (6.1)
12.1 (7.1)
12.2 (8.6)

Two different definitions of response groups where made post-hoc because no influence of light intensity on therapy response could be seen.

The first response group definition was made due to the fact that there were considerable changes in Hamilton depression scores between day 8 and day 15. To represent the time course of therapy, we defined three groups (Table 2). Patients were defined as (1) nonresponders when their improvement (decrease in Hamilton depression score) was less than 25 % at day 8 and/or day 15 (n = 8). This 25 % cut-off was previously used by Nagayama et al. (1991). The patients with a Hamilton depression score decrease of 25 % or more at day 8 and/or day 15 were defined as (2) responders as long as their score at day 15 was lower than or equal to that at day 8 (n = 21). The intermediate group (3) consisted of those patients whose Hamilton depression scores decreased (25 % or more) from day 1 to day 8 but increased again from day 8 to day 15 (n = 10).

There were no significant differences in day 1 Hamilton depression scores between the three groups and no differences in terms of age, gender or medication (Tables 3 and 4).

Secondly, a 'classical' two-group definition was made according to Terman et al. (1989). The cut-off for grouping patients as responders (n = 16) or nonresponders (n = 23) was a 50 % decrease in Hamilton depression score from day 1 to day 15. When comparing the two group definitions, 15 subjects of the responder and one of the intermediate group were 'Terman-responders'.

Table 2: Group means and standard deviations of Hamilton depression scores.

day 1
day 8
day 15
Responder group (n=21)
18.2 (4.6)
9.5 (5.3)
5.2 (4.3)
Intermediate group (n=10)
19.5 (4.3)
10.2 (4.1)
14.8 (4.7)
Nonresponder group (n=8)
22.2 (6.5)
20.3 (5.5)
19.4 (6.1)

Table 3: Distribution of gender and age among response groups.

Mean age (SD)
Responder group49.1 (10.4) 165
Intermediate group46.9 (10.8) 82

Nonresponder group

46.5 (11.9)



Table 4: Classification of patients with or without psychotropic drugs in the three therapy response groups and in groups with or without SAD symptoms.

Antidepressants + Neuroleptics
Antidepressants + Lithium
Responder group105 230 1
Intermediate group82 000 0
Nonresponder group52 001 0
SAD205 1301
non-SAD34 101 0

The patients with SAD symptoms had lower day 1 Hamilton depression scores than the non-SAD patients (17.9 vs. 23.7; SD 4.5/4.7) and a significantly better response to light therapy: day 1 to day 8: 44.4 % (SD 27.2) vs. 20.8 % (17); day 1 to day 15: 51.2 % (31.1) vs. 29.5 % (35.5); main effect for SAD group: F[1, 37] = 5.3, p = 0.027. Eighteen of 21 patients of the responder group, 8 of 10 patients of the intermediate group and 4 of 8 patients of the nonresponder group showed SAD symptoms.

The groups of patients with different antidepressant medication presented no significant differences in day 1 Hamilton depression scores or in therapy response.

ERP data

All subjects performed the task well. For day 1, ERP data of all 39 patients could be included in this analysis. The day 8 data set of one patient had to be excluded due to artifacts, while an EEG could not be recorded for another patient at day 15. So the complete sample consisted of the data of 37 subjects.

For statistical analysis, MANOVAs were computed with P300 amplitudes (or latencies respectively) as a dependent variable for each electrode site and day of recording (repeated measures factors). The independent variables are specified below. Due to the different sample sizes, MANOVAS were computed additionally for day 1 only for the variables response group, SAD symptoms and medication. For pairwise post-hoc comparisons, Fisher's-LSD test was used.

According to the clinical results, the light intensity pattern (as an independent variable) showed no influence on P300.

The comparison of the three response groups showed lower P300 amplitudes in the responder group in contrast to the intermediate and nonresponder groups. (Table 5). The MANOVA with the two independent variables response group and light-intensity pattern revealed significant main effects for response group: F(2, 31) = 3.8; p = 0.035 and electrode site: F(8, 248) = 37.4, p < 0.001. Post-hoc comparisons showed significantly lower amplitudes in the responder group compared with the nonresponder group (p = 0.01). The comparison between responder and intermediate group failed to reach significance (p = 0.13). Nonresponder and intermediate group did not differ from each other. There was no main effect for the day of recording and no interactions between any of these 4 variables.

A MANOVA for day 1 only, with all 39 subjects included, revealed this main effect for response group: F(2, 36) = 4.31, p = 0.021, where post-hoc comparisons showed lower amplitudes between the responder group and the nonresponder group (p = 0.033) and the intermediate group (p = 0.093), respectively.

The effect of antidepressant medication illustrated below did not affect these results, due to the distribution of subjects with or without medication in the three response groups (Chi-Square = 10.3; p = 0.41; n = 39; cf. Table 4).

To compare these results with the response group definition after Terman et al., a MANOVA was computed which showed nearly significantly lower P300 amplitudes in the better responding patients. Main effect group: F(1, 35) = 3.9, p = 0.056.

Table 5: Group means of P300 amplitudes before light therapy (day 1). R - Responder group, I - Intermediate group, N - Nonresponder group.

Elektrode site
R (n = 21)
I (n = 10)
N (n = 8)




































The comparison between the groups of patients with or without SAD symptoms showed lower P300 amplitudes in the SAD group at frontal and central electrodes (Table 6). A MANOVA with the SAD group as the independent factor yielded only trends for a group * electrode interaction: F(8, 280) = 1.7, p = 0.097, and for a group * electrode * day interaction: F(16, 560) = 1.5, p = 0.095. For day 1 there was also only a trend towards a group * electrode interaction: F(8, 296) = 1.75, p = 0.087. It should be noted that these results were probably masked by the effect of treatment with antidepressant medication shown below (67 % of the SAD patients were drug free but only 42 % of the non-SAD patients, cf. Table 4).

Table 6: Group means of P300 amplitudes before light therapy (day 1). Patients with vs. patients without SAD symptoms.

Electrode site
SAD (n = 30)
non-SAD (n = 9)



























Additionally, this study revealed an influence of treatment with antidepressant medication on P300 amplitude. The drug-free patients had larger amplitudes than the other patients. When dividing medication into several classes (Table 7), we found a main effect with F(5, 33) = 2.66, p = 0.04 (day 1 only). For day 1, 8 and 15 we found a medication group * electrode-site interaction with F(24, 248) = 1.61, p = 0.039. Post-hoc comparisons for both MANOVAs showed significant differences for all electrodes between the drug-free group (n = 23) and the antidepressant-only group (n = 9 for day 1 only, n = 7 for all three recordings).

For P300 latency no statistical difference could be found.

Age and gender of patients showed no influence on P300 amplitude.

Table 7: Means of P300 amplitudes of medication treatment groups before therapy.

Electrode site
Drug-free (n=23)
Antidepressants (n=9)
Antidepressants + Neuroleptics (n=2)
Antidepressants + Lithium (n=3)














































Thirty-nine depressive patients took part in our two-week light therapy study. The 30 patients with SAD symptoms showed a significantly better response to the therapy than the others. The result that some of the SAD patients did not respond to light therapy is in line with other studies (Duncan et al. 1990; review in Murphy et al. 1993). Concerning the applied light intensity pattern, we found no differences in therapy response between 300 or 2500 lux intensities. The light intensity which is necessary for successful therapy is still not yet clear and is a matter of controversy in the literature (reviewed in Wirz-Justice 1993). Perhaps there are methodological aspects to be considered, such as natural light exposure. To determine the light intensities used for a successful therapy it is necessary to measure the actual light intensity reaching the retina during treatment sessions, as mentioned by Wirz-Justice (1993), plus the natural light exposure.

We defined three response groups to represent the course of changes in psychopathology. Because this definition was made post-hoc, we compared it with a classical two-groups design, grouping patients according to their psychopathological states at the end of therapy.

The ERP results showed an overall (day 1, 8, 15) reduced P300 amplitude in the responder group compared with the intermediate and the nonresponder group (and in the responders compared with the nonresponders of the two-group design). P300 divided the responders from both the other groups before therapy, while the psychopathological states were equal. P300 changes did not occur over the course of therapy. Although the neurophysiological basis of P300 generation is still not yet clearly understood (Charles and Hansenne 1992, Polich and Kok 1995), our results point to a biological difference in depressive patients who do or do not respond to light therapy.

The P300 amplitudes of the intermediate group did not differ from those of the nonresponder group. This finding may imply a similar physiological basis of the disorder in the patients of both groups. Therefore the only temporary response to light therapy of the intermediate group might be due to a placebo and not to a physiological effect of light therapy.

It has to be noted that the intermediate group was not equal in terms of medication treatment compared to the two other groups. Only 2 patients received antidepressants, the other 8 patients were drug free. The effect of medication treatment might have shifted P300 amplitude in this group to higher values relative to the others, even though the interaction was not statistically significant.

The results of our study lead us to the conclusion that the biological trait marker P300 amplitude strengthens the definition of a subgroup of depressive patients which consists of light therapy responders. The results were not so clear when the patients were grouped according to seasonal symptoms instead of response to light therapy, which might be due to a masking effect of medication treatment. But it seems to be obvious that the diagnostic criteria SAD symptoms and the clinical criteria therapy response are not independent factors, but are strongly related to each other.

Although group differences became quite clear, there is too much variance between and within the response groups to use the P300 amplitude of a single measure for predicting the success of light therapy treatment.


We wish to thank Kerstin Herwig and Ute Klein for their skilful acquisition of EEG data and Daniela Roesch-Ely for discussing and proof-reading the draft of this paper.


American Psychiatric Association (1987). Diagnostic and Statistic Manual of Mental Disorders, 3rd Ed Rev. Washington DC: American Psychiatric Press.
Anderer P, Semlitsch HV, Saletu B (1989): Ein Korrekturverfahren zur Reduktion okularer Artefakte in der EEG- und ERP-Topographie. In: Saletu B, editor. Fortschritte in der Biologischen Psychiatrie. Stuttgart, New York: Thieme, pp 58-61.
Blackwood DHR, Whalley LJ, Christie JE, Blackburn IM, St Clair DM, McInnes A (1987): Changes in auditory P3 event-related Potential in Schizophrenia and Depression. Br J Psychiatry 150: 154-160.
Charles G, Hansenne M (1992): Le potentiel lent P300 - Intérêt clinique dans trois pathologies mentales et neurobiologie: une revue. L'Encéphale 18: 225-236.
Duncan CC, Mirsky AF, Deldin PJ, Skwerer RG, Jacobsen FM, Rosenthal NE (1990): P300 as an index of treatment response in seasonal affective disorder. In: Stefanis CN, editor: Psychiatry: A world perspective - Volume 2. Amsterdam: Elsevier, pp 398-401.
Eastman CI (1990): What the placebo literature can tell us about light therapy for SAD. Psychopharmacol Bull 26: 495-504.
Eastman CI, Lahmeyer HW, Watell LG, Good GD, Young MA (1992): A placebo-controlled trial of light treatment for winter depression. J Affect Disord 26: 211-221.

Gangadhar BN, Ancy J, Janakiramaiah N, Umapathy C (1993): P300 amplitude in non­bipolar, melancholic depression. J Affect Disord 28: 57­60.Hamilton M (1967): Development of a rating scale for primary depressive illness. Br J Soc Clin Psychol 6: 278-296.
Hansenne M, Pitchot W, Gonzalez-Moreno A, Torrecilas JG, Mirel J, Ansseau M (1994): Psychophysiological Correlates of Suicidal Behavior in Depression. Neuropsychobiology 30: 1-3.
Maurer K, Dierks T (1988): Topographie der P300 in der Psychiatrie - I. Kognitive P300-Felder bei Psychosen. Z EEG-EMG 18: 21-25.
Murphy DGM, Murphy DM, Abbas M, Palazidou M, Binnie C, Arendt J, et al. (1993): Seasonal Affective Disorder: Response to Light as measured by Electroencephalogram, Melatonin Suppression, and Cerebral Blood Flow. Br J Psychiatry 163: 327-331.
Nagayama H, Mitsuo S, Sadaaki I, Hanada K, Okawa M, Ohta T, Asano Y, Sugita Y (1991): Atypical depressive symptoms possibly predict responsiveness to phototherapy in seasonal affective disorder. J Affect Disord 23: 185-189.
Pfefferbaum A, Wenegrat BG, Ford JM, Roth WT, Kopell BS (1984): Clinical application of the P3 component of event-related potentials. II. Dementia, Depression and Schizophrenia. Electroencephalogr clin Neurophysiol 59: 104-124.
Pholien P, Charles G, Tilmant JJ, Wilmotte J (1987): Test de freination par la dexaméthasone et composantes endogènes du potentiel évoqué chez des sujets déprimés. Acta psychiatr belg 87: 694-703.
Polich J, Kok A (1995): Cognitive and biological determinants of P300: an integrative review. Biol Psychol 41: 103-146.
Rosenthal NE, Sack DA, Gillin JC, Lewy AJ, Goodwin FK, Davenport Y, et al. (1984): Seasonal affective disorder; a description of the syndrome and preliminary findings with light therapy. Arch Gen Psychiatry 41: 72-80.
Roth WT, Pfefferbaum A, Kelly AF, Berger PA, Kopell BS (1981): Auditory event-related potentials in schizophrenia and depression. Psychiatry Res 4: 199-212.
Santosh PJ, Malhotra S, Raghunathan M, Mehra YN (1994): A Study of P300 in melancholic Depression - Correlation with psychotic Features. Biol Psychiatry 35: 474-479.
Sara G, Gordon E, Kraiuhin, C, Coyle S, Howson A, Meares, R (1994): The P300 ERP component: an index of cognitive dysfunction in depression? J Affect Disord 31: 29-38.
Tam EM, Law RW, Levitt AJ (1995): Treatment of seasonal affective disorder: A review. Can J Psychiatry 40: 457-466.
Terman M, Terman JS, Quitkin FM, McGrath PJ, Stewart JW, Rafferty B (1989): Light therapy for seasonal affective disorder: a review of efficacy. Neuropsychopharmacology 2: 1-22.
Wirz-Justice A (1993): A decade of light therapy for seasonal affective disorder. In: Jung EG, Holick MF, editors. Biologic effects of Light 1993. Berlin New York: de Gruyter.