www.elsevier.com/locate/jad

Journal of Affective Disorders 80 (2004) 209–219

Research report

The reliability and validity of the Seasonal Pattern Assessment

Questionnaire: a comparison between patient groups

Peter Paul A. Merscha,*, Nanette C. Vastenburgb, Ybe Meestersa,Antoinette L. Bouhuysa, Domien G.M. Beersmac,

Rutger H. van den Hoofdakkera, Johannes A. den Boera

aDepartment of Psychiatry, University Hospital Groningen, Groningen, The NetherlandsbDr. F.S. Meijerskliniek, Forensic Clinic Utrecht, Utrecht, The Netherlands

cZoological Laboratory, University of Groningen, Groningen, The Netherlands

Received 24 September 2002; received in revised form 17 April 2003; accepted 24 April 2003

Abstract

Background: The Seasonal Pattern Assessment Questionnaire (SPAQ) is a frequently used screening instrument in the

research on Seasonal Affective Disorder (SAD). Nevertheless, studies on its reliability and validity are relatively scarce. In the

present study the reliability and the contrast validity of the SPAQ are investigated.Methods: SAD patients, selected by means of

a clinical interview, non-seasonal depressed out-patients, non-depressed out-patients, and a control group, are contrasted to

estimate the discriminating power of the SPAQ. Also, the reliability and factor structure of the seasonality and the climate

subscales are investigated. To study food intake the Seasonal Food Preference Questionnaire (SFPQ) was developed. Results:

The SAD criterion of the SPAQ shows good specificity (94%), but a low sensitivity (44%). Discriminant analysis shows

sufficient ability to classify subjects (81% correctly classified). The Global Seasonality Scale has a good internal consistency. It

consists of two factors, a psychological factor and a food factor. The SFPQ is sensitive for carbohydrate intake by SAD patients.

Limitations: Most SAD patients had received treatment and completed the SPAQ while they were not depressed, which may

have influenced the sensitivity. Conclusions: The SPAQ is not sensitive enough to be considered a diagnostic instrument for

SAD. Nevertheless, it is accurate enough to be used as a screenings instrument. The only false positives were found in the

depressive group. The accuracy of prevalence Figs. can be improved by completion of the SPAQ in the summer months,

combined with the completion of a depression scale.

D 2003 Elsevier B.V. All rights reserved.

Keywords: SPAQ; SAD; Reliability; Validity; Prevalence

0165-0327/$ - see front matter D 2003 Elsevier B.V. All rights reserved.

doi:10.1016/S0165-0327(03)00114-9

* Corresponding author. Department of Biological Psychiatry,

Groningen University Hospital, P.O. Box 30001, 9700 RB

Groningen, The Netherlands. Tel.: +31-50-361-4492; fax: +31-50-

361-9132.

E-mail address: p.p.mersch@acggn.azg.nl (P.P.A. Mersch).

1. Introduction

It is remarkable that one instrument, i.e., the

Seasonal Pattern Assessment Questionnaire (SPAQ;

Rosenthal et al., 1987) has such a prominent place

in the study of Seasonal Affective Disorder (SAD),

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219210

the more so since data on the psychometric qualities

of the SPAQ are relatively scarce. The SPAQ has

two objectives. The original purpose of the SPAQ

was to study a number of characteristics of SAD

like seasonal variation of mood, food intake and

weight gain, seasonal sleep duration, and sensitivity

to weather conditions. Although not developed as

such, the SPAQ was introduced as a diagnostic or

screening instrument for SAD by Kasper et al.

(1989a), who were the first to formulate the SPAQ

criteria for SAD. Since 1989 the SPAQ has been

widely used as a screening instrument. In fact, all

over the world the prevalence figures of SAD are

based on the ‘Kasper criteria’ of the SPAQ (Mersch

et al., 1999a).

Nevertheless, in only a few prevalence studies

responders who met the criteria of the SPAQ for

SAD were interviewed. Kasper et al. (1989a) con-

cluded that the SPAQ underestimated the number of

subjects that met the clinical criteria of SAD as

established in an interview (50% false negatives, 0%

false positives), while Magnusson (1996), evaluating

the results of the Magnusson and Stefansson (1993)

study, found that the clinical interview and the SPAQ

reached approximately the same number of SAD

subjects. The SPAQ in the latter study reached a

sensitivity of 94% and a specificity of 73%. Michalak

et al. (2001) interviewed 66 SAD cases detected by

the SPAQ in a prevalence study on 1999 residents of

the United Kingdom. No less than 55% of the SAD

cases were false positives. No false negatives were

detected in 23 interviewed non-SAD cases. In a study

by Raheja et al. (1996) the SPAQ only missed three

out of 47 SAD patients showing a very good sensi-

tivity (94%). Since only SAD cases were studied, the

specificity could not be assessed. At a follow-up

assessment 5–8 years later the sensitivity was 74%,

while the specificity was 46%.

There is accumulating evidence that the ‘Kasper

criteria’ of the SPAQ lead to overestimation of the

percentage of SAD cases in the general population. In

the Michalak et al. (2001) study the prevalence of

SAD was estimated by the SPAQ as 5.3% while a

DSM-IV interview estimated the prevalence at 2.4%.

In a population survey in the USA Blazer et al. (1998)

studied SAD in a sample of 8098 subjects by means

of a structured interview. Only 0.4% of the respond-

ents met the criteria for SAD. This figure is consid-

erably lower than the mean prevalence figure of SAD

in North America (6.2%), based on the SPAQ criteria

(Mersch et al., 1999a). Two studies investigated the

prevalence in Canada, both using a clinical interview

and the SPAQ in a telephone survey. Results of the

Levitt et al. (2000) study on a sample of 781

respondents showed that the SPAQ detected almost

twice as many SAD subjects (5.0 versus 2.9%) as a

clinical interview based on the DSM-IV (American

Psychiatric Association, 1994). In a second study on

1605 respondents designed to test the latitude hypoth-

esis of SAD (Levitt and Boyle, 2002), the SPAQ

overestimated the mean prevalence three to four

times, compared to a ‘golden-standard’ clinical inter-

view (7.4 versus 1.9%). So, validity studies that have

been performed show mixed results and do not

unequivocally support the confidence in the SPAQ’s

validity suggested by its widespread use (Mersch,

2001).

Another way to establish the validity is to study the

ability of the SPAQ to discriminate between different

groups (e.g., SAD and sub-SAD subjects, different

patient groups and normal control groups), composed

in a way other than by means of SPAQ criteria

(contrast validity). Regretfully, the few studies that

have compared different patient groups with the

SPAQ (Kasper et al., 1989b; Thompson et al., 1988;

Hardin et al., 1991) fail to provide data on the

sensitivity and specificity of the SAD criteria of the

SPAQ.

In the present study the ratings on the SPAQ of

SAD patients were compared with non-seasonal de-

pressed outpatients, non-depressed outpatients and a

control group of healthy volunteers. The main objec-

tive was to establish the validity of the SAD criteria of

the SPAQ. A secondary goal was to establish the

reliability and factor structure of the seasonality scale

(a replication of the Magnusson et al., 1997 study) and

the climate scale. Also, the differences between the

groups with respect to atypical SAD symptoms will

be studied.

2. Methods

2.1. Subjects

Four groups of subjects participated in the study.

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219 211

2.1.1. Winter depressives (SAD)

This group consisted of 45 patients; 32 women

(71.1%) and 13 men. The mean age was 40.7 years

(S.D.: 11.44; range: 19–68). Forty-three subjects

participated in the SAD treatment program at the

Department of Biological Psychiatry of the University

Hospital Groningen. They were diagnosed by an

experienced clinical psychologist according to the

DSM-III-R criteria. Most of these patients had re-

ceived light treatment for their complaints. At the

moment of completion of the questionnaire, two

patients were severely depressed, one moderately,

while three patients were mildly depressed. The mean

Beck Depression Inventory (BDI) score was 4.98

(S.D.: 7.10; range: 0–33).

2.1.2. Depressed outpatients (DP)

At application for treatment at the Psychiatric

Clinic, patients were interviewed by an experienced

clinical psychologist if they had a BDI score higher

than 16. The DSM-III-R interview (American Psy-

chiatric Association, 1987) was performed with the

MDCL (Hiller et al., 1990). Forty-eight subjects met

the criteria for an affective disorder and were willing

to participate in the study. Twenty-seven patients

were women (56.3%), 21 were men. The mean age

was 34.4 years (S.D.: 11.77; range: 16–54). The

mean BDI score was 27.2 (S.D.: 8.44; range: 17–

49). Forty-one patients were diagnosed as having a

major depression; in nine of these patients the

depressive episode was superimposed on a dysthy-

mic disorder, and in one patient on a cyclothymic

disorder. Four patients had a dysthymic disorder, one

a cyclothymic disorder, one a bipolar disorder, while

one patient was diagnosed with an adjustment dis-

order. Two patients were diagnosed as suffering from

SAD and were added to the SAD group. Twenty-one

(43.8%) of the patients had a secondary diagnosis, in

most cases an anxiety disorder. Seventeen patients

(35.4%) used benzodiazepines, anti-depressive med-

ication, or both.

2.1.3. Non-depressive outpatients (NDP)

This group consisted of 46 outpatients who applied

for treatment and had a BDI score lower than 10.

Twenty-three patients were women (50%). The mean

age was 34.7 years (S.D.: 13.33; range: 19–70). The

mean BDI score was 5.0 (S.D.: 2.92; range: 0–9).

2.1.4. Control group (CO)

The 37 subjects of this group consisted of 25

women (67.6%) and 12 men, who responded to an

advertisement in a local newspaper. The subjects had

not had a major depressive disorder in the prior year.

The minimum age was set at 26 years to reduce over-

representation of students. The mean age was 42.2

years (S.D.: 10.39; range: 26–65). The mean BDI

score was 3.4 (S.D.: 2.71; range: 0–5).

2.2. Comparisons between groups

There were no significant differences between the

groups in gender [X 2(3) = 5.39; P= 0.146]. Subjects in

the SAD and CO group were significantly older than

the subjects in both other groups [ F(3) = 4.95;

P= 0.0025; Student–Newman–Keuls (SNK): SAD,

CO>DP, NDP]. The educational level differed signif-

icantly between the groups. In the SAD and the CO

group, respectively, 82.2 and 91.9% of the subjects,

had a medium or higher education, while the2 percen-

tages in the DP and the NDP groups were 65.9 and

77.3%, respectively [X 2(6) = 15.59; P= 0.016]. As

expected the depressive group scored higher than

the other three groups on the BDI [F(3) = 166.06;

P < 0.0001; SNK: DEP>SAD, NDP, CO].

2.3. Instruments

2.3.1. Beck Depression Inventory

This self-report (Beck et al., 1961) instrument

measures the level of depression and consists of 21

questions with four answer possibilities (range 0–3)

each. The score on the total scale ranges from 0 to 63.

Oliver and Simmons (1984) used four categories to

delineate the level of depression: 0–9 nondepressed;

10–15 mildly depressed; 16–23 moderately de-

pressed; 24 and higher severely depressed.

2.3.2. Seasonal Pattern Assessment Questionnaire

(Rosenthal et al., 1987)

The criteria for SAD on the SPAQ have been

formulated in Kasper et al. (1989a). The SPAQ

applies three criteria for SAD:

1. The Global Seasonality (GS) scale provides a

composite measure for change across the seasons

of mood, social activities, appetite, sleep, weight

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219212

and energy. Item scales range from (0) ‘no change’

to (4) ‘extremely marked change’. The total scale

ranges from 0 to 24. The suggested cut off score for

caseness on this criterion is 11 for the self-report

version of the SPAQ.

2. A second criterion for SAD is based on the

question whether seasonal changes are considered

a problem. The response possibilities range from

0= no problem to 5 = a disabling problem. A score

of at least 2 (a moderate problem) is necessary to

reach the SAD threshold.

3. The final criterion is the ‘window’, i.e., the time

interval within which the problems should recur.

The timing of the problems is determined by

asking in what months subjects feel worst. Kasper

et al. (1989a) suggested that subjects should feel

worst in December and/or January and/or February

in order to fulfill the criteria for winter SAD. To

meet the criteria for summer SAD, subjects should

feel worst in June and/or July and/or August.

Furthermore, this calendar section of the question-

naire covers all the seasonality items (except

energy).

Subsyndromal-SAD (S-SAD) is defined as a clus-

ter of seasonal complaints which are not severe

enough to allow for a diagnosis of SAD. The criteria

of S-SAD, defined by Kasper et al. (1989b), are: (1) a

GS score of at least 11 for the self-report version of

the SPAQ and ‘no’ or ‘mild’ problems with the

seasonal changes, or (2) a GS score of 9 or 10 for

the self-report version of the SPAQ and seasonal

changes are either a problem or not. The window is

the same as for SAD.

Furthermore, the SPAQ measures several other

aspects of SAD. Hours of sleep in each season are

scored. Weight change over the year is measured by a

six-point Likert-type scale, ranging from 1 (less than 2

kg) to 6 (more than 10 kg). The influence of clima-

tological conditions on mood and energy is assessed

by seven-point Likert-type scales, ranging from � 3

(very low spirits or markedly slowed down) to + 3

(markedly improved mood or energy level).

2.3.3. Seasonal Food Preference Questionnaire

(SFPQ)

For this study the SFPQ was developed to ask

about seasonal food preference (see Appendix A).

Food categories are listed and subjects are asked in

what season (or seasons) they like each category most.

Scoring is 0 = not present and 1 = present for each

season.

2.4. Procedure

Both the SAD and the CO group completed the

questionnaires in summer. The patients in the DP and

the NDP group completed the BDI and the SPAQ over

a period of 10 months, at the moment of admission at

the clinic.

3. Results

3.1. Seasonality

The level of seasonal variation of the four groups

on the seasonality items is calculated by multiplying

the score on each item with the difference of the

corresponding items of the calendar section, a method

similar as used by Thompson et al. (1988) and Wirz-

Justice et al. (1992). For instance, the score of each

subject on the seasonality item ‘sleep’ is multiplied by

the difference between the scores on the question

‘when do you sleep most’ (0 or 1) and ‘when do

you sleep least’ (0 or 1) on each month. This way a

mean intensity rating for each group for each month of

the year could be calculated for 5 seasonality items

(calendar items on energy do not exist in the SPAQ).

See Fig. 1.

To test the variation over the year the sum of

the absolute values of the mean score on each

seasonality item were compared between the groups

by means of an analysis of variance. Post-hoc

comparisons were done with an SNK at an

a < 0.05 level. On all variables, the SAD group

shows the largest seasonal variation, while on mood

and social activity the DP group shows a larger

variation than normal controls and non-depressive

patients (Table 1).

3.2. Reliability of the Global Seasonality Score

Correlations between the items range from 0.30 to

0.80. The correlations between weight and appetite

(0.70) and between mood and energy (0.80) are high,

Fig. 1. Mean seasonality rating for each month of the year on the five seasonality items of all four groups (double plot).

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219 213

Table 1

Comparisons between groups on seasonal variability represented by the curves of Fig. 1

SAD DP NDP CO ANOVA Post-hoc test

S(Mabs) S.D. S(Mabs) S.D. S(Mabs) S.D. S(Mabs) S.D. df F P(SNK)

Sleep length 13.76 13.53 4.56 6.27 3.00 5.31 3.41 3.91 3,172 16.91 *** SAD>DP, NDP, CO

Social activity 13.13 13.04 7.88 12.32 2.72 6.11 3.86 5.49 3,171 9.62 *** SAD>DP>NDP, CO

Mood 21.64 11.64 10.65 12.29 3.91 7.30 3.89 5.03 3,171 31.80 *** SAD>DP>NDP, CO

Weight 4.91 5.39 2.29 6.04 1.33 2.79 2.76 4.15 3,172 4.53 ** SAD>DP, CO>NDP

Appetite 5.47 5.94 2.83 6.99 1.00 2.58 3.16 5.59 3,172 4.98 ** SAD>DP, CP>NDP

SNK=Student–Newman–Keuls. * P< 0.05, ** P < 0.01, ***P < 0.001.

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219214

while the correlations between weight and social

activity and between appetite and social activity are

low (both 9% explained variance).

Internal consistency analysis shows a high alpha for

the total scale, but somewhat lower correlations with

the total scale for weight and appetite (see Table 2).

Principal component factor analysis offers a two-

factor solution after varimax rotation: a psychological

factor (energy, mood and social activity) and a food

factor (weight and appetite). ‘Sleep length’ loads on

both factors, but since the loading on the first factor is

much higher this variable is placed in the psycholog-

ical factor (see Table 2). The psychological factor

Table 2

Internal consistency and factor analysis of the Global Seasonality scale

Reliability analysis

M S.D. Corr

item-

corre

Sleep length 1.09 1.07 0.69

Social activity 1.21 1.18 0.62

Mood 1.56 1.35 0.73

Energy level 1.62 1.31 0.80

Weight 0.67 0.83 0.50

Appetite 0.74 0.87 0.54

Overall Cronbach’s alpha = 0.85.

Bold, italic figures denote factor loadings higher than 0.40 on the respec

Table 3

Number of subjects of the four groups selected by the SPAQ criteria for

SAD (n= 45) DP (n= 48)

N % N

Winter-SAD 20 44.4 7

Summer-SAD 0 – 1

Winter-sub-SAD 5 11.1 4

explains 58.3% of the variance (eigenvalue 3.50) and

has an internal consistency coefficient of a = 0.87. Thefood factor explains 18.5% of the variance (eigenval-

ue 1.11), while a = 0.82.

3.3. Validity of the SAD criteria of the SPAQ

Of the 45 SAD patients, diagnosed by a clinical

interview, 20 (44.4%) were identified as such by the

SPAQ. Five subjects (11.1%) met the criteria for sub-

SAD winter pattern (see Table 3).

There were 55.6% false negatives in the SAD

group and 6.1% false positives in the three non-

Factor analysis

ected Alpha Rotated factor solution

total if item Factor 1 Factor 2

lation deleted

0.82 0.69 0.40

0.83 0.80 0.12

0.81 0.86 0.19

0.80 0.90 0.21

0.85 0.18 0.90

0.85 0.24 0.89

tive factor.

SAD

NDP (n= 46) CO (n= 37)

% N % N %

14.6 0 – 0 –

2.1 0 – 0 –

8.3 2 4.3 0 –

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219 215

SAD groups combined. The sensitivity of the SPAQ is

44% and the specificity is 94%. The positive predic-

tive value is 71%, while the negative predictive value

is 83%. All the false positive cases were in the

depressive outpatients group showing that the SPAQ

has difficulty differentiating SAD and non-SAD de-

pressive patients. This is also shown by the decline in

specificity to 20% and in negative predictive value to

62% if the analysis is reduced to the DP group.

To study the discriminative ability of the SPAQ

criteria further, a stepwise discriminant analysis was

performed on the SAD group and the other three

groups of non-SAD subjects combined. When the

GS-score and the score on the question whether

problems with the seasons are experienced were

entered in the analysis 80.84% of the subjects were

correctly classified. If both seasonality factors are

entered in the analysis together with the ‘problems

with the seasons’ question the correctly classified

subjects reached 81.8%, while the food factor was

not selected in the analysis.

3.4. Climate

Reliability analysis shows a medium overall alpha

(see Table 4). Principal component factor analysis

with varimax rotation resulted in a three factor solu-

tion (see Table 4). The first factor, explained variance

27.3%, consisted of the variables ‘long days’, ‘dry

Table 4

Internal consistency and factor analysis of the climate scale

Reliability analysis

M S.D. Corrected

item-total

correlation

Hot weather 0.76 1.66 0.24

Sunny days 1.97 1.28 0.27

Dry days 1.02 1.16 0.37

Long days 1.09 1.49 0.37

Cold weather � 0.36 1.28 0.12

Grey cloudy days � 0.94 1.20 0.35

Foggy, smoggy days � 0.70 1.26 0.39

Short days � 0.52 1.29 0.06

Humid weather � 0.84 1.06 0.12

High pollen count 0.02 0.90 0.21

Overall Cronbach’s alpha = 0.56.

Bold, italic figures denote factor loadings higher than 0.40 on the respect

days’, ‘hot weather’ and ‘sunny days’ represented a

summer or good weather factor (Cronbach’s a = 0.77).The second factor (explained variance 21.2%) con-

tains the variables ‘cold weather’, ‘grey cloudy days’,

‘foggy, smoggy days’ and ‘short days’ and represents

a winter or bad weather factor (Cronbach’s a = 0.69).The third factor (explained variance 11.5%) included

‘High pollen count’ and ‘humid weather’ (Cronbach’s

a= 0.26). Both on statistical grounds and in content

this latter factor is not applicable.

SAD patients score more extreme on ‘humid

weather’, ‘sunny days’, ‘grey, cloudy days’, ‘long

days’, ‘foggy, smoggy days’ and ‘short days’, while

the depressive patients score more extreme than the

CO-group and the NDP-group.

Analysis of variance on the four groups on the

summer factor shows that the SAD-group experienced

a significantly greater positive influence on mood

and energy level than the other three groups

[F(3,172) = 5.68, P = 0.001; SNK: SAD>DP, NDP,

CO]. On the winter factor both the SAD patients and

the depressive patients experienced a greater negative

influence of bad weather onmood and energy than both

non-depressive groups [F(3,172) = 9.29, P < 0.0001;

SNK: SAD, DP<NDP, CO].

To study which variables discriminated best be-

tween the SAD group and the other three groups,

stepwise discriminant analysis was performed. Three

variables were selected. At the first step the item

Factor analysis

Alpha Rotated factor solution

if item Factor 1 Factor 2 Factor

deleted

0.54 0.62 � 0.17 0.32

0.53 0.81 � 0.16 � 0.04

0.50 0.80 � 0.09 � 0.12

0.49 0.83 � 0.00 0.06

0.57 � 0.09 0.61 � 0.10

0.50 0.11 0.71 0.26

0.49 0.02 0.80 0.28

0.58 � 0.22 0.73 � 0.16

0.56 � 0.22 0.26 0.72

0.54 0.19 � 0.06 0.70

ive factor.

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219216

‘short days’ was selected, on the second step ‘humid

weather’ and on the third step ‘long days’. Together

they correctly classified 81.3% of the subjects.

3.5. Sleep

On the question of sleep duration over the seasons

all groups showed the same pattern: subjects sleep

most in winter and least in summer (see Table 5).

Between-group analysis showed that winter depres-

sives sleep significantly more in the winter than the

other groups.

3.6. Weight

Weight fluctuation during the course of the year

differed significantly between the four groups (see

Table 5). The weight in both groups depressive

patients varied more than in the group of non-depres-

sive patients.

3.7. The Seasonal Food Preference Questionnaire

A total of 31.3% of the subjects in the study

reported that their food preference, measured with

the SFPQ (see Appendix A), did not vary with the

seasons. There were no significant differences be-

tween the groups [X2(3) = 5.26; P= 0.15]. Of the

68.8% of the subjects who did experience variation

Table 5

Seasonal variation in sleep, weight and food

SAD DP

M S.D. M S.D.

Hours sleep Winter 9.04 1.91 8.13 1.76

Autumn 8.31 1.65 7.76 1.82

Spring 7.80 1.31 7.59 1.48

Summer 7.38 0.78 7.07 1.54

Weight change 1.98 1.06 2.02 1.37

SFPQ Seasonal preference 73.3% 62.5%

Variation in food items 6.79 2.46 4.90 2.51

Winter–summer variation

Total variation 2.27 1.89 1.63 1.57

Carbohydrates 0.73 1.64 0.10 1.02

Non-carbohydrates � 0.73 1.39 � 0.85 1.22

of food preference with the seasons SAD subjects

reported significantly more food items that varied with

the seasons than subjects in the other three groups (see

Table 5).

To study whether there are differences between the

groups between summer and winter in the consump-

tion of food, the summer score was subtracted from

the winter score for all 13 food items. There were no

significant differences between the groups. To test

whether there were differences between the groups in

carbohydrate preference, the kinds of food which

contain high levels of carbohydrate (at least 25%

carbohydrate per 100 g) were clustered and analysed

separately. High carbohydrate foods were bread and

rolls; potatoes; pasta and rice; chocolate, jam and

honey; pastry and biscuits. The difference between

winter and summer on the carbohydrate cluster was

highly significant between the four groups. SAD

subjects indicated more often that they preferred high

carbohydrate foods in winter compared to the sum-

mer. In The Netherlands, the increased consumption

of potatoes and pastry in winter contributed most to

this cluster (57.2%).

4. Discussion

The main goal of the present study was to establish

the capability of the SPAQ and the SAD criterion

NDP CO ANOVA Post-hoc test

M S.D. M S.D. df F p(SNK)

8.27 1.27 8.19 0.85 3,169 3.48 0.017 SAD>DP,

NDP, CO

8.00 1.13 8.03 0.73 3,169 1.14 ns

7.78 1.11 7.76 0.72 3,169 0.29 ns

7.38 0.91 7.49 0.77 3,169 1.27 ns

1.38 0.79 1.75 1.27 3,163 2.83 0.04 SAD, DP>NDP

60.9% 81.1%

3.96 2.63 4.80 2.19 3,118 7.59 0.0001 SAD>DP,

NDP, CO

1.63 1.58 1.97 1.44 3,172 1.62 ns

� 0.15 0.84 � 0.08 0.89 3,172 5.43 0.0014 SAD>DP,

NDP, CO

� 1.00 1.21 � 1.30 1.22 3,172 1.49 ns

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219 217

developed by Kasper et al. (1989a) to discriminate

between different patient groups and a normal control

group. In general, the SPAQ performed quite well.

The specificity was excellent, but the sensitivity is less

impressive. Discriminant analysis showed that the

criteria of the SPAQ were reasonably able to place

subjects in the right group. However, compared to the

Magnusson (1996) and the Raheja et al. (1996)

studies the sensitivity of the SPAQ criteria is consid-

erably lower in the present study, while the specificity

is better.

Especially the ability of the SPAQ to detect SAD

cases in the present study is questionable. A reason for

the high percentage false negatives may be that in the

SAD group the SPAQ was administered in the sum-

mer. Although, Mersch et al. (1999b) found seasonal

influences in completion of the seasonality scale, the

prevalence was not influenced because of the stability

of the other two elements of the criterion. Recent

studies of Levitt and Boyle (2002), Levitt et al. (2000)

and Lund and Hansen (2001) showed however, that

the time of year in which the data on the prevalence of

SAD are collected may influence these data consid-

erably. Also, the fact that most patients had success-

fully received light treatment for their complaints may

have influenced the ratings. Most of the subjects were

patients of our out-patient clinic for several years and

received light treatment every winter. The disappear-

ance of seasonal problems by an effective treatment in

the early stage of their depressive episode may have

led these patients to answer the ‘problems with the

seasons’ question negatively. Indeed, of the 25 SAD

patients who were not identified by the SPAQ, 13 did

not report seasonal problems. The lower sensitivity at

the follow-up assessment compared to the first assess-

ment in the Raheja et al. (1996)) study may in part

also be explained by such factors. A solution for

future research may be to put more emphasis on the

instruction and to ask patients to answer the question

on the basis of their complains before they ever

received treatment for SAD or to instruct them to

report seasonal problems if they required light thera-

py, even if the therapy prevented the occurrence of

severe problems.

Concerning the specificity of the SPAQ the situ-

ation is less problematic since the overall number of

false positives over the three non-SAD groups is low

(6.1%). An interesting finding is that the only false

positives in the study were found in the group of

depressive patients (16.7%). It may be that the

overestimation of SAD in prevalence studies can

be attributed in part to respondents who suffer from

non-seasonal depressive symptoms. In the prevalence

study in the Netherlands (Mersch et al., 1999b) the

SPAQ was sent each month to a sample of the

population together with a depression scale, the

CES-D (Ensel, 1986). It was remarkable that even

when the questionnaires were completed in the

summer months the respondents who met the SPAQ

criteria for SAD had a mean depression score well

above the cut-off score for possible caseness. Also,

all of the respondents who met the SPAQ criteria for

SAD reported recurrent depressive episodes indepen-

dent of season. In view of the results of the present

study, it may be that a number of the respondents

who met the SAD criteria were in fact subjects who

suffered from depressive episodes independent of

season. If this is the case it is imaginable that the

winter season increases the depressive symptoms and

may lead to a confirming answer on the ‘problems

with the seasons’ question. An indication that this

may be the case is given by the significantly greater

negative influence of the ‘bad weather’ factor on

mood of both the SAD group and the DP group

compared to the NDP and CO group. In future

research a more reliable measure of prevalence

may be reached by performing the study in the

summer months and by accompanying the SPAQ

by a validated depression scale. Cases identified as

SAD cases by the SPAQ should be excluded if they

have a depression score above the cut-off score of

possible caseness. Reanalysing the data of the

Mersch et al. (1999b) study by only including those

respondents who completed the questionnaire in

summer, met the SAD criteria of the SPAQ and

had a CESD score lower than 16, we found a

lowering of the prevalence of winter SAD in The

Netherlands from 3.1% to 1.4%. In view of recent

findings (Blazer et al., 1998; Levitt et al., 2000;

Levitt and Boyle, 2002; Michalak et al., 2001) this

figure is more realistic.

The Global Seasonality scale shows a good internal

consistency (a = 0.85), which is in line with the results

of the Magnusson et al. (1997) study (a = 0.82). Theresults of the present study confirm the existence of a

two-factor structure of the seasonality scale, which

Food Spring Sum-

mer

Aut-

umn

Win-

ter

No pre-

ference

1. Bread, rolls

2. Potatoes

3. Pasta, rice

4. Fruit

5. Meat, Fish

6. Meat products,

sausages

7. Boiled vegetables

8. Raw vegetables

9. Chocolate, jam, honey

10. Ice cream

11. Pastry, biscuits

12. Coffee, tea

13. Alcohol

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219218

can be characterised as a ‘psychological factor’ and a

‘food factor’, with approximately the same explained

variance for both subscales. Together with the ‘sea-

sonal problems’ question the psychological factor is a

better predictor than the food factor.

Results on the atypical symptom items of the

SPAQ show that they reliably discriminate between

SAD and non-SAD subjects. As expected, the items

‘short days’ and ‘long days’ are, together with

‘humid weather’, the best climatologic discriminators

between SAD and non-SAD subjects. On both the

summer as well as the winter factor SAD patients

were found to be more influenced by the climate. As

expected the patients in the SAD group sleep more

in the winter than the other groups. The question on

weight change seems not specific enough to differ-

entiate between the SAD and the DP group. Since

weight change is also one of the symptoms in

unipolar depression, a more detailed question is

necessary. The characteristic difference between

SAD patients and unipolar depressive patients is

the atypical carbohydrate craving by the former

patients. The Seasonal Food Preference Question-

naire, developed for the present study, appeared to

differentiate well between SAD and non-SAD sub-

jects on the high carbohydrate items. The stronger

overall seasonal variation in food items of the SAD

group compared to that of the other three groups

appeared to be determined mainly by the preference

of high-carbohydrate foods in winter.

In view of the low sensitivity, the conclusion is that

the SPAQ fails as a diagnostic instrument, although it

may perform better with patients who apply for

treatment the first time. The SPAQ seems accurate

enough to be used as a screening instrument but the

earlier discussed adjustments: completion of the

SPAQ in summer, together with the completion of a

depression scale and exclusion of possible depressive

cases, are indispensable.

Appendix A. Seasonal Food Preference

Questionnaire (SFPQ)

Do you notice any change in your preference of

certain foods over the seasons?

5No 5Yes

If Yes, in what season do you feel most like the

following foods: (You may select more than one

season)

References

American Psychiatric Association, 1987. Diagnostic and Statistical

Manual of Mental Disorders, 3rd Edition. American Psychiatric

Association, Washington, DC.

American Psychiatric Association, 1994. Diagnostic and Statistical

Manual of Mental Disorders, 4th Edition. American Psychiatric

Association, Washington, DC.

Beck, A.T., Ward, C.H., Mendelson, M., Mock, J.E., Erbaugh, J.K.,

1961. An inventory for measuring depression. Arch. Gen. Psy-

chiatry 4, 561–571.

Blazer, D.G., Kessler, R.C., Swartz, M.S., 1998. Epidemiology of

recurrent major and minor depression with a seasonal pattern.

The National Comorbidity Survey. Br. J. Psychiatry 172,

164–167.

Ensel, W.M., 1986. Measuring depression: the CES-D scale.

In: Lin, N., Dean, A., Ensel, W.M. (Eds.), Social Support,

Life Events, and Depression. Academic Press, Orlando, FL,

pp. 51–70.

Hardin, T.A., Wehr, T.A., Brewerton, T., Kasper, S., Berrettini, W.,

Rabkin, J., Rosenthal, N.E., 1991. Evaluation of seasonality in

six clinical populations and two normal populations. J. Psy-

chiatr. Res. 25, 75–87.

Hiller, W., von-Bose, M., Dichtl, G., Agerer, D., 1990. Reliability

of checklist-guided diagnoses for DSM-IIIR affective and anxi-

ety disorders. J. Affect. Disord. 20, 235–247.

P.P.A. Mersch et al. / Journal of Affective Disorders 80 (2004) 209–219 219

Kasper, S., Rogers, S.L.B., Yancey, A., Schultz, P.M., Skwerer,

R.G., Rosenthal, N.E., 1989b. Phototherapy in individuals with

and without subsyndromal seasonal affective disorder. Arch.

Gen. Psychiatry 46, 837–844.

Kasper, S., Wehr, T.A., Bartko, J.J., Gaist, P.A., Rosenthal, N.E.,

1989a. Epidemiological changes in mood and behavior. A tele-

phone survey of Montgomery County, Maryland. Arch. Gen.

Psychiatry 46, 823–833.

Levitt, A.J., Boyle, M.H., 2002. The impact of latitude on the

prevalence of seasonal depression. Can. J. Psychiatry 47,

361–367.

Levitt, A.J., Boyle, M.H., Joffe, R.T., Baumal, Z., 2000. Estimated

prevalence of the seasonal subtype of major depression in a

Canadian community sample. Can. J. Psychiatry 45, 650–654.

Lund, E., Hansen, V., 2001. Responses to the Seasonal Pattern

Assessment Questionnaire in different seasons. Am. J. Psychia-

try 158, 316–318.

Magnusson, A., 1996. Validation of the Seasonal Pattern Assess-

ment Questionnaire (SPAQ). J. Affect. Disord. 40, 121–129.

Magnusson, A., Friis, S., Opjordsmoen, S., 1997. Internal consis-

tency of the Seasonal Pattern Assessment Questionnaire

(SPAQ). J. Affect. Disord. 42, 113–116.

Magnusson, A., Stefansson, J., 1993. Prevalence of seasonal affec-

tive disorder in Iceland. Arch. Gen. Psychiatry 50, 941–946.

Mersch, P.P.A., Middendorp, H., Bouhuys, A.L., Beersma, D.G.M.,

Van den Hoofdakker, R.H., 1999a. Seasonal Affective Disorder

and latitude: a review of the literature. J. Affect. Disord. 53,

35–48.

Mersch, P.P.A., Middendorp, H., Bouhuys, A.L., Beersma, D.G.M.,

Van den Hoofdakker, R.H., 1999b. The prevalence of Seasonal

Affective Disorder in The Netherlands: a prospective and retro-

spective study of seasonal mood variation in the general pop-

ulation. Biol. Psychiatry 45, 1013–1022.

Mersch, P.P.A., 2001. Prevalence from epidemiological surveys. In:

Partonen, T., Magnusson, A. (Eds.), Seasonal Affective Disor-

der: Practice and Research. Oxford University Press, Oxford,

pp. 121–141.

Michalak, E.E., Wilkinson, C., Dowrick, C., Wilkinson, G., 2001.

Seasonal affective disorder: prevalence, detection and current

treatment in North Wales. Br. J. Psychiatry 179, 31–34.

Oliver, J.M., Simmons, M.E., 1984. Depression as measured by the

DSM-III and the Beck Depression Inventory in an unselected

adult population. J. Consult. Clin. Psychol. 52, 892–898.

Raheja, S.K., King, E.A., Thompson, C., 1996. The seasonal pat-

tern assessment questionnaire for identifying seasonal affective

disorders. J. Affect. Disord. 41, 193–199.

Rosenthal, N.E., Genhart, M., Sack, D.A., Skwerer, R.G., Wehr,

T.A., 1987. Seasonal affective disorder: relevance for treatment

and research of bulimia. In: Hudson, E.L., Pope, H.G. (Eds.),

Psychobiology of Bulimia. American Psychiatric Press, Wash-

ington, DC.

Thompson, C., Stinson, D., Fernandez, M., Fine, J., Isaacs, G.,

1988. A comparison of normal, bipolar and seasonal affective

disorder subjects using the Seasonal Pattern Assessment Ques-

tionnaire. J. Affect. Disord. 14, 257–264.

Wirz-Justice, A., Krauchi, K., Graw, P., Schulman, J., Wirz, H.,

1992. Seasonality in Switzerland: an epidemiological survey.

Soc. Light Treat. Biol. Rhythms 4, 33, Abstract.