VU Research Portal

Combination treatment for depressed outpatients: efficacy and prediction of outcome

Blom, M.B.J.

2007

document versionPublisher's PDF, also known as Version of record

Link to publication in VU Research Portal

citation for published version (APA)Blom, M. B. J. (2007). Combination treatment for depressed outpatients: efficacy and prediction of outcome.

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?

Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

E-mail address:vuresearchportal.ub@vu.nl

Download date: 06. Apr. 2021

https://research.vu.nl/en/publications/9c92c672-b2d5-4912-8e13-a846dff765c4

105

6

Consistency of the NEO-FFI during

an episode of major depression

Blom M.B.J., Hoffman, T., Spinhoven, Ph. Jonker, K., Hoencamp, E., Haffmans P.M. J.,

and van Dyck, R. Manuscript submitted for publication.

106

ABSTRACT Five ways of assessing continuity of personality factors were examined in 138

outpatients treated for major depressive disorder. Structural, absolute, individual-level,

ipsative and differential continuity, using the NEO-FFI, were all shown to be stable over

time. Changes in Neuroticism and Agreeableness were significant over time, but much

smaller compared to changes in depression severity. Changes on the HAMD and BDI

showed important differences in overlap with changes on personality dimensions, of

which Neuroticism was foremost. In this study it is shown that personality factors can be

reliably assessed in the course of treatment effective in reducing depression severity, but

that correlations between changes in depression scores and changes in personality

dimensions are more pronounced on a self-rated compared to an expert-rated instrument.

107

Reliable assessment of personality factors during an episode of major depressive disorder

(MDD) is thought as questionable. The discussion has been much influenced by an early

study by Hirschfeld et al. (1983b). In this study personality factors were found to be

strongly influenced by state factors. Hence it was concluded that personality factors

cannot be reliably assessed during an episode of MDD. This view is supported by a large

number of studies showing that scores on personality scales change significantly after

treatment (Coyne & Gotlib, 1983; Bagby, Joffe et al., 1995; Griens et al., 2002).

Especially for the personality dimension Neuroticism (Ne), described as a “general

tendency to experience negative affects…” (Costa & McCrae, 1992, page 14), this has

been found. If Ne and symptoms of depression show a significant overlap, Ne cannot be

evaluated separately from current symptoms. Clinicians then should be cautioned to draw

conclusions from assessments made during an episode of MDD (Enns & Cox, 2005;

Griens et al., 2002).

Since then several authors have argued that although personality factors are influenced

by symptom level, this does not imply that personality factors cannot be reliably assessed

during a depressive episode. For instance Santor, Bagby et al. (1997) found that in a

sample of 71 depressed outpatients measures of Ne and Extraversion (E) did change in

relationship with depression severity, but at the same time changes in these scores were

only moderately accounted for by changes in depression scores. Costa, Bagby et al.

(2005a) compared test-retest correlations of the NEO-PI-R in a sample of 250 outpatients

with MDD. Their study showed that scores on the Ne scale of the NEO-PI-R decreased

as did scores on the Openness (O) and Conscientiousness (C) scales in patients who

responded to treatment. Scores on the E-scale increased. The authors however concluded

that since changes were small and the results of factor analysis of the NEO-PI-R was

similar before and after treatment, this did not mean that measurements were unreliable

but that an accentuation of personality factors occurred during an episode of MDD.

Also, as has been shown before, Ne scores remained high in remitted patients (Bagby et

al., 1995) suggesting either a scar effect or a true personality characteristic.

Much of the confusion about whether or not personality factors can or cannot be reliably

measured during an episode of MDD is due to the nature of personality change. Both

108

Roberts and DelVecchio (2000) and De Fruyt, Van Leeuwen et al. (2006) distinguish

several types of personality change. In most studies in this area only one of these types of

consistency is studied.

The five types of personality change described by De Fruyt et al. (2006) are differential

continuity, referring to relative differences among individuals over time, absolute

continuity, referring to the change over time of a group of patients, individual-level

continuity, referring to the magnitude of change of each individual on any trait, ipsative

continuity, referring to stability of the ordering and organization of traits within the

individual and structural continuity referring to the stability of the co-variation among

traits over time. Each of these five types of continuity describes a unique facet of stability

of personality dimensions over time. Studying all these five types of change and

continuity has several important advantages. It provides insight into the nature of change

on a group level as well as on an individual level. Ideally patients should be examined

before the onset of depression to get a true estimate of personality factors. In everyday

clinical practice this is almost never possible. However if it is shown that personality

factors can be measured reliably during an episode of MDD, the study of the role of

personality factors in treatment outcome, maintenance of symptoms and recurrence could

be more easily elucidated.

The currently most often used model to describe normal personality is the Five Factor

Model (FFM). The FFM can be reliably measured using the NEO-PI-R or the NEO-FFI

(Costa & McCrae, 1992). In the study by de Fruyt et al (2002) a French instrument, the

Système de Description en Cinq Dimensions (DSD) (Rolland & Mogenet, 2001) was

used for the measurement of the big Five. They found that the DSD showed consistency

on all five types of possible personality change and concluded that “personality states

remain stable … during the course of different antidepressant therapies”. They however

recommend that a replication be carried out using a more commonly used instrument to

assess the Five Factor Model. They also pointed to an inconsistency between several

studies (Du, Bakish et al., 2002), and their own (De Fruyt et al., 2006), in the correlation

between depression change scores during treatment and change scores of personality

traits. They hypothesized that this discrepancy could partly be due to a different method

109

of scoring: i.e. self-rating versus expert-rated instruments. To solve this important

clinical question we compared in this study a self-rated instrument for the measurement

of depression severity (Beck Depression Inventory (BDI) (Beck et al., 1961) with an

expert rated instrument (Hamilton Rating Scale for Depression (HAMD) (Hamilton,

1960).

We hypothesized that we would confirm the findings by de Fruyt et al. (2006) in showing

continuity of personality dimensions during treatment of MDD. We further hypothesized

that these findings would be independent from the method of administering an instrument

(self-rating versus expert-rating). To our knowledge this is the first study using the NEO-

FFI before and after treatment of depression which investigates all five types of

dimension consistency for this instrument while assessing depression severity with both a

self-rated and expert-rated instrument.

Measurement of dimension consistency Intra-individual continuity can be analyzed using the residualized change index (RCI)

(Jacobson & Truax, 1991) or growth curves estimates of change (Tate & Hokanson,

1993). In this article the method described by De Fruyt et al (2006) using RCI will be

performed.

Ipsative continuity is as the name implies, exclusively person-oriented and refers to the

stability of the ordering of dimensions within the individual. For this, participants are

assessed at least two times on measures of personality and are further examined on

consistency of the configuration of the personality dimensions.

Ipsative personality is examined using dissimilarity (D) indices discussed by Cronbach

and Gleser (1953): D2, D’2 and D’’2. Dissimilarity indices describe the similarity of two

personality dimensions of the same person by quantifying this relationship. A large value

indicates great dissimilarity and a small value indicates less dissimilarity. These indices

are used to describe different parts of profile variation. These variations can be divided in

elevation (the mean level of dissimilarity for a given person), scatter (the individual

deviation about his own mean level of dissimilarity), and shape (the residual information

110

indicating a persons’ individual patterning of scores). D2 is sensitive to all three parts of

variation (elevation, scatter and shape), D’2 is only sensitive to scatter and shape by

removing the effect of elevation, and D’’2 is only sensitive to shape.

Mean-level consistency is the most well known type of personality change and refers to

the extent in which personality scores of the cohort change over time.

Rank-order consistency or differential consistency refers to the degree in which the

relative difference among individuals remain stable over time (De Fruyt et.al, 2006).

Method

Design and procedure All subjects participated in a randomized clinical efficacy study. The full design and

outcome of the study have been described elsewhere (Blom, et al., 2007). Participants

were required to be 18 years or older, with non-psychotic, non-bipolar major depressive

disorder (MDD) and a score ≥ 14 on the Hamilton Depression Rating Scale (HAMD, 17-

item) (Hamilton, 1960). Participants fulfilled the criteria of MDD as specified in the

DSM-IV (APA, 1994) and were assessed with the Structured Clinical Interview for Axis

I DSM-IV Disorders, Dutch translation (SCID) (Groenestijn, Akkerhuis et al., 1998).

Excluded from the study were participants with substance abuse, serious medical

condition, organic psychiatric disorder, severe suicidality, history of psychotic disorder

or schizophrenia, bipolar disorder, current use of psychotropic medication and ongoing

psychotherapy. The study was approved by an independent Ethical Commission. All

participants gave written informed consent before entering the study.

Participants Participants who were referred to treatment either by their primary care physician or by

another mental health professional were assessed for the study. No participants were

contacted through advertisements.

Measures Depression severity was measured using the Hamilton Depression Rating Scale

(Hamilton, 1960), 17-item version (HAMD). The HAMD is an expert rated measure of

111

depression severity. Before entering treatment, the SCID and the HAMD were

administered by experienced clinical raters not informed of the treatment condition.

Monthly interrater trainings were held to ensure good interrater reliability. Internal

consistency has been found to be between .46 and .97 (Bagby et al., 2004;Santor et al.,

1997). In the present study Crohbach’s alpha was .533, .716 and .845 for the three points

of measurement.

The Beck Depression Inventory (BDI) (Beck et al., 1961), a 21-item self report measure

of depression severity, was also completed at baseline, after 6 weeks and upon

completion of the study. The BDI has a high internal consistency, with alpha coefficients

of .86 for psychiatric populations (Beck, Steer, & Garbin, 1988). In this study

Cronbach’s alpha was .825, .879 and .925 respectively.

Personality dimensions were studied using the NEO-Five Factor Inventory (NEO-FFI)

(Costa & McCrae, 1992). It is the short form of the NEO-PI-R and has excellent

correlation (.88-.94) with the longer NEO PI-R (Costa & McCrae, 1992; Hoekstra, Ormel

et al., 1996). Correlations with the individual scales of the NEO-PI-R were found to be

for respectively Ne: .93; E: .90; O: .94; A: .88 and C: .89 (Costa & McCrae, 1992). The

NEO-FFI was administered at baseline and at endpoint.

Analytic Method Differential continuity was examined by using Pearson correlation coefficients between

the two assessments periods and structural continuity was examined by using structural

equations modelling (SEM) as described by (Robins, Fraley et al., 2001). After a visual

inspection of the intercorrelations, this was followed by a formal test of equivalence of

the correlation matrix. This was done by comparing the fit of a model in which the

intercorrelations among the NEO-FFI were freely estimated (baseline model) and a

model in which the intercorrelations were constrained to be equivalent across the two

assessment periods. If the fit of the constrained model compared to the fit of the baseline

model (unconstrained model) resulted in decline of the overall fit, it would indicate

significant changes in the intercorrelations among the NEO-FFI. The baseline model was

112

specified as a single-indicator latent variable model with one latent variable associated

with each of the 10 scale scores. The model was identified by fixing the variances of the

latent variables to 1 and the variances of the residuals to 0. The paths among all latent

variables and between each latent variable and its scale score were also estimated. As a

result, the matrix of covariances among the latent variables is equivalent to the matrix of

correlations. The baseline model is fully saturated. The constrained model was estimated

by placing 10 pairwise equality constraints between the scale scores across the

assessment periods.

Absolute continuity or mean-level change was examined by descriptive statistics,

Pearson correlations between the personality dimensions within and across time periods.

The correlations were corrected for unreliability (Hoekstra et al., 1996; Schmitt, 1996).

Next, a repeated measures analysis on the two assessment periods was used to test the

mean-level change. Effect sizes were calculated using an adjusted Cohen’s d for repeated

measures designs (Dunlop, Cortina et al., 1996). A regression analyses with partial

change scores of the personality measures as dependent variables and partial change

scores for depression as the independent variable was used to predict the personality

change by accounting for changes in depression. Partial change scores were calculated to

account for the time-dependent interrelation of same personality dimensions (Cohen,

Cohen et al., 2006). Rather than using simple change scores, partial change scores were

used. If we had used simple change scores, this would have suggested that regression of

the post-scores on the pre-scores had a slope of 1.0 instead of the regression coefficient

beta. This would risk an overcorrection of the postscores by the prescores (Cohen et al.,

2006).

Partial change scores of depression were regressed on partial change scores of all

personality dimensions using hierarchical multiple regression. The order of inclusion of

independent variables was as follows: Ne, E, A, C, and finally O. This order of

independent variables was based on Mulder (2002) indicating that Ne is the best

predictor followed by E.

113

Individual-level continuity was assessed by using the Reliable Change Index (RCI)

(Jacobson & Truax, 1991). This index was used to describe participants showing

decreased, unchanged or increased personality dimension scores. Participants were

classified as having a reliable increase or decrease in NEO-FFI scores when the

probability of the RCI was less than 5%. These categories were further tested as in

Robins et al. (2001) from the expected frequencies of 2.5% (reliable decrease), 95% (no

change), and 2.5% (reliable increase) for each of the personality dimension scores by

using chi-square analysis. Using the RCI requires reliability estimates, Cronbach’s alpha

and descriptive statistics from a normative sample. These data were obtained from the

Dutch NEO-FFI manual (Hoekstra, et al., 1996).

Ipsative continuity was examined by using D2, D’2 and D’’2 indices (Cronbach & Gleser,

1953). Participants were classified as having changed as the probability of the D2, D’2 and

D’’2 indices were less than 5%. Parallel to De Fruyt et al. (2006), these probabilities were

estimated by simulating dimension scores on a sample of 100,000 persons in which there

was no change in means, variances, covariances and coefficient alpha reliabilities as

estimated from the real data.

Results The sample consisted of 193 participants. Of these, 138 (71.5%) completed the study.

For a description of clinical and demographical variables see table 1. Some participants

were not able or refused to complete the NEO FFI either at baseline or at endpoint. A

total of 126 (91.3%) completed the NEO-FFI both at baseline and upon completion of the

study..

Participants who dropped out (DO) differed in some aspects significantly from the

completer sample: they were more often single (chi-square (1) = 4.473; p = 0.034) and

were more often from a non-European (NE) background (chi-square (1) = 6.840; p =

0.009). Illness-related factors, such as duration and severity of the index episode and

number of previous episodes, did not differ between DO and completers (F (1, 178) =

.003; p = .956; F (1, 190) = 1.069; p = .302, respectively chi-square (1) = .769; p = .380).

114

DO scored significantly higher on the Ne-dimension of the NEO-FFI (F=7.821; p =

0.006), but not on any of the other four NEO-FFI dimensions.

Table 1: Baseline demographic and clinical characteristics of all

participants who completed the trial.

Mean (SD) / %

(frequency) N

Age 40.82 (11.18) 137Sex (% female) 64.2 (88) 137Age first diagnosis 35.8 (12.58) 130Length index episode 11.69 (13.73) 132First episode (% first) 72.6 (98) 135Marital status (% single) 39.6 (53) 134Cultural background (% non-European)

14.5 (20) 138

Melancholic features (% melancholic)

56.6 (73) 129

Baseline Neuroticism 36.63 (5.48) 126Baseline Extraversion 35.77 (4.72) 125Baseline Openness 34.10 (3.85) 124Baseline Agreeableness 36.92 (4.45) 126Baseline Conscientiousness 38.65 (4.21) 125Baseline HAMD 21.14 (4.73) 138Baseline BDI 31.20 (9.33) 138

Differential continuity and structural stability Table 2 shows the intercorrelations among the Big Five Dimensions assessed at 0 weeks

and 12 weeks. On visual inspection one sees that the intercorrelations reported above the

diagonal (week 12) strongly parallel those below the diagonal (week 0). To formally test

the structural stability of the Big Five, SEM was used to compare the correlation matrices

on equivalence. A chi-squared difference test indicated that the constrained model did not

lead to a significant reduction in fit (chi-square (10) = 14.0255; p = .172 and CFI = .98;

115

SRMS = .026; RMSEA = .063). The intercorrelations are invariant across the two

assessment periods.

116

Tabl

e 2:

Str

uctu

ral C

ontin

uity

of P

erso

nalit

y: In

terc

orre

latio

ns A

mon

g B

ig F

ive

Dim

ensi

ons

Ass

esse

d at

Bas

elin

e an

d En

dpoi

nt.

N

E

O

A

C

Neu

rotic

ism

-

-.02

.07

.41*

**

-.06

Ext

rave

rsio

n .0

5 -

.37*

**

-.07

.53*

**

Ope

nnes

s .1

2 .2

7***

-

.05

.36*

* A

gree

able

ness

.26*

* .1

5 .1

6**

- .2

0*

Con

scie

ntio

usne

ss

.20*

.4

1***

.2

8**

.23*

-

Not

e In

terc

orre

latio

ns a

t bas

elin

e ar

e re

port

ed b

elow

the

diag

onal

, and

in

terc

orre

latio

ns a

t end

poin

t are

repo

rted

abo

ve th

e di

agon

al.

*

p >

05

** p

> .0

1 **

* p

> .0

01

Abso

lute

Con

tinui

ty o

r Mea

n-Le

vel C

hang

e fo

r HAM

D

The

desc

riptiv

e st

atis

tics a

nd c

orre

latio

ns a

mon

g th

e va

riabl

es a

re d

escr

ibed

in T

able

3.

117

Tabl

e 3:

Des

crip

tive

Stat

istic

s an

d C

orre

latio

ns A

mon

g Va

riabl

es

1 2

3 4

5 6

7 8

9 10

11

12

13

14

15

M

SD

1 : B

DI_

0T

1

31

,09

9,47

2:

BD

I_6T

.6

5 1

25

,89

10,3

3 3:

BD

I_12

T

.51

.76

1

21

,65

11,5

4 4

: HA

MD

_0T

.5

3 .4

5 .3

6 1

20

,95

4,40

5:

HA

MD

_6T

.4

0 .7

2 .5

4 .5

5 1

17,3

4 5,

86

6: H

AM

D_1

2T

.36

.56

.73

.41

.64

1

14,8

0 8,

34

7: N

_0T

.3

4 .2

4 .0

7 .2

0 .0

8 -.0

5 1

36,4

1 4,

95

8: E

_0T

-.1

5 -.1

5 -.1

6 -.1

9 -.1

5 -.0

9 .0

5 1

35

,50

4,81

9:

O_0

T

-.03

.02

.08

-.06

-.03

-.05

.12

.27

1

33

,89

3,76

10

: A_0

T

.21

.10

.05

-.02

.08

-.04

.26

.15

.16

1

36,8

4 4,

42

11: C

_0T

-.1

0 -.0

7 -.1

7 -.1

3 -.0

3 -.1

4 .2

0 .4

1 .2

8 .2

3 1

38,2

0 4,

53

12: N

_12T

.2

2 .3

6 .3

8 .1

2 .2

5 .1

8 .4

7 .1

1 .0

5 .1

4 -.0

7 1

35

,08

4,35

13

: E_1

2T

-.20

-.30

-.48

-.13

-.33

-.45

.11

.55

.25

.08

.39

-.02

1

35

,42

4,31

14

: O_1

2T

-.24

-.19

-.25

-.16

-.06

-.27

.10

.22

.53

.13

.21

.07

.37

1

34,4

7 3,

44

15: A

_12T

.2

5 .3

1 .3

1 .1

2 .1

9 .1

1 .1

9 -.0

4 .1

0 .4

7 .0

8 .4

1 -.0

7 .0

5 1

35,4

9 3,

88

16: C

_12T

-.1

6 -.0

5 -.2

1 -.0

9 .0

4 -.1

8 .0

4 .2

9 .2

1 .1

5 .6

1 -.0

6 .5

3 .3

6 .2

0 37

,89

3,64

BD

I: B

eck

Dep

ress

ion

Inve

ntor

y; H

AM

D: H

amilt

on D

epre

ssio

n R

atin

g Sc

ale,

Ne:

Neu

rotic

ism

; E: E

xtra

vers

ion;

O:

Ope

ness

; A: A

gree

able

ness

; C: C

onsc

ient

ious

ness

; 0t:

base

line;

6T:

ratin

g at

6 w

eeks

; 12T

: end

poin

t rat

ing.

118

Ta

ble

4: A

bsol

ute

chan

ge in

per

sona

lity

and

depr

essi

on

Wee

k 0

Wee

k 12

Scor

es

M

SD

M

SD

N

F (1

, 94

) p

Part

ial-e

ta

squa

red

Adj

uste

d C

ohen

’s d

N

euro

ticis

m

36.4

1 4.

9535

.08

4.35

95

7.25

90.

008

0.07

2 0.

285

Ext

rave

rsio

n 35

.50

4.81

35.4

24.

31

950.

035

0.85

1 0.

000

0.01

8 O

penn

ess

33.8

9 3.

7634

.47

3.44

95

2.46

60.

120

0.02

6 0.

159

Agr

eeab

lene

ss

36.8

4 4.

4235

.49

3.88

95

10.2

00.

002

0.09

8 0.

324

Con

scie

ntio

usne

ss

38.2

0 4.

5337

.89

3.64

95

0.69

70.

406

0.00

7 0.

075

Dep

ress

ion

seve

rity

(HA

MD

) 20

.95

4.40

14.8

08.

34

9554

.33

0.00

0 0.

366

0.82

0

Dep

ress

ion

seve

rity

(BD

I)

31.0

9 9.

4721

.65

11.5

495

81.3

40.

000

0.46

4 0.

884

119

Personality factors before and after treatment are highly correlated, with correlations of

.469, .552, .527, .472 and .609 for Ne, E, O, A and C respectively. After correction for

unreliability (Hoekstra et al., 1996) these correlations achieved high levels of .558, .756,

.798, .694 and .883.

Absolute continuity was assessed by comparing mean-level scores at week 0 and week

12 using repeated measures ANOVA. Absolute changes in personality and depression are

reported in Table 4.

Depression scores decreased considerably after treatment, but participants described

themselves on average also as slightly less neurotic, more open to experience, and less

agreeable after treatment. The changes in personality represent small changes because

none of the effect sizes indicated large values (small = .02, medium = .15, large = .35).

Transformed to Cohen’s d for repeated measures designs (Dunlop et al., 1996) the effect

sizes lead to the same conclusion (see table 4). Personality dimensions of the Big Five

remain stable across the two assessments. The reduction in depression score represents a

large and significant change (adj. Cohen’s d = 0.845).

Regressing personality partial change scores on depression partial change scores shows

that personality change is only to a limited extent accounted for changing levels of

depression. A and C are unrelated to changes in depression scores (on the HAMD) (p =

.484 respectively p = .807). Ne (p = .016), E (p < .0001), and O (p = .022) are related to

the Depression Scores and explain 4.6% of variance, 12.2% of variances, and 4.2% of the

variance of the HAMD score respectively. Regressing HAMD change scores on

personality traits change scores resulted in explaining 18.0% of variance with only E

showing significance (p < .001). Change scores of Ne, O, A, and C showed no significant

prediction on depression change scores (p = .071, p = .060, p = .827, and p = .104).

Absolute Continuity or Mean-Level Change for the BDI The same procedure as described above for the HAMD was applied for the BDI. The

descriptive statistics and correlations among the variables are described in table 3.

After correction for unreliability, the FFM before and after treatment is highly correlated,

with correlations of .56, .76, .80, .70, and .88 for Ne, E, O, A and C respectively.

120

Regressing personality partial change scores on BDI partial change scores shows that

personality change is only to an extent accounted for changing levels of the BDI. C is

unrelated to changes in BDI scores (p = .872). Ne (p

121

expected frequencies of 2.5% (reliable decrease), 95% (no change), and 2.5% (reliable

increase) for each of the dimensions scores (Ne: p = .099; E: p = .641; A: p = .187; C: p

= .385). O was not tested, for this dimension was a constant.

Ipsative Continuity D2 correlated .872 with D’2, and .393 with D’’2. D’2 correlated .522 with D’’2. The

distribution of D2 ranged from 5.0 to 398.0, with a mean of 87.014 and a standard

deviation of 77.961. The 25th , 50th, and 75th quartile values were 35.25, 60.00 and

113.75. Only 1.9% had D2 values greater than expected on the basis of chance. The

distribution of D’2 ranged from 3.2 to 297.10, with a mean of 61.399 and a standard

deviation of 63.222. The 25th , 50th, and 75th quartile values were 25.20, 39.40 and 70.70.

None of the participants exceeded levels of significance. The distribution of D’’2 ranged

from -.11 to 3.04, with a mean of .765 and a standard deviation of .767. The 25th , 50th,

and 75th quartile values were .174, .479 and 1.171. Of the participants 12.5% had D’’2

values greater than would be expected by measurement error alone. Consequently, none

of the profile changes reflected changes due to changes in elevation and scatter. Most of

the profile changes reflected changes due to changes in shape.

Discussion Our study is principally a replication of the findings of De Fruyt et al (2006), although

the instrument to measure the big Five used by De Fruyt et al (2006) was different from

the NEO-FFI used in this study. We also extended the findings by comparing results on

the HAMD as well as on the BDI.

As in the aforementioned study, we found that the structure of the Big Five personality

factors remained stable during treatment for depression. Treatment did not have a

significant influence on the intercorrelations of the Big Five.

Participants did, however change over time: Ne and A scores were lower compared to

baseline. Although changes were significant, the absolute change was very small. For

instance, Ne scores went from the 98th percentile down to the 97th percentile, A moved

from the 86th to the 77th percentile and the other three factors did not move at all. As

122

stated by Costa et al (2005) some personality traits are more enhanced during a period of

MDD, but remain essentially the same. Of note is that participants differed substantially

from normative data obtained in the general population (Costa & McCrae, 1992). Since

we did not perform a prospective study we do not know if this reflects a vulnerability for

depression prior to the onset of the episode or a ‘scar’ effect (Akiskal, Hirschfeld et al.,

1983). In one of the few prospective studies (Ormel et al., 2004) Ne was high before the

onset of depression, even higher during an episode of MDD and returned to the original

level upon recovery.

We found that changes in depression scores on the HAMD were relatively independent

from changes in personality traits across the duration of the study. This is completely in

line with the finding by De Fruyt et al (2006) and differs somewhat from other studies in

this area (Du et al., 2002; Santor et al., 1997). The reason for this difference could be due

to the method of administering the measurement (self-rating vs. independent rating). In

our study we both examined a self-rating scale (the BDI) and an expert-rated scale

(HAMD). We did indeed find important differences in correlations between both scales.

Overlap between the HAMD and the NEO-FFI is not strong, with HAMD scores

predicting 18.8% of variance in NEO-FFI scores of which only E, and not Ne, has a

significant relationship with depression change score. Using the BDI, on the other hand,

Ne does have a large and significant relationship with depression change scores (on the

BDI) explaining 30.0% of the variance. Vice versa Ne explains 17.1% of variance in

depression change scores. Compared to the HAMD, the use of the BDI may give serious

distortions which can to a large extent be contributed to differences in personality

characteristics (Enns et al., 2000). It has also been argued that the BDI includes not only

state variables but also trait variables (Groth-Marnat, 1990). If this is true, an overlap

with personality factors is to be expected.

Differential or rank-order stability was also studied. Stability coefficients were high. In

an earlier study by Vaidya, Gray et al. (2002) comparable coefficients were found.

Differences between personality traits were very small. This is noteworthy since Ne is

often found to be less stable than the four other traits (Robins et al., 2001). This finding is

different from that in the De Fruyt et al. (2006) study. The difference could be accounted

123

for by the shorter duration of our study and the somewhat smaller effectsize for reduction

of depression in our study. If Ne is influenced by state effects, one would presume that a

smaller effectsize would mean a smaller change in (especially) Ne scores. Future studies

should take this into account. A second reason for the difference between our study

results and those of De Fruyt et al. (2006) could lie in the relative high mean score on Ne.

As has been found before (Brown, 2007), high initial scores of negative affect tend to

predict less change over time as compared to lower baseline scores. This could point to a

higher trait versus state relationship in these participants.

Individual-level continuity was also found in our study. By far the majority (86.1%) of

participants stayed stable on all five personality factors. Six point five percent showed a

reliable change on one personality factor and only 1.4% of participants showed a reliable

change on two factors. Of the five factors, Ne was the least stable, and O the most stable

factor.

In the same line, ipsative continuity was also found to be high. Only variability in shape

was in 12.5% of participants greater than expected. Variability in scatter and elevation

were either not significant or in a small minority of participants (1.9%) greater than

expected. Again this is in accordance with the earlier study by de Fruyt et al (2006).

In conclusion our study demonstrates that personality factors remain stable during

treatment of depression. Only a small minority of participants exhibit significant changes

in scores on any personality factor. We did however find large and significant differences

between the two instruments most widely used for measuring depression severity, the

HAMD and the BDI, with respect to the association with Ne scores in particular. When

using the BDI as (only) outcome measure this could mean that scores are strongly

influenced by personality characteristics of participants studied. The question remains if

this difference is due to difference in the method of administration (self-rated or expert-

rated) or difference in items of the two instruments. This problem could be solved using

an instrument having both a self-rated as an expert rated version (e.g. Inventory of

Depressive Symptoms, (Rush et al.,1996).

124

Since personality factors do not predict short term outcome in the treatment of depression

(Blom et al., 2007), but more likely influence the long term prognosis of MDD (Du et al.,

2002; Jang, Clay et al., 2004), measurement of personality factors can be relevant for

clinicians in predicting long term vulnerability to the occurrence of relapse in MDD. We

have shown in our study that personality factors can be reliably assessed by the NEO-FFI

during an episode of MDD.

125

126