European Neuropsychopharmacology (2013) 23, 1057–1066

www.elsevier.com/locate/euroneuro

Evaluation of a pharmacist intervention on

patients initiating pharmacological treatment
for depression: A randomized controlled
superiority trial
Maria Rubio-Valeraa,b,n, Marian March Pujolc, Ana Ferna ndeza,b,d,
M. Teresa Peñarrubia-Marı́ab,e, Pere Trave c, Yolanda López del
Hoyob,f, Antoni Serrano-Blancoa,b

a
Research and Development Unit, Fundació Sant Joan de De u, Sant Boi de Llobregat, Barcelona, Spain
b
Red de Investigación en Actividades Preventivas y Promoción de la Salud en Atención Primaria RedIAPP
(RD06/0018/0017), Spain
c
Estades en Pra ctiques Tutelades Unit, School of Pharmacy, University of Barcelona, Barcelona, Spain
d

Clinical and Health Psychology Department, Universitat Autonoma de Barcelona, Barcelona, Spain
e
Primary Care Health Centre Bartomeu Fabres 
 Anglada, DAP Baix Llobregat Litoral, Ambit Costa de
Ponent, Institut Catala de la Salut, Gava, Spain
f
Department of Psychology and Sociology, University of Zaragoza, Spain

Received 2 July 2012; received in revised form 18 September 2012; accepted 13 November 2012

KEYWORDS Abstract
Depressive disorder; Major depression is associated with high burden, disability and costs. Non-adherence limits the
Medication adher- effectiveness of antidepressants. Community pharmacists (CP) are in a privileged position to
ence; help patients cope with antidepressant treatment. The aim of the study was to evaluate the
Antidepressive impact of a CP intervention on primary care patients who had initiated antidepressant
agents;
treatment. Newly diagnosed primary care patients were randomised to usual care (UC) (92)
Pharmaceutical ser-
or pharmacist intervention (87). Patients were followed up at 6 months and evaluated three
vices;
Primary health care times (Baseline, and at 3 and 6 months). Outcome measurements included clinical severity of
depression (PHQ-9), health-related quality of life (HRQOL) (Euroqol-5D) and satisfaction with
pharmacy care. Adherence was continuously registered from the computerised pharmacy
records. Non-adherence was defined as refilling less than 80% of doses or having a medication-
free gap of more than 1 month. Patients in the intervention group were more likely to remain
adherent at 3 and 6 months follow-up but the difference was not statistically significant.
Patients in the intervention group showed greater statistically significant improvement in
HRQOL compared with UC patients both in the main analysis and PP analyses. No statistically
significant differences were observed in clinical symptoms or satisfaction with the pharmacy

n
Correspondence to: Dr Antoni Pujades 42, 08830 Sant Boi de Ll., Barcelona, Spain. Tel.: +34 936406350x12546; fax: +34 685923432.
E-mail addresses: [email protected], [email protected] (M. Rubio-Valera).

0924-977X/$ - see front matter & 2012 Elsevier B.V. and ECNP. All rights reserved.
http://dx.doi.org/10.1016/j.euroneuro.2012.11.006
1058 M. Rubio-Valera et al.

service. The results of our study indicate that a brief intervention in community pharmacies
does not improve depressed patients’ adherence or clinical symptoms. This intervention helped
patients to improve their HRQOL, which is an overall measure of patient status.
& 2012 Elsevier B.V. and ECNP. All rights reserved.

1. Introduction plus CPI. A detailed description of the study protocol has been
provided elsewhere (Rubio-Valera et al., 2009).
Almost 13% of Europeans will suffer major depression at
some point in their life (Alonso et al., 2004). Depressive 2.2. Participant recruitment and randomisation
disorders are associated with considerable disability
(Mathers and Loncar, 2006) and with increased suicide rates Participants were recruited at 4 Primary Care Health Centres
(Angst et al., 1999). This results in a high burden for (PCHC) (30 GPs) from two satellite towns in the Barcelona
patients and society and is costly to the system, mainly metropolitan area (Gava and El Prat) (October 2008-May 2011). At
due to patients’ inability to work (Salvador-Carulla et al., first, only the PCHC from Gava participated in the study but to
accelerate patient inclusion, a population from El Prat was included
2011; Wade and Haring, 2010).
in March 2010. Eligible participants were patients aged between 18
The detection, prevention and treatment of depression and 75 who had been prescribed an antidepressant by a GP due to a
should improve to minimize relapse. Antidepressants depressive disorder. Patients who had taken any antidepressants or
decrease risk of relapse, especially in adherent patients consulted a mental health specialist in the previous 2 months; those
(Geddes et al., 2003). Low adherence to antidepressants with a history of psychotic, bipolar disorder or drug abuse; and
has been systematically reported (Lingam and Scott, 2002) those with cognitive impairment, were not included.
and, in primary care, largely explains low concordance of Spanish patients can choose any pharmacy countrywide to fill
real practice with clinical guidelines for depression (Pinto- their prescription and can switch from one to another in successive
Meza et al., 2008). visits. Patients were asked to refill their antidepressant prescrip-
Community pharmacists (CPs) are easily accessible to tions at the same pharmacy during the study. Those who agreed
were included.
ambulatory patients and can help improve adherence.
GPs invited eligible patients to participate and obtained signed
A systematic review evaluating the effectiveness of informed consent. To ensure allocation concealment, every GP
pharmacist intervention in improving adherence to antide- received a set of 10 sequentially numbered, opaque, sealed
pressants identified six relevant studies (Rubio-Valera envelopes generated by an external investigator (MRV) containing
et al., 2011). Although most of the individual studies had patient assignment. Randomisation was generated at the patient
shown non-statistically significant results, when pooled, a level by a computerised random-number generator following a
statistically significant effect was observed favouring phar- permuted block design (1:1). As patients were enroled, the GP
macist intervention. The review included interventions sequentially stapled one of the envelopes to the prescription.
conducted by pharmacists in hospital services and CPs and When the patient gave the prescription to their CP, the pharma-
sub-group analyses showed that, when pooled separately, cist opened the envelope and created a patient study chart
distinguishing between UC and CPI groups. Interventions performed
CPs studies produced non-statistically significant results.
by the pharmacists, both in the UC and CPI groups, were recorded
This sub-group analysis included only 3 studies, implying on the patient study chart. Blinding of participants and pharmacists
that the power of the meta-analysis to detect differences was not possible but outcome assessors were blind to the allocation.
may be limited. Patients were asked to avoid discussing the study among them.
Only one of the studies had been conducted in a European
country (Brook et al., 2005). In the per protocol (PP)
analysis, Brook found that patients who received a CPI with 2.3. Intervention
an informative videotape showed better antidepressant
All the pharmacies in the towns (39) were invited to participate but
adherence. It is not possible to isolate the relative impact
6 declined; citing heavy workload (n=2) or lack of interest in the
of each intervention component.
study (n=4). To homogenise the intervention, pharmacists received
The aim of the present study was to evaluate the an 8-h training session focused on implementation and information-
effectiveness of a community pharmacist intervention collection guidelines. Only 24 of the participating pharmacies were
(CPI) compared with usual care (UC) in improvement of finally approached by patients and took part in the study (58 CPs).
adherence to antidepressants and patient wellbeing in a Two pharmacies dropped out: one because the pharmacy closed and
population initiating pharmacological treatment following one because the CP responsible for the study left.
diagnosis of depression by their general practitioner (GP). Patients received the CPI on visiting the pharmacy where they
received their first prescription of the 6-months antidepressant
course. The CPI consisted of an educational intervention centred on
improving patients’ knowledge of antidepressants and awareness of
the importance of adherence. In patients with a sceptical attitude
2. Experimental procedures
towards the medication, the intervention aimed to reduce stigma,
reassure the patient about possible side-effects, and stress the
2.1. Study design importance of following GPs’ advice.
As patients were beginning treatment with antidepressants, the
This was a six-months follow-up naturalistic parallel-group con- first contact was considered crucial. During the first visit,
trolled trial with random allocation of participants into UC and UC the pharmacist provided the patient with information about the
Pharmacist intervention on patients initiating pharmacological treatment for depression 1059

medicine and briefly discussed various aspects of the illness to observations in longitudinal studies, it has been suggested that
improve understanding of the treatment, eliminate erroneous applying multilevel analysis is a good option. Multilevel analysis is
preconceptions and reinforce the concept of illness to the patient. very flexible in handling missing data and it has been shown that it
In subsequent visits, the pharmacist conducted a short review of is better to apply multilevel analysis to an incomplete dataset than
some points covered in the first visit and checked patient progress applying imputation methods (Twisk, 2006). Consequently, missing
(improvement, appearance of side-effects, or queries). First and data was not imputed.
subsequent contacts took a mean of 14.4 and 7.7 min, respectively. A second analysis was conducted according to the PP principle.
Control patients received UC from their GP and CP. UC varied Participants in the intervention group were excluded if they had
from one pharmacy to another but mainly consisted of dispensing never received the pharmacist intervention (never bought medica-
the medication; answering patients’ questions and giving some tion or did so at a non-participating pharmacy). Participants in the
basic advice about how to take the medication. First and subse- control group that never bought medication (i.e. did not receive
quent visits took a mean of 7.8 and 7.7 min, respectively. usual pharmaceutical care) were also excluded from the PP
analyses.
2.4. Measurements The models were fitted using Restricted Maximum Likelihood. To
account for correlation among several observations for each sub-
ject, an unstructured correlation matrix was used. In all models the
Three assessments (baseline, 3 and 6 months) were conducted by
gender and the interaction term ‘time  group’ were included in
8 trained psychologists.
the model as covariates. When the interaction was significant in the
Adherence to antidepressants was assessed using the compu-
model, the effect of the intervention was considered to vary during
terised pharmacy records that registered all the information about
the course of the study (HRQOL models only). When this interaction
medication in the patient’s clinical history at the time of purchase.
term was not significant, the model without the interaction term
Non-adherence was defined as refilling o80% of the prescribed
was used.
doses; a definition that has a reasonable balance between sensitiv-
Other sociodemographic and clinical characteristics that could
ity and specificity (Hansen et al., 2009) or having a treatment gap
plausibly affect the outcome were tested using a likelihood ratio
41 month (Peterson et al., 2007).
test (LR-test). We compared the models with and without these
Clinical severity of depression was measured with the Patient
variables and included them if the LR-test was positive (pr0.10).
Health Questionnaire, 9-item depression module (PHQ-9) (Kroenke
Number needed to treat (NNT) was calculated for the main outcome
et al., 2001; Spitzer et al., 1999).
(adherence) by computing the inverse of the differences between
HRQOL was evaluated using the EuroQol-5D (EQ-5D) and Spanish
groups in the probability of being adherent. For the continuous
tariffs or utility indexes were calculated (Badia et al., 1998, 1999;
outcome variables showing statistically significant differences
Dolan et al., 1995; The EuroQol Group, 1990).
between groups, effect size (Cohen’s d) was calculated by means
Satisfaction with the pharmacist service was measured with a
of standardised mean difference between the two populations using
patient-satisfaction questionnaire (Armando et al., 2008).
the pooled standard deviation of the two groups at baseline. The
During recruitment, clinical diagnosis was made by the GP and,
effect size was categorised as small (0.2), medium (0.5) and large
at the baseline assessment, was confirmed using the research
(0.8) (Cohen, 1988).
version of the Structured Clinical Interview for DSM-IV Axis I
All analyses were conducted with STATA 11.0.
Disorders (SCID-I) (First et al., 1996, 1999). GPs were blind to the
DSM-IV diagnosis and patient inclusion was performed according to
their usual practice.
Chronic physical conditions were assessed using a ‘‘yes’’ or ‘‘no’’
check-list. 3. Results

2.5. Sample size calculation and data analysis 3.1. Participants and drop-outs

Sample size calculation was based on the main study objective (i.e. Figure 1 shows the study flow chart. A total of 234 patients
improving adherence to antidepressants). The sample size was were referred by the GPs. Finally, 179 patients were
calculated for an expected difference between groups of at least randomised to UC (92) and CPI (87), were evaluated at
17 points in the percentage of medication intake, which is in baseline and included in the main analysis. Only 87 (95%)
agreement with the study by Brook et al. (2005). It was estimated and 64 (74%) in the control and intervention group, respec-
that a sample of 162 patients would have a power of 80% at a
tively, received the intervention as allocated and were
significance level of 5% to detect these differences. There were no
missing values for our main outcome.
included in the PP analysis.
Pre-treatment comparability between groups was assessed All assessment visits were completed by 120 (67%)
applying the w2-test or Fisher exact test for categorical data, the patients. Nineteen (11%) participants were only evaluated
Student’s t-test for continuous variables and the non-parametric at baseline. Forty (23%) patients missed 1 follow-up assess-
equality-of-medians test for biased numerical data. ment (7 at 3-months and 33 at 6-months) because they
To evaluate intervention effectiveness, multilevel mixed-effects could not be contacted or refused to attend.
linear and logistic models were fitted that allowed the inclusion of
all available data. A two-level longitudinal multilevel structure was
used where observations were clustered within patients. The
models predict treatment response using group as a fixed factor, 3.2. Baseline data
time point (baseline, 3 and 6 months) as a within-participants
repeated factor, and participants as a random factor. Models with
variables not assessed at baseline (adherence and satisfaction)
Table 1 shows the participants’ baseline characteristics.
included only two time points. Most participants were women (75.4%), with mean age of
For the main analyses, all participants were included as rando- 46.6 years. Fifty-one per cent of the participants met DSM-IV
mised regardless of whether they received the intervention or had criteria for major depression. The mean baseline severity of
incomplete follow-up data. To deal with the problem of missing depression (PHQ-9) was 15.9 (moderately severe symptoms).
1060 M. Rubio-Valera et al.

Assessed for eligibility (n=234)

Excluded (n=55)
Not meeting inclusion criteria (n=11)
Substance abuse (n=6)
Age over 75 (n=1)
Psychotic or bipolar disorder (n=2)
ADs had not been prescribed (n=1)
Consulting a psychologist in the past 2 months(n=1)
Declined to participate (n=44)

Randomized (n=179)

Allocated to intervention group (n=87) Allocated to control group (n=92)

Received allocated intervention (n=64) Received usual care as intended (n=87)
Did not receive allocated intervention (n=23) Did not receive usual care as intended (n=5)
(decided not to take antidepressants) (n=6) (decided not to take antidepressants) (n=5)
(not identified as study participants when
boughtthe medication) (n=17)

Follow-up

Lost to follow-up (n=26) Lost to follow-up (n= 26)

Refused to attend to follow-up interview (n=16) Refused to attend to follow-up interview (n=16)
Were unable to contact (n=10) Were unable to contact (n=10)

Analysis

Analysed by intent-to-treat (n=87) Analysed by intent-to-treat (n=92)

Analysed per protocol (n=64) Analysed per protocol (n=87)

Figure 1 Study flow chart.

Differences existed in the proportion of women between 3.4. Satisfaction with pharmacy service
groups; all analyses were adjusted for gender.
Overall, patient satisfaction with the pharmacy service was
high in both groups. No differences were observed between
groups at 3 or 6 months (Tables 2 and 3).
3.3. Adherence to antidepressants
3.5. Clinical severity of depression
Table 2 shows the patients’ probability of remaining adher-
ent, models-based mean satisfaction, severity of depression
Both groups showed an improvement in symptoms at 3 and
and HRQOL. Table 3 shows the regression models for
6 months (Table 2). Table 4 shows the depression severity
adherence and satisfaction.
regression models (PHQ-9) and HRQOL. No differences in
Eleven (6%) patients never bought medication (non-
symptom severity were observed between groups (Table 4).
initiators) and a high proportion of patients discontinued
at 3 (48.0%) and 6-months follow-up (57.0%).
In the main analyses, CPI group patients seemed more 3.6. HRQOL
likely to remain adherent both at 3- (67.7% vs. 83.3%) and
6-months (46.3% vs. 67.3%) follow-up (Table 2) but the trend Figure 2 shows the multilevel-based mean utilities (EQ-5D
did not reach statistical significance (OR =2.24; p=0.209) tariffs) and the improvement in HRQOL in the control and
(Table 3). intervention groups.
In the per PP analysis, the same trend was observed In both analyses, a significant time  group interaction
(Table 2) and differences between groups were close to was found in EQ-5D tariffs in favour of the intervention
statistical significance but did not reach it (OR= 3.44; group (Table 4). Overall improvement was higher in the
p=0.055) (Table 3). NNT was 5, indicating that to prevent intervention group in both the main (0.25 vs. 0.14) and PP
non-adherence in one patient, we needed to implement the (0.27 vs. 0.16) analyses. The effect size was small to
intervention in 5 (Table 2). medium in both analyses (0.31 and 0.33 respectively).
Pharmacist intervention on patients initiating pharmacological treatment for depression 1061

Table 1 Sample socio-demographic and clinical baseline characteristics.

Usual care (n = 92) Pharmacist’s intervention (n= 87) p-Value

Gender (% women (n)) 83.7% (77) 66.7% (58) 0.008

Age (mean (95% CI)) 46.3 (43.3, 49.2) 46.9 (44.0, 48.6) 0.742

Marital status (% (n)) 0.881

Never married 14.1% (13) 18.4% (16)
Married or living with someone 64.1% (59) 59.8% (52)
Previously married 10.9% (10) 10.3% (9)
Widow 10.9% (10) 11.5% (10)

Education(% (n)) 0.676

No studies 7.6% (7) 5.8% (5)
Primary 22.8% (21) 23.0% (20)
Graduated 23.9% (22) 19.5% (17)
Secondary 26.1% (24) 31.0% (27)
University 19.6% (18) 19.0% (34)
Other – 2.3% (2)

Working status(% (n)) 0.493

Househusband/housewife 13.0% (12) 17.2% (15)
Paid employment 40.2% (37) 29.9% (26)
Paid employment but on sick leave 21.7% (20) 24.1% (21)
Unemployed 17.4% (16) 16.1% (14)
Retired 7.6% (7) 9.2% (8)
Other – 2.3% (2)
NS/NC (missing) 1.2% (1)

Clinical severity according to PHQ-9(mean (95% CI)) 15.8 (14.6, 16.9) 16.1 (14.7, 17.4) 0.776
Number of co-morbidities 37.0% (34) 40.2% (35) 0.653
(% of cases over the median (median = 3) (n))

Table 2 Multilevel model-based probabilities of remaining adherent and multilevel model-based mean satisfaction and
severity of depression at 3 and 6 months follow-up in the control and intervention groups for the main analysis and
PP analyses.

Main analysis PP

Baseline 3-Months 6-Months Baseline 3-Months 6-Months

Probability of remaining adherent (95% CI) and number needed to treat (NNT)a
Usual care NA 61.9% 40.2% NA 43.8% 25.7%
(26.4, 88.1) (12.9, 75.3) (15.7, 76.5) (7.4, 59.8)
Intervention NA 78.4% 60.1% NA 72.9% 54.4%
(48.0, 93.5) (28.4, 85.11) (41.7, 91.0) (24.6, 81.4)
NNT 6.1 5.0 3.4 3.5

Mean satisfaction (95% CI)b

Usual care NA 38.3 39.0 NA 37.5 37.9
(32.8, 43.8) (33.4, 44.5) (31.4, 43.6) (31.8, 44.0)
Intervention NA 40.1 40.8 NA 39.2 39.6
(35.1, 45.1) (35.7, 45.8) (33.4, 44.9) (33.8, 45.4)

Mean severity of depression (95% CI)c

Usual care 14.0 6.8 5.0 14.0 7.1 5.1
(12.3, 15.6) (5.2, 8.5) (3.2, 6.7) (12.3, 15.8) (5.2, 8.9) (3.2, 7.0)
Intervention 14.5 7.4 5.5 14.8 7.8 5.9
(13.0, 15.9) (5.8, 8.9) (3.9, 7.1) (13.1, 16.4) (6.1, 9.5) (4.1, 7.6)

NA= not applicable.

a
Values for male patients of age 45.5 with a baseline severity of depressive symptoms of 16 (moderately-severe symptoms).
b
Values for male patients of age 45.5,never married and without comorbidities.
c
Values for male patients of age 45.5.
1062 M. Rubio-Valera et al.

Table 3 Multilevel model-based odds ratio (95% confidence interval) and p-values of the variables included in the models for
adherence to antidepressants.

Adherence to antidepressants Satisfaction with the pharmacy service

(odds ratio (95% CI) and p-value) (b coefficients (95% CI) and p-value)

Main analysis PP analysis Main analysis PP analysis

Constanta 1.63 (0.36, 7.37) 0.529 0.78 (0.19, 3.26) 0.734 38.3 (32.8, 43.8) 0.001 37.5 (31.4, 43.6) 0.001
Group
Control Reference Reference Reference Reference
Intervention 2.24 (0.64, 8.60) 0.209 3.44 (0.97, 12.22) 0.055 1.8 ( 0.9, 4.5) 0.20 1.7 ( 1.3, 4.7) 0.270

Gender
Men Reference Reference Reference Reference
Women 0.37 (0.08, 1.63) 0.188 1.21 (0.29, 4.97) 0.796 1.2 ( 4.6, 2.2) 0.48 1.8 ( 5.6, 1.9) 0.339
b
Age 1.06 (1.01, 1.11) 0.013 1.04 (1.00, 1.09) 0.070 0.03 ( 0.1, 0.2) 0.70 0.02 ( 0.2, 0.1) 0.849

Time
3-Months Reference Reference Reference Reference
6-Months 0.41 (0.21, 0.83) 0.012 0.44 (0.22, 0.90) 0.024 0.7 ( 1.1,2.5) 0.45 0.4 ( 1.6, 2.4) 0.673

Depression baseline 0.99 (0.89, 1.11) 0.911 0.98 (0.88, 1.09) 0.708 ni ni
severity(PHQ9)b

Comorbidities ni ni 0.9 (0.3, 1.5) 0.01 0.8 ( 5.6, 1.9) 0.339

Marital status
Never married ni ni Reference Reference
Married ni ni 2.7 ( 1.7, 7.2) 0.23 4.8 ( 0.3, 9.9) 0.065
Divorced ni ni 2.7 ( 3.0, 8.5) 0.35 4.5 ( 2.0, 11.0) 0.178
Widow ni ni 5.1 ( 11.9, 1.7) 0.14 2.4 ( 9.8, 5.0) 0.529

ni=Variables not included in the model (negative LR-test).

a
Constant or reference value corresponds to male patients of age 45.5 in the control group at baseline and with a baseline severity
of depressive symptoms of 16 (moderately-severe symptoms) in the model for adherence and to never-married male patients of age
45.5 without comorbidities in the control group at baseline in the model for satisfaction.
b
Centered in the median. One-year or 1-point increase.

4. Discussion received a minimum of 3 pharmacist contact sessions. Never-

theless, those patients who received one (or two) interventions
CPI group patients tend to have a higher probability of before abandoning the medication may not have wanted to
remaining adherent at 3 and 6 months than those receiving receive a second (or third) intervention session. Consequently,
UC. In the PP analysis, this result did not quite reach intervention group patients who abandoned the medication
statistical significance (p=0.055). However, the difference early may have been excluded from the PP analysis so
was clinically relevant since the NNT was relatively small for increasing the difference between groups. As such, there may
a fairly quick, easy-to-implement intervention (intervention be some difficulty in generalizing from this result.
implementation needed in 4 patients in order to help one In our case, first patient contact was crucial and
extra patient to remain adherent at 3 and 6-months follow- demanded a more flexible protocol stipulation.
up). One possible reason for not achieving statistical Despite having detected statistically significant differ-
significance would be a lack of statistical power. The large ences in the degree of adherence to antidepressants, none
amount of drop-outs and patients not following study of the previous studies found differences in clinical
protocol could have reduced the study power. Furthermore, improvement (Adler et al., 2004; Brook et al., 2005;
we did not take into account the clustering effect of the Capoccia et al., 2004; Finley et al., 2003; Rickles et al.,
multilevel mixed-effects analyses when the sample size was 2006). However, a powerful meta-regression based on
calculated, which could have limited our capacity to detect collaborative care in depression showed a positive associa-
differences between groups. tion between improved adherence and improvement in
In general, our study results are consistent with those of depressive symptomatology (Bower et al., 2006). The lack
Brook 2005 and Rickles 2006. In the intent-to-treat analysis, of difference in clinical improvement could imply that its
between-group differences were not found although these relationship with adherence is not as direct as it may appear
studies did observe statistically significant adherence differ- and is affected by diverse factors such as pharmacological
ences between groups in the PP analysis. Both studies used a efficacy or other environmental or social elements. Another
protocol that considered exclusively those patients that had factor could be diagnostic accuracy, as only half the
Pharmacist intervention on patients initiating pharmacological treatment for depression 1063

Table 4 Multilevel model based b-coefficients (95% confidence interval) and p-values of the variables included in the models
for clinical severity of depression and health-related quality of life.

Severity of depressive symptoms Health related quality of life

(PHQ-9) (EuroQol-5D tariffs)

Main analysis PP analysis Main analysis PP analysis

Constanta 13.95 0.001 14.0 0.001 0.67 0.001 0.66 0.001

(12.33, 15.58) (12.27, 15.78) (0.59, 0.74) (0.58, 0.75)

Group
Usual care Reference Reference Reference Reference
Intervention 0.51 0.432 0.77 0.297 0.061 0.108 0.09 0.038
( 0.77, 1.79) ( 0.67, 2.21) ( 0.14, 0.01) ( 0.17, 0.01)

Gender
Male Reference Reference Reference Reference
Female 2.37 0.002 2.19 0.011 0.031 0.386 0.031 0.438
(0.85, 3.89) (0.49, 3.89) ( 0.10, 0.04) ( 0.11, 0.05)

Ageb 0.04 0.067 0.03 0.237 0.003 0.008 0.003 0.005

( 0.09, 0.003) ( 0.08, 0.02) ( 0.01, 0.00) ( 0.01, 0.00)

Time
Baseline Reference Reference Reference Reference
3-Months 7.12 0.001 7.0 0.001 0.133 0.001 0.13 0.001
( 8.21, 6.03) ( 8.1, 5.8) (0.07–0.20) (0.07, 0.20)
6-Months 9.00 0.001 8.9 0.001 0.142 0.001 0.16 0.001
( 10.17, 7.80) ( 10.2, 7.7) (0.08–0.21) (0.09, 0.23)

Depression baseline ni ni 0.012 0.001 0.01 0.001

severity (PHQ-9)b ( 0.02– 0.01) ( 0.02, 0.01)

Time  groupinteraction
Baseline Reference Reference
Intervention group at ni ni 0.06 0.204 0.07 0.145
3-months ( 0.03, 0.15) ( 0.03, 0.17)
Intervention group at 6-months ni ni 0.10 0.034 0.11 0.042
(0.01, 0.20) (0.004, 0.22)

ni=Variables not included in the model.

a
Constant or reference value corresponds to male patients of age 45.5 in the control group at baseline in the PHQ-9 model and to
male patients of age 45.5 in the control group and with a baseline severity of depressive symptoms of 16 (moderately-severe
symptoms) in the EuroQol-5D model.
b
Centered in the median. One-year or 1-point increase.

patients met major depression SCID criteria and antidepres- bias. Nevertheless, both groups showed very high levels of
sants have only shown an effect on moderate to severe pharmacy-service satisfaction with no statistically signifi-
major depression. This could explain the lack of correlation cant differences between groups. As part of the interven-
between adherence and clinical outcomes. The analysis tion, the pharmacist discussed the nature of the treatment
performed on the major-depression patient subsample and illness with the patient. This may have helped the
(SCID) showed no statistically significant differences patient to cope with the new diagnosis, reducing stigmati-
between groups (data available on request), although this zation and even, in some cases, modifying inappropriate
study was not designed to observe differences in this health beliefs. This could manifest itself as an improvement
population and the power of the study was insufficient to in self-perception with respect to HRQOL (the constructs of
draw conclusions on subsample behaviour. quality of mental life).
In contrast to Capoccia et al. (2004), statistically sig- Although not directly related to intervention, we
nificant HRQOL differences between groups were observed, observed a high proportion of non-initiators (around 6%).
indicating that patients who received extra pharmacy care These patients had agreed to participate in a study to
perceived improved HRQOL. Effect size was small to mod- improve the use of antidepressants and, as such, we
erate, which led us to question the clinical relevance of this concluded that the proportion of non-initiators would
difference. It could be due to placebo effect or desirability be much higher in normal practice. In previous studies,
1064 M. Rubio-Valera et al.

Figure 2 Multilevel based mean utility and overall improvement in the EQ-5D (95% CI) at 3 and 6 months follow-up for the main and
PP analyses.

non-initiation rates reached 15% (Bull et al., 2002). We Fourth, only 74% of intervention group patients received
consider that non-initiators motives require detailed study. at least one pharmacy intervention and this may have
This study had a number of limitations. Firstly, enrolment limited its impact and affected the power to detect
may have been biased against patients unwilling to take differences.
antidepressants. However, the figures regarding treatment Fifth, patients could change pharmacy in successive
discontinuation correspond to those found previously in visits. Consequently, even those patients who received the
Catalonia (Serna et al., 2010) and we conclude that our intervention attended very few sessions. However, this
sample can be extrapolated to the primary care population. leads us to believe that even with a single, relatively
Secondly, inclusion criteria were very broad which may simple, although slightly more intense, intervention applied
have created great variability among subjects, although this at the point of initiating the medication, we can obtain
did favour generalisation of the results and the study’s significant improvements in adherence and patients’
external validity. HRQOL; although this would require further exploration.
Third, the pharmacists attended both UC and CPI patients, Sixth, as a result of the financial crisis, shortly after study
which could have led to some contamination. This could have commencement, a series of economic adjustments were
been prevented by performing a cluster randomisation at the made which affected the viability of pharmacies in Catalo-
pharmacy level. The pharmacists were asked to exercise great nia (Spanish Resolution, 2008, 2009, 2010). In addition, the
care and to register the intervention carried out on control low incidence of new cases meant that the inclusion period
patients. Also, although they were asked not to share had to be extended. These two factors, taken together, may
information with other participants, patients could have have demotivated and/or exhausted our pharmacists. This
transmitted information among themselves. may be reflected in the results although the pharmacists
Pharmacist intervention on patients initiating pharmacological treatment for depression 1065

recorded the interventions carried out and, as such, we and completion of the study. MRV and AF designed the statistical
believe that the impact was minimal. analyses and MRV undertook them. MRV wrote the first draft of the
Finally, the use of pharmacy registers as a measure of manuscript. All authors have corrected draft versions and approved
adherence involves a series of limitations. Patients may the final version of the manuscript.
acquire the tablets but not take them and this measure does
not provide us with information with respect to the time of Conflict of interest
taking the medication or reasons for non-adherence. How-
ever, it showed relatively good agreement with electronic All authors declare that they have no conflicts of interest.
pill container, especially in depressed patients (Hansen
et al., 2009). Moreover, this measure allowed us to collect Acknowledgements
information without the patient being aware that he or she
was being assessed even when the patients did not keep We are grateful to the patients, community pharmacists and
their evaluation appointments. Consequently, we had no general practitioners from Gava and El Prat the Llobregat for their
missing data in our main study variable. selfless participation in the study.
Despite all these limitations, this study is the first We thank redIAPP ‘‘Red de Investigación en Actividades Preven-
performed in Europe which focuses specifically on a com- tivas y Promoción de la Salud’’ (Research Network on Preventative
munity pharmaceutical intervention to improve adherence Activities and Health Promotion) (RD06/0018/0017) for its support
to antidepressants. In addition, it represents the largest in the development of this study.
study sample of patients undertaking a community phar-
macy intervention. In spite of being low, adherence to the References
protocol is higher than that reported in previous studies
(Brook et al., 2002). Finally, the naturalistic nature of the Adler, D.A., Bungay, K.M., Wilson, I.B., Pei, Y., Supran, S., Peckham, E.,
study design benefited the results external validity. Cynn, D.J., Rogers, W.H., 2004. The impact of a pharmacist
The study results indicate that a brief intervention in intervention on 6-month outcomes in depressed primary care
community pharmacy is not effective in improving patients’ patients. Gen. Hosp. Psychiatry 26 (3), 199–209 (available from:
PM:15121348).
adherence to antidepressants or clinical symptomatology.
Alonso, J., Angermeyer, M.C., Bernert, S., Bruffaerts, R., Brugha,
Though not statistically significant, there was a clinically
T.S., Bryson, H., De Girolamo, G., Graaf, R., Demyttenaere, K.,
important improvement in the degree of adherence in the Gasquet, I., Haro, J.M., Katz, S.J., Kessler, R.C., Kovess, V.,
intervention group. Furthermore, this type of intervention Lepine, J.P., Ormel, J., Polidori, G., Russo, L.J., Vilagut, G.,
does help patients with a new depressive episode to improve Almansa, J., Arbabzadeh-Bouchez, S., Autonell, J., Bernal, M.,
their HRQOL. As such, we believe that further studies are Buist-Bouwman, M.A., Codony, M., Domingo-Salvany, A., Ferrer,
required to investigate the pharmaceutical intervention’s M., Joo, S.S., Martinez-Alonso, M., Matschinger, H., Mazzi, F.,
active components with the aim of increasing the impact on Morgan, Z., Morosini, P., Palacin, C., Romera, B., Taub, N.,
improvements in quality of life. The greatest limitation on the Vollebergh, W.A., 2004. Prevalence of mental disorders in
CPI was the lack of continuity in the service. We would Europe: results from the European Study of the Epidemiology
of Mental Disorders (ESEMeD) project. Acta Psychiatr. Scand.
recommend designing a single but more intensive intervention
Suppl. 109 (420), 21 (available from: PM:15128384).
to be applied at the beginning of the treatment, making a
Angst, J., Angst, F., Stassen, H.H., 1999. Suicide risk in patients
greater effort to attempt to modify patients’ health concepts with major depressive disorder. J. Clin. Psychiatry 60 (Suppl. 2),
and beliefs about the treatment and the disease. Motivations 57–62 (discussion 75–76, 113–116, 57–62).
for non-initiation of the treatment with antidepressants should Armando, P.D., Martinez Sr., P., Marti, P.M., Sola Uthurry, N.H., Faus
be assessed in order to develop interventions that may be Dader, M.J., 2008. Development and validation of a Spanish
helpful in the recovery of these patients. language patient satisfaction questionnaire with drug dispen-
sing. Pharm. World Sci. 30 (2), 169–174 (available from:
PM:17885819).
Role of funding source Badia, X., Roset, M., Montserrat, S., Herdman, M., Segura, A., 1999.
The Spanish version of EuroQol: a description and its applica-
Funding for this study was provided by Carlos III Health Institute tions. European Quality of Life scale. Med. Clin. (Barc.) 112
Grant (Spanish Ministry of Health and Consumer Affairs) (FIS (Suppl. 1), 79–85.
PI070546); the Carlos III Health Institute had no further role in Badia, X., Schiaffino, A., Alonso, J., Herdman, M., 1998. Using the
study design; in the collection, analysis and interpretation of data; EuroQoI 5-D in the Catalan general population: feasibility and
in the writing of the report; and in the decision to submit the paper construct validity. Qual. Life Res. 7 (4), 311–322.
for publication. AF is grateful to the ‘‘Ministerio de Sanidad, Polı́tica Bower, P., Gilbody, S., Richards, D., Fletcher, J., Sutton, A., 2006.
Social e Igualdad, Instituto de Salud Carlos III’’ (Red RD06/0018/ Collaborative care for depression in primary care. Making sense

0017) for a predoctoral contract. MRV is grateful to the ‘‘Agencia de of a complex intervention: systematic review and meta-
Gestió d’Ajuts Universitaris i de Recerca’’ for a predoctoral grant regression. Br. J Psychiatry 189, 484–493 (available from:
(FI-DGR 2011). PM:17139031).
Brook, O., van Hout, H., de Haan, M., Nieuwenhuyse, H., 2002.
Pharmacist coaching of antidepressant users, effects on adher-
Contributors ence, depressive symptoms and drug attitude; a randomized
controlled trial. Eur. Neuropsychopharmacol. 12 (3), S201–S202
ASB is the principal investigator and developed the original idea for (available from: ISI:000178905000201).
the research. The study design was done by ASB, MM, MTP and PT. Brook, O.H., van Hout, H., Stalman, W., Nieuwenhuyse, H., Bakker, B.,
These authors, with the help of MRV and YLH, designed and planned Heerdink, E., de Haan, M., 2005. A pharmacy-based coaching
the intervention. All authors have participated in the implementation program to improve adherence to antidepressant treatment among
1066 M. Rubio-Valera et al.

primary care patients. Psychiatr. Serv. 56 (4), 487–489 (available Pinto-Meza, A., Fernandez, A., Serrano-Blanco, A., Haro, J.M.,
from: PM:15812103). 2008. Adequacy of antidepressant treatment in Spanish primary
Bull, S.A., Hu, X.H., Hunkeler, E.M., Lee, J.Y., Ming, E.E., Markson, care: a naturalistic six-month follow-up study. Psychiatr. Serv. 59
L.E., Fireman, B., 2002. Discontinuation of use and switching of (1), 78–83.
antidepressants: influence of patient–physician communication. Rickles, N.M., Svarstad, B.L., Statz-Paynter, J.L., Taylor, L.V.,
J. Am. Med. Assoc. 288 (11), 1403–1409 (available from: Kobak, K.A., 2006. Improving patient feedback about and out-
PM:12234237). comes with antidepressant treatment: a study in eight commu-
Capoccia, K.L., Boudreau, D.M., Blough, D.K., Ellsworth, A.J., nity pharmacies. J. Am. Pharm. Assoc. (2003) 46 (1), 25–32
Clark, D.R., Stevens, N.G., Katon, W.J., Sullivan, S.D., 2004. (available from: PM:16529338).
Randomized trial of pharmacist interventions to improve depres- Rubio-Valera, M., Serrano-Blanco, A., Magdalena-Belio, J., Fernan-
sion care and outcomes in primary care. Am. J. Health Syst. dez, A., Garcia-Campayo, J., Pujol, M.M., del Hoyo, Y.L., 2011.
Pharm. 61 (4), 364–372 (available from: PM:15011764). Effectiveness of pharmacist care in the improvement of adher-
Cohen, J., 1988. Statistical Power Analyses for the Behavioral ence to antidepressants: a systematic review and meta-analysis.
Sciences. Erlbaum, Hillsdale, New Jersey. Ann. Pharmacother. 45 (1), 39–48.
Dolan, P., Gudex, C., Kind, P., Williams, A., 1995. A Social Tariff for Rubio-Valera, M., Serrano-Blanco, A., Trave, P., Penarrubia-Maria,
EuroQol: Results From a UK General Population Survey. Discussion M.T., Ruiz, M., Pujol, M.M., 2009. Community pharmacist
Paper 138. Centre for Health Economics, University of York. . intervention in depressed primary care patients (PRODEFAR
Finley, P.R., Rens, H.R., Pont, J.T., Gess, S.L., Louie, C., Bull, S.A., study): randomized controlled trial protocol. BMC Public Health
Lee, J.Y., Bero, L.A., 2003. Impact of a collaborative care model 9, 284.
on depression in a primary care setting: a randomized controlled Salvador-Carulla, L., Bendeck, M., Fernandez, A., Alberti, C.,
trial. Pharmacotherapy 23 (9), 1175–1185 (available from: Sabes-Figuera, R., Molina, C., Knapp, M., 2011. Costs of
PM:14524649). depression in Catalonia (Spain). J. Affect. Disord. 132 (1–2),
First, M.B., Spitzer, R.L., Gibbon, M., William, J.B., 1996. Struc- 130–138.
tured Clinical Interview Axis I DSM-IV Disorders, Research Version Serna, M.C., Cruz, I., Real, J., Gasco, E., Galvan, L., 2010. Duration
(SCID-RV). American Psychiatric Press, Inc., Washington. and adherence of antidepressant treatment (2003 to 2007)
First, M.B., Spitzer, R.L., Gibbon, M., William, J.B., 1999. Entre- based on prescription database. Eur. Psychiatry 25 (4), 206–213.
vista clı́nica estructurada para los trastornos del eje I del DSM-IV. Spanish Resolution, 18 February 2008. /http://www.boe.es/boe/
Masson ed., Barcelona, Spain. dias/2008/02/18/pdfs/A09002-09003.pdfS.
Geddes, J.R., Carney, S.M., Davies, C., Furukawa, T.A., Kupfer, Spanish Resolution, 16 February 2009. /http://www.boe.es/boe/
D.J., Frank, E., Goodwin, G.M., 2003. Relapse prevention with dias/2009/02/19/pdfs/BOE-A-2009-2874.pdfS.
antidepressant drug treatment in depressive disorders: a sys- Spanish Resolution, 27 February 2010. /http://www.boe.es/boe/
tematic review. Lancet 361 (9358), 653–661. dias/2010/02/27/pdfs/BOE-A-2010-3231.pdfS.
Hansen, R.A., Kim, M.M., Song, L., Tu, W., Wu, J., Murray, M.D., Spitzer, R.L., Kroenke, K., Williams, J.B., 1999. Validation and
2009. Comparison of methods to assess medication adherence utility of a self-report version of PRIME-MD: the PHQ primary
and classify nonadherence. Ann. Pharmacother. 43 (3), 413–422. care study. Primary care evaluation of mental disorders. Patient
Kroenke, K., Spitzer, R.L., Williams, J.B., 2001. The PHQ-9: validity health questionnaire. J. Am. Med. Assoc. 282 (18), 1737–1744.
of a brief depression severity measure. J. Gen. Intern. Med. 16 The EuroQol Group, 1990. EuroQol—a new facility for the measure-
(9), 606–613. ment of health-related quality of life. Health Policy 16 (3),
Lingam, R., Scott, J., 2002. Treatment non-adherence in affective 199–208.
disorders. Acta Psychiatr. Scand. 105 (3), 164–172. Twisk, J.W.R., 2006. Applied Multilevel Analysis: A Practical Guide.
Mathers, C.D., Loncar, D., 2006. Projections of global mortality and Cambridge University Press, Cambridge.
burden of disease from 2002 to 2030. PLoS Med. 3 (11), e442 Wade, A.G., Haring, J., 2010. A review of the costs associated with
(available from: PM:17132052). depression and treatment noncompliance: the potential benefits
Peterson, A.M., Nau, D.P., Cramer, J.A., Benner, J., Gwadry-Sridhar, of online support. Int. Clin. Psychopharmacol. 25 (5), 288–296.
F., Nichol, M., 2007. A checklist for medication compliance and
persistence studies using retrospective databases. Value Health
10 (1), 3–12.