Content uploaded by Ulrika K. Stigsdotter
Author content
All content in this area was uploaded by Ulrika K. Stigsdotter on Sep 17, 2019
Content may be subject to copyright.
Content uploaded by Ulrika K. Stigsdotter
Author content
All content in this area was uploaded by Ulrika K. Stigsdotter on Sep 17, 2019
Content may be subject to copyright.
Efficacy of nature-based therapy for
individuals with stress-related
illnesses: randomised controlled trial
Ulrika Karlsson Stigsdotter, Sus Sola Corazon, Ulrik Sidenius, Patrik Karlsson Nyed, Helmer Bøving Larsen
and Lone Overby Fjorback
Background
Stress-related illnesses are a major threat to public health, and
there is increasing demand for validated treatments.
Aims
To test the efficacy of nature-based therapy (NBT) for patients
with stress-related illnesses.
Method
Randomised controlled trial (ClinicalTrials.gov ID NCT01849718)
comparing Nacadia
®
NBT (NNBT) with the cognitive–behavioural
therapy known as Specialised Treatment for Severe Bodily
Distress Syndromes (STreSS). In total, 84 participants were ran-
domly allocated to one of the two treatments. The primary out-
come measure was the mean aggregate score on the
Psychological General Well-Being Index (PGWBI).
Results
Both treatments resulted in a significant increase in the PGWBI
(primary outcome) and a decrease in burnout (the Shirom–
Melamed Burnout Questionnaire, secondary outcome), which
were both sustained 12 months later. No significant difference in
efficacy was found between NNBT and STreSS for primary out-
come and secondary outcomes.
Conclusions
The study showed no statistical evidence of a difference
between NNBT and STreSS for treating patients with stress-
related illnesses.
Declaration of interest
None.
Copyright and usage
© The Royal College of Psychiatrists 2018.
Stress-related illnesses are a major threat to public health in the
Western world.
1
In Denmark, it has been estimated that stress-
related illnesses are linked to 300 000 admissions to hospital and
1400 premature deaths every year with an annual cost to society
amounting to DKK 14.7 billion.
2
Consequently, there is growing
international and national demand for interventions both to
prevent and to treat stress-related illnesses.
1,3–5
The Danish
Health Authority recommends various forms of cognitive–behav-
ioural therapy (CBT) treatments to help patients recover from
stress.
5
Many alternative treatments exist, but evidence of the effi-
cacy of these various stress treatments is limited. One alternative
treatment is nature-based therapy (NBT), which refers to a thera-
peutic intervention that incorporates natural elements and nature-
related activities often in a specially designed natural environment.
6
NBT has its origins in Anglo-American countries and has a rela-
tively long history of use in healthcare.
7
State of the art
A number of review articles have been published focusing on
the effect of NBT for specific patient groups, such as elderly
people
8
and war veterans.
9
In 2011, Annerstedt & Währborg
published what remains the most comprehensive literature
review of the effects of various forms of NBT on different
patient groups, including both controlled and observational
studies.
10
Annerstedt & Währborg agree with other researchers
(for example Stigsdotter and colleagues
11
) that current research
on NBT focuses too much on qualitative data, and they recom-
mended that future studies be conducted as randomised con-
trolled trials (RCTs).
Study aim
This study aims to test the efficacy of a specific NBT treatment called
Nacadia® NBT (NNBT) by conducting a RCT (ClinicalTrials.gov ID:
NCT01849718) that compares NNBT with a CBT called Specialised
Treatment for Severe Bodily Distress Syndromes (STreSS). The null
hypothesis is: the efficacy of NNBT will not differ significantly from
the efficacy of STreSS, measured by mental health metrics. The
research questions are as follows: (a) has the mental health status of
the participants improved after treatment (compared with baseline),
immediately afterwards, and then 3, 6 and 12 months after the end
of the treatment; (b) are there significant differences between the
two forms of treatment in terms of improved mental health?
Method
Trial design
The efficacy of treatments relying on CBT has been widely docu-
mented,
12
and CBT is recommended by the Danish Health
Authority for treating stress-related illnesses.
5
The efficacy of
NNBT was, thus, compared with a validated CBT treatment.
13
The trial was designed as an equivalence trial to assess whether
NNBT (intervention) would be as effective as STreSS (control) in
treating patients with stress-related illnesses (the full trial protocol
is provided as Supplementary File 1, available at https://doi.org/
10.1192/bjp.2018.2). Participants were individually and equally
randomised to one of the two parallel treatments, and the same
outcome measures were employed. No changes were made to the
methods after the trial had commenced.
Participants
Potential participants were referred to the project by the municipal-
ities, health practitioners (private practice doctors, psychologists
and psychiatrists), and insurance companies in the area around
the Nacadia® therapy garden, which is located in the Municipality
of Hørsholm in Denmark (material on recruitment is provided as
The British Journal of Psychiatry (2018)
Page 1 of 8. doi: 10.1192/bjp.2018.2
1
Supplementary File 2). The municipalities, health practitioners and
insurance agencies had been notified about the project by an infor-
mation letter describing the project and the two treatments. The
letter was followed up by personal contact by the managing therap-
ist at the Nacadia therapy garden. Potential participants also con-
tacted the University of Copenhagen directly in response to brief
notices in newspapers and on the university webpage informing
the public about the project. In total, 200 potential participants
were assessed for eligibility. All individuals needed to provide a
health certificate from their medical doctor and approval to partici-
pate from their municipal case officer. A PhD student conducted a
telephone interview and those participants who were considered
likely to meet the inclusion criteria were then invited to undergo a
clinical assessment. The assessment was conducted by one of four
clinical psychologists, and the diagnoses were supervised and con-
firmed by a psychiatrist.
To be eligible for participation, individuals had to be incapable
of working and have one of the ICD-10 codes
14
F43.0, 2–9 as their
primary diagnosis (psychiatric diagnosis of adjustment disorder and
reaction to severe stress). The level of severe stress was considered to
correspond to 3–24 months of inability to work according to the
Danish Health Service. Adults aged 20–60 years were included.
Individuals with severe psychiatric morbidity, psychotic disorders,
personality disorders, suicidal tendencies and drug or alcohol pro-
blems were excluded. For practical reasons, individuals who were
not fluent in the Danish language were excluded. Based on the inclu-
sion and exclusion criteria, 84 participants were enrolled in the
study.
The study followed the ethical principles of the World Medical
Association Declaration of Helsinki. It was approved by the Danish
Data Protection Agency (J.no. 2013-54-0344) and by the National
Committee on Health Research Ethics (P.no. H-1-2013-038).
Participants received both oral and written information about the
study and signed an informed consent acknowledgement before
enrolment.
Settings
The treatments took place in two different types of settings: STreSS
took place indoors at private clinics of the practising psychologists
located in the city of Hørsholm, and NNBT took place mainly out-
doors at the University of Copenhagen’s therapy garden called
Nacadia. This garden has been designed as a 1.4-hectare wild
forest garden and is located within an arboretum containing the
largest collection of trees and shrubs in the country. The design of
the garden follows the model for evidence-based health design,
and the design process has been transparently described and
documented.
15
Treatments
Apart from the difference in environments, the two treatments also
differed with regard to staff, hours, content and set-up: in NNBT,
two therapists and a gardener were involved in the therapy, which
consisted of 3 h group sessions with individual therapeutic con-
versation and individual nature-based activities introduced by the
gardener. STreSS was conducted as individual 1 h therapeutic con-
versation sessions with one therapist. Although NNBT was carried
out with a group of participants present in the Nacadia garden, most
of the activities were individual, and participants were encouraged
not to interact during the sessions and to keep talk to a
minimum. Therefore, one can argue that NNBT is a hybrid
between a group-based and an individual treatment.
The corresponding elements between the two treatments were
as follows: the therapists were all licensed clinical psychologists
with formal training in CBT; the length of treatment time was
10 weeks; the individual treatment conversations were based on
CBT; and the two treatments took place simultaneously all year
round from August 2013 to March 2015, except during the winter
and summer holidays. (For an overview of differing and corre-
sponding elements, see Supplementary File 3).
CBT
The control group received a form of CBT called STreSS, which was
originally developed at Aarhus University.
16
STreSS consists of nine
modules of manualised CBT, and the efficacy of this treatment has
been tested in an RCT.
13
STreSS was originally designed for group
treatments. In this project, the manual was adjusted to individual
treatments in order to better to match the individual parts of
NNBT. The adjustments were elaborated in dialogue with the
leader of the Aarhus University group, who developed the treatment
manual (in Danish) under the supervision of the psychiatrist in the
project (from Aarhus University) and the person responsible for
conducting CBT in the project (from the University of
Copenhagen). The individual STreSS sessions relied on the same
nine modules as did the group treatment.
16
NNBT
NNBT was originally developed by Corazon and collegues.
6
It builds
upon elements from mindfulness-based stress reduction (MBSR)
17
and CBT, integrated with theories from environmental psychology,
especially attention restoration theory, which emphasises sensory
stimulation from natural environments as a means of restoring fati-
gued cognitive resources.
18
NNBT is a 10-week programme, and participants met 3 days a
week for 3 h. Every day has the same structure, and every week has a
specific theme in accordance with a MBSR manual.
19
NNBT consists of the following five components.
(a) Therapeutic conversations: individual conversations based on
CBT and psychoeducation based on MBSR.
(b) Awareness exercises: individual and group physical and mental
awareness exercises in accordance with MBSR and related to
nature experiences, such as mindful walking in the garden.
(c) Nature-based activities: individual gardening activities, depen-
dent on the season. Before each session, the gardener, who also
maintains the therapy garden, makes a list of possible activities
from which the patients choose together with the therapist.
The choice of activity is guided by what is comprehensible, man-
ageable and meaningful for the individual.
20
Mindful awareness
is integrated into the activities. That is, participants are trained to
notice non-beneficial behavioural patterns that could lead to
stress.
(d) Reflection and relaxation time: individual time for reflection
and relaxation in the garden.
(e) Homework: individual homework to practice the different
techniques and methods.
Outcome measures
Participants in both treatments completed the same self-rated mea-
sures in the form of two questionnaires at five different points in
time: at baseline (first week of treatment), at the end of treatment
(primary end-point measure), and at 3 months, 6 months and
12 months after treatment had ended. The three measures after
the primary end-point measure were included to assess whether
an effect was sustained after treatment had ended. The question-
naires at baseline and at end of treatment were given to participants
by the therapists. The subsequent questionnaires (after 3 months,
6 months, and 12 months) were sent by post to the participants.
Stigsdotter et al
2
The primary outcome was the aggregate score (global score) on
the Psychological General Well-Being Index (PGWBI), which mea-
sures the self-assessed perceived well-being and distress.
21,22
The
PGWBI includes 22 questions in relation to six subscales: anxiety,
depressed mood, positive well-being, self-control, general health
and vitality. Each item has six response options, ranging from
0 to 5. The aggregate score was calculated as the sum of scores on
the six subscales,
21
giving an aggregate score range of 0–110 with
higher values indicating greater well-being.
The secondary outcome measure was the mean total score of
the Shirom–Melamed Burnout Questionnaire (SMBQ). It consists
of 22 items, divided into four subscales: physical fatigue, cognitive
weariness, tension and listlessness. The questionnaire is validated
through research.
23
Additionally, participants completed a background question-
naire at baseline. The background questionnaire contained ques-
tions regarding demography, educational level, employment status
and health.
Sample size and randomisation
The minimum sample size was determined in a power calculation
(2-sample t-test: α= 0.05, 95% CI; expected s.d. = 14.5; β= 0.8;
P< 0.05) (program: Excel). A sample of 40 participants in each
treatment (80 participants in total) was estimated to be able to iden-
tify a mean difference in efficacy of 9.2 between treatments. Since
recruitment took place throughout the intervention on a running
basis, it was possible to account for individuals who dropped out
during the recruitment phase, by recruiting more individuals if
drop out occurred.
Participants were evenly and randomly allocated to one of the
treatments using a computer algorithm. Each participant was allot-
ted a random number between zero and one. Participants with a
number below the median value were randomised to receive
NNBT (n= 43), and participants with a number above the
median value received STreSS (n= 41). The allocation sequence
was generated by a statistician and administered by an independent
research assistant, who also assigned participants to the two differ-
ent treatments.
Statistical analysis
The data was analysed based on intention-to-treat (ITT), whereby
participants were analysed within the treatment to which they
were allocated, irrespective of non-adherence or deviations from
protocol.
24
ITT is the recommended approach in RCT trials
Enrolment
200 participants assessed for eligibility
84 randomised
116 Excluded
• Not meeting inclusion criteria or meeting
exclusion criteria (n= 115)
• Declined to participate (n=1)
• Lost to follow-up (n= 4 did not provide baseline data) • Lost to follow-up (n= 3 did not provide baseline data,
n= 1 withdrew consent)
• Received treatment (n= 42)
Misdiagnosed (n=1)
• Did not receive intervention (n=1)
(Did not want NBT)
43 allocated to NNBT
• 2 (2*) with baseline data only
• 2 (2*) with baseline data and one other time point
• 2 (3*) with data at baseline and two other time points
• 4 (7*) with data at baseline and three other time points
• 29 (25*) with data at all time points
39 included in ITT analysis
• 1 (1*) with baseline data only
• 7 (7*) with baseline data and one other time point
• 4 (5*) with data at baseline and two other time points
• 8 (8*) with data at baseline and three other time points
• 17 (16*) with data at all time points
37 included in ITT analysis
• Received treatment (n= 40)
• Did not received treatment (n=1)
(Did not want CBT)
41 allocated to STreSS
Note. *Provided data on secondary outcome
Follow-up
Analysis
Allocation
Fig. 1 Participant flow diagram. NNBT, Nacadia nature-based therapy; STreSS, Specialised Treatment for Severe Bodily Distress Syndromes;
CBT, cognitive–behavioural therapy; ITT, intention-to-treat.
Nature-based therapy for stress-related illnesses
3
provided by the Consolidated Standards Of Reporting Trials
(CONSORT) guidelines.
25
The CONSORT 2010 checklist can be
found in Supplementary File 4. However, there is no consensus
regarding how to handle missing outcome data within ITT.
24,25
In
the present trial, we used the numerical imputation strategy of
last observation carried forward (LOCF), which is a common
approach in RCT trials
26
and recommended by Hollis &
Campbell in a survey of RCTs.
27
The LOCF strategy was chosen
in order to include participants who happened not to fill out a ques-
tionnaire at some time points. As a complementary analysis, a com-
plete case analysis was also performed in which individuals with
missing outcome data were excluded from the analysis.
The primary outcome data had 36 missing responses to single
questions within the PGBWI questionnaire out of 8360 questions
(0.4%). The missing responses were estimated using the approach
suggested in Chassany et al.
21
The Shapiro–Wilk test, skewness
and kurtosis values, together with a visual inspection of the histo-
grams and the box plots of the sample distributions were used to
detect any deviations from normal distribution. Levene’s test (the
parametric or non-parametric version) was used to reveal any het-
erogeneity of variance among the five ‘points in time’samples
within a treatment or the between-treatment samples at a given
point in time.
Primary outcome (PGWBI)
A two-way mixed-design ANOVA (2 × 5) was employed to detect
any significant effect within a treatment over time and any differ-
ence between the two treatments. Effect sizes for each treatment
were calculated (partial η
2
and ω
2
) based on the one-way
repeated-measures ANOVA of each treatment. Within groups, we
compared each timepoint with baseline to detect by which point
in the study any change may have occurred. Multiple comparisons
were corrected for using Dunnett’s method.
Secondary outcome (SMBQ)
Friedman’s test was used to reveal any significant effect over time
within a treatment. Wilcoxon signed-rank tests were performed
for post hoc multiple pairwise comparisons between points in time
(including adjustment of the significance level to P< 0.0071, using
the Bonferroni correction for seven pairwise comparisons).
Mann–Whitney U-tests were performed for post hoc pairwise com-
parisons between the two treatments at each point in time (includ-
ing adjustment of the significance level to P< 0.01, using the
Bonferroni correction for five pairwise comparisons).
The statistical analyst was first introduced to the project once
the data from the questionnaires had been compiled. The statistical
analyst was not informed whether either of the treatment abbrevia-
tions was preferred over the other.
Results
Participant flow
Recruitment of participants occurred from 2 months before the
treatments started until the beginning of the last treatment period,
that is, from June 2013 to January 2015. In total, seven NNBT treat-
ments were conducted (median group size: 6). Of the 43 participants
assigned to NNBT and 41 patients assigned to STreSS, 42 participants
completed NNBT and 40 participants completed STreSS. In each
treatment, four participants were lost to follow-up (see Fig. 1 for
participant flow).
Baseline characteristics
The baseline characteristics of participants in relation to demography,
education, employment status and sick leave are presented in Table 1.
The vast majority of participants were women. The majority of
respondents held a bachelor’s degree or higher university education.
The respondents were mainly employed but on full-time sick leave.
The most common previous rehabilitation initiatives had involved
visiting a general practitioner or a psychologist.
Primary outcome (PGWBI): sample characteristics
The samples showed no significant deviations from the normal dis-
tribution (P> 0.05). There was neither a significant heterogeneity of
variance among samples within a treatment nor any heterogeneity
of variance between samples of the two treatments at any given
point in time (P> 0.05).
Primary outcome (PGWBI): results from the LOCF
analysis
There was a significant effect of treatment on perceived general
well-being (PGWBI) over time, F(4,144) = 5.23, P< 0.01, overall
partial η
2
= 0.13, and no significant difference between treatments,
F(1,36) = 0.39, P> 0.05, overall partial η
2
= 0.01 (see Fig. 2(a–c)).
NNBT had effect size estimates indicating a medium to large effect
of treatment
28
(η
2
= 0.144; ω
2
= 0.125). STreSS had effect size estimates
indicating a medium effect of treatment (η
2
= 0.088; ω
2
= 0.067).
Table 1 Participants’baseline characteristics
NNBT,
n=39
a
STreSS,
n=37
a
Nationality, n(%)
Danish 38 (97) 32 (97)
Other 1 (3) 1 (3)
Age, mean (s.d.) 47.9 (7.8) 44.9 (8.8)
Gender: female/male, n31/7 27/6
Marital status, n(%)
Married 18 (49) 10 (33)
Divorced 5 (14) 3 (10)
In relationship 8 (21) 4 (13)
Single 6 (16) 13 (43)
Education, n(%)
Primary school (7–10 years) 2 (5) 2 (6)
Secondary school 3 (8) 2 (6)
Vocational training 4 (11) 6 (19)
University bachelor degree 10 (26) 9 (29)
Higher education (>15 years) 19 (50) 12 (39)
Employment status, n(%)
Employed (part-time sick leave) 3 (8) 1 (3)
Employed (full-time sick leave) 23 (62) 24 (77)
Unemployed (on sick leave) 6 (16) 3 (10)
Unemployed 4 (11) 1 (3)
Student on sick leave 0 (0) 2 (6)
Other (revalidation, early retirement 1 (3) 0 (0)
Sick leave because of stress-related illness
prior to treatment start, months: median
(min/max)
6 (3/31) 6 (3/17)
Types of previous rehabilitation (multiple answers), n
None 6 3
General practitioner 22 19
Psychiatrist 11 14
Psychologist 22 18
Psychotherapy 6 6
Medication 9 10
a. The data from the background questionnaires included missing responses to single
questions; therefore, the sum of responses for the different variables does not match the
total number of respondents for Nacadia nature-based therapy (NNBT) (n= 39) and
Specialised Treatment for Severe Bodily Distress Syndromes (STreSS) (n= 37),
respectively.
Stigsdotter et al
4
The post hoc test showed there was a significant difference in
PGWBI scores between the start of treatment and the end of treat-
ment for both the STreSS (P< 0.05) and NNBT groups (P< 0.001)
(Table 2). There was also a significant difference in PGWBI scores
between the start of treatment and each subsequent time point (i.e.
at 3, 6 and 12 months after the treatment). Furthermore, there was
no significant difference between the end of treatment and each sub-
sequent time point for either NNBT or STreSS (Table 2).
The results from the complete case analysis (from which parti-
cipants with missing outcome data were excluded) largely
corresponded with the results from the LOCF analysis. These are
presented in Supplementary File 5.
Secondary outcome (SMBQ): sample characteristics
The samples showed significant deviations from the normal distri-
bution (P< 0.05 for several of the sample distributions). Efforts to
establish normal distributions using various types of transforma-
tions were unsuccessful. There was no significant heterogeneity of
variance among samples within a treatment or between samples
of the two treatments at specific points in time (P> 0.05).
(a)
(b) PGWBI – NNBT (c) PGWBI – STreSS
Mean SD
Min/max
values Mean SD
Min/max
values
Start 46,59 15,38 19/87 Start 49,24 16,64 19/84
End 61,44 15,51 19/98 End 59,62 18,87 19/87
Month 3 63,31 18,58 19/99 Month 3 63,38 21,51 29/96
Month 6 63,28 14,47 21/107 Month 6 65,92 19,91 36/91
Month 12 63,51 16,81 16/102 Month 12 64,86 21,87 29/96
(d)
80
75
70
65
60
55
50
45
40
START END Month 3 Month 6 Month 12
START END Month 3 Month 6 Month 12
5
4.8
4.6
4.4
4.2
4
3.8
3.6
3.4
3.2
3
StreSS
NNBT
STreSS
NNBT
Fig. 2 The effect of Nacadia nature-based therapy (NNBT) and Specialised Treatment for Severe Bodily Distress Syndromes (STreSS) on (a–c)
Psychological General Well-Being Index (PGWBI) scores and on (d) Shirom–Melamed Burnout Questionnaire (SMBQ) scores. (a) The effect of NNBT
on the primary outcome(PGWBI) with 95% confidenceintervals. The mean aggregatescore, standard deviation and minimum and maximum values
for PGWBI in the (b) NNBT group and (c) STreSS group. (d) The effect of NNBT on the secondary outcome (SMBQ) with 95% confidence intervals.
Nature-based therapy for stress-related illnesses
5
Secondary outcome (SMBQ): results
There was a significant effect of treatment over time for STreSS
(χ
2
(4) = 33.15, P< 0.001) and for NNBT (χ
2
(4) = 45.35, P< 0.001)
(see Fig. 2(d) for visual representation).
A significant effect of treatment was observed between the
start of treatment and all succeeding points in time (i.e. end of
treatment or after 3, 6 or 12 months) for both the STreSS and NNBT
groups (see Supplementary File 6). There was no significant pair-
wise difference in SMBQ scores between any of the remaining
time points.
Discussion
Main findings
Both the intervention (NNBT) and the control treatment (STreSS)
showed significant effects at the end of treatment (main end-point
measure). These were expressed as a higher aggregate general
well-being score (PGWBI, primary outcome) and a lower average
burnout score (SMBQ, secondary outcome). No significant differ-
ence was found between the two treatments with regard to
primary and secondary outcomes. The effect was sustained at all
time points after treatment. The mean aggregate score of the
PGWBI 12 months after end of treatment (NNBT: mean 63.51;
STreSS: mean 64.86) approached the reported value for a Danish
healthy sample (mean 73.14), provided by the MAPI Institute.
21
Having a RCT with long-term follow-up in the field of NBT pro-
vides a good opportunity to validate its efficacy and promote recog-
nition of the field, which has until recently relied mostly on
qualitative research lacking long-term follow-up.
10,11
Several muni-
cipalities in Denmark have expressed interest in including NNBT in
their treatment of citizens with stress-related illnesses, but lack of
evidence of its efficacy has been an obstacle.
It should be emphasised that the various NBT treatments on the
market differ substantially in content,
10,11
meaning that results
cannot be generalised to the entire field of NBT for stress-related ill-
nesses. The results relate, instead, to the present form of NNBT used
in the intervention. NNBT can be seen as a complex intervention;
29
and, as a consequence, it is difficult to deduce what caused the effect.
One could argue that some of the possible mechanisms involved in
accommodating the effect of NNBT are shared among NBT treat-
ments in terms of spending time in a natural environment when
feeling stressed.
37
Other mechanisms are potentially related to the
interrelationship of the specific psychotherapeutic approach and
activities in the environment. Such potential mechanisms require
additional investigation.
Limitations
There are a number of limitations to the study. The first is related to
the lack of evidence in the field, which made it difficult to recruit
participants through municipalities, health practitioners and insur-
ance agencies. As a result, we could not be strict in our recruitment
terms, and notices in newspapers and on the university website
represented the main recruitment sources. This meant that partici-
pants who contacted the university directly already had an interest
in the project, which may have been the reason that women were
significantly overrepresented in the sample. However, only two par-
ticipants (one from each treatment) declined to participate after
randomisation because they did not want the particular treatment
to which they were referred. So, we do not consider prior interest
in NBT to be a major bias. The results should nevertheless be
regarded as gender-dependent.
Second, there is a relatively low number of participants provid-
ing data at all time points in the NNBT group (n= 29) and an even
lower number of participants providing data at all time points in the
STreSS group (n= 17). The LOCF analysis seeks to compensate for
the absence of full data-sets by carrying the last observation forward.
A recommendation for future projects is to have fewer time points
and more intensive follow-up for participants who do not provide
data at all time points and to include more participants in the
sample size to account for individuals lost to follow-up.
NNBT was designed to be as individual as possible with minimal
interaction between participants even though they were present in the
therapy garden at the same time. Interviews with participants in the
NNBT group were conducted during the intervention. The results
thereof support the conclusion that the participants considered the
treatment to be very individual thanks to the therapy form and the
garden design.
31
However, because of the difference in treatment
set-ups between STreSS and NNBT (see supplementary File 3), the
NNBT participants could be regarded as clustered as opposed to
the STreSS participants. This issue inflicts a major limitation since
it was not accounted for in the study design, which also raises the
question of how to set-up future study designs entailing a hybrid
between individual and group therapy.
There is always the possibility that treatment in groups may
result in a correlation of individual responses within a group.
32
Table 2 Multiple comparisons between time points
a
Control group
(time point i)
Comparison group
(time point j)
Mean
difference (j–i)
s.e. P95% CI
Lower Upper
NNBT, 1-sided > control Start End 14.85 3.67 <0.001 6.86
Start Month 3 16.72 3.67 <0.001 8.73
Start Month 6 16.69 3.67 <0.001 8.71
Start Month 12 16.92 3.67 <0.001 8.94
NNBT, 2-sided ≠control End Month 3 1.87 3.72 0.923 −6.95 10.69
End Month 6 1.85 3.72 0.926 −6.97 10.67
End Month 12 2.08 3.72 0.900 −6.74 10.90
STreSS, 1-sided > control Start End 10.38 4.62 0.042 0.34
Start Month 3 14.14 4.62 0.005 4.09
Start Month 6 16.68 4.62 0.001 6.64
Start Month 12 15.62 4.62 0.002 5.58
STreSS, 2-sided ≠control End Month 3 3.76 4.78 0.772 −7.60 15.11
End Month 6 6.30 4.78 0.414 −5.06 17.65
End Month 12 5.24 4.78 0.558 −6.11 16.60
NNBT, Nacadia nature-based therapy; STreSS, Specialised Treatment for Severe Bodily Distress Syndromes.
a. The within-group comparison of treatment effect over time, using a particular time point as reference. Time point i is the reference point in time (control), to which each of the subsequent
time points (j) is compared. CI (confidence interval) is relating to the range around the calculated Mean Difference (j-i) value.
Stigsdotter et al
6
Therefore, a larger sample size in combination with more groups
may have been required in the NNBT arm.
33
In order to make an
estimate post hoc as to whether the NNBT treatment in groups
resulted in an intracluster effect, the intracluster correlation coeffi-
cient (ICC) was calculated on the change in the NNBT participants’
PGWBI scores (before minus after NNBT treatment) (ρ= 0.11). The
ICC estimate indicated a lower level of similarity within a group,
compared with between groups. One factor explaining the relatively
low ICC estimate may be the small number of participants within
each group in comparison with the relatively large number of
groups (clusters).
Research shows that ICC estimates vary a great deal in second-
ary care studies,
33,34
and it is difficult to say when an ICC can be
considered minor. However, even a low ICC estimate may still
have a profound design effect, measured as the extent to which a
sample size ought to be inflated to accommodate for homogeneity
in the clustered data. Based on the post hoc estimate of the ICC
and the group size, the design effect was 1.52, leading to an effective
sample size of 26 participants for the NNBT treatment.
35
Based on
the design effect, the required sample size should have been inflated
to 61 participants.
Since the research design did not account for clustering, a sub-
sequent clustered analysis would most likely be underpowered since
a group RCT requires more groups and much larger sample sizes
than an individual RCT.
36
Since the NNBT treatment was consid-
ered to be mainly individual, a decision was made to base the stat-
istical analyses on the sample of individuals rather than to take the
more conservative approach of using the number of groups as a
sample.
37
The research design and statistical analysis, therefore,
inflicts a limitation on the interpretation of the results in relation
to the potential cluster effect in NNBT.
When comparing treatments, there is always the question of
how to balance their differences so as to create a ‘fair’trial
that does not favour either treatment. In the present case, we tried
to balance as many factors as possible, the most important one
being to add more hours to STreSS than is usually provided in
CBT. Since STreSS had already proven its efficacy in an RCT
study
13
and CBT is the recommended approach to treating
stress-related illnesses in Denmark,
5
we considered it a strong
match to NNBT.
In conclusion, the results contribute to the validation of the
present NNBT as a treatment equally effective as STreSS for
people with stress-related illnesses. NNBT and the garden design
may provide research-based guidelines for practitioners, therapists
and landscape architects who are working within the field of NBT
for people with stress-related illnesses.
Ulrika Karlsson Stigsdotter, MSO, PhD, Sus Sola Corazon, PhD, Ulrik Sidenius, PhD,
Patrik Karlsson Nyed, PhD, Department of Geosciences and Natural Resource
Management, Faculty of Social Sciences, University of Copenhagen, Denmark; Helmer
Bøving Larsen, PhD, Department of Psychology, Faculty of Social Sciences, University of
Copenhagen, Denmark; Lone Overby Fjorback, MD, PhD, Department of Clinical
Medicine, University of Aarhus, Denmark
Correspondence: Sus Sola Corazon, Rolighedsvej 23, 1959 Frederiksberg C, Denmark.
Email: suoe@ign.ku.dk
First received 25 Sep 2017, final revision 3 Nov 2017, accepted 22 Dec 2017
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjp.2018.2.
Funding
The project was funded by the TRYG Foundation (grant number 7206-08), a non-profit founda-
tion with core interests in the areas of safety, health and well-being.
References
1World Health Organization. Mental Health, Facing the Challenges, Building
Solutions: Report from the WHO Ministerial Conference. WHO Regional Office
for Europe, 2005.
2Juel K, Sorensen J, Bronnum-Hansen H. Risk Factors and Public Health in
Denmark [in Danish]. National Institute of Public Health, 2006.
3European Foundation for the Improvement of Working and Living Conditions.
Work Related Stress. European Foundation for the Improvement of Working
and Living Conditions, 2010 (https://www.eurofound.europa.eu/sites/default/
files/ef_files/docs/ewco/tn1004059s/tn1004059s.pdf).
4Eplov FL, Lauridsen S. Mental Health Promotion –Background, Concepts and
Determinents [in Danish]. The Danish Health Authority, 2008.
5The Danish Health Authority. Long-term Stress –Current Knowledge and
Suggestions for Stress Prevention. Advice for General Practice. The Danish
Health Authority, 2007.
6Corazon SS, Stigsdotter UK, Jensen AG, Nilsson K. Development of the nature-
based therapy concept for patients with stress-related illness at the Danish
healing forest garden Nacadia. J Am Soc Hortic Sci 2010; 20:34–51.
7Marcus C, Barnes M. Healing Garden - Therapeutic Benefits and Design
Recommendations (1st edn). Wiley, 1999.
8Detweiler MB, Sharma T, Detweiler JG, Murphy PF, Lane S, Carman J, et al. What
is the evidence to support the use of therapeutic gardens for the elderly?
Psychiatry Investigat 2012; 9: 100–10.
9Poulsen VD, Stigsdotter UK, Refshauge AD. Whatever happened to the soldiers?
Nature-assisted therapies for veterans diagnosed with post-traumatic stress
disorder: a literature review. Urban For Urban Green 2015; 14: 438–45.
10 Annerstedt M, Währborg P. Nature assisted therapy: systematic review of
controlled and observational studies. Scand J Publ Health 2011; 39: 371–88.
11 Stigsdotter UK, Palsdottir AM, Burls A, Chermaz A, Ferrini F, Grahn P. Nature-
based therapeutic interventions. In Forests, Trees and Human Health (eds K
Nilsson and M Sangster): 309–42. Springer, 2011.
12 Butler AC, Chapman JE, Forman EM, Beck AT. The empirical status of cognitive-
behavioraltherapy. A review of meta-analyses.Clin Psychol Rev 2006; 26:17–31.
13 Schröder A, Rehfeld E, Ørnbøl E, Sharpe M, Licht RW, Fink P. Cognitive−
behavioural group treatment for a range of functional somatic syndromes: ran-
domised trial. Br J Psychiatry 2012; 200:499–507.
14 World Health Organization. The ICD-10 Classification of Mental and Behavioural
Disorders: Clinical Descriptions and Diagnostic Guidelines. WHO, 1992.
15 Forest & Landscape. Concept Manual for the Therapy Garden Nacadia
®
[In
Danish]. University of Copenhagen, 2008.
16 Rehfeld E, Schröder A, Fink P. Specialized Treatment for Severe Bodily Stress.
Aarhus University Hospital, 2009.
17 Kabat-Zinn J. Full Catastrophe Living: Using the Wisdom of Your Body and Mind
to Face Pain, Stress and Illness. Dell Publishing, 1990.
18 Kaplan R, Kaplan S. The Experience of Nature (1st edn). Cambridge University
Press, 1989.
19 Fjorback LO. Mindfulness Manual. PsykiatriFonden, 2009.
20 Antonovsky A. The salutogentic model as a theory to guide health promotion.
Health Promot Int 1996; 11:11–8.
21 Chassany O, Diderot D, Dubois ZD. The Psychological General Well-being Index
(PGWBI) User Manual. MAPI Research Institute, 2004.
22 Lundgren-Nilsson A, Jonsdottir IH, Ahlborg G, Tennant A. Construct validity of
the psychological general wellbeing index (PGWBI) in a sample of patients
undergoing treatment for stress-related exhaustion: a rash analysis. Health
Qual. Life Outcomes 2013; 11:2.
23 Lundgren-Nilsson A, Jonsdottir IH, Pallant J, Ahlborg G. Internal construct val-
idity of the Shirom-Melamed Burnout Questionnaire (SMBQ). BMC Public Health
2012; 11:2.
24 LindquistEF. Statistical Analysisin Educational Research.Houghton Mifflin, 1940.
25 Alshurafa M, Briel M, Akl EA, Haines T, Moyyedi P, Gentles SJ, et al. Inconsistent
definitions for intention-to-treat in relation to missing outcome data: system-
atic review of the methods literature. PLoS One 2012; 7: e49163.
26 Schulz KD, Altman DG, Moher D. The 2010 CONSORT statement: updated guide-
lines for reporting parallel group randomized trials. BMJ 2010; 340:698–703.
27 Hollis S, Campbell F. What is meant by intention to treat analysis? Survey of
published randomised controlled trials. BMJ 1999; 319: 670–74.
28 Cohen J. Statistical Power Analysis for the Behavioral Sciences. Routledge
Academic, 1988.
29 Campbell M, Fitzpatrick R, Haines A, Kinmonth AL, Sandercock P,
Spiegelhalter D, et al. Framework for design and evaluation of complex inter-
ventions to improve health. Br Med J 2000; 321: 694.
Nature-based therapy for stress-related illnesses
7
30 Berto R. The role of nature in coping with psycho-physiological stress. A lit-
erature review on restorativeness. Behav Sci 2014; 4: 394–409.
31 Sidenius U, Stigsdotter UK, Poulsen DV, Bondas T. ‘I look at my own forest and
fields in a different way’: the lived experience of nature-based therapy in a
therapy garden when suffering from stress-related illness. Int J Qual Stud
Health Well-Being 2017; 12:1–14.
32 Adams G, Gulliford MC, Ukoumunne OC, Eldridge S, Chinn S, Campbell MJ.
Patterns of intra-cluster correlation from primary care research to inform study
design and analysis. J Clin Epidemiol 2004;57: 785–94.
33 Campbell MK, Fayers PM, Grimshaw M. Determinants of the intracluster cor-
relation coefficient in cluster randomized trials: the case of implementation
research. Clin Trials 2005; 2:99–107.
34 Baldwin S, Murray DM, Shadish WR, Pals SL, Holland JM, Abramowitz JS, et al.
Intraclass correlation associated with therapists: estimates and applications in
planning psychotherapy research. Cogn Behav Ther 2011; 40:15–31.
35 Killip S, Mahfoud Z, Pearce K. What is an intracluster correlation coefficient?
Crucial concepts for primary care researchers. Ann Med 2004; 2: 204–8.
36 Christie J, O’Halloran P, Stevenson M. Planning a cluster randomized controlled
trial: methodological issues. Nurs Res 2009; 58: 128–34.
37 Donner A, Birkett N, Buck C. Randomization by cluster: sample size require-
ments and analysis. Am J Epidemiol 1981; 114: 906–14.
Stigsdotter et al
8
