Supported self-management for people with type 2 diabetes: a meta-review of quantitative systematic reviews

1Captieux M, et al. BMJ Open 2018;8:e024262. doi:10.1136/bmjopen-2018-024262
Open access
Supported self-management for people
with type 2 diabetes: a meta-review of
quantitative systematic reviews
Mireille Captieux,1 Gemma Pearce, 2 Hannah L Parke,3 Eleni Epiphaniou,4
Sarah Wild,1 Stephanie J C Taylor, 5 Hilary Pinnock1
To cite: Captieux M, Pearce G,
Parke HL, et al. Supported
self-management for people
with type 2 diabetes: a
meta-review of quantitative
systematic reviews. BMJ Open
2018;8:e024262. doi:10.1136/
bmjopen-2018-024262
► Prepublication history and
additional material for this
paper are available online. To
view these files, please visit
the journal online (http:// dx. doi.
org/ 10. 1136/ bmjopen- 2018-
024262.
Received 17 May 2018
Revised 31 August 2018
Accepted 12 September 2018
1
Usher Institute of Population
Health Sciences and Informatics,
The University of Edinburgh,
Edinburgh, UK
2
Coventry University, Centre
for Advances in Behavioural
Science, Coventry, UK
3
University of Exeter Biomedical
Informatics Hub, Exeter, Devon,
UK
4
University of Nicosia,
Department of Social Sciences,
Nicosia, Cyprus
5
Centre for Primary Care and
Public Health, Barts and the
London School of Medicine and
Dentistry, Queen Mary University
of London, London, UK
Correspondence to
Dr Hilary Pinnock;
Hilary. Pinnock@ ed. ac. uk
Research
© Author(s) (or their
employer(s)) 2018. Re-use
permitted under CC BY.
Published by BMJ.
AbstrACt
Objectives Self-management support aims to give people
with chronic disease confidence to actively manage their
disease, in partnership with their healthcare provider. A
meta-review can inform policy-makers and healthcare
managers about the effectiveness of self-management
support strategies for people with type 2 diabetes, and
which interventions work best and for whom.
Design A meta-review of systematic reviews of
randomised controlled trials (RCTs) was performed
adapting Cochrane methodology.setting and participants Eight databases were
searched for systematic reviews of RCTs from January
1993 to October 2016, with a pre-publication update in
April 2017. Forward citation was performed on included
reviews in Institute for Scientific Information (ISI)
Proceedings. We extracted data and assessed quality with
the Revised-Assessment of Multiple Systematic Reviews
(R-AMSTAR).
Primary and secondary outcome measures Glycaemic
control as measured by glycated haemoglobin (HbA1c)
was the primary outcome. Body mass Index, lipid
profiles, blood pressure and quality of life scoring were
secondary outcomes. Meta-analyses reporting HbA1c were
summarised in meta-forest plots; other outcomes were
synthesised narratively.results 41 systematic reviews incorporating data
from 459 unique RCTs in diverse socio-economic and
ethnic communities across 33 countries were included.
R-AMSTAR quality score ranged from 20 to 42 (maximum
44). Apart from one outlier, the majority of reviews
found an HbA1c improvement between 0.2% and 0.6%
(2.2–6.5 mmol/mol) at 6 months post-intervention, but
attenuated at 12 and 24 months. Impact on secondary
outcomes was inconsistent and generally non-significant.
Diverse self-management support strategies were
employed; no single approach appeared optimally
effective (or ineffective). Effective programmes tended to
be multi-component and provide adequate contact time
(>10 hours). Technology-facilitated self-management
support showed a similar impact as traditional approaches
(HbA1c MD −0.21% to −0.6%).
Conclusions Self-management interventions using a
range of approaches improve short-term glycaemic control
in people with type 2 diabetes including culturally diverse
populations. These findings can inform researchers, policy-
makers and healthcare professionals re-evaluating the
provision of self-management support in routine care.
Further research should consider implementation and
sustainability.
IntrODuCtIOn
The burden of type 2 diabetes is a promi-
nent global health challenge currently esti-
mated to affect 415 million adults worldwide1
with greatest prevalence among socio-eco-
nomically deprived populations and those
of African, Afro-Caribbean, South Asian
and Middle Eastern ethnicity. 2 An increas-
ingly obese, sedentary, ageing population is
expected to drive this number up to an esti-
mated 642 million (one adult in 10) by 2040.2
Healthcare service providers, commissioners
and policy-makers must meet the increasingly
complex needs and expectations of diverse
patient populations with type 2 diabetes
despite limited resources.
Supported self-management aims to give
people with chronic disease confidence in
taking an active role in all aspects of theirstrengths and limitations of this study
► Meta-reviews provide a high-level overview of evi-
dence ideal for informing policy and health service
development, but fine-grained detail is lost as ran-
domised controlled trials (RCTs) are synthesised into
systematic reviews and then meta-reviews.
► A comprehensive search strategy in line with a pre-
defined protocol was used to gather a large evidence
base examining the impact of diverse self-manage-
ment support interventions on different type 2 dia-
betes populations from 1993 to 2017.
► Individual RCTs may be included in multiple sys-
tematic reviews; this precludes meta-analysis and
means that that some RCTs may be over-represent-
ed in our synthesis; we have identified and report
this overlap.
► The research team encompassed public health,
statistics, epidemiology, primary care and health
psychology expertise, enabling a multi-disciplinary
approach to interpretation.
on 22 April 2019 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2018-024262 on 14 December 2018. Downloaded from

2Captieux M, et al. BMJ Open 2018;8:e024262. doi:10.1136/bmjopen-2018-024262
Open access
disease management, and health behaviours,3 in partner-
ship with their care-providers. 4 It is promoted as a strategy
that can cost-effectively enable patients to contribute to
the improvement of their own outcomes and plays a key
role in the WHO’s Innovative Care for Chronic Condi-
tions (ICCC) framework.5 The increasing literature in this
area may overwhelm decision-makers seeking to under-
stand how best to support patients with type 2 diabetes.6
A meta-review of systematic reviews can provide a broad,
high-level, over-arching synthesis of the existing evidence
base in a single manuscript to inform policy, research and
practice. 6 The review questions were: Do self-manage-
ment support interventions improve glycaemic and other
physiological outcomes for people with type 2 diabetes in
comparison to usual care? What works, for whom and in
what contexts?
MethODs
We adapted Cochrane methodology to conduct a meta-re-
view of systematic reviews of randomised control trials
(RCTs) examining self-management support in people
with type 2 diabetes. 7 Reporting follows the Preferred
Reporting Items for Systematic Reviews and Meta-Anal-
yses (PRISMA) guidelines. 8 The initial search (January
1993 to June 2012), undertaken as part of the Practical
Systematic Review of Self-Management Support for long-
term conditions (PRISMS) meta-review,9 was updated in
October 2016, and a pre-publication update completed
in April 2017. Meta-reviews cannot be registered with the
International Prospective Register of Ongoing System-
atic Reviews (PROSPERO) but the PRISMS protocol is
available online: https://www. journalslibrary. nihr. ac. uk/
programmes/ hsdr/ 11101404/#/.
Data sources and search strategy
The participants, interventions, comparators, outcomes
and settings (PICOS) search strategy8 (table 1) combined
terms for: ‘self-management support’ AND ‘diabetes’ AND
‘systematic review’ and limits specified (human subjects,
English language, published after 1 st January 1993)
(online supplementary table 1). We searched MEDLINE,
EMBASE, CINAHL, PsychINFO, AMED, BNI, Cochrane
Database of Systematic Reviews and Database of Abstracts
for Reviews of Effectiveness (DARE). A forward citation
was carried out on all included reviews in ISI Proceedings
(Web of Science) at the time of the database searches and
subsequently as a pre-publication update. This approach
is an efficient way to update searches.10
Table 1 PICOS search strategy and sources for the review
Definition
Population Adults with type 2 diabetes from all social and demographic settings. Multi-condition studies included if
possible to extract type 2 diabetes data separately.
Intervention Self-management support interventions.
We defined self-management as: 'The tasks that individuals must undertake to live with one or more
chronic conditions. These tasks include having the confidence to deal with medical management, role
management and emotional management of their conditions’.3 This definition implies action on the part of
the individual.
We defined self-management support interventions as ‘any interventions that facilitates self-
management’, that is, professional or non-professional care-givers collaboratively assisting individuals
to manage the medical, role or emotional components of their type two diabetes. Interventions that
solely provide one-way instructions to participants were not classified as self-management support
interventions.
We specified that supported self-management interventions would be multi-component, so that a
mono-component intervention (eg, exercise training) would be excluded unless it also offered (say) self-
management education giving people confidence to exercise in everyday life.
Comparator Generally usual care or less intense self-management interventions.
Outcomes Primary: HbA1c, Secondary: biomedical markers: body mass index/weight, lipids, complications. Patient
reported: quality of life. Intermediate: self-efficacy, self-management behaviours.
Settings Any healthcare settings.
Study Design Systematic review of randomised control studies.
Dates Initial database search: January 1993 to August 2012; Update search October 2016; Pre-publication
forward citation April 2017.
Databases MEDLINE, EMBASE, CINAHL, PsychINFO, AMED, BNI, Cochrane Database of Systematic Reviews,
Database of Abstracts of Review of Effects and ISI Proceedings (Web of Science).
Forward citations On all included systematic reviews. Bibliographies of eligible reviews.
In progress studies Abstracts were used to identify recently published trials.
Other exclusions Previous versions of updated reviews.
Papers not published in English.
on 22 April 2019 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2018-024262 on 14 December 2018. Downloaded from

3Captieux M, et al. BMJ Open 2018;8:e024262. doi:10.1136/bmjopen-2018-024262
Open accessstudy selection
Table 1 gives the definitions that we used to identify relevant
reviews: in summary, we included reviews of interventions that
supported individuals to actively manage the medical, role or
emotional components of their type 2 diabetes.3 4 Following
training, title and abstracts from the original PRISMS search
were screened using the exclusion criteria online supple-
mentary table 2 (HLP) with a 10% random check (GP, EE)
with 96% agreement; the update search was screened (MC)
with a 1% check (GP) with 97% agreement. Disagreements
were discussed with a third reviewer (HLP, SJCT or SW)
until consensus was reached. The full texts were screened
(original: HLP, GP, EE, update: MC) with 10% check in the
original review (HLP or SJCT) with 89% agreement, and
100% checked in the update (HLP) with 93% agreement.
Any disagreements were resolved in discussion with a third
reviewer (HLP, SJCT or GP).
Data extraction and quality assessment
Using a piloted form, data were extracted on: review
rationale, review methodology, inclusion criteria, partic-
ipant demographics and intervention details, outcomes
and conclusions as synthesised by the review authors.
Only data provided in systematic reviews were extracted;
data were not extracted from individual RCTs within
systematic reviews. Data extraction was undertaken (HLP
original; MC update) with a 10% check of extraction and
quality assurance (GP, EE) and a 100% check of numer-
ical data extracted (GP, HLP). Methodological quality was
assessed (HLP, MC) using the R-AMSTAR tool (Revised -
A MeaSurement Tool to Assess systematic Reviews) 11 with
a 10% check (GP, EE). Papers were defined as very high
quality if their score was ≥40, high quality if their score
was ≥35, medium quality if their score was ≥30 and low
quality if their score was less than 30. Publication bias, if
reported in systematic reviews, was noted.
Data synthesis and analysis
The primary outcome was HbA1c (or other measure of
glycaemic control). Secondary outcomes included: other
biomedical markers of disease (blood pressure (BP), lipid
profile, weight and body mass index (BMI); quality-of-life;
intermediate outcomes (health behaviour or self-efficacy).
In addition to the definition of self-management and
self-management support that were used to select relevant
studies (table 1), we also used the PRISMS Taxonomy of
Self-Management Support 12 to identify self-management
components within systematic reviews, even if the term
‘self-management’ was not used explicitly. The taxonomy
also provided a consistent language to describe the inter-
ventions in the included RCTs and to identify components
used. Meta-analysis is inappropriate at the meta-review
level because of overlap of RCTs included in the system-
atic reviews; therefore narrative synthesis was undertaken.
For the primary outcome (HbA1c), the summary data
from the meta-analyses in the included reviews were illus-
trated using meta-forest plots.
Patient and public involvement and stakeholder engagement
Our lay collaborator, people with long-term conditions,
representatives of patient organisations as well as profes-
sional stakeholders (clinicians, healthcare managers and
policy-makers) contributed to workshops throughout
the PRISMS programme of reviews.9 Their opinions
informed the decision about the focus of core reviews.
At an end of project workshop, patients and other stake-
holders provided feedback on the findings, informed
our interpretation and suggested practical approaches to
dissemination.results
The PRISMA diagram (figure 1) details the search
and selection process. We identified 28 143 references
(14 839 in the original PRISMS search and 13 304 in
the 2016 update). After screening, 41 systematic reviews
were included in the review: 17 papers from the orig-
inal review,13–22 24 papers from the update23–46
; and two
identified from other sources47 48 ; in addition, two of the
originally included systematic reviews were replaced by
updates. 49 50 See online supplementary table 3 for the
reviews excluded at the Update full text screening. There
were 459 unique RCTs reported in the included system-
atic reviews; the overlap of RCTs between the reviews is
illustrated in online supplementary figure 1.summary of included reviews
The 41 included systematic reviews encompassed RCTs
from 33 countries: Argentina, Australia, Austria, Bahrain,
Canada, China, Costa Rica, Croatia, Cuba, Denmark,
Finland, Germany, Hong Kong, Iceland, India, Iran,
Ireland, Israel, Italy, Japan, Mexico, New Zealand, South
Korea, Spain, Sweden, Taiwan, Thailand, the Nether-
lands, Turkey, UK, USA, Vietnam and the West Indies.
Year of publication ranged from 2001 to 2016, with the
RCT publications ranging from 1981 to 2015 (online
supplementary table 4). The majority of reviews (26/39)
included a meta-analysis,13–15 19 22–24 27–33 35–38 40 45–48 51–53
with the remaining 15 presenting a narrative synthesis.
Intervention duration and follow-up duration were
not always clearly defined. Where recorded, the average
number of sessions ranged from 1 to 10 sessions,
average contact time ranged from 30 min to 58 hours,
over 6 weeks to 2 years (online supplementary table
4). 15–18 21 24 26 28 31 32 35 36 40–48 51–54 Twenty-one systematic
reviews explicitly documented the follow-up duration of
their included RCTs.19 22 24 25 27 29–37 39 40 45 46 48 52 53 The
modal follow-up ranged from immediately after the inter-
vention to 5 years.
Quality assessment
The quality of the reviews ranged from 2047 to 4224 from a
R-AMSTAR total of 44 (online supplementary table 4 and
5). Four systematic reviews were very high quality,18 24 26 27 12
were judged high quality,14 15 19 23 28 35 37 43 45 48 52 53 15 reviews
were judged medium quality13 17 22 29–31 33 36 38 39 41 42 44 46 54
on 22 April 2019 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2018-024262 on 14 December 2018. Downloaded from

4Captieux M, et al. BMJ Open 2018;8:e024262. doi:10.1136/bmjopen-2018-024262
Open access
and 10 were low quality. 16 20 21 25 32 34 40 47 51 55 Total number
of patients in each review ranged from 64 to 33 124. Overall
nine systematic reviews stated no publication bias had
been found.14 23 29 36 38–40 45 48 Bolen et al found publication
bias but noted no change after sensitivity analysis, 12 iden-
tified possible publication bias13 15 19 24 25 28 30 33 37 39 43 46 and
16 did not assess publication bias16 17 20–22 31 32 34 41 42 47 51–55
;
three reviews stated insufficient studies to carry out mean-
ingful assessment of publication bias.18 26 27
Overview of results
Does supported self-management improve outcomes for
people with type 2 diabetes?
Primary outcome: HbA1c
Thirty-five of 41 systematic reviews assessed glycaemic
control, 24 of these presented meta-analyses of HbA1c
data (online supplementary table 6). Follow-up periods
varied between 0 and 24 months and were undefined in
eight of the 22 reviews.13 15 23 28 30 33 37 38 Eleven system-
atic reviews presented narrative findings on glycaemic
control.17 20 21 25 26 34 41 42 44 54 55 Ten of the 11 narrative reviews
were low or medium quality17 20 21 25 34 41 42 44 54 55
while
18 of the 24 meta-analyses were medium or high
quality. 13–15 19 23 28–31 33 35–38 45 48 52 53
All but one meta-analysis53 found a statistically signif-
icant improvement in HbA1c following a self-manage-
ment intervention (figure 2). The HbA1c decrease in 17
of these reviews was less than 0.5% (5 mmol/mol); three
reviews reported a decrease between 0.5% (5 mmol/
mol) and 1% (11 mmol/mol). 19 22 28 One low-quality
review reported an decrease of 1.2% (13 mmol/mol)
with wide confidence intervals. 40 Three reviews reported
effect sizes (thus were not included in the meta-forest
plot) showing a significant reduction in HbA1c. 30 45 47 Six
of the 11 narrative reviews confirmed a positive effect on
HbA1c 17 20 21 25 34 41
; five reported an inconsistent effect
on HbA1c.
The comparator group in the RCTs varied both within
and between systematic reviews and ‘usual care’ was not
always specified. Two reviews performed sub-set analyses
based on the nature of the control intervention.38 48 Both
found a greater mean difference (intervention/control)
when control was usual care than when the control was a
minimal self-management intervention. However, classi-
fying reviews based on whether they specified a usual care
comparator as opposed to a minimal care intervention
showed no obvious pattern in HbA1c (online supplemen-
tary figure 2a,b).short-term, medium-term and long-term hbA1c outcomes
Where follow-up times were differentiated in the system-
atic reviews, they are illustrated in figure 3a-c. This
series of forest plots illustrates that the effect on HbA1c
attenuated with time; a statistically significant effect
persisted for 6 months in four of six reviews 19 24 27 52 and
for 12 months in three of six reviews. 24 45 52 Attridge et
al (the highest quality systematic review 42/44) was one
of two reviews showing an improvement in HbA1c that
persisted at 24 months follow-up. 24 52 Fewer RCTs were
included in the meta-analyses for long-term outcomes;
at the 24 month follow-up, only one meta-analysis
included data from more than 4 RCTs. 14 Three narra-
tive reviews 17 21 22 reported decreasing effectiveness over
time.secondary outcomes
Biomedical markers
Nine systematic reviews presented meta-analysis data of
biomedical markers13 15 24 27 35 48 52 53 ; eight presented
narrative data.17 21 25 26 34 42 44 54 Self-management support
generally had no significant effect on BMI, weight and
Figure 1 Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram.
on 22 April 2019 by guest. Protected by copyright.http://bmjopen.bmj.com/BMJ Open: first published as 10.1136/bmjopen-2018-024262 on 14 December 2018. Downloaded from