How can clinical practices pragmatically increase physical activity for patients with type 2 diabetes? A systematic review

Abstract

Although regular physical activity (PA) is a cornerstone of treatment for type 2 diabetes (T2D), most adults with T2D are sedentary. Randomized controlled trials (RCTs) have proven the effectiveness of PA behavioral interventions for adults with T2D but have rarely been conducted in healthcare settings. We sought to identify PA interventions that are effective and practical to implement in clinical practice settings. Our first aim was to use the valid Pragmatic-Explanatory Continuum Indicator Summary 2 (PRECIS-2) tool to assess the potential for future implementation of PA interventions in clinical practice settings. Our second aim was to identify interventions that effectively increased PA and glycemic control among the interventions in the top tertile of PRECIS-2 scores. We searched PubMed MEDLINE from January 1980 through May 2015 for RCTs of behavioral PA interventions coordinated by clinical practices for patients with T2D. Dual investigators assessed pragmatism by PRECIS-2 scores, and study effectiveness was extracted from original RCT publications. The PRECIS-2 scores of the 46 behavioral interventions (n = 13,575 participants) ranged from 3.0 to 4.8, where 5 is the most pragmatic score. In the most pragmatic tertile of interventions (n = 16) by PRECIS-2 scores, 30.8 and 31.3% of interventions improved PA outcomes and hemoglobin A1c, respectively. A minority of published evidence-based PA interventions for adults with T2D were both effective and pragmatic for clinical implementation. These should be tested for dissemination using implementation trial designs.

INTRODUCTION

Between 1980 and 2014, the prevalence of diabetes mellitus almost quadrupled [1], and recent estimates suggest that one third of all Americans born after the year 2000 will develop diabetes [2]. Among adults, type 2 diabetes (T2D) accounts for over 90% of diabetes diagnoses [3]. T2D more than doubles the risk of cardiovascular disease, including myocardial infarction and stroke [4–7]. Even if people with T2D do not develop cardiovascular disease or microvascular complications of T2D, they remain at increased risk for early mortality and morbidity, including disability and other functional problems [8, 9].

To mitigate these serious risks and successfully manage T2D, a healthy lifestyle is important. In particular, regular physical activity (PA) is a cornerstone of T2D management due to its major health benefits [10]. Observational data have linked regular PA to improved all-cause mortality, lower rates of cardiovascular disease, lower rates of breast and colon cancer, improved symptoms of depression and anxiety, and better functional outcomes in people with and without T2D [11]. Data from randomized controlled trials (RCTs) of lifestyle interventions that included both PA and weight loss have also yielded better health outcomes, such as lower incidence of stroke, improved fitness, and improved mobility [12–14]. Based on observational data, it also appears that healthcare costs are lower for people with T2D who report regular PA [15, 16]. For example, among patients with T2D whose health insurance provided a health club membership, those who attended health clubs at least twice weekly over 12 months had $1252 lower mean healthcare costs than patients with T2D who attended PA classes less than once weekly (P < 0.001) [15]. The potential cost savings of regular PA are particularly important given that the annual medical costs of T2D in the US were recently estimated at $176 billion in direct medical costs and $69 billion in reduced productivity [17].

Based on the substantial benefits of PA, the American Diabetes Association, the American College of Sports Medicine, and the US Physical Activity Guidelines all recommend that people with diabetes engage in at least 150 min of moderate to vigorous intensity PA, such as brisk walking, as well as two to three bouts of resistance training each week [11, 18]. However, most patients with T2D do not meet these standards [19, 20]. In fact, people with T2D are less likely to engage in regular PA than their peers without diabetes, even though the majority of people with T2D recognize that regular PA is important [19, 20]. This disconnect between knowledge and action highlights the potential for implementing evidence-based interventions in clinical care settings to increase reach and address personal barriers to PA for patients with T2D.

Currently, delivery of evidence-based PA interventions for people with T2D is not common practice in most clinical care settings [21, 22]. One possible explanation for this is that large-scale interventions which improved PA were not fully integrated into clinical practice settings; thus, they are not easy to translate into real-world settings [12, 14]. Studying interventions which are clinically integrated is critically important because limited clinic resources pose challenges to the translation of effective PA interventions into primary care [23, 24]. Difficulties in bringing evidence-based interventions to clinical practice are not limited to diabetes and PA behavior—the field of implementation science has developed rapidly to address concerns that <10% of evidence-based interventions are implemented in real-world settings [25].

There are valuable real-world alternatives to integrating interventions into clinical practice: community-based programs are able to deliver evidence-based interventions, often tailored to specific community needs. Community-based interventions have been translated broadly for the Diabetes Prevention Program, and economic analyses suggest that these programs are highly cost-effective [26]. In contrast to diabetes prevention, there are more limited examples of community-based programs independently delivering diabetes self-management separately from clinical practices [27–29]. The National Diabetes Education Program has sought to clarify the key roles that communities can play to support patients with diabetes in their self-management [30], and a recent review summarizes how communities and clinical practices may collaborate to improve diabetes self-management [29]. While the relevance of community-based programs for diabetes self-management should not be ignored, the focus of this review will be on the delivery of evidence-based physical activity counseling programs into clinical practice. The rationale for focusing on programs integrated within clinical practices is because this provides the benefit of utilizing clinician’s judgment to identify patients who may participate safely; clinics also have the potential to receive reimbursement for these programs through health insurance.

To address the limited translation of effective physical activity programs into clinical practices, stakeholders with experience in clinical care, healthcare financing, and clinical trial design developed a tool to help clinical trial designers plan RCTs to test the delivery of evidence-based interventions in pragmatic real-world settings, as compared with planning RCTs of interventions to test efficacy that should be delivered in more idealized research settings [31]. This measurement tool was named the PRagmatic Explanatory Continuum Indicator Summary (PRECIS) model. In 2015, >80 international trialists, clinicians, and policy-makers revised and validated an updated version: the PRECIS-2 model [32]. The PRECIS-2 model may serve two purposes—to inform future trial design and to allow researchers to assess the “real-world applicability” of published studies [33].

There is a lack of published information regarding the potential for clinical translation of effective PA interventions for patients with T2D [10]. To address this gap, the goal of this review is to identify existing effective and pragmatic PA interventions for patients with T2D that may be translated into clinical practice. Our specific aims are (1) to use the PRECIS-2 tool to assess the pragmatism for implementation of PA interventions into clinical practice settings and (2) to identify interventions that effectively increased PA and glycemic control, respectively, among interventions ranked in the highest tertile by PRECIS-2 for pragmatism. By addressing these aims, we hope to spur further study of the benefits and costs of implementing pragmatic and effective PA interventions for patients with T2D into diverse clinical care settings.

METHODS

Data sources and searches

Our study team developed an a priori set of relevant inclusion and exclusion criteria (Appendix Table 2). Generally, these criteria led us to include RCTs that tested an intervention which included PA counseling for patients with T2D who were cared for in a clinical care practice. We required that PA counseling was conducted consistently for all patients in the intervention group. However, in keeping with real-world practice needs, we also included interventions that targeted additional T2D self-management behaviors, such as diet, medication adherence, and smoking cessation. Study interventions needed to be clinically integrated, which we defined a priori as regular bidirectional communication between the research and clinical care teams. Furthermore, for studies in which the intervention was delivered by a research team member rather than a clinical care team member, we required that the research team interventionists must be healthcare professionals (e.g., physicians, nurses, pharmacists, social workers, psychologists, or behavioral therapists) and that any face-to-face intervention visits must take place in a longitudinal clinical care setting, in order to ensure that this intervention had the potential for future translation into a clinical practice setting.

Table 2.

Inclusion and exclusion criteria for inclusion in systematic review

	Inclusion criteria	Exclusion criteria
Topic area	Physical activity counseling OR exercise prescription	Supervised exercise training Acute effects of exercise Not in humans Breathing/stretching exercises Type 1 diabetes Gestational diabetes Metabolic syndrome
Setting	Ambulatory (outpatient) Healthcare/clinical Integration of intervention with ambulatory clinical setting	Community-based Based in school, workplace, home, or inpatient (hospital) settings Advertising in community
Study design	Clinical trial	Case series/case control Cohort Qualitative Pilot study not powered for outcome Small sample size (N < 20) Pretest-posttest design
Target audience	Adults with type 2 diabetes	Type 1 diabetes Pre-diabetes Gestational diabetes Children
Language	Article in English	Article not in English

Study selection

Working with our library informatics staff, we developed a search strategy for behavioral interventions with a PA component targeted to patients with T2D. The PubMed MEDLINE database was considered to be a comprehensive source for our needs, given our interest in trials of behavioral interventions that were integrated into clinical care. Our complete literature search strategy included terms for behavioral interventions, PA and T2D (Appendix Fig. 4).

Fig. 4.

Terms used in literature search strategy

One author (KAL) preliminarily screened the search results by reviewing the manuscript title and abstract. For this initial screen, we applied our eligibility criteria conservatively to select RCTs conducted among adults with T2D which utilized a self-management intervention that might possibly contain PA, including diet and weight loss interventions. After this initial screen, the remaining articles were reviewed independently in a secondary screen by dual raters (KAL, AGH, and/or IML). For this secondary screen, raters independently applied the study eligibility criteria (Appendix Table 2) to the full-text articles to identify qualifying RCTs. In instances of rater disagreement on whether a study met inclusion/exclusion criteria, these disagreements were resolved through discussion and arbitration by another co-author, when necessary. When both raters felt that there was insufficient information to determine eligibility, we contacted the RCT authors for clarification.

Data extraction and quality assessment

In order to rank interventions by their “real-world” applicability [33, 34], we rated each study in terms of pragmatism, as measured by PRECIS-2. PRECIS-2 contains nine domains (Fig. 1): participant eligibility criteria, participant recruitment, trial setting, organization of intervention delivery, flexibility of delivery, flexibility of adherence, participant follow-up, primary analysis, and relevance of primary outcome [32]. The details of the PRECIS-2 rating scale, including pragmatic and explanatory extremes of each domain, have been described extensively elsewhere [32]. We adapted an existing rating tool that was used to assess PRECIS domains in a prior review article [33]. Our rationale was that highly pragmatic clinical trials more closely mimic real-world circumstances in terms of design. As mentioned earlier, the PRECIS-2 model is a revised version of the original PRECIS tool designed by Thorpe et al. to provide a means of evaluating a study design in terms of pragmatism [31].

Fig. 1.

Factors measured by PRECIS-2 wheel. PRECIS-2 Pragmatic-Explanatory Continuum Indicator Summary 2. Reproduced with permission from BMJ Publishing Group Ltd. from “The PRECIS-2 tool: designing trials that are fit for purpose”; K Loudon, S Treweek, F Sullivan, P Donnan, KE Thorpe, M Zwarenstein; 350:h2147; 2015

Because the PRECIS-2 tool is not the only tool to assess the translation potential for clinical trials, we also assessed other pragmatic factors based on the widely used Reach, Effectiveness, Adoption, Implementation, Maintenance (RE-AIM) framework. RE-AIM was designed with the intention of increasing reporting on the robustness, translatability, and public health impact of healthcare-related trials [33, 34]. We identified important practical feasibility factors derived from the RE-AIM framework that are not represented in the PRECIS-2 domains. These practical feasibility factors that we identified include reports of participant engagement, adaptation/change of intervention, program sustainability, unintended effects of intervention, and monetary costs of intervention [32–34]. We adapted an existing RE-AIM rating tool that was used to assess these practical feasibility factors in a prior review article [33]. Scoring for both RE-AIM and PRECIS-2 factors was assessed from the perspective of a patient-centered medical home (PCMH), rather than a traditional primary care clinic, as the PCMH model of care is rapidly spreading [35] and provides a model for population health teams to deliver behavior change interventions more optimally than traditional primary care practices [36, 37].

We also extracted study data according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) standards of patient population and study outcomes [38]. One author (KAL or IML) extracted the key contextual components of each study. To ensure uniformity, the same author extracted the information from all studies for each PRISMA criterion. In addition to the PRISMA standards, we further extracted data regarding other external validity factors that may relate to future implementation decisions, such as the level of training of interventionists, the clinical setting, whether PA tracking was incorporated, the use of electronic health/mobile health resources, and whether interventions effectively improved PA outcomes and HbA1c outcomes, respectively.

Data synthesis and analysis

For the studies that met our inclusion/exclusion criteria, each was scored independently for the PRECIS-2 domains and RE-AIM practical feasibility factors by at least two raters (KAL, IML, ALN, JCM, AGH). To ensure clarity of the scoring rubric and calibration of raters, the five raters performed two iterative cycles of review of other RCTs by the PRECIS-2 and RE-AIM practical feasibility factors. During the iterative cycles of review, any areas of discrepancy were resolved through team discussions with senior co-authors (JGR, ALD, REG). In addition, when dual rater scores differed by >1 point on any given PRECIS-2 domain/RE-AIM factor, the raters reassessed that domain/factor together and came to consensus on scores that differed by no more than 1 point. A third rater arbitrated any ratings that still differed by >1 point.

For each intervention (n = 46), we averaged the raters’ scores to calculate mean numerical scores for each domain/factor. We averaged the mean score of all nine PRECIS-2 domains to create a composite PRECIS-2 score that we used to rank the interventions by tertiles. Across all interventions, we also assessed the prevalence of reporting on practical feasibility factors and the mean score averaged across all practical feasibility factors.

We defined effective interventions as those that led to a statistically significant increase in a valid PA behavior outcome over the intervention period in the intervention group, as compared to the control group (P < 0.05). We also defined separately the interventions that improved HbA1c in the intervention vs. control groups (P < 0.05). A single author extracted the information on intervention effectiveness from the relevant publication, and a second author confirmed this information (KAL, IML).

We used logistic regression (SAS software version 9.4) to conduct a post hoc analysis of the relationship between PRECIS-2 scores and PA outcomes across all RCTs, considering the PRECIS-2 score as the independent variable and the PA outcome measure as the dichotomous dependent variable (yes/no for statistically significant change in PA). We used the identical method in a separate regression model to compare the relationship between PRECIS-2 and HbA1c outcomes across all RCTs.

RESULTS

The PubMed literature search revealed 4813 citations (Fig. 2). After screening by title and abstract, 4121 articles were excluded, leaving 692 full-text articles that were considered for inclusion. Of these, 651 did not meet the study inclusion criteria. The most common reasons for exclusion were supervised PA (183), lack of clinical integration (110), and study design other than RCT (92). The final RCTs selected (n = 38) represented 46 independent interventions, as some trials tested >1 intervention [14, 39–72]. The final RCTs selected were published between 1996 and 2015, as we did not identify any articles meeting our inclusion criteria that were published from 1980 to 1995.

Fig. 2.

Study flow diagram

PRECIS-2 scores

The composite PRECIS-2 score for each of the 46 interventions ranged from 3.0 to 4.8. The top tertile of interventions (n = 16) had composite PRECIS-2 scores ≥4.08. Mean scores varied across the nine PRECIS-2 domains: we found less pragmatic scores for PRECIS-2 domains of organization (3.12 [0.82]) and setting (3.24 [1.18]) and higher PRECIS-2 domain scores for flexibility of adherence (4.62 [0.50]) and primary analysis (4.83 [0.63]), with values expressed as (mean [SD]).

The majority of interventions studied (52.9%) reported effective improvements in PA outcomes. However, in the top tertile of interventions (n = 16) ranked by PRECIS-2 score, only 30.8% of interventions reported effective improvements in PA outcomes (Fig. 3a). As illustrated in Fig. 3a, our post hoc analysis suggested a possible inverse relationship between level of pragmatism by PRECIS-2 and PA effectiveness outcomes, albeit not statistically significant (P = 0.06). Just over one third of interventions studied (37.5%) effectively decreased HbA1c. In the top PRECIS-2 tertile, 31.3% of interventions effectively decreased HbA1c (Fig. 3b). The relationship between PRECIS-2 score and HbA1c effectiveness was not significant (P = 0.92).

Fig. 3.

PRECIS-2 score is plotted separately with regard to the RCT intervention effectiveness for physical activity (a) and hemoglobin A1c (b), respectively. PRECIS-2 Pragmatic-Explanatory Continuum Indicator Summary 2

Practical feasibility scores derived from RE-AIM

The majority of studies (62%) did not report on any of the practical feasibility factors that we assessed and thus received the lowest score of 1. Of the 38% of studies that reported on at least one practical feasibility domain, rates of reporting were highest for participant engagement (32%) and monetary costs of intervention (8%) (Appendix Fig. 5). No study authors reported on adaptation/change of the intervention to accommodate to clinical site needs or on unintended effects of the intervention. The composite practical feasibility for each intervention (n = 46) ranged from 1 to 2.2 on a possible five-point scale.

Fig. 5.

Prevalence of studies reporting on RE-AIM practical feasibility factors

Intervention characteristics

Given this review’s focus on identifying RCTs of therapies that may be translated into clinical practice, we report on both the typical PRISMA intervention characteristics as well as additional factors related to external validity (see Table 1 for top tertile and Appendix Table 3 for bottom two tertiles). Of the 46 interventions included, 38% used PA tracking and/or feedback methods as an intervention strategy. The PA tracking/feedback methods used were diverse, ranging from web-based communication to one-on-one visits with study personnel to review PA diaries. Participants were given pedometers, accelerometers, or other unspecified “activity monitors” to use for PA tracking in 21% of interventions. Only 11% of interventions utilized electronic health (eHealth) modalities such as the internet or mobile phones as part of the intervention. Studies took place in more than 10 different countries between 1990 and 2015, and the study duration ranged from 2.8 months to 8 years.

Table 1.

Characteristics related to external validity for trials in PRECIS-2 top tertile

First author, year, location	Participants (n), age, females (%), non-white (%), HbA1c (mean)	Clinical setting	Intervention duration, intervention frequency, follow-up	Intervention staff type, level of training
Sperl-Hillen (IE) et al., 2011, USA	I: n = 246, 61.6 yr., 49.4% f, 37.3% n-w, A1c—NR C: n = 134, 63.3 yr., 46.3% f, 32.8% n-w, A1c—NR	2 large medical groups in NM and MN	Duration: 3.8 mo. Frequency: Three monthly 1-h sessions Follow-up: 6.8 mo.	CDEs (RNs and RDs)
Nielsen et al., 2006, Denmark	I: n = 459, 65.5 yr., 47.6% f, NR n-w, A1c—10.2% C: n = 415, 65.3 yr., 46.9% f, NR n-w, A1c—10.2%	187 solo practices and 124 group practices	Duration: 6 yrs. Frequency: consultations every 3 mo. Follow-up: 6 yrs.	GPs
Sperl-Hillen (GE) et al., 2011, USAa	I: n = 243, 61.2 yr., 51.0% f, 33.5% n-w, A1c—NR C: n = 134, 63.3 yr., 46.3% f, 32.8% n-w, A1c—NR	2 large medical groups in NM and MN	Duration: 2.8 mo. Frequency: Four 2-h sessions at 1-wk. intervals Follow-up: 6.8 mo.	CDEs (RNs and RDs)
Adachi et al., 2013, Japan	I: n = 100, 60.4 yr., 55% f, NR n-w, A1c—NR C: n = 93, 62.3 yr., 58% f, NR n-w, A1c—NR	Multiple clinics in Kanagawa prefecture, Japan	Duration: 6 mo. Frequency: 3–4 sessions Follow-up: 6 mo.	Trained RDs
Babamoto (CM) et al., 2009, USA	I: n = 60, 50.0 yr., 52% f, NR n-w, A1c—8.5% C: n = 54, 50.0 yr., 78% f, NR n-w, A1c—9.5%	3 inner city family health centers in Los Angeles, CA	Duration: 6 mo. Frequency: “usually seen on a monthly basis,” follow-up calls “as determined by the case manager” Follow-up: 6 mo.	Two linguistically competent and culturally sensitive RNs
Davies et al., 2008, England and Scotland	I: n = 437, 59.0 yr., 47% f, 6% n-w, A1c—8.3% C: n = 387, 60.0 yr., 43% f, 6% n-w, A1c—7.9%	13 primary care sites	Duration: 12 mo. Frequency: One 6-h session delivered once at diagnosis Follow-up: 12 mo.	Registered healthcare professionals
Mash et al., 2014, South Africa	I: n = 710, 55.8 yr., 71.5% f, NR n-w, A1c—8.9% C: n = 860, 56.4 yr., 75.7% f, NR n-w, A1c—9.3%	34 community health centers in the working class areas of Cape Town Metropole, South Africa	Duration: approximately 6 mo. Frequency: Four 60-min sessions Follow-up: 12 mo.	Health promoters recruited from district health services and trained over 6 days
Shibayama et al., 2007, Japan	I: n = 67, 61.0 yr., 34.8% f, NR n-w, A1c—7.3% C: n = 67, 62.0 yr., 34.8% f, NR n-w, A1c—7.4%	OP clinic, Department of Diabetes and Metabolism, University of Tokyo Hospital	Duration: 1 yr. Frequency: monthly counseling Follow-up: 1 yr.	CEN in diabetes nursing
Irwig et al., 2012, USA	I: n = 52, 54.8 yr., 50% f, 86.5% n-w, A1c—10.2% C: n = 51, 57.2 yr., 56.9% f, 74.5% n-w, A1c—9.7%	Single academic multi-specialty clinic	Duration: 9 mo. Frequency: every 3 mo. (total of 4 visits) Follow-up: 9 mo.	MDs (2 endocrinologists and 3 internists)
Jansink et al., 2013, Netherlands	I: n = 229, 64.1 yr., 44.1% f, NR n-w, A1c—7.8% C: n = 292, 63.9 yr., 46.0% f, NR n-w, A1c—7.7%	58 general practices	Duration: 14 mo. Frequency: Four half-day training sessions spread over 6 mo; also received quarterly phone calls Follow-up: 14 mo.	Primary care nurses
De Greef et al. (GP), 2011, Belgium	I: n = 22, 66.6 yr., 22.7% f, NR n-w, A1c—7.23% C: n = 24, 66.0 yr., 29.2% f, NR n-w, A1c—7.0%	Three Belgian general practices	Duration: 12 wks. Frequency: Three individual 15-min face-to-face consultations Follow-up: 12 wks.	GPs
Christian et al., 2008, USA	I: n = 155, 53.0 yr., 65% f, 100% n-w, A1c—8.08% C: n = 155, 53.4 yr., 68% f, 100% n-w, A1c—8.29%	2 large urban community-based health centers in CO	Duration: 12 mo. Frequency: visits every 3 mo. Follow-up: 12 mo.	MDs
Trento et al., 2010, Italy	I: n = 421, 69.0 yr., 52.0% f, NR n-w, A1c—7.75% C: n = 394, 69.6 yr., 46.4% f, NR n-w, A1c—7.81%	13 hospital-based diabetes clinics	Duration: 4 yrs. Frequency: 7 1-h group care sessions every 3 months over 2 yr. and repeated; individual consultations at least yearly Follow-up: 4 yrs.	MDs and an educationist (M.T.)
Lee et al., 2011, Hong Kong	I: n = 84, NR yr., 60.9% f, NR n-w, A1c—8.18% C: n = 73, NR yr., 63.1% f, NR n-w, A1c—8.04%	General OP clinic, Hospital Authority New Territory East Cluster of Hong Kong	Duration: 6 wks. Frequency: weekly 2.5-h DM self-management course Follow-up: 28 wks.	Social worker accredited as a trainer for the self-management program
DiLoreto et al., 2003, Italy	I: n = 182, 62.0 yr., 51.6% f, NR n-w, A1c—7.6% C: n = 158, 61.6 yr., 53.8% f, NR n-w, A1c—7.7%	OP diabetes clinic	Duration: 2 yrs. Frequency: initial 30-min counseling session, home phone call 1 mo. later; 15-min appointments in clinic every 3 mo. Follow-up: 2 yrs.	MDs
Glasgow (CASM) et al., 2012, USA	I: n = 169, 58.7 yr., 44.6% f, 25.9% n-w, A1c—NR C: n = 132, 58.7 yr., 51.5% f, 29.4% n-w, A1c—8.16%	Five primary care clinics within Kaiser Permanente Colorado	Duration: 12 mo. Frequency: initial website orientation; action plan at 6 wks; periodic motivational calls Follow-up: 12 mo.	Research staff member
First author, year, location	PA tracking method/feedback provided	eHealth	Control group	Effectiveness (PA), effectiveness (HbA1c)	PRECIS-2 score
Sperl-Hillen (IE) et al., 2011, USA	N/A	N/A	Usual care	PA: yes HbA1c: yes	4.78
Nielsen et al., 2006, Denmark	N/A	N/A	Usual care	PA: no HbA1c: yes	4.61
Sperl-Hillen (GE) et al., 2011, USAa	N/A	N/A	Usual care	PA: no HbA1c: no	4.50
Adachi et al., 2013, Japan	N/A	N/A	General advice from RDs	PA: NR HbA1c: yes	4.50
Babamoto (CM) et al., 2009, USA	N/A	N/A	Usual care	PA: no HbA1c: no	4.44
Davies et al., 2008, England and Scotland	N/A	N/A	Enhanced standard care	PA: no HbA1c: no	4.33
Mash et al., 2014, South Africa	N/A	N/A	Usual care	PA: no HbA1c: no	4.33
Shibayama et al., 2007, Japan	Feedback provided at monthly sessions	N/A	Usual care	PA: NR HbA1c: no	4.28
Irwig et al., 2012, USA	N/A	N/A	Completed questionnaires only	PA: no HbA1c: no	4.28
Jansink et al., 2013, Netherlands	N/A	N/A	Usual care	PA: no HbA1c: no	4.22
De Greef et al. (GP), 2011, Belgium	Pedometers used to track progress and to encourage discussions with the GP	N/A	Usual care	PA: no HbA1c: no	4.17
Christian et al., 2008, USA	Results from computer-based assessment used to tailor intervention	“Computer expert system”	Usual care	PA: yes HbA1c: no	4.17
Trento et al., 2010, Italy	N/A	N/A	Usual care	PA: NR HbA1c: yes	4.17
Lee et al., 2011, Hong Kong	N/A	N/A	Usual care	PA: no HbA1c: no	4.11
DiLoreto et al., 2003, Italy	Phone calls at 1 month to determine level of PA and reinforce instructions	N/A	Usual care	PA: yes HbA1c: yes	4.11
Glasgow (CASM) et al., 2012, USA	Participants recorded progress on daily goals and received feedback using website	“My Path toHealthy Life” (Mi Camino a La Vida Sana) website	Enhanced usual care	PA: yes HbA1c: no	4.08

C control group, CDE(s) certified diabetes educator(s), hr./hrs. hour/hours, I intervention group, MD(s) doctor(s), mo. month/months, N/A not applicable, NR not reported, OP outpatient, PA physical activity, PCP(s)/GP(s) primary care physician(s)/general practitioner(s), RD(s) registered dietitian(s), RN(s) registered nurse(s), wk./wks. week/weeks, yr./yrs. year/years

^aA secondary intervention from an already noted study

Table 3.

Characteristics related to external validity for trials in PRECIS-2 bottom two tertiles

First author, year, location	Participants (n), age, females (%), non-white (%), HbA1c (mean)	Clinical setting	Intervention duration, intervention frequency, follow-up
Middle tertile
Babamoto (CHW) et al., 2009, USAa	I: n = 75, 51 yr., 64% f, NR n-w, A1c—8.6% C: n = 54, 50 yr., 78% f, NR n-w, A1c—9.5%	3 inner city family health centers in Los Angeles, CA	Duration: 6 mo. Frequency: 11.3 sessions (in person and telephone contact) Follow-up: 6 mo.
De Greef (Group) et al., 2011, Belgiuma	I: n = 21, 70.0 yr., 38.1% f, NR n-w, A1c—7.12% C: n = 24, 66.0 yr., 29.2% f, NR n-w, A1c—7.00%	Three Belgian general practices	Duration: 12 wks. Frequency: three 90-min group counseling sessions Follow-up: 12 wks.
Van der Weegen (SSP w/ tool) et al., 2015, Netherlands	I: n = 65, 57.5 yr., 52.3% f, NR n-w, A1c—NR C: n = 68, 59.2 yr., 54.4% f, NR n-w, A1c—NR	24 general practices in the south of the Netherlands	Duration: 6 mo. Frequency: 4 consultations with practice nurse occurring during first wk., after 2 wks., after 8–12 wks., and after 16–24 wks. Follow-up: measured at baseline after the intervention (4–6 mo.), and 3 mo. thereafter
Keyserling (Group B) et al., 2002, USA	I: n = 66, 59.8 yr., 100% f, 100% n-w, A1c—11.1% C: n = 67, 59.2 yr., 100% f, 100% n-w, A1c—11.3%	Primary care practices in central NC (5 community health centers, 1 staff model health maintenance organization, 1 general medicine clinic at an academic health center)	Duration: 6 mo. Frequency: 4 monthly visits Follow-up: 12 mo.
Van der Weegen (SSP) et al., 2015, Netherlandsa	I: n = 66, 56.9 yr., 47.0% f, NR n-w, A1c—NR C: n = 68, 59.2 yr., 54.4% f, NR n-w, A1c—NR	24 general practices in the south of the Netherlands	Duration: 6 mo. Frequency: 4 consultations with practice nurse occurring during first wk., after 2 wks., after 8–12 wks., and after 16–24 wks. Follow-up: measured at baseline after the intervention (4–6 mo.), and 3 mo. thereafter
Edelman et al., 2015, USA	I: n = 193, 57.8 yr., 54.4% f, 50.8% n-w, A1c—9.2% C: n = 184, 59.6 yr., 54.9% f, 49.5% n-w, A1c—9.0%	9 primary care practices in the Duke Clinical Research Institute Primary Care Research Consortium	Duration: 2 yrs. Frequency: phone calls every 2 mo. (12 total) Follow-up: 2 yrs.
Naik et al., 2011, USA	I: n = 45, 63.82 yr., NR f, 33.3% n-w, A1c—8.86% C: n = 42, 63.45 yr., NR f, 28.6% n-w, A1c—8.74%	Michael E. DeBakey Veterans Affairs Medical Center in Houston, Texas	Duration: 3 mo. Frequency: 4 group sessions every 3 wks. Follow-up: 3 mo.
Glasgow (CASM+) et al., 2012, USAa	I: n = 162, 58.7 yr., 53.7% f, 29.3% n-w, A1c—NR C: n = 132, 58.7 yr., 51.5% f, 29.4% n-w, A1c—8.16%	Five primary care clinics within Kaiser Permanente Colorado	Duration: 12 mo. Frequency: three 120-min group sessions after week 6; follow-up telephone calls at wks. 2 and 8; periodic motivational calls Follow-up: 12 mo.
Francosi et al., 2011, Italy	I: n = 46, 48.9 yr., 30.4% f, NR n-w, A1c—7.9% C: n = 16, 48.7 yr., 12.5% f, NR n-w, A1c—7.9%	3 OP diabetes clinics	Duration: 6 mo. Frequency: Face-to-face encounter every 3 mo. with diabetes nurses and additional telephone contact monthly Follow-up: 6 mo.
Maindal et al., 2014, UK	I: n = 322, 62.0 yr., 47.2% f, NR n-w, A1c—6.3% C: n = 187, 62.0 yr., 46.0% f, NR n-w, A1c—6.2%	33 general practices in a large Danish county	Duration: 3 mo. Frequency: 2 individual counseling interviews and 8 group sessions, totaling 18 h. Follow-up: 3 yrs.
Jarab et al., 2012, Jordan	I: n = 85, 63.4 yr., 42.4% f, NR n-w, A1c—8.5% C: n = 86, 65.3 yr., 44.2% f, NR n-w, A1c—8.4%	OP diabetes clinic at the 762-bed Royal Medical Services Hospital	Duration: 8 wks. Frequency: baseline face-to-face objective directed education and weekly telephone follow-up calls Follow-up: 6 mo.
Schillinger (ATSM) et al., 2009, USA	I: n = 112, 55.9 yr., 58.0% f, 87.5% n-w, A1c—9.3% C: n = 114, 55.8 yr., 55.3% f, 93.8% n-w, A1c—9.8%	9 clinics in the San Francisco Department of Public Health’s Community Health Network (CHNSF)	Duration: 9 mo. Frequency: weekly automated phone calls w/ follow-up calls from care manager if necessary Follow-up: 1 yr.
Trento et al., 2008, Italy	I: n = 25, 64.6 yr., 48.0% f, NR n-w, A1c—8.0% C: n = 24. 68.1 yr., 33.3% f, NR n-w, A1c—8.0%	Department of Internal Medicine, University of Turin	Duration: 2 yrs. Frequency: 40–50 min. Group sessions very 3–4 mo. Follow-up: 2 yrs.
Trento et al., 2002, Italy	I: n = 56, 62.0 yr., 51.8% f, NR n-w, A1c—7.4%’ C: n = 56, 61.0 yr., 39.3% f, NR n-w, A1c—7.4%	Diabetes clinic, Department of Internal Medicine, University of Turin	Duration: 4 yrs. Frequency: group educational sessions once every 3 mo. for 2 yrs.; 7 sessions spread over yrs. 3–4 Follow-up: 4 yrs.
Barratt et al., 2008, UK	I: n = 27, NR yr., NR f, NR n-w, A1c—9.6% C: n = 26, NR yr., NR f, NR n-w, A1c—9.7%	Two tertiary hospitals in southeast England	Duration: 6 mo. Frequency: one 90-min. session and five 30-min. sessions Follow-up: 6 mo.
Bottom tertile
Kim (PM) et al., 2006, South Korea	I: n = 22, NR yr., NR f, NR n-w, A1c—7.51% C: n = 23, NR yr., NR f, NR n-w, A1c—7.87%	OP diabetes clinic at a large university hospital in South Korea	Duration: 12 wks. Frequency: 2× in first 2 wks., again at 6 wks. Follow-up: 12 wks.
Kim (WB) et al., 2006, South Koreaa	I: n = 28, NR yr., NR f, NR n-w, A1c—7.99% C: n = 23, NR yr., NR f, NR n-w, A1c—7.87%	OP diabetes clinic at a large university hospital in South Korea	Duration: 12 wks. Frequency: 2× in first 2 wks., again at 6 wks. Follow-up: 12 wks.
Keyserling (Group A) et al., 2002, USAa	I: n = 67, 58.5 yr., 100% f, 100% n-w, A1c—10.8% C: n = 67, 59.2 yr., 100% f, 100% n-w, A1c—11.3%	Primary care practices in central NC (5 community health centers, 1 staff model health maintenance organization, 1 general medicine clinic at an academic health center)	Duration: 12 mo. Frequency: 4 monthly visits, 3 group sessions, and 4 monthly phone calls Follow-up: 12 mo.
Schillinger (GMV) et al., 2009, USAa	I: n = 113, 56.5 yr., 63.7% f, 91.1% n-w, A1c—9.4% C: n = 114, 55.8 yr., 55.3% f, 93.8% n-w, A1c—9.8%	9 clinics in the San Francisco Department of Public Health’s Community Health Network (CHNSF)	Duration: 9 mo. Frequency: monthly group meetings (90 min. each) Follow-up: 1 yr.
Anderson et al., 2010, USA	I: n = 146, NR yr., 58.9% f, 72.6% n-w, A1c—7.6% C: n = 149, NR yr., 57.1% f, 73.8% n-w, A1c—8.4%	2 largest Community Health Center, Inc. locations in CT	Duration: 1 yr. Frequency: weekly/bi-weekly/monthly (depending on risk stratification) Follow-up: 1 yr.
Lim et al., 2015, South Korea	I: n = 50, 64.3 yr., 20% f, NR n-w, A1c—8.1% C: n = 50, 65.8 yr., 30% f, NR n-w, A1c—7.9%	OP clinic, Bundang Hospital, Seoul National University	Duration: 6 mo. Frequency: 1-h diet and exercise counseling at baseline, 3, and 6-month visits Follow-up: 6 mo.
Thoolen et al., 2009, Netherlands	I: n = 78, 62.0 yr., 36.0% f, NR n-w, A1c—NR C: n = 102, 61.9 yr., 45.0% f, NR n-w, A1c—NR	General practices in southwest Netherlands	Duration: 12 wks. Frequency: two individual and four group sessions (2 h./session) with nurse Follow-up: 3 and 12 mo.
Taylor et al., 2003, USA	I: n = 84, 55.5 yr., 50% f, 33.3% n-w, A1c—9.5% C: n = 85, 54.8 yr., 44.7% f, 43.5% n-w, A1c—9.5%	Kaiser Permanente Medical Center in Santa Clara, CA	Duration: 1 yr. Frequency: 90 min. RN consultation; weekly 1- to 2-h group classes for 4 wks; telephone follow-up before 4th group session and at wks. 5, 8, 12, 16, 20, 28, 36, and 44 (∼15 mins./call) Follow-up: 1 yr.
Moriyama et al., 2009, Japan	I: n = 42, 66.4 yr., 59.5% f, NR n-w, A1c—7.5% C: n = 23, 65.2 yr., 43.5% f, NR n-w, A1c—7.4%	2 hospitals with less than 200 beds in a Japanese city	Duration: 1 yr. Frequency: monthly ∼30-min. sessions; phone calls from nurse educator every 2 wks. Follow-up: 1 yr.
Van Dyck et al., 2011, Belgium	I: n = 60, 62.37 yr., NR f, NR n-w, A1c—NR C: n = 32, 60.59 yr., NR f, NR n-w, A1c—NR	Endocrinology Department of the Ghent University Hospital	Duration: 24 wks. Frequency: face-to-face 30-min. session with psychologist; phone call every 2 wks. For first month, every 4 wks. For next 20 wks. (7 calls, ∼20 mins. each) Follow-up: 1 yr.
Kim et al., 2011, South Korea	I: n = 21, 56.62 yr., 47.6% f, 100% n-w, A1c—7.40% C: n = 22, 54.68 yr., 31.8% f, 100% n-w, A1c—7.41%	OP diabetic center at a large university hospital in South Korea	Duration: 16 wks. Frequency: 60–90-min. initial counseling session; 30–40-min. follow-up counseling session every 2 mo; 10–30-min. weekly telephone calls Follow-up: 16 wks.
Sone et al., 2010, Japan	I: n = 1017, 58.5 yr., 46.0% f, NR n-w, A1c—7.8% C: n = 1016, 58.6 yr., 47.0% f, NR n-w, A1c—7.9%	59 university and general hospitals that specialize in diabetes care	Duration: 8 yrs. Frequency: 15-min. telephone counseling sessions at least once every 2 wks; 5–10-min. extra during each clinic visit Follow-up: 7.8 yrs.
Uusitupa et al., 1996, Finland	I: n = 38, NR yr., 44.7% f, NR n-w, A1c—8.4% C: n = 40, NR yr., 40.0% f, NR n-w, A1c—9.0%	OP clinic of Department of Medicine, Kuopio University Hospital	Duration: 1 yr. Frequency: six visits to OP clinic at 2-mo. intervals Follow-up: 1 yr.
Gaede et al., 1999, Denmark	I: n = 80, 54.9 yr., 21.3% f, NR n-w, A1c—8.4% C: n = 80, 55.2 yr., 30.0% f, NR n-w, A1c—8.8%	Steno Diabetes Center	Duration: approximately 4–5 yrs., or when study endpoint was reached Frequency: every 3-mo. follow-up: At 2 and 4 yrs.
Huffman et al., 2010, USA	I: n = 38, NR yr., 0% f, NR n-w, A1c—NR C: n = 46, NR yr., 0% f, NR n-w, A1c—NR	Durham Veterans Affairs Medical Center in NC	Duration: 1 yr. Frequency: baseline counseling session; 3 biweekly phone calls over first 6 wks., monthly phone calls for study duration Follow-up: 1 yr.
First author, year, location	Intervention staff type, level of training	PA tracking method/feedback provided	eHealth
Middle tertile
Babamoto (CHW) et al., 2009, USAa	CHWs	N/A	N/A
De Greef (Group) et al., 2011, Belgiuma	Clinical psychologist with a background in behavior change strategies	Pedometers used to track progress and to encourage discussions with the behavioral expert	N/A
Van der Weegen (SSP w/ tool) et al., 2015, Netherlands	Practice nurse trained in delivery of the intervention	Participants used activity monitors to track activity on mobile phone and web app. Personal activity goals set based on dialog sessions and activity results	It’s LiFe! monitoring and feedback tool consisting of a three-dimensional activity monitor, a mobile phone app, and a web app
Keyserling (Group B) et al., 2002, USA	Nutritionists	N/A	N/A
Van der Weegen (SSP) et al., 2015, Netherlandsa	Practice nurse trained in delivery of the intervention	Participants recorded PA in diaries. They discussed their progress and made individual goals during sessions with practice nurses.	N/A
Edelman et al., 2015, USA	RN with extensive experience in case management	N/A	N/A
Naik et al., 2011, USA	PCPs trained in goal setting and action planning methodology	Participants received feedback on specific goals during one-on-one sessions with PCP	N/A
Glasgow (CASM+) et al., 2012, USAa	“Research project staff member,” diabetes care coordinator, nutritionist, bilingual family MD	Participants recorded progress on daily goals and received feedback using website. Patients also received follow-up calls at 2 and 8 weeks to discuss the their action plans	“My Path to Healthy Life” (Mi Camino a La Vida Sana) website
Francosi et al., 2011, Italy	Diabetes RNs trained in diabetes care during a 1-day session	Feedback provided during monthly phone calls	N/A
Maindal et al., 2014, UK	RNs, RDs, physiotherapists, and GPs who received formal training in autonomy support, participant-centered communication and action plan support	N/A	N/A
Jarab et al., 2012, Jordan	MDs and pharmacists	N/A	N/A
Schillinger (ATSM) et al., 2009, USA	Nurse care manager	Participant responses to automated phone calls triggered automated health education messages and/or nurse phone follow-up	N/A
Trento et al., 2008, Italy	RD and RNs trained in management of T2D	N/A	N/A
Trento et al., 2002, Italy	MDs and an educationist	N/A	N/A
Barratt et al., 2008, UK	RDs	Progress on personal goals discussed at each appointment	N/A
Bottom tertile
Kim (PM) et al., 2006, South Korea	Research RNs	PA frequency, intensity, and duration monitored weekly by diaries and kcal-pedometer. Research nurses provided feedback at clinic visits or with phone calls	N/A
Kim (WB) et al., 2006, South Koreaa	Research RNs	PA frequency, intensity, and duration monitored weekly by diaries and kcal-pedometer. Research nurses provided feedback at clinic visits or with phone calls. Web-based psychological and physical readiness questionnaires used to assess current stages of PA	Web site included stage-based personalized sections on goal setting, activity planning, determining target heart rates, and questionnaires
Keyserling (Group A) et al., 2002, USAa	Community diabetes advisors and nutritionists	N/A	N/A
Schillinger (GMV) et al., 2009, USAa	PCP and health educator	N/A	N/A
Anderson et al., 2010, USA	RNs trained in intervention delivery	N/A	N/A
Lim et al., 2015, South Korea	RD, exercise specialist, exercise physiologist	PA monitor results linked to the main server to provide tailored service	Dedicated website containing a glucose control section, diet control section, physical activity section, and an integrated widget. PA monitors and glucometers linked with the u-healthcare system
Thoolen et al., 2009, Netherlands	RN	N/A	N/A
Taylor et al., 2003, USA	RN	N/A	N/A
Moriyama et al., 2009, Japan	Nurse educators	Nurse educators contacted participants every 2 weeks to monitor exercise goals	N/A
Van Dyck et al., 2011, Belgium	Psychologists	Participants asked to keep a pedometer diary to track progress and encourage discussion with psychologist	N/A
Kim et al., 2011, South Korea	RN/MD Researchers	Weekly exercise and dietary logs indicated frequency, duration, and kilocalories for energy expenditure by an accelerometer	N/A
Sone et al., 2010, Japan	MDs, RNs, RDs, psychotherapists, and other co-medical staff	Pedometers used for objective assessment of PA	N/A
Uusitupa et al., 1996, Finland	MDs, RNs, and nutritionists	PA monitored by daily activity records	N/A
Gaede et al., 1999, Denmark	MD, RN, and RD	N/A	N/A
Huffman et al., 2010, USA	Health counselors and PCPs	Pedometers used to track PA. Regular phone calls to check in with participants and adjust PA goals as needed.	N/A
First author, year, location	Control group	Effectiveness (PA); effectiveness (HbA1c)	PRECIS-2 score
Middle tertile
Babamoto (CHW) et al., 2009, USAa	Usual care	PA: no HbA1c: no	4.06
De Greef (Group) et al., 2011, Belgiuma	Usual care	PA: yes HbA1c: no	4.06
Van der Weegen (SSP w/ tool) et al., 2015, Netherlands	Usual care	PA: yes HbA1c: NR	4.05
Keyserling (Group B) et al., 2002, USA	Educational pamphlets sent by mail	PA: yes HbA1c: no	4.00
Van der Weegen (SSP) et al., 2015, Netherlandsa	Usual care	PA: no HbA1c: NR	4.00
Edelman et al., 2015, USA	Non-tailored phone calls	PA: no HbA1c: no	3.89
Naik et al., 2011, USA	“Traditional education intervention”	PA: NR HbA1c: yes	3.89
Glasgow (CASM+) et al., 2012, USAa	Enhanced usual care, including additional tailored web-based feedback on health behaviors	PA: yes HbA1c: no	3.85
Francosi et al., 2011, Italy	Standard counseling and routine follow-up	PA: NR HbA1c: yes	3.83
Maindal et al., 2014, UK	Usual care	PA: no HbA1c: no	3.83
Jarab et al., 2012, Jordan	Usual care	PA: yes HbA1c: yes	3.83
Schillinger (ATSM) et al., 2009, USA	Usual care	PA: yes HbA1c: no	3.83
Trento et al., 2008, Italy	Usual care	PA: NR HbA1c: yes	3.83
Trento et al., 2002, Italy	Usual care	PA: NR HbA1c: no	3.81
Barratt et al., 2008, UK	Usual care	PA: NR HbA1c: no	3.78
Bottom tertile
Kim (PM) et al., 2006, South Korea	Usual care	PA: yes HbA1c: yes	3.78
Kim (WB) et al., 2006, South Koreaa	General information sheet from the clinic. Participants visited the clinic once at the beginning of the intervention period	PA: yes HbA1c: yes	3.72
Keyserling (Group A) et al., 2002, USAa	Educational pamphlets sent by mail	PA: yes HbA1c: no	3.69
Schillinger (GMV) et al., 2009, USAa	Usual care	PA: no HbA1c: no	3.67
Anderson et al., 2010, USA	Usual care	PA: NR HbA1c: no	3.67
Lim et al., 2015, South Korea	Dedicated website containing a glucose control section and diet control section	PA: yes HbA1c: yes	3.67
Thoolen et al., 2009, Netherlands	Brochure on diabetes self-management	PA: yes HbA1c: NR	3.63
Taylor et al., 2003, USA	Educational pamphlets and instructions to continue regular care with PCP	PA: NR HbA1c: yes	3.61
Moriyama et al., 2009, Japan	Educational textbook and asked to consult their physician as usual	PA: NR HbA1c: yes	3.56
Van Dyck et al., 2011, Belgium	Usual care	PA: yes HbA1c: NR	3.56
Kim et al., 2011, South Korea	Usual care	PA: yes HbA1c: no	3.50
Sone et al., 2010, Japan	Usual care	PA: yes HbA1c: no	3.46
Uusitupa et al., 1996, Finland	Usual care	PA: no HbA1c: no	3.29
Gaede et al., 1999, Denmark	Usual care	PA: no HbA1c: yes	3.17
Huffman et al., 2010, USA	Usual care	PA: yes HbA1c: NR	3.00

C control group, CHW(s) community health worker(s), hr./hrs. hour/hours, I intervention group, MD(s) doctor(s), min minute/minutes, mo. month/months, N/A not applicable, NR not reported, OP outpatient, PA physical activity, PCP(s)/GP(s) primary care physician(s)/general practitioner(s), RD(s) registered dietitian(s), RN(s) registered nurse(s), wk./wks. week/weeks, yr./yrs. year/years

^aA secondary intervention from an already noted study

Risk of bias

For each of the seven areas within the Cochrane risk of bias tool [73], two authors (IML and AGH) independently assessed the risk of bias for each included study. We resolved differences in reviewer ratings through discussion. In each of the seven categories, less than 6% of studies had a high risk of bias (Appendix Fig. 6). As no studies had a high risk of bias on all seven categories, we did not exclude any based on these bias risk assessments.

Fig. 6.

Prevalence of Cochrane risk of bias categories

CONCLUSIONS

This systematic review identified behavioral interventions that are feasible to implement in clinical practice settings and that effectively increased PA for people with T2D. The range of PRECIS-2 scores from 3.0 to 4.8 across the 46 interventions studied demonstrates a range of pragmatism from modestly pragmatic to highly pragmatic. It is of possible concern that the level of study pragmatism tended to be inversely related to improvements in PA, albeit of borderline statistical significance (P = 0.06). This suggests that PA interventions conducted in more real-world settings may be less effective when compared to studies in more idealized and heavily resourced contexts. Nevertheless, among the behavioral interventions that we ranked in the most pragmatic tertile by composite PRECIS-2 scores, four interventions by Sperl-Hillen et al. [39], Christian et al. [50], Di Loreto et al. [53], and Glasgow et al. [54] effectively improved PA outcomes—two of these four effective PA interventions also improved HbA1c outcomes [31, 45].

A key question that emerges from this review is how do we identify the moderators and key components of successful interventions that are highly pragmatic and translational? The existing literature on predictors of effective PA interventions has already identified several key factors that predict improved PA outcomes, including tracking/self-monitoring of PA, setting PA goals, increasing self-efficacy for PA, enlisting social support to be active, and utilizing behavioral theory-informed intervention techniques [74–76]. What remains unclear is how these evidence-based techniques may be optimally implemented with pragmatic approaches in clinical care. This review was not a meta-analysis and cannot statistically dissect the factors that moderated success among the four interventions in the top tertile by PRECIS-2 scores that also improved PA outcomes. Nevertheless, we did note some common themes among the four highly pragmatic and effective interventions that may be instructive. The majority of these interventions used PA tracking that allowed PA intervention content to be tailored to individual activity levels; interventions also tended to be delivered over a duration of 12 months or longer. Other than the regularity of PA tracking/feedback, the intervention approaches were fairly diverse, but commonly used simple tools to help participants and healthcare providers jointly identify and track behavior patterns and behavioral predictors of PA levels. These tools included a simplified form of in-person motivational interviewing [53], a computerized self-assessment that automatically generated individualized and tailored feedback [50, 54], or the use of a “conversation map” of diabetes self-management challenges [39].

While we already know much about what factors are necessary for effective interventions, we need to better understand how to ensure that these techniques are effective when delivered pragmatically in clinical settings. For example, both the effectiveness on PA/HbA1c outcomes and the pragmatism of Sperl-Hillen’s intervention were superior when delivered 1:1 rather than using a group visit delivery format [39]. This finding stands in stark contrast to the existing literature that suggests that group and individual diabetes self-management education are equivalent [77] and suggests that further pragmatic trials to assess the comparative effectiveness of individual and group self-management education of T2D in usual practice may be warranted. Sperl-Hillen’s findings from their pragmatic trial may suggest that individual interventions to improve T2D self-management are especially important when individual motivators and barriers are more heterogeneous, but another possibility is that this was a chance finding.

As another example, in the study conducted by De Greef et al., the effects of the same behavioral intervention were significantly different based on who delivered the intervention and the dose of intervention—a trained behavioral expert providing counseling over three separate 90-min sessions effectively increased PA but a primary care clinician providing the same counseling content over three separate 15-min clinic visits did not increase PA [49]. Other review articles have lamented the lack of data on fidelity of intervention delivery in RCTs and have noted this as a limitation to identifying key moderators of PA interventions [76]. The highly pragmatic and effective intervention by Di Loreto et al. that we studied in the present review balanced the competing clinical demands and fidelity to intervention techniques by condensing several effective behavioral change techniques into a simple yet tailored counseling checklist which identified enjoyable and appropriate PA options at the individual level and encouraged patients to continue with these programs through the use of social support, problem-solving, and PA tracking [45]. Another important component of this intervention that used the counseling checklist was to deliver the intervention directly after a 30-min clinical in-person assessment of T2D that allowed the opportunity to address competing clinical concerns immediately prior to delivering the behavioral counseling [53].

In recent years, the proliferation of mobile phone users coupled with the increasing popularity of PA tracking devices that sync with phones/computers have provided novel opportunities for eHealth applications. Diabetes researchers have also been encouraged by the promise of eHealth to enhance the reach and effectiveness of T2D self-management interventions, including PA interventions [78, 79]. Thus, we were intrigued to notice that only 6 of the 46 interventions that we studied utilized any measures related to eHealth [50, 54, 55, 64, 80]. However, it is important to note that most of the interventions we studied were conducted before the recent explosion of smartphone and wearable technology for tracking PA [81, 82]; in the coming years, these percentages may change considerably as more researchers begin to embrace eHealth. Our review covered a broad range of publication dates (1996–2015), which introduces a concern for any secular PA trends over that timeframe. PA guidelines for patients with type 2 diabetes over that time span did remain relatively unchanged, promoting PA of 30 min of moderate intensity on most days of the week (≥120 min/week) in 1996 and recommendations of 150 min of weekly moderate intensity activity in 2015. Perhaps more importantly, the increase in PA tracking use over the past decade may have allowed patients’ greater ability to simply self-monitor their PA levels. While they represented the minority of interventions that we studied, 100 and 40% of the aforementioned eHealth interventions improved PA and HbA1c outcomes, respectively. Although the eHealth interventions we studied uniformly improved PA outcomes, it is interesting to note that only two of the six eHealth interventions that we studied were in the top tertile by PRECIS-2 scores. To improve the pragmatism of future PA interventions using eHealth methods, one opportunity may be to enhance the linkage of PA tracking to existing electronic medical records [83]. This linking would allow clinicians and patients to more readily monitor physical activity levels and communicate regarding new activity goals. As wearable PA tracking devices and technology use continuously expand [81, 82], this is an area that will see growth and constant adaptations year after year.

To enhance the translation of effective studies into real-world settings, it is also important for researchers to report on outcomes other than effectiveness that will influence future implementation. The PRECIS-2 tool incorporates many outcomes that predict the future implementation of interventions, but we also assessed certain additional practical feasibility factors derived from RE-AIM, such as costs and program sustainability, that are not measured by the PRECIS-2 tool. It was concerning that almost no studies reported on these dimensions important to potential adopters of PA interventions in clinical settings. As in this review, the other literature assessing the feasibility to translate interventions into clinical practice found that authors severely underreport important measures from RE-AIM and PRECIS-2, such as intervention costs, “Reach” in RE-AIM that corresponds to the PRECIS-2 domains of eligibility and recruitment, as well as the organizational burden of the staff, training, resources, and infrastructure required to deliver an intervention [33, 84–87]. Without transparent reporting on the cost, adaptations made, and sustainability of PA interventions, it will be impossible for health systems to determine the value of these interventions for their clinical populations with T2D. Recently, the Standards for Reporting Implementation Studies (StaRI) initiative developed a checklist of items to report on for implementation science research [88]; with the future adaptation of these standards by researchers and journal editors, transparent reporting on factors relevant to stakeholders should allow health systems to more readily identify feasible programs for adoption.

This systematic review is innovative in its approach to categorize effective PA-related interventions for individuals with T2D by their level of pragmatism. One limitation, however, is that the variability in the number of participants in each study may have led us to underestimate the number of studies that were effective for PA or HbA1c outcomes. To mitigate this limitation, our a priori criteria were set to exclude pilot studies with <40 participants or that reported <80% power to address the primary RCT outcome. Also, because this is not a meta-analysis, we could not determine the moderators of effectiveness in the interventions that were both highly pragmatic and improved PA outcomes. Regarding our assessment of the PRECIS-2 criteria, we conducted scoring from the perspective of a PCMH rather than a standard primary care clinic. This is a limitation, as it would be more challenging for clinics without a population health team to implement interventions that rely on staff contact outside of face-to-face clinician visits. However, two of the effective interventions in the top tier of pragmatism would be pragmatic to deliver in standard primary care clinics, as well as PCMH models, as they did not rely on a population health team for intervention delivery [50, 53]. Finally, we only assessed interventions that were clinically integrated; a prior review has identified important strategies for partnership between clinics and community organizations to improve diabetes self-management [29], and future research should identify pragmatic and effective interventions that can be delivered in partnership with community organizations.

If models of healthcare continue to shift from fee-for-service to patient-centered medical homes and accountable care organizations [37, 89], the emphasis to deliver high-value care that improves important health outcomes should continue to grow. An additional trend that may further the translation of PA and other behavioral interventions into clinical practice is the recent Medicare approval of Chronic Care Management funding codes in 2015 to reimburse clinical counseling and coordination of care that is delivered outside of face-to-face clinic visits [90]. Health system administrators may consider implementing the interventions that we identified as effective and pragmatic when the costs are deemed reasonable and the characteristics of the trials are a good match with their health systems. To enhance implementation further, the interventions that we identified as both pragmatic and effective merit further study in diverse real-world settings. In addition to reporting effectiveness, such future RCTs of PA interventions in real-world settings should report on intervention fidelity and adaptation, in order to inform the assessment of intervention moderators. Perhaps most importantly, such future RCTs of PA interventions in real-world settings must report on key practical feasibility factors that have been typically ignored, including intervention costs and sustainability.

Compliance with ethical standards

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. This article does not contain any studies with animals performed by any of the authors. Informed consent was obtained from all individual participants included in the studies. No IRB approval was required for the completion of this review study.

Findings reported here have not been previously published, and this manuscript is under no consideration at other journals. Data here is original and has not been previously published. All authors had full control of all primary data and agree to allow Translational Behavioral Medicine to review this data, if requested. This study was primarily funded by the National Heart, Lung, and Blood Institute (5K23HL118133).

Conflict of interest

The authors declare that they have no conflict of interest.