Obesity management in primary care: systematic review exploring the influence of therapeutic alliance

Abstract

Purpose

To identify the influence of the therapeutic alliance on the effectiveness of obesity interventions delivered in primary care.

Method

Systematic review of randomized controlled trials of primary care interventions for adult patients living with obesity. Comprehensive search strategy using the terms ‘obesity’, ‘primary care’ and ‘intervention’ of seven databases from 1 January 1998 to March 2018. Primary outcome was difference in weight loss in interventions where a therapeutic alliance was present.

Results

From 10 636 studies, 11 (3955 patients) were eligible. Only one study had interventions that reported all aspects of therapeutic alliance, including bond, goals and tasks. Meta-analysis was not included due to high statistical heterogeneity and low numbers of trials; as per our protocol, we proceeded to narrative synthesis. Some interventions included the regular primary care practitioner in management; very few included collaborative goal setting and most used prescriptive protocols to direct care.

Conclusions

We were surprised that so few trials reported the inclusion of elements of the therapeutic alliance when relational aspects of primary care are critical for effectiveness. Interventions could be developed to maximize therapeutic relationships and research reports should describe interventions comprehensively.

Systematic review registration number

CRD42018091338 in PROSPERO (International prospective register of systematic reviews).

Consultation, continuity of care, doctor–patient relationship, lifestyle modification/health behaviour change, obesity, primary care

Key Messages

• Obesity interventions rarely include support for strong therapeutic alliances.

• Interventions could be developed to maximize therapeutic relationships.

• Research reports should describe interventions comprehensively.

Background

Obesity affects over 650 million adults worldwide (1) and is one of the world’s most pressing public health challenges (2). Obesity, defined as a body mass index (BMI) ≥30 kg/m², causes numerous health harms, including: impaired glucose metabolism, hypertension, musculoskeletal pain or mental health distress (3). Obesity also impacts all parts of health systems, including primary care.

Primary care is the first point of contact for the health system, and it seeks to provide holistic, person-centred, longitudinal care. In Australia, most primary care services are delivered within community general practice (4). In general, primary care has the potential to provide the most efficient and cost-effective means for accessible and effective chronic disease care in health systems (5). But despite primary care being well placed to assist patients with obesity (6,7), the role of primary care health professionals in assisting individuals living with the health implications of obesity remains hampered by a lack of effective strategies (8).

Primary care clinicians are well positioned to build and sustain longer-term relationships with patients. This is important because multiple trials and systematic reviews of continuity of care indicate that strong relationships between clinicians and patients contribute to improve outcomes for patients (9–13). The concept of ‘Working Alliance’ was established in the theoretical framework of Bordin in the late 1970s (14). Bordin extended the concept of the therapeutic relationship from warmth and empathy to include the practical and collaborative nature of a helping relationship that he termed the ‘Working Alliance’. This three-part model frames the relationship as: ‘Bond’—respect, empathy, trust and warmth; ‘Goals’—collaborative goal setting and ‘Tasks’—agreed steps to be undertaken to reach the goals (14).

This model has since been used to develop the Working Alliance Inventory for measuring the practitioner–client alliance in psychology; a high score is associated with better client outcomes (15,16). Further applications of the tool to other areas of medicine and primary care have also shown the strength of the alliance to be associated with better patient outcomes (17–19), and it has validity and can be reliably tested and pragmatically applied in the primary care setting (20).

This review focuses on ‘lifestyle interventions’, which include programs to support improved nutrition, increased physical activity and psychosocial care. While there are many intensive strategies for managing obesity, such as bariatric surgery, medications and very low energy diets, systematic reviews and guidelines support the use of behavioural and lifestyle changes to improve the effectiveness of all interventions (21,22). Yet, the influence of the therapeutic alliance between primary care clinicians and patients on the effectiveness of lifestyle interventions for obesity remains unknown.

This study aimed to answer:

1. What effect does therapeutic alliance have on the effectiveness of interventions for obesity in adults in primary care?

2. What effect does therapeutic alliance have on the rates of loss to follow-up in obesity management trials in primary care?

Methods

Data from eligible studies were examined for statistical heterogeneity with the aim of conducting meta-analyses where appropriate. As per our protocol, we planned to do a systematic review with meta-analysis but reverted to a narrative review because of statistical heterogeneity (23).

Search strategy and selection criteria

The protocol of this systematic review was published prospectively (23) and registered in PROSPERO (CRD42018091338). In summary, a systematic strategy was used to search: MEDLINE, EMBASE, PsycINFO, CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL) and Scopus for interventions for patients over the age of 18 years with obesity that were based in primary care.

Randomized controlled trials of interventions involving adults (over 18 years) with obesity delivered in primary care were eligible for inclusion. Full eligibility criteria have been published (23). Interventions for nutrition, physical activity and psychosocial care delivered wholly within primary care were included. Intensive approaches, such as medications and surgery, were excluded.

The presence or absence of therapeutic alliance within the intervention was defined on the basis of Bordin’s framework for working alliance:

1. the bond between the two parties: is there a pre-existing relationship with the provider of the intervention? (Yes/No) OR will there be an ongoing relationship with the provider of the intervention after the trial? (Yes/No);

2. collaborative goal setting: did the intervention involve collaborative goal setting? (Yes/No);

3. agreement on the required tasks to reach the goals: did the intervention involve agreement on the tasks to be undertaken (e.g. not protocol driven, room for individualization for the patient)? (Yes/No).

Data were independently identified by ES, NE and SA, extracted by KO and then checked by SA. Two members of the research team then independently categorized interventions into three groups:

• Group 1—none of the three components of therapeutic alliance (i.e. goal setting, task agreement and bond);

• Group 2—one to two components of therapeutic alliance (e.g. bond plus either tasks or goals);

• Group 3—all three components and/or the intervention specifically measured therapeutic alliance (although must not have ‘bond’ as the only part they describe).

Two papers were identified with duplicate data; the first published of these was included (24) and the second excluded (25). When appropriate, these groups were pooled for the purposes of assessing statistical heterogeneity (Supplementary Table S2). Studies were excluded if both trials arms had the same grouping for therapeutic alliance as the aim was to compare the impact of therapeutic alliance (23). To maintain quality, the review was done according to PRISMA quality criteria (26).

Data analysis

Our primary aim was to determine whether there was a measurable difference in weight loss and secondary metabolic outcomes achieved through interventions where a therapeutic alliance was present. We compared interventions where a therapeutic alliance was present (Groups 2 and 3) against those where it was not (Group 1). Group 1 included those studies where a therapeutic alliance was not present and typically included conventional health care or an alternative intervention that did not invoke a therapeutic alliance.

Interventions with two of the three components of the therapeutic alliance model were classified as Group 2 and those with all three components as Group 3. Only one article (27) included an intervention classified as Group 3. Therefore, interventions in Groups 2 and 3 were pooled and classified as having a therapeutic alliance present.

To determine whether to proceed with meta-analysis, we calculated the statistical heterogeneity. Data were entered into Review Manager 5.3 for assessing statistical heterogeneity for the meta-analyses. Data were entered as continuous data using the inverse variance method and random effects model. Heterogeneity was assessed using the inconsistency statistic (I²). I² of 50–90% represents substantial heterogeneity and of 75–100% considerable heterogeneity (28) (Supplementary Table S1). There was no funding source for this study.

Results

From 12 109 papers, after removing duplicates, 10 636 papers were screened for inclusion. From a preliminary scan of the titles and abstracts, 215 papers were selected for full screening. This resulted in the selection of 30 papers for the meta-analysis, of which 11 (24,27,29–37) met the inclusion criteria (Fig. 1).

Figure 1.

PRISMA flow diagram outlining study selection and exclusions.

Statistical heterogeneity

Overall statistical heterogeneity (I²) was high (most over 90%; Supplementary Table S1), and there were very few eligible studies in each subgroup once the trials were pooled. Following our protocol, we were, therefore, unable to proceed to the reporting of a meta-analysis. Studies did not consistently differentiate between loss to follow-up and withdrawal, so loss to follow-up was reported as a single category. Furthermore, subgroup analyses (time intensiveness, whether the intervention was delivered in regular practice and usual provider involvement) were also not possible due to the small number of studies identified and high heterogeneity.

Description of interventions

Characteristics of the 11 studies are shown in Table 1. Across the 11 studies, there were 34 different trial arms (control or intervention) with nine studies including a usual care arm (27,29–32,34–37). All studies used BMI in their definition for obesity, with two complementing this with a waist circumference (32,36) and three also including health impairment as part of their obesity definition (27,31,37).

Table 1.

Studies meeting inclusion criteria for the systematic review of obesity management in primary care

Study	Population	Study type	Primary outcome	Study arms with a therapeutic alliancea (n)	Study arms without a therapeutic allianceb (n; no elements of therapeutic alliance reported)	Multiple practitioners involved in delivery	Delivered at the usual place of primary care provision	Weight loss measures	Other markers
Anderson et al. (1)	USA Age 20–65 years BMI 30–39.9 kg/m2	RCT	Mean weight loss % at 16 weeks	Control group (received weight management counselling from an experienced dietician; n = 22; agreement on tasks)	Intervention group (participated in scheduled behavioural weight loss classes; n = 23)	Yes—intervention group only	No	Change in waist circumference (8/16/24 weeks) Percentage weight change (8/16/24 weeks)	Glucose LDL cholesterol
Beeken et al. (2)	England Age ≥18 years BMI ≥30 kg/m2	RCT	Change in measured weight (kg) between baseline and 3 months	10TT—leaflet plus two appointments with nurse/assistant (n = 267; bond: pre-existing relationship and opportunity for ongoing relationship)	Usual care (n = 270)	No	Yes—both arms	Average weight loss (kg; 3/6/12/18/24 months) Average BMI points lost (kg/m2; 3/6/12/18/24 months) Proportion losing >5% body weight (3/6/12/18/24 months) Change in waist circumference (3/6/12/18/24 months)	Glucose SBP DBP Total cholesterol LDL cholesterol
Bennett et al. (3)	USA Age 25–65 years BMI 30–40 kg/m2 with: hypertension and on treatment No smoking 6 months prior June 2005–June 2006	RCT	Change in body weight at 12 weeks	Step Up Trim Down website plus health coach (n = 51; agreement on tasks)	Usual care (n = 50)	Yes—Step Up Trim Down group	Uncertain	Average weight loss (kg; 3 months) Average BMI points lost (kg/m2; 3 months) Proportion losing >5% body weight (3 months) Change in waist circumference (3 months) Percentage weight change (3 months)	SBP DBP
Jakobsen et al. (4)	Denmark Age ≥18 years with schizophrenia, schizoaffective disorder or persistent delusional disorder, and waist circumference >102 cm for men and 88 cm for women Dec 2012–May 2014	Multi-centre RCT	10-year risk CVD (%) at 2 years	Control (n = 148; bond: pre-existing relationship and opportunity for ongoing relationship) Change (coach weekly, coordinated with GP; n = 138; bond: pre-existing relationship and opportunity for ongoing relationship)	Care coordination (n = 142)	No	Uncertain	Average weight loss (kg; 24 months) Average BMI points lost (kg/m2; 24 months) Change in waist circumference (24 months)	HbA1c SBP Total cholesterol HDL cholesterol
Jolly et al. (5)	England Age ≥18 years Those with no comorbidities: BMI ≥28 kg/m2 (South Asian) BMI ≥30 kg/m2 (all other ethnicities) Those with comorbidities: BMI ≥23 kg/m2 (South Asian) BMI ≥25 kg/m2 (all other ethnicities)	Eight-arm RCT	Weight loss at 3 months follow-up	GP (n = 70; bond: opportunity for ongoing relationship, prior relationship not recorded; agreement on tasks) Pharmacy (n = 70; bond: opportunity for ongoing relationship, prior relationship not recorded; agreement on tasks)	Weight watchers (n = 100) Slimming world (n = 100) Rosemary Conley (n = 100) Size down (n = 100) Choice (n = 100) Exercise/comparator (n = 100)	No	Yes—GP arm only	Average weight loss (kg; 3/12 months) Average BMI points lost (kg/m2; 12 months) Proportion losing >5% body weight (3/12 months)	n.r.
Little et al. (6)	England Age ≥18 years BMI ≥30 kg/m2 (or ≥28 kg/m2 with hypertension, hypercholesterolaemia or diabetes) Jan 2013–Mar 2014	RCT	Weight loss over 12 months	Evidence based dietetic advice; evidence-based dietetic advice to swap foods for similar, but healthier, choices and increase fruit and vegetable intake, in addition to 6 monthly nurse follow-up (control group; n = 279; bond: pre-existing relationship and opportunity for ongoing relationship) POWER+F; web-based intervention and face-to-face nurse support; up to seven nurse contacts over 6 months (n = 269; bond: pre-existing relationship and opportunity for ongoing relationship)	POWER+R; web-based intervention and remote nurse support; up to five emails or brief phone calls over 6 months (n = 270)	No	Yes	Average weight loss (kg; 6/12 months) Proportion losing >5% body weight (6/12 months)	Glucose HbA1c SBP DBP Total cholesterol
McDoniel et al. (7)	USA Age 18–70 years BMI ≥30 kg.m-2 Non-smoker Access to home computer and email	RCT	Bodyweight and arterial blood pressure (i.e. systolic and diastolic) were assessed at baseline and at week 12	Usual care (n = 56; bond: pre-existing relationship and opportunity for ongoing relationship)	SMART group (n = 55)	No	Yes	Average weight loss (kg; 12 weeks) Proportion losing >5% body weight (12 weeks)	SBP DBP
Munsch et al. (8)	Switzerland Age not specified BMI ≥30 kg/m2 Exclusions: severe mental illness, IDDM, hypothyroidism, terminal diseases	RCT	Change in BMI at 1 year	GP BASEL (n = 53; bond: pre-existing relationship and opportunity for ongoing relationship) GP control (n = 17; bond: pre-existing relationship and opportunity for ongoing relationship)	Clinic BASEL (n = 52)	Yes—both BASEL groups	Yes—both GP groups	Average BMI points lost (kg/m2; 12 months) Percentage weight change (12 months)	n.r.
Ross et al. (9)	Canada Age: 25–75 years BMI 25–39.9 kg/m2 Waist circumference: >102 cm (men) or >88 cm (women) Sedentary	RCT	Waist circumference measured at the superior edge of the iliac crest at 24 months	Usual care (n = 241; bond: pre-existing relationship and opportunity for ongoing relationship)	Behavioural intervention (n = 249)	No	Yes	Average weight loss (kg; 6/12/18/24 months) Average BMI points lost (kg/m2; 6/12/18/24 months) Change in waist circumference (6/12/18/24 months)	Glucose SBP DBP Total cholesterol LDL cholesterol Triglycerides HDL cholesterol
Wadden et al. (10)	USA Age ≥21 years BMI 30–50 kg/m2 ≥2/5 components of the metabolic syndrome	RCT	change in body weight at month 24	Usual care (n=43) (bond: pre-existing relationship and opportunity for ongoing relationship)	Brief lifestyle counselling (n=131) Enhanced lifestyle counselling (n=129)	Yes – two counselling groups; uncertain for usual care group	Yes	Average weight loss (kg) (6/12/18/24 months) Average BMI points lost (kg.m-2) (6/12/18/24 months) Percentage weight change (6/12/18/24 months)	n.r.
Yardley et al. (11)	England Age ≥ 18 years BMI ≥30 kg.m-2 (or ≥28 kg.m-2 with hypertension, hypercholesterolaemia or diabetes) May 2011–Dec 2012	RCT	Weight 12 months	Web intervention (n = 45; collaborative goal setting; agreement on tasks) Plus basic nursec (n = 44; bond: pre-existing relationship and opportunity for ongoing relationship; collaborative goal setting; agreement on tasks) Plus regular nursec (n = 47; bond: pre-existing relationship and opportunity for ongoing relationship; collaborative goal setting; agreement on tasks)	Usual care (n = 43)	No	Yes—nursing arms	Average BMI points lost (kg/m2; 6/12 months)	n.r.

BMI, body mass index; DBP, diastolic blood pressure; HDL, high-density lipoprotein cholesterol; HbA1c, glycated haemoglobin; LDL, low-density lipoprotein cholesterol; n.r., not reported; RCT, randomized controlled trial; SBP, systolic blood pressure.

^aGroup 2 (unless otherwise specified)—one to two components of therapeutic alliance (e.g., bond plus either tasks or goals).

^bGroup 1—none of the three components of therapeutic alliance (i.e., goal setting, task agreement, and bond).

^cGroup 3—all three components and/or the intervention specifically measured therapeutic alliance (although must not have ‘bond’ as the only part they describe).

The studies examined intensifying lifestyle-based education, advice and support delivered by health care practitioners. Only six of the programs (five active intervention arms (29,35,37) and one control arm (37)) involved more than one practitioner in the delivery of the health care.

Of the 34 different programs, nine (29,33,35) were unequivocally delivered away from the usual place of primary care provision; provision of five studies could not be determined due to vague intervention descriptions (31,32). One UK study examined commercial weight loss providers in three of their eight arms (33).

Effectiveness was most commonly measured as weight loss, but change in BMI, percentage weight loss and waist circumference were also used (Table 1). There was variability in the time point of the primary outcome (6 months to 2 years; Table 1). Overall, studies resulted in small changes in weight, from a small weight gain to approximately 5–7 kg of weight loss. None of the studies that reported weight at multiple time points showed an improvement over time and most showed weight returning towards the baseline (Supplementary Table S3). As I² was commonly over 90%, we are not able to make rigorous comparison between the studies regarding effectiveness. Further information can be found in the Supplementary Material.

Studies were also pooled to assess differences in loss to follow-up (Supplementary Table S3.3). Statistical heterogeneity was also substantial to considerable and, thus, we are unable to make a meaningful comparison, and we proceeded to narrative review as per our protocol.

Therapeutic alliance—all three elements

Only one study had interventions containing all three elements of the therapeutic alliance (i.e. bond, collaborative goal setting and agreement on tasks) (27). This four-arm trial tested the effectiveness of a web-based intervention, or web-based intervention with basic nursing support, or more intensive nursing support, versus a control arm. There was clear description of collaborative goal setting, individualized tasks plus the potential for ongoing care from the nursing provider. Mean weight loss at 12 months was highest in the arm with web plus basic nursing support (4.31 kg, n = 44), and the authors emphasized the feasibility of the intervention for primary care delivery.

Therapeutic alliance—bond

We classified study arms as having the bond element when it was explicitly reported that the usual primary care practitioner was involved in the intervention or there was potential for ongoing care with the practitioner. Of the nine control arms, six explicitly reported using the regular practitioner (66.7%) (25,32,34–37), and 8 of the 25 active interventions (32.0%) (27,30,32,33,35) used the regular primary care clinician(s) or there would be the possibility of ongoing care from the clinician. Reported primary outcomes were diverse from 10-year cardiovascular risk at 2 years (32); change in weight at 3 months (33) to 12 months (27); reduction in waist circumference (36) and change in weight and blood pressure at 12 weeks (34). This diversity makes it impossible to sensibly compare the effectiveness based on the presence of bond in the intervention.

Therapeutic alliance—goals

Studies rarely reported collaborative goal setting with patients. The study by Yardley et al., mentioned above, was the only study that clearly described collaboratively goal setting (27); one study with patients with severe and enduring mental health disorders potentially involved goal setting but it was unclear from the reporting (32).

Therapeutic alliance—task

Within the therapeutic alliance theoretical framework, tasks are the steps to be taken to reach the agreed upon goals. Of the 34 different study arms, eight [Anderson control (29); Bennett active (31); Yardley mentioned above (27); Jolly (GP, pharmacy and choice arm) (33)] explicitly reported opportunities for the patient to be involved in deciding the type of tasks to be taken. The majority of the studies described protocols that directed the specific approach to lifestyle change. It was unclear from the reporting of one trial (32) as to whether an individualized approach to tasks was possible.

Loss to follow-up and dropout

Studies did not consistently report whether patients withdrew or were lost to follow-up (see Supplementary Table S3.3 for the raw numbers of participants whose data were not reported at specific time points). Due to the low number of studies and high I², we are not able to determine the influence of the therapeutic alliance on withdrawal or loss to follow-up.

Risk of bias

The risk of bias is moderate to high with most studies being unable to blind participants and study personnel to the participant’s study arm due to the nature of the intervention. Allocation concealment was often absent and it was often unclear if studies had selective outcome reporting as they did not report the outcomes in a prospective protocol (Supplementary Table S4).

Conclusions

This systematic review identified randomized controlled trials of obesity interventions in primary care and classified both control and intervention arms based on the presence or absence of elements of the therapeutic alliance. Our aim was to determine the influence of the therapeutic alliance on the effectiveness of the interventions. The testing of our hypothesis was limited as only one study included an intervention that could be categorized as Group 3 (strong therapeutic alliance) (27). This intervention was convincingly described as including elements of collaborative goal setting, agreement on tasks and respectful bond between the parties. This intervention did not involve family doctors and was delivered by primary care nurses, who may or may not have had an existing relationship with the patient but had the potential to provide ongoing health care (27).

No trial directly measured the therapeutic alliance, so we had to deduce from the reporting of the design as to whether it was likely or unlikely to be present. Only one study had all three elements (bond, tasks and goals). The most important outcome of this study is that we have identified a significant gap in research on relational aspects of primary care and intervention developed for obesity management.

We began this endeavour as there is a growing body of evidence that shows improved health outcomes when there is a strong therapeutic relationship (9,10,13,14). Our review does not appear to strongly support relationship-based case; however, it is not uncommon for primary care trials to generate results that do not support hypotheses that are based on the values and mechanisms of effective primary care (38,39). Stange has highlighted that contrary results can lead us to raise doubts about the fundamental values of primary care or to question the evidence that has been presented (40). Our systematic review leaves us with more questions than answers when it comes to relational aspects of care in obesity management. It would be ideal to repeat our method using another primary care intervention where perhaps the interventions are designed to include the alliance and are reported accurately.

The reported primary outcomes varied greatly (Table 1), and this contributed to the high statistical heterogeneity. The difficulties associated with variability in effectiveness outcomes have been recognized and the ‘COMET’ initiative is working to develop agreed minimum outcomes for a variety of health conditions (41). A Delphi study is currently underway to develop core outcomes for behavioural weight management programs (42), and this would cover the type of studies included in this review.

Strengths and limitations

The included studies were not designed with assessment of the therapeutic alliance in mind and the data needed to be interpreted from the research report. The lack of detail in some of the primary reports of research made the extraction difficult and we accounted for this with two primary reviewers and follow-up with a third if needed. We concentrated on the specific construct of therapeutic alliance, and it is possible that papers made mention of other forms of therapeutic relationships that we did not capture in this synthesis. We were unable to publish results of meta-analyses due to considerable statistical heterogeneity of the data extracted from eligible studies.

Furthermore, our observations are limited by the low number of studies and vague descriptions of interventions. Future reports of research should contain comprehensive and detailed descriptions of interventions (43), including the clinicians involved (44), and descriptions of the study contexts.

Implications for research and practice

It was surprising to see so few interventions that had been developed with elements of the therapeutic alliance in mind. The clinician–patient relationship is one of the fundamental factors for high-quality primary care (13) and we had expected that interventions would be developed taking this into consideration. This speaks to the overall design, development, and implementation of interventions in primary care—are we incorporating factors known to be necessary for effective primary care? (Box 1) The lack of support for therapeutic alliance in these obesity trials has implications for primary care research beyond the study of obesity management, in particular for researching aiming to improve outcomes for patients living with other chronic and complex diseases.

Box 1—Recommendations for future trials of obesity management in primary care

Researchers are encouraged to incorporate elements of the therapeutic relationship when developing new interventions.

When evaluating interventions, consider theoretical frameworks and incorporate them into the evaluations for effectiveness.

Cleary describe interventions and controls using the TIDIER framework (43) as guidance.

Primary care trials could consider standard outcome sets to assist with meta-analyses (41).

Supplementary material

Supplementary material is available at Family Practice online.

Table S1: Statistical pooling – statistical heterogeneity (I²) and number of studies

Table S2: Papers contributing to each determination of statistical heterogeneity

Table S3: Data contributing to determination of statistical heterogeneity (I²)

Table S4—Risk of bias analysis

Declaration

Funding: this research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors. The study was funded by departmental resources.

Ethical approval: not applicable

Conflict of interest: the authors declare no actual or potential conflicts of interest relevant to this work.

Data availability

The data underlying this article are available in the article and in its online supplementary material.