Behaviour change techniques

We used the CALO-RE taxonomy [10], a reliable 40-item taxonomy of standardised BCT definitions used in physical activity and healthy eating interventions, which has been used recently to evaluate potentially active ingredients in complex behaviour change of interventions [11],[13],[14],[16]. Intervention content from each of the 22 articles was reviewed and coded for the presence (or absence) of each of the 40 BCTs in the CALO-RE taxonomy. When a protocol of the study was published, this was also evaluated. Two reviewers (JL and EE) were trained by an experienced health psychologist (NH) in the use of the BCT-taxonomy. The two reviewers (JL and EE) coded the intervention content independently and discrepancies between them were resolved through discussion with a third reviewer (NH) at each stage of the review process and a consensus approach used. Behaviour change theories used in these studies, if explicitly declared, were also recorded.

Statistical analysis

Review Manager (RevMan Version 5.1 for Windows Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2011) and Stata (Stata/SE 11.2 for Windows; StataCorp LP, College Station, TX, USA) were used to pool and analyse results from the individual studies.

Pooled results are reported as mean differences with 95% confidence intervals (Cis) and with two-tailed P-values. A random effects model accounting for inter-study variation was used, thereby minimising potential bias due to methodological differences between studies. Multiple dietary intervention arms from three studies were included in the meta-analysis. As suggested by Higgins et al. [40] excessive weightings from `double counts’ originating from the `shared’ group (that is, control group), were controlled by splitting the sample size of the shared group into approximately equal smaller groups for the comparisons; the means and standard deviations were left unchanged. When available, we used results from multivariate models with the most complete adjustment for potential confounders reported in original studies. Analysis was performed to investigate associations between BCTs identified in complex interventions and in the effectiveness of the interventions in increasing F and V intake by comparing effectiveness of those studies which did, and those which did not, employ the specified BCT. To increase statistical power and reduce the likelihood of type-1 error associated with multiple comparisons we limited this analysis to the most frequently reported BCTs (that is, those BCTs which were reported in at least five studies).

Statistical heterogeneity was evaluated by using the I ² statistic [40],[41]; as proposed by Ioannidis et al.[42] the 95% CIs for I ² were calculated using the Higgins et al. method [43]. Where I ² was >50%, the degree of heterogeneity was considered high.

Publication bias was appraised by visual inspection of a funnel plot of effect size against the standard error (SE) for each study, with asymmetry assessed formally with Egger’s regression test [44].

We performed meta-regression analysis with restricted maximum likelihood estimation to assess the relationship of the number of behaviour techniques extracted from each study with changes in F and V consumption.

Behaviour change techniques meta-analysis

Meta-analysis of change in intake of F and V in response to intervention was carried out based on the presence or absence of specific BCTs which were reported in at least five RCTs. Results showed that studies incorporating the BCTs `barrier identification/problem solving’ (93 g, 95%CI 48 to 137 greater F and V intake) (Figure 1), `plan social support/social change’ (78 g, 95%CI 24 to 132 greater F and V intake) (Figure 2), `use of follow-up prompts’ (66 g, 95%CI 10 to 123 greater F and V intake) and `goal setting (outcome)’ (55 g 95%CI 7 to 103 greater F and V intake) (Table 2) were associated with conclusive and clinically important improvements in F and V intakes. In addition, there was weaker evidence that studies employing the BCT `provide feedback on performance’ were associated with greater F and V intake than studies not incorporating this BCT in their interventions (between groups difference 39 g 95%CI ≥2 to 81) (Table 2). Conversely, studies reporting use of the BCTs `motivational interviewing’ and `provide information on consequences of behaviour to the individual were associated with significantly smaller improvements in F and V intakes in comparison with studies not using these BCTs (Table 2). Forest plots for each of the BCTs investigated in Table 2 are provided as supplementary material [see Additional file 1].

Meta-regression showed that one additional BCT led to 8.3 g (95%CI 0.006 to 16.6 g) increase in F and V intake (Figure 3).

Theoretical basis of interventions

Fourteen studies reported the use of behaviour change theories in the development of their intervention and eight studies did not report use of any theory. Eight of the 14 studies reported use of a combination of two or more behaviour change theories while the other six studies reported using a single theory. The behaviour change theories declared in the reviewed studies included social cognitive theory (n = 5), social learning theory (n = 4), the transtheoretical model (n = 4), theory of planned behaviour (n = 2), social support (n = 2), and health behaviour model (n = 2), self-management (n = 1). Subgroup analysis showed that studies (n = 14) reporting the use of behaviour change theories reported changes in F and V intake which were nearly 50% greater (94.7 g/day 95%CI 49.1 to 140.3 g) than in those studies not reporting use of a behaviour change theory (66.5 g/day 95%CI -7.7 to 140.7 g) but the difference between these groups (28.2 g 95%CI -58.9 to 115.3) was not statistically significant. Studies reporting the use of a single behaviour change theory achieved 67% greater increase mean F and V intakes (n = 6) (121.6 g/day 95%CI 56.8 to 186.5 g) than those studies reporting use of a combination of behaviour change theories (n = 8) (72.8 g/day 95%CI 11.9 to 133.7) but the difference between these groups (48.8 g 95%CI -40.1 to 137.7) was not statistically significant.

Statement of principal findings

This analysis suggests that use of BCTs, including barrier identification/problem solving, plan social support/social change, use of follow-up prompts and goal setting (outcome), is associated with enhanced effectiveness of dietary interventions. Using these BCTs in the intervention design resulted in increases in F and V intakes which were 60 to 90 g/day greater than when these BCTs were not utilised in the intervention.

Strengths and weaknesses of the study

To our knowledge, this is the first study evaluating the association of BCTs use in dietary interventions in promoting sustained increases in F and V intake among individuals of retirement transition age. The strengths of this study include a rigorous methodology in the systematic review of the literature and the application of a well-established BCT taxonomy to identify BCT use. Although it is possible that the findings of this study might be confounded by unexplored covariates, previous analysis published elsewhere [5] showed that other aspects of the populations or interventions reviewed, such as mode of delivery of the intervention, health status of participants, follow-up time or ethnicity, were not related to F and V intakes and, therefore, are unlikely to confound the present findings. In addition, we limited the effect of multiple comparisons by restricting the comparisons to only those BCTs observed in five or more studies. Among the limitations are the small number of studies (n = 22) available and the high level of unexplained heterogeneity in this meta-analysis. In addition, from the studies reviewed we cannot confirm that the BCTs identified in these interventions were delivered as intended. The subgroup analyses are exploratory and the results obtained should be interpreted with caution because of the small numbers of studies in each sub-group and the post hoc nature of the analysis. The limited number of studies available determined that our analysis focused on examining the associations between F and V intake in studies with the presence or absence of specific BCTs and associations of combinations of BCTs with intervention effects on F and V intake could not be evaluated.

Scientific analysis of findings

Research into the effectiveness of including specific BCTs in dietary interventions is still limited. Using an earlier version of the BCT taxonomy comprising 26 BCTs, Michie et al. [9], reported that the inclusion of techniques derived from Control Theory (`prompt intention formation’, `prompt specific goal setting’, `prompt review of behavioural goals’, `prompt self-monitoring of behaviour’, `provide feedback on performance’) was associated with increased effectiveness of interventions of healthy eating and physical activity compared with those which did not include such techniques.

None of these systematic reviews, including ours, replicated the findings of a clear clustering of BCTs associated with the Control Theory as the key BCTs associated with better dietary outcomes reported by Michie et al. [9]. Of the five BCTs identified in the present study, only `plan social support/social change’ was also identified by Olander et al. [11] as associated with positive changes in physical activity self-efficacy among obese individuals. The median number of BCTs, as well as the most frequently reported BCTs, identified in the present study are similar to the BCTs identified in these studies [11],[13],[14],[16]. Differences in the effective BCTs identified in different reports might result from differences in the population groups studied; therefore, specific BCTs useful in a particular population group might not be equally useful in a different population group.

BCTs congruent with the Control Theory were associated with greater effectiveness of lifestyle interventions in the report by Michie et al. [9] but the present study revealed four BCTs which were effective in increasing F and V and which may also be part of several behaviour change theories. The two BCTs which were more strongly associated with improved F and V intakes in our systematic review, `barrier identification/problem solving’ and `social support’ are congruent with the Social Cognitive Theory [45]. In addition, significant, but more moderate, effects on intervention effectiveness were observed for BCTs congruent with the Operant Learning Theory [46]. However, it is important to note that individual BCTs are not exclusive to specific behaviour change theory.

Analysis of current evidence on the impact of the extent and type of theory on diet and physical activity interventions suggests that theory is not often used or reported and that its relationship with intervention effectiveness is weak [47]. Hill et al. [14] found that studies based on theory were as effective as non-theory-based studies at limiting gestational weight gain. These findings are supported by our observation of no significant differences in effectiveness of theory-based versus non theory-based interventions.

Our results showed a significant trend towards greater improvements in F and V intake with employment of more BCTs per study. This finding, however, contradicts previous evidence suggesting that it might be more important to focus on a few effective BCTs, rather than a greater number of BCTs, when designing dietary interventions [9],[13],[14].

Twelve BCTs from the CALO-RE taxonomy were not identified in any of the dietary interventions reviewed here, while 16 BCTs were reported by fewer than 5 studies and, therefore, were not considered in this analysis. More research on the potential usefulness of this sub-set of less frequently used BCTs in improving the effectiveness of dietary interventions is required.

Previously, we reported a significant positive association between number of contacts within the intervention and greater increases in F and V intakes, suggesting the importance of features of interventions on their effectiveness [5]. These results are supported by a recent report from the U.S. Preventive Services Task Force (USPSTF) on behavioural counselling to promote a healthful diet and physical activity in adults without pre-existing cardiovascular disease (CVD) or its risk factors suggesting that interventions of moderate to high-intensity, but not low-intensity, were associated with better results [6],[7].

Implications for health and policymaking

Worldwide increases in life expectancy are increasing the proportion of older adults in most populations with consequent increases in the proportion of retirees [48]. Retirement has been associated with increases in body weight and waist circumference, well known risk factors for CVD [49], and health promotion at retirement is of considerable importance to retirees and policy makers [50]. The findings of this study will contribute to the development of effective dietary interventions incorporating BCTs that have proven useful among adults of retirement age. Interventions which incorporated `barrier identification/problem solving’ and `plan social support/social change’ BCTs resulted in approximately 93 and 78 g/day extra F and V, respectively, (equivalent to about a portion of F and V according to UK standards [51]) compared with those not using these BCTs. Based on average population intakes of F and V in the UK (4.2 portions or 336 g/day, in adults 19 to 64 years; 4.4 portions or 352 g/day in older adults 65-years old and older), interventions incorporating more effective BCTs could make a significant difference in achieving the WHO recommendation of at least 400 g F and &V/day [52].

Several systematic reviews and meta-analyses of cohort studies indicate that small differences in F and V intake, such as the ones observed in the present study, are associated with reductions in the risk for cardiovascular events with significant potential for population impact. Increasing F and V consumption by around 100 to 150 g/day decreases the risk of stroke by 11% [53],[54], coronary heart disease by 4% to 7% [54],[55] and diabetes by 10% to 14% [56],[57]. Future research should test the effect of tailored interventions using the BCTs identified in this systematic review. Our study focussed on those 54- to 70-years old, and whether the BCTs identified in this study are applicable to other age groups remains to be tested. The CALO-RE taxonomy of BCTs [10] is a useful tool and its adoption should enhance the description of dietary and other lifestyle interventions, thus allowing more objective comparisons between studies.