Scaling-up from an implementation trial to state-wide coverage: results from the preliminary Melbourne Diabetes Prevention Study

The Greater Green Triangle Diabetes Prevention Project (GGT DPP) was the first Australian implementation trial demonstrating that a lifestyle modification programme could be effectively implemented in a primary care setting, resulting in an estimated 40% diabetes risk reduction[1]. Following this success and a subsequent Healthy Living Course DPP randomised control trial[2], the Victorian government in 2007 announced funding for the Life! Taking Action on Diabetes (Life!) programme, a state-wide group-based lifestyle intervention targeting 25,000 Victorian residents aged over 50 at high risk of type 2 diabetes (T2DM)[3]. Programme goals are based on modifications to diet and physical activity. The Life! programme has direct lineage from the Finnish Diabetes Prevention Study[4], the Good Ageing in Lahti region Implementation trial[5] and the GGT DPP[1]. Over 10,000 people have enrolled in Life! courses, run by 208 group facilitators employed by 178 providers in Victoria (population 4,932,422[6], area 237,629 km²).

Moving from trials to large-scale implementation has long been fraught with failure[7], so evaluation of the scaled-up programme is important to assess its outcome. The Melbourne Diabetes Prevention Study (MDPS) is an evaluation of the effectiveness and cost-effectiveness of the Life! programme while it is being rolled out state-wide. This evaluation is being undertaken to see the effect of scaling-up to state level from the small GGT DPP implementation trial.

The specific objectives of the MDPS are to evaluate a structured primary-care based diabetes prevention programme (Life!) being implemented in Victoria for people aged over 50 at high risk of developing T2DM by: monitoring clinical and behavioural outcomes before and after the intervention (particularly reduction in diabetes risk, weight loss and quality of life); comparing the Life! programme with another cohort receiving usual care; and undertaking an economic assessment of the Life! programme.

The MDPS was designed to recruit 1,300 individuals (650 intervention, 650 control) from a number of sources including primary healthcare practices and the general community. The simultaneous large-scale rollout of the Life! programme in the community meant that the Life! implementers had to address issues such as recruitment, changes to eligibility criteria, referral processes, structure and content, as well as alternative programme delivery for different population groups. This complicated our evaluation of the Life! programme. For these reasons the initial phase of the MDPS was converted to a preliminary pilot study (pMDPS). We present here the effectiveness results from the pMDPS. The information obtained from this pilot phase was then used to inform a subsequent larger MDPS that has now commenced.

After exclusion of seven individuals (six with newly diagnosed T2DM at baseline OGTT), 92 participants were randomised (n%20=%2049 intervention, n%20=%2043 control) – of whom 80 (n%20=%2038 intervention, n%20=%2042 control) completed the study (Figure1). Information at 12 months was available for 38 intervention and 42 control participants. Baseline characteristics are shown in Table2. The control group contained a higher proportion of women and lower-income individuals.

Main outcomes are shown in Table1. Despite randomisation, differences between intervention and control subjects were observed in some baseline measures, especially weight, waist circumference and OGTT glucose at 2 hours.

The intervention group showed significant reductions at both 3 and 12 months in mean body mass index, weight, waist and hip circumferences and systolic blood pressure. Significant improvements were also seen at 12 months in high-density lipoprotein-cholesterol and all diet measures. In control subjects, OGTT glucose at 2 hours deteriorated and only waist circumference reduced significantly.

The intervention group significantly outperformed control subjects in changes over 12 months for body mass index reduction and fibre intake according to GEE results. After adjusting for baseline in analysis of covariance analyses, results were similar to the GEE estimates, except that the intervention group also showed greater weight loss and reduction in saturated fat intake (P=%200.035 and P%20=%200.003, respectively; data not shown). Although other differences were not significant, all favoured the intervention group, apart from Hospital Anxiety and Depression Scale depression score.

Table3 shows goal attainment. Intervention participants were more likely to achieve at least one goal (P%20=%200.013). At least 5% weight loss was achieved by 32% of the intervention group compared with none of the controls. Few attained the physical activity goal. While there were favourable improvements in total and saturated fat and fibre consumption (Table1) towards goals, only the fibre-intake goal (at least 15 g/1,000 kcal energy) was frequently achieved (Table3).

This preliminary study shows that it is possible to scale-up from an implementation trial for diabetes prevention (GGT DPP) and from the Victorian-based Healthy Living Course DPP[2] to a state-wide programme (Life!) with comparable outcomes. The results are tending to show that the system design for Life! will be fit for the purpose of scaling-up to a state-wide programme[3], but we will have to wait for the final MDPS results to confirm this. Participants achieved an estimated 40% diabetes risk reduction based on a decrease in waist circumference and 24% based on weight reduction, an outcome comparable with GGT DPP[1] and better than the FIN-D2D Finnish National Diabetes Prevention Program[16] and the Good Ageing in Lahti region trial[5].

Better outcomes were seen in the intervention group for anthropometric and biochemical measures, although some were not statistically significant. Goal achievement was not as successful as for the GGT DPP[1], the Finnish Diabetes Prevention Study[4] and the Good Ageing in Lahti region trial[5].

At this stage, the small sample size has reduced the generalisability of our findings. Groups were not particularly well matched at baseline and 22% of the intervention group dropped out, which may have confounded the results. Measurement bias may also have contributed to the high waist circumference reduction. Valuable lessons about methodology, especially recruitment and quality control for anthropometric measurements, have been learnt. These have been addressed in the forthcoming MDPS, which has now commenced. One benefit of this study was early diagnosis of six new cases of T2DM during recruitment.

Implementation failure is commonplace[7] because trials often provide insufficient information to facilitate successful scale-up. Moderating variables and issues of generalisability are frequently unreported and external validity is untested or uncertain[17].

Specifically, the challenges of translating clinical trials into effective population programmes have been reported in the literature[18, 19]. There has been a call to identify the barriers to diabetes prevention in the real world, with a special focus on identifying efficient intervention methods and delivery mechanisms[19]. Additionally, the importance of taking into account local circumstances and establishing partnerships and collaborations across sectors has been noted[18]. This has been achieved through the development of recommendations and guidelines for prevention of T2DM[18, 19]. These frameworks should be useful in our evaluation processes as we continue with diabetes prevention in Australia and with our evaluation of the state-wide Life! programme.