The PROPELS study found that a pragmatic, 3-hour group-based behavioural intervention, when combined with tailored text messages and telephone calls, can be used to initiate increased ambulatory activity of over 500 steps per day during the first 12 months of intervention delivery in people with prediabetes in primary care. However, the effects were not sustained at 48-month follow-up and the results were consistent across ethnic groups. Secondary outcomes were also largely unchanged. However, there were small but sustained reductions in triglycerides in the Walking Away Plus study arm, as well as in body weight and waist circumference in the Walking Away study arm, which were accompanied by improvements in liver function at 48 months.
It is increasingly recognised that even small changes to physical activity may have important health benefits. It has recently been proposed, using a triangulation of different data sources, that 500 steps per day or 5–6 minutes of brisk walking is the minimum clinically meaningful difference in ambulatory activity, conveying a relative reduction in the risk of cardiovascular disease and all-cause mortality of between 2% and 9% for inactive populations or a difference in life expectancy of around 4 years.126 These findings are reflected in the recently updated Physical Activity Guidelines for Americans,127 which acknowledges that even 5 minutes of moderate-intensity physical activity – equivalent to around 500 steps of brisk walking – has health benefits. The increase in ambulatory activity seen in the Walking Away Plus study arm relative to controls at 12 months, although modest, is, therefore, likely to be clinically meaningful.
Although, in this study, changes to physical activity in the Walking Away Plus study arm were not accompanied by changes to most of the assessed cardiometabolic risk factors, triglycerides were reduced, with the effects sustained after 48 months. Triglycerides are one of most consistently observed lipid profile factors that are associated with levels of physical activity.128 The change in triglycerides (0.15 mmol/l) observed in the Walking Away Plus study arm at 12 months has previously been associated with a 2% reduction in the relative risk of CVD-related mortality.129
Although there was no evidence of changes in objectively assessed physical activity at 48 months, there was an increase in self-reported energy expenditure of 4.4 (97.5% CI 0.0 to 8.8) kJ/kg/day. It has previously been reported that each additional 1 kJ/kg/day per year increase in physical activity energy expenditure (equivalent to the 4 kJ/kg/day increase seen in the Walking Away Plus study arm seen over 4 years) is associated with a 24% reduction in all-cause mortality, a 29% reduction in cardiovascular disease mortality and an 11% reduction in cancer mortality.130 Although the objective measures of physical activity suggest that overall physical activity volume was no different between the study arms at 48 months, the context in which the activity occurred may have varied. Objective measures of physical activity record all movement, including movement undertaken in incidental habitual activities, such as household chores, shopping and walking between meetings at work. Self-reported questionnaires, on the other hand, are more likely to capture purposeful physical activity that the individual has cognitively assigned as physical activity, such as choosing to go for a walk. Therefore, participants in the Walking Away Plus study arm may have engaged in greater amounts of structured physical activity. This would be consistent with the findings that self-monitoring behaviours continued to be well adhered to after 48 months in the Walking Away Plus study arm, with over three-quarters of participants reporting that they set themselves physical activity goals and almost two-thirds reporting using a pedometer at least some of the time. Nevertheless, the high levels of self-reported energy expenditure could also be a result of response bias. The added health benefits of undertaking greater levels of self-reported physical activity, without accompanying changes to objectively assessed physical activity, are, therefore, uncertain and require further research.
To our knowledge, this is the first trial designed to assess the effectiveness of a pragmatic physical activity intervention over 4 years in a population with prediabetes. However, the results are consistent with other recent pragmatic physical activity trials that were published after the commencement of PROPELS (see Appendix 16, Table 74).
A previous evaluation of the standard Walking Away programme in those with risk factors for T2D demonstrated an increase in ambulatory activity of 411 steps per day at 12 months, but with no longer-term effects observed at 24 or 36 months.28 Similarly, the Let’s Prevent Programme, which incorporated the key components of Walking Away into a broader lifestyle diabetes prevention programme, demonstrated an increase in ambulatory activity of 552 steps per day in the intervention study arm compared with the control study arm at 12 months, with an increase of 535 steps maintained at 36 months.131 A clinician-led physical activity intervention in those with established T2D reported similar findings to the Walking Away-based prevention studies, with a 6.8-minutes-per-day increase in moderate to vigorous physical activity after 12 months, but with effects reducing to 3.6 minutes after 36 months. The PACE-UP pedometer intervention for inactive adults demonstrated increases in ambulatory activity of between 600 and 700 steps per day over 36 months, but the effect for ambulatory activity was not sustained when the intervention was evaluated in older adults aged 60–75 years over 48 months.132 The recently published EuroFit intervention, which employed a more intensive intervention structure than Walking Away, reported a 678 steps/day increase in the intervention study arm, relative to controls after 12 months in overweight and obese men.133 These studies are in contrast to the earlier ProActive intervention, which found no effect after 12 months of a behavioural intervention designed to increase physical activity in those with a family history of diabetes,134 while the Early ACTID dietitian- and nurse-led physical activity and dietary intervention for those with T2D reported a 5.6-minute increase in moderate or vigorous physical activity after 12 months relative to standard care.135
This mounting evidence from trials evaluating pragmatic physical activity interventions with objective measures of physical activity are also consistent with the wider diabetes prevention and management literature. An earlier review of the seminal diabetes prevention programmes concluded that either no or only small changes to physical activity were seen when the trials were evaluated using self-reported physical activity. Even the highly intensive LookAHEAD lifestyle intervention for those with T2D, which was designed to maximise behaviour change rather than to be implemented into routine care, reported a difference of only 4 minutes per day in moderate to vigorous physical activity after 4 years in the intervention study arm compared with the control arm, with no difference seen in the overall volume of physical activity undertaken.136 Taken together, these results suggest that small, but nevertheless clinically meaningful, increases in physical activity are possible at 12 months after receipt of a behavioural intervention designed for inactive adults or those with metabolic dysfunction within primary care, but that such changes may be difficult for individuals to sustain in the longer term.
The findings from the process measures and embedded qualitative study highlighted important themes that help provide context for the difficulties in promoting sustained physical activity behaviour change. Importantly, the results did not seem to be explained by intervention attendance or engagement with behaviour change techniques. The majority (80%) of those randomised to Walking Away and Walking Away Plus attended the initial group-education session, with 78% of those in the Walking Away Plus arm also registering with the text messaging service. In addition, the interpretation of the results was not affected by a per-protocol analysis, suggesting that engagement with all elements of the intervention did not lead to substantially increased effectiveness. In addition, self-reported rate of engagement with the key behaviour change techniques centred on pedometer use remained relatively high in the intervention arms compared with the control arm, even after 48 months. However, our qualitative research highlighted a key theme whereby major illnesses, injury or life events that occurred over the 4-year trial period in this older population caused large reductions in physical activity levels and the ability to move around freely, from which individuals found it hard to fully recover. This suggests that to be sustained, a focus on rehabilitation from illness or injury could be considered in future physical activity programmes in older populations with a high risk of chronic disease.
As far as we are aware, no longer-term (i.e. longer than 2 years), objectively assessed, physical activity interventions have incorporated mobile technology. A recent review of the literature137 in adults aged ≥ 50 years, reflective of the PROPELS population, identified 52 relevant physical activity, sedentary behaviour or sleep studies that had employed mobile technology, of which text messaging was the most common component. Among these studies, the largest sample size was 710 and the longest follow-up duration was 12 months. The authors concluded that, although there was some evidence for the effectiveness of mobile health interventions, higher-quality trials were needed. PROPELS suggests that incorporating mobile health technologies into a low-resource, pragmatic, group-based behaviour change intervention can moderately enhance the degree of behaviour change after the first 12 months, but it may not help maintain behaviour change in the longer term. Successful behaviour change maintenance has been suggested to require at least one sustained motivator.113 Whether or not the greater short-term effectiveness in the Walking Away Plus study arm at 12 months helped to provide this sustained motivation is not investigated by the current statistical analysis plan. Future post hoc analysis using the PROPELS data will address these questions further.
Although the Walking Away programme was focused on physical activity, the programme did include a short dietary component that was designed to introduce participants to broad dietary concepts, including macronutrient composition and different types of fats. Participants in both study arms reported making modest dietary changes by increasing fruit and vegetables intake at both 12 and 48 months and actively trying to limit the amount of total and saturated fat consumed on more days of the week. In addition, participants in the Walking Away study arm lost 1 kg in weight and reduced their waist circumference by 1.6 cm compared with those in the control arm at 48 months. Although sustained, these changes were relatively modest, with smaller effects than interventions that are specifically aimed at long-term weight loss.138 Although the impact of this degree of weight loss on mortality outcomes is uncertain,138 the Diabetes Prevention Programme reported that each additional kilogram of weight loss was associated with a 16% reduction in diabetes risk,139 suggesting that this degree of weight loss may have conferred some cardiometabolic benefits to those in the Walking Away study arm. Indeed, there was evidence that markers of liver function (ALT and ALP) were improved at 48 months in the Walking Away study arm, which is consistent with previous research showing a strong effect of weight loss on markers of liver function.140 Interestingly, no such changes were observed in the Walking Away Plus study arm, in which markers of weight and adiposity were unchanged compared with control throughout the trial period. In Walking Away Plus, the mHealth follow-on support was specifically focused on physical activity only, which may have acted to deflect a focus on diet.
The real-world costs of delivering Walking Away and Walking Away Plus were estimated as £257 and £322 per person, respectively. The probabilistic lifetime costs of Walking Away and Walking Away Plus (£22,945 and £23,018, respectively) remained higher than those for standard care (£22,598). Lifetime cost-effectiveness modelling suggested that standard care had the highest probability of being cost-effective below a willingness-to-pay threshold of £20,000 per QALY. It was further estimated that, to reach a threshold of £20,000 per QALY, the Walking Away Plus study arm would have to be delivered at a maximum cost of £116 per person. However, there was a high level of uncertainty in these estimates, with the total value to the UK of research to eliminate all uncertainty estimated at £279,559,484. Few other studies have conducted lifetime cost-effectiveness modelling for behavioural interventions. A study141 of brief behavioural interventions suggested that those that incorporated pedometer use, although not cost saving overall, did have a high probability of being cost-effective at a threshold of £20,000 per QALY. However, consistent with our study, the value of eliminating uncertainty was estimated at £1.85B.141 The PACE-UP study142 found that simply sending pedometers with accompanying motivational material by post was likely to be cost saving over a lifetime.
At the other end of the spectrum, intensive lifestyle interventions for diabetes prevention have also been reported to have a high probability of being cost saving.143 It is likely that our results differ from these results for three key reasons. First, the only clinical effect of Walking Away Plus was on step count and triglycerides, without associated changes in other variables (e.g. HbA1c and weight), which were included in the brief interventions study.141 Second, the PROPELS study showed that the effect of Walking Away Plus on physical activity had diminished by 4 years post intervention, so a constant improvement in physical activity was not included in our modelling. Third, the observed data from the PROPELS study suggested that the incidence of diabetes diagnoses was higher, but statistically insignificant, in the Walking Away and Walking Away Plus arms. The relationship between step count and diabetes diagnoses is in the opposite direction of those used the other analyses.141,142 The inconsistent finding and the high value of eliminating future uncertainly for the cost-effectiveness of physical activity intervention suggest that further research is needed to inform commissioning decisions, especially around any assumed surrogate relationships based on change in physical activity alone and on the duration of effect of any clinical benefit. The within-trial analysis results mirrored these findings (see Appendix 14, Table 70). Although the ICER for Walking Away was estimated at £9639, which is largely consistent with previous group-based interventions for those with or at high risk of diabetes,25,144 there was a high level of uncertainty associated with it, whereas all study conditions had a near equal chance of being cost-effective at a threshold of £20,000 per QALY.
The PROPELS trial has notable strengths and limitations. Its strengths include a population that was predominantly recruited from primary care and had coded HbA1c or glucose values in the prediabetes range, making the population reflective of those referred to currently available diabetes prevention programmes. The large, diverse multiethnic population recruited from urban and rural locations with large variations in social deprivation is also reflective of the modern UK. Indeed, the distribution of material deprivation within our sample almost exactly mirrored the national average, with an equal spread of individuals across nationally defined deprivation quintiles. To the best of our knowledge, PROPELS is also the largest and longest-running physical activity or lifestyle intervention in those with prediabetes that has been evaluated using an objective measure of physical activity.
These strengths also come with potential limitations. The length and the nature (RCT) of the trial may have acted to discourage some from taking part, thereby limiting the generalisability of the sample. For example, the relatively high levels of ambulatory activity and physical activity self-efficacy at baseline may have limited the effectiveness of the intervention in promoting further behaviour change. It is also of note that only around 40% of participants were confirmed to have prediabetes based on HbA1c values at baseline. It has previously been shown that there are high levels of regression to normal glycaemia between paired HbA1c tests when the first is within the prediabetes range,145 with a recommendation that two HbA1c tests are needed to confirm prediabetes status. Along with the possibility of regression to normoglycaemia between measurement and inclusion, it is possible that this relatively small number of participants with prediabetes based on HbA1c was because inclusion was based not only on previous HbA1c values, but also on post-challenge and fasting glucose levels. This is important as it has been shown that HbA1c, fasting glucose and post-challenge values identify discordant high-risk populations.145,146 However, the fact that only HbA1c was measured within this trial and that participants were informed of their HbA1c status may have influenced the trial results. For example, for individuals whose HbA1c was found to be within normal ranges, this information may have acted to de-emphasise their need for behaviour change. However, the sensitivity analysis for the primary outcome did not find that results were modified by baseline HbA1c prediabetes status, suggesting that similar levels of behaviour change at follow-up were observed in those with and those without prediabetes at baseline. Furthermore, the strategy for HbA1c assessment within this study is consistent with that specified in Healthier You, the NHS Diabetes Prevention Programme, in which the HbA1c assessment has to be repeated by the provider if the original referral value is older than 3 months.147 However, a normal value on the repeat test is not an exclusion criterion and such individuals are still accepted on the programme.
Nevertheless, the degree of engagement with interventions in this study (52% and 62% compliance with the per-protocol definition of the Walking Away Plus and Walking Away interventions, respectively) is consistent with previous implementation studies. For example, in the DEPLOY study148 conducted in the USA, participants attended 57% of available sessions, and in Finland 56% of individuals reported attending all group-based sessions in the GOAL Implementation Trial.149 Although using objective measures of physical activity risks biasing the results as a result of the Hawthorn effect (measurement reactivity) and poor compliance, recent evidence has suggested that the Hawthorn effect is likely to be minimal in adults for moderate or vigorous physical activity.150 In addition, the randomised design makes the results more robust, given that the Hawthorne effect by itself would not be enough to bias results for change; rather, a further assumption would be needed, namely that the Hawthorne effect differed across randomised study arms.
At 48 months, 993 (73%) participants in the sample had valid accelerometer data, which is consistent with other well-conducted RCTs over the longer term (see Appendix 16, Table 74). For example, the PACE-UP study, which recruited 1023 inactive adults from primary care, reported 67% compliance with the primary outcome, namely objectively assessed physical activity, after 36 months.132 Given that participants in PROPELS were followed up only at 12 and 48 months, the trajectory of change between these time points was not evaluated, making it unclear whether or not the clinically meaningful change in the Walking Away Plus study arm was maintained beyond 12 months. Given that a specific aim of the trial was to explore how best to promote physical activity in South Asian communities, it was disappointing that there was not greater participation by South Asian individuals in the focus groups investigating the uptake and experience of the mHealth intervention. Finally, the rapid and continuing evolution of potential mHealth technologies and the changing landscape in terms of their take-up by the general public complicated the task of selecting the optimal technology for our purposes. Smartphone use became increasingly common during the study; when PROPELS was in development, smartphone ownership was lower. Moreover, access to communications technologies is not distributed equally: lower income households and over 54s continue to be less likely to have smartphones, laptops and tablets. Only mobile telephones and televisions have near-universal reach in the UK (96% of households). This suggested that interventions integrating SMS text messaging technology would have the greatest potential for maximal reach. However, this will change over time and other technologies will need to be considered in the future.
Conclusions
In conclusion, the PROPELS study demonstrated that combining a pragmatic physical activity intervention with text messaging and telephone support can result in modest, but clinically meaningful, changes to ambulatory activity over 12 months, but that such changes were not maintained at 48 months.
The economic evaluation showed that a pragmatic intervention for promoting physical activity to those at risk of T2D was not cost-effective and did not lead to longer-term behaviour change, even when combined with mHealth technology.
Implications for decision-makers
One interpretation of the findings from the PROPELS study suggests that individual-level approaches to physical activity behaviour change may not be effective over the longer term, however good the intervention. It is possible that, in the long term, the environmental drivers of inactivity will always triumph over efforts to address individual motivation unless these are integrated with wider systems-level interventions and policies that target the wider built environment and policies that help make physical activity more accessible, enjoyable and safe.
However, given that the population in this trial was already physical active, which may have limited generalisability to those referred to diabetes prevention programmes within routine care, these findings do not in and of themselves constitute sufficient evidence to support a change in direction for diabetes prevention policy in England or elsewhere. Rather, they support the need for routinely delivered diabetes prevention to be subject to rigorous ongoing evaluation, and for policies to be revised if they are found to have limited longer-term benefits.
Recommendations for future research
Further research is needed to:
Investigate which methods for maintaining physical activity behaviour change are effective over the long term in different ethnic populations. This includes investigating which intervention types, components and features – such as behaviour change techniques, real-time support and environmental change – can help maintain physical activity behaviour change and are cost-effective over the longer term and whether or not different solutions are needed for different populations or ethnic groups.
Evaluate the long-term effectiveness and cost-effectiveness of routinely delivered national diabetes prevention programmes. Given the results of the PROPELS programme, it is important that research frameworks are wrapped around routinely delivered diabetes prevention programmes to determine their effectiveness and cost-effectiveness over the longer term.
Test a stepped prevention programme of initial lifestyle intervention before offering pharmacological interventions (e.g. metformin) to those who do not adhere to or are unable to take up lifestyle interventions.
The PROPELS trial demonstrated that up to half of individuals offered a pragmatic behavioural physical activity intervention will not adhere to all elements, with around 20% not attending even the first educational session. This indicates a need to investigate whether or not offering alternative pharmacological therapies to those who cannot or do not want to engage with lifestyle interventions is effective or cost-effective compared with no intervention or either intervention alone.
Explore the importance of risk status and risk communication to behaviour change. This includes investigating how feeding back biochemical data in the normal ranges influences participants’ perception of their own risk, and whether or not this discourages behaviour change.
Illuminate the importance of integrating rehabilitation from illness or injury as a core intervention component to sustain long-term physical activity behaviour change.
The findings from this qualitative research aspects of this study suggest that future interventions should consider integrating rehabilitation from illness or injury as a core intervention component in order that initial behaviour change success is more resilient to derailment by injury or illnesses and so is better sustained over time.
