Background

The number of people with type 2 diabetes is reaching epidemic proportions all over the world [1]. In Western societies, a social gradient in the prevalence of type 2 diabetes and its risk factors is also well-documented, vulnerable population groups having greater burden than those in higher social strata [2,3,4]. The clinical manifestation of type 2 diabetes usually appears later in life, but many of the risk factors and behaviours develop much earlier and many disparities in health are rooted already early in life. The prevalence of the disease is growing also in younger individuals as a result of increasing obesity rates, unhealthy diet and physical inactivity already present during childhood [1].

Tackling the type 2 diabetes epidemic is a major public health challenge. Most of the diabetes prevention interventions so far have been targeted at middle-aged people who already have non-diabetic hyperglycaemia [5,6,7]. These interventions are highly warranted, as identified high-risk individuals cannot be left untreated and the achieved risk reduction has been shown to be most pronounced among individuals who are already close to the diagnostic limit [8]. However, to achieve largest impact on population level, prevention emphasizing healthy lifestyle should be started already during childhood and continued throughout the life course. A new challenge is to learn from the previous type 2 diabetes interventions and tailor them for younger individuals who have traditionally been considered as low risk and who therefore have not received the appropriate attention [9].

The objective of the Feel4Diabetes project was to develop, implement and evaluate a school-, community- and family-based intervention program for the prevention of type 2 diabetes among vulnerable families with children in primary school, in six European countries during 2015–2019 [10]. The 2-year intervention included two components: “all families component” and “high-risk families component”. The all families component was targeted at school-aged children and their families and tailored to improve the diet, physical activity patterns, and body weight according to national guidelines. A school setting was chosen to reach families with different socioeconomic backgrounds and to utilize the school as an intervention venue. Feel4Diabetes also focused on areas with lower socioeconomic status to reach the most vulnerable populations for type 2 diabetes.

The parents of the participating families filled FINDRISC-questionnaire and the parents with high risk scores were invited to participate in type 2 diabetes prevention study, e.g. the high-risk families component. These parents got in addition to school based activities more intensive intervention, including individual and group sessions about type 2 diabetes, healthy eating and exercising following a SMS intervention based on tips and reminds about healthy lifestyle. Feel4Diabetes-study was registered in clinicaltrials.gov with registration number NCT02393872.

As part of the PRECEDE phase of the PRECEDE-PROCEED model of Feel4Diabetes [11] several literature searches were completed, to guide the development of evidence-based implementation of the Feel4Diabetes intervention. In addition to this review focusing on adults, a review focusing on studies implemented in school setting aiming to enhance healthy lifestyle in children was conducted [12]. The aim of the work presented in this paper was to systematically review the available research literature on type 2 diabetes prevention strategies targeted at adult high-risk individuals and find state-of-the-art methods in all phases from risk identification to implementation and maintenance to use in Feel4Diabetes high-risk families component. The primary literature search was conducted in 2015 before beginning of the Feel4Diabetes-study and updated in 2019 to provide a comprehensive review of the subject. Specifically, our aim was to pinpoint effective type 2 diabetes prevention strategies regarding vulnerable population groups, as well as strategies that have been successfully implemented among under middle-aged population groups.

Methods

Search strategy

For the primary literature search conducted in 2015 a search strategy was developed in consultation with an information specialist. The information specialist completed three searches using search terms related to ‘diabetes’, ‘prevention’, ‘intervention’ and ‘efficacy’. The first search was a general search for diabetes prevention interventions from scientific literature in OvidSP (MEDLINE), Web of Science, EBSCOhost and Cochrane databases. The second search was targeted to interventions on vulnerable populations and in addition to search terms used in search one, the term ‘vulnerability’ and related terms were used. The third literature search was completed to find grey literature using Open Grey, Greylit -reports, NICE Evidence Search and Google search engines with same search terms as in the first one. The question about identification of high risk adults was explored in the context of preventive interventions. The information specialist did preliminary selection according to the search strategy. Two reviewers independently examined titles and abstracts and selected relevant articles according to inclusion criteria. In case of disagreement, inclusion was resolved through discussion.

The complementary literature search was done in 2019 to update the original work with the most recent type 2 diabetes prevention studies. The search terms used were the same as in first search for the primary literature review and the search was done using PubMed (MEDLINE). The complementary search was done by one reviewer and the search was complemented using cross-references in the already included publications and reviews to ensure coverage.

Inclusion and exclusion criteria

Inclusion criteria based on title and abstract were:

  1. 1.

    Type of study: Randomized controlled studies (RCT) or pre-post intervention studies that considered the effectiveness of a lifestyle intervention (diet and/or exercise).

  2. 2.

    The stated aim of the study: type 2 diabetes risk reduction or prevention of type 2 diabetes.

  3. 3.

    Population: Adults (18 years and over) identified as being at high risk of developing type 2 diabetes identified having prevalent risk factors (for example obesity, sedentary lifestyle, family history of diabetes, metabolic syndrome, impaired glucose tolerance (IGT), impaired fasting glucose (IFG), prediabetes, hyperlipidemia, gestational diabetes, cardiovascular disease, elevated diabetes risk score or elevated cardiovascular risk score).

  4. 4.

    Outcome of the study: Development of diabetes or change in diabetes risk, measured by a reliable and scientifically approved risk marker like weight, body mass index (BMI), fasting glucose or glucose tolerance.

  5. 5.

    Study published: in the English language and as full-length articles between January 1st, 2000 and January 29th, 2015 in primary search OR between January 29th, 2015 and February 28th 2019 in complementary search

  6. 6.

    Follow-up time of at least 12 months OR at least 6 months if median/mean age of participants was < 45 years

After the selection of relevant publications based on abstracts (n = 232 in the primary search), the publications originating from the same study were combined, the full-length papers were acquired and read. Studies were excluded if they did not meet the inclusion (if the study was not RCT or pre-post study, the aim was not to prevent type 2 diabetes, participants inclusion was not based on type 2 diabetes risk, the outcome was not a measured risk marker for type 2 diabetes or the article was not in English and published before January 2000) criteria or the study population included a large proportion of people with diabetes (over one fourth), the results of the primary endpoints were not published or follow-up time was less than 6 months. Originally, we decided to exclude studies with less than 12 months follow-up time, to emphasize the evidence on long-term effectiveness of the intervention. However, as the research including younger participants (< 45 years of age) proved to be scarce, we modified the criteria to include studies with at least 6 months follow-up time if they included participants within the age range of 18 to 45 years.

Data synthesis

The selected publications (n = 80) in the primary search showed that the majority of the published diabetes prevention studies have been targeted at older population groups than the target group of the Feel4Diabetes intervention (parents with school-aged children). It is known that increasing age is a significant risk factor for type 2 diabetes and the study by Deeks et al. [13] found age dependent differences in health beliefs and screening participation rates. Older people were more likely to participate in specific health checks including blood glucose and cholesterol measurement than younger people. Presumably older people have different life circumstances and thus different barriers for participation and changing lifestyles compared with younger ones. Relying on studies with mainly older participants (as those with sole number override the studies on younger people) might have steered the conclusions off target. Therefore, two different review approaches were conducted. In first approach the studies with the participants aged ≥18 years and minimum follow-up of 12 months (n = 27) were reviewed. In second approach the studies with mean or median age of participants less than 45 years and minimum follow-up of 6 months (n = 9) were included. In the complementary search all studies had follow-up time over 12 months (n = 10). A flow chart of the selection of relevant studies is presented in Fig. 1.

Fig. 1
figure 1

Flow chart of the literature review process

Full size image

To present the findings from the literature reviews in a systematic way, we used the Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) framework designed for assessing interventions and public health programs [14]. The overall goal of the RE-AIM framework is to encourage implementers to pay more attention to core elements, which can improve adoption and implementation interventions.

The summary tables of the selected studies (see Tables 1, 2 and 3) were prepared and reviewers independently evaluated the clinical significance of the results presented for each study, to facilitate interpretation of the effectiveness versus the design, methods, delivery, and costs of intervention. The clinical significance of the study results was scored as follows: meaningful reduction in diabetes risk; meaningful improvement in (most) target risk factors; meaningful improvement in some/few risk factors; or no effect.

Table 1 Overview of the studies targeting participants aged ≥ 18 years and minimum follow-up of 12 months
Full size table
Table 2 Overview of the studies targeting participants age 18–45 years and with minimum follow-up time 6 months
Full size table
Table 3 Overview of the studies in the complementary search for studies published between January 2015 and January 2019
Full size table

Results

In the primary literature review of publications dated Jan 2000-Jan 2015, searches identified 663 potentially relevant publications, of which 80 studies met the initial inclusion criteria. For the first approach, altogether 27 studies targeted at population aged ≥18 years were reviewed after discarding the studies with the follow-up time under 12 months (Table 1). Of these, 12 were completed in Europe [6, 26, 29, 30, 32, 33, 35, 37, 41, 48, 55, 86], five in the USA [36, 44, 45, 50, 87], three in China [5, 47, 54], four in Japan [22, 28, 51], two in India [52, 88] and one in Australia [31].

In the second approach, the inclusion criteria of participants’ mean age ≤ 45 and follow-up at least 6 months were fulfilled only in nine studies, of which two were completed in Europe [26, 48], five in the USA or Canada [36, 57, 58, 60, 62], and two in China [5, 54] (Table 2). There were five studies which fulfilled inclusion criteria for both reviews [5, 26, 36, 48, 54].

The complementary search found 12 studies published after January 2015 (Table 3). In brief, six studies were conducted in USA [63, 64, 66,67,68,69,70,71,72, 75,76,77] (most were based on the DPP intervention implemented with adaptations in various settings), two in Spain [65, 78,79,80], two in India [74, 81,82,83,84], one in Israel [85] and one in Malaysia [73]. All 12 studies had a mean follow-up of at least 12 months and 5 included younger individuals (mean age ≤ 45 years old) [65, 72, 74,75,76, 85] and another three reported a mean participant age of ≤50 years old [66,67,68,69,70,71]. The core elements of the implementation of the high-risk intervention were identified through a synthesis of selected studies using RE-AIM model [14].

The core elements of implementation

Reach

The total number of participants was over 13,000 in the studies with the participants aged ≥18 years and minimum follow-up of 12 months. Most common inclusion criteria in these studies were IGT or IFG based on laboratory tests (mentioned in 16 studies). In addition to laboratory tests, the FINDRISC diabetes risk score was used as an inclusion method in five studies [30, 32, 35, 37, 41] and used as a pre-screening instrument in one study [30]. Three studies used other risk scores (e.g. AUSDRISC) or risk algorithms for inclusion [26, 33, 44]. In addition, previous gestational diabetes was the basis of inclusion in one study [54], and BMI in another study [36]. It appears that none of the risk identification methods was superior as regards to the subsequent effectiveness of the intervention. The process of screening and recruitment was often reported to be laborious and more time-consuming than expected. When initial contact was not targeted at a risk population, the final inclusion was in general less than 10%, for example 0.5% in Da Qing [5], 2% in the Diabetes Prevention Program (DPP) [87] and approximately 10% in the Diabetes Prevention Study (DPS) [6]. The proportion of included subjects was lower in studies that had higher baseline risk (e.g. IGT) as an inclusion criterion. Only two studies stated that they were targeted at “underserved” or socioeconomically “vulnerable” population groups [36, 44]. Both of them showed significant improvement in target risk factors. In addition, two studies [37, 48] either included or were targeted solely at minority groups such as immigrants.

From the studies targeted at less than 45 years old participants, the major difficulties were related to participation. Lack of interest to take part in the studies was common and the drop-out rates tended to be in general high and participation in interventions in general inadequate. For example in The Families united study only 18 participants, of 90 eligible screened, attended the 12 month follow-up [58] and in DH!AAN only 22–26% of participants attended the lifestyle sessions. Some studies had completed focus group interviews and engaged the community already in the planning phase [49], but that did not have an unambiguous effect on the actual participation or the achieved lifestyle changes. Inviting people from e.g. registers as opposed to asking volunteers with self-perceived risk to sign up has been especially challenging and the response rate has generally been low. The participant selection method has also been reflected in attendance in interventions and evaluation measurements (especially in control groups or control areas), resulting in higher attrition. Five out of the nine studies, with the participants aged ≤45 years and the minimum follow-up of 6 months, were targeted at “vulnerable” population groups, such as native North Americans [57], public housing communities [60] or underserved Latino population in USA [36], or immigrants in the Netherlands [48]. The results from these studies were comparable to other 4 studies targeting general high diabetes risk population.

In complementary search 5 studies were targeted on vulnerable groups, one on economically disadvantaged adults [64], one on African-Americans in Georgia US [66], one on low-income Hispanic women [75], one on Socioeconomically disadvantaged Hispanic females in Philadelphia [75, 76] and one in developing country [81].

Adoption and implementation

In most of the studies with follow-up time over 12 months, the lifestyle goals were based on DPP or DPS and were related to body weight (reduction 5–7% recommended), changes in diet and increase in physical activity. Frequent dietary goals were to increase fibre, whole grains, fruit and vegetables, and to reduce total, saturated and/or trans-fat, sugar, refined carbohydrates, starch, alcohol and/or total calorie consumption. Also diet related lifestyle targets such as “reducing the frequency of eating out” were mentioned as target [22]. The studies differed from each other in relation to what and how much each target was emphasized. The participants in Chinese studies were not in general overweight at baseline, so weight reduction target was typical only for the subgroup of overweight or obese people [5]. In the PREDIMED-Reus study completed in Spain and emphasizing Mediterranean diet enriched with extra-virgin olive oil or nuts, there was neither weight reduction (despite high rate of obesity) nor physical activity target but yet a significant reduction in diabetes risk was observed [33].

Coach-delivered, face-to-face, individual (n = 11), group (n = 7), or group-individual-combination (n = 5) interventions were the most common delivery modes. In addition, in one study [28], intervention was delivered in hospital (in-patient) and in one study [37] intervention personnel went to the family homes. Short message system (SMS) was used in two studies [47, 52] as the primary intervention method. In addition, some studies used phone calls and telefax. SMSs proved to be a promising way to deliver intervention in a cost-effective way. The intensity of intervention (number and frequency of counselling sessions/contacts between the personnel and the participants) appeared to be more important for effective intervention than the mode. The coach has most often been either a nurse or a dietician. As long as the coach is trained appropriately to do the intervention and applies a structured intervention curriculum, there does not seem to be a big difference between professions.

Several theories of behaviour change were applied as the basis for the interventions. The most often mentioned theory was the stages of change/trans-theoretical model [24, 36, 41, 51, 52, 89] (applied in 6 interventions), followed by the theory of self-regulation (in three interventions) [26, 30, 31], the theory of planned behaviour (in 2 interventions) [26, 47], social cognitive theory of behaviour change (in two interventions) [44, 47], and the health action process approach, HAPA (in one intervention) [31]. None of the theories seemed to be unequivocally better or worse than the other and many of the highly successful studies have not stated specifically relying on an overarching theory of behaviour change. The reason for this might be that even though the theories have differing background presumptions about the process of behavioural change they utilize more or less the same components, techniques and tools, such as motivational interviewing, social and peer support, interactive learning strategies, behavioural support on goal-setting and self-monitoring, problem-solving and feedback. Of the two studies using SMS, the Indian study [53] utilized the tailoring of the messages based on individuals’ estimated stage of change and showed significantly better results compared with the Chinese study [47] using generic messages. The studies where motivation of participants was emphasized were successful. In many studies, self-monitoring and personal goal setting were seen as very important. In less successful studies there was a high drop-out rate, which might be related to motivation. In the studies targeted at less than 45 years old participants, the results of the intervention did not seem to depend on the theoretical framework or, whether there was a theoretical framework mentioned at all. However, several authors still emphasized the need for a theoretical framework and structured intervention. Most studies, however, described components, techniques, and tools that are included in many of these theories.

The feedback from those who actually participated in the interventions was in general very positive. The authors’ recommendation in several studies was not to cut down the number of contacts/sessions and topics, as this would lead to dissatisfaction by the participants, but to increase the number of sessions and offer the sessions with shorter intervals, paying special attention to accessibility such as timing of the sessions and e.g. offering child care when needed. Also offering a variety of intervention modes (individual, group, SMS, telephone, DVD, internet) to choose from was considered a feasible strategy to reach participants in studies with mean age under 45 years old. Incentives for participation were recommended by some authors. Community partnership was considered important to train and support community health workers. However, employing full-time project staff as opposed to expecting local community workers to do the project in addition to their normal work was emphasized. One study (Families United) [58] aimed at recruiting a family member as a support person but that proved to be a challenge.

The complementary search showed that the basic contents of the interventions in the recent studies had not changed from the first diabetes prevention projects. Most studies applied traditional intervention modes, i.e. group and/or individual counselling sessions based on behavioural change techniques delivered by health professionals or trained non-medical community members. As reported in previously published studies, non-medical individuals delivered efficient interventions. Many studies highlighted the importance of attendance for intervention efficacy and the need of strategies to increase and sustain patient engagement; among studies targeted on women with GDM and on people with low socioeconomic status, adequate attendance was promoted through family involvement and childcare offer [70,71,72]. Also the community, in many innovative ways, was utilised as an intervention or recruiting place, for example the church as an intervention state for African-Americans [66], a workplace for employees in information technology industry as the recruiting and intervention venue [74] and peer-led intervention as a way to involve community in a study in India [84]. In the study using telephone/newsletter [70, 71] and the study using SMS/email technology [74] to deliver the intervention the results were comparable with others. Lifestyle advice through telecommunication was considered as an efficient, low-cost and potentially scalable intervention for technology-literate individuals. Even though different mobile phone applications, activity trackers and other modern technology have become widely available, the technology was not systematically used in modern studies.

Efficacy and maintenance

Of the 27 studies including the participants aged ≥18 years old and minimum follow-up of 12 months, eight [5,6,7, 22, 23, 28, 33, 40] were rated highly successful and showing meaningful reduction in diabetes incidence. In seven studies [30, 32, 35, 36, 44, 45, 50], meaningful improvement in (most) target risk factors were seen; in those studies reduction of diabetes incidence was either not a target or not achieved. In eight studies [29, 37, 41, 51, 54, 55], meaningful reduction in some/few risk factors was achieved. Only four studies [26, 31, 47, 48] failed to show any effect on risk factors or diabetes risk. In most studies including younger participants (18–45 years old), achieved results/changes in predefined outcome variables were less-pronounced than in studies with older participants. The exception for this rule was the Chinese Da Qing study [5], where a significant and highly meaningful reduction in diabetes risk in 6 years was achieved. However, also in that study there was no reduction in body weight in general, and the measured changes in diet and physical activity were modest. In general the achieved lifestyle changes were minor compared with changes in the studies on older participants. This might be due to lower self-perceived risk and life situation in general, such as “demands of motherhood and family life” as stated in one of the studies [54].

In the complementary search significant reduction in diabetes risk was distinguished in the two studies conducted in Spain of all 12 reviewed studies [4, 79, 80]. Both of these studies were conducted at a medical environment (primary health-care centres or hospital) by medical personnel, applied intensive interventions in terms of patient contact and reinforcement and had a long follow-up period of 2 and 3 years. In 8 studies significant reductions in most of targeted diabetes risk factors were achieved but reduction in diabetes risk was not stated. In two studies significant reductions were achieved only in some diabetes risk factors [63, 64, 75, 76]. In those studies, diabetes risk assessment for recruitment was based on Hb1Ac levels, which may have led to recruiting participants with relatively low baseline fasting glucose values and without room for improvement. Significant changes in glycaemic control were not seen, although interventions were efficient in achieving weight loss and improving body composition.

Discussion

The systematic literature reviews revealed and highlighted several important aspects that were subsequently taken into account while developing the Feel4Diabetes high-risk intervention. To improve effectiveness as well as sustainable adoption and implementation of interventions, they should be targeted at people with increased type 2 diabetes risk. Risk identification can be based on fasting or 2 h blood glucose measurement, however, also non-invasive methods can be used. Since the publication of the FINDRISC in 2003 it has been used in several studies as the first-line or even sole risk screening tool. In the major type 2 diabetes prevention trials [6, 90] the oral glucose tolerance test was used as the screening method and IFG or IGT as inclusion criteria. In the studies included in this review, the risk identification method appeared not to be associated with the effectiveness of the intervention. The selection of risk identification method may thus be based on pragmatic issues such as cost, acceptability, and accessibility, especially when completing an implementation project. Of note, the measurement of glycated haemoglobin (HbA1c) has been found clearly inferior to oral glucose tolerance test (OGTT) in identification of prediabetes based on the meta-analysis by Barry et al. [91]. An important finding was that the process of screening and recruitment is often laborious and more time-consuming than expected; especially people with lower socioeconomic status may require additional effort and action. Thus, screening is a critical step of preventive interventions and should not be overlooked. Most studies included in this review show that changes in diabetes risk factors are similar regardless of whether the intervention is delivered by experts (clinically trained health professionals) or lay educators; therefore, costs associated with diabetes prevention can be lowered without sacrificing intervention effectiveness, involving nonmedical personnel.

Equally important is to arrange the intervention so that it is easily and conveniently accessible. In the younger age-group, the most important reasons for non-participation were lack of time and difficulties in participating in the scheduled counselling sessions. Furthermore, in most studies including younger participants (18–45 years old), achieved results/changes in predefined outcome variables were less-pronounced than in studies with older participants. Obvious reason for this is that a longer follow-up is needed to see effect on diabetes risk in younger people. The complementary literature search showed that the same challenges of recruitment and participant engagement continue in recent studies as in previous studies. Wider use of modern technology could help participants to commit on the intervention and the use of different community settings were seen helping the recruitment. Even in the most recent studies technology was quite rarely used, even though it was seen as efficient as the classic face-to-face counselling. Targeting intervention earlier in life might be a trend in most recent studies, 40% of studies conducted between 2015 and 2019 was targeted to people under 45 years old. Technology could be part of a solution to engage and reach younger participants.

The focus and general goal of intervention should be clearly specified and communicated. Cultural adjustments to the intervention goals probably increase the participation and motivation to make the suggested lifestyle changes. The intensity of intervention (number and frequency of counselling sessions/contacts between the personnel and the participants) is more important for effective intervention than the mode of delivery. Moderate but comprehensive changes in several lifestyles seem to lead to a good intervention effect. At least 3 years follow-up seemed to be required to show actual reduction in diabetes risk in high-risk individuals.

Theoretical model is considered important as a framework for the intervention. However, as long as it facilitates the understanding of the complexity of behavioural change, it doesn’t matter which model is used. Tools and methods shown to be efficacious include motivational interviewing, social and peer support, interactive learning and motivation and self-efficacy building strategies, support on individualized goal-setting, self-monitoring, problem solving, relapse management, and feedback.

Importantly, research on prevention interventions targeted at younger adults or vulnerable population groups such as people with lower socioeconomic position proved to be surprisingly scarce. More research is warranted, and Feel4Diabetes is an important example of projects aiming to fill this research gap.

Conclusions

This narrative review highlighted several important aspects that subsequently guided the development of the Feel4Diabetes high-risk intervention. Research on diabetes prevention interventions targeted at younger adults or vulnerable population groups is still relatively scarce. Feel4Diabetes is a good example of a project aiming to fill this research gap.