Skip to main content

PREVention of diabetes through lifestyle Intervention and population studies in Europe and around the World

Final Report Summary - PREVIEW (PREVention of diabetes through lifestyle Intervention and population studies in Europe and around the World)

Executive Summary:
PREVIEW addressed potential solutions to the massive problems associated with the global diabesity epi-demic (obesity and type-2 diabetes, T2D). PREVIEW included partners from Europe (East, West, North and South) and Australia, New Zealand, and Canada. The primary goal was to identify the most efficient life-style pattern for the prevention of T2D in a population of pre-diabetic overweight or obese individuals. The project comprised two distinct lines of evidence, both embracing European and overseas countries: 1) A multicentre, randomized, controlled intervention trial (RCT) with a total of 2,500 participants with over-weight or obesity and pre-diabetes, including children, adolescents, adults, and elderly. The duration was 3 years for adults and elderly, and 2 years for children and adolescents. 2) Large population studies using multinational data sets from all age groups. Focus for both lines of evidence was on diet composition (spe-cifically protein and glycaemic index, GI) and intensity of physical activity (PA), the interaction with the life-style factors: habitual stress, sleeping patterns as well as behavioural, environmental, cultural, and socio-economic variables.
The RCT for adults started with a 2 months weight loss phase on a low-calorie diet (LCD, the Cambridge Weight Plan). After successful weight loss (≥ 8%), participants were instructed to eat one of 2 healthy diets (high-protein/low-GI or medium-protein/medium GI) and undertake one of 2 exercise strategies (medium or high-intensity PA). Over the period of 34 months involving 18 face-to-face visits, they were provided with a high level of counselling and support in a group setting. A total of 15,611 adults were pre-screened, 5,472 screened, 2,326 enrolled, and 962 completed the 3-y trial. The most important finding was that 96% of the participants who successfully lost ≥ 8% of their total body weight (n = 1,857) and went on to com-plete the trial did not develop T2D within 3 years, despite regaining some weight. There were no differ-ences in outcomes between the two diets or two PA regimes. Only 4% of those who completed the 3-y in-tervention developed T2D (n=62), whereas, based on findings from similar lifestyle interventions, it was estimated that about 13% would progress to T2D in that time. One key is likely the substantial weight loss at the start of intervention (average 11% vs 5-7% in other similar studies) followed by the healthy lifestyle, weight maintenance programme.
In the children and adolescent part, 126 were included in the RCT, 86 completed 1-y and 49 completed the 2-y study. There were no differences in the main outcomes between the two diet groups. In the adult part, control of food intake, sustained PA, prevention of sleepiness, stress, mood disturbance, disinhibition of eating and hunger were all related to the prevention of T2D. Further, in children sleep duration was in-versely related with BMI z-score, while BMI z-score was positively related to insulin resistance, emphasis-ing a role for sleep duration in insulin sensitivity. A relationship between diminished brain reactivity in food-reward related brain areas implied that more control of food intake and increases in dietary restraint, protein intake and insulin sensitivity may contribute to body weight management.
The PREVIEW Behaviour Modification Toolbox (PREMIT), designed to support lifestyle habit changes in the RCT, reinforced participants' understanding of behavioural control, thus supporting maintenance of healthy PA and diet behaviours. However, younger participants and those with higher BMI appeared to be at higher risk of missing out on the support offered by the intervention. Health Impact Assessments con-cluded that especially in participants who were successful at losing body weight and maintaining it, poten-tial health impacts included not only reduced risk of developing T2D, but potentially lower risk of develop-ing cardiovascular and musculoskeletal diseases. Finally, the moderate protein diet was more cost-effective.
The population studies included ≈ 162,000 participants from the Lifelines (NL), NQPlus (NL), Cardiovascular Risk in Young Finns Study (FI), New Zealand Adult Nutrition Survey 2008/09 (NZ), and Quebec Family Study (CAN). The results were inconsistent across cohorts, but the overall findings suggested that lower GI and higher protein, moderate PA and longer sleep are important in prevention of diabetes and related diseas-es.

Project Context and Objectives:
Type-2 diabetes (T2D) is a disease associated with serious comorbidities, including microvascular (retinopathy, nephropathy, neuropathy) and macrovascular (cardiovascular) events. Reducing the personal, societal and economic burden of T2D is one of the most important challenges facing almost all nations. An im-portant risk factor for T2D is obesity (BMI ≥ 30 kg/m2) predicting a more than 10-fold increase in risk of T2D relative to normal weight. Previous long-term studies have shown that a lifestyle intervention (diet, PA, weight loss) may reduce the incidence of T2D by 28–59%. These studies were all designed to produce weight loss by prescribing a high-carbohydrate (CHO) (>50 percent of energy (E%) from CHO), low-fat (<30 E%) diet, reflecting a healthy diet at the time. Glycaemic index (GI) was not considered and, to date, no studies have investigated the role of GI for the prevention of T2D. More recently, the multicentre trial ‘Di-OGenes’ (EU-FP6) found that a combined higher-protein, moderate-CHO, low-GI diet was superior to 4 other diets of varying macronutrient and GI composition in preventing weight regain over 6 months after initial rapid weight loss in overweight, healthy individuals. Other studies have also suggested that a rela-tively high (0.8 -1.2 g/kg daily) vs. moderate (< 0.8 g/kg daily) protein diet promotes body weight mainte-nance for 6 – 12 months.
In general, exercise is connected with beneficial health effects. To date, there is, however, no evidence that one type exercise regime is superior to another in T2D prevention. Sleeping pattern may also be an important risk factor for T2D. Thus, a deviation from normal sleeping patterns (7-8 h sleep per night), par-ticularly short sleep has been found to increase appetite and promote obesity and its related diseases (e.g. T2D and cardiovascular diseases). The long-term effects are, however, still unknown.

Main project objective
The main objective of PREVIEW was to determine the extent to which a high-protein (HP), low-GI diet in combination with moderate or high intensity PA could reduce the incidence of T2D in pre-diabetic over-weight children, adolescents, adults, and elderly compared with a moderate-protein (MP), higher-GI diet. Furthermore, the interaction of lifestyle factors (diet, PA) with sleep, habitual stress and other behavioural and sociological factors in the prevention of T2D was investigated. The objectives were achieved by using data both from a multinational intervention study and from large population studies.
The project consisted of 6 work packages (WP): WP1: Multicentre intervention: Randomized, controlled, multicentre trial (RCT). WP2: Population studies (PS). WP3: The role of sleep and stress in interaction with the role of diet and PA. WP4: Other lifestyle variables: Behavioural, sociological, environmental, cultural, socio-ecological, and socio-economic components. WP5: Dissemination and exploitation. WP6: Management.
Specific objectives - long-term randomized controlled trial (RCT) (WP1).

A randomised, controlled, multicentre trial (RCT) was conducted in participants with overweight (BMI ≥ 25 kg/m2) and pre-diabetes. The trial was conducted in 9 intervention centres in 8 different countries: Univer-sity of Copenhagen (UCPH, DK), University of Helsinki (HEL, FI), University of Maastricht (UM, NL), Univer-sity of Nottingham (UNOTT, UK), University of Navarra (UNAV, ES), Medical University of Sofia (MU, BG), Swansea University (UK), University of Sydney (UNSYD, AUS), and University of Auckland (UOA, NZ).
Our primary hypothesis was that the HP, low-GI diet would be superior in preventing T2D compared with the MP, higher-GI diet. Further, we hypothesized that high-intensity PA would be superior in preventing T2D compared to moderate-intensity PA. The primary endpoint was incidence of T2D at 3 years according to diet, based on a 75 g oral glucose tolerance test (OGTT). For children and adolescents, the primary end-point was change in homeostatic model assessment of insulin resistance, HOMA-IR, at 2 years. Secondary endpoints were changes in body weight and waist circumference, insulin sensitivity, risk factors for cardio-vascular disease (e.g. blood pressure, lipids, C-reactive protein), changes in perceived quality of life and satisfaction with work, habitual well-being and stress, appetite sensations, and PA. Finally, adverse events and concomitant medication were registered (Clinicaltrials.gov Identifier: NCT01777893).
The primary power calculation was based on two intervention arms (i.e. HP vs. MP diet). From previously published data (Lindström et al; The Finnish Diabetes Prevention Study; Diabetes Care 2003;26:3230–6. Knowler et al; Diabetes Prevention Program Research Group; N Engl J Med 2002 Feb 7;346:393-403), the 3-y incidence of T2D in people aged 40 y with a diabetes risk score ≥ 15 is 21%. We hypothesized that a risk reduction of 25% in the MP (control) group would reduce the diabetes incidence in this group to 15.8%. The estimated 25% risk reduction was based on outcomes achieved in published diabetes prevention trials. We hypothesized that the HP group would achieve an overall 50% diabetes risk reduction from a baseline risk of 21% to a 3-y risk of 10.5%. Using these assumptions, the sample size required to detect this differ-ence in incidence (15.8% vs. 10.5%), with a power of 80% and alpha of 0.05 was at least 649 per arm or 1,298 adult participants in total. We estimated a 30% drop-out rate (similar to the FP6-project, “DiOGenes”) during intervention. Thus, we needed at least 1,854 participants to start the weight-maintenance interven-tion. To further allow for a drop-out after inclusion and for participants not losing 8% weight during the 8 weeks’ weight loss phase, a total of 2,500 adult participants were to be recruited.
All adults started with an 8-wk weight loss phase on a low-calorie diet (LCD, Cambridge Weight Plan) providing 800-1000 kcal/d. Those who achieved the target weight loss (≥ 8% of initial body weight) were in-cluded in the second phase, the 34-month weight-maintenance phase. For this phase, participants were centrally cluster-randomized in blocks so that the 4 dietary and PA arms were equally represented in each centre:
Group 1: HP-HI, High-protein, high-intensity PA
Group 2: HP-MI, High-protein, moderate-intensity PA
Group 3: MP-HI, Moderate-protein, high-intensity PA
Group4: MP-MI, Moderate-protein, moderate-intensity PA
For children and adolescents, a slightly modified intervention was followed: an initial 8-wk run-in phase on an individually calculated healthy diet for maintaining body weight and a second phase with a long-term weight-maintenance intervention (in total 22 months).
Dietary counselling and exercise instruction visits took place at community or university based centres in groups (n = 8-12) every second week the first 8 wks, monthly to 6 mo, quarterly to 12 mo, and then at mo 18, 24 and 36. At each visit, a dietitian/lifestyle counsellor advised on weight control and reinforced the diet composition through recipes, cooking advice, and behaviour modification. The dietary instructions were standardised for all centres, but reflected local dietary customs and food products. All participants were given written and oral instructions to achieve the target diet composition. For the children and adoles-cents, the study included a more flexible visit schedule, and individual counselling.
The main data collection points were at the clinical investigation days, (CID), at baseline (beginning of weight-reduction phase, CID1) and at month 2 (CID2), 6 (CID3), 12 (CID4), 18 (CID5), 24 (CID6), and 36 (CID7) of the intervention. Several procedures were performed, including sampling of blood, urine, faeces (3-d in a sub-group) as well as a range of questionnaires for subjective evaluations (e.g. appetite, stress, sleep). For data storage, a central project database (datahub) was created at UCPH. All data, including lab analyses, questionnaire data gathered via the specially designed Questionnaire Delivery Platform (QDP) and anthropometric assessments, were continuously uploaded to the datahub. The food consumption and nutrient intake data were derived from repeated 4-d food records. The laboratory analyses were done from serum, plasma, blood and urine specimen at each CID. To obtain the same quality throughout the study, all key laboratory analyses for each CID were carried out in central laboratories.
The sub-objectives were: To finalize the preparation phase; To screen eligible participants for the study; To complete the 8 weeks weight-reduction phase (including participant supervision and data collection) and randomize participants for the weight maintenance phase; To complete the 34 month weight maintenance intervention (including participant supervision and data collection); To collect, store and analyse data.
Specific objectives - population studies (PS) (WP2)
The main objective of the PS was to evaluate whether the intake of protein and glycaemic index (GI), as well as PA, were predictive of the development of diabetes and its cardiovascular consequences in pre-diabetic respondents of all ages and both sexes in the various epidemiological studies included in PREVIEW. Data from the PS covering the entire lifespan in European and overseas countries was analysed. Focus was on dietary components (protein, glycaemic index), type of PA, sleep duration, and stress. Country-specific epidemiological studies included were: Lifelines (NL), NQPlus (NL), Cardiovascular Risk in Young Finns Study (FI), New Zealand Adult Nutrition Survey 2008/09 (NZ), and the Quebec Family Study (CAN). Our hypothe-sis was that a diet with lowered carbohydrate intake (about 45 per cent energy, (E%)), high protein intake (25 E%) and low GI foods (<55) would be related to reduced incidence of obesity and T2D. The sub-objectives were:
• To compile a single tailored database from the participating epidemiological studies.
• To develop a dedicated GI questionnaire.
• To assess differences in prevalence and incidence of diabetes in participants with high-low protein/GI diets, high-low PA, high-low sleep duration, and high-low (perceived) stress in the participating cohorts.
• To assess the impact of protein/GI/PA in the context of other risk factors of diabetes (ie population at-tributable fraction (PAF) of abnormal glucose tolerance (AGT) and diabetes cases.
Specific objectives - the role of sleep and stress in interaction with diet & PA (WP3)
The main objective was to evaluate the role of sleeping pattern, chronic stress and neural changes in the development of T2D in pre-diabetic participants – and the interaction of these variables with diet and PA. Our hypothesis was that a high-protein, low-GI diet in interaction with high-intensity PA would improve the sleeping pattern and reduce the level of chronic stress during the weight maintenance phase - ulti-mately reducing the risk for diabetes type-2 in pre-diabetic participants. The sub-objectives were:
• To obtain approval from the Medical Ethics committee for (i) local performance of our part of the tasks in WP1; (ii) research on sleep architecture in a sub-cohort of 40 children and adolescents; (iii) research on brain imaging in a sub-cohort of 40 adults.
• To recruit 40 children and adolescents (10-18 y) for sleep architecture and 40 adults for brain imaging.
• To identify the mediating role of changes in sleeping pattern (sleep-quality and sleep duration) and in stress in possible changes in insulin sensitivity, body-composition, waist- circumference and body-weight changes due to the interventions in the cohort described under WP1.
• To identify, in a representative sub-cohort (n=40), the role of sleep architecture and components thereof (duration of sleep latency, sleep phases 1, 2, REM sleep and slow wave sleep) in intervention related changes as described above.
• To identify, in a representative sub-cohort (n=40), possible changes in brain responsiveness of food re-ward-related brain areas to visual food cues with functional magnetic resonance imaging (fMRI), related to changes in insulin sensitivity following the intervention as described above.
• To identify the relationship between sleep duration and protein intake, PA, body-composition, and in-sulin sensitivity in the population based Quebec family study.
Specific objectives - moderating or mediating influence of social-ecological variables (WP4).
The goal was to evaluate the moderating and/or mediating influence of socio-cognitive and environmental variables such as social-cognitive determinants of behavioural change and habitual behaviour, social and environmental influences, cultural habits, socio-ecological and socio-economic components for individuals at risk of developing diabetes using data from both the RCT and the PS. Socio-ecological variables include cognitive variables such as self-efficacy, habits, perception of physical and social environment, as well as demographic variables such as education and employment status. There were two main objectives: (1) To guide and support behaviour change of RCT participants using a theory and evidence based intervention tool. (2) To evaluate the moderating and/or mediating effects of psycho-socio-environmental variables in individuals at risk of developing T2D using data from both the RCT and PS. In addition, the effect of “other life style factors” (eg self-regulation and demographic variables of age and sex) in the cohort studies and reports were also evaluated, as was the cost-effectiveness and health impact of the RCT. In particular, so-cio-cognitive factors influencing behaviour change were examined. Emphasis was on how these complex, dynamic, and multilevel factors influenced not only behaviours, but also outcomes for the individuals. Col-lection of socio-cognitive variables was enabled by development of the QDP The sub-objectives were:
• To identify and provide validated questionnaires
• To make questionnaires available online.
• To enable intervention staff of the 8 RCT centres to help participants of the RCT to initiate health behav-iour change and maintain a healthy life style and to create a logic model regarding the intervention.
• To analyse the influence of “other life style variables” in the RCT.
• To analyse the influence of “other life style variables” in the PS.
• To analyse the cost-effectiveness of the RCT.
• To assess the public-health-impact of the RCT.
Our hypothesis was that the recommended behaviour change and its maintenance would be influenced by the socio-ecological variables. Thus, low self-efficacy and lack of social support could both by themselves and in combination cause a relapse back to an earlier, unhealthy behaviour.

Specific objectives - dissemination and exploitation (WP5)
The main objective was to communicate and exploit the findings of PREVIEW in order to convert new knowledge into socioeconomic benefits, in this case better strategies in public health, products and ser-vices which would reduce the risk of obesity and T2D. The dissemination activities were targeted to have an influence on health and research policy both in Europe and globally. The sub-objectives were:
• To cultivate a dialogue with beneficiaries of the research (consumers, health professionals, food indus-try and the media, including new social media) and sustain two-way communication throughout the project.
• To cultivate and maintain communication between researchers, health professionals and policy makers so that the findings of PREVIEW were translated into a reduction in the incidence of obesity and T2D.
• To identify, secure and manage the transfer of exploitable knowledge and IP generated by PREVIEW to SMEs, the food industry and service providers in the area of health and diet.
• To foster knowledge transfer to beneficiaries and among PREVIEW participating organisations by IT-supported learning modules, laboratory workshops, and interchange of PhD students & Post Docs.

Project Results:
Main results and achievements
The long-term randomized controlled trial (RCT) (WP1).
Preparation phase

The first months of the project (and the preceding several months) were mainly devoted to preparing the large RCT and the protocol, to be approved by the national Ethical Committees for each intervention cen-tre. The protocol was registered at Clinicaltrials.gov Identifier: NCT01777893 and subsequently published (Fogelholm et al, Nutrients 2017; 2017; 9, 632; doi:10.3390/nu9060632).
The two diet interventions were: MP = moderate-protein diet: protein 15% of total energy intake (E%), carbohydrates 55 E%, dietary glycaemic index (GI) >56; HP = high-protein diet: protein 25 E%, carbohy-drates 45 E%, GI <50. Both diets were composed by using healthy food items.
The two exercise interventions were: MI = moderate-intensity: 60—75% of maximal heart rate (HRmax), e.g. brisk walking for adults; HI = high-intensity: 76—90% HRmax, e.g. running for adults. Since the partici-pants were randomized into one diet as well as one exercise group, the total number of combinations (in-tervention groups) was four: MP-MI, MP-HI, HP-MI and HP-HI. The randomization for the adults was strati-fied by sex and age group, and for the children and adolescents by sex.
Also as part of the preparation, guidelines for the two intervention diets including cooking books, and for the two intervention exercise prescriptions, Standard Operation Procedures (SOPs) and Instructions to participants were written and translated into local languages. The paper forms, as well as the electronic case report forms (eCRF) and e-platform were developed. To enforce a consistent data collection, to en-sure the data quality and efficient and secure data storage across centres, a system for electronic data cap-ture (EDC), OpenClinica developed by OpenClinica LLC (https://www.openclinica.com/) was chosen. All CRF’s were matched by eCRF coded in OpenClinica. OpenClinica was a server based solution accessed over the web with a web browser. Data was entered in the web browser and stored centrally immediately upon entry when the investigator pressed "save" in the browser. The server was physically located in Copenha-gen, ensuring that data was stored under European jurisdiction. Furthermore, a questionnaire delivery platform (QDP) including all questionnaires for participants was developed by NetUnion.
The staff in the different intervention centres were trained centrally at a training workshop in Copenhagen which focussed on protocol related procedures, sampling of biological material, measurements, data col-lection, and Good Clinical Practice (GCP). Moreover, a morning session was included on cooking skills and techniques at the partner “Meyers Madhus” (DK). This was meant to provide new ideas for the group counsellors, to facilitate the dietary guidance and to improve adherence to the new dietary programs. Apart from the organized training workshop and subsequent local training/mentoring, an Instructors’ Net-work was formed early in the project to deal with day-to-day practical questions from the intervention centres. Another smaller network of medical doctors was also formed to assist with medical questions which arose during the study. Furthermore, the principal investigators of the RCT had monthly or bi-monthly telephone meetings throughout the whole study to discuss strategies for recruitment and reten-tion of participants and all other relevant matters.

Recruitment of study population
The task of the preparation of instructions and the standard operation procedures, SOPs (in total ≈ 35) was shared between the different intervention centres that had relevant expertise in the field. Within the SOPs, all measurements and procedures were described in detail, so that all trial staff in the 8 intervention centres could follow these in a coherent way and conduct the study as similarly as possible. All partners in WP1 and WP4 participated in writing or reviewing the SOPs - depending on their expertise and experience - to ensure high-quality and up-to-date procedures in the study. All centres started pre-screening and screening potential participants during year 2013 after local Ethical Approval had been obtained.
The main inclusion criteria for adult participants were that they were 25-70 years, had a BMI ≥ 25 kg/m2, and had pre-diabetes. Criteria for assessing pre-diabetes were those recommended by the American Dia-betes Association. Thus, a fasting venous plasma glucose concentration of 5.6-6.9 mmol/L and/or venous plasma glucose concentration of 7.8-11.0 mmol/L at 2 hours after intake of 75 g glucose during an oral glu-cose tolerance test (OGTT), with fasting plasma glucose (FPG) less than 7.0 mmol/L. HbA1c was not used to determine eligibility. A pre-screening was done by FINDRISC (Finnish Diabetes Risk Score). A number of other in- and exclusion criteria were also applied (Clinicaltrials.gov Identifier: NCT01777893).
In total, 15,611 adult participants were pre-screened, 5,472 were screened and 2,326 were eventually in-cluded. The recruitment phase had to be extended by 1 year, mainly since the pre-screening and screening failure rates were much higher than anticipated. To obtain enough participants it was also decided to in-clude the age group 46-54 y, which was not part of the original study plans.

Intervention
Eligible participants started the 8 weeks weight-reduction (LCD) phase with the Cambridge Weight Plan©. All adult participants in the PREVIEW intervention trial were required to achieve a weight loss of ≥ 8 % of in-itial weight to continue to the 34 months weight maintenance phase. The average weight loss success was 79% or 1,857 participants in total, and thereby the sample size derived from the power calculation (n=1,854) was achieved.
Huge effort was invested by all intervention centres during the weight maintenance intervention. Alt-hough the group sessions were less frequent during the weight-maintenance sessions (as planned) com-pared to the low-calorie diet (LCD) weight-reducing period, the challenges to keep the participants moti-vated were evidently much larger. All centres used various methods to support the group sessions, such as Facebook groups and general evening meetings with the participants. Great care was, however, taken not to cross-contaminate the groups with the program of the alternative intervention. Most importantly, the specifically developed behaviour modification tool for PREVIEW (“PREMIT”, please see WP4) was used at all group visits during the intervention to keep the participants motivated to change and maintain a new lifestyle.
For all 4 groups together, the % adult participants who left the trial during the 3 years (including those with diagnosed T2D) at each stage of the study was:
• CID2 to CID3 (months 2 to 6): 19.6%
• CID3 to CID4 (months 6 to 12): 15.2%
• CID4 to CID5 (months 12 to 18): 10.0%
• CID5 to CID6 (months 18 to 24): 13.1%
• CID6 to CID7 (months 24 to 36): 12.0%
The overall cumulative drop-out was larger than anticipated. However, there were no differences in drop-out rate between the four intervention groups. This is a positive result, since intervention-dependent bias related to participant drop-out is therefore unlikely.
For children and adolescents, the study ended at 2 years. The drop-outs were as follows:
• CID2 to CID3 (months 2 to 6): 9.7%
• CID3 to CID4 (months 6 to 12): 8.6%
• CID4 to CID5 (months 12 to 18): 7.1%
• CID5 to CID6 (months 18 to 24): 22.8%

A clear increase in the proportion of drop-outs was seen after month 18, showing how difficult it was to keep the children and adolescents during their teenage years interested and committed to a long-term study.

Data collection and analyses
All deviations from the protocol and SOPs were recorded in a centre-specific diary. Other quality checks in-cluded accelerometer compliance (hours errorless recording daily), dietary recording (discussion with a die-titian, completeness of the recording) and questionnaires (ensuring all questions have been answered).
Data cleaning was performed for all CIDs on an ongoing basis by the data management team in UCPH. If missing variables or obvious outliers were discovered, queries were sent to the relevant intervention cen-tre for corrective action. A statistical analysis plan was developed before data analysis was performed. Breaking the code for the 4 intervention groups was not done until all data were cleaned and analysed. Thus, the researchers performing the statistical analyses knew there were four treatment codes, but not which exact intervention group these refereed to. After the main statistical analyses had been performed, the actual group allocation was revealed.

Dietary intake
The food consumption and nutrient intake data were derived from repeated weighed 4-d food records. Since many food-items were local and their nutrient composition could vary between countries, all dietary records were first transformed into food consumption and nutrient intake including GI, protein (quantity and quality), fibre, whole grain products by using local food composition data bases and software. Any problems in data entry (e.g. incomplete description of food items, unavailability to find certain food items in the food composition data base,etc.) were recorded in a centre-specific diary. The food-consumption and nutrient intakes were then exported to the central datahub. The completeness of data entry and the preliminary quality check (identification of impossible data, etc.) were verified by examining the initial de-scriptive results. The Instructors Network was important for discussing and harmonizing the dietary data, so that the data would be as uniform as possible in the datahub.
Altogether, 5,608 acceptable dietary records from adult participants and 215 from children and adolescents were obtained. The preliminary results for the adults showed a decreased energy intake throughout the study, compared to baseline. Thus, the mean energy intake at baseline was 7.6 MJ/d, whereas it varied be-tween 5.6 and 6.2 MJ/d at months 6, 12, 24, and 36. This decrease was apparent in all macronutrient clas-ses, but not in total fibre intake. This is positive, since fibre intake is a strong indicator of the total dietary quality.
Despite decreased energy intake, the intake of e.g. vegetables and fruits showed a marginal increase, again reflecting healthier dietary habits throughout the study. In the aims of PREVIEW, it was clearly stated that both diets should be compatible with dietary guidelines and healthy eating. The preliminary examina-tion of our data gives strong indication that this has been the case, although we do not have the data anal-yses on a group level at this time point.
The dietary data from children and adolescents was more constant during the study. The intake of energy and macronutrients did not vary very much at different CIDs. However, there was a trend for a decreased dietary GI and glycaemic load, GL, during the intervention.

Lab analyses
The lab analyses were done from blood, urine and faecal specimen. To improve quality, all main analyses were done in a central laboratory (THL) from specimens taken by the local study centres. Urinary nitrogen analyses (an indicator of dietary protein intake and a quality check for dietary compliance) were done local-ly. The completeness of data entry and the preliminary check (identification of impossible data, etc.) were verified by examining the initial descriptive results.
In total, more than 70,000 laboratory analyses were carried out for the RCT. The mean and standard devia-tion of all the analyses were calculated for each CID for descriptive purposes and for quality check. The val-ues that were outside the measurement range were excluded from the calculation. Analyses were per-formed for the main outcomes from the whole intervention group (at THL) as well as for sub-group anal-yses on e.g. kidney safety in elderly participants (analyses done in UCPH), colon cancer risk markers in a sub-group (UCPH), and urine metabolomics for metabolic profile (UCPH). Samples were also taken in single centres for later DNA and RNA analyses.

Results
Results from the LCD period - adults
Of the 2,224 individuals (1,504 women, 720 men) who attended the baseline visit, 2,020 (90.8%) completed the 8 wks LCD period. Following the LCD, mean body weight decreased by 11%. The weight loss was 16% greater in men than in women (11.8% vs 10.3%, respectively), but improvements in insulin resistance were similar. Following the LCD, 694 participants (35.8%) had reverted to normo-glycaemia based on fasting plasma glucose alone. This number increased to 40.2% among participants who met the target weight loss (≥ 8%). HOMA-IR decreased by 1.50 ± 0.15 units in men and by 1.35 ± 0.15 in women (ns). After adjusting for differences in weight loss, men had larger reductions in metabolic syndrome Z-score, C-peptide, fat mass, and heart rate, while women had larger reductions in HDL cholesterol, fat-free mass, hip circumfer-ence and pulse pressure. It was concluded that the 8-week LCD induced less beneficial effects in women than in men. These findings are clinically important and suggest gender-specific changes after weight loss. It is imperative to investigate whether the greater decreases in FFM, hip circumference and HDL choles-terol in women after rapid weight loss compromise weight loss maintenance and future cardiovascular health (Christensen et al Diabetes, Obesity and Metabolism 2018; 2018;1–12. https://doi.org/10.1111/dom.13466).

Results from the 3-y intervention - adults
The statistical analysis of the primary endpoint, incidence of T2D during 3 years according to diet, was a semi-parametric Cox proportional hazards regression model. A similar model was used for incidence of T2D according to PA and according to all 4 intervention groups (diet and PA combined). These models included adjustments for age, BMI, gender, ethnicity, intervention centre, fasting glucose and 2-h glucose. Addi-tionally, the model for comparing the 2 diets included adjustment for PA group, and the model for compar-ing PA groups included adjustment for diet group. Continuous secondary outcomes (e.g. blood chemistry, anthropometrics), data were analyzed using an ANCOVA-type linear mixed model (repeated measure-ments analysis), adjusted for fixed effects (baseline outcome, age, gender) and random effects (interven-tion centre).
In total 2,326 adults were eligible, 2,223 started and 962 completed the 3-y intervention study (43%). The total number of T2D cases after 3 y was substantially lower than predicted, amounting to only 62 cases and the incidence was 4%. There were no differences between the two diets (Hazard ratio, HR, HP vs MP: 1.22 p = 0.45). T2D incidence did also not differ between the two PA regimes (HI vs MI, HR = 1.35 p = 0.27) or the 4 different intervention groups. As mentioned, body weight decreased by 11% after the weight loss phase, and it was still 5% lower than at baseline after 3 years (completers). All outcomes related to glucose or lipid metabolism, anthropometrics, blood pressure and heart rate decreased significantly after the LCD phase. After 3 years BMI, fat mass, fat-free mass, waist, hip and thigh circumference, 2-h glucose, insulin, C-peptide, HOMA-IR, and triglycerides were still significantly lower compared with baseline values. A group difference was seen for C-reactive protein and diastolic blood pressure (p < 0.05) but otherwise there were no differences between the two diet groups, two PA groups, or the four groups combined (p > 0.05).
Compliance for dietary intake and PA (PA) was assessed by 24-h urinary nitrogen collection (to estimate protein intake) and 7-d accelerometer data. Compliance to the 2 diets showed a significant interaction be-tween time and diet (p<0.0001) with a higher protein intake in HP than MP at 6 months and 1 y, but not at 2 y and 3 y. Mean daily protein intake in completers over 3 years was in HP 0.95 ± 0.09 g/kg and in MP 0.84 ± 0.08 g/kg (p<0.001). Regarding compliance to the 2 PA regimes, there were no significant differences be-tween the groups at any time point, for total counts per minute, moderate-to-vigorous PA, vigorous PA, or sedentary time.
In conclusion, the main hypothesis could not be confirmed. Thus, a high protein-low GI diet was not supe-rior to a moderate-protein, moderate-GI diet in relation to prevention of T2D. It can be speculated, that the large initial weight loss followed by an overall healthier lifestyle including a higher protein intake and specific behaviour modification tools, resulted in the present outcomes. Further analyses are needed to verify this.
A draft manuscript on the main results has been prepared and the final version will be submitted to a high-ranking journal in 2019.

Other outcomes from the RCT – adult
Colon cancer risk markers
A total of 94 participants from HEL and UOA were involved in this sub-study. Putative colorectal cancer re-lated markers were identified from 24-h fecal samples collected over 3 consecutive days at baseline and af-ter 1 year. These markers were concentration of total short-chain fatty acids, acetate, propionate and bu-tyrate, phenols, ammonia, and pH. Unadjusted analyses were conducted as well as analyses adjusted for potential confounders including: age, gender, change in weight loss from baseline to one-year follow-up, dietary fiber intake, study site. A positive association was found between change in total protein intake (E%) and change in faecal ammonia concentration. A higher intake of protein from processed meat was as-sociated with a decrease in the concentration of total SCFAs (p = 0.02) and acetate (p = 0.04) and an in-crease in butyrate (p = 0.047) but only in the group with the highest intake of protein. No other associa-tions were found.
In conclusion, there were few indications of adverse changes in putative colorectal cancer related markers after one year of increased protein intake in overweight pre-diabetic adults who had underwent an initial 8-wks weight loss period.

Respiratory Chamber study
Effects of a HP vs MP diet on satiety, energy expenditure and substrate oxidation have been shown be-fore. In this sub-study, associations of hunger and satiety with GLP-1 and PYY concentrations, adaptive thermogenesis, and metabolic flexibility was assessed after a 34 month on the 2 intervention diets. A total of 40 adult participants from UM were included. The results are expected to be published in 2019.

Visceral fat and liver fat
The effects of the RCT on intrahepatic lipid (IHL) content and implications for insulin sensitivity were exam-ined by assessing changes in body fat deposition and liver-fat content in relation to the diet and PA arms in one of the intervention centres (UM). Magnetic Resonance Imaging (MRI) and Proton Magnetic Reso-nance Spectroscopy (HMRS) were used. A total of 25 adults with pre-diabetes (13 on MP, 12 on HP diet) participated. The results showed that a 2 year relatively high- or medium protein diet reduced IHL and vis-ceral adipose tissue and increased insulin sensitivity. The reduction in IHL was inversely related to insulin sensitivity, independent of BMI. These findings stress the clinical implications and potential benefits of in-creased protein intake after weight loss for people with non-alcoholic fatty liver disease at risk of develop-ing diabetes (Drummen et al Am J Physiol Endocrinol Metab 2018; 2018; 315:E885-891).

Kidney safety
Concerns about detrimental renal effects of a high-protein intake have been raised, due to an induced glomerular hyperfiltration, which may accelerate the progression of kidney disease. The aim of this sub-study was to assess the effect of a higher intake of protein on kidney function in pre-diabetic men and women, aged 55 years and older. Analyses were done on baseline and 1 year data in a sub-group of 310 participants. Protein intake was estimated from 4-day dietary records and 24-hour urinary urea excretion. Linear regression was used to assess the association between protein intake after one year of intervention and kidney function markers: creatinine clearance, estimated glomerular filtration rate (eGFR), urinary al-bumin/creatinine ratio (ACR), urinary urea/creatinine ratio (UCR), serum creatinine, and serum urea be-fore and after adjustments for potential confounders. The results showed that a higher protein intake was associated with a significant increase in UCR (p = 0.03) and serum urea (p = 0.05) after one year. There were no associations between increased protein intake and creatinine clearance, eGFR, ACR, or serum creatinine.
In conclusion, no indication of impaired kidney function was seen after one year with a higher protein in-take in pre-diabetic older adults (Møller et al, Nutrients 2018; 10; 54; doi:10.3390/nu10010054).

Metabolomics
The evaluation of urine metabolomics (metabolic profile) was performed in a subgroup using urine speci-mens from baseline and 1 y. Five 5 intervention centres were included (UCPH, HEL, UNOTT, UNAV and UOA) and a total 768 participants provided urine samples. Together with quality control samples and standards, the metabolic profiling in positive and negative ionization mode included more than 3,500 anal-yses. The quality of the data was assured, as no analytical variation was seen after correcting for the usual Liquid chromatography mass spectrometry (LC-MS) batches drifts and the difference in individual urine concentration. Such data handling is standard procedure in LC-MS untargeted metabolomics. At this stage, no major difference was observed between baseline corrected and non-corrected data with regard to the performance of the models. Mean centred normalized data seemed to perform better than Loess correct-ed data (ie normalization by polynomial regression). However, more detailed work should be performed to understand the quality of the features selected as discriminative metabolites. Further analyses are planned, including comparing results from the 4 intervention groups.

Physical activity, sedentary time, and cardiometabolic risk
The associations among PA, sedentary time (ST) (from 7-d accelerometry) and cardiometabolic risk were investigated in the adult population of the RCT. Multiple linear regression revealed that moderate-to-vigorous physical activity (MVPA) was significantly negatively associated with HOMA-IR, waist circumfer-ence, fasting insulin, 2-hour glucose, triglycerides, and C-reactive protein. Sedentary time was positively associated with HOMA-IR WC, fasting insulin, triglycerides, CRP, systolic and diastolic blood pressure. Asso-ciations reported between total PA and all risk factors were comparable or stronger than for MVPA.
To conclude, in adults with prediabetes, objectively measured PA and ST were associated with cardiomet-abolic risk markers. Total PA was at least as strongly associated with cardiometabolic risk markers as MVPA, which may imply that the accumulation of total PA over the day is as important as achieving the intensity of MVPA (Swindell et al, Diabetes Care. 2018;41:562-69).

Main results - children and adolescents
Baseline data
Insulin resistance (IR) in adolescence is associated with T2D. In 126 children and adolescents (mean ± SD age 13.6 ±2.2 years, BMI z-score 3.04 ± 0.66 HOMA-IR 3.48 ± 2.28) anthropometrics, fat mass percentage (FM%), metabolic characteristics, physical activity, food intake and sleep were measured. Baseline charac-teristics did not differ between the groups. IR was higher in pubertal adolescents with morbid obesity than in prepubertal children with morbid obesity (5.41 ± 1.86 vs 3.23 ± 1.86; P = 0.007) and prepubertal and pu-bertal children and adolescents with overweight/obesity (vs 3.61 ± 1.60 P = 0.004 and vs 3.40 ± 1.50 P < 0.001 respectively). IR was associated with sex, Tanner stage, BMI z-score and FM%. Fasting glucose con-centrations were negatively associated with Baecke sport score (r = −0.223 P = 0.025) and positively with daytime sleepiness (r = 0.280 P = 0.016) independent of sex, Tanner stage, BMI z-score and FM%.
In conclusion, IR was most severe in pubertal children/adolescents with morbid obesity. The associations between fasting glucose concentration and Baecke sport score and sleepiness suggest these might be pos-sible targets for diabetes prevention (Dorenbos et al; Diabetes, Obesity and Metabolism 2018;20:1096–1101).

2-y data
A total of 126 adolescents were included, 86 completed 1 y (66%), and 49 participants completed 2 y (39%). The results showed that changes in protein intake were not significantly different between intervention groups at any time point. For both diet groups combined, BMI z-score decreased after 1 y (-0.16 p<0.001) and 2 years (-0.19 p=0.03). Cognitive restraint and moderate-to-vigorous PA increased at 1 and 2 years. At 1 year, change in BMI z-score was inversely associated with change in dietary restraint (p=0.025) and posi-tively associated with change in HOMA-IR (p=0.004). In conclusion, there were no differences between the two diet groups in the main outcome, possibly due to lack of compliance to the diet. Dietary restraint relates to decreasing BMI z-score in adolescents with overweight or obesity, while decreased BMI z-score relates to attenuating pubertal IR (manuscript submitted for publication).

Population studies, PS (WP2)
Compiling a single database from the participating epidemiological studies
The first step was to produce a document describing the construction of the common database from the participating epidemiological studies: The Netherlands (LifeLines: n=152,180 participants and 5-y follow-up measurements; NQPlus: n= 2,048 and 2- follow-up), Finland (Young Finns Study: n=3,600 and 4-5 y follow-up), New Zealand (New Zealand Adult Nutrition Survey: n=3,348 with cross-sectional data only) and Cana-da (The Quebec Family Study: n=952 and 5-y follow-up). The cohort characteristics including study design and population, cohort size, and measurements were discussed and an extensive overview of the assess-ment of all variables of interest was given. Data in PREVIEW have been collected within the framework of independent population studies, with different protocols for data collection and distinct original research foci. Therefore, data harmonization was a major task of PREVIEW. Harmonized variables were created for all parameters of interest for the PREVIEW data-analysis. A list of common variables with uniform defini-tions was constructed according to the following clusters of variables: general (age, sex, socio-economic status), anthropometry (weight, height, waist circumference), diet, PA, other health and lifestyle variables (blood pressure, cholesterol, smoking behaviour, alcohol consumption, stress, and sleep), and diabetes. For each cohort, a uniform dataset was built based upon this list of common variables. From these dataset, we were able to perform cohort-specific data-analyses and random-effect meta-analyses from which pooled-estimates were derived.

Developing a dedicated GI questionnaire
In the first period of the project a draft food frequency questionnaire (FFQ) aimed at assessing GI and GL of the diet was developed. We selected specific food items to include in the FFQ based on literature and upon which food items contributed most to the absolute level and the between-individual variation in GI and GL of the Dutch population. The draft food item list included mainly bread, potatoes, dough, chips, pasta, rice, cake, sugar-sweetened beverages, milk, beer, and fruit. To this end, GI and GL values were sys-tematically assigned to ca. 2,500 foods as present in the Dutch food composition table based upon a six-step methodology using four different databases. This food composition table served as the calculation ba-sis for the GI-FFQ, but also the analysis of GI and GL in the Dutch cohorts of WP2. After preparing the draft GI-questionnaire, the draft Frequency Food Questionnaire (FFQ) was finalized and improved. After pre-paring the final food item list, the number of questions, items and sub-items, and the lay-out were opti-mized within the Dutch FFQ tool. In April 2015, the online FFQ was sent out to ca. 1,600 participants en-rolled in the NQplus study. The relative validity of the questionnaire was examined. The intake of GI and GL from this FFQ was evaluated against multiple 24-hour recalls, a general FFQ and HbA1c levels (a longer-term marker of glycaemic control). Mean intake estimates for total carbohydrates, mono/disaccharides, polysaccharides, bread, cereals, potatoes, pasta, rice, fruit, dairy, cakes/cookies, and sweets were compa-rable in the three dietary assessment methods. With regard to GI, mean estimates were also comparable across methods while mean GI-FFQ GL was slightly lower than the general-FFQ GL and 24h-recalls GL. Clas-sification of GI in quartiles was identical for the GI-FFQ and general-FFQ for 43% of the population and with 24h-recalls for 35% of the population. For GL this was 48% and 44%, respectively. With all methods, no measure, including carbohydrate and dietary fibre, predicted HbA1c
To conclude, the validation results of the GI-FFQ concurred with the results of other studies in this re-search area as well as other nutritional factors commonly studied using FFQ. This supports the use of the novel GI-FFQ to estimate dietary GI and GL and related dietary factors (Brouwer-Brolsma et al. Nutrients 2019, 11, 13; doi:10.3390/ nu11010013. Online 21-Dec-2018).

Assessing differences in prevalence and incidence of diabetes in participants with high-low protein/GI diets, high-low PA, high-low sleep duration and high-low (perceived) stress in participating cohorts
A data-analysis plan was developed to describe the methodological decisions to examine the study ques-tions of WP2. Decisions were made regarding the definition of the cohorts, exposure variables (protein, GI, GL, PA, sleep, and stress), outcome variables (diabetes variables), and the statistical methods.

Protein
The associations between protein as a single factor in relation to diabetes risk was explored in the cohorts of the New Zealand Adult Nutrition Survey, Quebec Family Study, Young Finns Study, NQplus, and Life-lines. At the time of the first analysis, which was cross-sectional, a total of 67,535 participants were availa-ble for inclusion, of which most (91%) were on the Lifelines cohort. Analysed as a single dietary factor, a higher absolute and relative protein intake was cross-sectionally associated with a higher odds of having diabetes in all population studies combined. The association with pre-diabetes was less pronounced. The prospective analyses on protein showed that a higher protein intake expressed in grams per kilogram body weight was associated with a lower risk of pre-diabetes and diabetes. Analysing plant-based and animal-based protein separately in relation to pre-diabetes and diabetes indicated that the association was most pronounced for plant-based protein. Associations were substantially attenuated after adjustments for BMI and waist circumference. This demonstrates the crucial role of adiposity and may account for previous con-flicting findings in the literature. No interactions were observed between protein intake and sex or BMI categories. The stratified analyses by sex and BMI did not identify substantially different associations. Excluding persons with cardiovascular disease, hypertension and/or high cholesterol, while maintaining suffi-cient power, did not alter the findings either (Sluik et al, In press, Am J Clin Nutr 2019).

GI, GL and carbohydrates
Cross-sectional analyses into GI and GL as a single factor in relation to (pre-)diabetes prevalence and inci-dence in the five WP2 cohorts showed mixed results, but a trend towards a positive association was ob-served. In the New Zealand Adult Nutrition Survey and the Quebec Family Study, dietary GI was not signifi-cantly associated with measures of glucose tolerance. In the Young Finns Study, higher dietary GI and GL were associated with lower levels of fasting glucose and lower odds of having pre-diabetes or diabetes. On the other hand, in NQplus and Lifelines higher dietary GI was associated with higher HbA1c levels and high-er odds of having diabetes. Prospectively, meta-analyses of the results of the four prospective cohort studies indicated that GI was modestly associated with pre-diabetes, but not diabetes. These findings were not reflected in the prospective associations for GL or carbohydrate intake in grams per day. Data did, however, suggest an association between higher intake of carbohydrates consumed at the expense of fat or protein and lower incidence of pre-diabetes and diabetes. Although modest, stratification by BMI indi-cated the strongest positive associations between GI and incidence of pre-diabetes (P for interaction 0.06) and diabetes (P for interaction 0.18) among overweight and obese participants. No clear patterns were ob-served for GL and carbohydrates. Stratified analyses for sex pointed towards more pronounced positive associations between GL and pre-diabetes (P for interaction 0.07) and diabetes (P for interaction 0.11) in women. Significant interactions were also observed for GL and carbohydrates with protein intake in the as-sociation with pre-diabetes (P for interaction 0.05 for both metrics). The data and these analyses are de-scribed in a paper, which has been submitted for publication (Brouwer-Brolsma et al.).

PA
Moderately physically active persons had a lower HbA1c level and a lower odds of abnormal glucose toler-ance in most cohorts, compared to less active persons. The associations were attenuated by additional ad-justment for BMI and waist.

Stratification- and interaction analyses
The interactions between the factors GI, GL, and protein, as well as between diet and physical activity in relation to HbA1c/glucose levels and prevalence and incidence of type-2 diabetes in the population studies included in WP2. Cross-sectionally, no convincing signs of interactions were observed between protein and GI, protein and GL, and between protein, GI, and intense PA. However, prospective stratified analyses in-dicated that the highest risks of pre-diabetes were among those in the upper GL or carbohydrate group and the upper protein group. Indications were also found for an interaction between protein intake and moderate intense physical activity and to a lesser extent between GI and moderate intense physical activi-ty (p-interaction<0.05). However, these interactions were not consistent across cohorts and methods (stratified vs. product term). Although we did not find convincing evidence for an interaction between diet and moderate intense PA, we cannot rule it out and inconsistent finding across cohorts underline the im-portance of including diet and PA in all statistical analyses. Future analyses and publications into protein and GI should include stratified analysis for low and high amounts of moderate intensity PA.

Sleep and stress
Across all cohorts, a longer sleep duration was associated with lower HbA1c or fasting glucose levels, which remained after adjustment for important confounders including energy intake, BMI and waist. A longer sleep duration was associated with a lower odds of an abnormal glucose tolerance. When analysed in cate-gories, a sleep duration of 5 hours or less per day was related to higher odds of an abnormal glucose toler-ance compared to 6-8 hours per day. No associations were observed between perceived stress level and diabetes prevalence. Prospective data suggested lower risks of pre-diabetes and diabetes among partici-pants with 8-9 hours of sleep. A paper on these results has been drafted and will be finalized and submit-ted in 2019.

Assessing the impact of protein/GI/PA in the context of other risk factors of diabetes
The aim was to investigate what proportion of the diabetes cases within the population could be attributed to intake of protein, GI/GL and levels of PA. Analyses showed that a higher protein intake in grams per day could not consistently be attributed to a higher or lower proportion of abnormal glucose tolerance (AGT) or diabetes cases in any of the cohorts studied. For protein intake in grams/kilogram body weight, a lower proportion of AGT and diabetes cases was observed, both cross-sectional and prospectively. This finding may indicate that a higher relative protein intake (in gram per kilogram body weight) is more important for AGT or diabetes prevention compared to absolute protein intake (in gram per day), but as concluded in the previous analysis, it could also be the result of the protective effect of lower body weight. In the case of GI and GL neither the cross-sectional, nor prospective analyses, demonstrated that a lower proportion of AGT or diabetes could be attributed to a higher or lower GI or GL intake. For PA examined cross-sectionally, be-ing both more moderately or more intensively physically active could be attributed to a lower proportion of AGT cases in some cohorts studied. However, this could not be confirmed in the prospective analysis. An explanation for these findings could be the lack of variation in exposure variables within the cohorts stud-ied.

Extra outcome
Based on quantity and quality of protein consumed, a diet protein scoring tool was developed as an extra outcome from the project in WP2. The association between the protein score and HbA1c as well as renal function (eGFR) was examined. Analyses were based on the 3 European population studies, NQplus, Life-lines, and The Young Finns Study. A higher protein score was associated with a lower HbA1c and an increase in estimated Glomerular filtration rate (eGFR) in some studies but not all. Further studies are needed to validate this newly developed protein score (Poulsen et al, Nutrients 2017; 9, 763; doi:10.3390/ nu9070763).
Highlights and conclusions of WP2
The first substantial result of WP2 includes the development and administration of the GI-FFQ among ca. 1,600 participants from the NQplus study. The performance of this novel questionnaire was validated against a general FFQ and a 24h-recall and results support the use of this FFQ to correctly rank participants according to their dietary GI, GL and related factors (i.e. total carbohydrates, carbohydrate fractions, and food groups).
Another important component of WP2 comprised the data-analysis of all factors, i.e. protein, GI, PA, sleep, and stress, in relation to measures of glucose tolerance in the five cohorts. However, the results from the five cohorts were not consistent for all determinants under study and outcomes. The multiple association analyses have shown that a higher absolute and relative intake of protein was cross-sectionally associated with higher odds of having diabetes and prospectively with a lower risk, while GI and GL showed mixed re-sults in the different cohorts. Moderate PA and longer sleep were associated with lower odds of having di-abetes. The impact of protein/GI/PA in the context of other risk factors of diabetes was also evaluated. It could not be clearly demonstrated that AGT or diabetes cases could be attributed to protein intake, GI/GL or levels of PA. Findings could indicate that a higher relative protein intake (in gram per kilogram body-weight) is more important for AGT or diabetes prevention compared to absolute protein intake (in gram per day), although this could also be due to a protective effect of lower body weight. An important high-light of the aforementioned analyses is that during the final phase of the PREVIEW project, the prospective data of the Lifelines cohort were released, which allowed us to conduct well-powered prospective anal-yses (full sample as well as stratified for age, sex, BMI and other relevant variables depending on determi-nants under study) with sample size over 70,000 participants. The added value of these data are accentu-ated by the fact that peer-reviewers recommended publication of our data on protein, pre-diabetes and diabetes in the American Journal of Clinical Nutrition, one of the most highly regarded peer-reviewed journals in the field. Moreover, our paper presenting the prospective data of GI and GL in relation to pre-diabetes and diabetes has just been submitted for publication; and during this final project period the pa-per on the prospective data on sleep in relation to pre-diabetes and diabetes has been drafted.
In conclusion, results were inconsistent across cohorts. Nevertheless, the findings still suggest that GI, pro-tein, PA and sleep may have a positive impact in the prevention of diabetes and related conditions. Alt-hough a very systematic harmonization of the data was carried out to ensure comparability, the use of dif-ferent methods for collection and calculation of variables, specifically for GI, could explain some of the var-iability in the results, in addition to low variability in some exposures. Further investigation in other popula-tion studies as well as intervention studies will be necessary to produce more convincing evidence of the relative importance of GI, protein, PA and sleep. A further potentially valuable outcome is the newly de-veloped GI-FFQ and the protein score which could be useful tools to apply in future studies.

The role of sleep and stress in interaction with the role of diet and PA (WP3)
Associations from the complete RCT population showed that control of food intake, sustained PA and pre-vention of sleepiness, stress, mood disturbance, disinhibition of eating and hunger were relevant for pre-vention of T2D. In a representative sub-cohort (n=40), assessment of the role of sleep architecture and components thereof in intervention related changes showed that sleep duration appeared to be inversely associated with BMI z-score in children and adolescents with overweight and obesity, independent of pu-bertal stage, while HOMA-IR was positively associated with BMI z-score. Research on possible changes in brain responsiveness of food reward-related brain areas to visual food cues with fMRI, related to changes in insulin sensitivity following the intervention in a representative adult sub-cohort (n=40), showed that a higher intervention induced protein intake, and more controlled eating during weight maintenance was as-sociated with less brain activation in food-reward related areas, implying greater control of food intake. Fi-nally, results obtained in order to identify the relationship between sleep duration and protein intake, PA, body-composition, and insulin sensitivity in the population based Quebec family study (QFS) showed no difference in sleep duration between individuals differing in their HOMA-IR profile in the QFS. The insulin response to glucose was highly correlated with body composition in men and in women in the QFS, after adjustment for BMI, protein intake, PA participation and sleep duration. Protein intake, PA participation, and sleep duration did not contribute to the insulin response to glucose (HOMA-IR).
The role of sleep architecture and components thereof in intervention induced changes in insulin sensitivi-ty was assessed in a sub-cohort of 40 children and adolescents in the RCT.
As the prevalence of childhood obesity and obesity-related comorbidities increases, more research is per-formed to identify possible targets for obesity prevention and therapy. Lifestyle interventions, mainly fo-cussing on increasing PA and control of food intake, are currently the cornerstone of prevention and treatment of overweight and obesity-related comorbidities. However, more evidence is emerging that sleep may be a third modifiable contributor to energy balance, and consequently, to obesity and related comorbidities. Identifying the relationship of sleep with obesity status might aid in optimizing treatment strategies and prevent development of morbidities for children with overweight and obesity, especially when during puberty weight gain is a considerable risk, and sleep duration declines.
A growing body of evidence identified inadequate sleep duration and quality as an independent risk factor for weight gain in children, even after correcting for contributing factors such as body mass index at the start of puberty. In addition, decreased sleep duration and changes in sleep architecture, especially de-crease of slow wave sleep (SWS), seem to be related to increased insulin resistance, hypertension, dyslipi-daemia, and inflammatory factors. Although the exact mechanism linking inadequate sleep with obesity and obesity-related comorbidities is not yet known, the most important mechanisms appear to be related to endocrine stress regulation. Sleep is a refractory period for stress hormones such as cortisol, norepi-nephrine and epinephrine. Loss of sleep or decrease of sleep quality may lead to increased endocrine stress. Experimental studies in children found that both increased cortisol concentrations and sympathetic nervous system activity were associated with unfavourable changes in glucose metabolism. In addition, short sleep duration is associated with higher levels of the orexigenic hormone ghrelin and lower concen-trations of the anorexigenic hormone leptin, which promotes hunger and food intake. One experimental study found increased food intake after sleep restriction, and several observational studies reported in-creased intake of specifically high-energy and sugar-rich foods. Reduced sleep and subsequent daytime tiredness may also contribute to decreased PA and exercise in general.
Although the majority of studies point to an inverse relation between sleep duration and the development of cardiometabolic risk, including obesity, evidence in children is limited and often conflicting. Few studies have researched the relationship between sleep and cardiometabolic risk in children in longitudinal de-signs. Moreover, studies were performed with different methods of sleep assessment e.g. polysomnog-raphy, accelerometry, self-reporting with questionnaires and/or parental sleep assessment, which hinders comparison of studies. Lastly, many studies did not correct for variables like pubertal stage, sex and obesity status, all of which are known to be related to sleep and cardiometabolic risk factors.
The children and adolescents part of the RCT aimed to assess the effect of a lifestyle intervention on BMI z-score and insulin resistance in children and adolescents with overweight/obesity. The aim of the study on sleep was to identify possible associations between sleep duration and sleep architecture with anthropo-metric characteristics, parameters of glucose metabolism, cardiovascular risk, and inflammation in prior to and after one year of the PREVIEW intervention. In addition, the study aimed to assess the association be-tween objective and subjective sleep assessment outcomes. It was hypothesized that sleep duration was negatively associated with BMI z-score and HOMA-IR, and that change in sleep duration would be inversely related to change in BMI z-score after one year.
In a representative sub-cohort of 67 children and adolescents, sleep was objectively measured at baseline, during an overnight stay at the paediatric intensive care unit of Maastricht UMC, using polysomnography. Analyses with BrainRT (v2.1 OSG, Rumst, Belgium) yielded durations of Total Sleeping Time (TST), Wake After Sleep Onset (WASO), and the different sleep stages Rapid Eye Movement (REM) sleep, and non-REM sleep phases N1, N2 and N3 (also known as Slow Wave Sleep (SWS)). In addition, Quality Sleep (QS, calculated as (REM+SWS)/TST) was determined. Moreover, habitual TST was measured during 4 consecu-tive nights at home with the Actisleep GT3X (Actigraph Corp., Pensacola, FL, USA), applying a fully auto-mated algorithm developed for use in 24-h waist worn accelerometer protocols. Subjectively experienced sleep was assessed using the Pittsburgh Sleep Quality Index (PSQI), where higher scores indicate poorer sleep quality and which for clarity will be formulated as PSQI (poor) sleep quality in this summary.
Results on the assessment of the relationship between sleep parameters with anthropometric characteris-tics and cardiometabolic parameters and changes herein over time showed that after correcting for puber-tal stage and sex, habitual sleep duration was negatively related to BMI z-score. No other sleep parame-ters were associated with anthropometry or cardiometabolic risk. PSQI (poor) sleep quality scores were in-versely associated with PSG-measured Quality Sleep, habitual TST, and positively with phase N2 sleep and WASO (wake-up after sleep onset). Accelerometry and polysomnography measured sleep durations were not associated.
At baseline the inverse relation between habitual TST and BMI z-score, corrected for Tanner stage and sex, indicated that children and adolescents with a higher BMI z-score had shorter sleep duration, independent of pubertal stage. This observation confirms observations from earlier studies in children, adolescents and adults of all weight classes. Sleep architecture parameters were not related to the severity of overweight and body composition.
A total of 29 out of the 67 children and adolescents, who were recruited to the sleep study, participated in a second polysomnography 1 year later. Both objectively measured and self-reported sleep parameters did not had not changed significantly. It can be speculated that the reduction in BMI z-score may have counteracted the previously observed age and Tanner stage related reduction in sleep duration.
While polysomnography is the gold standard for assessment of sleep architecture, accelerometry allows assessment of daily sleep during multiple nights, and sleep questionnaires assess subjectively experienced sleep quality. In this study both objective and subjective sleep measurement methods were combined to assess a full range of objectively measured and experienced sleep parameters. It appeared that self-reported sleep duration was positively associated with accelerometer-measured TST, while higher self-assessed sleep quality was related to higher percentages of SWS and REM sleep. Similarly, higher WASO was positively related with higher self-reported daytime dysfunction scores. These relationships suggest that PSQI scores are indicative for polysomnography outcomes. On the other hand accelerometer-measured TST and polysomnography-measured TST were not interrelated, possibly explained by meas-urements during different nights: accelerometry-TST was measured during 4 consecutive nights at home while polysomnography-TST was measured during one night at an in-hospital setting. The use of both field and self-assessed sleep measurement methods may complement each other when combined.
In conclusion, the inverse association of sleep duration with BMI z-score in children and adolescents with overweight/obesity confirms earlier observations of sleep duration with obesity. Previously reported rela-tionships with insulin resistance, cardiometabolic risk or inflammation were not confirmed in children and adolescents with overweight/obesity. PSQI (poor) sleep quality scores were indicative of polysomnogra-phy outcomes, while accelerometry and polysomnography measured sleep duration were not related to each other. More studies are needed to assess changes in sleep duration and architecture parameters, and the effects of sleep hygiene interventions, on cardiometabolic health in children and adolescents without sleep syndromes. We would recommend future sleep studies to include both objective as well as self-reported sleep measurement methods.
Possible changes in brain responsiveness of food reward-related brain areas to visual food cues were as-sessed with fMRI, and related to changes in insulin sensitivity following the intervention in an adult sub-cohort.
Insulin resistance is one of the most important factors linking obesity with chronic disease risk, including T2D and cardiovascular disease. Although consequences of increased body fat and systemic insulin re-sistance have been investigated for the periphery, consideration of the brain as an insulin-sensitive tissue is comparatively recent. Both obesity and T2D were shown to be associated with altered brain signalling in areas relevant for food-intake motivation and food reward. However, due to the frequent co-occurrence of both, it is difficult to discern whether their contribution to alterations in neural, cue-initiated responses is individual or synergistic in nature. Not only does insulin relay information about peripheral energy status for homeostatic control to the hypothalamus, insulin receptors also densely populate brain regions associ-ated with motivated behaviours and reward processing. Furthermore, insulin is thought to play a role in brain reward regulation through its pivotal role in dopamine release and reuptake. Uninhibited brain re-ward received from food intake may be one of the mechanisms underlying continued overeating in obese individuals. In healthy and lean participants, studies that used intranasal insulin administration showed an attenuation of brain activation in response to food pictures and reduced food intake. Two studies in chil-dren with overweight showed brain activation in response to food images to be inversely associated with either insulin sensitivity or waist circumference. Those results corroborated studies suggesting the co-occurrence of peripheral and central insulin resistance, an idea further supported by studies that used magneto-encephalography. In addition, behavioural factors were shown to play a role in the modulation of brain reward signalling in people with obesity or T2D. Eating behaviour and PA affect food reward, yet it is not clear whether their impact is direct or rather indirect through their relation with either obesity or insu-lin resistance. In order to identify primary intervention targets, it is vital to discern the relative contribution of weight, insulin, and behavioural factors to the observed alterations in brain reward signalling, especially for people at risk of T2D. Almost all of the insulin in the brain is derived from the periphery, posing periph-eral insulin resistance as a particularly interesting target to tackle the mechanisms underlying altered brain responses in individuals at risk.
The aim here was to investigate the respective association of insulin resistance, weight status, and behav-ioural factors with brain reward activation in response to visual food cues on a whole-brain level and specif-ically for the nucleus accumbens, a central region of the dopaminergic reward system at baseline. Given the evidence so far, the hypothesis was that brain reward signalling would be primarily associated with in-sulin resistance rather than weight status. Thus, the first study investigated the respective association of body weight, insulin sensitivity, and behavioural factors with brain activation contrasting food images with nonfood images presented to overweight and obese individuals with IFG, IGT, or both, while undergoing functional Magnetic Imaging (fMRI), at baseline. Food reward–related brain activation was positively asso-ciated with insulin resistance during the anticipation of food reward, specifically in the nucleus accumbens, left and right insula, and right cingulate gyrus, independently of BMI. Brain activation in these regions is likely connected through the dopaminergic pathway originating in the ventral tegmental area. Generally, but especially in people with IFG or IGT, disturbance of food reward processing may be an important accel-erator for the transition to T2D through overeating. The catabolic action of insulin in the brain is an im-portant signal to the homeostatic circuit, but is also relevant to the termination of reward experience from pleasurable behaviours, including food intake. Taking the brakes off reward perception may consequen-tially support hyperphagia. Furthermore, a positive association between food reward–related brain activa-tion and emotional eating and disinhibition in the right supramarginal gyrus was found, which has been previously associated with reward, motivation, and drug abuse.
In conclusion, in participants with overweight or obesity and limited insulin sensitivity, insulin resistance was positively associated with brain reactivity to food cues in the nucleus accumbens, insula, and cingulate gyrus and emotional eating and disinhibition were positively associated with brain reactivity to food cues in the supramarginal gyrus, which may lead to changes in food preference and hyperphagia. The results are especially relevant for individuals with increased risk of the development of T2D. The lifestyle factor of PA was inversely associated with food reward–related brain signalling, posing as a potentially protective factor to counteract the effects of insulin resistance on brain reactivity to food cues. In addition to the well-known peripheral benefits of PA, the results corroborate an additional route for PA in the prevention of T2D and obesity (Drummen et al, Am J Clin Nutr. 2018, Dec 26. doi: 10.1093/ajcn/nqy2521).
Given that insulin binds to receptors expressed on neurons in areas of reward processing, central insulin resistance may contribute to impaired reward signalling and a reduction in insulin resistance may help to normalize the reward response. One of the most potent strategies to reduce insulin resistance is weight loss. It has been shown that food-cue reactivity of reward regions was suppressed along with extended ca-loric restriction. Patients with T2D who underwent bariatric surgery showed decreased reward activation along with a normalization in glycaemic control, compared to T2D patients without surgery and impaired glycaemic control. Besides surgery, lifestyle intervention is a tool to reduce body weight and increase insu-lin sensitivity. The intervention requires profound changes in eating behaviour but it remains unclear how changes in aspects of eating behaviour, such as cognitive restraint, relate to changes in neuronal activation. Considering the well-established effect of protein on satiety, diet-induced thermogenesis and the preser-vation of fat-free mass during weight loss, it seems to be the fundamental macronutrient in the nutritional component of a lifestyle intervention. While largely unknown, the beneficial effects of protein for weight loss and weight maintenance may in part be mediated through an effect on reward signalling.
The aim of the second study was to investigate the effects of weight loss and subsequent weight mainte-nance with higher protein intake versus moderate protein intake on brain reward activity in response to visual food cues in participants of the RCT. It was hypothesized that participants with higher protein intake had reduced brain reactivity to food cues after 2 years compared to participants with moderate protein in-take. However, results showed that in practice, no differences in protein intake nor in brain reactivity to food cues occurred between the groups. Therefore, the whole group was taken together, to assess ef-fects of protein intake on brain reactivity to food cues. Protein intake during weight maintenance was neg-atively related to changes in brain activation contrasting high-calorie with low-calorie images in regions as-sociated with cognitive control and visual processing. Changes in BMI and body fat percentage during weight maintenance were positively related to changes in brain reactivity to high-calorie versus low-calorie images, especially in brain regions associated with reward and dietary self-control. The healthier PREVIEW lifestyle, implying increases in dietary restraint, protein intake, and PA resulting in reduced insulin re-sistance and reduced body weight and fat mass, may have contributed to these results. Reduction in brain reactivity implying a greater control of food intake was shown, as well as an increase of cognitive dietary restraint and a higher protein intake, resulting in reduction of insulin resistance, body weight and fat mass. The relationships between dietary restraint, protein intake and brain response to food images could be possible mechanisms via which healthier lifestyles lead to more control of food intake, and thereby leading to beneficial body weight management and insulin sensitivity.
In conclusion, the relationship between diminished brain reactivity in food-reward related brain areas, im-plying more control of food intake, and increases in dietary restraint, protein intake, and insulin sensitivity may be mechanisms that play a role in a lifestyle approach of body weight management (Published in: Drummen et al, Nutrients. 2018 Nov 15;10(11).
Finally, the relationship between sleep duration and protein intake, PA, body-composition, and insulin sensitivity in the population based Quebec family study (QFS) was assessed.
Results of assessment of the relationship between sleep duration and protein intake, PA, body-composition, and insulin sensitivity in the population based Quebec family study (QFS) show no difference in sleep duration between individuals differing in their HOMA-IR profile in the QFS. The insulin response to glucose was highly correlated with body composition in men and in women in the QFS, after adjustment for BMI, protein intake, PA participation and sleep duration. Protein intake, PA participation, and sleep du-ration did not contribute to the differences in insulin response to glucose (HOMA-IR).
Other lifestyle variables: Behavioural, sociological, environmental, cultural, socio-ecological, and socioeconomic components (WP4)
Individuals’ behaviour is regularly guided by routines and habits, and is thus often unconscious and unin-tended. Modifying established behavioural routines and habits is challenging and can fail. Interventions based on theoretical knowledge are an important basis for making behaviour change more likely. The PREMIT (PREVIEW Behaviour Modification Toolbox) manual was developed to support the PREVIEW RCT by bringing together current evidence and theoretical knowledge of behaviour modification (Kaehlert et al, Front Psychol 2016;7:1136). The focus of the PREMIT Toolbox was on supporting PA and healthy diet changes expected of the PREVIEW participants. PREMIT incorporated a logic principle so that the determi-nants of behaviour change were mapped to the most effective techniques and “tools” to change these de-terminants. A manual was written for the PREVIEW counsellors in order to give them a structure as well as the psychological content for of group sessions. In addition, consistent application of the PREMIT Toolbox, manual, and learning materials at each intervention centre was ensured through workshops. During the workshops, counsellors were taught about the application of different behaviour modification techniques that were to be used during the group visits with the PREVIEW participants. Further, the counsellors were trained to train other counsellors on the techniques (train the trainers). Workshops were held locally with EU partners in Stuttgart and via video conference with the overseas partners (AU and NZ). In addition, reli-able questionnaires were identified to measure “Other lifestyle variables” and associations with behaviour change variables in the PREVIEW-study.

Socio-cognitive factors - baseline results
Baseline data analyses of the psycho-socio-environmental variables of adult PREVIEW-participants de-scribed demographic, socio-economic/ecological variables, and the secondary outcome variables for PA, quality of life and workability. Second, analyses of variances (ANOVAs) were used to analyse between-country differences of participants with regard to the secondary outcome variables. Results showed that participants were mainly highly educated with an average household income, for instance between 27,560 £ - 32,5799 £ per year for the UK and 22,800 € - 27,600 € per year for Spain. Analyses of the psycho-socio-environmental variables for behaviour change indicated that participants were well prepared to take part in the intervention. Analysing the secondary outcome variables, differences between participants from different countries were found for PA, weight and quality of life, and controlled for in the subsequent re-gression analyses. Regression analyses showed significant associations between PA/ weight/ quality of life and personal, social and environmental variables indicating the importance of such variables for behaviour change.

Socio-cognitive factors - Results after LCD phase
Participants losing at least eight percent (8%) of their body weight during the Low-Calorie Diet (LCD) were included in the analyses completed with the data after the LCD phase. First, descriptive statistics were used to describe the data regarding attendance to the supporting behaviour change intervention, de-mographics, and socio-economic and environmental variables. Second, analyses of variances were used to explore psycho-socio-environmental and cross-country differences between participants. Emphasis was also placed on further exploration of the results highlighted in the previous analyses, especially associa-tions between age and weight loss and educational achievement and weight loss.
Results indicated that younger age was associated with greater weight change, as was male gender. The behaviour change intervention (PREMIT) between the visits was well attended and more frequent attend-ance appeared to be associated with higher likelihood of achieving the weight loss target during the LCD phase. During the LCD phase participants reported lessened habits for inactivity and unhealthy eating. Re-flecting on the preparation for the next study phase, participants reported increased intention to exercise and eat healthy, with a significant interaction between intervention centres and intention. Although higher educational achievement was reported to be associated with lower weight in the baseline data, this asso-ciation was not found after the LCD phase. Whilst a significant difference was found overall between edu-cational achievement and weight, there was no significant difference between those with lower and high-er educational achievement. A significant difference was found between those reporting “University level” education and “Other”. Participants also tended to feel positive about their capabilities to maintain their preferred body weight in a year’s time (Hansen et al, Int J Behav Med 2018;25:682-92).
Concerning the socio-economic and environmental variables for behaviour change, participants appeared well prepared to take part in the next intervention phase. Participants had already success in losing weight and felt positive about their capabilities to achieve a long-term behaviour change (Huttunen-Lenz et al, Psychol Res Behav Manag 2018;11:383-94).

Socio-cognitive factors – Results after 1 year, Part 1
Selection of the social-cognitive variables for analysis in this part was both theory-driven and based on the previous results. Descriptive statistics were used to describe the data regarding attendance at the PREMIT sessions, demographic characteristics of the participants, and social-cognitive variables of self-efficacy, mo-tivation, and outcome expectancies. Analyses of variance were used to explore differences, especially in social-cognitive variables within and between participants. Analyses were also conducted to examine longi-tudinal changes in social-cognitive variables and whether attendance at group counselling sessions has in-fluence on social-cognitive variables.
Analyses explored characteristics between those who dropped out and those who continued in the trial. Results suggested that younger participants were at higher risk of dropping out of the trial between the visits than older participants. There were also indications that younger participants were less likely to at-tend the group counselling sessions, thus missing out on the available support. Those with higher BMI were also more likely to drop out between the visits. This means that younger participants with higher BMI were especially at risk of not completing the trial.
Overall, the results showed that the participants had high levels of motivation to make lifestyle changes and described high levels of both self-efficacy and coping self-efficacy. Importantly, participants expected that outcomes of diet and PA changes would be positive. Analyses of the social-cognitive variables identi-fied a decreasing trend over time, i.e. participants reported lower values for socio-cognitive variables (eg motivation) on average as the study progressed. This might be associated with changes in habits and re-duced need for self-regulation and self-efficacy to maintain behaviour. This aspect, however, was not ex-amined here. Finally, the results highlighted greater attrition of younger participants in the trial. While rea-sons for this are likely to be multifactorial, it highlights a challenge for future interventions, i.e. how young-er participants can be motivated to persist with preventive measures.

Socio-cognitive factors – Results after 1 year, Part 2
In this part descriptive statistics were used to describe the data regarding demographic characteristics of the participants, and social-cognitive variables of causal attributions and self-regulation of goal orientation. In particular, analyses of self-regulation of goal adjustment were exploratory, as there is limited knowledge about how the process of goal adjustment as a part of a long-term behaviour modification in-tervention. Analysis of variance was used to explore differences in causal attributes and self-regulation of goal adjustment within and between participants. Analyses were also conducted to examine longitudinal changes and whether attendance at the PREMIT behaviour modification intervention had influence on causal attributions and self-regulation of goal adjustment.
Analyses were largely exploratory, but suggested that participants’ explanations for their current weight remained relatively stable. Participants tended to attribute their weight to internal rather than external causes, which is considered conducive to behaviour change. Participants considered PA and food prefer-ences as attributes to body weight that could be changed and controlled by an individual’s own actions. At-trition in the RCT was not associated with body weight attributions. However, data suggested that attend-ance at the PREMIT behaviour change intervention supported participants especially in understanding that they have control over their food and PA behaviours, thus supporting maintenance of new PA and diet behaviours. Self-regulation of goal adjustment appeared not to be directly influenced by the PREMIT be-haviour modification intervention attendance. Instead, the analyses indicated that participants appeared to protect their wellbeing by engaging in the self-regulation of goal adjustment when unsure about their ability to reach the different intervention goals. Thus, the PREMIT intervention appears to have indirectly influenced self-regulation of goal adjustment by supporting successful behaviour change and consequent-ly successful goal achievement.

Population studies
This part relates to analysis of the effect of “other life style factors” in the PS. The association between the following “other lifestyle factors” and diabetes was assessed: age, sex, and educational attainment. More-over, the associations between the following questionnaires and general and lifestyle factors, dietary in-take, and HbA1c were assessed: the Dutch Eating Behaviour Questionnaire (DEBQ), Consideration of Future Consequences (CSFC), and the Brief Scale of Self Control (BSCS).
The DEBQ assesses three factors related to eating behaviour, i.e. restrained eating, emotional eating, and external eating. The CFCS assesses to which extent individuals consider the potential distant outcomes of their current behaviours and the extent to which individuals are influenced by these potential outcomes. The BSCS assesses five dimensions of self-control: a general capacity for self-discipline, an inclination to-wards deliberate or non-impulsive action, a range of healthy habits, self-regulation in service of a work ethic, and reliability.
A higher age was associated with a higher level of HbA1c and fasting glucose and higher odds of having an abnormal glucose tolerance. Women tended to have lower odds of having abnormal glucose tolerance than men, but these associations were not significant in all cohorts. Furthermore, higher education was as-sociated with lower odds of abnormal glucose tolerance.
No associations were observed between restrained, emotional, and external eating, the consideration of distant outcomes of their own behaviour, as well as self-control and HbA1c levels. For the psycho-social and behavioural factors, most cross-sectional associations with nutrient and food group intakes were observed for the restrained, emotional, and external eating scores. Individuals with a higher restrained eating score had an overall healthier diet, including a lower GI and GL. Those with a higher emotional eating score had a lower intake of protein, GI, GL, potatoes, cereals, dairy, and meat and a higher intake of fat, nuts and seeds and chocolate. Those with a higher external eating score had an unhealthier dietary pattern, includ-ing a higher GL. For the CFCS and BSCS, no strong associations with dietary intake were observed.
In conclusion, a higher age, male sex and a lower education were associated with higher odds of having abnormal glucose tolerance. With respect to psycho-social behaviour, the most interesting associations with dietary intake were observed for restrained, emotional, and external eating behaviour as assessed with the DEBQ. Especially the findings with respect to protein, GI, and GL were found worthwhile to fur-ther investigate within the project.

Cost-effectiveness of the RCT
Cost-effectiveness is a crucial consideration in intervention planning. Here, the cost-effectiveness of the RCT for adult participants is described. Quality of Life (QoL) was measured during the RCT at baseline, after 1, 2, and 3 years. The costs of the LCD, MP and HP diets were investigated for the UK situation, as exten-sive dietary cost data were available there. Additionally, the impact of the diets on QoL and predicted dia-betes risk as assessed by the Finnish diabetes risk score (FINDRISC) were evaluated. Subsequently, the cost-effectiveness ratios for each diet was analysed.
Costs per diet were £ 43.18 for an average weekly UK diet, £ 50.40 /week for the LCD diet, £ 31.98 for the MP diet and £ 42.62 for the HP diet. At baseline, QoL scores in all domains ranged from 62.0 to 72.3. After adjustments for sex, age, intervention centre and baseline QoL levels, QoL scores ranged from 68.1 to 75.5 after 12 months and 66.0 to 76.1 after 36 months. Although differences were small, mean QoL was signifi-cantly higher for the HP diet compared to the MP diet in the physical and psychological domains (+1.6 p=0.03 and +1.7 p=0.01 respectively). In the overall population, the chance of having a high diabetes risk at follow-up compared to baseline was lower. However, when comparing the chances between the HP and MP diet there was no significant difference in diabetes risk (p=0.8).
Finally, compared to a null situation (no intervention), both diets showed a lower cost-effectiveness ratio (CER) expressed in UK £ as ratio of 3-year health benefit in terms of QoL or FINDRISC, namely £ 31.05 /3 y-QoL benefit and £ 27.8 /3 y-FINDRISC reduction, respectively for the MP diet, and £ 40.59 /3 y-QoL and £ 35.2 /3 y-FINDRISC risk reduction in the HP diet vs. £ 43.18 for an average UK diet.
Taking costs into account, the moderate protein diet showed better cost-effectiveness. However, the high protein diet was slightly more favourable than the medium protein diet within the PREVIEW intervention to maintain physical and psychological quality of life after a weight loss phase. As the HP is similar in costs to an average diet in the UK, the HP diet could also be advised as a means to improve health outcomes with no additional expense compared to a nationally average diet in 2018.

Health Impact Assessment of the RCT
The purpose of the Health Impact Assessment (HIA) was to aid policy development. The results indicated that participation in the PREVIEW intervention had a favourable impact on health, especially if a participant was successful in losing weight and maintaining weight-loss. Potential health impacts include reduced risk of developing T2D, cardiovascular and musculoskeletal diseases. However, the HIA suggested that the benefit of the PREVIEW intervention is unlikely to be similar among different socio-demographic groups, and policy action is needed to ensure not only equal access, but also to promote continued participation amongst under-represented groups. Poorer participation was associated with lower educational achieve-ment and male gender. Of those who started the intervention, younger participants, those of non-Caucasian ethnicity, with lower educational background, higher BMI, not married or in a civil partnership, and those with less social support were at higher risk of attrition and worse health outcomes.

Dissemination and exploitation (WP5)
A project logo and livery were developed and utilised throughout the project for all project communica-tions, including the project website, flyers, brochures, newsletters, reports and stationery. The simple de-sign portrays a beta-cell (the blue circle) secreting three molecules of insulin. A beta-cell was considered appropriate because of its key role in the development of T2D, through the mechanisms of insulin re-sistance, hyperinsulinemia and eventual beta-cell dysfunction.
A project website was developed with interactive features for promoting public and health professional engagement and a password-protected member section for storing and sharing all project documents. We chose the NING platform for the website in 2013 and it has fulfilled its purposes very well. All the main documents generated by the Consortium over the 6 year life of the study can be accessed with ease. To date (14 Jan 2019), according to Google Analytics, there have been 11,555 users of the PREVIEW website for a total of 18,567 sessions and 30,099 page views. The average session duration has been 1.28 minutes. Most of the users came from the United Kingdom (17.5%), followed by United States (12.7%), France (9.1%), Australia (7.2%), and Denmark (6.2%).
Our dissemination activities (both lay and scientific) at various stages in the life of the PREVIEW project have been compiled in detailed tables showing activities in chronological order, including scientific publica-tions, conference presentations and radio, TV, newspapers and magazine interviews and official newsletters. From the start to the end of the project period (6 years), at least 396 dissemination activities were specifically related to the PREVIEW project. Conference activities such as oral and poster presentations, abstracts and keynote speeches accounted for > 30% of the dissemination activities, suggesting that the members of the consortium have been successful in promoting the PREVIEW project to scientists and healthcare professionals, despite the fact the main results from the large RCT were only available at the end of the project (Oct-Dec 2018). To date, there have been 16 publications in different scientific journals with another 2 currently in press or under revision.
Considerable work was also done amongst the general public to promote the knowledge gathered as part of the PREVIEW project, including more than 31 TV and radio interviews from the consortium members. Many more are likely when the main paper is published in 2019. We distributed 11 official PREVIEW news-letters (on average 2 per year) to the public and health professionals who had registered their interest on the PREVIEW website (~600 addresses). Special attention was paid in ensuring that the PREVIEW partici-pants in WP1 (the large intervention study) received regular information including newsletters and other communications from the research teams.

Intellectual property rights
New knowledge generated in PREVIEW has been organized for protection and disclosure, taking into ac-count existing prior art and known current R&D activities outside PREVIEW. We have successfully created an environment that encourages and expedites the dissemination of discoveries, creations and new knowledge generated by PREVIEW for the greatest public benefit. Our IPR strategy involves:
• The policy document for IPR in the project, reflecting the rights and commitments in the EC.
• The Consortium Agreement reflecting the EC’s Guidelines on IPR in FP7 projects.
• The intention to give training where appropriate, so that all partners are aware of basic IPR principles and processes.
• The establishment of several processes for both good housekeeping on IPR-related matters, and a pro-active approach to identifying exploitable outputs and pursuing patenting opportunities.
The IPR-related matters include scientific works; electronic learning modules, video performances, hand-books, cooking recipes, inventions in all fields of scientific endeavour, trademarks, services marks, com-mercial names and designations. In the case of jointly owned results, the consortium intends to reach an agreement for the effective management with details, for example, on shares, exploitation and licensing to third parties.

Policy Brief
The policy brief is aimed at policy makers in particular and was prepared in conjunction with health policy specialists. It comprises a set of recommendations developed from the early findings and health impact as-sessment of the RCT. Recommendations are intended to be adapted into current practice, consistent with nutritional science, based on the strength of evidence required to change practice, considering consistent recent research findings and good clinical sense.
To reduce the incidence of T2D, the major recommendations based on the PREVIEW RCT are:
• That governments and health care providers enable and support achievement of substantial (and rap-id) weight loss with proven, safe methods (e.g. total meal replacements) in high risk, overweight and obese people.
• Consistent with PREVIEW experience, that governments and health care providers enable, promote and subsidise the high level of personal and group support required for successful lifestyle and behav-iour change.
• In keeping with this goal, governments, community institutions and health providers enable and sup-port exercise programs which are appropriate, accessible and safe, to increase activity levels and pro-mote health.
Additional recommendations inferred from PREVIEW and relevant to policy were also developed, including government messages around a healthy body weight, weight loss programs, personal support to achieve behaviour change, whole-population initiatives, co-ordination of multi-disciplinary efforts and compensat-ing for social and economic disadvantage. It is intended that Consortium partners will modify the recom-mendations in the brief, incorporating nation-specific issues in major languages.

E-learning module
Two E-learning modules were developed by the Consortium. The first was developed in Year 1 of the pro-ject when findings were still to be realized. As President of the Glycemic Index Foundation, WP5 leader Jennie Brand-Miller supported the development of an attractive IT-supported digital learning module for Australian consumers in conjunction with the Commonwealth Scientific and Industrial Research Organisa-tion (CSIRO) Australia. This learning module was aligned to the outcomes of the DiOGenes Study (an EU FP6 project) which formed the primary rationale for undertaking the PREVIEW project. The intellectual property rights inherent in this e-learning module belong to the GI Foundation and CSIRO. This module has been evaluated for efficacy in achieving behaviour change by GIF and CSIRO in parallel to the PREVIEW pro-ject.
The second interactive digital learning module was developed in Years 4 and 5 by USTUTT with the support of UCPH-, HEL-, WU-, UM-, SU- and UNSYD-partners. The main aim is to show how, through a healthy life-style, individuals in risk of T2D can reduce their risk and/or prevent the development of T2D. It is accessed through the website home page (under the E learning button) and designed to be attractive, easy to ac-cess, and interactive for consumers to learn about diabetes prevention. While the main target group is overweight pre-diabetic people, the learning module is designed to be relevant for the general public and those concerned about their weight or developing diabetes. Further topics relevant for the prevention of T2D may be added later as new evidence emerges from the PREVIEW study. The learning module is not in-tended for those with medical conditions requiring a special diet such as those with metabolic disorders, celiac disease, and Crohn’s disease or for those with T2D.

Student and post-doctoral exchanges
Several student and post-doctoral exchanges have taken place under the PREVIEW project. As an integral part of dissemination, the aim has been to foster knowledge transfer and research collaboration via direct face-to-face contact, including laboratory workshops, summer schools and other activities. In total, 50 exchanges have taken place. In addition, two workshops for PREVIEW partners and students were highly successful, one in Copenhagen, DK, in 2013, and a summer school in San Sebastian, ES, in 2016. There are also plans for a second summer school in San Sebastian in 2020.

Exploitation of foreground
Licence agreements
This part describes the PREVIEW project’s actual and potential licence agreements for protection and ex-ploitation of the study’s findings. As the main findings became available only recently (from Oct 2018), only two agreements have been initiated at this stage:
1) A memorandum of understanding (MOU) between the PREVIEW Consortium and the Glycemic Index Foundation (Australia) to translate the PREVIEW intervention into the ‘real’ world for people with pre-diabetes, and
2) An exploitable toolkit for PA developed by Swansea University (SU).

Potential licence agreements
Stemming from its relationship to UNSYD, UNOTT and PREVIEW, the Glycemic Index Foundation (GIF) has negotiated a non-exclusive Agreement with Nutritics (UK) to license the AUSNUT 2011-13 nutrient data base compiled with appended GI values. AUSNUT 2011–13 is a set of files that enables food, dietary sup-plement and nutrient intake estimates to be made from the 2011-13 Australian Health Survey (AHS). It in-cludes foods and dietary supplements consumed as part of the 2011-12 National Nutrition and PA Survey (NNPAS) and the 2012-13 National Aboriginal and Torres Strait Islander Nutrition and PA Survey (NATSINPAS) components of the Australian Health Survey. The GI values were appended to the AusNut2011-13 data base using the procedure detailed in a peer reviewed article “Methodology for adding glycaemic index values to an Australian food composition database”. GIF has agreed to licence the data base to Nutritics for the purpose of incorporating the GI values for Nutritics Data Base for nutrient analysis includes a Web App and Libro smartphone applications. Nutritics have launched the Nutritics Data Base in Australia in 2017. The License Agreement is for a 5-year period to 2023 globally.
Other potential licence agreements are currently being explored as described below. PREVIEW members have also identified additional sectors beyond just the food industry, which ought to be potential exploi-ters of new knowledge from the PREVIEW programme. These included:
• food and health ingredient companies, in addition to manufacturers/marketers
• food retailers and supermarkets
• food and agriculture associations
• government policy makers
• obesity-related NGOs
• diabetes-related NGOs
• health care and health insurance organizations
• health advice organisations
• professional organisations, (e.g. GPs, dietitians, obesity specialists)
• slimming clubs and organisations (e.g. Weight Watchers, Jenny Craig)
• pharma including diagnostics
• media organisations


Potential Impact:
Potential Impact
Potential socio-economic impact
For many reasons, the PREVIEW Project is a landmark study in diabetes prevention and weight loss maintenance. It is the largest study of its kind incorporating a randomised clinical trial in ~2,500 individuals as well as population studies in ~162,000 individuals. PREVIEW is the first diabetes prevention study to use total diet replacements to achieve weight loss (≥ 8% in 8 weeks), and then compare two diets of different composition, and two physical activity intensities for weight loss maintenance. No other study of this na-ture has incorporated multiple nations, adults across a wide age range, and children and adolescents. Most importantly, its findings suggest that it is possible to cost-effectively prevent T2D in most individuals at high risk.
The PREVIEW intervention trial (RCT) was designed to reduce the risk of T2D in children, adolescents and adults with pre-diabetes irrespective of treatment group. The intervention (3 years for adults and 2 years for children and adolescents) was designed to be cost-effective and delivered to groups of people with pre-diabetes by nutrition and exercise counsellors, not necessarily by fitness instructors. Of note, the in-tervention gradually reduced the support (fading contact) given by counsellors and required an approach that helped to empower participants to change their behaviour and maintain this change over 3 years.
The intervention was also designed as a public health intervention and did not provide individualised, pre-cise exercise prescriptions for participants but was based on current physical activity recommendations from the Australia, Canada, UK, USA and WHO. These recommendations included guidance on the amount and intensity of physical activity required to promote health including moderate and vigorous intensity ac-tivity (MVPA), and activities for bone and muscle health.
The PREVIEW RCT was conducted in 8 countries, with a wide age range of participants starting their pro-gramme at different times of the year, in different seasons, widely ranging climates, infrastructures and political landscapes. The PREVIEW RCT included the integration of dietary instruction behaviour change and physical activity guidance. Participants were offered a high level of group counselling and support. The ed-ucational materials were developed in multiple languages. The approach taken by nutrition and exercise counsellors was to use their knowledge and expertise in the context of their participants, and the envi-ronment and context within which they lived. All these components increase the extent to which the find-ings can be generalised and actioned by other population groups.
The most important finding is that the PREVIEW RCT showed that 96% of the participants who successfully lost 8% or more of their total body weight (n = 1857) and went on to complete the trial (n = 962), did not develop T2D within 3 years, despite regaining some weight. Eight in ten study participants were able to rapidly lose ≥ 8% of their total body weight in 8 weeks using total meal replacement diets (~800 calo-ries/day). After successful weight loss, participants were encouraged to eat one of 2 healthy diets and un-dertake one of 2 regular exercise strategies. Over a period of 34 months involving 18 face-to-face visits, they were provided with a high level of counselling and support in a group setting. Only 4% of those who completed the 3-year lifestyle intervention developed T2D, whereas, based on the findings of similar life-style interventions, it was estimated that about 13% would progress to T2D in that time. The key is likely to be the substantial amount of weight loss achieved at the beginning of intervention (an average of 11% vs 5-7% in other prevention studies).

Because of the low overall number of T2D cases (n=62), we were unable to detect a difference between the 2 diets or the 2 physical activity intensities or any interaction between diet and physical activity. Cost-effectiveness is a crucial consideration in intervention planning. In our cost-effectiveness analysis compar-ing the 2 diets, the moderate protein diet showed better cost-effectiveness. However, the high protein diet seemed to be slightly more favourable than the medium protein diet to maintain physical and psycho-logical quality of life after the weight loss phase. As the high protein diet was found to be similar in costs to an average diet in the UK, a higher protein intake could also be advised as a means to improve health out-comes with no additional expenses compared to an average diet.

Wider societal implications
Worldwide, diabetes is a significant cause of mortality and morbidity, invoking considerable costs for healthcare systems (NCD Risk Factor Collaboration (NCD-RisC), 2016). The most common form of diabetes is T2D, caused by a combination of reduced insulin secretion and increased resistance to insulin-mediated glucose disposal, leading to hyperglycaemia. Individuals with hyperglycaemia have an increased risk of de-veloping cardiovascular and renal diseases. While the risk of developing T2D is influenced by interactions between genetic predisposition, environment, and lifestyle choices, sedentary lifestyle and obesity are considered as major risk factors (McGuire et al., 2016; World Health Organization, 2016). Furthermore, es-pecially in high income countries, risk of developing T2D appears to be higher among lower socio-economic groups, whether measured by educational level, income, or occupation (e.g. Agardh et al., 2007, 2011).
PREVIEW findings in the RCT imply that substantial weight loss (facilitated by total meal replacements) fol-lowed by partial weight loss maintenance over 3 years can be the single most important factor in the delay and avoidance of T2D in overweight individuals with pre-diabetes. Reducing weight in populations will also reduce cardiovascular disease (CVD), muscular-skeletal and joint diseases. However, there is a small per-centage of overweight pre-diabetic people for whom substantial weight loss and exercise will not achieve delay or remission of T2D and who will benefit from medication even when successful in achieving lifestyle changes.
The PREVIEW RCT has important messages for policy makers and other stakeholders in the diet and health space. The first is that governments and organisations responsible for healthcare delivery should enable and support achievement of substantial and rapid weight loss with proven, safe methods e.g. total meal replacement diets in overweight and obese people.
The second is that governments and organisations responsible for health care delivery, should enable, promote and subsidise the high level of personal and group support required for successful lifestyle and behaviour change. Indeed, the UK recently announced that its National Health Service will subsidise the cost of total meal replacements for weight loss. PREVIEW broadens the evidence that such a policy is based on robust scientific evidence and applies to people at high risk of developing T2D.
Governments and stakeholders should also provide messaging around the maintenance of a healthy weight as a core life still and a goal for generations that are living longer. Obesity and inactivity make us vul-nerable to costly diseases. Government funding is needed to encourage the high level of support that is needed to ‘renovate’ lifestyles in family, group, community and individual professional and clinical settings.
Specifically, to reduce the incidence of T2D, the societal implications of the PREVIEW RCT include:
• That governments and health care providers enable and support achievement of substantial and rapid weight loss, with proven, safe methods (e.g. total meal replacements) in high risk, overweight and obese people.
• Consistent with PREVIEW experience, that governments and providers enable, promote and subsidise the high level of personal and group support required for successful lifestyle and behaviour change.
• In keeping with this goal, that governments, community institutions and health providers enable and support exercise programs which are appropriate, accessible and safe, to increase activity levels and promote health.

References
Agardh, E., Allebeck, P., Hallqvist, J., Moradi, T. and Sidorchuk, A. (2011) ‘T2D incidence and socio-economic position: A systematic review and meta-analysis’, International Journal of Epidemiology, 40(3), pp. 804–818. doi: 10.1093/ije/dyr029.
McGuire, H., Longson, D., Adler, A., Farmer, A. and Lewin, I. (2016) ‘Management of T2D in adults: summary of updated NICE guidance’, BMJ. BMJ Publishing Group Ltd, 353. doi: 10.1136/bmj.i1575.
NCD Risk Factor Collaboration (NCD-RisC) (2016) ‘Worldwide trends in diabetes since 1980: a pooled analysis of 751 population-based studies with 4.4 million participants’, Lancet (London, England). NCD Risk Factor Collaboration. Open Access article distributed under the terms of CC BY, 387(10027), pp. 1513–1530. doi: 10.1016/S0140-6736(16)00618-8.
World Health Organization (2016) ‘Global Report on Diabetes’, Isbn, 978, p. 88. doi: ISBN 978 92 4 156525 7.


List of Websites:
http://preview.ning.com/