Participants were recruited in January 2008 by 48 GPs and 24 nurse practitioners from an association of 14 primary care practices in the Netherlands. A Dutch translation of the Finnish FINDRISC [17
] was sent to GP patients aged ≥40 and ≤70
16032). The FINDRISC was validated in three Dutch cohorts and was found to be a reasonably good predictor of incident diabetes in the Netherlands [18
]. All individuals with a score ≥13 points (n
1533) were invited to participate in the intervention. Randomization was performed on the level of the individual. In total, 479 individuals were allocated to the intervention group and 446 individuals to the usual care group. Details of participant recruitment and intervention reach were described previously [19
The APHRODITE intervention was based on the transtheoretical model [20
]. Stage transition was supported by using behavioural change techniques to influence motivation (motivational interviewing, decisional balance), action (goal setting, action planning, barrier identification) and maintenance (relapse prevention) [21
]. Research objectives focused on weight loss, increasing the amount of physical activity and improving dietary composition. Clinical and lifestyle measurements were performed at baseline and after 6, 18 and 30
months. Details of participant measurement taking were described previously [23
To ascertain regular contact with health care providers, 11 consultations of 20 minutes were scheduled over 2.5
years with alternately the nurse practitioner and the GP. To decrease workload for providers and stimulate contact between participants, five group-meetings of 1 hour were organised by dieticians and physiotherapists to provide more detailed information on diet and exercise. Moreover, all individuals were invited for a 1-hour personal consultation with the dietician. Participants were invited for the group-meetings in their own town directly by the dietician or physiotherapist. Participants could call the dieticians’ or physiotherapists’ office to join another meeting if the day or time didn’t suit them. Details of the planning of the intervention and content of the group-consultations are shown in Table .
Planning of the APHRODITE intervention and content of the group-consultations
The programme was free of charge for all participants. Providers received financial reimbursement for all consultations with their participants according to Dutch payment standards. The intervention was registered with the Dutch Trial Register (NTR1082). The Medical Ethical Review Committee of the Catharina Hospital in Eindhoven gave ethical approval to the study (M07-1705). All participants gave informed consent for participation.
During the admission interview, participants in the usual care group received oral and written information about type 2 diabetes and a healthy lifestyle. The nurse practitioner was visited only for measurements (10 minutes) at baseline and after 6, 18 and 30
months. Apart from the admission interview participants did not have study-related encounters with the GP.
Training of GPs and nurse practitioners
Before the start of the study, all nurse practitioners and GPs received a two-evening training on the theoretical framework of the intervention and its translation into practice. In addition, all nurse practitioners received a five-evening course in motivational interviewing (MI) [24
]. During the project, two return-meetings were organised with the GPs (once a year) and four with the nurse practitioners (every half a year). All nurse practitioners in our study were certified and had previously obtained a degree on higher vocational education level.
Drop-out and attendance rates
A list of individuals ending participation was kept by the project assistant. Presence was registered by the GP or nurse practitioner (individual consultations) or by the dietician or physiotherapist (group-consultations). Attendance rates were calculated based on all persons participating at the particular time-point, excluding drop-outs (intervention: N
%); usual care: N
%) and individuals diagnosed with diabetes (intervention: N
%); usual care: N
%)). One practice was left out of attendance rate calculations as presence at individual consultations was not accurately registered.
Participant and provider questionnaires
Questionnaires were developed and reviewed by an expert panel of epidemiologists, GPs and nurse practitioners. Provider questionnaires were filled out within one month after finishing the project. Response to the questionnaires was 80
% within GPs and 100
% within nurse practitioners. Participant questionnaires (intervention group) were filled out during the 30-month data collection. Response to this questionnaire was 84
%. As follow-up ended as soon as persons were diagnosed with diabetes or withdrew from the study, no records were available from drop-outs (N
%)) and individuals with diabetes (N
Confidence of professionals in diabetes prevention in primary care, confidence in diabetes prevention by lifestyle intervention and chance of success of diabetes prevention in primary care were assessed on 5-point Likert scales. Satisfaction of professionals with individual consultations was assessed on a 1 to 10 scale, questioning: ‘how much pleasure did you experience in consultations with intervention group participants?’. For analysis, 1–5 was categorised as ‘low’, 6–7 as ‘average’ and 8–10 as ‘high’.
Opinions of providers and participants on knowledge of providers were assessed with Yes/No-questions asking ‘do you think you have / your GP/nurse practitioner has enough knowledge on the following topics?’. Topics questioned were type 2 diabetes, complications of diabetes, a healthy weight, intakes of (saturated) fat and dietary fibre, benefits of exercise, and the amount of exercise needed. Participant satisfaction with lifestyle change guidance and providers perceptions of their suitability were assessed on 5-point Likert scales. Usefulness and desirability of the MI-course were also assessed on 5-point Likert scales.
Satisfaction of participants and professionals with the frequency and duration of consultations, usefulness of group-consultations, and opinions of participants on desirability of guided exercise programs and individual dietary counseling were assessed on 5-point Likert scales. Opinions of providers on desirability of guided exercise programs, on suitability of primary care for group-consultations, and on perceived organizational barriers for group-consultations were assessed with open questions.
Sample size calculation
Sample size calculation was based on the main outcome diabetes incidence. Studies in experimental settings had shown reductions in diabetes incidence between study groups up to 58
]. However as implementation of lifestyle interventions in real life settings is challenging [6
], more modest differences in diabetes incidence were expected for this study. To detect small differences in diabetes incidence (Cohen’s conventional effect size of 0.1 [25
]), with a power of 0.8, 393 individuals were needed in each arm. When a post-hoc correction for correlation on the nurse practitioner level was applied (variance 0.03), [26
], this number changed to 405. As in total 925 individuals could be included, this allowed for a dropout rate of approximately 15
%, which was in line with others [2
Differences in baseline characteristics between the study groups were evaluated with independent samples’ t-tests or chi-square tests. Other differences between study groups and between subgroups of participants were evaluated using multilevel analysis (level 1: participant; level 2: nurse practitioner). As the clustering effects on the GP level (level 3) were neglectable after accounting for the effects of the nurse practitioner, the GP level was omitted. Changes in clinical outcomes of drop-outs were calculated from the last available measurement before withdrawal from the project. Analyses were performed using SPSS version 18.0 and SAS version 9.2. A p-value of <0.05 was considered significant.