Participants were recruited among the inhabitants (males and females aged 40–74 years) of the working area of two Municipal Health Services in the Southwest Region of the Netherlands. Their name and address information was obtained from municipal authority records. An invitation letter together with the study material, comprising of an information brochure, a consent form, a tape measure and a questionnaire were sent by mail to the target population. The questionnaire contained questions on demographic features such as marital status and education, self-perceived health, weight and height, lifestyle, symptoms and diabetes related risk factors, family history of diabetes and personal history of selected diseases, including cardiovascular diseases, stroke, and diabetes.
The individuals were requested to measure their waist circumference twice with the tape measure sent to their home addresses and fill out their measurements on the questionnaire, in addition to data on sociodemographic characteristics, lifestyle and risk factors. After completion of the questionnaire, the potential participants returned it in the enclosed postage-free reply envelope. Eligibility, which was based on the absence of pre-existing diabetes and a self-reported waist circumference indicating abdominal obesity, was ascertained upon receipt of the questionnaires. Abdominal obesity was defined as a waist circumference of ≥80 cm for women and ≥94 cm for men, following the cut-off points of the International Diabetes Federation [8
The validity of the self-reported waist circumference was assessed in a subset of 160 persons by comparing the self-measurements with measurements taken by a trained nurse. The Cohen's kappa, which was calculated as a proxy for the reliability between self-reported and professionally-measured waist circumference, was an acceptable 0.64.
A flowchart of the study is presented in Figure . The study protocol and materials were approved by the Medical Ethical Review Committee of Erasmus MC and the RCT was registered with Current Controlled Trials (ISRCTN75983009).
Flowchart of the screening trial.
Eligible consenting respondents were stratified by gender and individually randomized at a 1:1 ratio to the intervention (screening) arm or the control arm using a random number generator. Participants in the screening arm received an invitation to attend screening, which consisted of fasting plasma glucose (FPG) measurement and information about the importance of a healthy lifestyle (Netherlands Nutrition Centre). Those in the control arm only received the healthy lifestyle information.
FPG was used as the screening criterion in accordance with the Dutch College of General Practitioners’ guidelines [22
]. The FPG cut-off values were 7.0 mmol/L or higher for diabetes and between 6.1 and 6.9 mmol/L for Impaired Fasting Glucose (IFG) [22
]. As dyslipidemia are often present in overweight and obese subjects, and it is clinical practice to measure blood lipids in patients with diabetes, we opted to additionally measure serum lipids (total cholesterol, high-density lipoprotein, HDL, cholesterol and triglycerides). FPG concentrations were determined using the hexokinase method. Fasting serum total cholesterol, HDL cholesterol and triglycerides were quantified using an enzymatic colorimetric method. All analyses were performed on the Beckman AU2700 chemical analyzer (Beckman Coulter Nederland BV, Woerden, Netherlands), using kits supplied by Beckman. Low-density lipoprotein (LDL) was calculated using the Friedewald formula [23
Individuals with a FPG level of 7.0 mmol/L or higher were referred to their general practitioners (GP) for confirmatory diagnostic testing and treatment, which included glycemic control by lifestyle intervention or antihyperglycemic medication (oral agents and eventually insulin if necessary), and management of any present cardiovascular risk factors. Individuals with IFG were also referred to their GP for FPG monitoring and treatment of possible cardiovascular risk factors according to the GP guidelines. Participants were told their blood glucose level; this information was also given to their GP. Normoglycemic individuals with an estimated absolute cardiovascular disease risk of 5% or higher, according to the Systematic Coronary Risk Evaluation (SCORE) risk model, were also referred to their GP [24
]. The SCORE risk estimates are calculated based on age, blood pressure level, current smoking status and the ratio of total cholesterol to HDL cholesterol.
Because all participants in this RCT received the same disease information, those assigned to the control arm might take opportunistic testing after learning about the simplicity of the screening test. Therefore, FPG testing in the control arm was monitored through their GPs and linkages with laboratory databases, similar to a previous prostate cancer screening trial [25
The primary endpoint of the trial was the first occurrence of a fatal or non-fatal cardiovascular event within the follow-up period after randomization. A cardiovascular event was defined as non-fatal coronary heart disease, non-fatal stroke or death from any disease of the circulatory system (among diabetic cases and those at intermediate to high cardiovascular risk). Estimates of prevented microvascular complications are also important, but this study primarily focuses on macrovascular complications, because these are by far the most important contributor to premature death among patients with diabetes. A secondary analysis will include the primary non-fatal endpoint, as well as non-cardiovascular diabetes-related morbidity and mortality, all-cause mortality and cardiovascular interventions (e.g. revascularization).
Intermediate endpoints include the prevalence/incidence of IFG and newly screen-detected type 2 diabetes, screening performance (attendance, referral and detection rates, and test characteristics), number of individuals with an absolute cardiovascular disease risk of 5% or higher (intermediate to high), and temporal changes in the levels of blood parameters (glucose, HbA1c, lipids and blood pressure).
Follow-up and data collection
For each participant, a minimum follow-up period of 10 years is planned. During follow-up, information on the time of diabetes diagnosis, cardiovascular risk profiles, cardiovascular morbidity, diabetic complications and use of medication will be obtained from GPs and from the National Hospital Discharge Register. Data on cardiovascular risk profiles, cardiovascular morbidity and medication use will also be collected for those referred to their GP based on their absolute cardiovascular disease risk. Vital status will be ascertained through record linkage with the Cause of Death Registry of Statistics Netherlands.
In the sample size calculations we estimated the minimum number of participants required to identify a statistically significant 20% reduction in non-fatal and fatal cardiovascular events in the screening compared with the control arm. The expected rate of the composite outcome of cardiovascular events in an aging cohort of 40–74 years at entry was estimated at 2% per year, based on 2003 age and gender specific data obtained from the Hospital Admission Registry for non-fatal events and Statistics Netherlands for fatal events. Using age-specific abdominal obesity prevalence data, the number of events was estimated in the sub-population with abdominal obesity [26
]. Using the method developed by Baan and coworkers [27
] to assess the diabetes-related mortality in the Dutch population, we additionally accounted for the prevalence of pre-existing diabetes among individuals with abdominal obesity [6
], the relative risk of developing type 2 diabetes in this group [26
], and sex-specific hazard rates for cardiovascular events. We estimated that the required number of participants per study arm over 10 years was 30,885 to show a reduction of 20%, with a power of 80% and a two-sided significance level of 5%, assuming a contamination rate of 10% in the control arm, as previously reported for prostate-specific antigen screening [25
In the first year of the trial, we examined the feasibility of home-assessed and self-reported waist circumference measurements as a first-step screening tool for recruiting individuals at high risk. The sociodemographic characteristics of the respondents and the intermediate endpoints, particularly attendance and detection rate were assessed. Socioeconomic status was based on 2006 social status ranking of the Social and Cultural Planning Office. The ranking, from 1 (high status) to 3965 (low status) is estimated based on income, employment status and level of education in the households within postal code areas.
In 2006 and 2007, 79,142 inhabitants of the Dutch municipalities of Capelle aan den IJssel and Dordrecht, in the age group 40–74 years were invited to participate. The mailings were sent out in two different ways. Half of the population received the complete study material, consisting of a letter of invitation, information brochure, tape measure and consent form (one-step screening approach). The other half first received a letter of invitation containing a brief introduction to the study, an explanation of how to measure their waist size, and a tape measure. Individuals who responded and who were judged potentially eligible were then sent an accompanying letter, information brochure and the consent form (two-step screening approach). We aimed to find out which method produced the greatest response from people at high risk, and what the costs of each strategy were. In either approach strategy, those eligible for participation were randomized only if they gave their informed consent.
Differences between respondents and non-respondents and between the control arm and the screening arm were compared using Chi-square tests for categorical variables and unpaired t-tests for continuous variables.
To examine the effect of the variables age, waist size (continuous) and gender, marital status, education, smoking status, family history and country of birth (categorical) on the odds of attending screening, univariate logistic regression was used and then a multivariate model was constructed with the variables that had a univariate p-value of ≤0.10. The results are presented as odds ratio (OR) and corresponding 95% confidence interval (CI).
All analyses were performed using SPSS version 17 (SPSS, Chicago, IL, USA). A p-value of <0.05 was considered statistically significant.