PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • All participants gave written informed consent, and the study was approved by the local ethics committees in the participating countries and the Internal Review Board of the International Agency for Research on Cancer. The design and methods of the InterAct case-cohort study have previously been described [9]. InterAct is a case-cohort study nested within the EPIC cohort, and the project involves 29 institutions in nine European countries. Ascertainment of incident T2D involved a review of the existing EPIC datasets at each centre using multiple sources of evidence including self-report, linkage to primary-care registers, secondary-care registers, medication use (drug registers), hospital admissions, and mortality data. Information from any follow-up visit or external evidence with a date later than the baseline visit was used. To increase the specificity of the case definition, we sought further evidence for all cases with information on incident T2D from fewer than two independent sources, including seeking information via individual medical records review in some centres. Cases in Denmark and Sweden were not ascertained by self-report, but identified via local and national diabetes and pharmaceutical registers, and hence all ascertained cases were considered to be verified. Follow-up was censored at the date of diagnosis, 31 December 2007, or the date of death, whichever occurred first. All ascertained cases with any evidence of diabetes at baseline were excluded. Prevalent diabetes was identified on the basis of baseline self-report of a history of diabetes, doctor-diagnosed diabetes, diabetes drug use, or evidence of diabetes after baseline with a date of diagnosis earlier than the baseline recruitment date. A total of 340,234 participants of European descent were followed up for 3.99 million person-years (mean [range] follow-up of 11.7 [0–17.5] y), during which 12,403 verified incident cases of T2D were identified [1]. Individuals without stored blood (n = 109,625) or without reported diabetes status (n = 5,821) were excluded. A centre-stratified, random sub-cohort of 16,835 individuals was selected. After exclusion of 548 individuals with prevalent diabetes and 133 with unknown diabetes status, the sub-cohort included 16,154 individuals for analysis. By design, because of the random selection, this sub-cohort also included a set of 778 individuals who developed incident T2D during follow-up. Participants in the random sub-cohort were similar to all EPIC participants eligible for inclusion in InterAct [9]. InterAct cases were followed-up for a mean (standard deviation [SD]) of 6.9 (3.3) y, and 49.8% were men. The overall incidence of T2D in InterAct was 3.8 per 1,000 person-years of follow-up. Weight and height were measured with participants not wearing shoes and in light clothing or underwear in the majority of centres [10]. Waist circumference (WC) was measured either at the narrowest circumference of the torso or at the midpoint between the lower ribs and the iliac crest. Hip circumference was measured horizontally at the level of the largest lateral extension of the hips or over the buttocks. For a subset of the Oxford participants (n = 363), only self-reported waist and hip circumferences were available. Each participant's body weight and waist and hip circumferences were corrected for the clothing worn during measurement in order to reduce heterogeneity due to protocol differences among centres. Correction included adjustment for self-reporting in Oxford participants using a prediction equation based on a comparison of self-reported and measured data in a sample of 5,000 of the Oxford general population [10],[11]. Body mass index (BMI) was calculated as weight (kg)/height (m) squared. Waist–hip ratio was calculated and expressed as a percentage. Measures of waist and hip circumference were not performed in Umeå, Sweden (n = 1,845), and were missing in an additional 173 and 193 InterAct participants, respectively [12]. Standardised information was collected by questionnaire at baseline on education, smoking status [13], and diabetes family history [14]. Physical activity was based on a brief questionnaire covering occupation and recreational activity, which was summarised into an ordered categorical overall physical activity index (inactive, moderately inactive, moderately active, and active) that has been validated in the populations participating in EPIC [15],[16]. In one of the centres (Umeå, Sweden), a slightly different questionnaire was used to assess physical activity. From this questionnaire we derived a four-category index similar to that derived from all other study locations based on two questions on occupational and leisure time physical activity [16]. Usual food intake was estimated using country-specific validated dietary questionnaires. Estimated individual nutrient intakes were derived from foods included in the dietary questionnaires through the standardised EPIC Nutrient Database [17]. Participants in the lowest and highest 1% of the cohort distribution of the ratio of reported total energy intake to energy requirement were excluded from the current study (n = 736). The Mediterranean dietary pattern as used here is characterised by a high consumption of unrefined cereals, fruits, vegetables, olive oil, and legumes; a moderate consumption of dairy products (mostly cheese and yogurt); moderate wine consumption; a moderate-to-high consumption of fish; and a low consumption of meat and meat products [18],[19]. Adherence to the Mediterranean diet was assessed using the relative Mediterranean diet score that has previously been associated with the risk of incident T2D in InterAct [20]. This score included nine nutritional components characteristic of the Mediterranean diet: seven potentially beneficial components (vegetables, legumes, fruits and nuts, cereals, fish and seafood, olive oil, and moderate alcohol consumption) and two potentially detrimental components (meat and meat products, and dairy products). The overall relative Mediterranean diet score was divided into categories reflecting low (0–6 points), medium (7–10 points), and high (11–18 points) adherence to the Mediterranean diet on the basis of previously published cutoff points [21]. DNA was not available for Danish (n = 4,037) participants, leaving a total maximum sample size of 10,348 incident cases and 14,671 random sub-cohort participants with DNA available, including 13,394 non-diabetic InterAct sub-cohort participants. Hence, of the original 27,779 InterAct participants, a maximum of 23,742 were eligible for genetic analyses. Of these, a total of 19,651 participants, including 8,582 incident cases and 11,069 non-diabetic sub-cohort participants, had DNA available for genotyping (Table S1). DNA was extracted from up to 1 ml of buffy coat for each individual from a citrated blood sample. Standard procedures on an automated Autopure LS DNA extraction system (Qiagen) with PUREGENE chemistry (Qiagen) were used, and the DNA was hydrated overnight prior to further processing. DNA samples were quantified by PicoGreen assay (Quant-iT) and normalised to 50 ng/ µl. A total of 10,027 participants (4,644 cases) were selected across all except the Danish centres for genome-wide genotyping using the Illumina 660W-Quad BeadChip at the Wellcome Trust Sanger Institute. Samples were randomly selected from those successfully genotyped on Sequenom or Taqman platforms (based on DNA concentration, call rate, and gender matching sex chromosome genotype), with the number of individuals selected per centre being proportional to the percentage of total cases in that centre. Of these, a total of 9,431 samples passed quality control criteria following genome-wide genotyping (call rate >95%, no conflict between gender and X chromosome heterozygosity, concordant candidate genotyping, not an outlier for autosomal heterozygosity or ethnicity), with 99.9% and 99.5% of included samples at call rates of 97% and 99%, respectively. In addition, 9,794 InterAct participants with available DNA and not selected for genome-wide measurement were genotyped using the Illumina Cardio-Metabochip [16]. Genotyping was completed in 9,467 InterAct samples, with 99.8% and 98.2% of samples at call rates of 97% and 99%, respectively. Genotype information and quality metrics for the 49 T2D loci in the InterAct random sub-cohort are included in Table S4. Genotype distributions were in Hardy-Weinberg equilibrium using a Bonferroni-adjusted significance level of p<0.001, with the exception of rs11063069 (CCND2) in the Illumina 660 W subset (p = 7.84×10−13). We selected all top-ranked SNPs from loci reaching genome-wide significance for association with T2D in European-descent populations in the latest DIAGRAM meta-analysis [22]. From a total of 66 reported T2D-associated variants, we excluded the DUSP8 locus, which had a parent-of-origin-specific effect [23], in addition to 15 variants that were significant genome-wide in Asian populations only. The top-ranked SNP at DUSP9 on the X chromosome was also unavailable and without a suitable proxy, and was therefore not included. Hence, a total of 49 variants were selected for the InterAct genetic score, including two established obesity loci (FTO and MC4R) and two loci that reached genome-wide significance in sex-differentiated meta-analyses (CCND2 and GIPR) [22]. The top-ranked SNP at HNF1B (rs11651052) was not available on the Illumina 660 W-Quad BeadChip, and a proxy in high linkage disequilibrium (rs4430796; r2 = 0.97) was used instead. Risk alleles (Table S4) were summed into a genetic risk score, including imputation of missing genotypes. Characteristics of all InterAct participants and of the random sub-cohort are summarised, alongside those of individuals who had DNA available for genotyping, in Tables S2 and S3, respectively. Associations between the published T2D risk allele for each SNP (Table S4) and incident T2D were estimated using Prentice-weighted Cox regression models, separately within each country, with age as the underlying time scale, adjusted for sex and centre and assuming additive genetic effects with the T2D risk allele as the effect allele [9]. The hazard ratio (HR) for each SNP was combined across countries using random effects meta-analysis. Sensitivity analyses were performed replacing centre by linearized (i.e., expressed in kilometres) latitude and longitude of the centre [24],[25], and also with additional inclusion of BMI (continuous) in the sex- and centre-adjusted model. A genetic risk score was constructed by summing the number of risk alleles across all 49 loci. To maximise sample size, missing genotypes were imputed by assigning the mean genotype in the overall dataset at each locus for cases and non-cases separately. This was done only for individuals successfully genotyped for at least 47 of the 49 loci, and allowed the inclusion of 18,890 rather than 18,390 individuals in analyses of the genetic score. The HR for T2D per 1-SD increase in the score (SD calculated in the sub-cohort) was estimated as described above. Sensitivity analyses were performed using the original non-imputed genetic risk score, and also a weighted version of the two scores, where the weights for each SNP were equal to the log odds ratio for that SNP from DIAGRAM replication samples [22]. Further sensitivity analyses were performed removing CCND2 from the risk score, and also removing CCND2 and GIPR (identified in sex-differentiated meta-analyses) specifically for the analysis of interaction with sex. Meta-regression models were used to explore whether average age, BMI, or WC by country in the sub-cohort explained any of the heterogeneity between countries. Interactions between the imputed, unweighted genetic risk score and each of the following risk factors previously shown to be associated with T2D in InterAct were assessed: sex [9], diabetes family history [14], BMI (three levels: <25, 25 to <30, ≥30 kg/m2) [12], WC (three levels: men, <94 cm [34.6 inches], 94 to <102 cm [34.6 to <40 inches], ≥102 cm [≥40 inches]; women, <80 cm [31.5 inches], 80 to <88 cm [31.5 to <35 inches], ≥88 cm [≥35 inches]) [12], age (continuous) [9], physical activity (four levels: inactive, moderately inactive, moderately active, active) [26], and Mediterranean diet score (integer scale from 0–18, included as a continuous variable) [20]. To estimate p-values for interaction with either the genetic risk score or individual SNPs, a parameter representing the interaction between the score or SNP and the variable of interest was included in country-specific Prentice-weighted Cox regression models, with additional adjustment for centre and sex and using age as the underlying time scale (except for analyses of baseline age, where calendar time was used). The interaction parameter estimates were then combined across countries using random effects meta-analysis, and observed versus expected p-values were plotted for individual SNP interactions (Figure S1). Numerical p-values were included in tables and figures, but Bonferroni-adjusted levels of significance were used to draw inferences about statistical significance, to account for the number of tests performed for the score (score by seven T2D risk factors, equivalent to seven tests, with p<0.007 ensuring control of family-wise error rate at level α = 0.05) or individual SNPs (49 SNPs by seven T2D risk factors, equivalent to 343 tests, with p<1.46×10−4 ensuring control of family-wise error rate at level α = 0.05). HRs were also calculated by level for each risk factor, as described above (age at baseline <50, 50 to <60, ≥60 y; Mediterranean diet score 0–6, 7–10, and 11–18). We additionally grouped T2D cases according to their age of diagnosis (<55, 55 to <65, ≥65 y) and fit different weighted Cox models using each of these groups as a separate outcome. To estimate the cumulative incidence of T2D within strata defined by quartiles of the genetic risk score (cutoffs derived from the distribution in the sub-cohort) and modifiable risk factors, we used the Stata bsample command to recreate the full cohort by resampling with replacement from the sub-cohort, according to the distributions of the stratum variables within the sub-cohort. This made it possible to estimate absolute cumulative incidences (one minus the Kaplan-Meier estimate of the survivor function).
rdf:type