[Show abstract][Hide abstract] ABSTRACT: Non-compliance with food record submission can induce bias in nutritional epidemiological analysis and make it difficult to draw inference from study findings. We examined the impact of demographic, lifestyle and psychosocial factors on such non-compliance during the first 3 years of participation in a multidisciplinary prospective paediatric study.
The Environmental Determinants of Diabetes in the Young (TEDDY) study collects a 3 d food record quarterly during the first year of life and semi-annually thereafter. High compliance with food record completion was defined as the participating families submitting one or more days of food record at every scheduled clinic visit.
Three centres in the USA (Colorado, Georgia/Florida and Washington) and three in Europe (Finland, Germany and Sweden).
Families who finished the first 3 years of TEDDY participation (n 8096).
High compliance was associated with having a single child, older maternal age, higher maternal education and father responding to study questionnaires. Families showing poor compliance were more likely to be living far from the study centres, from ethnic minority groups, living in a crowded household and not attending clinic visits regularly. Postpartum depression, maternal smoking behaviour and mother working outside the home were also independently associated with poor compliance.
These findings identified specific groups for targeted strategies to encourage completion of food records, thereby reducing potential bias in multidisciplinary collaborative research.
Public Health Nutrition 06/2015; DOI:10.1017/S1368980015001883 · 2.48 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Tissue transglutaminase autoantibodies (tTGAs) represent the first evidence of celiac disease (CD) development. Associations of HLA-DR3-DQA1*05:01-DQB1*02:01 (i.e., DR3-DQ2) and, to a lesser extent, DR4-DQA1*03:01-DQB1*03:02 (i.e., DR4-DQ8) with the risk of CD differ by country, consistent with additional genetic heterogeneity that further refines risk. Therefore, we examined human leukocyte antigen (HLA) factors other than DR3-DQ2 for their contribution to developing tTGAs.
The Environmental Determinants of Diabetes in the Young (TEDDY) study enrolled 8,676 infants at an increased HLA-DR-DQ risk for type 1 diabetes and CD into a 15-year prospective surveillance follow-up. Of those followed up, 21% (n=1,813) carried DR3-DQ2/DR3-DQ2, 39% (n=3,359) carried DR3-DQ2/DR4-DQ8, 20% (n=1701) carried DR4-DQ8/DR4-DQ8, and 17% (n=1,493) carried DR4-DQ8/DQ4. Within TEDDY, a nested case-control design of 248 children with CD autoimmunity (CDA) and 248 matched control children were genotyped for HLA-B, -DRB3, -DRB4, -DPA1, and -DPB1 genes, and the entire cohort was genotyped for single-nucleotide polymorphisms (SNPs) using the Illumina ImmunoChip. CDA was defined as a positive tTGA test at two consecutive clinic visits, whereas matching in those with no evidence of tTGAs was based on the presence of HLA-DQ2, country, and sex.
After adjustment for DR3-DQ2 and restriction to allele frequency (AF) ≥5%, HLA-DPB1*04:01 was inversely associated with CDA by conditional logistic regression (AF=44%, odds ratio=0.71, 95% confidence interval (CI)=0.53-0.96, P=0.025). This association of time to CDA and HLA-DPB1*04:01 was replicated with statistical significance in the remainder of the cohort using imputation for specific HLA alleles based on SNP genotyping (hazard ratio=0.84, 95% CI=0.73-0.96, P=0.013).
HLA-DPB1*04:01 may reduce the risk of tTGAs, an early marker of CD, among DR3-DQ2 children, confirming that additional variants in the HLA region influence the risk for CDA.Am J Gastroenterol advance online publication, 26 May 2015; doi:10.1038/ajg.2015.150.
The American Journal of Gastroenterology 05/2015; 110(6). DOI:10.1038/ajg.2015.150 · 9.21 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Early childhood environmental exposures, possibly infections, may be responsible for triggering islet autoimmunity and progression to type 1 diabetes (T1D). The Environmental Determinants of Diabetes in the Young (TEDDY) follows children with increased HLA-related genetic risk for future T1D. TEDDY asks parents to prospectively record the child’s infections using a diary book. The present paper shows how these large amounts of partially structured data were reduced into quantitative data-sets and further categorized into system-specific infectious disease episodes. The numbers and frequencies of acute infections and infectious episodes are shown.
Study subjects (n = 3463) included children who had attended study visits every three months from age 3 months to 4 years, without missing two or more consecutive visits during the follow-up. Parents recorded illnesses prospectively in a TEDDY Book at home. The data were entered into the study database during study visits using ICD-10 codes by a research nurse. TEDDY investigators grouped ICD-10 codes and fever reports into infectious disease entities and further arranged them into four main categories of infectious episodes: respiratory, gastrointestinal, other, and unknown febrile episodes. Incidence rate of infections was modeled as function of gender, HLA-DQ genetic risk group and study center using the Poisson regression.
A total of 113,884 ICD-10 code reports for infectious diseases recorded in the database were reduced to 71,578 infectious episodes, including 74.0% respiratory, 13.1% gastrointestinal, 5.7% other infectious episodes and 7.2% febrile episodes. Respiratory and gastrointestinal infectious episodes were more frequent during winter. Infectious episode rates peaked at 6 months and began declining after 18 months of age. The overall infectious episode rate was 5.2 episodes per person-year and varied significantly by country of residence, sex and HLA genotype.
The data reduction and categorization process developed by TEDDY enables analysis of single infectious agents as well as larger arrays of infectious agents or clinical disease entities. The preliminary descriptive analyses of the incidence of infections among TEDDY participants younger than 4 years fits well with general knowledge of infectious disease epidemiology. This protocol can be used as a template in forthcoming time-dependent TEDDY analyses and in other epidemiological studies.
[Show abstract][Hide abstract] ABSTRACT: Islet autoantibodies, in addition to elevated blood glucose, define type 1 diabetes. These autoantibodies are detectable for a variable period of time before diabetes onset. Thus, the occurrence of islet autoantibodies is associated with the beginning of the disease process. The age at, and order in, which autoantibodies appear may be associated with different genetic backgrounds or environmental exposures, or both.
Infants with HLA-DR high-risk genotypes (DR3/4, DR4/4, DR4/8 and DR3/3) were enrolled and prospectively followed with standardised autoantibody assessments quarterly throughout the first 4 years of life and then semi-annually thereafter.
Autoantibodies appeared in 549/8,503 (6.5%) children during 34,091 person-years of follow-up. Autoantibodies at 3 (0.1%) and 6 (0.2%) months of age were rare. Of the 549, 43.7% had islet autoantibodies to insulin (IAA) only, 37.7% had glutamic acid decarboxylase autoantibodies (GADA) only, 13.8% had both GADA and IAA only, 1.6% had insulinoma antigen-2 only and 3.1% had other combinations. The incidence of IAA only peaked within the first year of life and declined over the following 5 years, but GADA only increased until the second year and remained relatively constant. GADA only were more common than IAA only in HLA-DR3/3 children but less common in HLA-DR4/8 children.
Islet autoantibodies can occur very early in life and the order of appearance was related to HLA-DR-DQ genotype.
[Show abstract][Hide abstract] ABSTRACT: Role of the funding source: Funding from the NIH was used for support of the participating clinical centers and the coordinating center. The funding source did not participate in the collection or the analysis of the data. BACKGROUND. The β cell killing that characterizes type 1 diabetes (T1D) is thought to begin years before patients present clinically with metabolic decompensation; however, this primary pathologic process of the disease has not been measured. METHODS. Here, we measured β cell death with an assay that detects β cell-derived unmethylated insulin (INS) DNA. Using this assay, we performed an observational study of 50 participants from 2 cohorts at risk for developing T1D from the TrialNet Pathway to Prevention study and of 4 subjects who received islet autotransplants. RESULTS. In at-risk subjects, those who progressed to T1D had average levels of unmethylated INS DNA that were elevated modestly compared with those of healthy control subjects. In at-risk individuals that progressed to T1D, the observed increases in unmethylated INS DNA were associated with decreases in insulin secretion, indicating that the changes in unmethylated INS DNA are indicative of β cell killing. Subjects at high risk for T1D had levels of unmethylated INS DNA that were higher than those of healthy controls and higher than the levels of unmethylated INS DNA in the at-risk progressor and at-risk nonprogressor groups followed for 4 years. Evaluation of insulin secretory kinetics also distinguished high-risk subjects who progressed to overt disease from those who did not. CONCLUSION. We conclude that a blood test that measures unmethylated INS DNA serves as a marker of active β cell killing as the result of T1D-associated autoimmunity. Together, the data support the concept that β cell killing occurs sporadically during the years prior to diagnosis of T1D and is more intense in the peridiagnosis period. TRIAL REGISTRATION. Clinical Trials.gov NCT00097292. FUNDING. Funding was from the NIH, the Juvenile Diabetes Research Foundation, and the American Diabetes Association.
[Show abstract][Hide abstract] ABSTRACT: Abstract We recently developed new electrochemiluminescence (ECL) insulin autoantibody (IAA) and glutamic acid decarboxylase 65 autoantibody (GADA) assays that discriminate high-affinity, high-risk diabetes-specific autoantibodies from low-affinity, low-risk islet autoantibodies (iAbs) detected by radioassay (RAD). Here, we report a further validation of the ECL-IAA and -GADA assays in 3,484 TrialNet study participants. The ECL assay and RAD were congruent in those with prediabetes and in subjects with multiple autoantibodies, but only 24% (P<0.0001) of single RAD-IAA-positive and 46% (P<0.0001) of single RAD-GADA-positive were confirmed by the ECL-IAA and -GADA assays, respectively. During a follow-up (mean, 2.4 years), 51% of RAD-IAA-positive and 63% of RAD-GADA-positive subjects not confirmed by ECL became iAb negative, compared with only 17% of RAD-IAA-positive (P<0.0001) and 15% of RAD-GADA-positive (P<0.0001) subjects confirmed by ECL assays. Among subjects with multiple iAbs, diabetes-free survival was significantly shorter if IAA or GADA was positive by ECL and negative by RAD than if IAA or GADA was negative by ECL and positive by RAD (P<0.019 and P<0.0001, respectively). Both positive and negative predictive values in terms of progression to type 1 diabetes mellitus were superior for ECL-IAA and ECL-GADA, compared with RADs. The prevalence of the high-risk human leukocyte antigen-DR3/4, DQB1*0302 genotype was significantly higher in subjects with RAD-IAA or RAD-GADA confirmed by ECL. In conclusion, both ECL-IAA and -GADA are more disease-specific and better able to predict the risk of progression to type 1 diabetes mellitus than the current standard RADs.
[Show abstract][Hide abstract] ABSTRACT: Rationale: The relationship between clinical phenotype of childhood primary ciliary dyskinesia (PCD) and ultrastructural defects and genotype is poorly defined. Objectives: To delineate clinical features of childhood PCD and their associations with ultrastructural defects and genotype. Methods and Measurements: 118 participants <19 years old with PCD were evaluated prospectively at 6 centers in North America using standardized procedures for diagnostic testing, spirometry, chest computed tomography, respiratory cultures, and clinical phenotyping. Main Results: Clinical features included neonatal respiratory distress (82%), chronic cough (99%), and chronic nasal congestion (97%). There were no differences in clinical features or respiratory pathogens in subjects with outer dynein arm (ODA) defects (ODA alone, n=54) and ODA plus inner dynein arm (IDA) defects (ODA+IDA; n=18) versus subjects with IDA and central apparatus defects with microtubular disorganization (IDA/CA/MTD; n=40). Median FEV1 was worse in the IDA/CA/MTD group (72% predicted) versus the combined ODA groups (92% predicted; p=0.003). Median BMI was lower in the IDA/CA/MTD group (46th percentile) versus the ODA groups (70th percentile; p=0.003). For all 118 subjects, median number of lobes with bronchiectasis was 3 and alveolar consolidation was 2. However, the 5-11 year old IDA/CA/MTD group had more lobes of bronchiectasis (median= 5, p=0.0008) and consolidation (median=3, p=0.0001) compared to the ODA groups (median=3 and 2, respectively). Similar findings were observed when limited to participants with biallelic mutations. Conclusions: Lung disease was heterogeneous across all ultrastructural and genotype groups, but worse in those with IDA/CA/MTD ultrastructural defects, most of whom had biallelic mutations in CCDC39 or CCDC40.
American Journal of Respiratory and Critical Care Medicine 12/2014; 191(3). DOI:10.1164/rccm.201409-1672OC · 11.99 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Background:
In modeling a N by m data matrix, i.e. N samples on a m dimensional space, the issue arises when
m is bigger than N. The sample size cannot be increased, especially in medical research, due to the limited number of diseased subjects. Feature selection is often used to select a subset of relevant m variables, often lower than N, for use in model construction.
A multiple step bootstrap method is proposed to quantify relevance of candidate predictors with the
outcome based on their areas under the Receiver Operating Characteristic curve (ROCAUCs) from bootstrap
resamples and then select only significant variables, which meet pre-specified criteria, as a feature selection process.
Extensive simulation was conducted using thousands of predictor variables and 5 levels of prediction
ability between the true predictor and the outcome. The results from the simulation data indicate that the mean of ROCAUCs from bootstrap samples is close to the true ROCAUC. Even with only 30 cases and 30 controls, 25 out of 25 listed predictor variables provide the correct level of classification ability by using mean of bootstrapped ROCAUCs. The proposed bootstrapped ROCAUCs method outperforms the single ROCAUC. The standard error of mean of bootstrapped ROCAUCs was 20% to 50% smaller than the standard error of the single ROCAUC estimate from the original sample. An illustrative example is presented to apply the proposed methodology to identify the gene expressions that could predict clinical survival in breast cancer patients, using the Van’t Veer study’s breast cancer data.
We conclude that the bootstrapped ROCAUCs methodology is intuitive and attractive for use in
feature selection problems when the goals of the study are to identify important predictors and to provide insight regarding the discriminative or predictive ability of individual predictor variables. Such goals are common among microarray studies and new biomarker discovery.
Journal of Proteomics & Bioinformatics 10/2014; S9. DOI:10.4172/jpb.S9-006
[Show abstract][Hide abstract] ABSTRACT: Established in 2003 by the Office of Rare Diseases Research (ORDR), in collaboration with several National Institutes of Health (NIH) Institutes/Centers, the Rare Diseases Clinical Research Network (RDCRN) consists of multiple clinical consortia conducting research in more than 200 rare diseases. The RDCRN supports longitudinal or natural history, pilot, Phase I, II, and III, case-control, cross-sectional, chart review, physician survey, bio-repository, and RDCRN Contact Registry (CR) studies. To date, there have been 24,684 participants enrolled on 120 studies from 446 sites worldwide. An additional 11,533 individuals participate in the CR. Through a central data management and coordinating center (DMCC), the RDCRN's platform for the conduct of observational research encompasses electronic case report forms, federated databases, and an online CR for epidemiological and survey research. An ORDR-governed data repository (through dbGaP, a database for genotype and phenotype information from the National Library of Medicine) has been created. DMCC coordinates with ORDR to register and upload study data to dbGaP for data sharing with the scientific community. The platform provided by the RDCRN DMCC has supported 128 studies, six of which were successfully conducted through the online CR, with 2,352 individuals accrued and a median enrollment time of just 2 months. The RDCRN has built a powerful suite of web-based tools that provide for integration of federated and online database support that can accommodate a large number of rare diseases on a global scale. RDCRN studies have made important advances in the diagnosis and treatment of rare diseases.
Journal of General Internal Medicine 07/2014; 29(3). DOI:10.1007/s11606-014-2894-x · 3.42 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: The Environmental Determinants of Diabetes in the Young (TEDDY) planned biomarker discovery studies on longitudinal samples for persistent confirmed islet cell autoantibodies and type 1 diabetes (T1D) using dietary biomarkers, metabolomics, microbiome/viral metagenomics and gene expression.
This paper describes the details of planning the TEDDY biomarker discovery studies using a nested case-control design that was chosen as an alternative to the full cohort analysis. In the frame of a nested case-control design, it guides the choice of matching factors, selection of controls, preparation of external quality control samples, and reduction of batch effects along with proper sample allocation.
Our design is to reduce potential bias and retain study power while reduce the costs by limiting the numbers of samples requiring laboratory analyses. It also covers two primary end points (the occurrence of diabetes-related autoantibodies and the diagnosis of T1D). The resulting list of case-control matched samples for each laboratory was augmented with external quality control (QC) samples. This article is protected by copyright. All rights reserved.
Diabetes/Metabolism Research and Reviews 07/2014; 30(5). DOI:10.1002/dmrr.2510 · 3.59 Impact Factor