Background Although human leukocyte antigen (HLA) DQ and
DR loci appear to confer the strongest genetic risk for
type 1 diabetes, more detailed information is required for other loci within the
HLA region to understand causality and stratify additional risk factors. The
Type 1 Diabetes Genetics Consortium (T1DGC) study design included
high-resolution genotyping of HLA-A, B,
C, DRB1, DQ, and
DP loci in all affected sibling pair and trio families, and
cases and controls, recruited from four networks worldwide, for analysis with
clinical phenotypes and immunological markers.Purpose In this article, we present the operational strategy of training,
classification, reporting, and quality control of HLA genotyping in four
laboratories on three continents over nearly 5 years.Methods Methods to standardize HLA genotyping at eight loci included: central
training and initial certification testing; the use of uniform reagents,
protocols, instrumentation, and software versions; an automated data transfer;
and the use of standardized nomenclature and allele databases. We implemented a
rigorous and consistent quality control process, reinforced by repeated
workshops, yearly meetings, and telephone conferences.Results A total of 15,246 samples have been HLA genotyped at eight loci to
four-digit resolution; an additional 6797 samples have been HLA genotyped at two
loci. The genotyping repeat rate decreased significantly over time, with an
estimated unresolved Mendelian inconsistency rate of 0.21%. Annual
quality control exercises tested 2192 genotypes (4384 alleles) and achieved
99.82% intra-laboratory and 99.68% inter-laboratory
concordances.Limitations The chosen genotyping platform was unable to distinguish many allele
combinations, which would require further multiple stepwise testing to resolve.
For these combinations, a standard allele assignment was agreed upon, allowing
further analysis if required.Conclusions High-resolution HLA genotyping can be performed in multiple laboratories
using standard equipment, reagents, protocols, software, and communication to
produce consistent and reproducible data with minimal systematic error. Many of
the strategies used in this study are generally applicable to other large
multi-center studies.