Record linkage to enhance consented cohort and routinely collected health data from a UK birth cohort.

BACKGROUND: In longitudinal health research, combining the richness of cohort data to the extensiveness of routine data opens up new possibilities, providing information not available from one data source alone. In this study, we set out to extend information from a longitudinal birth cohort study by linking to the cohort child's routine primary and secondary health care data. The resulting linked datasets will be used to examine health outcomes and patterns of health service utilisation for a set of common childhood health problems. We describe the experiences and challenges of acquiring and linking electronic health records for participants in a national longitudinal study, the UK Millennium Cohort Study (MCS). METHOD: Written parental consent to link routine health data to survey responses of the MCS cohort member, mother and her partner was obtained for 90.7% of respondents when interviews took place at age seven years in the MCS. Probabilistic and deterministic linkage was used to link MCS cohort members to multiple routinely-collected health data sources in Wales and Scotland. RESULTS: Overall linkage rates for the consented population using country-specific health service data sources were 97.6% for Scotland and 99.9% for Wales. Linkage rates between different health data sources ranged from 65.3% to 99.6%. Issues relating to acquisition and linkage of data sources are discussed. CONCLUSIONS: Linking longitudinal cohort participants with routine data sources is becoming increasingly popular in population data research. Our results suggest that this is a valid method to enhance information held in both sources of data.

Early academic achievement in children with isolated clefts: a population-based study in England.

OBJECTIVES: We used national data to study differences in academic achievement between 5-year-old children with an isolated oral cleft and the general population. We also assessed differences by cleft type. METHODS: Children born in England with an oral cleft were identified in a national cleft registry. Their records were linked to databases of hospital admissions (to identify additional anomalies) and educational outcomes. Z-scores (signed number of SD actual score is above national average) were calculated to make outcome scores comparable across school years and across six assessed areas (personal development, communication and language, maths, knowledge of world, physical development andcreative development). RESULTS: 2802 children without additional anomalies, 5 years old between 2006 and 2012, were included. Academic achievement was significantly below national average for all six assessed areas with z-scores ranging from -0.24 (95% CI -0.32 to -0.16) for knowledge of world to -0.31 (-0.38 to -0.23) for personal development. Differences were small with only a cleft lip but considerably larger with clefts involving the palate. 29.4% of children were documented as having special education needs (national rate 9.7%), which varied according to cleft type from 13.2% with cleft lip to 47.6% with bilateral cleft lip and palate. CONCLUSIONS: Compared with national average, 5-year-old children with an isolated oral cleft, especially those involving the palate, have significantly poorer academic achievement across all areas of learning. These outcomes reflect results of modern surgical techniques and multidisciplinary approach. Children with a cleft may benefit from extra academic support when starting school.

Concomitant use of ibuprofen and paracetamol and the risk of major clinical safety outcomes.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT: Non-steroidal anti-inflammatory drugs (NSAIDs) and paracetamol are widely used analgesics in the prescription and non-prescription settings. Although both classes of drug are generally well tolerated, they can lead to well-characterized adverse effects. Both drugs are widely co-prescribed and it is of interest to understand better safety outcomes when the two drugs are taken concomitantly. WHAT THIS STUDY ADDS?: Relative rates and hazard ratio patterns of safety outcomes were broadly similar for patients prescribed ibuprofen alone, paracetamol alone and concomitant ibuprofen and paracetamol. The risks of the various safety outcomes examined do not appear to be modified by concomitant use of ibuprofen and paracetamol compared with paracetamol or ibuprofen alone. AIMS: To evaluate and compare the risk of specific safety outcomes in patients prescribed ibuprofen and paracetamol concomitantly with those in patients prescribed ibuprofen or paracetamol alone. The outcomes were evaluated according to dose, duration and exposure. METHODS: The study used a retrospective longitudinal cohort design with data from the UK General Practice Research Database (GPRD). The study population included patients aged 18 years or over who were prescribed ibuprofen alone, paracetamol alone or concomitant ibuprofen and paracetamol (tablets or capsules only). The safety outcomes evaluated were upper gastrointestinal events, myocardial infarction, stroke, renal failure (excluding chronic), congestive heart failure, intentional or accidental overdose, suicidal behaviour and mortality. Time-dependent Cox regression was used to estimate relative rates for the safety outcomes, by treatment group. A further analysis evaluated whether the hazard rates (i.e. absolute risks) varied over time with changes in drug exposure. RESULTS: The study population included 1.2 million patients. There was considerable heterogeneity in both patient and exposure characteristics. When comparing with past users, for most safety outcomes, current users of concomitant paracetamol and ibuprofen had relative rates between those for current users of ibuprofen alone and paracetamol alone. The hazard rates were generally proportional over time, from current to past exposure, following a prescription for concomitant paracetamol and ibuprofen compared with ibuprofen alone or paracetamol alone. CONCLUSIONS: The known risk of the safety outcomes examined does not appear to be modified by concomitant use of ibuprofen and paracetamol compared with paracetamol or ibuprofen alone.

The pattern of risk of myocardial infarction in patients taking asthma medication: a study with the General Practice Research Database.

AIM: To describe the patterns of risks of acute myocardial infarction (MI) during exposure to long-acting beta2-agonists (LABA). METHODS: The study population consisted of patients aged 18+ years prescribed LABA or short-acting beta2-agonists (SABA) in the UK General Practice Research Database (GPRD). The outcomes included acute MI as recorded in GPRD and hospitalization for acute MI as obtained from the national registry of hospital admissions in England. The patterns of the hazard rates over time (i.e. absolute risks) were evaluated. RESULTS: The study population included 507,966 patients, who received a total of 5.5 million inhaled SABA, 4.0 million inhaled corticosteroids (ICS) and 1.3 million LABA prescriptions. In patients who recently started asthma medication, there were substantial changes in the hazard rates of MI over time: hazard rates were increased shortly following the prescription and then decreased. The hazard rates of MI in GPRD and of MI hospitalizations were proportional over time between inhaled SABA, LABA and ICS. Heavy long-term users (13+ Rx of the same asthma drug in the 1 year before) had increased risks of MI both with inhaled SABA and ICS. The relative rate in the heavy long-term users was 1.6 with inhaled SABA, 1.1 with LABA and 1.7 with ICS. The pattern of risk was similar between LABA with and without concomitant ICS use. CONCLUSION: The patterns of risks of MI were broadly similar between inhaled SABA, LABA and ICS, suggesting that there were no major differences between these drugs.

A genome-wide DNA microsatellite association screen to identify chromosomal regions harboring candidate genes in diabetic nephropathy.

In an effort to accelerate the identification of susceptibility genes in diabetic nephropathy, the first genome-wide fluorescence-based DNA microsatellite (n=6000) association screen was performed, using pools of genomic DNA derived from Irish patients with (cases; n=200) and without (controls; n=200) type 1 diabetic nephropathy. Allele image profiles were generated for 5353 (89.2%) microsatellite markers for both case and control pools. Allele counts (estimated from allele image profiles) were compared in case versus control groups, and empirical P values were generated. Markers then were ranked on the basis of their empirical P values (lowest to highest). Repeat PCR amplification and electrophoresis of pooled samples were performed systematically on ranked markers until the 50 most associated markers with consistent results were identified. DNA samples that composed the pools then were genotyped individually for these markers. Two markers on chromosome 10, D10S558 (Pcorrected=0.005) and D10S1435 (Pcorrected=0.016), revealed statistically significant associations with diabetic nephropathy. An additional four markers (D6S281, D4S2937, D2S291, and D17S515) also are worthy of further investigation. Relevant functional candidate genes have been identified in the vicinity of these markers, demonstrating the feasibility of low-resolution genome-wide microsatellite association screening to identify possible candidate genes for diabetic nephropathy.

Logistic regression protects against population structure in genetic association studies.

We conduct an extensive simulation study to compare the merits of several methods for using null (unlinked) markers to protect against false positives due to cryptic substructure in population-based genetic association studies. The more sophisticated "structured association" methods perform well but are computationally demanding and rely on estimating the correct number of subpopulations. The simple and fast "genomic control" approach can lose power in certain scenarios. We find that procedures based on logistic regression that are flexible, computationally fast, and easy to implement also provide good protection against the effects of cryptic substructure, even though they do not explicitly model the population structure.

A genome wide linkage disequilibrium screen in Parkinson's disease.

Whole genome screening is increasingly used to identify genetic risk factors for complex diseases. In this study, a genome wide linkage disequilibrium (LD) screen was performed in a cohort of Parkinson's disease (PD) patients from the UK (n = 195) using pooled DNA to facilitate efficient genotyping of 5546 microsatellite markers. Allele frequencies were compared with those found in 2 previously typed disease free control populations, and the most interesting markers were selected for multiple repeat testing among the 3 pools. Markers were then individually genotyped in our original PD cohort and one of the original control groups, and independently in a second cohort of UK PD patients (n = 179), and additional controls. Using this 2-stage approach, we have been unable to find evidence for consistent association of any markers with sporadic PD. Subgroup analysis of the most promising marker shows some evidence that microsatellite marker D1S2886 is associated with familial forms of the disease.

Refining the analysis of a whole genome linkage disequilibrium association map: the United Kingdom results.

Individual genotyping of the 10 most promising markers identified in our previously reported screen for linkage disequilibrium (LD) in multiple sclerosis identified a number of effects which confound the analysis and are of general importance in the interpretation of results obtained using microsatellite markers typed in pooled DNA. In order to identify and characterise these effects, we individually genotyped 529 promising markers in 16 trio families. We then devised adapting factors, which were designed to correct for these confounders. This more extensive analysis of the previously published UK data set and the repeat analyses incorporating these adaptations led to the identification of two novel markers that may be associated with multiple sclerosis in this population, providing a close correlation between the results of pooled analysis and individual typing.

New candidate loci for multiple sclerosis susceptibility revealed by a whole genome association screen in a Belgian population.

We have completed a whole genome screen for association with multiple sclerosis (MS) in a Belgian population. The 6000 microsatellite markers provided through the Genetic Association of Multiple Sclerosis in EuropeanS (GAMES) collaborative were genotyped in case-control and family-based samples. The 20 most promising markers included three markers (D6S1615, D6S2444 and TNFa) from the classically established HLA class II cluster and one (D6S265) from the recently re-emphasized HLA class I cluster. In other highlighted regions, preliminary candidate genes from the immune system have been identified: e.g. the integrin ligand EDIL3, the high-mobility group box protein TOX, neutral sphingomyelinase activating factor (NSMAF) and the B-cell specific transcription factor POU2AF1.

A genome-wide German screen for linkage disequilibrium in multiple sclerosis.

We report on a genome-wide screen for association with multiple sclerosis (MS) in the German population performed using 6000 microsatellite markers. These markers were typed in four DNA pools consisting of 234 MS patients (cases), 209 unrelated controls, 68 index patients from trio families and their 136 parents (related controls). Stringent analysis identified 11 markers showing apparent evidence for association. Five from regions previously identified in linkage studies and two from the MHC region on chromosome 6p21. These MHC markers are known to be in linkage disequilibrium with HLA class II alleles influencing susceptibility to MS. The identification of these markers serves as an important positive control.

A genome-wide screen for association in Hungarian multiple sclerosis.

Although the pathogenesis of multiple sclerosis (MS) is not fully understood, substantial evidence points to the involvement of genetic factors. We report on a genome-wide screen for disease association in the Hungarian population using 5532 microsatellite markers. These markers were typed in DNA pools that consisted of 88 MS patients (cases), and 128 unrelated controls. Based on a stringent selection criterion, we obtained 33 markers suggesting association with the disease.

A genome wide scan for association with multiple sclerosis in a N. Irish case control population.

In order to screen the genome for linkage disequilibrium (LD) in multiple sclerosis (MS), we typed 2537 microsatellite markers in separately pooled DNA from 200 cases and 200 controls from N. Ireland. Twenty two markers showing significant evidence of association were identified including three from the HLA region on chromosome 6p21. Putative candidate genes mapping close to the 19 novel markers include the IL10RA and CD3E genes on 11q23 (which both lie close to the marker D11S1998). Individual typing of the marker D11S1998 confirmed its association.

A whole genome screen for linkage disequilibrium in multiple sclerosis performed in a continental Italian population.

We have systematically screened the genome for evidence of linkage disequilibrium (LD) with multiple sclerosis (MS) by typing 6000 microsatellite markers in case-control and family based (AFBAC) cohorts from the Italian population. DNA pooling was used to reduce the genotyping effort involved. Four DNA pools were considered: cases (224 Italian MS patients), controls (231 healthy Italians), index (185 index cases from trio families) and parents (the 370 parents of the patient included in the Index pool), respectively. After refining analysis of the most promising 14 markers to emerge from this screening process, only marker D2S367 retained evidence for association.

Two genome-wide linkage disequilibrium screens in Scandinavian multiple sclerosis patients.

We report the first two genome-wide screens for linkage disequilibrium between putative multiple sclerosis (MS) susceptibility genes and genetic markers performed in the genetically homogenous Scandinavian population, using 6000 microsatellite markers and DNA pools of approximately 200 MS cases and 200 controls in each screen. Usable data were achieved from the same 3331 markers in both screens. Nine markers from eight genomic regions (1p33, 3q13, 6p21, 6q14, 7p22, 9p21, 9q21 and Xq22) were identified as potentially associated with MS in both screens.

A whole genome screen for association in Polish multiple sclerosis patients.

We have performed the first systematic search for MS susceptibility genes completed in the Polish population. This screen was performed using 6000 microsatellite markers typed in pooled DNA from cases (n=200), controls (n=200) and trio families (n=129). Five associated markers are identified, one (D6S2444) from the HLA region and four are from novel regions not previously associated with MS, 2p16 (D2S2153), 3p13 (D3S3568), 7p22 (D7S2521) and 15q26 (D15S649).

A genome-wide screen for linkage disequilibrium in Sardinian multiple sclerosis.

Using indirect whole genome association screening, we have searched for multiple sclerosis susceptibility genes in the genetically isolated high risk Sardinian population. Two screens were performed; the first was based on 229 cases and 264 unrelated controls, and the second on 235 trio families. Each screen employed a dense set of microsatellite markers and DNA pooling. Data from both screens were available from 2764 markers. Nine markers showed nominally significant results in both screens independently. Five of these markers-D2S408 (2q36), D6S271 (6p21), D6S344 (6p25), D7S1818 (7p12) and D16S420 (16p12)-remained nominally significant in both studies after conservative refining analysis.

A genomic screen of Spanish multiple sclerosis patients reveals multiple loci associated with the disease.

In order to identify the genomic regions that might confer susceptibility to multiple sclerosis (MS) in the Spanish population, we have performed a genome-wide screen for association in patients with MS using pooled DNA from 200 clinical cases and 200 healthy controls. The pools were typed using 5546 microsatellites. The typing was repeated for the most promising 1269 markers after which 191 potentially associated markers were identified. Eleven of these markers map to the MHC region, and 14 to non-MHC regions identified in previous linkage screens. Our results provide support for the presence of multiple coding regions that contain MS susceptibility genes of small or moderate effect.

A whole genome screen for linkage disequilibrium in multiple sclerosis confirms disease associations with regions previously linked to susceptibility.

Linkage analysis in multiplex families has provisionally identified several genomic regions where genes influencing susceptibility to multiple sclerosis are likely to be located. It is anticipated that association mapping will provide a higher degree of resolution, but this more powerful approach is limited by the substantial genotyping effort required. Here, we describe the first use of DNA pooling to screen the whole genome for association in multiple sclerosis based on a 0.5 cM map of microsatellite markers and using four DNA pools derived from cases (n = 216), controls (n = 219) and trio families (n = 745 affected individuals and their 1490 parents). The 10 markers showing the greatest evidence for association with multiple sclerosis that emerge from this analysis include three from the HLA region on chromosome 6p (D6S1615, D6S2444 and TNFa), providing a positive control for the method, four from regions previously identified by linkage analysis in UK multiplex families (two mapping to chromosome 17q GCT6E11 and D17S1535; one to chromosome 1p GGAA30B06; and one to 19q D19S585), and three from novel sites with respect to linkage analysis (D1S1590 at 1q; D2S2739 at 2p; and D4S416 at 4q). Our results thus provide further supporting evidence for the candidature of 6p, 17q, 19q and 1p as regions encoding susceptibililty genes for multiple sclerosis. The protocol used in this UK-based study is now being extended to 18 additional sites in Europe in order to search for susceptibility genes shared between populations of common ancestry, as well as those that exert ethnically more restricted effects.