Scientific
publications

The regression discontinuity (RD) design is a quasi-experimental design that estimates the causal effects of a treatment by exploiting naturally occurring treatment rules. It can be applied in any context where a particular treatment or intervention is administered according to a pre-specified rule linked to a continuous variable. Such thresholds are common in primary care drug prescription where the RD design can be used to estimate the causal effect of medication in the general population. Such results can then be contrasted to those obtained from randomised controlled trials (RCTs) and inform prescription policy and guidelines based on a more realistic and less expensive context. In this paper, we focus on statins, a class of cholesterol-lowering drugs, however, the methodology can be applied to many other drugs provided these are prescribed in accordance to pre-determined guidelines. Current guidelines in the UK state that statins should be prescribed to patients with 10-year cardiovascular disease risk scores in excess of 20%. If we consider patients whose risk scores are close to the 20%  risk score threshold, we find that there is an element of random variation in both the risk score itself and its measurement. We can therefore consider the threshold as a randomising device that assigns statin prescription to individuals just above the threshold and withholds it from those just below. Thus, we are effectively replicating the conditions of an RCT in the area around the threshold, removing or at least mitigating confounding. We frame the RD design in the language of conditional independence, which clarifies the assumptions necessary to apply an RD design to data, and which makes the links with instrumental variables clear. We also have context-specific knowledge about the expected sizes of the effects of statin prescription and are thus able to incorporate this into Bayesian models by formulating informative priors on our causal parameters.

BACKGROUND: Allopurinol is the most commonly used urate-lowering therapy, with rare but potentially fatal adverse effects. However, its impact on overall mortality remains largely unknown. In this study, we evaluated the impact of allopurinol initiation on the risk of mortality among individuals with hyperuricaemia and among those with gout in the general population. METHODS: We conducted an incident user cohort study with propensity score matching using a UK general population database. The study population included individuals aged ≥40 years who had a record of hyperuricaemia (serum urate level >357 μmol/L for women and >416 μmol/L for men) between January 2000 and May 2010. To closely account for potential confounders of allopurinol use and risk of death, we constructed propensity score matched cohorts of allopurinol initiators and comparators (non-initiators) within 6-month cohort accrual blocks. RESULTS: Of 5927 allopurinol initiators and 5927 matched comparators, 654 and 718, respectively, died during the follow-up (mean=2.9 years). The baseline characteristics were well balanced in the two groups, including the prevalence of gout in each group (84%). Allopurinol initiation was associated with a lower risk of all-cause mortality (matched HR 0.89 (95% CI 0.80 to 0.99)). When we limited the analysis to those with gout, the corresponding HR was 0.81 (95% CI 0.70 to 0.92). CONCLUSIONS: In this general population study, allopurinol initiation was associated with a modestly reduced risk of death in patients with hyperuricaemia and patients with gout. The overall benefit of allopurinol on survival may outweigh the impact of rare serious adverse effects.

OBJECTIVE: Gut microbiota influence metabolic pathways related to the pathogenesis of obesity, insulin-resistance and diabetes. Antibiotic therapy can alter the microbiota, and is commonly used in western countries. We sought to evaluate whether past antibiotic exposure increases diabetes risk. RESEARCH DESIGN AND METHODS: We conducted a nested case-control study using a large population-based database from the UK. The cases were defined as those with incident diagnosis of diabetes. For every case, four eligible controls matched on age, sex, practice-site, and duration of follow-up before index-date were selected using incidence-density sampling. Exposure of interest was antibiotic therapy >1 year before index-date. Odds ratios (ORs) and 95% CIs were estimated using conditional logistic regression. The risk was adjusted for BMI, smoking, last glucose level, and number of infections before index-date, as well as past medical history of coronary artery disease and hyperlipidaemia. RESULTS: The study included 208 002 diabetic cases and 815 576 matched controls. Exposure to a single antibiotic prescription was not associated with higher adjusted diabetes risk. Treatment with two to five antibiotic courses was associated with increase in diabetic risk for penicillin, cephalosporins, macrolides and quinolones with adjusted OR ranging from 1.08 (95% CI 1.05-1.11) for penicillin to 1.15 (95% CI 1.08-1.23) for quinolones. The risk increased with the number of antibiotic courses and reached 1.37 (95% CI 1.19-1.58) for more than 5 courses of quinolones. There was no association between exposure to anti-virals and anti-fungals and diabetes risk. CONCLUSIONS: Exposure to certain antibiotic groups increases diabetes risk.

BACKGROUND: Smoking increases the risk of developing type 2 diabetes. However, several population studies also show a higher risk in people 3-5 years after smoking cessation than in continuing smokers. After 10-12 years the risk equates to that of never-smokers. Small cohort studies suggest diabetes control deteriorates temporarily during the first year after quitting. We examined whether or not quitting smoking was associated with altered diabetes control in a population study, for how long this association persisted, and whether or not this association was mediated by weight change. METHODS: We did a retrospective cohort study (Jan 1, 2005, to Dec 31, 2010) of adult smokers with type 2 diabetes using The Health Improvement Network (THIN), a large UK primary care database. We developed adjusted multilevel regression models to investigate the association between a quit event, smoking abstinence duration, change in HbA1c, and the mediating effect of weight change. FINDINGS: 10 692 adult smokers with type 2 diabetes were included. 3131 (29%) quit smoking and remained abstinent for at least 1 year. After adjustment for potential confounders, HbA1c increased by 0·21% (95% CI 0·17-0·25; p<0·001; [2·34 mmol/mol (95% CI 1·91-2·77)]) within the first year after quitting. HbA1c decreased as abstinence continued and became comparable to that of continual smokers after 3 years. This increase in HbA1c was not mediated by weight change. INTERPRETATION: In type 2 diabetes, smoking cessation is associated with deterioration in glycaemic control that lasts for 3 years and is unrelated to weight gain. At a population level, this temporary rise could increase microvascular complications. FUNDING: National Institute for Health Research School for Primary Care Research.

INTRODUCTION: The aim of this study was to assess the efficacy of co-administering sitagliptin to patients with inadequate glycemic control following treatment with metformin (MET), sulfonylurea (SU), or MET + SU. METHODS: A cohort of 25,386 patients with type 2 diabetes mellitus (hemoglobin A1c [HbA1C] >53 mmol/mol or 7%), newly treated with sitagliptin between 2007 and 2013, was sourced from UK general practices via The Health Improvement Network database. Among these, eligible patients were segregated into three groups: MET (n = 3364), SU (n = 509), or MET + SU therapy (n = 5929). The relative efficacy of sitagliptin added to SU or MET + SU compared with sitagliptin added to MET monotherapy was assessed with regards to HbA1c and body weight changes from baseline up to 52 weeks. The glycemic efficacy was a measure of average treatment effects obtained from multivariable linear regression models and propensity score-matching analysis. RESULTS: A total of 9802 patients were included in the study. Overall, addition of sitagliptin 100 mg once daily resulted in 5.5 mmol/mol (0.5%) HbA1c reduction (P < 0.001) and 0.8 kg weight reduction at 1 year (P < 0.001). Efficacy was similar across the treatment groups, but in patients with baseline HbA1c ≥9% adding sitagliptin to MET + SU produced a significantly smaller reduction in HbA1c when compared to the reference group MET (MET + SU vs. MET only: -0.5% vs. -0.7%, P < 0.001). The mean HbA1c reduction from baseline within this subgroup of patients was not significantly different between SU and MET monotherapies (-0.8% vs. -0.7%, respectively, P = 0.4). Across treatment groups, HbA1c reductions with add-on sitagliptin occurred after 24 weeks of treatment with a peak reduction occurring between 36 and 48 weeks, and receded after week 48. CONCLUSION: In a real-world general practice setting, sitagliptin was effective in patients with suboptimal glycemic control with MET, SU or dual therapy, maximum between 36 and 48 weeks, but in patients with HbA1c of >9% receiving MET + SU therapy, adding sitagliptin, as a third agent, conferred minimal benefit.

OBJECTIVE: The study described initial pharmacological treatment of children in the United Kingdom diagnosed as having ADHD and assessed predictors of medication persistence. METHODS: U.K. children ages 3-16 diagnosed as having ADHD between 1994 and 2006 were identified from primary care practice data. Child characteristics, prescription patterns, and initial medication prescribed were described over the study period. The associations of child and clinical factors with medication persistence (defined as initial treatment length greater than six months) were estimated by using binomial regression. RESULTS: Of 2,878 children with an ADHD diagnosis, 46% (N=1,314) received at least one prescription for ADHD medication within two years of diagnosis. The mean initial treatment length was 10.7±.5 months. Only 35% (N=464) of pharmacologically treated children had a treatment length greater than six months after initial medication prescription when the analysis used a 30-day grace period; 57% were persistent in treatment when a less stringent 60-day grace period was used. Children who were initially prescribed long-acting methylphenidate were more likely to persist in treatment than those prescribed standard methylphenidate (risk ratio=1.2, 95% confidence interval=1.1-1.4). CONCLUSIONS: A large proportion of children who received medication for ADHD in primary care did not continue in initial treatment for more than six months. Few child or clinical factors were associated with treatment persistence. Epidemiological research about the effects of long-term ADHD medication use should account for the observed limited persistence in medication treatment.

BACKGROUND: An algorithm that detects errors in diagnosis, classification or coding of diabetes in primary care computerised medial record (CMR) systems is currently available. However, this was developed on CMR systems that are episode orientated medical records (EOMR); and do not force the user to always code a problem or link data to an existing one. More strictly problem orientated medical record (POMR) systems mandate recording a problem and linking consultation data to them. OBJECTIVE: To compare the rates of detection of diagnostic accuracy using an algorithm developed in EOMR with a new POMR specific algorithm. METHOD: We used data from The Health Improvement Network (THIN) database (N = 2,466,364) to identify a population of 100,513 (4.08%) patients considered likely to have diabetes. We recalibrated algorithms designed to classify cases of diabetes to take account of that POMR enforced coding consistency in the computerised medical record systems [In Practice Systems (InPS) Vision] that contribute data to THIN. We explored the different proportions of people classified as having type 1 diabetes mellitus (T1DM) or type 2 diabetes mellitus (T2DM) and with diabetes unclassifiable as either T1DM or T2DM. We compared proportions using chi-square tests and used Tukey's test to compare the characteristics of the people in each group. RESULTS: The prevalence of T1DM using the original EOMR algorithm was 0.38% (9,264/2,466,364), and for T2DM 3.22% (79,417/2,466,364). The prevalence using the new POMR algorithm was 0.31% (7,750/2,466,364) T1DM and 3.65% (89,990/2,466,364) T2DM. The EOMR algorithms also left more people unclassified 11,439 (12%), as to their type of diabetes compared with 2,380 (2.4%), for the new algorithm. Those people who were only classified by the EOMR system differed in terms of older age, and apparently better glycaemic control, despite not being prescribed medication for their diabetes (p < 0.005). CONCLUSION: Increasing the degree of problem orientation of the medical record system can improve the accuracy of recording of diagnoses and, therefore, the accuracy of using routinely collected data from CMRs to determine the prevalence of diabetes mellitus; data processing strategies should reflect the degree of problem orientation.

BACKGROUND: Serological studies indicate that evidence of coeliac disease (CD) exists in about 1% of all children, but we lack estimates of current diagnostic patterns among children and how they vary by socioeconomic group. METHODS: We identified all children aged 0-18 years between 1993 and 2012 who were registered with general practices across the UK that contribute to a large population-based general practice database. The incidence of CD was evaluated in each quintile of the Townsend index of deprivation and stratified by age, sex, country and calendar year. RESULTS: Among 2,063,421 children, we identified 1247 CD diagnoses, corresponding to an overall CD incidence of 11.9 per 100,000 person-years, which was similar across the UK countries and higher in girls than in boys. We found a gradient of CD diagnosis across socioeconomic groups, with the rate of diagnosis being 80% higher in children from the least-deprived areas than in those from the most-deprived areas (incident rate ratio 1.80, 95% CI 1.45 to 2.22). This pattern held for both boys and girls and across all ages. Across all four countries of the UK, we found similar associations between CD and socioeconomic status. While CD incidence up to age 2 remained stable over the study period, diagnoses at older ages have almost tripled over the past 20 years. CONCLUSIONS: Children living in less socioeconomically deprived areas in the UK are more likely to be diagnosed with CD. Increased implementation of diagnostic guidelines could result in better case identification in more-deprived areas.

AIM: To investigate risk factors for first long-bone fractures in children up to 5 years old in order to provide evidence about which families could benefit from injury prevention interventions. METHODS: Population-based matched nested case-control study using The Health Improvement Network, a UK primary care research database, 1988-2004. Maternal, household and child risk factors for injury were assessed among 2456 children with long-bone fractures (cases). 23,661 controls were matched to cases on general practice. Adjusted ORs and 95% CIs were estimated using conditional logistic regression. RESULTS: Fractures of long-bones were independently associated with younger maternal age and higher birth order, with children who were the fourth-born in the family, or later, having a threefold greater odds of fracture compared to first-born children (adjusted OR 3.12, 95% CI 2.08 to 4.68). Children over the age of 1 year had a fourfold (13-24 months, adjusted OR 4.09 95% CI 3.51 to 4.76) to fivefold (37+ months, adjusted OR 4.88 95% CI 4.21 to 5.66) increase in the odds of a long-bone fracture compared to children aged 0-12 months. Children in families with a history of maternal alcohol misuse had a raised odds of long-bone fracture (adjusted OR 2.33, 95% CI 1.13 to 4.82) compared to those with no documented history. CONCLUSIONS: Risk factors for long-bone fractures in children less than 5 years old included age above 1 year, increasing birth order, younger maternal age and maternal alcohol misuse. These risk factors should be used to prioritise families and communities for injury prevention interventions.

Using data from The Health Improvement Network and Hospital Episode Statistics, we investigate smoking prevalence, number of smokers treated and opportunities for cessation intervention among patients treated in National Health Service (NHS) hospitals in England from April 2010 to March 2011. Our results show that approximately 1.1 million smokers are treated in English hospitals each year, receiving a total of 2.6 million episodes of care. These findings suggest that delivering smoking cessation as a routine component of hospital care, as recommended by recent National Institute for Health and Care Excellence guidance, could achieve marked reductions in the prevalence of smoking and improve the cost-effectiveness of NHS hospitals.

PURPOSE: To explore whether and how longitudinal medical records could be used as a source of reference in the early phases of signal detection and analysis of novel adverse drug reactions (ADRs) in a global pharmacovigilance database. METHODS: Drug and ADR combinations from the routine signal detection process of VigiBase® in 2011 were matched to combinations in The Health Improvement Network (THIN). The number and type of drugs and ADRs from the data sets were investigated. For unlabelled combinations, graphical display of longitudinal event patterns (chronographs) in THIN was inspected to determine if the pattern supported the VigiBase combination. RESULTS: Of 458 combinations in the VigiBase data set, 190 matched to corresponding combinations in THIN (after excluding drugs with less than 100 prescriptions in THIN). Eighteen percent of the VigiBase and 9% of the matched THIN combinations referred to new drugs reported with serious reactions. Of the 112 unlabelled combinations matched to THIN, 52 chronographs were inconclusive mainly because of lack of data; 34 lacked any outstanding pattern around the time of prescription; 24 had an elevation of events in the pre-prescription period, hence weakened the suspicion of a drug relationship; two had an elevated pattern of events exclusively in the post-prescription period that, after review of individual patient histories, did not support an association. CONCLUSIONS: Longitudinal medical records were useful in understanding the clinical context around a drug and suspected ADR combination and the probability of a causal relationship. A drawback was the paucity of data for newly marketed drugs with serious reactions.

OBJECTIVES: Studies report that the risks of significant neurologic complications (including stroke, cerebral abscess, and migraine) and hemorrhagic sequelae are high in patients with hereditary hemorrhagic telangiectasia (HHT), and that life expectancy in this cohort is reduced. However, most published cohorts derive from specialist centers, which may be susceptible to bias. METHODS: We used a population-based approach to estimate the risks of developing neurologic and hemorrhagic complications of HHT, the association of a diagnosis of HHT with common cardiovascular and malignant comorbidities, and also long-term survival of those with the disease. RESULTS: From a UK primary care database of 3.5 million patients (The Health Improvement Network), we identified 675 cases with a diagnosis of HHT and compared them with 6,696 controls matched by age, sex, and primary care practice. Risks of stroke (odds ratio [OR] 1.8, 95% confidence interval [CI] 1.2-2.6), cerebral abscess (OR 30.0, CI 3.1-288), and migraine (OR 1.7, CI 1.3-2.2) were elevated over controls. Bleeding complications including epistaxis (OR 11.6, CI 9.1-14.7) and gastrointestinal hemorrhage (OR 6.1, CI 2.8-13.4) were more common in cases with HHT. Survival of cases with HHT was poorer than controls with a hazard ratio for death of 2.0 (CI 1.6-2.6) and a median age at death 3 years younger. CONCLUSIONS: Patients with HHT are at substantially increased risk of serious neurologic and hemorrhagic complications of the disease. Because a diagnosis of HHT is associated with a significantly poorer survival compared with those who have no disease, evaluation of new strategies to improve clinical management is required.

CONTEXT: Evidence suggests that the human skeleton might be involved in the regulation of glucose homeostasis. OBJECTIVE: The objective of the study was to investigate the effect of exposure to bisphosphonates on the risk of incident type 2 diabetes mellitus (T2DM). DESIGN: This was a population-based, retrospective, open cohort study over the period 1995-2010. SETTING: The study was conducted from The Health Improvement Network database from the United Kingdom in a primary care setting. PATIENTS: A total of 35 998 individuals aged older than 60 years, without diabetes at baseline and with more than 1 year's exposure to bisphosphonates, and 126 459 age-, gender-, body mass index- and general practice-matched unexposed individuals participated in the study. INTERVENTIONS: There were no interventions. MAIN OUTCOME MEASURE: A new diagnosis of T2DM during the 16-year-long observation period, determined by Read codes and adjusted incidence rate ratio in bisphosphonate-exposed compared with unexposed groups, was the main outcome measure. RESULTS: The risk of incident T2DM was significantly lower in patients exposed to bisphosphonates compared with matched controls [adjusted incidence rate ratio 0.52, 95% confidence interval (CI) 0.48-0.56, P < .0001]. In subgroup analyses, the findings remained consistent in males [0.77 (95% CI 0.66-0.89)], females [0.49 (95% CI 0.45-0.53)], obese [0.54 (95% CI 0.50-0.59)], individuals exposed to steroid treatment [0.47 (95% CI 0.34-0.64)], and over different types of bisphosphonate medication. Analysis of duration of treatment suggested a brief increase in the risk of T2DM (1 to 2.5 y of exposure), followed by a progressive, sustained decrease as the years of exposure accumulated. CONCLUSIONS: This observational evidence suggests exposure to bisphosphonates was associated with a significant 50% reduction in the risk of incident T2DM.