Search

Brendan Hallam,1 Irene Petersen,1 Claudia Cooper,2 Christina Avgerinou,1 Kate Walters1 1UCL Research Department of Primary Care & Population Health, University College London, London, UK; 2Division of Psychiatry, University College London, London, UKCorrespondence: Brendan Hallam, UCL Research Department of Primary Care & Population Health, University College London, London, UK, Email Brendan.hallam.18@ucl.ac.ukPurpose: To investigate time trends in incidence of recorded memory concerns (MC) and cognitive decline (CD) in a UK older population presenting to primary care with no prior diagnosis of dementia. To determine the risk of developing dementia in people with recorded memory concern and cognitive decline.Patients and methods: We included individuals aged 65– 99 years who contributed to data within the IQVIA medical research database from 1st January 2009 to 31st December 2018. We reported crude incidence rates for MC (study population n=1,310,838) and CD (n=1,348,796). We conducted survival analysis to estimate the risk of developing dementia using fine-grey sub-distribution hazard model with competing risk of death.Results: We identified 55,941 individuals (4.3%) with a record of incident MC; rates were fairly stable over the decade of study. We identified 14,869 people (1.1%) with a record of incident CD, and these rates increased from 1.29/1000 PYAR (95% CI 1.21 to 1.38) in 2009 to 3.49/1000 PYAR (95% CI 3.30 to 3.68) in 2018. Within 3 years of follow up from the first record of MC, 45.5% of individuals received a diagnosis of dementia, while of those with a record of CD, 51.7% received a dementia diagnosis. Women, people in older age groups and those living in more deprived areas were more likely to have a record of MC or CD, and their symptoms were more likely to progress to a dementia diagnosis.Conclusion: Incidence rates of MC and CD estimated from routinely collected primary care data are lower than those reported in community surveys, suggesting that a minority of people who experience memory loss consult their GP and have it recorded. Our findings indicate that those who do report concerns to primary care, especially women, those in older age groups and those in more deprived areas, are at a higher risk for developing dementia.Keywords: primary care, memory, dementia, incidence rate, survival analysis

Irene Petersen,1 Alexander Crozier,2 Iain Buchan,3 Michael J Mina,4,5 Jonathan W Bartlett6 1Department of Primary Care & Population Health, University College London, London, UK; 2Division of Biosciences, University College London, London, UK; 3Institute of Population Health, University of Liverpool, Liverpool, UK; 4Department of Epidemiology, Department of Immunology and Infectious Diseases, Harvard T.H. Chan School of Public Health, Boston, MA, USA; 5Department of Pathology, Clinical Microbiology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA; 6Department of Mathematical Sciences, University of Bath, Bath, UKCorrespondence: Irene Petersen Email i.petersen@ucl.ac.ukABSTRACT: Testing for SARS-CoV-2 internationally has focused on COVID-19 diagnosis among symptomatic individuals using reverse transcriptase polymerase chain reaction (PCR) tests. Recently, however, SARS-CoV-2 antigen rapid lateral flow tests (LFT) have been rolled out in several countries for testing asymptomatic individuals in public health programmes. Validation studies for LFT have been largely cross-sectional, reporting sensitivity, specificity and predictive values of LFT relative to PCR. However, because PCR detects genetic material left behind for a long period when the individual is no longer infectious, these statistics can under-represent the sensitivity of LFT for detecting infectious individuals, especially when sampling asymptomatic populations. LFTs (intended to detect individuals shedding SARS-CoV-2 antigens) validated against PCR (intended to diagnose infection) are not reporting against a gold standard of equivalent measurements. Instead, these validation studies have reported relative performance statistics that need recalibrating to the purpose for which LFT is being used. We present an approach to this recalibration. We derive a formula for recalibrating relative performance statistics from LFT vs PCR validation studies to give likely absolute sensitivity of LFT for detecting individuals who are shedding shedding SARS-CoV-2 antigens. We contrast widely reported apparent sensitivities of LFT with recalibrated absolute sensitivity for detecting individuals shedding SARS-CoV-2 antigens. After accounting for within-individual viral kinetics and epidemic dynamics within asymptomatic populations we show that a highly performant test for SARS-CoV-2 antigen should show LFT-to-PCR relative sensitivity of less than 50% in conventional validation studies, which after re-calibration would be an absolute sensitivity of more than 80%. Further studies are needed to ascertain the absolute sensitivity of LFT as a test of infectiousness in COVID-19 responses. These studies should include longitudinal series of LFT and PCR, ideally in cohorts sampled from both contacts of cases and the general population.Keywords: rapid test, PCR, validation, recalibration, lateral flow tests

Juan Carlos Bazo-Alvarez,1,2 Tim P Morris,3 James R Carpenter,3,4 Irene Petersen1,5 1Research Department of Primary Care and Population Health, University College London (UCL), London, UK; 2School of Medicine, Universidad Cesar Vallejo, Trujillo, Peru; 3MRC Clinical Trials Unit at UCL, London, UK; 4Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, UK; 5Department of Clinical Epidemiology, Aarhus University, Aarhus, DenmarkCorrespondence: Juan Carlos Bazo-AlvarezResearch Department of Primary Care and Population Health, University College London (UCL), Rowland Hill Street, London, NW3 2PF, UKTel +44 7376076260Email juan.alvarez.16@ucl.ac.ukObjective: Missing data can produce biased estimates in interrupted time series (ITS) analyses. We reviewed recent ITS investigations on health topics for determining 1) the data management strategies and statistical analysis performed, 2) how often missing data were considered and, if so, how they were evaluated, reported and handled.Study Design and Setting: This was a scoping review following standard recommendations from the PRISMA Extension for Scoping Reviews. We included a random sample of all ITS studies that assessed any intervention relevant to health care (eg, policies or programmes) with individual-level data, published in 2019, with abstracts indexed on MEDLINE.Results: From 732 studies identified, we finally reviewed 60. Reporting of missing data was rare. Data aggregation, statistical tools for modelling population-level data and complete case analyses were preferred, but these can lead to bias when data are missing at random. Seasonality and other time-dependent confounders were rarely accounted for and, when they were, missing data implications were typically ignored. Very few studies reflected on the consequences of missing data.Conclusion: Handling and reporting of missing data in recent ITS studies performed for health research have many shortcomings compared with best practice.Keywords: interrupted time series analysis, segmented regression, missing data, multiple imputation, scoping review

Albert Prats-Uribe,1 Junqing Xie,1 Daniel Prieto-Alhambra,1 Irene Petersen2,3 1Pharmaco- and Device Epidemiology, Centre for Statistics in Medicine - Nuffield Department of Orthopaedics, Rheumatology, and Musculoskeletal Sciences, University of Oxford, Oxford, OX3 7LD, UK; 2Department of Primary Care and Population Health, UCL, London, NW3 2PF, UK; 3Department of Clinical Epidemiology, Aarhus University, Aarhus N, 8200, DenmarkCorrespondence: Daniel Prieto-AlhambraPharmaco- and Device Epidemiology, Centre for Statistics in Medicine - Nuffield Department of Orthopaedics, Rheumatology, and Musculoskeletal Sciences, University of Oxford, Botnar Research Centre, Windmill Road, Oxford, OX3 7LD, UKEmail daniel.prietoalhambra@ndorms.ox.ac.ukBackground: Several papers have shown contradictory evidence about the relationship between smoking and COVID-19-related deaths. There is little evidence about smoking and risk of infection. We aim to examine association between smoking and COVID-19 infection and subsequent mortality.Methods: This was a prospective study with participants from the UK Biobank cohort. Participants who lived in England were followed up from 01/02/2020 to 28/06/2020 with data linked to hospital episode statistics, Office for National Statistics and Public Health England PCR tests. We compared current-smokers, previous-smokers with never-smokers and estimated risk ratio (RR) of COVID-19 infection and subsequent mortality using Poisson regression adjusting for age, sex, ethnicity, body mass index and socio-economic status. Interactions between smoking status and age and sex were tested for using multiplicative interactions, and analyses were stratified by median age (49– 68 years, 69– 86 years) and sex.Results: In total, 402,978 participants were included in the analyses. The majority were never smokers, 226,294 (56.2%), 140,090 (34.8%) were previous smokers, and 39,974 (9.9%) current smokers. COVID-19 infection was identified in 1591 (0.39%) people, and 372/1591 (23.4%) died. Amongst the younger participants, smokers were nearly twice as likely to become infected with COVID-19 than never smokers (RR 1.88 [1.49– 2.38]) whereas there was no difference for those aged 69+ (RR 1.05 [0.82– 1.34]). In contrast, amongst the older participants, smokers were twice as likely to die from COVID-19 compared to non-smokers (RR 2.15 [1.11– 4.16]) whereas there was no difference for those under the age of 69 (RR 1.22[0.83– 1.79]). Similar patterns were observed for previous smokers. The impact of smoking was similar in men and women.Conclusion: The association between smoking and COVID-19 infection and subsequent death is modified by age. Smokers and previous smokers aged under 69 were at higher risk of COVID-19 infection, suggesting the risk is associated with increased exposure to SARS-COV-2 virus. Once infected, older smokers were twice as likely to die from COVID-19 than never smokers, possibly mediated by increased risk of chronic conditions/illnesses.Keywords: smoking, COVID-19, UK Biobank

Juan Carlos Bazo-Alvarez,1,2 Tim P Morris,3 Tra My Pham,3 James R Carpenter,3,4 Irene Petersen1,5 1Research Department of Primary Care and Population Health, University College London (UCL), London, UK; 2Instituto de Investigación, Universidad Católica Los Ángeles de Chimbote, Chimbote, Peru; 3MRC Clinical Trials Unit at UCL, London, UK; 4Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, UK; 5Department of Clinical Epidemiology, Aarhus University, Aarhus, DenmarkCorrespondence: Juan Carlos Bazo-AlvarezResearch Department of Primary Care and Population Health, University College London (UCL), Rowland Hill Street, London NW3 2PF, UKTel +44 7376076260Email juan.alvarez.16@ucl.ac.ukBackground: In the interrupted time series (ITS) approach, it is common to average the outcome of interest at each time point and then perform a segmented regression (SR) analysis. In this study, we illustrate that such ‘aggregate-level’ analysis is biased when data are missing at random (MAR) and provide alternative analysis methods.Methods: Using electronic health records from the UK, we evaluated weight change over time induced by the initiation of antipsychotic treatment. We contrasted estimates from aggregate-level SR analysis against estimates from mixed models with and without multiple imputation of missing covariates, using individual-level data. Then, we conducted a simulation study for insight about the different results in a controlled environment.Results: Aggregate-level SR analysis suggested a substantial weight gain after initiation of treatment (average short-term weight change: 0.799kg/week) compared to mixed models (0.412kg/week). Simulation studies confirmed that aggregate-level SR analysis was biased when data were MAR. In simulations, mixed models gave less biased estimates than SR analysis and, in combination with multilevel multiple imputation, provided unbiased estimates. Mixed models with multiple imputation can be used with other types of ITS outcomes (eg, proportions). Other standard methods applied in ITS do not help to correct this bias problem.Conclusion: Aggregate-level SR analysis can bias the ITS estimates when individual-level data are MAR, because taking averages of individual-level data before SR means that data at the cluster level are missing not at random. Avoiding the averaging-step and using mixed models with or without multilevel multiple imputation of covariates is recommended.Keywords: interrupted time series analysis, segmented regression, missing data, multiple imputation, mixed effects models, electronic health records, big data

Irene Petersen,1,2 Andrew Phillips3 1Research Department of Primary Care and Population Health, University College London, London, UK; 2Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark; 3Institute for Global Health, University College London, London, UKCorrespondence: Irene PetersenResearch Department of Primary Care and Population Health, University College London, Rowland Hill St., London NW3 2PF, UKTel +44 020 801 68032Email i.petersen@ucl.ac.ukBackground: To reduce transmission of SARS-CoV-2, it is important to identify those who are infectious. However, little is known about what proportion of infectious people are asymptomatic and potential “silent” transmitters. We evaluated the value of COVID-19 symptoms as a marker for SARS-CoV-2 infection from a representative English survey.Methods: We used data from the Office for National Statistics Coronavirus (COVID-19) Infection Survey pilot study. We estimated sensitivity, specificity, the proportion of asymptomatic cases (1 – sensitivity), positive predictive value (PPV) and negative predictive value (NPV) of COVID-19 symptoms as a marker of infection using results of the SARS-CoV-2 test as the “gold standard”.Results: In total, there were 36,061 individuals with a SARS-CoV-2 test between 26 April and 27 June 2020. Of these, 625 (1.7%) reported symptoms on the day of the test. There were 115 (0.32%) with a positive SARS-CoV-2 test result. Of the 115, there were 27 (23.5%) who were symptomatic and 88 (76.5%) who were asymptomatic on the day of the test. Focusing on those with specific symptoms (cough, and/or fever, and/or loss of taste/smell), there were 158 (0.43%) with such symptoms on the day of the test. Of the 115 with a positive SARS-CoV-2, there were 16 (13.9%) reporting symptoms. In contrast, 99 (86.1%) did not report specific symptoms on the day of the test. The PPV for all symptoms was 4.3% and for the specific symptoms 10.1%. The specificity and NPV of symptoms were above 98%.Conclusion: COVID-19 symptoms are poor markers of SARS-CoV-2. Thus, 76.5% of this random sample who tested positive reported no symptoms, and 86.1% reported none of those specific to COVID-19. A more widespread testing programme is necessary to capture “silent” transmission and potentially prevent and reduce future outbreaks.Keywords: COVID-19 symptoms, SARS-CoV-2, sensitivity, asymptomatic SARS-CoV-2

Irene Petersen,1,2 Sia Kromann Nicolaisen,2 Federico Ricciardi,3 Manuj Sharma,1 Reimar W Thomsen,2 Gianluca Baio,3 Lars Pedersen2 1Department of Primary Care and Population Health, University College London, London, UK; 2Department of Clinical Epidemiology, Aarhus University, Aarhus, Denmark; 3Department of Statistical Science, University College London, London, UKCorrespondence: Irene PetersenDepartment of Primary Care and Population Health, University College London, London NW3 2PF, UKTel +44 20 801 68032Email i.petersen@ucl.ac.ukBackground: Individuals with type 2 diabetes (T2D) have a twofold increased risk for cardiovascular events (CVE), and CVE is responsible for nearly 80% of the mortality. Current treatment guidelines state that individuals should immediately initiate antidiabetic treatment and cardiovascular risk-factor management from T2D diagnosis. However, the evidence base is sparse, and randomized trials are unlikely to be conducted. We examined the impact of being eligible for T2D treatment, as determined by the threshold of HbA1c ≥ 6.5% (≥ 48 mmol/mol), on all-cause mortality and CVE. We hypothesised that individuals who were just above this threshold had a lower risk of CVE and all-cause mortality than individuals just below.Methods and Findings: We used the regression discontinuity design (RDD), a quasi-experimental design, comparing rates of all-cause mortality and CVE in people just below and just above the eligibility for treatment threshold. We included Danish healthcare records from 43,070 individuals aged 40– 80 years with no previous T2D record and the first record of HbA1c in the range of 6.0– 7.0% (42– 53 mmol/mol) between 2006 and 2014. In total, 36,360 individuals had the first record of HbA1c between 6.0% and 6.4% (42– 47 mmol/mol), and 6710 individuals had a first record between 6.5% and 7.0% (48– 53 mmol/mol). Individuals with a measurement just above 6.5% (48 mmol/mol) had a 21% lower rate of death or CVE, compared to those just below (hazard ratio: 0.79 (95% CI 0.69– 0.90)). Few individuals received early metformin treatment. However, the chance of metformin treatment initiation within 3 months was substantially higher for individuals with an HbA1c measurement above (14%) than below (1%) the threshold.Conclusion: Individuals with first record of HbA1c measure just above treatment threshold experienced a 21% lower rate of death or CVE than those just below. Lifestyle modifications and cardiovascular risk-factor management may contribute to this reduced rate.Keywords: type 2 diabetes, glycated hemoglobin A1c, regression discontinuity design, cardiovascular event, mortality

Mary Elizabeth Jones, 1 Irene Petersen, 1 Kate Walters, 1 Cini Bhanu, 1 Jill Manthorpe, 2 Rosalind Raine, 3 Naaheed Mukadam, 4 Claudia Cooper 4 1Department of Primary Care and Population Health, University College London, London, UK; 2NIHR Health and Social Care Workforce Research Unit, King’s College London, London, UK; 3Epidemiology and Public Health, University College London, London, UK; 4Division of Psychiatry, University College London, London, UKCorrespondence: Claudia CooperDivision of Psychiatry, University College London, 6th Floor, Maple House, Tottenham Court Road, London W1T 7BN, UKTel +44 7759703235Email claudia.cooper@ucl.ac.ukPurpose: To test hypotheses that minority ethnic people with dementia in the UK receive fewer anti-dementia drugs and more psychotropic and anticholinergic drugs associated with harms.Patients and Methods: We analyzed UK primary care electronic health records from The Health Improvement Network (THIN) database (2014– 2016), comparing psychotropic drug prescribing initiation and duration between people with dementia from White, Black, and Asian ethnic groups. We repeated analyses in people (aged 50+) without dementia, to explore whether any differences found reflected prescribing patterns in the general older population, or were specific to dementia.Results: We included 53,718 people with and 1,648,889 people without dementia. Among people with dementia, compared to White ethnic groups, Asian people were less likely to be prescribed anti-dementia drugs when they were potentially indicated (adjusted prevalence rate ratio 0.86 (95% Confidence Interval 0.76– 0.98)), and received them for on average 15 days/year less. Compared to White groups, Asian and Black individuals with dementia were no more likely to take an antipsychotic drug, but those that had were prescribed them for 17 and 27 days/year more, respectively (190.8 (179.6– 199.1) and 200.7 (191.1– 206.5) days). Black people were less likely to be prescribed anxiolytics/hypnotics (0.60 (0.44– 0.8)), but the duration these drugs were prescribed was similar across ethnic groups. Asian people were more likely to be prescribed anticholinergic drugs (1.43 (1.19– 1.73)), in analyses unadjusted for cardiovascular comorbidities. Among people without dementia, those in the Asian and Black ethnic groups were less likely to be prescribed psychotropic drugs, relative to people from White groups.Conclusion: Among people with dementia, Asian groups received less potentially beneficial symptomatic treatments, and Asian and Black groups were prescribed antipsychotic drugs for longer than White ethnic groups. Our findings may indicate care inequalities.Keywords: ethnicity, prevalence rate ratio, medication, prescription duration

By 2050 more than 1.6 billion women worldwide will be of post-reproductive age, with >75% reporting severe menopausal symptoms. The last few years saw a gradual uplift in public awareness reaffirming the health needs of women with menopause. Still, effective translation of available evidence on menopause treatments is hindered by several methodological limitations and poor research conduct. We argue that a paradigm shift is required in menopause research to address the remaining knowledge gap and guide safe evidence-based care provision. A critical misconception across studies on menopause is the assumption that women represent a homogeneous group who respond similarly to a particular therapy irrespective of their exposure and individual risk factors. We highlight potential solutions to optimize the quality of future research in menopause including adopting robust trial methodology, standardize outcome reporting to capture quality-of-life measures, and improve lay patient and public involvement in future research. [ABSTRACT FROM AUTHOR]

Tra My Pham,1,2 James R Carpenter,1,3 Tim P Morris,1 Manuj Sharma,2 Irene Petersen2,4 1MRC Clinical Trials Unit at UCL, London WC1V 6LJ, UK; 2Department of Primary Care and Population Health, University College London, London NW3 2PF, UK; 3Department of Medical Statistics, London School of Hygiene & Tropical Medicines, London WC1E 7HT, UK; 4Department of Clinical Epidemiology, Aarhus University, Aarhus N 8200, DenmarkCorrespondence: Tra My PhamMRC Clinical Trials Unit at UCL, 90 High Holborn, London WC1V 6LJ, UKTel +44207 670 4626Email tra.pham.09@ucl.ac.ukAims/Hypothesis: Type 2 diabetes mellitus is associated with high levels of disease burden, including increased mortality risk and significant long-term morbidity. The prevalence of diabetes differs substantially among ethnic groups. We examined the prevalence of type 2 diabetes diagnoses in the UK primary care setting.Methods: We analysed data from 404,318 individuals in The Health Improvement Network database, aged 0–99 years and permanently registered with general practices in London. The association between ethnicity and the prevalence of type 2 diabetes diagnoses in 2013 was estimated using a logistic regression model, adjusting for effect of age group, sex, and social deprivation. A multiple imputation approach utilising population-level information about ethnicity from the UK census was used for imputing missing data.Results: Compared with those of White ethnicity (5.04%, 95% CI 4.95 to 5.13), the crude percentage prevalence of type 2 diabetes was higher in the Asian (7.69%, 95% CI 7.46 to 7.92) and Black (5.58%, 95% CI 5.35 to 5.81) ethnic groups, while lower in the Mixed/Other group (3.42%, 95% CI 3.19 to 3.66). After adjusting for differences in age group, sex, and social deprivation, all minority ethnic groups were more likely to have a diagnosis of type 2 diabetes compared with the White group (OR Asian versus White 2.36, 95% CI 2.26 to 2.47; OR Black versus White 1.65, 95% CI 1.56 to 1.73; OR Mixed/Other versus White 1.17, 95% CI 1.08 to 1.27).Conclusion: The prevalence of type 2 diabetes was higher in the Asian and Black ethnic groups, compared with the White group. Accurate estimates of ethnic prevalence of type 2 diabetes based on large datasets are important for facilitating appropriate allocation of public health resources, and for allowing population-level research to be undertaken examining disease trajectories among minority ethnic groups, that might help reduce inequalities.Keywords: ethnicity, type 2 diabetes, primary care database, electronic health records, multiple imputation, missing not at random

Irene Petersen,1,2 Catherine A Welch,3 Irwin Nazareth,1 Kate Walters,1 Louise Marston,1 Richard W Morris,4 James R Carpenter,5,6 Tim P Morris,5 Tra My Pham1 1Department of Primary Care and Population Health, University College London, London NW3 2PF, UK; 2Department of Clinical Epidemiology, Aarhus University, 8200 Aarhus N, Denmark; 3Department of Health Sciences, University of Leicester, Leicester LE1 7RH, UK; 4Department of Population Health Sciences, Bristol Medical School, University of Bristol, Bristol BS8 2PS, UK; 5MRC Clinical Trials Unit at UCL, London WC1V 6LJ, UK; 6Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London WC1E 7HT, UK Background: Clinical databases are increasingly used for health research; many of them capture information on common health indicators including height, weight, blood pressure, cholesterol level, smoking status, and alcohol consumption. However, these are often not recorded on a regular basis; missing data are ubiquitous. We described the recording of health indicators in UK primary care and evaluated key implications for handling missing data.Methods: We examined the recording of health indicators in The Health Improvement Network (THIN) UK primary care database over time, by demographic variables (age and sex) and chronic diseases (diabetes, myocardial infarction, and stroke). Using weight as an example, we fitted linear and logistic regression models to examine the associations of weight measurements and the probability of having weight recorded with individuals’ demographic characteristics and chronic diseases.Results: In total, 6,345,851 individuals aged 18–99 years contributed data to THIN between 2000 and 2015. Women aged 18–65 years were more likely than men of the same age to have health indicators recorded; this gap narrowed after age 65. About 60–80% of individuals had their height, weight, blood pressure, smoking status, and alcohol consumption recorded during the first year of registration. In the years following registration, these proportions fell to 10%–40%. Individuals with chronic diseases were more likely to have health indicators recorded, particularly after the introduction of a General Practitioner incentive scheme. Individuals’ demographic characteristics and chronic diseases were associated with both observed weight measurements and missingness in weight.Conclusion: Missing data in common health indicators will affect statistical analysis in health research studies. A single analysis of primary care data using the available information alone may be misleading. Multiple imputation of missing values accounting for demographic characteristics and disease status is recommended but should be considered and implemented carefully. Sensitivity analysis exploring alternative assumptions for missing data should also be evaluated. Keywords: primary care, EHRs, recording, QOF, multiple imputation, statistics, epidemiology, research methods, data analysis

Leukocyte telomere length (LTL) is associated with a number of common age-related diseases and is a heritable trait. Previous genome-wide association studies (GWASs) identified two loci on chromosomes 3q26.2 (TERC) and 10q24.33 (OBFC1) that are associated

Manuj Sharma,1 Irwin Nazareth,1 Irene Petersen1,2 1Department of Primary Care and Population Health, University College London, London, UK; 2Department of Clinical Epidemiology, Aarhus University, Aarhus, Denmark Abstract: Observational studies which evaluate effectiveness are often viewed with skepticism owing to the fact that patients are not randomized to treatment, meaning that results are more prone to bias. Therefore, randomized controlled trials remain the gold standard for evaluating treatment effectiveness. However, it is not always possible to conduct randomized trials. This may be due to financial constraints, for example, in identifying funding for a randomized trial for medicines that have already gained market authorization. There can also be challenges with recruitment, for example, of people with rare conditions or in hard-to-reach population subgroups. This is why observational studies are still needed. In this manuscript, we discuss how researchers can mitigate the risk of bias in the most common type of observational study design for evaluation of treatment effectiveness, the cohort study. We outline some key issues that warrant careful consideration at the outset when the question is being developed and the cohort study is being designed. We focus our discussion on the importance of deciding when to start follow-up in a study, choosing a comparator, managing confounding and measuring outcomes. We also illustrate the application of these considerations in a more detailed case study based on an examination of comparative effectiveness of two antidiabetic treatments using data collected during routine clinical practice. Keywords: epidemiology, therapeutics, diabetes mellitus, public health, effectiveness

Stine Bakkensen Bruun,1 Irene Petersen,1,2 Nickolaj Risbo Kristensen,1 Deirdre Cronin-Fenton,1 Alma Becic Pedersen1 1Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark; 2Department of Primary Care and Population Health, University College London, London, UK Purpose: To examine the association between selective serotonin reuptake inhibitor (SSRI) use and mortality, postoperative complications, and quality of in-hospital care in hip fracture patients. Patients and methods: The study was a nationwide cohort study based on individual-level linked, prospectively collected data from Danish population-based national registries covering all hospitals in Denmark. The health care system in Denmark is tax-funded, and all citizens have equal access to health care services. We included patients with first-time hospitalization due to hip fracture undergoing surgery from 2006–2016. We estimated the risk of 30-day mortality, any unplanned readmission, any reoperation, specific postoperative complications including cardiovascular events and major bleeding, and quality of in-hospital care using Cox and Poisson regression analyses comparing current and former SSRI users with non-users. Results: In 68,487 hip fracture patients, 13,272 (19%) were current SSRI users, 2,777 (4%) were former SSRI users, and 52,438 (77%) were SSRI non-users. The 30-day mortality risk was 13% in current SSRI users (HR 1.16, 1.10–1.21) and 12% in former (HR 1.15, 1.04–1.27) compared with 10% in non-users. The HR for any unplanned readmission was 1.11 (1.02–1.20) in current and 1.13 (1.01–1.27) in former SSRI users and for any reoperation 1.21 (1.11–1.31) in current and 1.04 (0.84–1.28) in former SSRI users compared with non-users. The risk of venous thromboembolism, myocardial infarction, stroke, and bleeding were similar irrespective of SSRI use. No association between current and former SSRI use and quality of in-hospital care was found. Conclusion: In patients undergoing hip fracture surgery, 30-day mortality and overall readmission risk were elevated in both current and former SSRI users compared with non-users. Those currently using SSRI had a 26% increased reoperation risk compared with non-users. However, SSRI use was not associated with increased risk of other postoperative complications and lower quality of in-hospital care. A limitation of this study was the inability to control for potential confounding of social deprivation. Keywords: cohort studies, hip fracture, mortality, postoperative complications, quality of care, selective serotonin reuptake inhibitors

Tra My Pham,1 Irene Petersen,1,2 Kate Walters,1 Rosalind Raine,3 Jill Manthorpe,4 Naaheed Mukadam,5 Claudia Cooper5 1Department of Primary Care and Population Health, University College London, London, UK; 2Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark; 3Department of Applied Health Research, University College London, London, UK; 4Social Care Workforce Research Unit, King’s College London, London, UK; 5Division of Psychiatry, University College London, London, UK Objectives: We compared incidence of dementia diagnosis by white, black, and Asian ethnic groups and estimated the proportion of UK white and black people developing dementia in 2015 who had a diagnosis for the first time in a UK-wide study. Methods: We analyzed primary care electronic health records from The Health Improvement Network database between 2007 and 2015 and compared incidence of dementia diagnosis to dementia incidence from community cohort studies. The study sample comprised of 2,511,681 individuals aged 50–105 years who did not have a dementia diagnosis prior to the start of follow-up. Results: A total of 66,083 individuals had a dementia diagnosis (4.87/1,000 person-years at risk, 95% CI 4.83–4.90); this incidence increased from 3.75 to 5.65/1,000 person-years at risk between 2007 and 2015. Compared with white women, the incidence of dementia diagnosis was 18% lower among Asian women (adjusted incidence rate ratio (IRR) 0.82, 95% CI 0.72–0.95) and 25% higher among black women (IRR 1.25, 95% CI 1.07–1.46). For men, incidence of dementia diagnosis was 28% higher in the black ethnic group (IRR 1.28, 95% CI 1.08–1.50) and 12% lower in the Asian ethnic group (IRR 0.88, 95% CI 0.76–1.01) relative to the white ethnic group. Based on diagnosis incidence in The Health Improvement Network data and projections of incidence from community cohort studies, we estimated that 42% of black men developing dementia in 2015 were diagnosed compared with 53% of white men. Conclusion: People from the black ethnic group had a higher incidence of dementia diagnosis and those from the Asian ethnic group had lower incidence compared with the white ethnic group. We estimated that black men developing dementia were less likely than white men to have a diagnosis of dementia, indicating that the increased risk of dementia diagnosis reported in the black ethnic group might underestimate the higher risk of dementia in this group. It is unclear whether the lower incidence of dementia diagnosis in the Asian ethnic group reflects lower community incidence or underdiagnosis. A cohort study to determine this is needed. Keywords: dementia, ethnicity, primary care, electronic health records

Elisa Martín-Merino,1 Amaia Calderón-Larrañaga,2,3 Samuel Hawley,4 Beatriz Poblador-Plou,3 Ana Llorente-García,1 Irene Petersen,5,6 Daniel Prieto-Alhambra4,7 1Base de datos para la Investigación Farmacoepidemiológica en Atención Primaria, Division of Pharmacoepidemiology and Pharmacovigilance, Spanish Agency of Medicines and Medical Devices, Madrid, Spain; 2Aging Research Center, Karolinska Institutet, Stockholm University, Stockholm, Sweden; 3EpiChron Research Group on Chronic Diseases, Aragon Health Sciences Institute, Aragon Health Research Institute, Miguel Servet University Hospital, Zaragoza, Spain; 4Centre for Statistics in Medicine, Botnar Research Centre, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford, UK; 5Department of Primary Care and Population Health, University College London, London, UK; 6Department of Clinical Epidemiology, Aarhus University, Aarhus, Denmark; 7GREMPAL (Grup de Recerca en Malalties Prevalents de l’Aparell Locomotor) Research Group, Idiap Jordi Gol and CIBERFes, Instituto de Salud Carlos III, Universitat Autonoma de Barcelona, Barcelona, Spain Background: Missing data are often an issue in electronic medical records (EMRs) research. However, there are many ways that people deal with missing data in drug safety studies.Aim: To compare the risk estimates resulting from different strategies for the handling of missing data in the study of venous thromboembolism (VTE) risk associated with antiosteoporotic medications (AOM). Methods: New users of AOM (alendronic acid, other bisphosphonates, strontium ranelate, selective estrogen receptor modulators, teriparatide, or denosumab) aged ≥50 years during 1998–2014 were identified in two Spanish (the Base de datos para la Investigación Farmacoepidemiológica en Atención Primaria [BIFAP] and EpiChron cohort) and one UK (Clinical Practice Research Datalink [CPRD]) EMR. Hazard ratios (HRs) according to AOM (with alendronic acid as reference) were calculated adjusting for VTE risk factors, body mass index (that was missing in 61% of patients included in the three databases), and smoking (that was missing in 23% of patients) in the year of AOM therapy initiation. HRs and standard errors obtained using cross-sectional multiple imputation (MI) (reference method) were compared to complete case (CC) analysis – using only patients with complete data – and longitudinal MI – adding to the cross-sectional MI model the body mass index/smoking values as recorded in the year before and after therapy initiation. Results: Overall, 422/95,057 (0.4%), 19/12,688 (0.1%), and 2,051/161,202 (1.3%) VTE cases/participants were seen in BIFAP, EpiChron, and CPRD, respectively. HRs moved from 100.00% underestimation to 40.31% overestimation in CC compared with cross-sectional MI, while longitudinal MI methods provided similar risk estimates compared with cross-sectional MI. Precision for HR improved in cross-sectional MI versus CC by up to 160.28%, while longitudinal MI improved precision (compared with cross-sectional) only minimally (up to 0.80%). Conclusion: CC may substantially affect relative risk estimation in EMR-based drug safety studies, since missing data are not often completely at random. Little improvement was seen in these data in terms of power with the inclusion of longitudinal MI compared with cross-sectional MI. The strategy for handling missing data in drug safety studies can have a large impact on both risk estimates and precision. Keywords: missing data, electronic medical records, pharmacoepidemiology, multiple imputation, complete case analysis, longitudinal data

Kristina Laugesen,1 Henrik Støvring,2 Jesper Hallas,3 Anton Pottegård,3 Jens Otto Lunde Jørgensen,4 Henrik Toft Sørensen,1 Irene Petersen1,5 1Department of Clinical Epidemiology, Aarhus University Hospital, 2Department of Public Health, Aarhus University, Aarhus, 3Clinical Pharmacology and Pharmacy, Department of Public Health, University of Southern Denmark, Odense, 4Department of Endocrinology and Internal Medicine, Aarhus University Hospital, Aarhus, Denmark; 5Department of Primary Care and Population Health, University College London, London, UK Purpose: Glucocorticoids are widely used medications. In many pharmacoepidemiological studies, duration of individual prescriptions and definition of treatment episodes are important issues. However, many data sources lack this information. We aimed to estimate duration of individual prescriptions for oral glucocorticoids and to describe continuous treatment episodes using the parametric waiting time distribution.Methods: We used Danish nationwide registries to identify all prescriptions for oral glucocorticoids during 1996–2014. We applied the parametric waiting time distribution to estimate duration of individual prescriptions each year by estimating the 80th, 90th, 95th and 99th percentiles for the interarrival distribution. These corresponded to the time since last prescription during which 80%, 90%, 95% and 99% of users presented a new prescription for redemption. We used the Kaplan–Meier survival function to estimate length of first continuous treatment episodes by assigning estimated prescription duration to each prescription and thereby create treatment episodes from overlapping prescriptions.Results: We identified 5,691,985 prescriptions issued to 854,429 individuals of whom 351,202 (41%) only redeemed 1 prescription in the whole study period. The 80th percentile for prescription duration ranged from 87 to 120 days, the 90th percentile from 116 to 150 days, the 95th percentile from 147 to 181 days, and the 99th percentile from 228 to 259 days during 1996–2014. Based on the 80th, 90th, 95th and 99th percentiles of prescription duration, the median length of continuous treatment was 113, 141, 170 and 243 days, respectively.Conclusion: Our method and results may provide an important framework for future pharmacoepidemiological studies. The choice of which percentile of the interarrival distribution to apply as prescription duration has an impact on the level of misclassification. Use of the 80th percentile provides a measure of drug exposure that is specific, while the 99th percentile provides a sensitive measure. Keywords: glucocorticoids, pharmacoepidemiology, prescription duration, parametric waiting time distribution

L Fardet,1–3 I Nazareth,1 I Petersen1 1Department of Primary Care and Population Health, University College London, UK; 2Department of Dermatology, Henri Mondor Hospital AP-HP, Créteil, France; 3Equipe d’Accueil 7379 EpiDermE, Université Paris Est Créteil, Créteil, France Background: Hydroxychloroquine and chloroquine may reduce the risk of cancer as they inhibit autophagy, in particular, in people with connective tissue diseases.Methods: The hazard ratios of cancers, metastases, and death were assessed in adults with connective tissue diseases prescribed hydroxychloroquine/chloroquine for at least 1 year in comparison with unexposed individuals with the same underlying conditions. A competing risk survival regression analysis was performed. Data were extracted from the Health Improvement Network UK primary care database.Results: Eight thousand nine hundred and ninety-nine individuals exposed to hydroxychloroquine (98.6%) or chloroquine (1.4%) and 24,118 unexposed individuals were included in the study (median age: 56 [45–66] years, women: 76.8%). When compared to the unexposed group, individuals exposed to hydroxychloroquine/chloroquine were not at lower risk of non-skin cancers (adjusted sub-distribution hazard ratio [sHR]: 1.04 [0.92–1.18], p=0.54), hematological malignancies (adjusted sHR: 1.00 [0.73–1.38], p=0.99), or skin cancers (adjusted sHR: 0.92 [0.78–1.07], p=0.26). The risk of metastasis was not significantly different between the two groups. However, it was significantly lower during the exposure period when compared with the unexposed (adjusted sHR: 0.64 [0.44–0.95] for the overall population and 0.61 [0.38–1.00] for those diagnosed with incident cancers). The risk of death was also significantly lower in those exposed to hydroxychloroquine/chloroquine (adjusted HR: 0.90 [0.81–1.00] in the overall population and 0.78 [0.64–0.96] in those diagnosed with incident cancer).Conclusion: Individuals on long-term exposure to hydroxychloroquine/chloroquine are not at lower risk of cancer. However, hydroxychloroquine/chloroquine may lower the risk of metastatic cancer and death. Keywords: antimalarial drugs, cancer, death, connective tissue diseases

Alma B Pedersen,1 Ellen M Mikkelsen,1 Deirdre Cronin-Fenton,1 Nickolaj R Kristensen,1 Tra My Pham,2 Lars Pedersen,1 Irene Petersen1,2 1Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus N, Denmark; 2Department of Primary Care and Population Health, University College London, London, UK Abstract: Missing data are ubiquitous in clinical epidemiological research. Individuals with missing data may differ from those with no missing data in terms of the outcome of interest and prognosis in general. Missing data are often categorized into the following three types: missing completely at random (MCAR), missing at random (MAR), and missing not at random (MNAR). In clinical epidemiological research, missing data are seldom MCAR. Missing data can constitute considerable challenges in the analyses and interpretation of results and can potentially weaken the validity of results and conclusions. A number of methods have been developed for dealing with missing data. These include complete-case analyses, missing indicator method, single value imputation, and sensitivity analyses incorporating worst-case and best-case scenarios. If applied under the MCAR assumption, some of these methods can provide unbiased but often less precise estimates. Multiple imputation is an alternative method to deal with missing data, which accounts for the uncertainty associated with missing data. Multiple imputation is implemented in most statistical software under the MAR assumption and provides unbiased and valid estimates of associations based on information from the available data. The method affects not only the coefficient estimates for variables with missing data but also the estimates for other variables with no missing data. Keywords: missing data, observational study, multiple imputation, MAR, MCAR, MNAR

Irene Petersen,1,2 Shuk-Li Collings,1,3 Rachel L McCrea,1 Irwin Nazareth,1 David P Osborn,4 Phil J Cowen,5 Cormac J Sammon1 1Department of Primary Care and Population Health, University College London, London, UK; 2Department of Clinical Epidemiology, Aarhus University, Aarhus N, Denmark; 3OXON Epidemiology, London, UK; 4Division of Psychiatry, University College London, London, UK; 5University Department of Psychiatry, Warneford Hospital, Oxford, UK Objective: The aim of this study was to examine the prevalence of major congenital malformations associated with antiepileptic drug (AED) treatment in pregnancy.Patients and methods: Using data from The Health Improvement Network, we identified women who have given live birth and their offspring. Four subgroups were selected based on the AED treatment in early pregnancy, valproate, carbamazepine, lamotrigine and women not receiving AED treatment. We compared the prevalence of major congenital malformations within children of these four groups and estimated prevalence ratios (PRs) using Poisson regression adjusted for maternal age, sex of child, quintiles of Townsend deprivation score and indication for treatment.Results: In total, 240,071 women were included in the study. A total of 229 women were prescribed valproate in pregnancy, 357 were prescribed lamotrigine and 334 were prescribed carbamazepine and 239,151 women were not prescribed AEDs. Fifteen out of 229 (6.6%) women prescribed valproate gave birth to a child with a major congenital malformation. The figures for lamotrigine, carbamazepine and women not prescribed AEDs were 2.7%, 3.3% and 2.2%, respectively. The prevalence of major congenital malformation was similar for women prescribed lamotrigine or carbamazepine compared to women with no AED treatment in pregnancy. For women prescribed valproate in polytherapy, the prevalence was fourfold higher. After adjustments, the effect of estimates attenuated, but the prevalence remained two- to threefold higher in women prescribed valproate.Conclusion: The results of our study suggest that lamotrigine and carbamazepine are safer treatment options than valproate in pregnancy and should be considered as alternative treatment options for women of childbearing potential and in pregnancy. Keywords: pregnancy, valproate, lamotrigine, carbamazepine, adverse drug effects