Huub Straatman

Using electronic health care records for drug safety signal detection: a comparative evaluation of statistical methods.

Background:Drug safety monitoring relies primarily on spontaneous reporting, but electronic health care record databases offer a possible alternative for the detection of adverse drug reactions (ADRs). Objectives:To evaluate the relative performance of different statistical methods for detecting drug-adverse event associations in electronic health care record data representing potential ADRs. Research Design:Data from 7 databases across 3 countries in Europe comprising over 20 million subjects were used to compute the relative risk estimates for drug-event pairs using 10 different methods, including those developed for spontaneous reporting systems, cohort methods such as the longitudinal gamma poisson shrinker, and case-based methods such as case-control. The newly developed method “longitudinal evaluation of observational profiles of adverse events related to drugs” (LEOPARD) was used to remove associations likely caused by protopathic bias. Data from the different databases were combined by pooling of data, and by meta-analysis for random effects. A reference standard of known ADRs and negative controls was created to evaluate the performance of the method. Measures:The area under the curve of the receiver operator characteristic curve was calculated for each method, both with and without LEOPARD filtering. Results:The highest area under the curve (0.83) was achieved by the combination of either longitudinal gamma poisson shrinker or case-control with LEOPARD filtering, but the performance between methods differed little. LEOPARD increased the overall performance, but flagged several known ADRs as caused by protopathic bias. Conclusions:Combinations of methods demonstrate good performance in distinguishing known ADRs from negative controls, and we assume that these could also be used to detect new drug safety signals.

Withdrawal of Statins and Risk of Subarachnoid Hemorrhage

Background and Purpose— Vascular endothelium, which can be affected by statins, is believed to play a substantial role in subarachnoid hemorrhage (SAH). Our objective was to estimate the association between use and withdrawal of statins and the risk of SAH. Methods— We conducted a population-based case–control study within the PHARMO database. A case was defined as a person hospitalized for SAH (ICD-9-CM code 430) in the period January 1, 1998 to December 31, 2006. Ten randomly chosen controls were matched to each case on age, gender, and calendar date. Results— During the study period 1004 incident cases of SAH were identified. Current use of statins did not significantly decrease the risk of SAH (OR=0.77, 95% CI 0.55 to 1.07). The odds ratio for recent withdrawal compared to nonusers was 1.62 (95% CI 0.96 to 2.73). Compared to current use, recent withdrawal was associated with an increased risk of SAH (OR=2.34, 95% CI 1.35 to 4.05). Interaction analysis showed that the effect of statin withdrawal was highest in patients who had also recently stopped antihypertensive drugs (OR=6.77, 95% CI 2.10 to 21.8). Conclusions— Current use of statins seems to lower the risk of SAH, although the reduction was not significant in new users. Statin withdrawal increased the risk of SAH by a factor 2, even more in patients who had also recently stopped their antihypertensive treatment.

Replication of the OMOP Experiment in Europe: Evaluating Methods for Risk Identification in Electronic Health Record Databases

BackgroundThe Observational Medical Outcomes Partnership (OMOP) has just completed a large scale empirical evaluation of statistical methods and analysis choices for risks identification in longitudinal observational healthcare data. This experiment drew data from four large US health insurance claims databases and one US electronic health record (EHR) database, but it is unclear to what extend the findings of this study apply to other data sources.ObjectiveTo replicate the OMOP experiment in six European EHR databases.Research DesignSix databases of the EU-ADR (Exploring and Understanding Adverse Drug Reactions) database network participated in this study: Aarhus (Denmark), ARS (Italy), HealthSearch (Italy), IPCI (the Netherlands), Pedianet (Italy), and Pharmo (the Netherlands). All methods in the OMOP experiment were applied to a collection of 165 positive and 234 negative control drug–outcome pairs across four outcomes: acute liver injury, acute myocardial infarction, acute kidney injury, and upper gastrointestinal bleeding. Area under the receiver operator characteristics curve (AUC) was computed per database and for a combination of all six databases using meta-analysis for random effects. We provide expected values of estimation error as well, based on negative controls.ResultsSimilarly to the US experiment, high predictive accuracy was found (AUC >0.8) for some analyses. Self-controlled designs, such as self-controlled case series, IC temporal pattern discovery and self-controlled cohort achieved higher performance than other methods, both in terms of predictive accuracy and observed bias.ConclusionsThe major findings of the recent OMOP experiment were also observed in the European databases.

Platelet aggregation inhibitors, vitamin K antagonists and risk of subarachnoid hemorrhage

Summary.  Background: Use of platelet aggregation inhibitors and vitamin K antagonists has been associated with an increased risk of intracranial hemorrhage (ICH). Whether the use of these antithrombotic drugs is associated with an increased risk of subarachnoid hemorrhage (SAH) remains unclear, especially as confounding by indication might play a role. Objective: The aim of the present study was to investigate whether use of platelet aggregation inhibitors or vitamin K antagonists increase the risk of SAH. Methods: We applied population‐based case–control, case–crossover and case–time–control designs to estimate the risk of SAH while addressing issues both of confounding by indication and time varying exposure within the PHARMO Record Linkage System database. This system includes drug dispensing records from community pharmacies and hospital discharge records of more than 3 million community‐dwelling inhabitants in the Netherlands. Patients were considered a case if they were hospitalized for a first SAH (ICD‐9‐CM code 430) in the period between 1st January 1998 and 31st December 2006. Controls were selected from the source population, matched on age, gender and date of hospitalization. Conditional logistic regression was used to estimate multivariable adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the risk of SAH during use of platelet aggregation inhibitors or vitamin K antagonists. In the case–crossover and case–time–control designs we selected 11 control periods preceding the index date in successive steps of 1 month in the past. Results: In all, 1004 cases of SAH were identified. In the case–control analysis the adjusted OR for the risk of SAH in current use of platelet aggregation inhibitors was 1.32 (95% CI: 1.02–1.70) and in current use of vitamin K antagonists 1.29 (95% CI: 0.89–1.87) compared with no use. In the case–crossover analysis the ORs for the risk of SAH in current use of platelet aggregation inhibitors and vitamin K antagonists were 1.04 (95% CI: 0.56–1.94) and 2.46 (95% CI: 1.04–5.82), respectively. In the case–time–control analysis the OR for platelet aggregation inhibitors was 0.50 (95% CI: 0.26–0.98) and for vitamin K antagonists 1.98 (95% CI: 0.82–4.76). Conclusion: The use of platelet aggregation inhibitors was not associated with an increased SAH risk; the modest increase observed in the case–control analysis could be as a result of confounding. The use of vitamin K antagonists seemed to be associated with an increased risk of SAH. The increase was most pronounced in the case–crossover analysis and therefore cannot be explained by unmeasured confounding.

Risk of ischemic stroke and the use of individual non-steroidal anti-inflammatory drugs: A multi-country European database study within the SOS Project

Background and purpose A multi-country European study using data from six healthcare databases from four countries was performed to evaluate in a large study population (>32 million) the risk of ischemic stroke (IS) associated with individual NSAIDs and to assess the impact of risk factors of IS and co-medication. Methods Case-control study nested in a cohort of new NSAID users. For each case, up to 100 sex- and age-matched controls were selected and confounder-adjusted odds ratios for current use of individual NSAIDs compared to past use calculated. Results 49,170 cases of IS were observed among 4,593,778 new NSAID users. Use of coxibs (odds ratio 1.08, 95%-confidence interval 1.02–1.15) and use of traditional NSAIDs (1.16, 1.12–1.19) were associated with an increased risk of IS. Among 32 individual NSAIDs evaluated, the highest significant risk of IS was observed for ketorolac (1.46, 1.19–1.78), but significantly increased risks (in decreasing order) were also found for diclofenac, indomethacin, rofecoxib, ibuprofen, nimesulide, diclofenac with misoprostol, and piroxicam. IS risk associated with NSAID use was generally higher in persons of younger age, males, and those with a prior history of IS. Conclusions Risk of IS differs between individual NSAIDs and appears to be higher in patients with a prior history of IS or transient ischemic attack (TIA), in younger or male patients. Co-medication with aspirin, other antiplatelets or anticoagulants might mitigate this risk. The small to moderate observed risk increase (by 13–46%) associated with NSAIDs use represents a public health concern due to widespread NSAID usage.

Risk of acute myocardial infarction during use of individual NSAIDs: A nested case-control study in the SOS project

Background: Sudden discontinuation of some antihypertensive agents such as beta-blockers and centrally acting antihypertensive agents are associated with increased risk of acute coronary events. Objectives: The aim of this study was to assess the association between discontinuation of different antihypertensive agents and the risk of acute myocardial infarction (AMI). Methods: A nested case control study was performed in a cohort of antihypertensive drug users from the Utrecht Cardiovascular Pharmacogenetics (UCP) database. Within this cohort, patients who were hospitalized for first AMI were considered cases. Cases were matched (1 up to 4) to controls at the same AMI date (index date). Antihypertensive users were defined as current users if the index date fell within prescribed duration or as stoppers if this date fell outside the prescribed duration. According to recency of stopping, stoppers were divided into recent stoppers (≤90 days), intermediate-term stoppers (91-180 days), and longterm stoppers (>180 days). The study included only antihypertensive users who were specifically current users or stoppers of one antihypertensive agent. Logistic regression analysis was used to assess the association between the discontinuation of antihypertensive agents and the risk of AMI and to control for confounding. Results: We included 1245 cases and 4994 controls in our analysis. The risk of AMI was significantly increased with all stoppers of beta-blockers (adjusted OR: 1.54, 95%CI (1.25-1.90)), calcium channel blockers (CCBs) (adjusted OR: 2.25, 95%CI (1.53- 3.30)), and diuretics (adjusted OR: 1.76, 95%CI (1.24-2.48)) compared with current users. Moreover, the risk of AMI was significantly increased for longterm stoppers (beta-blockers, CCBs, angiotensinconverting enzyme inhibitors, and diuretics) and intermediate- term stoppers (beta-blockers and CCBs) versus current users. There was no difference in AMI risk between recent stoppers of antihypertensive agents versus current users. Conclusions: Discontinuation of antihypertensive agents increases the risk of AMI after more than 90 days of stopping. Adherence to antihypertensive agents plays an important role in reducing the risk of AMI in patients with hypertension.Background: It has been reported that patients with type 1 diabetes (T1DM) have a decreased lung function. Studies on the association of T1DM and asthma in children show controversial results. Objectives: The aim of this study was to quantify asthma medication use in children and adolescents with and without (reference cohort) T1DM 5 years before and after the onset of diabetes. Methods: A population-based cohort study was conducted in the Dutch PHARMO Record Linkage System. All children (1. Modelling of Endpoint Postponement for All-Cause Mortality in Statin Trials Morten Rix Hansen, Anton Pottegård, Asbjørn Hróbjartsson, Per Damkier, René D Christensen, Kasper Søltoft Larsen, Malene EL Kristensen, Palle M Christensen, Jesper Hallas. Clinical Pharmacology, University of Southern Denmark, Odense, Denmark; Department of Clinical Chemistry and Pharmacology, Odense University Hospital, Odense, Denmark; The Nordic Cochrane Centre, Rigshospitalet, Copenhagen, Denmark; Research Unit for General Practice, University of Southern Denmark, Odense, Denmark. Background: The average postponement of the outcome event has been proposed as a novel method to present the magnitude of effect for preventive medications. This measure has been shown to have better agreement with patient preferences than conventional outcomemeasures, including the “number needed to treat” (NNT), possibly because it is more intuitively understood. For some interventions, it may also provide a better theoretical frame for how benefit is distributed among participants than the NNT measure. The aim of this study was to present a novel method for modelling endpoint postponement (EP) from trial data and compare it with the usual approach of measuring the area between survival curves. We also present a formalized meta-analysis of modelled EP for all-cause mortality in statin trials. Methods: We identified 17 placebo-controlled statin trials that fulfilled our inclusion criteria. Eleven of these presented Kaplan–Meier curves for all-cause mortality. Average EP was calculated as the area between Kaplan–Meier curves by counting pixels on magnified prints for these 11 trials. The modelled EP was computed for all trials on the basis of (1) hazard ratio, relative risk or odds ratio; (2) the cumulative event rate in the untreated group; and (3) the trial’s running time. The underlying assumption was that the mortality was reasonably stable within the trials’ running time. The modelled EP was subjected to a meta-analysis, using inverse variance weighting in a random effect model. Results: EPswere generally small for estimates based on pixel-counting, 10 and 27days for trials both primary and secondary intervention that typically ran over 1.9– 6.1years. The modelled EPs varied between 2 and 34days. The difference between modeled EP and EP based on pixel-counting was between 8 and 12days. The results of the meta-analyses will be presented at the meeting. Conclusions: Based on these trial data, statin treatment results in a surprisingly small gain in average survival. Our modelled EP estimates agreed reasonably with EPs based on pixel-counting. The modeled EP is amenable to meta-analyses and may be a useful approach to presenting the benefit of preventive treatment. 2. Permanent User Bias in Case–Crossover Studies in Pharmacoepidemiology Jesper Hallas, Shirley V. Wang, Joshua J. Gagne, Sebastian Schneeweiss, Anton Pottegård. Clinical Pharmacology, University of Southern Denmark, Odense C, Denmark; Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA. Background: In pharmacoepidemiology, the case– crossover design is based on cases that have contrasting drug exposure at the time of an event and at a reference time in the past. If the drug in question should be taken permanently, only certain exposure patterns will occur. These patients cannot be unexposed at the event time and exposed at the reference time, while the opposite pattern can occur if the drug was initiated recently. The resulting odds ratio (OR) would thus be biased upward. As many drugs have a subpopulation of permanent users, this bias might pervade many case–crossover analyses of drug effects. Objectives: The aims of this study were to demonstrate this “permanent user bias” and to evaluate whether it can be remedied by including a control group (case–time–control design). Methods: Using nationwide Danish data resources, we conducted case–crossover and case–time–control analyses for combinations of three exposures that are

Individualized NSAID prescribing based on gastrointestinal and cardiovascular risks: A decision model in the SOS project

Background: Sudden discontinuation of some antihypertensive agents such as beta-blockers and centrally acting antihypertensive agents are associated with increased risk of acute coronary events. Objectives: The aim of this study was to assess the association between discontinuation of different antihypertensive agents and the risk of acute myocardial infarction (AMI). Methods: A nested case control study was performed in a cohort of antihypertensive drug users from the Utrecht Cardiovascular Pharmacogenetics (UCP) database. Within this cohort, patients who were hospitalized for first AMI were considered cases. Cases were matched (1 up to 4) to controls at the same AMI date (index date). Antihypertensive users were defined as current users if the index date fell within prescribed duration or as stoppers if this date fell outside the prescribed duration. According to recency of stopping, stoppers were divided into recent stoppers (≤90 days), intermediate-term stoppers (91-180 days), and longterm stoppers (>180 days). The study included only antihypertensive users who were specifically current users or stoppers of one antihypertensive agent. Logistic regression analysis was used to assess the association between the discontinuation of antihypertensive agents and the risk of AMI and to control for confounding. Results: We included 1245 cases and 4994 controls in our analysis. The risk of AMI was significantly increased with all stoppers of beta-blockers (adjusted OR: 1.54, 95%CI (1.25-1.90)), calcium channel blockers (CCBs) (adjusted OR: 2.25, 95%CI (1.53- 3.30)), and diuretics (adjusted OR: 1.76, 95%CI (1.24-2.48)) compared with current users. Moreover, the risk of AMI was significantly increased for longterm stoppers (beta-blockers, CCBs, angiotensinconverting enzyme inhibitors, and diuretics) and intermediate- term stoppers (beta-blockers and CCBs) versus current users. There was no difference in AMI risk between recent stoppers of antihypertensive agents versus current users. Conclusions: Discontinuation of antihypertensive agents increases the risk of AMI after more than 90 days of stopping. Adherence to antihypertensive agents plays an important role in reducing the risk of AMI in patients with hypertension.Background: It has been reported that patients with type 1 diabetes (T1DM) have a decreased lung function. Studies on the association of T1DM and asthma in children show controversial results. Objectives: The aim of this study was to quantify asthma medication use in children and adolescents with and without (reference cohort) T1DM 5 years before and after the onset of diabetes. Methods: A population-based cohort study was conducted in the Dutch PHARMO Record Linkage System. All children (1. Modelling of Endpoint Postponement for All-Cause Mortality in Statin Trials Morten Rix Hansen, Anton Pottegård, Asbjørn Hróbjartsson, Per Damkier, René D Christensen, Kasper Søltoft Larsen, Malene EL Kristensen, Palle M Christensen, Jesper Hallas. Clinical Pharmacology, University of Southern Denmark, Odense, Denmark; Department of Clinical Chemistry and Pharmacology, Odense University Hospital, Odense, Denmark; The Nordic Cochrane Centre, Rigshospitalet, Copenhagen, Denmark; Research Unit for General Practice, University of Southern Denmark, Odense, Denmark. Background: The average postponement of the outcome event has been proposed as a novel method to present the magnitude of effect for preventive medications. This measure has been shown to have better agreement with patient preferences than conventional outcomemeasures, including the “number needed to treat” (NNT), possibly because it is more intuitively understood. For some interventions, it may also provide a better theoretical frame for how benefit is distributed among participants than the NNT measure. The aim of this study was to present a novel method for modelling endpoint postponement (EP) from trial data and compare it with the usual approach of measuring the area between survival curves. We also present a formalized meta-analysis of modelled EP for all-cause mortality in statin trials. Methods: We identified 17 placebo-controlled statin trials that fulfilled our inclusion criteria. Eleven of these presented Kaplan–Meier curves for all-cause mortality. Average EP was calculated as the area between Kaplan–Meier curves by counting pixels on magnified prints for these 11 trials. The modelled EP was computed for all trials on the basis of (1) hazard ratio, relative risk or odds ratio; (2) the cumulative event rate in the untreated group; and (3) the trial’s running time. The underlying assumption was that the mortality was reasonably stable within the trials’ running time. The modelled EP was subjected to a meta-analysis, using inverse variance weighting in a random effect model. Results: EPswere generally small for estimates based on pixel-counting, 10 and 27days for trials both primary and secondary intervention that typically ran over 1.9– 6.1years. The modelled EPs varied between 2 and 34days. The difference between modeled EP and EP based on pixel-counting was between 8 and 12days. The results of the meta-analyses will be presented at the meeting. Conclusions: Based on these trial data, statin treatment results in a surprisingly small gain in average survival. Our modelled EP estimates agreed reasonably with EPs based on pixel-counting. The modeled EP is amenable to meta-analyses and may be a useful approach to presenting the benefit of preventive treatment. 2. Permanent User Bias in Case–Crossover Studies in Pharmacoepidemiology Jesper Hallas, Shirley V. Wang, Joshua J. Gagne, Sebastian Schneeweiss, Anton Pottegård. Clinical Pharmacology, University of Southern Denmark, Odense C, Denmark; Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA. Background: In pharmacoepidemiology, the case– crossover design is based on cases that have contrasting drug exposure at the time of an event and at a reference time in the past. If the drug in question should be taken permanently, only certain exposure patterns will occur. These patients cannot be unexposed at the event time and exposed at the reference time, while the opposite pattern can occur if the drug was initiated recently. The resulting odds ratio (OR) would thus be biased upward. As many drugs have a subpopulation of permanent users, this bias might pervade many case–crossover analyses of drug effects. Objectives: The aims of this study were to demonstrate this “permanent user bias” and to evaluate whether it can be remedied by including a control group (case–time–control design). Methods: Using nationwide Danish data resources, we conducted case–crossover and case–time–control analyses for combinations of three exposures that are

Sa1960 Risk of Upper Gastrointestinal Complications and the Use of Individual NSAIDs: A Nested Case-Control Study From the Sos Project

Background . Proton pump inhibitors (PPIs) are commonly coprescribed to reduce gastroduodenal injury caused by Non Steroidal Anti-Inflammatory Drugs (NSAIDs). However, while suppression of gastric acid secretion by PPIs is highly effective in reducing gastric injury, PPIs might worsen intestinal injury caused by non selective COX inhibitors. This so-called NSAIDs-PPIs enteropathy has been linked to development of an altered microbiota. Aims. To investigate whether probiotics prevent development of intestinal injury in a model of NSAIDs-PPIs enteropathy.Methods. Mice administered with omeprazole 10 mg/kg (ip twice daily) for 9 days, treated with naproxen (50 mg/kg per os) on the final 4 days. The probiotic VSL#3 at the dose of 18 x109 UCF daily (per os) was administered for 9 days. Small intestinal damage was scored, changes in hematocrit, gastric pH, intestinal expression of IL-1 β, and TNFαmRNA were measured. Change in small intestinal microbiota was assessed by terminal restriction fragment lenght polymorphisms (T-RFLP). To produce terminal restriction fragments the 16S rRNA was amplified and digested with HaeIII. Results. Treating mice with omeprazole increased significantly gastric pH from 2,43±0,29 to 3,66±0,2 (naproxen +omeprazole group) and 3,9 ±0,18 (naproxen + omeprazole +VSL#3 group) and reduced significantly the severity of gastric injury score (mm of lesion) from 20,17±4,8 (naproxen group) to 6±1,15 (naproxen+ omeprazole group) and 7±2,4 (naproxen+ omeprazole+VSL#3 group). In contrast, intestinal injury (mm of lesion) caused by naproxen was exacerbated by omeprazole administration from 37,5±12,12 (naproxen group) to 84,17±11,56 (naproxen + omeprazole group; P,0.05). This effect was reversed by VSL#3. Naproxen alone and naproxen + omeprazole increased the ileal expression of IL-1 β and TNF α, two markers of mucosal inflammation. This effect was reversed by VSL#3. Finally, VSL#3 administration resulted in a robust colonization of small intestine as assessed by T-RFLP. Conclusions. Omeprazole exacerbates naproxen -induced intestinal damage at least in part because of significant shifts in enteric microbiota. VSL#3 has been shown effective in reducing intestinal injury caused by indomethacin in human (APT32:209) and might hold utility in the prevention of NSAIDs-PPIs enteropathy.