ISoP October 2015, Prague

Conference Abstracts
15th Annual Meeting of the International Society of Pharmacovigilance (ISoP 2015), Prague, October 2015


Abstract ID: ISOP-0009
Utilisation and safety of deferasirox (Exjade®): Results from an observational cohort study in England. V. Osborne, M. Davies, D. Layton, S. Shakir.
Background: Deferasirox is an oral iron chelating agent (ICA) primarily used to reduce chronic iron overload in patients (pts) receiving blood transfusions for various chronic anaemias and some non- transfusion dependant anaemias. Use in patients 2yrs+ is licensed for certain indications. An identified safety concern is increased serum creatinine (Cr) during treatment (Rx); monitoring is therefore recommended prior to and during Rx
Aim: To examine the utilisation and safety of deferasirox used in general practice in England
Methods: Single exposure observational cohort study. Pts identified from dispensed prescriptions for deferasirox. Prescriptions collected Sep06-Sep14. Outcome data collected via postal questionnaires sent to prescribers ≥6mths after 1st dispensed prescription, including information on prior Cr measurements and prior use of alternative ICAs. Summary descriptive statistics calculated
Results: Evaluable cohort=122 pts (2-17yrs=51, 41.8%); Median age=23yrs (IQR11-61); 58.2% male. Frequent reasons for prescribing (underlying conditions leading to iron overload): sickle cell anaemia (27/103 where specified,26.2%) and beta thalassaemia (BT) (26,25.2%); 53.8% BT pts had frequent blood transfusions (≥7ml/kg/mth packed red blood cells). Most pts (43/51,84.3%) were prescribed licensed doses of 10 or 20mg/kg/day at start. Rx initiated by a specialist for 100 pts (100/103,97.1%). 18 serum Cr values reported prior to Rx; 4 in excess of reference range [median value of all prior serum Cr 69μmol/L (IQR51-95)]. Events reported in these 4 pts included raised ferritin and renal function decline. In total, 91 incident events were reported, including 2 raised serum Cr after starting Rx. 45.6% pts (26/57) used an alternative ICA in the 12 months prior to Rx; 80.8% desferrioxamine.
Conclusions: These results show that deferasirox is largely being prescribed for its licensed indications in general practice in England and events reported were consistent with the known safety profile. These results contribute to post-marketing information. However, considering the small cohort size, any conclusions from this study should be put into context with results from other studies.

Abstract ID: ISOP-0010
Utilisation of a once weekly injection for Type 2 Diabetes Mellitus: Interim results from an observational cohort study in England V. Osborne, N. Qayum, A. Coughtrie, D. Layton, S. Shakir.

Background: Bydureon® (exenatide) is indicated for the treatment (Rx) of Type 2 Diabetes Mellitus (T2DM) in combination with metformin, sulphonylurea (SU), thiazolidindione (TZD) alone, metformin & SU or metformin & TZD for patients (pts) who have not achieved adequate glycaemic control on maximally tolerated doses of these oral therapies alone. A post-marketing observational cohort study of Bydureon® was requested as part of the EU Risk Management Plan. The final target cohort size is 5000 pts. This study is ongoing
Aim: To describe the utilisation characteristics of pts prescribed Bydureon® at interim
Methods: An observational, population-based cohort design in primary care. Pts in the interim cohort were identified from all dispensed prescriptions for Bydureon® in England Sept11-Jan14 (interim data lock). Data were collected from prescribers via postal questionnaires sent ≥12mths after the 1st prescription was dispensed. Summary descriptive statistics were calculated
Results: Evaluable cohort at interim=520 pts; median age 58yrs (IQR 51-65); 58.1% male. T2DM and time since diagnosis was specified for 493 pts; the majority were diagnosed >10yrs prior to starting Rx (210, 42.6%). Where specified (n=441), 91.1% pts were classed as obese (BMI >30.0 kg/m2) immediately prior to Rx. On starting, 48.4% (134/277) had a HbA1c >9%, representing very poor diabetes control. Most pts had HbA1c ≥7.5% (223, 80.5%). The majority of pts used 2mg once weekly (489/492, 99.4%) and most pts were prescribed Bydureon® as either second line (158/477, 33.1%) or third line co-therapy (307, 64.4%). The majority of pts were prescribed Bydureon® with metformin (417/520, 80.2%). Co-prescribing of insulin was also reported (117, 22.5%)
Conclusion: These interim results characterise the utilisation of Bydureon® in primary care in England. The majority of pts had T2DM, used 2mg once weekly and were co-prescribed metformin. Most pts were obese which raises the possibility of channelling by prescribers due to the purported benefits of Bydureon® in weight loss. This interim analysis will be superseded when validation and follow-up are complete for the final analysis.

Abstract ID: ISOP-0168
Patient-generated data and digital engagement – Social Media: its impact and contribution to pharmacovigilance. E. Suggate, D. Layton, D. Brown.
Introduction: Patient adverse event reports (PAER) reports have been shown to positively contribute to the ongoing benefit: risk versus assessment of medicines. In response to the rapid expansion of social media, PAER from web-based sources have the potential to be exploited as alternative sources that may, if extracted and analysed appropriately, provide complementary data to that obtained from routine pharmacovigilance (PV) activities.
Aim: To collect and assess the attitudes of PV professionals regarding the utilisation of safety data from social media for PV purposes.
Methods: A cross-sectional survey was conducted between August to September 2014 inclusive using web-based tool (Survey MonkeyTM) using a format of open/closed and Likert scale based questions, to gather anonymous information at an international level of PV professional characteristics (profession, work place and PV experience) and opinions on the value, quality, benefits and challenges of utilising PAER data obtained via social media. For PV professionals working in pharmaceutical industry (pharma), information was also requested on operational activities such as how such data were being gathered. The University of Portsmouth Science Faculty Ethics Committee (ref 17/7/14) provided study approval. Eligible participants comprised 5329 PV professionals from pharma, regulator and academic backgrounds registered on a mailing contact database. Data confidentiality was assured by the survey being managed by the database manager independent of the investigators. Data analysis comprised qualitative thematic evaluation and quantitative descriptive statistics.
Results: 197 participants (3.7% eligible) responded from a wide range of professional backgrounds – the most frequent being from academia (n=44, 22.3% responders), from pharmaceutical industry sector (n=143, 72.6%), and with less than 10 yrs experience (n=118, 59.8% responders). Where specified, 74.8% (107/143) believed PAER data obtained via the internet have the potential to contribute valuable information to the benefits and risks of medicines in a real-world context; 68.2% (118/173) agreed that such data provided an opportunity to increase the amount of safety data and also specifically inform on how adverse events affect quality of life. Concerns included data quality, burden on resourcing and relevant guidance. Of 112 PV pharma professionals, 34 (30.4%) reported awareness of their company engaging in activities such as searching non-company sponsored websites (digital ‘listening’).
Conclusions: Internet derived patient-generated data clearly has the potential to contribute to ongoing risk versus benefit assessment of medicines. However a number of issues still need addressing. Continual improvement of methodological techniques is required with relevant regulatory driven guidance with regards to PAER signal detection purposes.

Abstract ID: ISOP-0169
Observational Assessment of Safety in Seroquel (OASIS) – rates and patterns of common events
D. Layton, S. Clarke, I. Ratcliffe, S. Shakir, T. Hale.
Introduction: The Observational Assessment of Safety in Seroquel (OASIS) study (ENCePP Study reg.5412) aimed to extend the post-authorisation safety knowledge of quetiapine extended release (Seroquel XL©) in the mental health secondary care setting, as prescribed by psychiatrists in new user adult (>18 years) patients with Schizophrenia or Bipolar Disorder compared to quetiapine immediate release (IR). The primary focus was on short-term (12-week) safety and high dose (>600mg/day) use.
Aim: To quantify event incidence and pattern after starting treatment and compare common event rates between users defined by dose and formulation
Methods: An observational cohort design using Specialist Cohort Event Monitoring (SCEM). Questionnaires completed by specialists collected data on patient characteristics, exposure and events between December 2009 and December 2012. Exposure (patient-weeks) was calculated for: total cohort and sub-group stratified by formulation. Period specific exposures were calculated for three 4-week periods for total cohort. Period specific exposures were further stratified by formulation and high dose use (where >600mg/day for >50% of each 4-week period). Crude event Incidence Densities (ID) per 1000 patient-weeks and ID differences (IDD+95%CI) were calculated within and between groups over total study period and 4 week exposure periods ; IDD 95%CI excluding the null (0) were considered signals of events associated with starting treatment and/or high dose.
Results: SCEM cohort comprised 845 patients; 471 (59%) were female and median age was 39 years (IQR 29,49); 40% (n=338) had Schizophrenia, 52% (n=442) had Bipolar disorder; 8% (n=65) had Other indications. High dose use was reported for 4% (28/631) in XL group and <1% (3/214) in IR group. The most frequently reported events weeks 1-12 inclusive (not associated with indication) were: sedation:ID 24, somnolence:ID 20, akasthesia: ID 3, and parkinsonism: ID 2. In the XL group, the IDDs for these 4 events were non-significantly lower for high vs standard dose for total study period and weeks 1-4. Within XL high dose group, sedation and somnolence were associated with starting treatment (IDD w1-4 – w5-8 : 25 (7,42) and 14 (14, 15) respectively); this pattern was also observed within XL and IR standard dose groups. Low counts in high dose group and IR cohort precluded reliable comparisons.
Conclusions: This study found that sedation & somnolence were common events associated with starting treatment, but not with high dose. Although the frequency of high dose use was low, OASIS provides important information on the safety and utilisation of Seroquel XL©.

Abstract ID: ISOP-0051
Potential for underdosing of antipsychotics in primary and mental health care: findings from post- authorisation safety studies on Seroquel XL© D. Layton, V. Osborne, M. Davies, I. Ratcliffe, S. Clarke, S. Shakir, J. Reilly, A. Hale.
Introduction: UK guidelines state that the lowest possible dose of antipsychotics should be used and titrated to the lowest effective dose. A risk management plan of quetiapine extended release (Seroquel XL©) had a need to describe long-term (12+ months) and short term (12 weeks) use and safety in primary and mental-health care setting, respectively. An M-PEM study monitored use and safety in all indications. A SCEM study monitored early-onset use and safety during Seroquel XL© titration and at higher doses (>600mg) in Schizophrenia and Bipolar Disorder adults (>18 years); a comparator group (immediate release quetiapine) was also included (ENCEPP Study 5412). M-PEM and SCEM study objectives included exploring posology.
Aim: An adhoc analysis to describe prevalence of potential underdosing in clinical practice.
Methods: Exposure, selected prior medical history and medications use data were collected for each study from forms sent to hospital specialists for SCEM December 2009 to December 2012 and to primary care physicians (GPs) for M-PEM September 2008 to February 2013. Descriptive statistics were calculated; doses were converted to percentage of relevant maximum dose according to each indication and titration stage as specified in the Summary of Prescribing Characteristics (SPC).
Results: In the M-PEM cohort (N=13276), at start of treatment potential underdosing was calculated for: 37%(785/2136) Schizophrenia patients, 3% (98/3500) Bipolar Disorder patients, and 6%(147/2646) patients with Major Depressive Disorder. At date maintenance treatment regimen was reportedly achieved, potential underdosing was calculated for 38% (509/1339) Schizophrenia patients, 33% (721/2165) Bipolar Disorder patients, and 32% (531/1648) with Major Depressive Disorder. In the SCEM XL cohort (n=646), at start of treatment potential underdosing was calculated for 56% (144/258) Schizophrenia patients and 59% (204/345) Bipolar Disorder patients. At date maintenance treatment regimen was reportedly achieved, potential underdosing was calculated for 86% (223/258) Schizophrenia patients and 91% (315/345) Bipolar Disorder patients.
Conclusions: Both studies found that potential underdosing occurred very commonly in all indications studied. Start dose data correlated poorly with SPC and expert guidelines to use lowest effective dose, but corresponded to UK prescribing guidelines which do not recommend excessive doses, unless other evidence–based strategies have failed. Possible explanations for more common potential underdosing in SCEM vs M-PEM is that patients treated by specialists may require more individualised therapy (including use of immediate release quetiapine) to initially stabilise their condition, whilst GPs tend to manage patients at later stages. Further work will explore impact of age and prior/concurrent psychotropic use on underdosing.

Abstract ID: ISOP-0047
Defining risk profiles in special populations -results from a post authorisation safety study of Seroquel XL© conducted in England. D. Layton, V. Osborne, M. Davies, S. Shakir
Introduction: A risk management plan of quetiapine extended release (XL) had a need to describe use & monitor long-term (12+ months) safety in the primary care setting. Its aim was to monitor for known risks (e.g. metabolic effects) and safety signals in patients of all indications. One of the study objectives included monitoring specials populations who may be more vulnerable to risks due to past medical history and/or concurrent conditions or medication use.
Aim: To describe risk profiles of special population groups (by indication, past medical history, prior psychoactive drugs, and age ≥65,<64 years)
Methods: An observational, single-exposure cohort design. Data were derived on exposure from dispensed prescriptions; on events from forms completed by physicians 12 months post first exposure. For general surveillance crude incidence densities (ID) per 1000 patient-months were calculated for events in months (m) 1, m2-6, m7-12 and m1-12 inclusive for cohort, and by special population group. Overdose events were grouped by suicidal ideation, suicidal behaviour, self- injurious behaviour. Event ID differences (IDD m1- m2-6 ) + 95%CI excluding the null (0) were signals of starting treatment. Survival methods estimated rates of overdose and hyperglycaemic events (+95%CI) per 1000 patient-months (where n>7 and excluding patients if event dates were missing).
Results: Cohort comprised 13276 patients; 59% female; median age 43 (IQR 33,55). Events with highest ID m1-12 inclusive were: Sedation, n=317: ID 40; Somnolence,n=288: ID 47; and Weight increased,n=198: ID 54. Somnolence & Sedation were associated with starting treatment overall (IDD m1-m2-6 :3.8(1.5,5.7) and 2.5(0.8,4.2) respectively) and in special population groups with Bipolar Bisorder or Non-licensed (off-label) indications; age ≤64 years; and a history of depression. For hyperglycaemic events, the highest rates in months 1-12 inclusive were: raised random blood sugar (n=122):1.0 (0.8,1.2) and raised fasting plasma glucose (n=104):0.8 (0.7,1.0). Patients with a history of impaired glucose tolerance and diabetes appeared to have highest rates of hyperglycaemic events.Of 104 overdose events, a history of mental illness was reported for all patients; event rates in months 1-12 inclusive were: suicidal behaviour(n=84):101.8 (82.2,126.1), suicidal ideation (n=1):4.0 (0.6,28.4), and self-injurous behaviour (n=15): 36.6 (22.0,60.1).
Conclusions: For frequently reported central nervous system, psychiatric and hyperglycaemic events, many patients had pre-existing risk factors that are likely to put them at elevated risk. This study demonstrates the ongoing importance of observational studies to support the risk:benefit evaluation of medications.

Abstract ID: ISOP-0048
Use of antipsychotics in mental health secondary care setting versus primary care: results from two post-marketing safety studies D. Layton, V. Osborne, M. Davies, I. Ratcliffe, S. Clarke, S. Shakir, J. Reilly, A. Hale.
Introduction: Generalisability of results from Modified Prescription-Event Monitoring (M-PEM) studies conducted in primary-care may be limited by excluding patients with complex characteristics (underlying/concurrent disease, medications) exclusively managed in hospital. Specialist Cohort Event Monitoring (SCEM) applies to secondary care. A risk management plan of quetiapine extended release (XL) had a need to describe use and monitor long-term (12+ months) and short term (12 weeks) safety in primary and secondary care, respectively. An M-PEM study monitored safety and use (all indications). A SCEM study monitored early-onset events during titration and at higher doses (>600mg) in Schizophrenia and Bipolar Disorder adults (>18 years) for whom the decision to treat was made by psychiatrists; a comparator (immediate release (IR) quetiapine) was included (ENCEPP Study 5412).
Aim: To explore the potential of bias by comparing M-PEM and SCEM patient characteristics.
Methods: Both studies used an observational cohort design. MPEM data derived from dispensed prescriptions September 2008-February 2013 and questionnaires sent to GPs 12+ months after each patient started treatment. SCEM data derived from questionnaires completed by psychiatrists 12+ weeks after each patient started treatment December 2009-December 2012. Descriptive statistics and Odds Ratios (OR) +95%CI (exact method) were calculated.
Results: SCEM XL cohort (n=646) included 258 (39.9%) with Schizophrenia, and 345 (53.4%) with Bipolar Disorder. M-PEM cohort (n=3276) included 2362 (17.8%) adults with Schizophrenia, 3820 (28.6%) with BD. In Schizophrenia, SCEM patients were significantly more likely than M-PEM patients to be < 30 yrs old [84vs488; OR 1. (1.2,2.2)], have a history of: depression [152vs446; OR 6. (4.8, 8.1)], extrapyramidal symptoms (EPS) [46vs103; OR 4. (3.2,7.0)] and prior antipsychotic use [175vs779; OR 4.2 (3.2,5.6)]; recent (<28 days prior) IR use was less likely [18vs632; OR 0.2(0.1,0.3)]. In BD, SCEM patients were more likely than M-PEM patients to have a history of: depression [258vs1344; OR 5. (4.2,7.0)], EPS [34v66; OR 6.2(3.9,9.7)], diabetes [37vs168; OR 2.6 (1.7, 3.8)] and prior antipsychotic use [182vs772; OR 4.4(3.5,5.5)]; recent IR use was less likely [17vs985; OR 0.2(0.1,0.2)].
Discussion: SCEM patients had a higher prevalence than M-PEM patients of depression, EPS, diabetes and antipsychotic use which are risk factors for some adverse events. Considerations include differences in: the data available within medical records; prescribing guidelines; method of identifying patients; and overlap of study populations. These findings suggest that selection bias may exist which may affect generalisability of primary-care based study results to all treated patients, thus there is a need for surveillance in secondary care.

Abstract ID: ISOP-0049
Antipsychotic use in older adults with dementia: results from a Post-Authorisation Safety Study. D. Layton, M. Davies, V. Osborne, S. Shakir
Introduction: The UK National Institute for Health & Care Excellence recommends that antipsychotics can be used in elderly patients under strict guidelines, however use is associated with serious safety concerns (including cerebrovascular accidents (CVA)). A Modified Prescription-Event Monitoring (M- PEM) study was conducted as part of the Risk Management Plan for Seroquel XL© to examine it’s safety and use, irrespective of indication, as prescribed in primary care in England.
Aim: An adhoc analysis to examine the risk of CVA in the elderly.
Methods: M-PEM uses an observational cohort design; data on exposure were derived from dispensed prescriptions that had been issued by primary care physicians (GPs) between September 2008 and February 2013; data on events were derived from forms completed by GPs sent 12+ months after the date of each patient’s first prescription. Age and sex adjusted Mantel-Haenszel Odds Ratios (ORs) plus 95% Confidence Intervals (CI) were calculated for all cause deaths and CVA (MedDRA PT: CVA, Cerebellar infarction, Cerebral haemorrhage, Haemorrhagic stroke) in elderly patients with or without dementia, and with or without psychosis.
Results: Final elderly cohort comprised 3127 patients; median age 77 yrs (IQR 69,84); 62% (n=1940) female; 29% (n=892) had indications associated with dementia, of which 17% (n=148) had concomitant psychosis. Within 12 months of starting Seroquel XL© 10% (n=301) died; commonly from bronchopneumonia (n=44). Deaths were more likely in elderly with dementia than without [15% (136/892) vs 8% (165/2070); adjOR 1.5(1.2, 1.9)] but not for dementia patients with psychosis versus those without [14% (21/148) vs 15% (115/744); adjOR 1.0 (0.6,1.6)]. At least one report of CVA was reported in 23 (<1%) elderly patients, of which 17 had a fatal outcome. CVA events were twice as likely in patients with dementia than without [48% (11/23) vs 28% (881/3104); adjOR 2.8 (1.3, 6.2).
Conclusions: Approximately one-third (29%) of this elderly cohort had dementia with or without psychosis. A higher rate of death was observed in elderly patients with dementia than without, but concomitant psychosis with dementia did not appear to be a risk factor. CVA was uncommon (<1%), but more likely in elderly with dementia than those without. Study limitations include low CVA counts, possible misclassification of depression and delirium as dementia and limited information on other possible factors (other modifiable medical and environmental factors).

Abstract ID: ISOP-0052
Applying the Ready Reckoner tool for assessing antipsychotic prescribing within a post- authorisation safety study. D. Layton.
Introduction: In the UK, the Ready Reckoner algorithm is a tool used in psychiatric clinical practice to monitor antipsychotic dosing in patients with complex antipsychotic regimens. For each patient, the dose of each antipsychotic can be converted to percentage of relevant maximum dose according to each indication and titration stage as specified in the Summary of Prescribing Characteristics (SPC). Total dose percentage (TD%) >100% identifies patients at risk. A nested case control (NCC) study was undertaken as part of a post authorisation safety study (PASS) program to explore possible dose- event response for somnolence/sedation (SS) after starting a new formulation of an antipsychotic, with a focus on evaluating total daily dose (TDD)> 600mg.
Aim: An exploratory analysis to evaluate use of the Ready Reckoner algorithm as an alternative indication-adjusted dose metric to model antipsychotic treatment effects.
Methods: An incidence density matched NCC study used data from a primary care based observational cohort study (N=13276) identified between September 2008 and February 2013. Of 756 cases, 212(28%) were randomly selected and 170 risk sets created. For all subjects, the reported TDD (start, maintenance, and event) were converted to TD% in accordance with the Ready Reckoner algorithm; for off label indications the SPC dose range for major depression was used. Fractional polynomial (FP) logistic models explored functional form; empirical and fitted within-person Ordinary least Squares (OLS) dose trajectories described patterns over time.
Results: SS cases tended have higher reported TDD vs controls at start [median 300 (IQR 200,600) vs 200mg (IQR 100,300), p<0.01] but the inverse was seen when the TD% metric was applied (median 50% (IQR 16,100) vs 75% (33,100), p=0.02). At maintenance & event, similar relationships were observed for reported TDD and TD% variables. SS risk was a negative function of TDD and TD%. No deviation from linear assumption was apparent for start and maintenance doses (FP2 vs linear, P>0.05), but was non-linear for TD% at event (p=0.04). In exploring pattern over time, the general trend was decreasing for both cases and controls, although slower for controls.
Conclusions: This exploratory analysis demonstrates the feasibility of the Ready Reckoner algorithm as an alternative method of calculating an indication-adjusted variable suitable for analysing dose- related effects, particularly where multiple indications may exist that use various dose ranges. Analytical advantages include avoiding creation of strata of small sample sizes. Limitations include possible metric under-estimation from missing data for other concomitant antipsychotics.

Abstract ID: ISOP-0040
Application of the International Society on Thrombosis and Haemostasis (ISTH) classification of major bleeds within a Post Authorization Safety Study. Miranda Davies, Vicki Osborne, Deborah Layton and Saad Shakir.
Background: Haemorrhage is a frequent complication of anticoagulant (AC) use. In order to compare incidences between trials a definition of major bleeds (MB) in non surgical studies was developed by the International Society on Thrombosis and Haemostasis (ISTH) in 2005. In recent years, the Committee for Medicinal Products for Human Use (CHMP) has recommended its use in studies for prevention of stroke and systemic embolic events (SEE) in patients (pts) with non valvular atrial fibrillation (NVAF) and prevention of deep vein thrombosis (DVT) and pulmonary embolus (PE). Bleeds reported in a Specialist Cohort Event Monitoring (SCEM) study on the oral AC rivaroxaban (conducted as part of Risk Management Plan) will be classified using this definition.
Objectives: To describe the methodological considerations of applying this definition to observational data.
Methods: Aim to collect data on 1700 pts treated for the prevention of SEE [n=561], and the treatment and prevention of recurrent DVT and PE [n=1005]. Recruitment Sep2013-2016. Information (info) was obtained on bleeds that occurred during initial 12 weeks; criteria for MB included: a fall in Hb of ≥ 2 g/dL, a transfusion of ≥ 2 units, critical organ site, or fatal outcome. Bleeds will be classified as clinically relevant non major (CRNM) if none of the MB criteria were met, but if medical attention was required and/or a change in antithrombotic therapy and/or any other bleed with clinical consequences.
Results: To minimise misclassification, supplementary info will be used to validate and confirm the type, obtain missing data and further details of the bleed (site, management, and outcome). All bleeds will be adjudicated by an expert, and interim results will be published.
Conclusions: By systematically applying the ISTH definition, we hope to gain better understanding of the type of bleeds reported in a cohort of AC users, associated risk factors and outcome details. This should enable more meaningful comparisons to be made between major and CRNM bleeding incidences obtained in this setting with those observed during trials.

Abstract ID: ISOP-0215
Development of a qualitative method to assist in interpretation of results from disproportionality analyses in the SAFEGUARD project. Lorna Hazell, Niklas Schmedt, Lorenza Scotti, Ingrid Leal, Gianluca Trifiro, Miriam Sturkenboom and Saad Shakir.
Introduction: Disproportionality analyses are used in spontaneous reporting systems to support signal detection. Limitations of these systems, however, present challenges for interpretation and comparisons with other types of data. The purpose of this study was to develop an output suitable for integration with other data, specifically epidemiological studies investigating the safety of diabetes drugs in the context of the SAFEGUARD project.
Aim: To categorise associations from disproportionality analyses in the SAFEGUARD project and assess their level of uncertainty.
Methods: Disproportionality analyses were performed for 29 diabetes drugs and 9 outcomes of interest using 8 different strategies: two databases (FAERS and Eudravigilance), two outcome definitions (broad and narrow) and two reference groups (all drugs and diabetes drugs only). An algorithm was defined, depending on which analysis strategies detected disproportionality, to categorise each drug-outcome pair (n=261) as low, medium, high or unclear evidence of association. Each categorisation was qualitatively assessed using 9 ‘uncertainty factors’: 1) specificity of outcome definition, 2) potential for stimulated reporting, 3) time on the market, 4) consistency between databases, 5) multiple testing, 6) reporter status, 7) level of drug use, 8) masking bias and, 9) use of statistical thresholds. An overall ‘uncertainty score’ for each categorisation from 0 (low uncertainty) to 1 (high uncertainty) was derived from the proportion of the 9 uncertainty factors that applied.
Results: Twenty-three pairs (8.8%) were classified as high evidence of an association, 48 (18.4%) as medium, 122 (46.7%) as low and 68 (26.1%) as unclear. Examples of those categorised as ‘high’ are shown in Table 1.
Table 1 Examples of Pairs Categorised as ‘High’ Evidence of Association with their Uncertainty Scores

Drug-outcome pair Uncertainty Scores
Incretin-based therapies & pancreatic outcomes From 0.25 for liraglutide to 0.625 for vildagliptin
Glitazones & heart failure 0.57 for pioglitazone; 0.14 rosiglitazone
Pioglitazone & bladder cancer 0.14
Rosiglitazone & cardiovascular outcomes 0.14

Conclusion: Disproportionality analyses should be interpreted in the context of their limitations due to potential biases in the data. We have used an uncertainty score, as a measure of these limitations, to qualify the categorisation of the associations detected. The assessment criteria were not exhaustive, were topic-specific and required considerable computation. Thus further refinement and evaluation is required before such methods could be recommended for routine use. Nevertheless the categorisation and uncertainty score provide an output that can be compared and integrated with similarly formatted results from independent studies.


Abstract ID: ISOP-0050
Evaluating prescriber concordance with prescribing: results from a post-marketing study in primary care setting. D. Layton, N. Qayum, C. Doe, S. Freemantle, S. Shakir.
Introduction: At launch (2005), ivabradine (Procoralan©) was indicated for treatment of chronic stable angina pectoris in patients with normal sinus rhythm, with a contraindication and/or intolerance for beta-blockers. A drug utilisation study was conducted to support risk management of the drug. Study objectives included examining contraindications, warnings for use and incidence of two adverse events: visual phosphenes and bradycardia (persistent heart rate <50bpm).
Aim: To develop a tool for use to support assessment of prescribing discordance with prescribing recommendations and explore impact on phosphenes and bradycardia incidence.
Methods: An observational single exposure cohort study. Exposure data were collected from dispensed prescriptions issued by primary care physicians November 2005 and May 2009; outcome data (including pre-existing patient medical conditions and/or drugs which carried contraindications and/or warnings for use) from forms sent to physicians ≥6 months after each patient’s first prescription. An algorithm based prescribing framework was developed and assisted with the assessment of available information according to the contraindications and warnings for use in the Summary of Product Characteristics. To explore impact of prescriber discordance, risk (%; + 95% Confidence Interval (CI)) of phosphenes and bradycardia were calculated by framework group (≥1 contraindication; ≥1 warnings for use; combination of >1 contraindication/>1 warning for use; concordant; unknown). Descriptive statistics were also calculated.
Results: Final cohort comprised 4624 patients, median age 68 years (IQR 60, 77); 57% (n=2663) male. Contraindications and warnings for use were assessed for 3357 patients: 74% (n=2491) were concordant, 21% (n=701) had warnings for use, 4% (n=124) had contraindications and 1% (n=41) had a combination of contraindications and warnings for use. Bradycardia was reported for 96 patients (2.1% (1.6, 2.5) cohort) of which 73 were assessed by framework group: n=4 (3.2% (0.9, 8.3)) were contraindicated, n=17 (2.4% (1.4, 3.9)) had warnings for use, n=1 had combination of both, and n=51 (2.0% (1.5, 2.7)) were concordant. Phosphenes were reported for 140 patients (3.0% (2.5, 3.6) cohort), of which 104 were assessed by framework group: n=2 (1.6% (0.2, 5.8)) were contraindicated, n=23 had warnings for use (3.3% (2.1, 4.9)), n=1 had a combination of both, and n=78(3.1% (2.5, 3.9)) were concordant.
Conclusions: In this study, the incidence of bradycardia and phosphenes was common in all concordance groups, although estimates lacked precision. This study demonstrates the feasibility of using a framework to assess prescribing discordance as reported in drug utilisation studies to identify at-risk populations, which may help support post-marketing risk:benefit evaluations.