22

Prescription–Event Monitoring (PEM): The Evolution to the New Modified PEM and its Support of Risk Management DEBORAH LAYTON AND SAAD SHAKIR Drug Safety Research Unit, Southampton, UK and University of Portsmouth, Portsmouth, UK

BACKGROUND profit organization that operates in association with the University of Portsmouth. The recognition that not all hazards could be The system’s key objective at inception in the known before a drug was marketed and that spon- 1980s was to recruit the first 10 000 patients who taneous adverse drug reaction (ADR) reporting received a new drug of interest so that any systems have limitations to identify all hazards led adverse event that occurred in more than one in to several proposals for schemes based on the iden- 1000 patients would be reliably identified. Based tification of patients by means of prescription data. on the success of these standard PEM studies These schemes were largely intended to provide over a period of 30 years, this methodology has information on populations of known size so that subsequently evolved in response to the require- the incidence of adverse reactions could be esti- ments for risk management of medicines to facil- mated with reasonable accuracy. This led to the itate more targeted safety surveillance. This has founding of a prescription-based monitoring been achieved through the technique known as system to monitor events regardless of relatedness modified (M)-PEM, which retains all the strengths to drug exposure (Prescription-Event Monitoring of the standard method with the same underly- (PEM)) by W. H. W Inman and the establishment ing process but also tries to overcome some of of the Drug Safety Research Unit (DSRU). The its limitations (further details in following DSRU is an independent registered medical non- sections).

Mann’s Pharmacovigilance, Third Edition. Edited by Elizabeth B. Andrews and Nicholas Moore. © 2014 John Wiley & Sons, Ltd. Published 2014 by John Wiley & Sons, Ltd. Companion Website: www.wiley.com\go\mann\pharmacovigilance 360 MANN’S PHARMACOVIGILANCE

GENERAL PRINCIPLES The evaluable cohort is also regarded as an inception cohort (where study drug is a new entity) STUDY DESIGN or a new user cohort (e.g., where the drug under study might be a new formulation or new indica- The general methodology common to both stand- tion). Here, the observation period begins as soon ard and modified approaches uses a retrospective as the patient starts the medication, which is par- noninterventional observational cohort design to ticularly important if the risk of an event is higher provide active surveillance of targeted medicines on in the early period after starting therapy. An advan- a national scale in England. Data analysis utilizes tage of an inception cohort is that potential con- several approaches that combine the application of founding factors can be measured before treatment epidemiological methods with medical evaluation starts and adjusted for in subsequent statistical to provide estimates of prevalence of selected drug analysis. Unlike standard PEM, the M-PEM utilization characteristics, incidence rates for events methodology offers greater scope to collect this reported in the exposed cohort ,exploration of risk baseline data. profiles within different subpopulations, and also A wide range of drugs have been studied using provides the opportunity for further clinical evalu- one of the two approaches, including agents to treat ation of selected events of interest. Details of the hypertension, angina, asthma and chronic obstruc- methodology of standard PEM, including the tive pulmonary disease (COPD), diabetes, epilepsy, methods of data coding, computerization, and depression, schizophrenia, erectile dysfunction, analysis, have been provided in a number of publi- urinary incontinence, and a number of nonsteroi- cations, and thus are not covered in detail in this dal anti-inflammatory drugs (including selective chapter (Freemantle et al., 1997; Layton and Shakir, COX-2 inhibitors). Thus, the drugs included in the 2011). The majority of all new studies are now system are those intended for widespread, long- constructed using the M-PEM approach, and term use, special emphasis being given to drugs for examples of bespoke analytical requirements neces- which treatment is likely to be both initiated and sary to achieve an M-PEM’s study aims and objec- continued by the general practitioner (GP) (Anon., tives are provided later. 1986; BMA, 2006). In addition to drugs that are The eligible cohort is identified based on a single taken regularly, it has also been possible to study common exposure identifier (a prescription for the products that are not used daily, such as sildenafil new medication under surveillance). The method is for erectile dysfunction (Shakir et al. 2001). noninterventional because the decision to prescribe Of the 119 studies listed in Table 22.1, an average has already been taken and there are no additional of 55.5% of the 108 standard PEM questionnaires constraints on care imposed by subsequent partici- sent out have been returned by the GPs to the pation in the study. Confirmation of exposure DSRU with an average evaluable cohort size of status and outcome are ascertained retrospectively 10 947. For the 11 M-PEM studies given in Table to assemble the evaluable cohort (i.e., the cohort 22.1, an average of 58.8% of M-PEM question- available for analysis). The design is also longitudi- naires sent have been returned, with a smaller nal because health outcomes can be examined over average final evaluable cohort size of 6876 patients. a span of time. Furthermore, since prescription data collection begins immediately after the new DATA SOURCE drug has been launched (and covers the national population in England), evaluable patient cohorts Within the British National Health Service (NHS) can be accrued rapidly, which provides the oppor- structure, all individuals are registered with a tunity to detect safety issues as early as possible primary-care GP. Medical records held by the GP after market launch, a fundamental principle in are generally lifelong, transferable when a patient pharmacovigilance. Furthermore, the study design relocates, and include information on healthcare is highly dynamic, such that newly emerging safety consultations and interventions provided by both issues can be investigated while a study is in progress. primary and secondary care. The sampling frame is MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 361 hierarchical, comprising of two levels: all GPs in Business Services Authority (NHSBSA). This oper- England who prescribe the study drug and their ates for a length of time necessary for the DSRU patients. This wide coverage aims to provide an to collect a sufficient number of prescriptions to evaluable cohort that is representative of the whole identify the required study sample size. The population of patients who are registered with an NHSBSA receives remuneration from the DSRU NHS GP in England who take the study drug for this service. These data are reconciled with GP during the study period. identifier records available from the NHS Organisa- tion Data Services (ODS), to obtain prescriber contact details and, with existing records on the DATA COLLECTION PROCESS DSRU customized PEM database, to ascertain This occurs through a two-phased approach, which whether the data pertain to an existing eligible is summarized in Figure 22.1. The first phase is the patient already within the DSRU PEM database. It collection of prescription data to capture patient should be noted that all relevant prescriptions are and prescriber details, and the second is the collec- collected, irrespective of whether they are a new or tion of exposure and outcome data. repeat course. The first phase of the identification of prescrib- The second phase involves secondary use of ers and patients relies on data from dispensed NHS medical records data that have been entered into prescriptions. Prescription data are provided to the medical records as part of routine clinical care DSRU under long-standing arrangements and (EMA, 2012). For each eligible patient identified, through secure transmission, by a central NHS pre- a questionnaire is sent by post (according to chron- scription processing center, known as the NHS ological order of prescription issue date) to the

Table 22.1 List of 119 completed studies, by type (standard, modified).

Generic name Drug name Group Response (%) Final cohort

Standard PEM studies Enalapril Innovace ACE-inhibitor 68.3 15 361 Lisinopril Zestril+Carace ACE-inhibitor 63.5 12 438 Perindopril Coversyl ACE-inhibitor 53.4 9 089 Ramipril Tritace ACE-inhibitor 47.3 1 371 Doxazosin Cardura Alpha-blocker 60.1 8 482 Tamsulosin Flomax mr Alpha-blocker 57.4 12 484 Donepezil Aricept Alzheimer’s treatment 58.9 1 762 Tramadol Zydol Analgesic 55.8 10 532 Bupropion Zyban Antismoking aid 51.5 11 735 Fosfomycin Monuril Antibacterial 45.6 3 363 Terodiline Terolin Anticholinergic 69.6 12 444 Tolterodine Detrusitol Anticholinergic 59.0 14 526 Mirtazapine Zispin Antidepressant 56.0 13 554 Nefazodone Dutonin Antidepressant 54.9 11 834 Venlafaxine Efexor Antidepressant 54.6 12 642 Acarbose Glucobay Antidiabetic 62.8 13 655 Repaglinide Novonorm Antidiabetic 42.6 5 729 Troglitazone Romozin Antidiabetic 60.3 1 344 Rosiglitazone Avandia Antidiabetic 54.2 14 418 Pioglitazone Actos Antidiabetic 54.7 12 772 Nateglinide Starlix Antidiabetic 50.2 4 557 (Continued) 362 MANN’S PHARMACOVIGILANCE

Table 22.1 (Continued)

Generic name Drug name Group Response (%) Final cohort

Vildagliptin Galvus Antidiabetic 47.3 4 828 Lamotrigine Lamictal Anti-epileptic 67.9 11 316 Vigabatrin Sabril Anti-epileptic 69.2 10 178 Gabapentin Neurontin Anti-epileptic 66.4 3 100 Oxcarbazepine Trileptal Anti-epileptic 60.7 2 243 Fluconazole Diflucan Antifungal 68.6 15 015 Itraconazole Sporanox Antifungal 63.5 13 645 Acrivastine Semprex Antihistamine 56.5 7 863 Cetirizine Zirtek Antihistamine 57.4 9 554 Fexofenadine Telfast Antihistamine 50.9 16 638 Loratadine Clarityn Antihistamine 50.7 9 308 Desloratadine Neoclarityn Antihistamine 44.7 11 828 Levocetirizine Xyzal Antihistamine 49.2 12 876 Irbesartan Aprovel Antihypertensive 59.4 14 398 Losartan Cozaar Antihypertensive 59.9 14 522 Valsartan Diovan Antihypertensive 54.7 12 881 Sildenafil Viagra Anti-impotence 54.7 22 473 Apomorphine Uprima Anti-impotence 57.1 11 185 Tadalafil – Cohort 1 Cialis Anti-impotence 47.0 6 266 Tadalafil – Cohort 2 Cialis Anti-impotence 39.5 16 129 Vardenafil Levitra Anti-impotence 46.1 15 656 Sumatriptan Imigran Antimigraine 70.8 14 928 Tiotropium Spiriva Antimuscarinic bronchodilator 54.0 13 892 Iatropium Bromide Atrovent Antimuscarinic bronchodilator 63.0 13 211 Orlistat Xenical Anti-obesity 50.1 16 022 Sibutramine Reductil Anti-obesity 56.3 12 336 Olanzapine Zyprexa Antipsychotic 68.9 8 858 Risperidone Risperdal Antipsychotic 64.7 7 684 Sertindole Serdolect Antipsychotic 78.2 436 Quetiapine Seroquel Antipsychotic 59.1 1 728 Cisapride Prepulsid Antispasmodic 62.4 13 234 Aciclovir Zovirax Antiviral 74.1 11 051 Famciclovir Famvir Antiviral 65.4 14 169 Valaciclovir Valtrex Antiviral 64.1 12 804 Buspirone Buspar Anxiolytic 54.1 11 113 Nedocromil Tilade Asthma prophylaxis 68.1 12 294 Bambuterol Bambec Beta2 agonist 50.8 8 098 Eformoterol Foradil Beta2 agonist 52.9 5 777 Salmeterol Serevent Beta2 agonist 61.9 15 407 Betaxolol Kerlone Beta-blocker 54.7 1 531 Alendronate Fosamax Bone disease 59.4 11 916 Strontium Ranelate Protelos Bone disease 52.7 10 865 Amlodipine Istin Ca-antagonist 58.7 12 969 Diltiazem Tildiem Ca-antagonist 67.3 10 112 Isradipine Prescal Ca-antagonist 51.3 3 679 Mibefradil Posicor Ca-antagonist 54.1 3 085 Nicardipine Cardene Ca-antagonist 62.6 10 910 Cefixime Suprax Cephalosporin 39.6 11 250 Famotidine Pepcid H2-antagonist 51.8 9 500 Nizatidine Axid H2-antagonist 44.7 7 782 Zolpidem Stilnoct Hypnotic 49.0 13 460 Zopiclone Zimovane Hypnotic 54.8 11 543 MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 363

Table 22.1 (Continued)

Generic name Drug name Group Response (%) Final cohort

Xamoterol Corwin Inotropic 68.7 5 373 Nicorandil Ikorel K-channel activator 58.3 13 620 Montelukast Singulair Leukotriene antagonist 53.6 15 612 Zafirlukast Accolate Leukotriene antagonist 42.3 7 976 Fluvastatin Lescol Lipid-lowering 63.2 7 542 Rosuvastatin Crestor Lipid-lowering 40.2 11 680 Azithromycin Zithromax Macrolide 52.4 11 275 Moclobemide Manerix MAOI 58.8 10 835 Celecoxib Celebrex NSAID 44.1 17 458 Etodolac Lodine NSAID 49.9 9 091 Etoricoxib Arcoxia NSAID 42.7 12 665 Meloxicam Mobic NSAID 52.0 19 087 Nabumetone Relifex NSAID 54.9 10 444 Rofecoxib Vioxx NSAID 38.9 15 268 Tenoxicam Mobiflex NSAID 44.5 10 882 Raloxifene Evista Osteoporosis 57.2 13 987 Risedronate Actonel Osteoporosis 58.6 13 643 Misoprostol Cytotec Prostaglandin analog 67.3 13 775 Finasteride Proscar Prostate treatment 63.0 14 772 Lansoprazole Zoton Proton pump inhibitor 51.0 17 329 Omeprazole Losec Proton pump inhibitor 62.4 16 204 Pantoprazole Protium Proton pump inhibitor 44.5 11 541 Esomeprazole Nexium Proton pump inhibitor 41.8 11 595 Ciprofloxacin Ciproxin Quinolone 60.0 11 477 Enoxacin Comprecin Quinolone 44.5 2 790 Norfloxacin Utinor Quinolone 50.0 11 110 Ofloxacin Tarivid Quinolone 45.7 11 033 Aliskiren Rasilex Renin inhibitor 52.4 6 285 Duloxetine Cymbalta+Yentreve SNRI 49.4 19 485 Fluoxetine Prozac SSRI 58.4 12 692 Fluvoxamine Faverin SSRI 59.9 10 983 Paroxetine Seroxat SSRI 61.6 13 741 Sertraline Lustral SSRI 60.2 12 734 Tacrolimus Protopic Topical immunomodulator 52.8 12895 Pimecrolimus Elidel Topical immunomodulator 45.1 10 660

Mean 55.8 10 974

M-PEM studies Modafinil Provigil ADHD treatment 60.1 2 092 Atomoxetine Strattera ADHD treatment 60.3 5 079 Quetiapine XL Seroquel XL Antipsychotic 55.9 14 616 Fluticasone Flixotide Evohaler Corticosteroid 63.9 13 413 Fluticasone/ Seretide Evohaler Corticosteroid 62.0 13 464 Salmeterol Pulmicort Budesonide Corticosteroid 55.4 10 408 Varenicline Champix Nicotinic receptor partial agonist 54.5 12 135 Lumiracoxib Prexige NSAID 42.9 285 Fentanyl buccal Effentora Opioid analgesic 56.2 556 Fentanyl citrate PecFent Opioid analgesic 53.0 63 Travoprost Travatan Prostaglandin analog 82.7 3 528

Mean 58.8 6 876 364 MANN’S PHARMACOVIGILANCE

DSRU noties NHS Prescription Services (NHSBSA) of study drug under surveillance ↓ DSRU receives data from dispensed NHS prescriptions issued in England by GPs from the date of market launch, in strict condence from the NHSRxS ↓ PEM (standard/modied) questionnaires sent to GPs (e.g., ≥3/6/12 months after rst primary care prescription issued for patient) ↓ Information requested on questionnaire includes baseline demographic data, drug exposure details, events and other outcomes, important risk factors , and prescribing patterns. ↓ PEM (standard/modied) questionnaires returned, scanned, reviewed, and data entered onto DSRU database ↓ Selected events of medical interest (ADRs, rare and iatrogenic adverse reactions events, deaths (where cause not known), pregnancies, and other outcomes which require further evaluation may be followed up [Patient condentiality maintained throughout]

Figure 22.1 Common process of a standard PEM or M-PEM study.

prescribing GP until the target sample size (usually information bias through misclassification. GPs are many thousands of patients) is achieved. In order offered a modest reimbursement to cover adminis- to avoid placing an unreasonable demand on the trative costs in recognition of the time spent com- prescribers, no more than four questionnaires for pleting the more detailed M-PEM data collection each M-PEM study are sent to each doctor in any forms. one month for any one study. Data collected include Within the DSRU, each questionnaire is scanned patient demographics (age, sex), prescribing infor- into the system and the image is reviewed by a mation, and details of all significant events that scientific member of the DSRU staff. This initial have been recorded in the patient’s medical records review aims to identify possible serious ADRs or during a specific time period after starting the study events requiring action (e.g., external communica- drug. Early modifications to the standard method- tions or expedited follow-up). An aggregate as­­ ology involved adding a small number of additional sessment of drug-relatedness, clinical features/ questions (with yes, no, don’t know answers) on the manifestations, clinical course, and prognosis of questionnaire. These focused on issues specific to clinical conditions may be performed (see later). the drug under study; for example, the standard Supplemental information may be sought from questionnaire for the standard PEM study on the GPs using targeted questionnaires, where such NSAID meloxicam included questions about previ- information is not obtained in the initial survey. ous history of gastrointestinal conditions and intol- Such questionnaires are sent within weeks of the erance to NSAIDs to identify possible confounding initial review; but in some cases, where an objective by indication (Martin et al., 2000). The customized of a study might be to monitor events with longer M-PEM format questionnaires were developed to time to possible onset, a lag period may be intro- collect relevant supplementary information in order duced (e.g., 12 months from the date of first occur- to perform more detailed exploration of specific rence of the event of interest), such as for androgenic safety issues. An example is shown in Figure 22.2. manifestations with testosterone use in women The M-PEM questionnaire collects more detailed using a testosterone patch for reduced sexual drive. information on outcomes (including specific events A list of medically serious events (ICH, 2003) that that comply with prespecified case definitions), have been associated with the use of medicines (e.g. drug exposure, and other relevant disease risk aplastic anemia) has been compiled by the DSRU; factors at the start of treatment. This improves data such events routinely undergo further evaluation. accuracy and quality, reducing the possibility of All pregnancies reported during treatment or within MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 365

Figure 22.2 Example of M-PEM questionnaire. 366 MANN’S PHARMACOVIGILANCE

Figure 22.2 (Continued) MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 367

3 months of stopping the drug are followed up known and there is an a priori hypothesis of the using a supplementary questionnaire to determine effect size, then it is possible to analyze the statisti- the outcome of the pregnancy. All reported deaths cal power of a study given a fixed sample size. For for which no cause is specified are also followed up example, assuming 5% (two-sided test) significance, to try to establish the cause of death, provided the the power of a study based on 10 000 subjects to reporting GP has supplied a practice identification detect as statistically significant an increase in inci- code for the patient. dence from 0.1% to 0.2% would be 80% (Machin For each report, trained coding staff prepare a et al., 1997: Table 7.2). Because of the customized computerized, longitudinal, chronological record nature of M-PEM studies, a specific sample size is of demographic, exposure, and outcome data asso- calculated depending on the research question of ciated with starting the study drug. All events interest, for which the outcomes are chosen and reported on questionnaires are now coded onto a defined through internal DSRU scientific -discus DSRU database using Medical Dictionary for Reg- sion as those which best reflect the research ques- ulatory Activities (MedDRA) terminology (this tion. For the majority of M-PEM studies that have replaced the DSRU bespoke dictionary in 2012). been undertaken to date, the sample size has been Selected attributes are linked to selected data. For smaller than the 10 000 required for standard PEM example, an event is coded as an ADR if the GP studies. specified that the event is attributable to a drug Importantly, the final evaluable cohort sizes and (either the study drug, or another drug taken during the duration of a study are dependent on the level the study observation period), if the event had a of prescribing of the study drug in England by GPs. fatal outcome, or if the event was a reason for stop- However, cohort accrual is likely to be faster and ping. Data quality are assured through a number larger than in postmarketing clinical trials or exist- of methods based on error prevention, data moni- ing longitudinal medical records databases that toring, data cleaning, and documentation. sample from a subset of the population.

SAMPLE SIZE PRINCIPLES OF GOOD PHARMACOVIGILANCE PRACTICE In standard PEM, the sample size of 10 000 exposed patients has been driven by the methodology’s orig- M-PEM studies are conducted according to inal objective to bridge the gap between randomized national and international guidelines for ethical trials and spontaneous reporting regarding sensi- conduct of research involving human subjects tivity to rare and uncommon events that can be (RCP, 1996; DOH, 2000; CIOMS and WHO, 2002; achieved by including a larger sample size than GMC, 2013). Following the principles of good premarketing studies. Based on the general “rule of pharmacovigilance practice (EMA, n.d.), a full 3”,1 it follows that the larger the sample size, the protocol is written for each study to monitor and rarer the event that can be detected (Strom, 1994). research the safety of medicines. In addition, under Thus, a sample size of 10 000 within a standard Section 251 of the NHS Act 2006, the DSRU has PEM should allow for the detection of at least three received support from the Ethics and Confidential- cases of an adverse event, with 85% power, if the ity Committee of the National Information Gov- event occurs at a rate of at least 1 in 2000 patients ernance Board2 to gain access to and process patient (assuming the background rate is zero) (Machin iden­tifiable information without consent for the et al., 1997: Table 7.1). If the background rate is purposes of medical research (October 2009) (Health Research Authority, n.d.). Patient informa- tion security is assured through strict measures 1 The rule for safety data is commonly referred to as the “rule of 3.” In many situations involving rare reactions it is assumed guided by DSRU policies. Highly confidential that the frequency of the event is small, so that the occurrence of the event follows a Poisson distribution, and the 95% confi- dence interval is calculated based on the number of events. If 2 The responsibility for Section 251 was transferred to the Health no events are observed in a study of X individuals, then one can Service Authority and the Confidentiality Advisory Group be 95% certain that the event occurs no more often that 3/X. (CAG) in April 2013. 368 MANN’S PHARMACOVIGILANCE patient data (name and address) supplied by the specific events, including those considered to require NHSBSA used to identify the patient to the pre- special monitoring by regulatory authorities. scriber are then made anonymous through use of a Through M-PEM it is possible to evaluate the unique study identifier code assigned by the DSRU safety of a medicine in particular subpopulations and separately one supplied by the GP on the ques- defined by particular prognostic characteristics or tionnaire at the point of return. The practice code risk factors at various points in time (pre exposure or number is used for subsequent correspondence and or at treatment index date and/or concurrent) if additional information is sought from the doctor. during treatment which are considered important At least one interim report is written to summa- for the events of interest. In such studies, patients rize the data for each study based on per-protocol may be identified according to prespecified criteria predefined milestones (e.g., annually, or number of (age, sex, indication) through use of an eligibility evaluable patients). These reports may include a questionnaire. This may be necessary to define the listing, by month since the beginning of treatment, incept cohort to exclude, for example, those who of all events reported, and evaluation of factors received the product prior to the approval of an that may affect cohort accrual and impact on the extension to license subgroups (Davies et al., 2007; ability to meet study objectives. They are, if possi- Aurich-Barrera et al., 2009) or following important ble, discussed with the marketing authorization changes in the product’s lifecycle (e.g., a licensing holder (MAH) so that reporting obligations to the or formulation change) (ENCEPP n.d.e). regulatory bodies can be fulfilled and any remedial Examples of modifications to provide targeted action undertaken. Wherever possible each study is safety surveillance for a specific ADR (and sequalae) undertaken in a collaborative, but always independ- are, for example, idiopathic bronchospasm in new ent relationship with the MAH. users of chlorofluorocarbon (CFC)-free formula- tions of inhaled corticosteroids or anticholinergics (Perrio et al., 2007), misuse and diversion in new CONTRIBUTION OF STANDARD AND users of opioid products (Layton et al., 2011; MODIFIED PRESCRIPTION-EVENT Osborne et al., 2013), and psychiatric events in new MONITORING TO PHARMACOVIGILANCE users of rimonabant (Buggy et al., 2011). Examples of exploration of drug utilization and compliance The methodology is recognized as a tool for phar- with recommended prescribing regimens include macovigilance and risk management contributing the use of ivabradine in patients with conditions to the monitoring of overall safety of newly mar- that are contraindicated or for whom special warn- keted medicines as used in real-life clinical practice. ings apply (Doe et al., 2010), or use of selected It is included within EU regulatory guidelines as a medicines which, for example, affect cytochrome pharmacoepidemiological method that can be used P450 metabolism and hence drug response, such as in post-authorization safety studies (PASS) (EMA, CYP3A4 moderate and strong inhibitors, which are 2012). contraindicated for use within 14 days prior to As describe earlier, a number of M-PEM studies starting a novel formulation of fentanyl (Osborne have been completed and several are ongoing. The et al., 2013). results of these studies have been published sepa- rately elsewhere or the studies are in process; SIGNAL DETECTION AND however, Tables 22.2 and 22.3 provide an overview HYPOTHESIS TESTING of the methods used to illustrate the potential applications of M-PEM in the context of pharma- Signal detection and evaluation are the primary covigilance and risk management. These studies concerns of pharmacovigilance. Common to both were designed to address specific research ques- standard and M-PEM study designs, several tions, including characterization of real-life drug methods are applied for signal detection, both qual- use, adherence to prescribing recommendations or itative and quantitative, not only to look for new guidelines, and targeted surveillance or analysis of unexpected adverse reactions, but also for further MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 369

Table 22.2 Examples of applications of M-PEM methodology: completed studies.

Targeted population Drug (n) Background Data collection or event surveillance Applications

Carvedilol UK license extended Patient Heart failure Assessment of (Eucardix™) to treat mild to demographics, subgroup compliance with (Aurich-Barrera moderate chronic treatment identified by prescribing et al., 2009) heart failure initiation and initial eligibility recommendations [Roche subject to supervision, dose questionnaire and clinical Products Ltd] supervision of titration, severity management (n = 1666) hospital specialist of heart failure, guidelines pretreatment tests, post-license past medical extension history, concomitant medication FlixotideTM Regulatory Patient Event rates Active surveillance EvohalersTM requirement to demographics, compared for post-formulation (Perrio et al., monitor severity of specific change from 2007) [Allen & introduction of indication, use of respiratory event metered dose Hanburys Ltd, CFC-free inhalers oral rates (paradoxical inhaler to Uxbridge, in Europe corticosteroids, bronchospasm) CFC-free Middlesex, UK] spacer devices, before and after EvohalersTM (n = 13 413) and other starting CFC-free Identification of respiratory inhalers off-label use in treatments COPD Travoprost eye License extension to Patient Eligibility Active surveillance drops first-line use in demographics, questionnaire post-licence (Travatan™) the treatment of hospital initiation used to identify extension (Davies et al., ocular and specific population of Quantification and 2007) [Alcon hypertension in questions on the patients who better Labs. UK Ltd] open-angle occurrence of started treatment understanding of (n = 1441) glaucoma abnormal eyelash post-license specific events of granted in 2003 growth, abnormal extension interest eyelid hair growth, Incidence of specific and iris or ocular events periocular skin reported in discoloration premarketing trials assessed Modafinil License extended to Prescribing patterns, Subcohort of users Enhanced (Provigil™) include the plus selected identified post- characterization (Davies et al., treatment of aspects of patient license extension. of real-life drug 2013) “excessive management in Analyses further use [Cephalon sleepiness terms of stratified by Active surveillance (UK) Ltd.] associated with contraception. indication post-license (post-licence chronic Data also collected extension extension pathological on risk factors for cohort; conditions” in cardiovascular and n = 1096) 2004. Low psychiatric adverse projected use events and serious skin reactions (Continued) 370 MANN’S PHARMACOVIGILANCE

Table 22.2 (Continued)

Targeted population Drug (n) Background Data collection or event surveillance Applications Rimonabant Anti-obesity drug Patient demographic Comparison of Assessment of risk (Acomplia™) launched in the data, health status specific of specific (Buggy et al., UK in 2006 (body mass index, psychiatric event psychiatric/ 2011) [Sanofi- (product weight, smoking), rates occurring in nervous system Aventis] withdrawn from past medical and the 6 months events of (n = 10 011) market during psychiatric history prior to and after regulatory course of this and specific starting treatment concern study) questions on events of depression, anxiety, insomnia and seizures Varenicline Smoking cessation Demographic data, Focused time-to- Characterization of (Champix™) therapy. past and current event analysis on real-life drug use (Kasliwal et al., Regulatory smoking habit, prespecified Hypothesis testing 2009) [Pfizer concern over past medical events of interest: on pre-specified Ltd] psychiatric events history, current myocardial events of (n = 12 159) (suicidal ideation) morbidities and infarction, particular reason for depression, concern stopping (if anxiety, stopped) aggression, suicidal ideation, and nonfatal self-harm Atomoxetine Licensed for Demographic data, Matched cohort Hypothesis testing (Strattera™) treatment of prescribing analysis on events on prespecified (Davies et al., attention-deficit patterns, targeted of interest events of 2010) [Eli Lilly hyperactivity capture of data particular and Co. Ltd] disorder. (both prior to and concern (n = 5079) Regulatory during usage) on concern over an psychiatric events, increased risk of convulsions, suicidal thinking abnormal liver (Layton et al., function, and 2011) selected cardiovascular events Fentanyl buccal Launched in the UK Data collected on Targeted capture of Enhanced tablets in January 2009, demographics, data (both prior characterization (Effentora™) licensed for the initiation of to and during of drug use and (Layton et al., management of therapy (setting usage), including misuse 2011; Osborne breakthrough and titration) and respiratory, renal Specific evaluation et al., 2013) pain in patients past opioid use. and hepatic of use of [Cephalon (UK) with cancer Specific questions conditions, and medicine in Ltd] already receiving to identify concomitant relation to (n = 556) and tolerant to potential misuse medication concomitant opioid therapy or inappropriate/ medication or off-label use diseases that are contraindicated or where precautions are advised MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 371

Table 22.2 (Continued)

Targeted population Drug (n) Background Data collection or event surveillance Applications Fentanyl nasal Launched in the UK Data collected on Targeted capture of Enhanced spray tablets in 2010, licensed demographics, data (both prior characterization (PecFent™) for the initiation of to and during of drug use and (Osborne et al., management of therapy (setting usage), including misuse 2013) breakthrough and titration) and respiratory, renal Specific evaluation [Archimedes pain in patients past opioid use. and hepatic of use of Pharma] with cancer Specific questions conditions, and medicine in (n = 63) already receiving to identify concomitant relation to and tolerant to potential misuse medication concomitant opioid therapy or inappropriate/ medication or off-label use diseases that are contraindicated or where precautions are advised Ivabradine Licensed in the UK Demographic data, Targeted data Specific evaluation (Procorelan™) in 2006 for information on capture and of use of (Doe et al., treatment of treatment analysis for ivabradine in 2010) chronic stable initiation, past selected ocular relation to [Servier Lab. angina pectoris in medical history, and diseases/ Ltd] patients with current cardiovascular conditions that (n = 4624) normal sinus morbidities, events are rhythm, who contraindications contraindicated have a for use, baseline or where contraindication and ongoing precaution is or intolerance for results of tests of advised β-blockers heart rate and Quantification and concomitant characterization medications of specific ocular and cardiovascular events of interest observed in premarketing clinical trials Quetiapine Extended-release Data collected on Nested matched Hypothesis testing extended formulation demographics, use case–control on prespecified release licensed for the of medication that study to explore events of (Seroquel treatment of may cause relationship particular XL™) (ENCEPP schizophrenia, somnolence or EPS between dose concern in risk n.d.e) manic episodes and other risk and events of management [AstraZeneca associated with factors for these somnolence and plan UK Ltd] bipolar disorder, events EPS (n = 13 276) or as add-on Targeted data therapy for major capture and depressive analysis of pattern disorder of events related to diabetes mellitus/metabolic syndrome over time

EPS: extrapyramidal symptoms; XL: extended release. Table 22.3 Examples of applications of M-PEM methodology: ongoing studies.

Targeted Drug (target population/event number for cohort) Background Data collection surveillance Applications

Exenatide Once weekly Data collected on Targeted data Quantification and (Bydureon™) injection launched demographics, capture and characterization of (ENCePP, n.d.a) in UK in April 2011 initiation of therapy analysis of specific [Eli Lilly and Co. for treatment of (setting and pattern of gastrointestinal Ltd] (n = 5000) diabetes mellitus titration) and events related events of interest current/past to diabetes and observed in antidiabetic pancreatitis premarketing medication use and clinical trials adherence. Specific questions to identify risk factors for pancreatitis and gallstones Asenapine A novel atypical Data collected on Self-controlled Enhanced (Sycrest™) antipsychotic demographics, case series characterization of (ENCePP, n.d.b) developed for initiation of therapy study to drug use and [N.V. Organon] treatment of (setting and explore misuse (n = 5000) moderate to severe titration), and past temporal Specific evaluation manic episodes antipsychotic use. relationship of use of medicine associated with Primary focus on between in relation to bipolar disorder somnolence and starting concomitant and schizophrenia sedation, weight treatment and medication or gain, oral oral events diseases that are hypoesthesia, Targeted data contraindicated or swelling of the capture and where precautions tongue and throat, analysis time to are advised and allergic event Quantification and reactions characterization of specific oral events of interest observed in premarketing clinical trials Rivaroxaban A highly selective Data collected on Specific Quantification and (Xarelto™) direct factor Xa patient evaluation of characterization of (ENCePP, n.d.c) indicated for the demographics, use in special haemorrhagic [Bayer Pharma prevention of medical history/ populations events and VTE A.G.] venous medication use, Targeted data events indicating (n = 10000) thromboembolism adherence, capture and failure of (VTE) in patients prescribing analysis time to anticoagulation undergoing decisions. Primary event. elective hip or knee focus on replacements, haemorrhagic prevention of events. Secondary stroke and systemic foci on drug embolism in utilisation, off-label non-valvular atrial use and quantifying fibrillation (AF) and incidence of other the treatment and events. prevention of deep vein thrombosis (DVT) and pulmonary embolism (PE)*

*Also indicated for secondary prevention following an acute coronary syndrome in combination with aspirin alone or aspirin plus clopidogrel or ticlopidine. MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 373

information regarding expected drug–adverse However, the principle statistic of interest IDt is events associations of interest that might affect the the crude incidence density (ID) (or rate) that can benefit–risk balance of a drug. The M-PEM study be calculated for a given fixed time period t for all design offers greater opportunity to systematically events reported in patients for a given time period, collect supplementary information at the patient and expressed usually in units of first event reports level for the whole cohort, facilitating the investiga- per 1000 patient-months of treatment (or observa- tion and exploration of a range of additional tion). Since there are a large number of health out- research questions beyond that originally explored comes of interest and the censoring would be by the standard PEM approach. different for each outcome, the denominator for the crude ID does not initially include censoring. These rates can be used to give estimates of the “real- world” frequency of reported events by estimating INCIDENCE RISK AND INCIDENCE the cumulative incidence rate over fixed time RATES (DENSITIES) periods. In standard PEM studies, these estimates Both standard and M-PEM approaches enable are crude (unadjusted). For example, consider the analysis of longitudinal data and examination of standard PEM study of drospirinone/ethinyl estra- temporal relationships of outcomes to new expo- diol (Yasmin®), which identified 13 cases (five of sures. The methodological approach provides a deep vein thrombosis and eight of pulmonary numerator (the number of reports of an event), embolism) in 15 645 females, each with possible risk denominators not only in terms of the number of factor(s). Applying complete case analysis, the patients, but also number of patient-months of crude incidence rate was 13.7 cases per 10 000 exposure to the drug, and a known time frame. This woman-years (95% confidence interval (CI): 7.3, allows for event profiles over time to be examined 23.4) (Pearce et al., 2005). In the PEM study of through application of various statistical methods. strontium, which was an early example of modifica- In standard PEM, the scope for this analysis is tion collecting information on prior history of a limited to crude estimates, since information cannot targeted outcome (venous thromboembolism be collected on all important risk factors that may (VTE)), the crude annualized incidence of VTE be confounding factors for all outcomes because of was (95% Poisson exact CI) 6.24 (4.60–8.27) per the nature of the simple questionnaire design bal- 1000 patient-years strontium treatment (Osborne anced with no remuneration to respondents. In et al., 2010). M-PEM, additional information is collected for all Compared with the “classic” cohort design with patients within the cohort regarding relevant multiple exposure groups, the methodology is more co-morbidities and other potential confounding efficient in terms of resources. However, the absence factors, which can, through statistical modeling of data on a contextual comparator can, in some techniques, provide adjusted estimates for selected cases, be a limitation. To attempt to address this, it outcomes. is possible to undertake calculations of measures of For both approaches, initial simple crude analy- effect (risk or rate ratios) for internal comparisons sis of the incidence (risk) of events for an evaluable within cohort according to time periods and/or cohort by month by system organ class is an effec- between subgroups according to different defined tive descriptive method in which one may observe particular characteristics, or external comparisons disproportionally higher counts than expected from to carefully selected data sources. Again, common summary of product characteristics or other drugs to both standard and M-PEM approaches, within- within the database. Examples of signals as seen in cohort estimates of crude ID rate differences or a standard PEM are gynaecomastia with finas- ratios can identify events that occur significantly teride (Wilton et al., 1996) and hallucinations with more frequently soon after starting the study drug. tramadol (Dunn et al., 1997). An example seen in The null hypothesis is that the incidence rates are an M-PEM study is that of psychiatric events with constant between the two groups being compared; varenicline (Buggy et al., 2013). the alternative hypothesis is that the incidence rates 374 MANN’S PHARMACOVIGILANCE are different. In rejecting the null hypothesis where high degree of correlation between these two sets substantial differences are observed, this could be of values. The values are presented for the standard explained by a number of factors, including drug PEM study of desloratadine in Table 22.4. These treatment. values can be used to compare and contrast drugs Most frequently, for signal generation purposes within one therapeutic class; for example, with anti- for general surveillance, for each reported event, the histamine drugs it shows that drowsiness and seda- difference or ratio between time periods is calcu- tion are the most frequently reported events likely lated to allow the examination of the null hypoth- to be a drug side effect with levocetirizine, whereas esis; that is, the IDs in the first month after starting this is far less common with desloratadine; simi- treatment and the IDs for months 2 to 6 (ID1 − larly, lower respiratory tract infection (which occurs

ID2–6). A 95% CI is applied to the rate difference or month in and month out in all cohorts and which ratio (based on the normal approximation). Thus, is, with many drugs, unlikely to be related to either where the ID1 − ID2–6 value for an event is positive, the drug or disease being treated) is fairly common or ID1/ID2–6 is above one and the confidence limits among the ID values but virtually never appears around the point estimate exclude the null value among the common reasons for drug with­­ (zero or one respectively), the null hypothesis is drawal. A more detailed exploration of associations rejected. This result can be considered to be a signal between patient characteristics and reasons for for an event associated with starting treatment with stopping is possible within M-PEM. A recent the study drug. If the rate of events in months 2 to example is the exploration of psychiatric events as 6 combined is considered to be significantly greater reasons for treatment withdrawal for rimonabant than during month 1, this result is considered to be (Willemen et al., 2012). a signal for a delayed-onset event. In comparing these two time periods, the assumption is made FURTHER QUANTITATIVE ENHANCEMENTS that, given an event, its reporting is equivalent in WITHIN MODIFIED PRESCRIPTION-EVENT both periods in a fixed cohort. These signals then MONITORING require confirmation or refutation by further study. An example of a signal as seen in standard PEM For signal strengthening and exploration of specific is the association of skin reactions with lamot­­ safety issues, a key characteristic and advantage of rigine (Mackay et al., 1997). For drugs where M-PEM is the possibility to conduct comparisons, pattern of use is intermittent and/or short term, such as pre- and post-exposure periods, that help such summaries are also produced, but there are control for within-subject change in disease severity several differences. First, the numerator is based on as well as reducing between-group differences. The total incident counts irrespective of treatment bespoke M-PEM design offers greater scope for status (whether recorded during/post treatment or analysis within the cohort using self-controlled whether “unknown”) and the denominator takes methodology, because it allows lines of enquiry into account the observation period (between start about possible fixed and time-variant confounders. date and end of survey date). Second, the compara- Consider the ID rate or difference statistic. A tor (reference) period may be restricted. For significant result may be appear to be a safety signal example, in a standard PEM study of an antihista- arising for the product under study, but such events mine (desloratadine) (Layton et al., 2009) intended may be associated with the indication for treatment for short term (<30 days) intermittent use, the (confounding by indication), and/or or channeling second month was considered most appropriate as (preferential prescribing to subsets of patients the reference period. defined by specific characteristics, such as having a Other complementary quantitative analyses condition that is resistant to previous therapy) common to both standard and modified approaches (Petri and Urquhart, 1991) and/or switching (past include capturing information on and ranking by experience with an alternative drug may modify the frequency the reasons for stopping and comparing risk of adverse events associated with current use with ranked IDs. In general, there appears to be a of the study drug) (Ray, 2003). Examples include MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 375

Table 22.4 Most frequently reported events during first 2 months of observation with the two antihistamine drugs desloratadine and levocetirizine, ranked in order of counts in first month N1.

Desloratadine Levocetirizine

DSRU dictionary higher term N1 N2 RFS DSRU dictionary higher term N1 N2 RFS

Condition improved 1384 606 1984 Condition improved 1470 434 1896 No further request 658 77 733 No further request 640 59 699 Not effective 537 238 772 Not effective 460 133 588 Course completed 177 24 201 Course completed 160 29 189 Upper respiratory tract 53 47 4 Other drug substituted 62 26 88 infection Patient request 40 16 56 Upper respiratory tract 56 25 4 infection Hospital referrals no admission 30 13 14 Drowsiness, sedation 46 4 43 Headache, migraine 28 7 9 Headache, migraine 22 9 6 Lower respiratory tract 24 17 1 Hospital referrals no admission 22 11 10 infection Other drug substituted 23 13 36 Noncompliance 21 2 18 Urinary tract infection 19 6 0 Rash 20 9 8 Effective 17 2 19 Pregnancy 11 2 5 Infection skin, unspecifieda/ 17 9 0 Urinary tract infection 18 12 0 local bacterial Rhinitis allergic 17 18 5 Anxiety 17 4 0 Asthma worse 16 12 2 Lower respiratory tract 17 21 0 infection

N1: total number of first reports of each event during observation in month 1; N2: total number of first reports of each event during observation in month 2. RFS: reasons for stopping. Desloratadine: total no. reports 3969 during months 1 and 2 of observation out of 5559 reports for whole study period in 5502 patients (46.5% of cohort) for whole study period. Levocetirizine: total no. reports 3732 during months 1 and 2 of observation out of 5509 reports for whole study period in 5453 patients (44.1% of cohort). aUnspecified: no event term currently exists in DSRU dictionary.

paradoxical increase in rates of gastrointestinal health outcomes and the population to whom the adverse effects in users of COX-2-selective inhibi- results may be applicable. Examples of particular tors at high baseline risk of gastrointestinal adverse design and analytical applications nested within effects. Selection bias introduced by these phenom- M-PEM studies are provided below. ena may affect the generalizability of the study results since the evaluable cohort may not be fully Drug Utilization representative of the postmarketing users of the product. Other factors that may introduce selection Drug utilization research describes the extent, bias are external influences on prescribing (such as nature, and determinants of drug exposure at the expert committee guidelines and/or decisions for patient level. Data from M-PEM studies can inform reimbursement). In M-PEM, whilst prescribing about prescriber adoption of new drugs. The demo- patterns of a new drug cannot be predicted or con- graphic and clinical characteristics of new users can trolled for, the issues of prescribing governance, be described and examined in relation to signals of channeling, and influence of previous therapy can off-label use; for example, indications, dose, and be examined through careful data capture and a conditions or other factors that are contraindicated variety of analytical methods to provide a better or special warnings for use. An example is the understanding of the cohort characteristics and M-PEM study of ivabradine (which is licensed for 376 MANN’S PHARMACOVIGILANCE chronic stable angina) and its utilization in patients The method itself overcomes some of the disadvan- under 40 years of age, in which use for other indica- tages associated with non-nested case–control tions was observed since the prevalence of angina studies while incorporating some of the advantages (which is the indication for this product) is low in of a cohort study (Flanders and Louv, 1986). As a this age group (Doe et al., 2010). In addition, pharmacoepidemiologic tool for risk management M-PEM studies can examine aspects of adherence plans, the design potentially offers reductions in to prescribing guidelines. One M-PEM study is costs and efforts of data collection and analysis underway to explore the impact of expert guide- compared with the full cohort approach, with rela- lines on adoption within clinical practice of a novel tively minor loss in statistical efficiency. M-PEM anticoagulant (ENCePP, n.d.c; Layton et al., 2013). cohorts provide opportunities to conduct such nested case–control studies, for example, for patients who develop selected ADRs and matched Before and After Studies patients who receive the same drug without devel- “Before and after” studies compare the rate of par- oping ADRs. Two prospectively designed nested ticular outcomes during a defined period of expo- case–control studies are underway to investigate the sure (or observation) after starting the study drug association between dose and the occurrence of with those rates in the same individuals during a two outcomes (extrapyramidal symptoms; somno- defined period of observation before starting, using lence and sedation) in users of a new formulation a repeated-measures matched-pair analysis. The of an atypical antipsychotic (ENCEPP, n.d.e). null hypothesis is that event rates are the same prior and post starting treatment. One example within an Self-Controlled Case Series Analysis M-PEM study was the examination of rates of res- piratory events with the introduction a CFC-free Other methodologic developments that are being formulation of an anticholinergic (ipratropium) introduced to M-PEM studies to examine temporal metered dose inhaler (MDI) in populations who associations between specific events of interest and were “switchers” from the original MDI and those starting treatment with a new drug include the naive to ipratropium treatment (Osborne et al., application of the method of self-controlled case 2011). The analyses suggested that characteristics series studies proposed by Farrington et al. (1996). of these two subpopulations differed such that The method was originally developed to study naive patients were more likely to be children, have adverse reactions to vaccines. The method uses only an indication of asthma, and have milder disease cases; no separate controls are required as the cases severity, while switchers were more likely to be act as their own controls, thus minimizing the effect adults, have an indication of COPD, and have more of confounding by factors that do not vary with severe disease. Such differences have an important time, such as genetics and gender. Each case’s given impact on ongoing evaluation of risk–benefit observation time is divided into control and risk balance of the new formulation. The matched anal- periods. Time-varying confounding factors such as ysis in each subset revealed that, in naive patients, age can be allowed for by dividing up the observa- dyspnea was shown to be significantly lower in the tion period further into age categories. Because the “before” reference period (relative risk (RR) 0.6 method requires time-varying covariate data on (95% CI 0.40, 0.88) for post- versus pre-treatment), cases only and not for the whole cohort, it is effi- while dyspnea for switchers was shown to be signifi- cient in terms of sample size and resource. The cantly higher in the “after” high-risk period (RR method requires that specific criteria are met (for 1.46 (95% CI 1.02, 1.81)). example, occurrence of the event of interest should not affect subsequent exposure history or increase mortality) and thus is not applicable to all out- Predictors of Risk comes. Using this approach, measures of effect The nested case–control design is particularly (risk or rate ratio estimates) are automatically advantageous for examining predictors of disease. adjusted for all fixed confounders. Non-cases can MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 377 be ignored without bias, while cases are self- important antidiabetic drugs. The null hypothesis matched. Conditional regression modeling will was that the risk of these outcomes was the same provide the adjusted estimate of relative incidence regardless of treatment. Pioglitazone may be (with 95% CIs) of the outcome for the high-risk used alone or in combination with a sulfonylurea, observation period of interest relative to the remain- metformin, or insulin as an adjunct to diet and ing observation time. M-PEM studies provide an exercise for the management of type 2 (noninsulin- ideal platform to enable the relative incidence of dependent) diabetes mellitus. Though the combina- newly diagnosed outcomes of interest to be studied tion of pioglitazone and insulin is licensed and between predefined high- and low-risk periods in allows improvement of glycemic control, this com- new users, thus enabling time-to-occurrence of bination is associated with increased risk of edema selected events to be explored and reviewed for evi- and may cause weight gain. The adjusted hazard dence of temporal patterns (ENCePP, n.d.b). ratios for each of the separate models based on PEM study data for patients taking pioglitazone– insulin combination compared with those taking Time to Event Analysis pioglitazone monotherapy and/or pioglitazone It is acknowledged in signal detection that the gen- with another antidiabetic (sulfonylurea or met- eralized approach to segregation of time periods formin) were: edema 2.28 (95% CI: 1.37, 3.78); may not be appropriate for all events with respect weight gain 2.03 (95% CI: 1.15, 3.58), and cardiac to their most relevant time periods of excess. failure 1.73 (95% CI: 0.63, 4.74). This suggests that However, it is possible to explore the time of occur- patients taking the pioglitazone–insulin combina- rence of an event by using statistical methods, tion had higher risks than pioglitazone mono- termed “time to event” analysis, based on survival therapy or pioglitazone combined with another methodology. Using these methods, a hazard func- antidiabetic drug. tion can be estimated using an appropriate distribu- tion (e.g., Weibull) that shows the instantaneous ASSESSMENT OF THE EFFECTIVENESS OF RISK risk of an event over time. The use of this technique MANAGEMENT (RISK MINIMIZATION) is now incorporated within M-PEM studies to PROGRAMS explore temporal relationships for targeted events of interest as an additional tool for signal genera- Risk management is attracting immense interest in tion purposes. Examples include calculation of pharmacovigilance. M-PEM methodology contrib- smoothed hazard functions in examining rates of utes not only to the identification and measurement hypoglycemia in thiazolidinedione antidiabetic of risks of medicines, but, with some additions, drugs (Vlckova et al., 2009) and neuropsychiatric can also examine how the risks of medicines are outcomes associated with varenicline (Buggy et al., being managed in real-world clinical settings. An 2013). example is theM-PEM study that was conducted to monitor the introduction of carvedilol for the treat- ment of cardiac failure (Aurich-Barrera et al., Modeling 2009). The product (combined alpha- and beta- Multiple regression modeling allows the simultane- adrenergic blockers) has been used for the treat- ous testing and modeling of multiple independent ment of angina and hypertension for some time, variables on an outcome of interest. An example of but there was concern about its appropriate use for a conditional logistic regression modeling was a cardiac failure in the community. The aim of the within-PEM study comparison to examine the risk study was to monitor how the product is being of pioglitazone treatment combinations (with managed in the community; for example: What insulin or other antidiabetic agents) on risk of investigations were undertaken prior to starting the edema, weight gain, cardiac failure, and anemia drug? Who supervised the dose titration (GP or (Kasliwal et al., 2008). This was a standard PEM specialist)? Was the drug given to patients with the with modification to include history of use of appropriate severity of heart failure? The design 378 MANN’S PHARMACOVIGILANCE included sending an eligibility questionnaire fol- dence of risk of 7.00 per 1000 patients (Wilton lowed by up to three detailed questionnaires for a et al., 1999). Other events that routinely undergo period of up to 2 years. evaluation include pregnancies and deaths. Data collected from reported pregnancies include the proportion and nature of congenital anomalies in EXPLORATION OF SIGNALS AND babies born to women exposed to newly marketed FOLLOW-UP OF IMPORTANT EVENTS drugs during pregnancy, in particular in the first trimester (Wilton et al., 1997). All deaths are fol- Analysis and evaluation of pharmacoepidemiologi- lowed up to ascertain cause of death where cause cal data should include medical assessment, both has not been reported. to improve the understanding of signals raised by epidemiological techniques and to raise (and evalu- SIGNAL STRENGTHENING ate) new signals or hypotheses by using medical judgment with appropriate systems for causal infer- Signal strengthening can also be conducted through ence. Once a signal has been recognized through a variety of comparisons using selections of the either the standard or M-PEM approach, supple- PEM database (within therapeutic class, specific mentary analysis is required to further characterize patient groups) (Layton et al., 2001, 2004, 2006; important attributes. Medical evaluation of reports Acharya et al., 2005). Such comparisons are appro- is an important component. Further information priate because the database is comprised of new on events of interest and/or signals may be obtained drug-user populations assembled at the same stage from the prescriber and a case series constructed. in time in the immediate postmarketing period As highlighted previously, M-PEM design provides since introduction of each product. As described the opportunity for the collection of detailed infor- previously, it is also possible to conduct external mation on targeted events of interest from the comparisons using demographic data of the popu- initial survey for the evaluable cohort, as opposed lation as a whole (Boshier et al., 2004). to case-only information for standard PEM studies. The DSRU also receives requests from regula- Important safety signals have been generated tory authorities and manufacturers to investigate and events of interest explored in this way. In the drug safety signals in the PEM database. Whenever standard PEM study of the anti-epileptic drug possible the DSRU conducts retrospective analyses vigabatrin, following published case reports of (which usually include follow up of reports for the visual field defects associated with the use of the drug in question and comparator drugs). Such product, four cases of visual field defects were iden- analyses contribute to the debates on these signals tified initially in the PEM cohort. In view ofthe and to regulatory and public health decisions. One importance of the signal, 7228 patients who were example used data from a standard PEM study on reported to be taking the product by the end of the sertindole (Wilton et al., 2001). Sertindole is an study were followed up by sending a simple ques- atypical antipsychotic known to be associated tionnaire to the GP to ask whether any serious with prolongation of the QTc interval. The product adverse events or changes in vision had been was withdrawn from markets in the EU following reported since the initial study form had been reports of sudden death and serious cardiac returned. In addition, if the patient has been seen arrhythmias. The comparative analyses of the PEM by an ophthalmologist for visual problems, the studies of sertindole and two other atypical antip- ophthalmologist was asked to complete a question- sychotics, risperidone and olanzapine, studied car- naire giving details of visual field testing before and diovascular events, deaths from cardiovascular during treatment with vigabatrin. In addition to the events, and deaths from other causes (such as initial four reports, the follow-up information suicide) and were considered to be a very important revealed 29 cases of visual field defects that were source of information for the regulatory decision considered by the ophthalmologist to be probably on the matter. Other M-PEM studies where results or possibly related to vigabatrin, giving an inci- informed regulatory decisions regarding ongoing MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 379 benefit–risk evaluations included those examining contributing to the monitoring of overall safety of cardiovascular and gastrointestinal safety of newly marketed medicines as used in real-life clini- COX-2 selective inhibitors (Kasliwal et al., 2005). cal practice. M-PEM studies combine the advan- Where appropriate, comparisons are made tages of standard PEM studies (in monitoring between patients identified within an evaluable general safety and identification of unexpected cohort and an external reference group, if a suitable risks of a medicine) with that of a more targeted internal reference cohort cannot be found within safety study that addresses specific questions (to the DSRU database and the research question better understand known or partially known risks requires the result to be contextual. An example is with a medicine). the analysis of cardiovascular events of the PEM The disadvantages and limitations of the meth- study on sildenafil (a phosphodiesterase type 5 odology, like those of most of the available tech- inhibitor used for erectile dysfunction) (Boshier niques of pharmacovigilance, are however real. et al., 2004). Reported deaths from myocardial inf- They include the following: arction and ischemic heart disease in users of silde- nafil in the PEM study were found to be no higher 1 Selection bias is possible, in that there is the than expected according to national mortality sta- potential that the PEM cohort is not representa- tistics. Similarly, death from ischemic heart disease tive of the general population using NHS serv- in the bupropion PEM (when used for smoking ices. This cannot be assessed because PEM does cessation) was compared with external data and not monitor an unexposed cohort concurrently. showed no difference in the standardized mortality Non-response bias is another form of selection ratio (Boshier et al., 2003). The precautions with bias that is possible, since not all questionnaires regard to possible sources of bias and confounding sent are returned. The mean returns of the stand- also apply to external comparisons, principally due ard PEM and M-PEM questionnaires sent out to differences in study design and data collection are 56% and 60%, respectively. These are signifi- methods. Therefore, results of external compari- cantly higher than the reporting rate in the yellow sons must be interpreted very carefully. card and similar schemes (Heeley et al., 2001; While such comparisons produce valuable addi- Hazell and Shakir, 2006), but it cannot be estab- tions to the understanding of the safety of medi- lished in each PEM study whether the patients cines, it is important to emphasize that comparisons whose doctors return the questionnaires are in of independent cohorts are subject to bias and con- any way different from those whose doctors fail founding that must be taken into consideration in to complete and return the questionnaire. We the analysis and evaluation process. However, the already know that the responding and nonre- paucity of postmarketing safety studies in large sponding GPs differ very little in the distribution populations makes the information provided by of ages in which they became principals or in these comparative studies very useful. Real benefit their geographical distribution (Mackay, 1998; can only be achieved when not only the limitations Key et al., 2002). of any postmarketing safety study are taken into 2 Until recently, the methodology did not involve consideration, but also when its results are consid- monitoring within the secondary care setting. ered in relation to other studies that had been con- Thus, a “survivor bias” can operate whereby ducted on the same product. patients who both started and stopped a drug under hospital care may never receive a GP pre- scription and may, therefore, be undetected by DISCUSSION PEM. None of the current methods of pharma- covigilance is ideal in respect of this problem – PEM, and its modern replacement M-PEM, is a hence the importance of extending PEM well-established postmarketing surveillance tech- methodology into hospital practice. The DSRU nique in England, and internationally recognized as has adapted the principle method to examine a tool for pharmacovigilance and risk management drug initiations by specialists in the secondary 380 MANN’S PHARMACOVIGILANCE

care setting (DSRU, 2013). An example of the is between capturing the real-world and general- specialist cohort event monitoring (SCEM) izable data through the observational design and design is the Observational Assessment of Safety randomization in clinical trials that in postmar- in Seroquel (OASIS) study (ENCEPP n.d.f). This keting settings have many logistical and even was designed to examine the short-term (up to 12 ethical difficulties, as well as limited external weeks) safety and use of quetiapine fumarate in a validity caused by exclusion criteria and other prolonged-release formulation (Seroquel XL™), restrictions. along with a comparator group started on the 5 One of the strengths of the PEM/M-PEM tech- quetiapine immediate-release (IR) formulation. nique is that it collects dispensed rather than Any patient seen by a psychiatrist in England in prescribed data. This is in contrast to other data the mental health care setting was considered sources, such as data collected in the General eligible for inclusion where a clinical decision was Practice Research Database. However, while made to prescribe either the XL or IR prepara- indeed compliance is not examined routinely in tion of quetiapine as part of normal clinical prac- PEM/M-PEM, it is possible, if necessary, to tice for schizophrenia or mania associated with monitor repeated dispensing for the same patient bipolar disorder. Other SCEM studies are under- as an indicator of compliance and provide esti- way as PASS that support RMP (ENCePP, 2013; mates of compliance with treatment regimen by ENCEPP n.d.d.; Layton et al., 2013). The SCEM calculating medication possession ratios. methodology facilitates the systematic collection and reporting of safety and utilization data on The advantages of PEM and M-PEM are: patients newly initiated within the secondary care setting and, thus, is complementary to M-PEM 1 It is noninterventional, and thereby minimizes PASS, which are based in primary care. the selection biases that occur when the study 3 Like other observational studies, PEM has design interferes with the doctor’s choice of drug limited ability to collect information on con- for the individual patient. This means that in founding factors that might be important for all PEM/M-PEM, data are collected on patients possible outcomes. The adoption of M-PEM who have received the study drug because the methods has provided considerable opportuni- doctor considered it the most appropriate treat- ties to enhance collection of supplementary ment for that patient, as in everyday “real-world” data on important risk factors. Furthermore, as clinical practice. described earlier, new techniques are being 2 It is national in scale, and the cohort comprises introduced to examine temporal associations, all patients given the drug usually immediately such as the application of the methods of self- after its launch into general practice. In Europe controlled case series studies. However, even it is the only database that can regularly identify without analysis, lists of reported events are cohorts of thousands of patients for newly useful to prescribing doctors for they show introduced medicines soon after launch. This is which events are reported in everyday clinical in contrast to some data sources that have practice and the relative frequency with which limited data on newly marketed products in this these events will be seen. They are perhaps more immediate postmarketing period because of useful than the unquantified long lists of - pos small size of because of population exposed in sible side effects given in the standard prescrib- the subset of the population monitored within ing information. these systems. The methodology prompts all 4 Whilst efficient in terms of resources, the single- prescribers to provide information on safety and group cohort design where evaluable patients are use because they automatically receive a ques- included on the basis of a single common expo- tionnaire for each patient prescribed the drug sure has the limitation that there is absence of an being monitored. It is probably this prompting unexposed comparator. Thus, comparisons need function that is responsible for the success of to be undertaken with great care. The “trade-off” the methodology; it does not rely on the doctor MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 381

taking the initiative to report happenings. These hazards and to further evaluate safety concerns features ensure that the studies are population identified by other pharmacovigilance methods based and that they disclose the real-life clinical or arising from regulatory concerns. Their cus- experience with the drug; there are no exclu- tomized sample size is advantageous in terms of sions, and all patients prescribed the drug are study conduct, limiting costs, and providing recruited even if they are very old, very young, timely information to the dynamic risk manage- or receiving several drugs concurrently for mul- ment process. Thus, they should be considered a tiple illnesses. valuable tool when developing a risk manage- 3 Exposure data are derived from dispensed pre- ment plan for the evaluation of the safety a new scriptions, with validation from prescribers medicine. through confirmation of such data on the ques- 7 M-PEM contributes to risk management plans tionnaires. Considering the large proportion of not only by the analysis and understanding of patients who do not get a prescription dispensed possible adverse events, including those consid- (Beardon et al., 1993), this is an advantage in that ered to be potential and identified risks in risk exposure data are more accurate than that management plans, but also by providing oppor- derived from records of physician-issued pre- tunities for studying drug utilization to answer scriptions (which are not always dispensed), as questions regarding missing information and the held in some databases. characteristics of postmarketing users of medi- 4 Because the data are concerned with events, the cines. Furthermore, M-PEM is being used to system could detect side effects that none of the study the effectiveness of risk minimization doctors have suspected to be due to the drug. The methods. information provided by event reporting does not require the doctor to decide whether or not an individual event in a single patient is drug FUTURE PLANS related. It thereby avoids a very difficult clinical decision, since as most reactions resemble fairly The mission of the DSRU is to continue to improve common clinical events, avoiding the doctor its research methods and foster the conduct of having to decide on causation may well encour- innovative national and international collaborative age reporting. The system allows direct contact research. between the doctors working in the DSRU and the GPs, so that follow-up surveillance of individual cases (including long-latency events) ACKNOWLEDGMENTS or deaths and all pregnancies is facilitated. 5 Additional advantages accrue from the increas- Special gratitude goes to Professor Ron Mann for ing size of the PEM database, which has been allowing the use material from the previous built up since 1984. The database now contains editions. information on 120 completed PEM/M-PEM studies and over 1 million patients. This has started to provide opportunities for comparing REFERENCES products and patient groups in the database. As time passes and more studies are completed, the Acharya, N.V., Pickering, R.M., Wilton, L.W., & Shakir, value of the database as a research tool increases S.A. (2005) The safety and effectiveness of newer progressively. antiepileptics: a comparative postmarketing cohort 6 A number of M-PEM studies have been under- study. J Clin Pharmacol, 45 (4), 385–393. taken to support the construction of risk man- Anon. (1986) Medicines (Adverse Reactions). Hansard 90(c620W). agement plans by providing opportunities for a Aurich-Barrera, B., Wilton, L.V., & Shakir, S.A. (2009) number of additional research applications that Use and risk management of carvedilol for the treat- can be used to generate signals of potential safety ment of heart failure in the community in England: 382 MANN’S PHARMACOVIGILANCE

results from a modified prescription-event monitoring Multi-centre research in the NHS – the process of study. Drug Saf, 32 (1), 43–54. ethical review when there is no local researcher. Beardon, P.H., McGilchrist, M.M., McKendrick, A.D. Department of Health. URL http://webarchive et al. (1993) Primary non-compliance with pre­­ .nationalarchives.gov.uk/20130107105354/ [accessed scribed medication in primary care. BMJ, 307 (6908), 10 December 2013]. 846–848. DSRU (2013) Studies in Secondary Care. URL: http:// BMA (2006) Reporting Adverse Drug Reactions. A Guide www.dsru.org/studies-for-risk-management/studies for Healthcare Professionals. British Medical Associa- -in-specialist-hospital-care [accessed 23 January tion Board of Science, British Medical Associa­­ 2014]. tion Science & Education, London. URL http:// Dunn, N., Wilton, L.V., Pearce, G., & Mann, R.D. (1997) bmaopac.hosted.exlibrisgroup.com/exlibris/aleph/ The incidence of hallucinations associated with trama- a21_1/apache_media/GYVFNJ1RT2PFNDUJ8IUK dol. Pharmacoepidemiol Drug Saf, 6 (Suppl. 2), abstract PUNTFVRF8G.pdf [accessed on 10 December 2013]. 042. Boshier, A., Wilton, L.V., & Shakir, S.A. (2003) Evalua- EMA (2012) Guideline on good pharmacovigilance prac- tion of the safety of bupropion (Zyban) for smoking tices. Module VIII –Post-authorisation studies, EMA/ cessation from experience gained in general practice 330405/2012, European Medicines Agency. URL use in England in 2000. Eur J Clin Pharmacol, 59 (10), http://www.ema.europa.eu/docs/en_GB/document_ 767–773. library/Scientific_guideline/2012/06/WC500129137 Boshier, A., Wilton, L.V., & Shakir, S.A. (2004) Evalua- .pdf Date [accessed on 20 July 2012]. tion of the safety of sildenafil for male erectile dysfunc- EMA (n.d.) Good pharmacovigilance practices. URL tion: experience gained in general practice use in http://www.ema.europa.eu/ema/index.jsp?curl=pages/ England in 1999. BJU Int, 93 (6), 796–801. regulation/document_listing/document_listing Buggy, Y., Cornelius, V., Wilton, L., & Shakir, S.A. _000345.jsp [accessed on 23 December 2013]. (2011) Risk of depressive episodes with rimonabant: a ENCePP (2013) ENCePP E-Register of Studies. DSRU. before and after modified prescription event monitor- An observational post-authorisation safety specialist ing study conducted in England. Drug Saf, 34 (6), cohort monitoring study (SCEM) to monitor the 501–509. safety and utilisation of asenapine (Sycrest™) in Buggy, Y., Cornelius, V., Fogg, C. et al. (2013) Neuropsy- the mental health trust setting in England. URL: chiatric events with varenicline: a modified prescription- http://www.encepp.eu/encepp/studiesDatabase.jsp. event monitoring study in general practice. Drug Saf, ENCePP (n.d.a) ENCePP E-Register of Studies. 36 (7), 521–531. DSRU. An observational post-authorization modified CIOMS & WHO (2002) International Ethical Guidelines prescription–event monitoring safety study to monitor for Biomedical Research Involving Human Subjects. the safety and utilization of exenatide once weekly Council for International Organizations of Medical (Bydureon™) in the primary care setting in England. Sciences(CIOMS) and World Health Organiza­­ URL http://www.encepp.eu/encepp./studiesDatabase tion (WHO), Geneva. URL http://www.cioms.ch/ .jsp. publications/layout_guide2002.pdf [accessed on 10 ENCePP (n.d.b) ENCePP E-Register of Studies. DSRU. December 2013]. An observational post-authorization modified Davies, M.N., Paiba, N., Wilton, L.W., & Shakir, S.A.W. prescription–event monitoring safety study to monitor (2007) A 12 month modified prescription-event moni- the safety and utilization of asenapine (Sycrest™) in toring report for travoprost. Drug Saf, 30 (10), 919– the primary care setting in England. URL http://www 990 (abstract P.020). .encepp.eu/encepp./studiesDatabase.jsp. Davies, M., Tong, E., Fogg, C. et al. (2010) Atomoxetine ENCePP (n.d.c) ENCePP E-Register of Studies. DSRU. end of modified PEM study report. Drug Safety An observational post-authorization modified Research Unit, Southampton. prescription–event monitoring safety study to monitor Davies, M., Wilton, L., & Shakir, S. (2013) Safety profile the safety and utilization of rivaroxaban (Xarelto™) in of modafinil across a range of prescribing indications, the primary care setting in England. URL http:// including off-label use, in a primary care setting in www.encepp.eu/encepp./studiesDatabase.jsp. England: results of a modified prescription-event mon- ENCEPP (n.d.d). ENCePP E-register of Studies. DSRU. itoring study. Drug Saf, 36 (4), 237–246. An Observational Post-Authorization Safety Specialist Doe, C., Fogg, C., Layton, D., & Shakir, S.A.W. (2010) Cohort Event Monitoring Study (SCEM) to Monitor Interim results of a modified prescription event moni- the Safety and Utilization of Rivaroxaban (XareltoTM) toring study on ivabradine: case series of utilisation in for the Prevention of Stroke in Patients with AF, Treat- patients under the age of 40 years. Pharmacoepidemiol ment of DVT and PE, and the Prevention of Recurrent Drug Saf, 19 (Suppl. 1), abstract 330. DVT and PE in the Secondary Care Hospital Setting in DOH (2000) Supplementary operational guidelines for England and Wales. URL http://www.encepp.eu/ NHS research ethics committees – November 2000. encepp/studiesDatabase.jsp MODIFIED PEM AND SUPPORT OF RISK MANAGEMENT 383

ENCEPP (n.d.e). ENCePP E-register of Studies. DSRU. doctors in a prescription event monitoring study of Modified Prescription-Event-Monitoring Study on drug safety. Pharmacoepidemiol Drug Saf, 11 (2), Extended-Release Quetiapine (Seroquel XL™) . URL 143–148. http://www.encepp.eu/encepp/studiesDatabase.jsp Layton, D. & Shakir, S.A.W. (2011) Prescription-event ENCEPP (n.d.f). ENCePP E-register of Studies. DSRU. monitoring. In: B.L. Strom, S.E. Kimmel, & S. Hen- Observational Assessment of Safety in Seroquel nessy (eds), Pharmacoepidemiology, 5 edn, pp. 301– (OASIS) Study. URL http://www.encepp.eu/encepp/ 330. John Wiley & Sons, Ltd, Chichester. studiesDatabase.jsp Layton, D., Clark, D.W., Pearce, G.L., & Shakir, S.A. Farrington, C.P., Nash, J., & Miller, E. (1996) Case series (2001) Is there an association between selective serot- analysis of adverse reactions to vaccines: a comparative onin reuptake inhibitors and risk of abnormal bleed- evaluation. Am J Epidemiol, 143 (11), 1165–1173. ing? Results from a cohort study based on prescription Flanders, W.D. & Louv, W.C. (1986) The exposure odds event monitoring in England. Eur J Clin Pharmacol, ratio in nested case–control studies with competing 57 (2), 167–176. risks. Am J Epidemiol, 124 (4), 684–692. Layton, D., Heeley, E., & Shakir, S.A.W. (2004) Identifi- Freemantle, S.N., Pearce, G.L., Wilton, L.W. et al. (1997) cation and evaluation of a possible signal of exacerba- The incidence of the most commonly reported events tion of colitis during rofecoxib treatment, using with 40 newly marketed drugs – study by prescription- prescription-event monitoring (PEM) data. J Clin event monitoring. Pharmacoepidemiol Drug Saf, 6 Pharm Ther, 29 (2) 171–181. (Suppl. 1), 1–8. Layton, D., Wilton, L., Boshier, A. et al. (2006) Com- GMC (2013) Good Medical Practice. General Medical parison of the risk of drowsiness and sedation between Council. URL http://www.gmc-uk.org/Prescribing_ levocetirizine and desloratadine: a prescription-event Guidance__2013__50955425.pdf [accessed on 23 monitoring study in England. Drug Saf, 29 (10), December 2013]. 897–909. Hazell, L. & Shakir, S.A. (2006) Under-reporting of Layton, D., Wilton, L., & Shakir, S.A. (2009) Examining adverse drug reactions: a systematic review. Drug Saf, the tolerability of the non-sedating antihistamine 29 (5), 385–396. desloratadine: a prescription-event monitoring study Health Research Authority (n.d.) Confidentiality Advi- in England. Drug Saf, 32 (2), 169–179. sory Group advice and approval decisions. Prescription– Layton, D., Osborne, V., Al-Shukri, M., & Shakir, S.A.W. event monitoring, CAG ref: ECC 5-07(b)/2009. URL: (2011) Indicators of dependence and aberrant behav- http://www.hra.nhs.uk/about-the-hra/our-committees/ iours during post-marketing use of fentanyl buccal section-251/cag-advice-and-approval-decisions/ tablets: risk scores in support of risk management [accessed on 23 December 2013]. [abstract OP22]. In: 11th Annual Conference of the Heeley, E., Riley, J., Layton, D. et al. (2001) Prescription- International Society of Pharmacovigilance (ISOP), event monitoring and reporting of adverse drug reac- 26–28 October, Istanbul, Turkey. tions. Lancet, 358 (9296), 1872–1873. Layton, D., Davies, M., Evans, A., & Shakir, S.A.W. ICH (2003) ICH Harmonised Tripartite Guideline. Post- (2013) Methodological considerations in evaluating Approval Safety Data Management: Definitions and the safety of novel anticoagulants in secondary care Standards for Expedited Reporting E2D. The Interna- setting in the UK: defining the contextual comparator tional Conference on Harmonisation of Technical cohort. Pharmacoepidemiol Drug Saf, 22 (Suppl. 1), Requirements for Registration of Pharmaceuticals for abstract 216. Human Use, Geneva. Machin, D., Campbell, M., Fayers, P. & Pinol, A. (1997) Kasliwal, R., Layton, D., Harris, S. et al. (2005) A com- Sample Size Tables for Clinical Studies. Blackwell parison of reported gastrointestinal and thromboem- Science, Ltd, Oxford. bolic events between rofecoxib and celecoxib using Mackay, F.J. (1998) Post-marketing studies: the work of observational data. Drug Saf, 28 (9), 803–816. the Drug Safety Research Unit. Drug Saf, 19 (5), Kasliwal, R., Cornelius, V., Wilton, L.W., & Shakir, 343–353. S.A.W. (2008) Combined pioglitazone-insulin use: Mackay, F.J., Wilton, L.V., Pearce, G.L. et al. (1997) analysis from the pioglitazone PEM study. Phar­ Safety of long-term lamotrigine in epilepsy. Epilepsia, macoepidemiol Drug Saf, 17 (Suppl. 1), abstract 38 (8), 881–886. 255. Martin, R.M., Biswas, P., & Mann, R.D. (2000) The Kasliwal, R., Wilton, L.V., & Shakir, S.A. (2009) Safety incidence of adverse events and risk factors for upper and drug utilization profile of varenicline as used in gastrointestinal disorders associated with meloxicam general practice in England: interim results from a use amongst 19,087 patients in general practice in prescription-event monitoring study. Drug Saf, 32 (6), England: cohort study. Br J Clin Pharmacol, 50 (1), 499–507. 35–42. Key, C., Layton, D., & Shakir, S.A. (2002) Results of a Osborne, V., Layton, D., Perrio, M. et al. (2010) postal survey of the reasons for non-response by Incidence of venous thromboembolism in users of 384 MANN’S PHARMACOVIGILANCE

strontium ranelate: an analysis of data from a Shakir, S.A., Wilton, L.V., Boshier, A. et al. (2001) Car- prescription-event monitoring study in England. Drug diovascular events in users of sildenafil: results from Saf, 33 (7), 579–591. first phase of prescription event monitoring in England. Osborne, V., Layton, D., Tong, E., & Shakir, S.A.W. BMJ, 322 (7287), 651–652. (2011) Safety of Atrovent® CFC-free MDI (metered Strom, B. (1994) Sample size considerations for pharma- dose inhaler): an example of protopathic bias in obser- coepidemiology studies. In: B. Strom (ed.), Pharma- vational cohort studies. Pharmacoepidemiol Drug Saf, coepidemiology, 2nd edn. John Wiley & Sons, Ltd, 20 (Suppl. 1), abstract 305. Chichester. Osborne, V., Al-Shukri, M., Gilchrist, A. et al. (2013) A Vlckova, V., Cornelius, V., Kasliwal, R. et al. (2009) modified prescription event monitoring study on fen- Hypoglycaemia with oral antidiabetic drugs: results tanyl pectin nasal spray (PecFent®) for the manage- from prescription-event monitoring cohorts of rosigli- ment of breakthrough cancer pain. Final Study tazone, pioglitazone, nateglinide and repaglinide. Drug Report. Drug Safety Research Unit, Southampton. Saf, 32 (5), 409–418. Osborne, V., Layton, D., Fogg, C. et al. (2013) Utilisation Willemen, M., Mantel-Teeuwisse, A.K., Buggy, Y. et al. of fentanyl buccal tablets in England: exploring off- (2012) Reasons for and time to discontinuation of label use reported in a cohort study. Eur J Pain, in rimonabant therapy. Drug Saf, 35 (12), 1147–1158. press. doi: 10.1002/j.1532-2149.2013.00389.x. Wilton, L., Pearce, G., Edet, E. et al. (1996) The safety Pearce, H.M., Layton, D., Wilton, L.V., & Shakir, S.A. of finasteride used in benign prostatic hypertrophy: a (2005) Deep vein thrombosis and pulmonary embo- non-interventional observational cohort study in lism reported in the prescription event monitoring 14,772 patients. Br J Urol, 78 (3), 379–384. study of Yasmin. Br J Clin Pharmacol, 60 (1), Wilton, L.V., Mackay, F., Pearce, G., & Mann, R.D. 98–102. (1997) The outcomes of pregnancy in women pre- Perrio, M.J., Wilton, L.V., & Shakir, S.A. (2007) A modi- scribed newly marketed drugs in 24 studies con­­ fied prescription-event monitoring study to assess the ducted by prescription–event monitoring (PEM). introduction of Flixotide Evohaler into general prac- Pharmacoepidemiol Drug Saf, 6 (Suppl. 2), abstract tice in England: an example of pharmacovigilance 072. planning and risk monitoring. Pharmacoepidemiol Wilton, L.V., Stephens, M.D., & Mann, R.D. (1999) Drug Saf, 16 (9), 969–978. Visual field defect associated with vigabatrin: ­obser Petri, H. & Urquhart, J. (1991) Channeling bias in the vational cohort study. BMJ, 319 (7218), 1165– interpretation of drug effects. Stat Med, 10 (4), 1166. 577–581. Wilton, L.V., Heeley, E.L., Pickering, R.M., & Shakir, Ray, W.A. (2003) Evaluating medication effects outside S.A. (2001) Comparative study of mortality rates and of clinical trials: new-user designs. Am J Epidemiol, cardiac dysrhythmias in post-marketing surveillance 158 (9), 915–920. studies of sertindole and two other atypical antipsy- RCP (1996) Guidelines on the Practice of Ethical Com- chotic drugs, risperidone and olanzapine. J Psychop- mittees in Medical Research involving Human Subjects. harmacol, 15 (2), 120–126. Royal College of Physicians of London.