Review For reprint orders, please contact [email protected]
Clinical perspectives in drug safety and adverse drug reactions Expert Rev. Clin. Pharmacol. 1(5), 695–705 (2008)
Peter Ian Pillans Princess Alexandra Hospital, Woolloongabba, Brisbane, 4102, Queensland, Australia Tel.: +61 073 240 2693 Fax: +61 073 240 7131 [email protected]
Adverse drug reactions (ADRs) remain a common clinical problem since they can mimic many diseases and cause significant morbidity and mortality. Judicious prescribing is important to minimize their occurrence. Apart from the recent identification of a few pharmacogenomic biomarkers for serious reactions, many remain unpredictable. Spontaneous reporting continues to play an important role in pharmacovigilance and the value of astute clinical observation and well-documented reports of suspicions of a causal link cannot be underestimated. Many national reporting schemes have developed considerable experience and expertise over many years and have large ADR databases, which are national assets. Despite advances in pharmacovigilance, numerous deficiencies have been identified; postmarketing surveillance remains the weakest link in the regulatory process. Regulatory authorities have tended to act later rather than sooner in response to safety signals, and this, when combined with under-reporting, may have led to exposure of a large number of patients to drug-related harm before restriction or withdrawal. In an attempt to improve vigilance, international surveillance may benefit by moving from its current passive/reactive mode toward active surveillance systems with a prospective, comprehensive and systematic approach to monitoring, collecting, analyzing and reporting data on ADRs. This will include increased pressure on pharmaceutical companies to conduct postmarketing studies. Such an active/proactive approach, while maintaining focus on ADR detection, could also aim to extend knowledge of safety, such that emerging changes in risk–benefit during a drug’s marketed life are effectively communicated to clinicians and patients. Drug safety monitoring and its regulation are now undergoing an overhaul and it is hoped that vigilance, public safety and trust will improve as a result. Keywords : adverse drug reaction • drug safety • pharmacovigilance • postmarketing surveillance
Over the past 5 years, a series of recalls of highprofile prescription medicines has aroused serious concerns regarding the safety of medicines [1–6] . Clinical trials and routine regulatory oversight, as practiced currently, often fail to uncover important adverse effects for widely marketed products [5] . Medical journals have published numerous articles and editorials relating to drug safety and regulation, with recommendations to overhaul drug safety monitoring, improve vigilance, ensure greater protection to the health of the public and restore trust [6–13] . The first part of this review will deal with adverse drug reactions (ADRs), with a brief overview of the epidemiology, detection of ADRs and pharmacovigilance, reporting of ADRs, the role of national monitoring schemes, handling of signals and the communication of www.expert-reviews.com
10.1586/17512433.1.5.695
safety information. The second part addresses drug-regulatory issues pertinent to drug safety in the current climate. Adverse drug reactions
An ADR is an unintended response, occuring at doses used in humans [14] . It may predict hazards for future administration or warrant prevention or specific treatment, alteration of the dosage regimen or withdrawal. ADRs are typically divided into those that are dose dependent and predictable, based on the pharmacology of the drug (Type A or augmented) and those that are unpredictable, dose independent, unrelated to the drug’s pharmacology and often immuno logically mediated (Type B or bizarre). Since, it is sometimes difficult to assign a reaction to one type, more-complex subdivisions have been proposed, based on dose relatedness, timing and
© 2008 Expert Reviews Ltd
ISSN 1751-2433
695
Review
Pillans
patient susceptibility [15] . In this comprehensive classification, other criteria are taken into account, including properties of the reaction (the time course of its appearance and its severity) and properties of the individual (the genetic, pathological and other biological differences that confer susceptibility). The definition of an ADR is distinct from an adverse drug event, which is an ‘umbrella term’ to describe any adverse experience that occurs in a person on a medication that may or may not be causally related. Epidemiology of ADRs
Adverse drug reactions are an important cause of iatrogenic disease [16] . They can involve any organ system, can present clinically in many different ways and have taken over from syphilis and tuberculosis as the disease mimic [17] . Although estimates of the incidence of ADRs vary, they are an important cause of morbidity, mortality and wasted expenditure [18] . In a large meta-analysis of 39 prospective studies from US hospitals from 1966 to 1996, the annual incidence of severe ADRs (defined as those that required hospitalization, were permanently disabling or resulted in death, occurring while in hospital or as the cause of admission to hospital) was 6.7% [19] . In a recent study involving a large population of hospitalized Medicare patients, 1.7% experienced an ADR [20] . A similar ADR prevalence of 1.5% was found in another recent study involving 4.3 million patients with adverse drug events who visited doctor’s offices, hospital outpatient clinics and emergency facilities [21] . A review of 14 Australian studies found that ADRs occurred in 2.4–3.6% of hospital admissions [22] . Despite programs to promote rational and safer use of medicines in West Australia, a study that was limited to ADRs of sufficient severity to warrant or extend hospitalization found the rate of ADRrelated hospital stays increased from 2.5 per 1000 person years in 1981 to 12.9 per 1000 patient years in 2002 [23] . The largest increases occurred in those aged over 80 years. In another study in Australian general practice, 10% of patients had experienced an adverse drug event in the previous 6 months [24] . Patients aged over 45 years, children aged 1–4 years and female patients were significantly more likely to have experienced an adverse drug event in this study. In the UK, 6.5% of hospital admissions were found to be medication related [25] . The median age of patients admitted with ADRs was 76 years, which was significantly older than patients without ADRs (median age: 66 years). In some studies, the public-health impact is likely to be underestimated, because many adverse events do not result in hospital admission [26] . Many ADRs are potentially preventable, particularly those that are predictable (Type A) or where the patient has a known drug hypersensitivity [19,27,28] . Clearly, judicious prescribing with a sound knowledge of the pharmacology of the drugs in question, including relevant pharmacokinetics, contraindications and interactions, together with consideration of the comordities in the patient at hand, is important in the prevention of ADRs. For a very small number of drugs, where in the past the ADR was considered unpredictable, there are now pharmacogenomic biomarkers for prediction of serious reactions [29] . Examples include the association 696
between the HLA‑B* 5701 gene variant and hypersensitivity reactions to abacavir [30] , HLA-B*1502 and carbamazepineinduced Stevens–Johnson syndrome (SJS) in Han Chinese [31] , HLA-B*5801 and allopurinol-induced SJS [32] , mutant alleles of cytochrome P450 (CYP)2C9 and bleeding with warfarin [33] and a thiopurine methyltransferase genotype associated with deficiency of the enzyme and severe myelotoxicity from azathioprine [34] . With some exceptions, routine pharmacogenetic testing for only two of these biomarkers, CYP2C9 and thiopurine methyltransferase, is conducted in some centers at the present time. Detection of ADRs & pharmacovigilance
Discovery in an observational science such as pharmacovigilance depends on the capacity to recognize and investigate unexpected clinical events that are manifest once a new drug is in use [35] . Pharmacovigilance is the science related to the detection, assessment, understanding and prevention of ADRs [11] . Its primary function is to improve the safety of marketed medicines. Data are derived from many sources, including published case reports, voluntary ADR reporting to national pharmacovigilance centers, postmarketing clinical and epidemiological studies, prescription event monitoring schemes morbidity and mortality databases [36] . The key stakeholders are patients, health professionals, governments and pharmaceutical companies. The current system of voluntary reporting plays an important role in identifying ADRs [37,38] . This voluntary scheme of ADR reporting has reached a high level of sophistication internationally, although the limitations are well known: poor quality of some submitted reports, significant under-reporting [39–41] , difficulty in calculating rates because of incomplete numerator data along with unreliable denominators and limited ability to establish cause and effect [42] . While this methodology has been responsible for the identification of numerous safety signals, there are more fundamental reasons to continue with a system of voluntary reporting. Promoting the concept that serious or unexpected adverse reactions should be reported to relevant national pharmacovigilance centers emphasizes the importance of drug safety issues to those involved. A unique feature is that it may detect reactions that might otherwise go unrecorded because it entails suspicions of cause and effect that health professionals consider potentially important. Furthermore, the reports allow for a qualitative description of how the reaction affects patient’s lives. Pharmaceutical companies play an important role in gathering ADR reports on their own products. Periodic Safety Update Reports (PSURs) are generated, which provide global safety updates. While detection of a new ADR is likely to lead to an amendment to the product information, the emphasis has been on the fulfilment of regulatory requirements, rather than focusing attention on the need to gather better evidence of safety. Currently, most systems rely on passive data collection and there are few that use active ascertainment of adverse events from health professionals. In the UK, reports of such events are actively solicited through the Prescription Event Monitoring Expert Rev. Clin. Pharmacol. 1(5), (2008)
Clinical perspectives in drug safety & adverse drug reactions
system, which surveys prescribers regarding any adverse experiences among the first 10,000 patients prescribed a given drug. In New Zealand, the Intensive Medicines Monitoring Programme actively solicits reports of adverse events and uses a methodology based on establishing cohorts of approximately 10,000 patients monitored for a mean of almost 5 years [43] . The aggregation of adverse events is from a combination of prescription follow-up and intensified spontaneous reporting. The objective is to undertake adverse-event monitoring on selected drugs in the early postmarketing period. Prescription information provides a denominator for calculating rates and the numerator from intensified reporting. Reasons for selection may include drugs with potential safety issues identified during development, new chemical entities or a relationship to a class that has caused problems. Reporting rates are much higher than with traditional voluntary reporting, which increases the chance of detecting unsuspected reactions, such as the identification of cough with captopril and agranulocytosis with mianserin [44,45] . The establishment of large cohorts of patients also provides the opportunity for follow-up studies of signals [46–48] . While more resource intensive, such programmes that actively solicit adverse events can complement spontaneous systems because of their potential for earlier identification and better quantification of risks. Controlled epidemiological studies provide an observational method for detecting and quantifying the frequency of adverse drug effects [49] . Over the past two decades, technological advances have increased the availability of automated databases and the capacity of epidemiologists to analyze them. The UK General Practice Research Database, with population coverage in excess of 3 million, collects population-based data [50] . Its size makes it possible to follow-up large cohorts of users of specific drugs, although it does not provide information on what occurs during hospitalization and is not useful for studying in-patient drug use. Observational healthcare data can describe healthcare encounters, including doctor visits, medicine dispensing, hospital admissions and deaths [51] . Ultimately, it should be possible to develop complete population databases, with well-documented exposures to medicines, outcomes and potential risk factors. Unfortunately, the drive towards increased confidentiality may counter the momentum of such initiatives. From an international perspective, there is merit in bringing all relevant information together in one place, although there is a major logistic difficulty. This is currently attempted with ADR reports by the WHO Collaborating Centre for International Drug Monitoring in Uppsala, Sweden [52] . Established 40 years ago, the WHO programme maintains the international ADR database ‘Vigibase’, which has accumulated more than 3.5 million case reports [53] . A large group of countries participate in the programme. Providing a collating center for a worldwide network of pharmacovigilance centers offers the potential to provide additional gains when compared with operating in isolation. The WHO network is also able to capitalize on the wealth of competence and experience at the disposal of countries involved in international pharmacovigilance. www.expert-reviews.com
Review
Under-reporting of ADRs
Under-reporting is an inherent weakness of adverse-reaction reporting, particularly with the current voluntary reporting schemes. Under-reporting can lead to long delays between marketing and detection/regulatory action of an adverse reaction. Almost 7 million patients were exposed to fenfluramine before the association with valvular heart disease led to its withdrawal from the market [54] . Even in clinical trials, different methods of collecting patient data regarding adverse events can lead to large differences in the reported rates of adverse events, potentially reducing the validity of comparisons between the side-effect profiles of drugs. For example, the reported rates of sexual side effects from selective serotonin-reuptake inhibitors range from 2 to 73%; much of this difference is probably attributable to different methods of data collection [55] . In a recent, randomized, controlled trial, the percentage of patients reporting any adverse event was 14% among patients who were asked an open-ended question, versus 77% of subjects assigned to a checklist [56] . The use of the Consolidated Standards of Reporting Trials (CONSORT) may help to make clinical trials more uniform and comparable [57,58] , and such efforts could be extended to include uniform methods for identifying adverse events. In a systematic review of 37 studies, there was no significant difference in the median under-reporting rates calculated for general-practice and hospital-based studies [59] . Although, in the general-practice studies in this review, there was lower underreporting for more serious or severe ADRs, other studies have found frequent under-reporting for serious and unlabelled reactions [60,61] and for suspected reactions to newly marketed (black triangle) drugs [40] . In general, adverse-event reporting is greatest during the first few years after marketing and reporting may be stimulated by media and scientific attention. The quality of spontaneous reports is also very important for the proper evaluation of drug safety signals and it is preferable to have a small number of well-documented reports than voluminous quantities of substandard quality. Important details enabling causality assessment may be missing from poor-quality reports with limited information. With regard to publishing case reports, the International Society for Pharmacoepidemiology has published guidelines, which provides a useful standard for ADR reports [201] . The denominator for reporting rates can be derived either from sales data, which may not accurately reflect usage levels, or by using the volume of prescriptions as a surrogate of population exposure [35] . It is important to remember that reporting rates are based only on the small percentage of adverse events actually reported. Whenever possible, true population-based incidence rates are the preferred method for describing the level of risk. Inadequate knowledge and attitudes among doctors have been found to be associated with under-reporting [62–65] . Doctors may not report because they believe that serious reactions will be well documented by the time a drug is marketed or that one case will not contribute to medical knowledge [64] . Factors that discourage reporting are trivial or well-known reactions, an uncertain association, lack of awareness of the requirement for reporting, lack of 697
Review
Pillans
understanding of the purpose of spontaneous reporting schemes or insufficient time [41,66,67] . Among doctors, an appropriate interpretation of ‘lack of time’ may be that ADR reporting is not seen as a high priority [66] . Uncertainty regarding a causal association and lack of time are also reasons for not reporting by hospital pharmacists [68] . Clearly, reducing barriers to reporting could have a positive impact on the quality and quantity of postmarketing surveillance data. Improving knowledge and awareness of ADR reporting and provision of a convenient system of submission can substantially increase reporting. Initiatives to increase reporting have included allowing pharmacists and nurses to report ADRs [69,70] and there is evidence that reporting by hospital pharmacists increases the variety and number of ADR reports [69] . Whilst pharmacists in the UK have been found to be less aware than their medical colleagues of the purpose and usefulness of the information collected [62,71] , an Australian study has found that hospital pharmacists are more familiar with the ADR reporting system than doctors and are more likely to report [65] . Reporting was more likely if pharmacists provided assistance with review of case records and completion of reports. A centralized system coordinated and driven by pharmacy departments was found to have a positive impact on hospital ADR reporting. Educational intervention can also improve physician reporting of ADRs, although the effects diminish over time [72,73] . A survey of UK medical and pharmacy schools found that fewer than half provided students with a guide to reporting ADRs [74] . One way to increase ADR reporting is for patients to act as a direct source of information. Consumer reporting of ADRs occurs in some countries such as in the USA, via the MedWatch scheme run by the US FDA [75] . However, the reliability and validity have been questioned because many events related to the underlying illness can be confused with adverse drug effects [76] . Despite an increase in the rate of reporting, it remains questionable whether direct patient reporting increases the ability to generate data regarding previously unsuspected ADRs. A preferred option may be direct patient reporting to trained healthcare interviewers with considerable experience in ADR reporting, who can act as a filter and submit only suspected ADRs to their national centers. In Australia, patient reporting has been facilitated by the Adverse Medicine Events line, a pharmacist-operated telephone helpline, which played an important role in facilitating reporting of bizarre sleep-related events with zolpidem [76,77] . Further work is needed to determine the impact of underreporting on regulatory decisions and the effects of initiatives to improve reporting, such as electronic reporting, pharmacist, nurse and direct consumer reporting, as well as improved training of healthcare professionals. It should be kept in mind, however, that the capacity for expert review of individual reports could be exceeded if the number of reports increases substantially. National ADR monitoring schemes
More than 70 countries, including a number of developing countries, have their own spontaneous reporting systems, which attempt to ensure that signals of possible ADRs are detected as 698
soon as possible after licensing, such as the yellow-card scheme in the UK [78] . Large databases have been developed that are clearly national assets. The Australian and New Zealand Schemes, for example, have played important roles internationally in pharma covigilance [79,80] . Rapport with healthcare professionals and a reporting mindset have been built up over 40 years. Expert medical review of all individual reports is undertaken by staff with clinical and diagnostic skills, which helps to ensure high-quality data. Important information regarding adverse reactions to drugs provides a sound foundation for regulatory and educational activities, as well as a valuable resource for research. Advisory Committees, such as the Australian Adverse Drug Reactions Advisory Committee (ADRAC), are an essential component. The committee comprises a variety of medical experts, meets eight times each year and provides ongoing review of safety issues from individual reports to national and international alerts. Quality is assured by two processes – formalized assessment and peer review. The essential steps involve analysis of the evidence, identification of options and policy decision making. The ADRAC model appears to work reasonably well. Clearly, experts should have a key role in evidence appraisal and developing and assessing assumptions in a formal decision analysis. It is likely, however, that such expert committees will be subjected to the increasing degree of scrutiny involved with issues involving public safety. The reporting of suspected adverse reactions by health professionals to the ADRAC has generated a number of important signals, helped to define the nature of the adverse reaction in question and some of the reporting doctors have then been able to assist in further research [81] . An understanding of factors involved in flucloxacillin-induced hepatitis, such as delayed time to onset, predominant cholestatic pattern and delayed recovery, arose from Australian reports [82,83] . Similarly, the large number and detail of ADRAC reports of amoxicillin/clavulanate-induced hepatitis, established that many cases have a delayed onset and typically a cholestatic pattern [81,84] . Handling of signals
A continuous systematic review of all combinations of drugs and suspected ADRs reported to spontaneous reporting centers is necessary to optimize signal detection. Quantitative procedures can be used to aid human reviewers by sifting data to determine the extent to which an ADR is reported disproportionally to a certain drug [85] . The WHO, FDA and some government agencies use statistical methods of data mining to help discern which combinations of drugs and adverse events in their database suggest a causal link. Data mining can trigger additional analysis of individual cases or population-based studies. Techniques such as proportional reporting ratios compare the proportion of an ADR for a specific drug within a database with the background proportion for that ADR for all drugs in the database [86] . This eliminates the need for denominator information and also counteracts some of the problems associated with unknown variation in under-reporting. However, the use of algorithms should be seen as a tool or an aid to signal detection. Clinical medicine, and not statistics, should continue to Expert Rev. Clin. Pharmacol. 1(5), (2008)
Clinical perspectives in drug safety & adverse drug reactions
play the leadership role in causal research [87] . Well-developed clinical skills and judgement remain crucial in the discovery and documentation of ADRs. The judgement made regarding the priority that should be given to a signal requires evaluation of its strength, whether it is new, the clinical importance and the potential for prevention. Currently used methods in prioritizing signals may still employ collective judgement of experienced pharmacovigilance personnel. An alternative tool is impact analysis, a method developed for use at the UK Medicines and Healthcare Products Regulatory Agency [88] . This uses a mathematical approach to analyze the impact of signals arising from spontaneous reporting data with the purpose of focusing detailed signal evaluation on those signals that are the strongest and are most likely to have an impact on public health. Filtering techniques are important for prioritizing signals in large databases and the WHO have developed triage algorithms for this purpose [53] . Factors considered to be relevant for signal priority include ADR unexpectedness, seriousness, disproportionality, reports from multiple countries and positive rechallenge. Frequency should be measured both in terms of risk to the individual and the population frequency, which requires knowledge of the level and duration of exposure. Absolute frequency can be categorized in four levels of magnitude, from one in ten or more, to one in 10,000 or less. The identification of risk factors is most important in developing strategies to prevent ADRs. It is important to consider not only who is at risk but when they are at risk (i.e., the temporal relationship between commencement of treatment and onset of the reaction). Recognition of a potential link between a drug and an adverse event may be aided by the degree of unexpectedness of the event. Highly unusual or infrequent outcomes, such as pure red cell aplasia associated with epoetin therapy [89] , are strong triggers of suspicion. Careful analysis of the initial 175 cases of epoetin-associated pure red cell aplasia revealed risk factors that may have contributed to the development of aplasia, which then enabled appropriate procedures to address the issue. Awareness may also be increased when an unexpected and otherwise unexplained adverse event is often drug related, such as agranulocytosis or pancytopenia, SJS, fulminant liver failure and Torsades de pointes. When an adverse event occurs at a high background rate or may have arisen from the underlying disease, attributing causality is more difficult and more importance has to be attached to the temporal relationship, outcome when the drug is discontinued or recurrence on re-exposure. An example to illustrate the handling of an unusual signal was that of an outbreak of hyoscine hydrobromide toxicity detected through the Australian pharmacovigilance system [90] . In a 7-day period in January 2003, six local reports from three Australian states of suspected adverse reactions to a product sold to prevent and treat travel sickness were received. Each tablet was stated to contain dimenhydrinate 50 mg, hyoscine hydrobromide 0.2 mg and caffeine 20 mg. Reactions were characterized by confusion, visual and auditory hallucinations, agitation, anxiety and unsteady gait, which were consistent with excessive doses of hyoscine. Adverse effects of hyoscine hydrobromide at therapeutic doses are usually limited to dry mouth, blurred vision and mydriasis. www.expert-reviews.com
Review
Analysis of the suspect batches identified that the hyoscine bromide content of individual tablets ranged from ‘not detected’ to 707% of the label claim. After 11 days following the first report being received, there was a national recall with further investigation into breaches of Good Manufacturing Practice. The recall stimulated a further 104 reports implicating two of the batches. Three factors contributed to the prompt recognition and action. First, the reports were reported directly to a national center. Second, early identification of a possible problem followed prompt clinical review of incoming reports in compliance with an in-house performance requirement. The importance of a central database where a clustering of a particular pattern of reactions can be readily identified, is well illustrated. Third, access to laboratory analyses provided a prompt explanation for the clinical observations. Communicating new safety information
After identification of drug safety issues, it is important to close the loop by communicating with and influencing users of medicines in order to promote safe use and protect public health. Changes to product information, drug safety bulletins, ‘Dear Doctor’ letters and the internet are methods employed currently. The ADRAC’s major vehicle for dissemination of information, including alerting health professionals to important safety issues, is the Australian Adverse Drug Reactions Bulletin, which is sent to all practitioners in the country. Numerous important associations are highlighted, most of which are stimulated by Australian reports. Product information serves as both a regulatory document and as a primary source of information for health professionals. At individual patient level, avoiding re-exposure with lifethreatening ADRs, such as anaphylaxis and fulminant liver failure, is of paramount importance. All patients and/or parents in the case of children, should be informed of their adverse reaction and the risks of rechallenge. For serious reactions, consideration should be given to the use of medical identification devices in the form of a bracelet, necklace or wallet card [91] . Hospitals have various strategies to prevent further misadventure. At the Princess Alexandra Hospital (Brisbane, Australia), a large, bright-yellow ADR sticker is applied to the chart of any patient that has experienced an important adverse reaction. A further, small, yellow ADR sticker is also placed on their medication chart, which includes a red box for details of the ADR(s). Patients also wear a red ADR arm bracelet. A copy of the ADR form, which is sent to the national reporting committee (ADRAC), is sent to the patient’s general practitioner (GP). While all these measures are designed to avoid inadvertent re-exposure, the loss of information on ADRs by poor communication between hospitals and primary care is a real risk. Van der Linden et al. found that approximately a quarter of the drug treatments withdrawn during hospitalization because of an ADR were represcribed within 6 months of discharge, irrespective of the seriousness of the ADR [92] . In addition, transfer of information to GPs and documentation by GPs was poor. Better communication on ADRs between different healthcare echelons should reduce inadvertent re-exposure, which may require the development of national electronic medication files. 699
Review
Pillans
Drug regulation & drug safety
A review of the list of drugs withdrawn from the market can provide useful insight into postmarketing surveillance, as well as the drug approval process. In a review of drug withdrawals for safety reasons in the USA, more than 75 drugs/drug products were removed during the 33-year period from 1969 to 2002. A further 11 drugs had special requirements for prescriptions or had restricted distribution programs [38] . Numerous other drugs required the addition of a ‘black box’ warning [93] . During a similar period, 41 products were identified that had been withdrawn from the Canadian market, potentially for safety reasons [37] . Examples of international drug withdrawals have included fulminant liver failure with troglitazone [94,95] , QT prolongation and ventricular arrhythmias with cisapride [96] , hemolytic anemia, coagulopathy and renal or hepatic dysfunction with temafloxacin [97] and fatal rhabdomyolysis with cerivastatin [98] . The annual number has increased over time, which may merely reflect the expanding number of marketed drugs or may, in part, be due to more sophisticated methods for identifying safety issues, less stringent approval criteria or increasing availability of international safety data. Regulatory authorities have tended to act later rather than sooner in response to safety signals and, when combined with under-reporting, this may have led to exposure of a large number of patients to drug-related harm before restriction or withdrawal. An estimated 4 million people were exposed to bromfenac (hepato toxicity), dexfenfluramine/fenfluramine (cardiac valvulopathy), mebefradil (drug interactions/cardiogenic shock) and terfenadine (ventricular arrhythmias/drug interactions) before they were withdrawn from the market between 1997 and 1998 [6] . Gatifloxacin was first marketed in 1999 in the USA and reports of dysglycemic effects appeared soon after the drug’s approval [99] . However, it was not until 2006 that the manufacturers announced changes to the drug’s label, which included mention of some cases with fatal outcomes [100] . An evaluation of safety withdrawals from worldwide pharmaceutical markets of 87 products over almost 40 years found that the median time on the market was 5.4 years, with only approximately a third withdrawn in the first 2 years [101] . This underscores the need for a comprehensive plan to respond appropriately and expeditiously to potential drug safety signals in order to reduce unnecessary complications and deaths [10] . The most well-known recent example of inadequate pharmaco vigilance is rofecoxib (Vioxx®), where both the manufacturers and the FDA ‘failed the public health’ [3] . Rofecoxib was marketed in 1999 as an effective, safer alternative to conventional NSAIDs. Despite the drug’s potential for adverse cardiovascular effects by altering the ratio of prostacyclin to thromboxane [102,103] , none of the intervention studies that constituted its new drug application were designed to evaluate cardiovascular risk [104,105] . A cautionary flag in 2001 regarding the risk of cardiovascular events associated with selective COX-2 inhibitors, based on an excess number of myocardial infarctions associated with rofecoxib in the Vioxx Gastrointestinal Outcome Research (VIGOR) trial [106] , led to the recommendation that it was ‘mandatory to conduct a trial to specifically assess cardiovascular risk and benefit’ [107] . Unfortunately, 700
such a trial was never conducted. Epidemiological studies confirmed and amplified the concern [108] . Only by happenstance, in a trial involving patients with colon polyps, the Adenomatous Polyp Prevention on Vioxx (APPROVe) trial, was it discovered that 3.5% of the patients assigned to rofecoxib had myocardial infarction or stroke compared with 1.9% of patients on placebo (p
Clinical perspectives in drug safety and adverse drug reactions Expert Rev. Clin. Pharmacol. 1(5), 695–705 (2008)
Peter Ian Pillans Princess Alexandra Hospital, Woolloongabba, Brisbane, 4102, Queensland, Australia Tel.: +61 073 240 2693 Fax: +61 073 240 7131 [email protected]
Adverse drug reactions (ADRs) remain a common clinical problem since they can mimic many diseases and cause significant morbidity and mortality. Judicious prescribing is important to minimize their occurrence. Apart from the recent identification of a few pharmacogenomic biomarkers for serious reactions, many remain unpredictable. Spontaneous reporting continues to play an important role in pharmacovigilance and the value of astute clinical observation and well-documented reports of suspicions of a causal link cannot be underestimated. Many national reporting schemes have developed considerable experience and expertise over many years and have large ADR databases, which are national assets. Despite advances in pharmacovigilance, numerous deficiencies have been identified; postmarketing surveillance remains the weakest link in the regulatory process. Regulatory authorities have tended to act later rather than sooner in response to safety signals, and this, when combined with under-reporting, may have led to exposure of a large number of patients to drug-related harm before restriction or withdrawal. In an attempt to improve vigilance, international surveillance may benefit by moving from its current passive/reactive mode toward active surveillance systems with a prospective, comprehensive and systematic approach to monitoring, collecting, analyzing and reporting data on ADRs. This will include increased pressure on pharmaceutical companies to conduct postmarketing studies. Such an active/proactive approach, while maintaining focus on ADR detection, could also aim to extend knowledge of safety, such that emerging changes in risk–benefit during a drug’s marketed life are effectively communicated to clinicians and patients. Drug safety monitoring and its regulation are now undergoing an overhaul and it is hoped that vigilance, public safety and trust will improve as a result. Keywords : adverse drug reaction • drug safety • pharmacovigilance • postmarketing surveillance
Over the past 5 years, a series of recalls of highprofile prescription medicines has aroused serious concerns regarding the safety of medicines [1–6] . Clinical trials and routine regulatory oversight, as practiced currently, often fail to uncover important adverse effects for widely marketed products [5] . Medical journals have published numerous articles and editorials relating to drug safety and regulation, with recommendations to overhaul drug safety monitoring, improve vigilance, ensure greater protection to the health of the public and restore trust [6–13] . The first part of this review will deal with adverse drug reactions (ADRs), with a brief overview of the epidemiology, detection of ADRs and pharmacovigilance, reporting of ADRs, the role of national monitoring schemes, handling of signals and the communication of www.expert-reviews.com
10.1586/17512433.1.5.695
safety information. The second part addresses drug-regulatory issues pertinent to drug safety in the current climate. Adverse drug reactions
An ADR is an unintended response, occuring at doses used in humans [14] . It may predict hazards for future administration or warrant prevention or specific treatment, alteration of the dosage regimen or withdrawal. ADRs are typically divided into those that are dose dependent and predictable, based on the pharmacology of the drug (Type A or augmented) and those that are unpredictable, dose independent, unrelated to the drug’s pharmacology and often immuno logically mediated (Type B or bizarre). Since, it is sometimes difficult to assign a reaction to one type, more-complex subdivisions have been proposed, based on dose relatedness, timing and
© 2008 Expert Reviews Ltd
ISSN 1751-2433
695
Review
Pillans
patient susceptibility [15] . In this comprehensive classification, other criteria are taken into account, including properties of the reaction (the time course of its appearance and its severity) and properties of the individual (the genetic, pathological and other biological differences that confer susceptibility). The definition of an ADR is distinct from an adverse drug event, which is an ‘umbrella term’ to describe any adverse experience that occurs in a person on a medication that may or may not be causally related. Epidemiology of ADRs
Adverse drug reactions are an important cause of iatrogenic disease [16] . They can involve any organ system, can present clinically in many different ways and have taken over from syphilis and tuberculosis as the disease mimic [17] . Although estimates of the incidence of ADRs vary, they are an important cause of morbidity, mortality and wasted expenditure [18] . In a large meta-analysis of 39 prospective studies from US hospitals from 1966 to 1996, the annual incidence of severe ADRs (defined as those that required hospitalization, were permanently disabling or resulted in death, occurring while in hospital or as the cause of admission to hospital) was 6.7% [19] . In a recent study involving a large population of hospitalized Medicare patients, 1.7% experienced an ADR [20] . A similar ADR prevalence of 1.5% was found in another recent study involving 4.3 million patients with adverse drug events who visited doctor’s offices, hospital outpatient clinics and emergency facilities [21] . A review of 14 Australian studies found that ADRs occurred in 2.4–3.6% of hospital admissions [22] . Despite programs to promote rational and safer use of medicines in West Australia, a study that was limited to ADRs of sufficient severity to warrant or extend hospitalization found the rate of ADRrelated hospital stays increased from 2.5 per 1000 person years in 1981 to 12.9 per 1000 patient years in 2002 [23] . The largest increases occurred in those aged over 80 years. In another study in Australian general practice, 10% of patients had experienced an adverse drug event in the previous 6 months [24] . Patients aged over 45 years, children aged 1–4 years and female patients were significantly more likely to have experienced an adverse drug event in this study. In the UK, 6.5% of hospital admissions were found to be medication related [25] . The median age of patients admitted with ADRs was 76 years, which was significantly older than patients without ADRs (median age: 66 years). In some studies, the public-health impact is likely to be underestimated, because many adverse events do not result in hospital admission [26] . Many ADRs are potentially preventable, particularly those that are predictable (Type A) or where the patient has a known drug hypersensitivity [19,27,28] . Clearly, judicious prescribing with a sound knowledge of the pharmacology of the drugs in question, including relevant pharmacokinetics, contraindications and interactions, together with consideration of the comordities in the patient at hand, is important in the prevention of ADRs. For a very small number of drugs, where in the past the ADR was considered unpredictable, there are now pharmacogenomic biomarkers for prediction of serious reactions [29] . Examples include the association 696
between the HLA‑B* 5701 gene variant and hypersensitivity reactions to abacavir [30] , HLA-B*1502 and carbamazepineinduced Stevens–Johnson syndrome (SJS) in Han Chinese [31] , HLA-B*5801 and allopurinol-induced SJS [32] , mutant alleles of cytochrome P450 (CYP)2C9 and bleeding with warfarin [33] and a thiopurine methyltransferase genotype associated with deficiency of the enzyme and severe myelotoxicity from azathioprine [34] . With some exceptions, routine pharmacogenetic testing for only two of these biomarkers, CYP2C9 and thiopurine methyltransferase, is conducted in some centers at the present time. Detection of ADRs & pharmacovigilance
Discovery in an observational science such as pharmacovigilance depends on the capacity to recognize and investigate unexpected clinical events that are manifest once a new drug is in use [35] . Pharmacovigilance is the science related to the detection, assessment, understanding and prevention of ADRs [11] . Its primary function is to improve the safety of marketed medicines. Data are derived from many sources, including published case reports, voluntary ADR reporting to national pharmacovigilance centers, postmarketing clinical and epidemiological studies, prescription event monitoring schemes morbidity and mortality databases [36] . The key stakeholders are patients, health professionals, governments and pharmaceutical companies. The current system of voluntary reporting plays an important role in identifying ADRs [37,38] . This voluntary scheme of ADR reporting has reached a high level of sophistication internationally, although the limitations are well known: poor quality of some submitted reports, significant under-reporting [39–41] , difficulty in calculating rates because of incomplete numerator data along with unreliable denominators and limited ability to establish cause and effect [42] . While this methodology has been responsible for the identification of numerous safety signals, there are more fundamental reasons to continue with a system of voluntary reporting. Promoting the concept that serious or unexpected adverse reactions should be reported to relevant national pharmacovigilance centers emphasizes the importance of drug safety issues to those involved. A unique feature is that it may detect reactions that might otherwise go unrecorded because it entails suspicions of cause and effect that health professionals consider potentially important. Furthermore, the reports allow for a qualitative description of how the reaction affects patient’s lives. Pharmaceutical companies play an important role in gathering ADR reports on their own products. Periodic Safety Update Reports (PSURs) are generated, which provide global safety updates. While detection of a new ADR is likely to lead to an amendment to the product information, the emphasis has been on the fulfilment of regulatory requirements, rather than focusing attention on the need to gather better evidence of safety. Currently, most systems rely on passive data collection and there are few that use active ascertainment of adverse events from health professionals. In the UK, reports of such events are actively solicited through the Prescription Event Monitoring Expert Rev. Clin. Pharmacol. 1(5), (2008)
Clinical perspectives in drug safety & adverse drug reactions
system, which surveys prescribers regarding any adverse experiences among the first 10,000 patients prescribed a given drug. In New Zealand, the Intensive Medicines Monitoring Programme actively solicits reports of adverse events and uses a methodology based on establishing cohorts of approximately 10,000 patients monitored for a mean of almost 5 years [43] . The aggregation of adverse events is from a combination of prescription follow-up and intensified spontaneous reporting. The objective is to undertake adverse-event monitoring on selected drugs in the early postmarketing period. Prescription information provides a denominator for calculating rates and the numerator from intensified reporting. Reasons for selection may include drugs with potential safety issues identified during development, new chemical entities or a relationship to a class that has caused problems. Reporting rates are much higher than with traditional voluntary reporting, which increases the chance of detecting unsuspected reactions, such as the identification of cough with captopril and agranulocytosis with mianserin [44,45] . The establishment of large cohorts of patients also provides the opportunity for follow-up studies of signals [46–48] . While more resource intensive, such programmes that actively solicit adverse events can complement spontaneous systems because of their potential for earlier identification and better quantification of risks. Controlled epidemiological studies provide an observational method for detecting and quantifying the frequency of adverse drug effects [49] . Over the past two decades, technological advances have increased the availability of automated databases and the capacity of epidemiologists to analyze them. The UK General Practice Research Database, with population coverage in excess of 3 million, collects population-based data [50] . Its size makes it possible to follow-up large cohorts of users of specific drugs, although it does not provide information on what occurs during hospitalization and is not useful for studying in-patient drug use. Observational healthcare data can describe healthcare encounters, including doctor visits, medicine dispensing, hospital admissions and deaths [51] . Ultimately, it should be possible to develop complete population databases, with well-documented exposures to medicines, outcomes and potential risk factors. Unfortunately, the drive towards increased confidentiality may counter the momentum of such initiatives. From an international perspective, there is merit in bringing all relevant information together in one place, although there is a major logistic difficulty. This is currently attempted with ADR reports by the WHO Collaborating Centre for International Drug Monitoring in Uppsala, Sweden [52] . Established 40 years ago, the WHO programme maintains the international ADR database ‘Vigibase’, which has accumulated more than 3.5 million case reports [53] . A large group of countries participate in the programme. Providing a collating center for a worldwide network of pharmacovigilance centers offers the potential to provide additional gains when compared with operating in isolation. The WHO network is also able to capitalize on the wealth of competence and experience at the disposal of countries involved in international pharmacovigilance. www.expert-reviews.com
Review
Under-reporting of ADRs
Under-reporting is an inherent weakness of adverse-reaction reporting, particularly with the current voluntary reporting schemes. Under-reporting can lead to long delays between marketing and detection/regulatory action of an adverse reaction. Almost 7 million patients were exposed to fenfluramine before the association with valvular heart disease led to its withdrawal from the market [54] . Even in clinical trials, different methods of collecting patient data regarding adverse events can lead to large differences in the reported rates of adverse events, potentially reducing the validity of comparisons between the side-effect profiles of drugs. For example, the reported rates of sexual side effects from selective serotonin-reuptake inhibitors range from 2 to 73%; much of this difference is probably attributable to different methods of data collection [55] . In a recent, randomized, controlled trial, the percentage of patients reporting any adverse event was 14% among patients who were asked an open-ended question, versus 77% of subjects assigned to a checklist [56] . The use of the Consolidated Standards of Reporting Trials (CONSORT) may help to make clinical trials more uniform and comparable [57,58] , and such efforts could be extended to include uniform methods for identifying adverse events. In a systematic review of 37 studies, there was no significant difference in the median under-reporting rates calculated for general-practice and hospital-based studies [59] . Although, in the general-practice studies in this review, there was lower underreporting for more serious or severe ADRs, other studies have found frequent under-reporting for serious and unlabelled reactions [60,61] and for suspected reactions to newly marketed (black triangle) drugs [40] . In general, adverse-event reporting is greatest during the first few years after marketing and reporting may be stimulated by media and scientific attention. The quality of spontaneous reports is also very important for the proper evaluation of drug safety signals and it is preferable to have a small number of well-documented reports than voluminous quantities of substandard quality. Important details enabling causality assessment may be missing from poor-quality reports with limited information. With regard to publishing case reports, the International Society for Pharmacoepidemiology has published guidelines, which provides a useful standard for ADR reports [201] . The denominator for reporting rates can be derived either from sales data, which may not accurately reflect usage levels, or by using the volume of prescriptions as a surrogate of population exposure [35] . It is important to remember that reporting rates are based only on the small percentage of adverse events actually reported. Whenever possible, true population-based incidence rates are the preferred method for describing the level of risk. Inadequate knowledge and attitudes among doctors have been found to be associated with under-reporting [62–65] . Doctors may not report because they believe that serious reactions will be well documented by the time a drug is marketed or that one case will not contribute to medical knowledge [64] . Factors that discourage reporting are trivial or well-known reactions, an uncertain association, lack of awareness of the requirement for reporting, lack of 697
Review
Pillans
understanding of the purpose of spontaneous reporting schemes or insufficient time [41,66,67] . Among doctors, an appropriate interpretation of ‘lack of time’ may be that ADR reporting is not seen as a high priority [66] . Uncertainty regarding a causal association and lack of time are also reasons for not reporting by hospital pharmacists [68] . Clearly, reducing barriers to reporting could have a positive impact on the quality and quantity of postmarketing surveillance data. Improving knowledge and awareness of ADR reporting and provision of a convenient system of submission can substantially increase reporting. Initiatives to increase reporting have included allowing pharmacists and nurses to report ADRs [69,70] and there is evidence that reporting by hospital pharmacists increases the variety and number of ADR reports [69] . Whilst pharmacists in the UK have been found to be less aware than their medical colleagues of the purpose and usefulness of the information collected [62,71] , an Australian study has found that hospital pharmacists are more familiar with the ADR reporting system than doctors and are more likely to report [65] . Reporting was more likely if pharmacists provided assistance with review of case records and completion of reports. A centralized system coordinated and driven by pharmacy departments was found to have a positive impact on hospital ADR reporting. Educational intervention can also improve physician reporting of ADRs, although the effects diminish over time [72,73] . A survey of UK medical and pharmacy schools found that fewer than half provided students with a guide to reporting ADRs [74] . One way to increase ADR reporting is for patients to act as a direct source of information. Consumer reporting of ADRs occurs in some countries such as in the USA, via the MedWatch scheme run by the US FDA [75] . However, the reliability and validity have been questioned because many events related to the underlying illness can be confused with adverse drug effects [76] . Despite an increase in the rate of reporting, it remains questionable whether direct patient reporting increases the ability to generate data regarding previously unsuspected ADRs. A preferred option may be direct patient reporting to trained healthcare interviewers with considerable experience in ADR reporting, who can act as a filter and submit only suspected ADRs to their national centers. In Australia, patient reporting has been facilitated by the Adverse Medicine Events line, a pharmacist-operated telephone helpline, which played an important role in facilitating reporting of bizarre sleep-related events with zolpidem [76,77] . Further work is needed to determine the impact of underreporting on regulatory decisions and the effects of initiatives to improve reporting, such as electronic reporting, pharmacist, nurse and direct consumer reporting, as well as improved training of healthcare professionals. It should be kept in mind, however, that the capacity for expert review of individual reports could be exceeded if the number of reports increases substantially. National ADR monitoring schemes
More than 70 countries, including a number of developing countries, have their own spontaneous reporting systems, which attempt to ensure that signals of possible ADRs are detected as 698
soon as possible after licensing, such as the yellow-card scheme in the UK [78] . Large databases have been developed that are clearly national assets. The Australian and New Zealand Schemes, for example, have played important roles internationally in pharma covigilance [79,80] . Rapport with healthcare professionals and a reporting mindset have been built up over 40 years. Expert medical review of all individual reports is undertaken by staff with clinical and diagnostic skills, which helps to ensure high-quality data. Important information regarding adverse reactions to drugs provides a sound foundation for regulatory and educational activities, as well as a valuable resource for research. Advisory Committees, such as the Australian Adverse Drug Reactions Advisory Committee (ADRAC), are an essential component. The committee comprises a variety of medical experts, meets eight times each year and provides ongoing review of safety issues from individual reports to national and international alerts. Quality is assured by two processes – formalized assessment and peer review. The essential steps involve analysis of the evidence, identification of options and policy decision making. The ADRAC model appears to work reasonably well. Clearly, experts should have a key role in evidence appraisal and developing and assessing assumptions in a formal decision analysis. It is likely, however, that such expert committees will be subjected to the increasing degree of scrutiny involved with issues involving public safety. The reporting of suspected adverse reactions by health professionals to the ADRAC has generated a number of important signals, helped to define the nature of the adverse reaction in question and some of the reporting doctors have then been able to assist in further research [81] . An understanding of factors involved in flucloxacillin-induced hepatitis, such as delayed time to onset, predominant cholestatic pattern and delayed recovery, arose from Australian reports [82,83] . Similarly, the large number and detail of ADRAC reports of amoxicillin/clavulanate-induced hepatitis, established that many cases have a delayed onset and typically a cholestatic pattern [81,84] . Handling of signals
A continuous systematic review of all combinations of drugs and suspected ADRs reported to spontaneous reporting centers is necessary to optimize signal detection. Quantitative procedures can be used to aid human reviewers by sifting data to determine the extent to which an ADR is reported disproportionally to a certain drug [85] . The WHO, FDA and some government agencies use statistical methods of data mining to help discern which combinations of drugs and adverse events in their database suggest a causal link. Data mining can trigger additional analysis of individual cases or population-based studies. Techniques such as proportional reporting ratios compare the proportion of an ADR for a specific drug within a database with the background proportion for that ADR for all drugs in the database [86] . This eliminates the need for denominator information and also counteracts some of the problems associated with unknown variation in under-reporting. However, the use of algorithms should be seen as a tool or an aid to signal detection. Clinical medicine, and not statistics, should continue to Expert Rev. Clin. Pharmacol. 1(5), (2008)
Clinical perspectives in drug safety & adverse drug reactions
play the leadership role in causal research [87] . Well-developed clinical skills and judgement remain crucial in the discovery and documentation of ADRs. The judgement made regarding the priority that should be given to a signal requires evaluation of its strength, whether it is new, the clinical importance and the potential for prevention. Currently used methods in prioritizing signals may still employ collective judgement of experienced pharmacovigilance personnel. An alternative tool is impact analysis, a method developed for use at the UK Medicines and Healthcare Products Regulatory Agency [88] . This uses a mathematical approach to analyze the impact of signals arising from spontaneous reporting data with the purpose of focusing detailed signal evaluation on those signals that are the strongest and are most likely to have an impact on public health. Filtering techniques are important for prioritizing signals in large databases and the WHO have developed triage algorithms for this purpose [53] . Factors considered to be relevant for signal priority include ADR unexpectedness, seriousness, disproportionality, reports from multiple countries and positive rechallenge. Frequency should be measured both in terms of risk to the individual and the population frequency, which requires knowledge of the level and duration of exposure. Absolute frequency can be categorized in four levels of magnitude, from one in ten or more, to one in 10,000 or less. The identification of risk factors is most important in developing strategies to prevent ADRs. It is important to consider not only who is at risk but when they are at risk (i.e., the temporal relationship between commencement of treatment and onset of the reaction). Recognition of a potential link between a drug and an adverse event may be aided by the degree of unexpectedness of the event. Highly unusual or infrequent outcomes, such as pure red cell aplasia associated with epoetin therapy [89] , are strong triggers of suspicion. Careful analysis of the initial 175 cases of epoetin-associated pure red cell aplasia revealed risk factors that may have contributed to the development of aplasia, which then enabled appropriate procedures to address the issue. Awareness may also be increased when an unexpected and otherwise unexplained adverse event is often drug related, such as agranulocytosis or pancytopenia, SJS, fulminant liver failure and Torsades de pointes. When an adverse event occurs at a high background rate or may have arisen from the underlying disease, attributing causality is more difficult and more importance has to be attached to the temporal relationship, outcome when the drug is discontinued or recurrence on re-exposure. An example to illustrate the handling of an unusual signal was that of an outbreak of hyoscine hydrobromide toxicity detected through the Australian pharmacovigilance system [90] . In a 7-day period in January 2003, six local reports from three Australian states of suspected adverse reactions to a product sold to prevent and treat travel sickness were received. Each tablet was stated to contain dimenhydrinate 50 mg, hyoscine hydrobromide 0.2 mg and caffeine 20 mg. Reactions were characterized by confusion, visual and auditory hallucinations, agitation, anxiety and unsteady gait, which were consistent with excessive doses of hyoscine. Adverse effects of hyoscine hydrobromide at therapeutic doses are usually limited to dry mouth, blurred vision and mydriasis. www.expert-reviews.com
Review
Analysis of the suspect batches identified that the hyoscine bromide content of individual tablets ranged from ‘not detected’ to 707% of the label claim. After 11 days following the first report being received, there was a national recall with further investigation into breaches of Good Manufacturing Practice. The recall stimulated a further 104 reports implicating two of the batches. Three factors contributed to the prompt recognition and action. First, the reports were reported directly to a national center. Second, early identification of a possible problem followed prompt clinical review of incoming reports in compliance with an in-house performance requirement. The importance of a central database where a clustering of a particular pattern of reactions can be readily identified, is well illustrated. Third, access to laboratory analyses provided a prompt explanation for the clinical observations. Communicating new safety information
After identification of drug safety issues, it is important to close the loop by communicating with and influencing users of medicines in order to promote safe use and protect public health. Changes to product information, drug safety bulletins, ‘Dear Doctor’ letters and the internet are methods employed currently. The ADRAC’s major vehicle for dissemination of information, including alerting health professionals to important safety issues, is the Australian Adverse Drug Reactions Bulletin, which is sent to all practitioners in the country. Numerous important associations are highlighted, most of which are stimulated by Australian reports. Product information serves as both a regulatory document and as a primary source of information for health professionals. At individual patient level, avoiding re-exposure with lifethreatening ADRs, such as anaphylaxis and fulminant liver failure, is of paramount importance. All patients and/or parents in the case of children, should be informed of their adverse reaction and the risks of rechallenge. For serious reactions, consideration should be given to the use of medical identification devices in the form of a bracelet, necklace or wallet card [91] . Hospitals have various strategies to prevent further misadventure. At the Princess Alexandra Hospital (Brisbane, Australia), a large, bright-yellow ADR sticker is applied to the chart of any patient that has experienced an important adverse reaction. A further, small, yellow ADR sticker is also placed on their medication chart, which includes a red box for details of the ADR(s). Patients also wear a red ADR arm bracelet. A copy of the ADR form, which is sent to the national reporting committee (ADRAC), is sent to the patient’s general practitioner (GP). While all these measures are designed to avoid inadvertent re-exposure, the loss of information on ADRs by poor communication between hospitals and primary care is a real risk. Van der Linden et al. found that approximately a quarter of the drug treatments withdrawn during hospitalization because of an ADR were represcribed within 6 months of discharge, irrespective of the seriousness of the ADR [92] . In addition, transfer of information to GPs and documentation by GPs was poor. Better communication on ADRs between different healthcare echelons should reduce inadvertent re-exposure, which may require the development of national electronic medication files. 699
Review
Pillans
Drug regulation & drug safety
A review of the list of drugs withdrawn from the market can provide useful insight into postmarketing surveillance, as well as the drug approval process. In a review of drug withdrawals for safety reasons in the USA, more than 75 drugs/drug products were removed during the 33-year period from 1969 to 2002. A further 11 drugs had special requirements for prescriptions or had restricted distribution programs [38] . Numerous other drugs required the addition of a ‘black box’ warning [93] . During a similar period, 41 products were identified that had been withdrawn from the Canadian market, potentially for safety reasons [37] . Examples of international drug withdrawals have included fulminant liver failure with troglitazone [94,95] , QT prolongation and ventricular arrhythmias with cisapride [96] , hemolytic anemia, coagulopathy and renal or hepatic dysfunction with temafloxacin [97] and fatal rhabdomyolysis with cerivastatin [98] . The annual number has increased over time, which may merely reflect the expanding number of marketed drugs or may, in part, be due to more sophisticated methods for identifying safety issues, less stringent approval criteria or increasing availability of international safety data. Regulatory authorities have tended to act later rather than sooner in response to safety signals and, when combined with under-reporting, this may have led to exposure of a large number of patients to drug-related harm before restriction or withdrawal. An estimated 4 million people were exposed to bromfenac (hepato toxicity), dexfenfluramine/fenfluramine (cardiac valvulopathy), mebefradil (drug interactions/cardiogenic shock) and terfenadine (ventricular arrhythmias/drug interactions) before they were withdrawn from the market between 1997 and 1998 [6] . Gatifloxacin was first marketed in 1999 in the USA and reports of dysglycemic effects appeared soon after the drug’s approval [99] . However, it was not until 2006 that the manufacturers announced changes to the drug’s label, which included mention of some cases with fatal outcomes [100] . An evaluation of safety withdrawals from worldwide pharmaceutical markets of 87 products over almost 40 years found that the median time on the market was 5.4 years, with only approximately a third withdrawn in the first 2 years [101] . This underscores the need for a comprehensive plan to respond appropriately and expeditiously to potential drug safety signals in order to reduce unnecessary complications and deaths [10] . The most well-known recent example of inadequate pharmaco vigilance is rofecoxib (Vioxx®), where both the manufacturers and the FDA ‘failed the public health’ [3] . Rofecoxib was marketed in 1999 as an effective, safer alternative to conventional NSAIDs. Despite the drug’s potential for adverse cardiovascular effects by altering the ratio of prostacyclin to thromboxane [102,103] , none of the intervention studies that constituted its new drug application were designed to evaluate cardiovascular risk [104,105] . A cautionary flag in 2001 regarding the risk of cardiovascular events associated with selective COX-2 inhibitors, based on an excess number of myocardial infarctions associated with rofecoxib in the Vioxx Gastrointestinal Outcome Research (VIGOR) trial [106] , led to the recommendation that it was ‘mandatory to conduct a trial to specifically assess cardiovascular risk and benefit’ [107] . Unfortunately, 700
such a trial was never conducted. Epidemiological studies confirmed and amplified the concern [108] . Only by happenstance, in a trial involving patients with colon polyps, the Adenomatous Polyp Prevention on Vioxx (APPROVe) trial, was it discovered that 3.5% of the patients assigned to rofecoxib had myocardial infarction or stroke compared with 1.9% of patients on placebo (p
