Eur J Clin Pharmacol (1990) 39:321-325


Springer-Verlag 1990

Special article Risk factors for adverse drug reactions - epidemiological approaches* R. Hoign6 ~, D. H. Lawson 2, and E. Weber 3 Comprehensive Hospital Drug Monitoring, Zieglerspital, Bern, Switzerland 2 ClinicalPharmacology,Royal Infirmary, Glasgow,UK and 3 Abteilung Klinische Pharmakologie, Medizinische Universit~tsklinik, Heidelberg, FRG Received: April 20, 1990/Accepted:July 2, 1990

Summary. Age by itself is not an important risk factor for ADRs. Age-related changes are the consequence of a number of individual factors, for example morbidity associated with polypharmacy, decline in renal or liver function in the elderly, hypoalbuminaemia, reduced body weight, etc. The relationship between gastrointestinal bleeding and non-steroidal anti-inflammatory drugs can be assessed globally in large cohort studies with access to computerized data, but complete accuracy requires access to the original patient records. The increase in the risk of GI bleeding in users of NSAIDs and aspirin was 50% above that in non-users. About a quarter of ADRs in hospitalized patients seem not to arise from purely pharmacological mechanisms. They are mainly due to allergic, anaphylactoid, or idiosyncratic reactions and to intolerance. In such nonpharmacological reactions, the time of exposure, reaction time, and even dosage may be important factors in identification of the causal drug. The use of benzodiazepines can be optimized by taking into account potency, time of action and the different syndromes encountered after withdrawal. Following long-term use problems of relapse and rebound are being increasingly recognized, in addition to organic withdrawal symptoms. In psychiatric patients extrapyramidal disorders due to neuroleptics are common. The rates of these ADRs differ markedly between various drugs, even after dosages and co-medications are taken into account. Epidemiological screening for potentially carcinogenic drugs can only be done in large cohorts of patients with pre-recorded full information sets as may be found in an HMO (Health Maintenance Organization). The findings of several such studies have been published in specialist cancer journals. However, most of the associations observed should only be viewed as hypotheses for further investigation. * All tables and figuresin this surveyhave been taken from the published proceedings [1].

In a classical twenty-year follow-up study of a cohort of phenacetin abusers, there was an excess of patients with renal insufficiency and death related to renal and urogenital causes.

Key words: risk factor, adverse drug reactions; epidemiological approach, aspirin, benzodiacepines

Despite extensive research in recent years, there is still a lack of information about the epidemiology of adverse drug reactions (ADRs). Patient risk profiles, for instance, are rarely investigated as comprehensively as one would wish. The participants in a satellite symposium of the World Congress of Clinical Pharmacology, in Mannheim/Heidelberg last year, described and discussed many important factors regarded as contributing to the development of adverse drug reactions, such as age, second generation effects, the influence of underlying pathological mechanisms, duration of treatment, and compliance. Some aspects of adverse responses to frequently used drugs, e. g. non-steroidal anti-inflammatory drugs, benzodiazepines, and neuroleptics, were also reviewed in depth. This editorial only summarizes some of the symposium's highlights, and the interested reader should consult the Proceedings for more details of this approach to adverse drug reactions [1]. Premarketing studies provide some information about common pharmacological reactions, but because of the small number of patients studied, particularly in the early and late stages of life, only spare information is available on the adverse reaction profile of a drug when it is first marketed. Thus, postmarketing studies are required, particularly for medicines which will be used for the longterm treatment of otherwise healthy subjects. Although the elderly are often said to be at greater than average risk of developing adverse drug reactions, there are few epidemiological data to substantiate this view. J. H. Gurwitz and J. Avorn emphasized how the interpretation of the results of such studies is often limited by inconsistent definition of the outcomes of interest and

R. Hoigne et el.: Risk factors for adverse drug reactions - epidemiological approaches



0-60 minutes 1-24 hours = acute reactions _J_.= s u b a c u t e reactions

1 day to several weeks = reactions of latent type

Acute severe d y s p n o e a ~

al including anaphylactoid shock and anaphylactoid reactions bl without thrombocytopenia, without clinical or histological evidence of vasculitis d including clinically typical cases without histology Fig. 1. Reaction time in relation to the nature of allergic adverse reactions to drugs: data from the Comprehensive Hospital Drug Monitoring Program Bern/St. Gallen [1]. Reaction time is defined as the period between the last exposure to an allergen and the onset of the allergic reaction

failure to control for important age-related covariates, including the clinical status of the patient, and the number of drugs being taken. To evaluate these problems accurately, postmarketing surveillance studies, using databases containing clinical data from large numbers of older patients, provide an optimal and cost-effective approach. The changes in pharmacokinetics and pharmacodynamics that occur with advanced age provide a theoretical basis for explaining the increased susceptibility to ADRs seen in some studies of the elderly. Of the four traditional components of pharmacokinetics - absorption, distribution, metabolism, and excretion - all except absorption appear to be substantially dependent on age [2]. Important pathways of drug metabolism in the liver may be impaired in advanced age [3]. Drug excretion by the kidney may be considerably prolonged in elderly patients, due to age-related impairment of renal function [4]. Age-related changes in drug disposition have been shown for benzodiazepines and opiates, as have changes in their pharmacodynamics. By contrast, in a recent prospective study, the risk of major bleeding was not found to be affected by patient age and the duration of oral anticoagulant therapy [5]. Individual physiological factors appear to be far more important than chronological age in determining whether an individual patient will tolerate a prescribed drug. However, research efforts in this field so far have not been extensive and much further detailed study is required. An account of current views on the relationship between gastrointestinal bleeding and non-steroidal anti-inflammatory drugs (NSAIDs) was presented by B. L. Strom and colleagues from Pennsylvania. In part because of the frequent use of these drugs, and in part because of the potential seriousness of these ADRs, the association between aspirin, NSAIDs, and gastrointestinal bleeding has been the subject of much postmarketing

pharmaco-epidemiological research. Strom et el. reviewed studies which had examined the relationship between NSAIDs and clinically apparent gastrointestinal bleeding and peptic ulcer disease. A case-control study by Collier et el. [6] in 269 patients with perforated peptic ulcer is a good example of such a study. A twofold increase in risk, which was unlikely to have arisen by chance, was found in patients over 65 y, whereas the risk was not increased in those under 65 y. By contrast, the four available cohort studies showed a smaller increase in risk (Table 1). In a retrospective cohort study of MEDICAID patients, Carson et al. found a 50% increase in the risk of upper gastrointestinal bleeding in NSAID users, with a linear dose-response and a quadratic duration-response relationship [7]. Those studies were based on pooled NSAID recipients. Assessing the relative potency of different NSAIDs in producing gastrointestinal ADRs represents an especially difficult problem. As the cohort of NSAIDexposed patients is divided into subjects exposed to each of the different NSAIDs, the available sample size is dramatically reduced and the confidence intervals become correspondingly enlarged. There is also a problem with selection bias as physicians may tend to choose one drug over others for patients with a high risk of GI toxicity. Using spontaneous reporting systems, the United Kingdom Committee on Safety of Medicines explored the serious gastrointestinal ADRs attributed to NSAIDs [8]. Overall, the NSAIDs could not be distinguished with any confidence, with the exception of ibuprofen, which appeared to be safer than the other agents in the doses commonly prescribed. Given the risk of misclassification bias inherent in the diagnoses in automated databases, particularly in the site of gastrointestinal bleeding, it is important that such studies be restricted to patients whose medical records can be obtained and reviewed. The time patterns seen in 32317admissions of 23935 patients suffering from generalized allergic drug reactions were reviewed by R.Hoign6 and colleagues, using information from the Comprehensive Hospital Drug Monitoring Program in Berne/St.Gallen. The pathological mechanisms in the patients were mainly assigned on the basis of clinical evidence available at the time of monitoring. The period between the last exposure and the onset of the reaction is especially important for elTable 1. Risks of major upper gastrointestinal events in non-steroidal anti-inflammatory drug users: results of large cohort studies [1] Authors


Study outcome

Odds ratio or relative risk (95% confidence intervals)

Carson et al. 1987 a Upper GI bleeding


(1.2- 2.0)

Beard et el.


Upper GI bleeding a


(0.9- 2.1)

Guess et el.


Fatal upper GI bleeding 3.1 or perforation 5.3

(0.1-23.2) b (1.6-17.4) c

Jick et al.


Peptic ulcer perforation

(0.7- 3.7)


a study restricted to the elderly u female patients < 75 y with a negative past history of GI disease c female patients > 75 y with a negative past history of GI disease

R. Hoignd et al.: Risk factors for adverse drug reactions - epidemiotogicalapproaches 2. Classificationof allergic drug reactions as seen in the Comprehensive Hospital Drug Monitoring Programme Bern/St. Gallen: 310 reactions [1] Table

1. Anaphylactic shock (2), anaphylactoid shock (4)L acute severe dyspnoea (7), and anaphylactoid reactions (2)a 2. Urticaria, angioedema (49), bronchial asthma attack (11), and drug fever (36) 3. Maculopapular rash and unspecified rash (154) b, special exanthemata (5)b, conjunctivitis(1), and vascular purpura (12) 4. Serum sickness and serum sickness syndrome (4) and allergic vasculitis,including typical cases without histological examination (12) 5. Thrombocytopenia (1), agranulocytosisand neutropenia (5), and PIE syndrome (pulmonary infiltration with eosinophilia) (2) 6. Liver (3) and kidney (0) reactions a "Anaphylactoid" reactions include "nonallergic" circulatory reactions. The expression "shock" is only used if the blood pressure was reduced b This group comprises only 159 (years 1985-1987) of the total of 889 reactions (years 1974-1987) Table 3. Extrapyramidal symptoms in recipients of neuroleptic drugs: % affected [1] No. of All All All patients grades s.d.i, a grade II/III All neuroleptics 754 34.7 32.6 29.3 High potency 476 51.1 46.8 43.5 Medium potency 551 11.8 7.1 8.3 Low potency 94 0 0 0 Haloperidol 395 55.9 51.6 48.6 Fluspirilene 71 16.9 7.0 15.5 Flupenthixol-D 32 12.5 9.4 12.5 Perazine 340 14.4 9.1 9.7 Levomepromazine 255 5.1 0.8 4.3 Thioridazine 102 6.9 5.9 4.9 Clozapine 54 0 0 0 a single drug imputations

All gradeIII 8.1 12.0 1.9 0 12.7 4.2 3.1 2.1 0.4 1.0 0

lergic reactions and is defined as the 'reaction time'. Three types of reactions can be distinguished: acute (reaction time RT 0-60 min), subacute (RT 1-24 h) and latent reactions (RT 1 day to several weeks) [9, 10], The study covered true allergic reactions as well as those based on intolerance of or idiosyncrasy to aspirin, pyrazolones, paracetamol, NSAIDs, quinidine, iodine-containing contrast media, and certain as yet ill-understood reactions to local anaesthetics. The frequencies of different groups of "allergic" reactions are shown in Table 2. The different syndromes and time patterns of their manifestations (reaction time) are presented in Fig. 1. Of the first group of A R D s (Table 2), most reactions were of the acute type, and this is also true for two-thirds of the attacks of bronchial asthma. In general manifestations of drug allergy only begin after an exposure period of at least several days, 5-7 days or more being the minimum time necessary for senitization and the appearance of symptoms. The first exposure in relation to aprevious period of drug treatment is especially important in the elicitation of acute reactions. The


rate at which exanthemata occur depends on the exposure time, as documented for long-term sulphonamide use by Bergoend et al. [11]. In aminopenicillin-associated exanthemata, the rate per day increases with the duration of exposure up to 12 days, and then decreases markedly during continued treatment. The subject of drug development optimization for the benzodiazepines was presented by M.H.Lader, of The Bethlem and Maudsley Hospitals, London. Sedative effects of those drugs are common, but tend to lessen after a few days, although cognitive effects may persist. On discontinuation of treatment, a variety of syndromes is encountered, including relapse, rebound and withdrawal. The latter follows a characteristic course and symptom pattern, perceptual hypersensitivity being common and distressing. The syndrome is worse after stopping shortacting than long-acting benzodiazepines. Even after short-term use, rebound, i.e. an increase in insomnia beyond the pre-treatment level, may be a problem. Dizziness, fainting, blackouts, and falls are more common in the elderly taking tranquillisers than in drug-free people [12]. Several studies have suggested a significantly increased risk of having a road accident in benzodiazepine users compared to controls, interaction with alcohol being particularly hazardous. In general, high-potency compounds are associated with more clinical reports of sedation and psychomotor, cognitive and memory impairment, and paradoxical effects than are lower-potency compounds. In addition, dependence problems are regarded as worse with drugs like lorazepam and alprazolam. In contrast, oxazepam, a lowpotency, shorter-acting compound, is not usually associated with a severe withdrawal syndrome. The effects of different benzodiazepines have been compared in two studies [13, 14]. Buspirone is a new drug, which is believed to act primarily on 5-HT pathways. It produces little, if any, sedation, and discontinuation so far does not seem to be associated with rebound or withdrawal phenomena. However, the onset of action of this non-benzodiazepine tranquillizer is rather slow, and nausea and headache can be troublesome adverse effects. Extrapyramidal disorders in patients taking neuroleptics in two psychiatric clinics (The Free University of Berlin and The University of Munich) were discussed by R. Grohmann et al. Of the 1107 patients admitted in 19791986, 754 received neuroleptics (68%). Extrapyramidal motor symptoms (EPMS) were observed in 35% of the patients on neuroleptics. Parkinsonism was the most frequent single syndrome, followed by acute dystonia and akathisia. Tardive dyskinesia, a very rare event, developed only after months or years of neuroleptic therapy. There were differences in extrapyramidal motor signs between the two most frequently used neuroleptics, haloperidol (a high-potency butyrophenone) and perazine (a medium-potency phenothiazine). They were seen in 56% of patients on haloperidol and were attributed to haloperidol alone in 92% of those cases. With perazine, the same effects occurred in 14% of cases. In 37% of those with EPMS attributed to perazine it had been given in combination with other neuroleptics. A D R rates for

R. Hoign6 et al.: Risk factors for adverse drug reactions - epidemiological approaches


Table 4. Screening for drug solutions cancer associations in a linked

record system: crude association between barbiturates and lung cancer [1] Drug

Number of users

Number of users Observed Expected


Pentobarbitone Phenobarbitone Secobarbitone Any of the above

2156 5834 2884 9816

34 44 27 87

2.8*** 1.5" 1.8"* 1.7"**

12.3 28.9 15.0 50.2

SMR: Standardized morbidity ratio * p

Risk factors for adverse drug reactions--epidemiological approaches.

Age by itself is not an important risk factor for ADRs. Age-related changes are the consequence of a number of individual factors, for example morbidi...
562KB Sizes 0 Downloads 0 Views