Anecdotes as evidence

Table 2

Table 3

PHARMA: Publishing Histories of Adverse Reactions to Medicaments Anecdotally
 

Table 1 Types of articles on suspected adverse drug reactions and interactions published in 2000 and cited, with tags, in the Side Effects of Drugs Annual 24^{w1 w2}
 

w2 Aronson JK, Derry S, Loke YK. Adverse drug reactions: keeping up to date. Fundam Clin Pharmacol 2002;16:49-56.

Table 2 Reasons for publishing anecdotal reports of adverse drug reactions or interactions, with examples
 

w1 Inman WHW. Don’t tell the patient. Behind the drug safety net. Bishops Waltham: Highland Park Productions, 1999.

w2 Orme ML’E. The Debendox saga. BMJ (Clin Res Ed) 1985;291:918-19.

w3 Mazzotta P, Magee LA. A risk-benefit assessment of pharmacological and nonpharmacological treatments for nausea and vomiting of pregnancy. Drugs 2000;59:781-800.

w4 Aronson JK, Grahame-Smith DG. Altered distribution of digoxin in renal failure—a cause of digoxin toxicity? Br J Clin Pharmacol 1976;3:1045-51.

w5 Wagner JG. Loading and maintenance doses of digoxin in patients with normal renal function and those with severely impaired renal function. J Clin Pharmacol 1974;14:329-38.

w6 Endoh Y, Hanai R, Uto K, Uno M, Nagashima H, Takizawa T, et al. Diagnostic usefulness of KL-6 measurements in patients with pulmonary complications after administration of amiodarone. J Cardiol 2000;35:121-7.

w7 Esato M, Sakurada H, Okazaki H, Kimura T, Nomizo A, Endou M, et al. Evaluation of pulmonary toxicity by CT, pulmonary function tests, and KL-6 measurements in amiodarone-treated patients. Ther Res 2001;22:867-73.

w8 Dessertenne F. La tachycardie ventriculaire à deux foyers opposés variables. Arch Mal Coeur Vaiss 1966;59:263-72.

w9 Buckley CD, Aronson JK. Prolonged half-life of verapamil in a case of overdose: implications for therapy. Br J Clin Pharmacol 1995;39:680-3.

w10 Gokel Y, Paydas S, Duru M. High-dose verapamil–trandolapril induced rhabdomyolysis and acute renal failure. Am J Emerg Med 2000;18:738-9.

w11 De Klerk GJ, Nieuwenhuis MG, Beutler JJ. Hypokalaemia and hypertension associated with use of liquorice flavoured chewing gum. BMJ 1997;314:731.
 

Table 3Desirable contents of a case report of an adverse drug reaction
 

*These 14 details are solicited on the yellow card for reporting suspected adverse drug reactions to the Committee on Safety of Medicines in the United Kingdom.

†These 14 details are solicited on the MedWatch adverse event forms in the United States.
 
 

PHARMA: Publishing Histories of Adverse Reactions to Medicaments Anecdotally

Reporting suspected adverse drug reactions and interactions: proposed guidelines for authors

Many reports of suspected adverse drug reactions and interactions are anecdotal. For example, of the 3252 citations in the 24th volume of the Side Effects of Drugs Annual,^w1 in which the world literature on adverse drug reactions and interactions that was published in 2000 was critically reviewed, there were almost as many anecdotal reports (1075 citations) as there were major randomised controlled trials or observational studies (1266 citations); in contrast, only 45 citations reported a systematic review (see web table 1).^w2

Reasons for publishing anecdotal reports of adverse drug reactions or interactions

There are several reasons for publishing anecdotal reports of adverse drug reactions and interactions (see web table 2):

• To describe a new adverse reaction or interaction Adverse effects are often discovered through anecdotal reports. After psoriasiform rashes were reported in patients taking practolol^w3 the Committee on Safety of Medicines received many reports of similar adverse events. The oculomucocutaneous syndrome caused practolol to be withdrawn.^w4
• To generate hypotheses A single case does not prove an association, but it does generate a testable hypothesis. The combination of doxylamine, dicycloverine, and pyridoxine, originally marketed in the UK as Debendox and in the USA as Bendectin, was implicated as a teratogen on the basis of a single case that provoked litigation in the USA.^w5 The combination was withdrawn prematurely in 1984, before the hypothesis was tested. Later, many trials and a meta-analysis showed that it is not teratogenic.^w6
• To test hypotheses A case can provide the single counterexample that vitiates or disproves a hypothesis.^w7 Although in severe renal insufficiency a low apparent volume of distribution of digoxin implies that the loading dose should be reduced, a theoretical analysis suggested otherwise.^w8 Subsequently, a set of three cases proved that the theoretical argument was wrong and that in severe renal insufficiency the loading dose of digoxin should indeed be reduced, to avoid toxicity.^w9
• To demonstrate diagnostic techniques Measurement of serum KL-6, a sialylated carbohydrate antigen, may help in the diagnosis of amiodarone-induced lung damage, according to two reports, one of seven cases^w10 and one of three.^w11 This hypothesis needs to be formally tested.
• To elucidate mechanisms The association between QT interval prolongation and drug-induced torsade de pointes was first observed in a single case.^w12
• To study methods of managing adverse drug reactions Elucidation of the pharmacokinetics of verapamil and its major metabolite in a case of verapamil self-poisoning suggested that repeated doses of activated charcoal would be effective in such cases^w13; this has since been successfully done.^w14
• To remind or educate Although an adverse drug reaction or interaction may be well known, a reminder is sometimes useful. That liquorice can cause hypertension with hypokalaemia has been known for many years, but in 1997 these complications in two patients who had used chewing gum flavoured with liquorice gave the BMJ the opportunity of publishing a reminder and listing the large number of products containing liquorice (including confectionery, health products, beers, pastis, and various types of anisette) that are widely available to consumers.^w15
• Systematic review of anecdotes Reviewing anecdotes may yield new findings. Inman discovered the association between mestranol and venous thrombosis by reviewing 88 individual reports of pulmonary embolism that had been submitted on yellow cards to the then Committee on Safety of Drugs.^w4
• A 49-year-old man with lip swelling. Angioedema associated with angiotensin II receptor blockade.^w19
• Pulmonary infarction associated with crack cocaine use in a previously healthy 23-year-old woman.^w20
• Vinorelbine-associated myelopathy in a patient who previously received paclitaxel.^w21
• Delusion of worm infestation associated with clarithromycin in a patient on peritoneal dialysis.^w22
• Adverse event.
• Drug implicated, the dosage used, the duration of therapy, and other drugs used.
• The patient (important details, including age, sex, and risk factors).
• Evidence that links the drug to the event.
• Management, if relevant.
• Mechanism, if known.
• Implications for use of the drug.
• Hypotheses to be tested as a result of the observation.
• the time-course in relation to the administration of the suspected drug;
• the effect of withdrawing the drug, including the time-course;
• the effect of rechallenge; if rechallenge was not performed, state the reasons;
• diagnostic tests, both those relevant to the diagnosis of the event and those relevant to the link with the drug; report evidence either as numbers with reference ranges (for example, liver function tests, blood counts) or as illustrations (for example, electrocardiograms, x rays, histological sections); also include the results of tests that, if positive, would have suggested an alternative diagnosis.
• the final outcome.
• measurements of plasma concentrations, including the parent compound and its main metabolites; in the case of a drug interaction both of the drugs involved and their relevant metabolites should be studied; when giving usual target ranges for plasma drug concentrations, use both mass and molar units, with the litre as the reference volume (for example, mg/l and μ mol/l)^w23;
• data from animal or in vitro studies that support the effect or the proposed mechanism.
• a conclusion on the likelihood that the reported event was an adverse drug reaction, based on Bayesian methods^w26 or, if there is insufficient information for that approach, an established algorithm,^{w26 w27} although the limitations of such algorithms must be recognised.^w28
• why the implicated drug was to blame and why other drugs that the patient took were not;
• how other possible causes were eliminated;
• a review of previous cases; this should include unpublished cases reported to the manufacturer^w29 or to the Committee on Safety of Medicines or other regulatory authorities; if unpublished cases of this sort have not been reported, that should be stated;
• when relevant, a discussion of methods of diagnosis;
• a discussion of possible mechanisms, including evidence, when available, from case reports, formal clinical studies, and animal and in vitro studies;
• a discussion of possible forms of management, based on similar evidence;
• a discussion of the implications of the report for clinical practice;
• a discussion of any hypotheses that need to be tested as a result of the report.

Proposed guidelines

I therefore propose that there should be a set of guidelines, comparable to the CONSORT guidelines for randomised controlled trials,^w16 for preparing such reports, which have specific features that need to be elucidated and described. The fact that an adverse event has been published and attributed to a drug or drugs takes it into the realm of a suspected adverse reaction rather than just an adverse event.

The ideal attributes of case reports in general have been well described.^w17 In this appendix I describe my own set of proposed guidelines for reporting suspected adverse drug reactions and interactions (see web table 3). I hope that this proposal will stimulate discussion and debate, resulting in a comparable consensus statement to that of CONSORT.

These guidelines are aimed at encouraging clinicians to report cases in a uniform fashion that will allow direct comparison of individual reports, and to obtain information that will in part help to assess whether an adverse event was actually an adverse drug reaction, and in part help to establish diagnostic techniques, mechanisms, and guidelines for management. Uniform presentation of relevant information will also facilitate systematic review of suspected adverse drug reactions. The guidelines do not deal with n-of-one studies, which should be dealt with separately.

The title

The importance of the title should not be underestimated. In one study, 25% of all papers in a known corpus dealing with adverse drug reactions could not be retrieved from Medline and EMBASE because their titles, abstracts, and indexing terms did not contain sufficient information relevant to adverse drug reactions.^w18 An accurate title affords a method for retrieving relevant publications.

The title should be a non-declaratory formulation that mentions the adverse event, the drug that was suspected to have caused it, and any important risk factors. The term "associated with" should be used in preference to "caused by" or "due to". Here are some examples of clear explicit titles:

• A 49-year-old man with lip swelling. Angioedema associated with angiotensin II receptor blockade.^w19
• Pulmonary infarction associated with crack cocaine use in a previously healthy 23-year-old woman.^w20
• Vinorelbine-associated myelopathy in a patient who previously received paclitaxel.^w21
• Delusion of worm infestation associated with clarithromycin in a patient on peritoneal dialysis.^w22
• Adverse event.
• Drug implicated, the dosage used, the duration of therapy, and other drugs used.
• The patient (important details, including age, sex, and risk factors).
• Evidence that links the drug to the event.
• Management, if relevant.
• Mechanism, if known.
• Implications for use of the drug.
• Hypotheses to be tested as a result of the observation.
• the time-course in relation to the administration of the suspected drug;
• the effect of withdrawing the drug, including the time-course;
• the effect of rechallenge; if rechallenge was not performed, state the reasons;
• diagnostic tests, both those relevant to the diagnosis of the event and those relevant to the link with the drug; report evidence either as numbers with reference ranges (for example, liver function tests, blood counts) or as illustrations (for example, electrocardiograms, x rays, histological sections); also include the results of tests that, if positive, would have suggested an alternative diagnosis.
• the final outcome.
• measurements of plasma concentrations, including the parent compound and its main metabolites; in the case of a drug interaction both of the drugs involved and their relevant metabolites should be studied; when giving usual target ranges for plasma drug concentrations, use both mass and molar units, with the litre as the reference volume (for example, mg/l and μ mol/l)^w23;
• data from animal or in vitro studies that support the effect or the proposed mechanism.
• a conclusion on the likelihood that the reported event was an adverse drug reaction, based on Bayesian methods^w26 or, if there is insufficient information for that approach, an established algorithm,^{w26 w27} although the limitations of such algorithms must be recognised.^w28
• why the implicated drug was to blame and why other drugs that the patient took were not;
• how other possible causes were eliminated;
• a review of previous cases; this should include unpublished cases reported to the manufacturer^w29 or to the Committee on Safety of Medicines or other regulatory authorities; if unpublished cases of this sort have not been reported, that should be stated;
• when relevant, a discussion of methods of diagnosis;
• a discussion of possible mechanisms, including evidence, when available, from case reports, formal clinical studies, and animal and in vitro studies;
• a discussion of possible forms of management, based on similar evidence;
• a discussion of the implications of the report for clinical practice;
• a discussion of any hypotheses that need to be tested as a result of the report.

Many journals do not require summaries in case reports, but they should. Summaries are searchable in databases such as Medline and EMBASE and can also provide helpful information for the reader who does not have easy access to the journal. A structured summary should include at least the first four of the following headings:

• Adverse event.
• Drug implicated, the dosage used, the duration of therapy, and other drugs used.
• The patient (important details, including age, sex, and risk factors).
• Evidence that links the drug to the event.
• Management, if relevant.
• Mechanism, if known.
• Implications for use of the drug.
• Hypotheses to be tested as a result of the observation.
• the time-course in relation to the administration of the suspected drug;
• the effect of withdrawing the drug, including the time-course;
• the effect of rechallenge; if rechallenge was not performed, state the reasons;
• diagnostic tests, both those relevant to the diagnosis of the event and those relevant to the link with the drug; report evidence either as numbers with reference ranges (for example, liver function tests, blood counts) or as illustrations (for example, electrocardiograms, x rays, histological sections); also include the results of tests that, if positive, would have suggested an alternative diagnosis.
• the final outcome.
• measurements of plasma concentrations, including the parent compound and its main metabolites; in the case of a drug interaction both of the drugs involved and their relevant metabolites should be studied; when giving usual target ranges for plasma drug concentrations, use both mass and molar units, with the litre as the reference volume (for example, mg/l and μ mol/l)^w23;
• data from animal or in vitro studies that support the effect or the proposed mechanism.
• a conclusion on the likelihood that the reported event was an adverse drug reaction, based on Bayesian methods^w26 or, if there is insufficient information for that approach, an established algorithm,^{w26 w27} although the limitations of such algorithms must be recognised.^w28
• why the implicated drug was to blame and why other drugs that the patient took were not;
• how other possible causes were eliminated;
• a review of previous cases; this should include unpublished cases reported to the manufacturer^w29 or to the Committee on Safety of Medicines or other regulatory authorities; if unpublished cases of this sort have not been reported, that should be stated;
• when relevant, a discussion of methods of diagnosis;
• a discussion of possible mechanisms, including evidence, when available, from case reports, formal clinical studies, and animal and in vitro studies;
• a discussion of possible forms of management, based on similar evidence;
• a discussion of the implications of the report for clinical practice;
• a discussion of any hypotheses that need to be tested as a result of the report.

The introduction should do no more than name the suspected drug and the adverse event with which it was associated, briefly mention previous similar reports, and state the purpose of the report. It need not include a detailed discussion of the pharmacology, use, and known adverse effects of the drug; however, a reference to a good review might be given.

The case report

Demographic information The patient’s age, sex, weight, and ethnic background.

Premorbid condition Details of all the patient’s diagnoses. Allergies (present or absent) should be specifically mentioned.

Drug therapy All current drugs, including dosages, durations of therapy, and indications; mention drugs that have been recently withdrawn if they have either a long half-life (for example, amiodarone) or effects that persist after drug withdrawal (for example, corticosteroids). The list of diagnoses (indications) and the list of drugs that the patient is currently taking can conveniently be combined.

Other relevant history Include relevant family history and social history (for example, drinking and smoking habits); include a negative history if relevant (for example, an absence of excess alcohol intake in a case of liver damage).

The adverse event A detailed clinical description of the event, including the following information:

• the time-course in relation to the administration of the suspected drug;
• the effect of withdrawing the drug, including the time-course;
• the effect of rechallenge; if rechallenge was not performed, state the reasons;
• diagnostic tests, both those relevant to the diagnosis of the event and those relevant to the link with the drug; report evidence either as numbers with reference ranges (for example, liver function tests, blood counts) or as illustrations (for example, electrocardiograms, x rays, histological sections); also include the results of tests that, if positive, would have suggested an alternative diagnosis.
• the final outcome.
• measurements of plasma concentrations, including the parent compound and its main metabolites; in the case of a drug interaction both of the drugs involved and their relevant metabolites should be studied; when giving usual target ranges for plasma drug concentrations, use both mass and molar units, with the litre as the reference volume (for example, mg/l and μ mol/l)^w23;
• data from animal or in vitro studies that support the effect or the proposed mechanism.
• a conclusion on the likelihood that the reported event was an adverse drug reaction, based on Bayesian methods^w26 or, if there is insufficient information for that approach, an established algorithm,^{w26 w27} although the limitations of such algorithms must be recognised.^w28
• why the implicated drug was to blame and why other drugs that the patient took were not;
• how other possible causes were eliminated;
• a review of previous cases; this should include unpublished cases reported to the manufacturer^w29 or to the Committee on Safety of Medicines or other regulatory authorities; if unpublished cases of this sort have not been reported, that should be stated;
• when relevant, a discussion of methods of diagnosis;
• a discussion of possible mechanisms, including evidence, when available, from case reports, formal clinical studies, and animal and in vitro studies;
• a discussion of possible forms of management, based on similar evidence;
• a discussion of the implications of the report for clinical practice;
• a discussion of any hypotheses that need to be tested as a result of the report.

Other desirable (but not compulsory) information includes:

• measurements of plasma concentrations, including the parent compound and its main metabolites; in the case of a drug interaction both of the drugs involved and their relevant metabolites should be studied; when giving usual target ranges for plasma drug concentrations, use both mass and molar units, with the litre as the reference volume (for example, mg/l and μ mol/l)^w23;
• data from animal or in vitro studies that support the effect or the proposed mechanism.
• a conclusion on the likelihood that the reported event was an adverse drug reaction, based on Bayesian methods^w26 or, if there is insufficient information for that approach, an established algorithm,^{w26 w27} although the limitations of such algorithms must be recognised.^w28
• why the implicated drug was to blame and why other drugs that the patient took were not;
• how other possible causes were eliminated;
• a review of previous cases; this should include unpublished cases reported to the manufacturer^w29 or to the Committee on Safety of Medicines or other regulatory authorities; if unpublished cases of this sort have not been reported, that should be stated;
• when relevant, a discussion of methods of diagnosis;
• a discussion of possible mechanisms, including evidence, when available, from case reports, formal clinical studies, and animal and in vitro studies;
• a discussion of possible forms of management, based on similar evidence;
• a discussion of the implications of the report for clinical practice;
• a discussion of any hypotheses that need to be tested as a result of the report.

The discussion

The discussion should contain the following:

• a conclusion on the likelihood that the reported event was an adverse drug reaction, based on Bayesian methods^w26 or, if there is insufficient information for that approach, an established algorithm,^{w26 w27} although the limitations of such algorithms must be recognised.^w28
• why the implicated drug was to blame and why other drugs that the patient took were not;
• how other possible causes were eliminated;
• a review of previous cases; this should include unpublished cases reported to the manufacturer^w29 or to the Committee on Safety of Medicines or other regulatory authorities; if unpublished cases of this sort have not been reported, that should be stated;
• when relevant, a discussion of methods of diagnosis;
• a discussion of possible mechanisms, including evidence, when available, from case reports, formal clinical studies, and animal and in vitro studies;
• a discussion of possible forms of management, based on similar evidence;
• a discussion of the implications of the report for clinical practice;
• a discussion of any hypotheses that need to be tested as a result of the report.

Conclusions

Anecdotal reports make up about one-third of the published literature on suspected adverse drug reactions and interactions. Despite the current emphasis on randomised controlled trials and systematic reviews in evidence-based medicine, such anecdotes are worth reporting, in order to describe new reactions, generate and test hypotheses, study diagnostic techniques, mechanisms, and management, and provide information and education. There is such a preponderance of anecdotal reports of adverse drug reactions that there is a case for a new journal devoted to reports of this sort.^w2 Such a journal could improve the quality of reporting and encourage the development of new methods of analysing the mass of reports, which are much needed.

Published case reports should ideally contain enough information to allow direct comparisons of cases in different reports, and there should be a standard method of presentation, comparable to the CONSORT guidelines for presentation of randomised controlled trials. Whenever possible, standardised terminology should be used.^w30

Table 3 gives a skeleton for a desirable structure of such reports. It is not my intention to suggest that all adverse drug reactions reports should adhere strictly to the proposed format, which might not be necessary for, for example, reports intended to remind or educate. Furthermore, national and international reporting schemes, such as the yellow card scheme run by the UK’s Committee on Safety of Medicines and the MedWatch Averse Event Forms, will continue to want to receive reports, however brief or incomplete. In passing, it should be noted that the amounts of information required on the yellow card and the MedWatch forms are slightly less extensive than suggested here (web table 3); the yellow card and MedWatch forms could usefully be expanded to include the additional items as optional extras.

Nevertheless, for many reports, particularly those that deal with new or unexpected events or unusually severe ones, complete reports will be useful. The format proposed here will form the basis for discussion in reaching a consensus on the appropriate format for reporting suspected adverse drug reactions and interactions.

Jeffrey K Aronson

consultant clinical pharmacologist and consultant physician

Department of Clinical Pharmacology, Radcliffe Infirmary, Woodstock Road, Oxford OX2 6HE, UK (jeffrey.aronson@clinpharm.ox.ac.uk)

w1 Aronson JK, ed. Side effects of drugs. Annual 24. Amsterdam: Elsevier, 2001.

w2 Aronson JK, Derry S, Loke YK. Adverse drug reactions: keeping up to date. Fundam Clin Pharmacol 2002; 16: 49-56.

w3 Felix R, Ive FA. Skin reactions to practolol. BMJ 1974;2:333.

w4 Inman WHW. Don’t tell the patient. Behind the drug safety net. Bishops Waltham: Highland Park Productions, 1999.

w5 Orme ML’E. The Debendox saga. BMJ (Clin Res Ed) 1985;291:918-19.

w6 Mazzotta P, Magee LA. A risk-benefit assessment of pharmacological and nonpharmacological treatments for nausea and vomiting of pregnancy. Drugs 2000;59:781-800.

w7 Lakatos I. Proofs and refutations. Br J Phil Sci 1963; XIV: 1-25, 123-39, 221-45, and 296-339.

w8 Wagner JG. Loading and maintenance doses of digoxin in patients with normal renal function and those with severely impaired renal function. J Clin Pharmacol 1974;14:329-38.

w9 Aronson JK, Grahame-Smith DG. Altered distribution of digoxin in renal failure—a cause of digoxin toxicity? Br J Clin Pharmacol 1976;3:1045-51.

w10 Endoh Y, Hanai R, Uto K, Uno M, Nagashima H, Takizawa T, et al. Diagnostic usefulness of KL-6 measurements in patients with pulmonary complications after administration of amiodarone. J Cardiol 2000;35:121-7.

w11 Esato M, Sakurada H, Okazaki H, Kimura T, Nomizo A, Endou M, et al. Evaluation of pulmonary toxicity by CT, pulmonary function tests, and KL-6 measurements in amiodarone-treated patients. Ther Res 2001;22:867-73.

w12 Dessertenne F. La tachycardie ventriculaire à deux foyers opposés variables. Arch Mal Coeur Vaiss 1966;59:263-72.

w13 Buckley CD, Aronson JK. Prolonged half-life of verapamil in a case of overdose: implications for therapy. Br J Clin Pharmacol 1995;39:680-3.

w14 Gokel Y, Paydas S, Duru M. High-dose verapamil–trandolapril induced rhabdomyolysis and acute renal failure. Am J Emerg Med 2000;18:738-9.

w15 De Klerk GJ, Nieuwenhuis MG, Beutler JJ. Hypokalaemia and hypertension associated with use of liquorice flavoured chewing gum. BMJ 1997;314:731.

w16 Altman DG, Schulz KF, Moher D, Egger M, Davidoff F, Elbourne D, et al. The revised CONSORT statement for reporting randomized trials: explanation and elaboration. Ann Intern Med 2001;134:663-94.

w17 Jenicek M. Clinical case reporting in evidence-based medicine. 2nd ed. London: Arnold, 1999.

w18 Derry S, Loke YK, Aronson JK. Incomplete evidence: the inadequacy of databases in tracing published adverse drug reactions in clinical trials. BMCMed Res Methodol 2001;1:7. (www.biomedcentral.com/1471-2288/1/7/ (accessed 1 May 2003).

w19 Itoh H, Mizuno S, Shimizu M, Ohnaka M. A 49-year-old man with lip swelling. Angioedema associated with angiotensin II receptor blockade. J Cardiol 2001; 38: 355-6.

w20 Delaney K, Hoffman RS. Pulmonary infarction associated with crack cocaine use in a previously healthy 23-year-old woman. Am J Med 1991;91:92-4.

w21 Han JY, Choi BG, Song DH, Ahn JG, Yoon JS, Lee KS. Vinorelbine-associated myelopathy in a patient who previously received paclitaxel: a case report. Med Oncol 2001;18 95-7.

w22 Tse KC, Li FK, Tang S, Lam MF, Chan TM, Lai KN. Delusion of worm infestation associated with clarithromycin in a patient on peritoneal dialysis. Perit Dial Int 2001;21:415-16.

w23 Aronson JK. When I use a word. Masses and masses. BMJ 2002;324:1521.

w24 Geaney DP, Carver JG, Davies CL, Aronson JK. Pharmacokinetic investigation of the interaction of azapropazone with phenytoin. Br J Clin Pharmacol 1983;15:727-34.

w25 Holford NH, Sheiner LB. Understanding the dose-effect relationship: clinical application of pharmacokinetic–pharmacodynamic models. Clin Pharmacokinet 1981;6:429-53.

w26 Lanctot KL, Naranjo CA. Comparison of the Bayesian approach and a simple algorithm for assessment of adverse drug events. Clin Pharmacol Ther 1995;58:692-8.

w27 Stephens MDB. The diagnosis of adverse medical events associated with drug treatment. Adv Drug React Ac Pois Rev 1987;1:1-35.

w28 Meyboom RH, Hekster YA, Egberts AC, Gribnau FW, Edwards IR. Causal or casual? The role of causality assessment in pharmacovigilance. Drug Saf 1997;17:374-89.

w29 Smith R. The BMJ, spin, and the stockmarket. BMJ 2000;320:1737.

w30 Edwards IR, Aronson JK. Adverse drug reactions—definitions, classification, diagnosis, management, surveillance. Lancet 2000;356:1255-60.