The Pharmaceutical Journal
Vol 270 No 7252 p797-798
7 June 2003

This article
Reprint
Photocopy

PDF* 100K

Royal College of Physicians

Better use of databases should be made in adverse drug reaction (ADR) detecting, according to Professor STEPHEN EVANS, professor of pharmacoepidemiology, London School of Hygiene and Tropical Medicine.

Is a new ADR classification system needed?

Current practice involves individual case assessment by manufacturers prior to marketing and spontaneous reports post-marketing. Such case assessment is always problematic, he said. "Individual case assessments in spontaneous reporting are a total waste of resource."

Instead he suggested that analysis of the database of spontaneous reports should be carried out to look for "signals" — suspicions of adverse reactions. Individual case assessments should only be carried out once a signal was identified. Such signals could be identified from putting all adverse reactions on a giant database. If 2,000 drugs are listed in the rows and 4,000 reactions identified in the columns then this gives 8,000,000 possible cells. "The question is where in the matrix are the hot spots," he said. However, he cautioned that there is potential for misinterpretation of signals so careful evaluation is needed using medical scientific knowledge.

"Of all reactions, 61 per cent only occur once, 15 per cent twice and about 7 per cent three times." He stressed: "Less than 1 per cent occur more than 50 times. If you do a vast amount of work on individual case assessments then it is a waste of your time."

Even large numbers of reports may not indicate causality, he said. Although it is thought that there is a problem with withdrawal from some selective serotonin re-uptake inhibitors, he warned that "basing decisions on rates of spontaneous reporting is not the way to do science". He added: "We need to establish the real reactions from the background noise." The observed number had to be compared with the number expected.

"A target of 100 per cent reporting is a total waste of time," he said. There should be more of a focus in finding new adverse reactions: in fact, he said, the Medicines and Healthcare products Regulatory Agency should give a prize to the person who is the first to find a new ADR. Lots of reports of the same reaction are not helpful, he added.

At present, adverse events that occur at a rate of more than 1 per cent in pre-
marketing trials are listed in the summary of product characteristics (SPC). "This is a great mistake," he said. Instead, he stated, the SPC should include comparisons of the rate of adverse events with the background rate in the placebo group. Use of a database of control groups should be made to calculate the number of expected adverse events. This could then be used to make a comparison to see if events picked up for the drug were statistically significant. Pharmacological knowledge should also be used to predict adverse reactions.

Asked about patients reporting adverse reactions, Professor Evans said "politically, we are going to have to do it". However, he thought that an automatic web-based system should be introduced which would allow patients to classify the reaction through a series of questions. This would reduce the human resource needed in processing what could be a large number of reports. He added that patient-reports could be used to identify "signals".

Responding to a question about a recent study comparing rates of adverse reactions with the MMR vaccine and with another vaccine (PJ, 24 May, p709), Professor Evans said it was "the most scientific nonsense I have even seen in a published paper". He explained: "You can't use this type of data to say that there is an absence of a problem or to demonstrate the presence of a problem."

Classification of ADRs

A new system of classifying adverse reactions was proposed by Dr JEFF ARONSON, university department of clinical pharmacology, Radcliffe Infirmary, Oxford. He called the system "DoTS" because it involved classification according to:

• Dose-responsiveness

• Timing

• Susceptibility

In terms of dose-responsiveness, reactions could be divided into those that occurred at doses higher than those used in therapy, those that occurred in the usual therapeutic range and those that occurred below the usual therapeutic range.

Under the current system, adverse reactions are classified into "type A" or "type B" reactions. This system is a form of classification by dose, explained Dr Aronson. Type A reactions tend to be dose-dependent and are the result of exaggerated but otherwise normal pharmacological actions when the dose is given at therapeutic doses. Type B are aberrant effects that are not expected from the known pharmacological actions of the drug. They tend to be non-dose-dependent.

"However, we have to think about time as well," said Dr Aronson. Reactions could be time-independent, occurring at any time during therapy. Bleeding due to warfarin and digoxin toxicity are two examples of this.

Time-dependent reactions could be split into immediate reactions, first-dose reactions, early reactions, intermediate reactions, late reactions and delayed reactions. Immediate reactions happen due to rapid administrations such as red-man syndrome occurring with vancomycin. This is different from first-dose reactions such as hypotension that occurs after the first dose of an ACE-inhibitor. Early reactions occur after the first few doses but the chance of them occurring reduces with time. Intermediate reactions are similar with higher risk of reaction for the first few weeks or months but then the risk reduces. The risk of late reactions increases as time goes by. This might be the result of the drug accumulating in the body such as in retinopathy with chloroquine. Delayed reactions occur after exposure, sometimes even after the drug has been withdrawn.

Dr Aronson also said that variability in susceptibility to reactions occurs. Sources of variability include genetics, age, sex, altered physiology, exogenous factors (such as interactions with drugs or food) and disease.

The DoTS systems together would provide information for drug development and regulation, for pharmacovigilance, for monitoring individual patients for ADRs and for the prevention, diagnosis and treatment of ADRs, he concluded.

Managing errors

Errors cannot be eliminated but they can be managed, said Professor JAMES REASON, University of Manchester. Effective management depends on understanding the varieties and origins of error. He explained that there are three levels of performance, each of which can result in error:

Skill-based Skill-based errors are routine tasks carried out habitually. Errors are slips or lapses when the person knows what they are doing but the actions do not go as planned, for example, a prescription written as 5mg rather than 0.5mg.

Rule-based Rule-based errors are tasks that a person is trained for and any problems that occur can be dealt with using a rule. Errors occur when the person fails to notice something and applies an inappropriate rule, it might be a "good" rule but in the wrong situation or a "bad" rule, for example, applying the wrong formula to adjust a dose of medicine.

Knowledge-based Knowledge-based errors require the level of highest demand where the person has to apply developed knowledge to novel situations, for example, failing to make a dose adjustment by not appreciating the need to do this.

Some errors keep recurring, Professor Reason said. "The same situations keep producing the same kinds of errors even though different people are involved. Therefore a reporting database is needed to identify recurrences of error."

A strong emphasis on safe use of drugs is part of the National Health Service's new approach to patient safety, according to Professor KENT WOODS, University of Leicester. "Things are going wrong with drug prescribing on an unacceptable scale," he said. Newly qualified doctors are the most error-prone so those doctors who do the most prescribing are those who make the most mistakes. As doctors gain more experience, their error rate drops indicating a learning curve. "The learning must be done in the undergraduate curriculum; doctors need to be better trained and skilled in prescribing before registration," he said.

Although drugs are getting safer, the prescribing environment in hospitals is getting more hazardous, said Professor Woods. This is a result of rising work intensity, loss of supervision and continuity, staff rotas in specialised areas and increasing therapeutic complexity.

Dr Wood has undertaken some work for the Department of Health looking at reducing medication errors. He analysed reports of intrathecal medication errors and found that certain things happened "again and again". These included a lack of standard procedures, variable organisation of care and hazards with the connection device. As a result he made a number of recommendations to the chief medical officer. Most of these — such as national guidance on intrathecal chemotherapy, identifying designated trained individuals and work on the dispensing form of vinca drugs — has happened or is ongoing. However, he believed that his recommendation that medical school curricula should cover drug safety had not been taken up yet. "Undergraduate education and training have a key role in preventing errors but innovative teaching methods are needed," he concluded.

The Home Office's current view on charging doctors with manslaughter following a medical error is rather different from the Department of Health view, according to Dr ROBIN FERNER, City Hospital, Birmingham. The DoH has stated that there should be a more open culture in which errors can be reported and discussed. However, since the 1980s, there has been an increasing trend in the number of doctors charged with manslaughter.

Pharmacogenetics: the future?

For pharmacogenetics to be effective it needs to be in general practice, said Professor MUNIR PIRMOHAMED, University of Liverpool. However, point-of-care testing is not a reality at the moment. "There's a lot of hype about point-of-care testing being present in five years time. I think it is unlikely to be present in five years; it is more likely to be 50 years."

Pharmacogenetics can currently help in three areas related to adverse drug reactions, said Professor Pirmohamed. These are in assessing dose requirements, as part of the drug development process and in idiosyncratic ADRs.

An example of the use of pharmacogenetics in dose requirement was warfarin. The population has a 20-fold variation in warfarin dose requirement, he explained. Warfarin is a racemate drug. The more potent racemate, S-warfarin, is metabolised by CYP2C9. Variations in the CYP2C9 genotype can lead to significant differences in the dose requirement of warfarin. However, it is not the only factor that determines warfarin dose requirement and this variability means that it cannot be used alone to determine dose.

Pharmacogenetics also has a role in drug development. A gene that is important in this area is the one that codes for cytochrome P450 2D6 which is involved in drug metabolism. Variations in this gene can result in reduced or no metabolism via cytochrome P450 2D6, increased metabolism or the production of different metabolites. When drugs are developed they are screened to see if they are metabolised by cytochrome P450 2D6 through the use of genotyped panels of volunteers. If the drug is found to be metabolised through this route then the magnitude of response and frequency of effect will be used to determine whether or not drug development should continue. "Sometimes development is not continued and other times appropriate warnings are placed in the drug's SPC," Professor Primohamed said.

Genetic variations could also be used to explain idiosyncratic ADRs. Carbamazepine hypersensitivity reactions occur 3.2 times more often in patients with a particular genetic variation in than in patients without this variation. In addition, the majority of cases of hypersensitivity to abacavir could be predicted by genetic variation. Unpublished data that compared the cost of treating the reaction with the cost of genetic testing and alternative therapy suggest that pre-prescribing genotyping is cost-effective, he said.