MHRA-LOGO.gif (6825 bytes) Medicines and Healthcare products
Regulatory Agency

Market Towers
1 Nine Elms Lane, London SW8 5NQ


Mr Charles Medawar
Social Audit Ltd
P.O. Box 111
London NW1 8XG

General enquiries
Telephone 020 7273 0000  Fax 020 7273 0353
E-mail [email protected]

Direct line:  020 7084 2400
Direct fax:  020 7084 2675

Email:   [email protected]


30 September 2004 

Dear Mr Medawar,

Yellow Card Scheme

I write with reference to your paper: A comparison of adverse drug reaction reports from professionals and users, relating to risk of dependence and suicidal behaviour with paroxetine. published in the International Journal of Risk and Safety of Medicine (2003. 16. 3-17), co-authored by Professor Herxheimer.

The MHRA has carefully reviewed the criticisms made in the article in relation to follow up and coding of reports of suspected adverse reactions received through the Yellow Card Scheme, and referred to in a Parliamentary Question answered on 21 July (Col.371W). I have attached to this letter the MHRA’s analysis of the issues raised in and by the article.

During the course of your research, in July 2003 the Independent Review of access to the Yellow Card Scheme was launched. This Review, to which you gave evidence. has produced a number of recommendations to widen access to Yellow Card data and also to strengthen the Scheme, in the interests of public health. As you know, the recommendation for the introduction of direct patient reporting has been accepted and plans are in hand for pilots of different mechanisms to gauge effectiveness. The MHRA and CSM have set up a dedicated Working Group to consider not only the practical aspects of how patient reporting can be introduced, but effective communications with patients and the public.

Decisions on the remaining recommendations will be made in the light of the consultation responses. But it is clear that if best use is to be made of the data, there must be precision and understanding of the data requested and provided. We made it clear, for example, in our letters to you that you were not provided with Anonymised Single Patient Printouts. The Review recommendations also address the need for formal procedures for pre-publication receipt of papers to ensure that there is adequate opportunity for MHRA to review research findings.

The procedures for collection and classification of data collected through the Yellow Card Scheme have been developing since 1964. The limitations of spontaneous reporting data are well recognised and the 500,000 or so reports held on the ADROIT database stretch back over 40 years and vary in quality and completeness. These limitations are taken into account by the staff involved in analysing the data from the scheme. The limitations do not detract from the utility of the data for the purposes of "signal" detection. There are robust procedures in place to support prompt analysis of data for this purpose, and when a potential risk is under review, the totality of the data from each suspected adverse reaction report is available for assessment.

Finally, you assert that flawed analysis of Yellow Cards has led to underestimation of the risk of suicidal behaviour. The "signal" of suicidal behaviour associated with SSRIs was detected in the 1990’s. Spontaneous ADR data such as Yellow Card reports cannot be used to confirm or refute this signal, which is why the MHRA has sought other data sources including epidemiological studies.

I reiterate my offer to meet with you to discuss the issues you have raised in your papers and to clarify any further questions you may have.

Yours sincerely

Dr J M Raine
Director, Post-Licensing Division


The Medicines and Healthcare products Regulatory Agency analysis of the issues raised in and by the paper: "A comparison of adverse drug reaction reports from professionals and users, relating to risk of dependence and suicidal behaviour with paroxetine", published in the International Journal of Risk and Safety of Medicine (2003, 16, 3-17): Mr Charles Medawar and Professor Andrew Herxheimer.

1. Data analysed in the paper
MHRA staff provided, following requests and subsequent correspondence with the authors in the six months between March and August 2003, line listings on an Excel spreadsheet showing information on each single case of drug withdrawal, drug dependence, injury and poisoning and suicidal reactions in association with paroxetine. MHRA staff did not filter the case information provided. Identifiable patient information was removed to protect patient anonymity and the spreadsheets were formatted to improve readability but this did not involve the removal of case information. As discussed with the authors in correspondence ASPPs were not provided. Anonymised Single Patient Printouts (ASPPs) are generated from the ADROIT database once all ADR information from an individual Yellow Card report has been transposed.

The data provided were individual case data by line listings for the different reactions requested. One patient may experience more than one reaction (e.g. drug withdrawal and suicidal ideation). The same case, therefore, can repeat on the different data sets and the ‘duplication’ identified appears to be a result of this feature of the data. There are procedures in place for identifying true duplication of cases in the database and merging duplicate cases into a single master report. The article correctly identifies one duplicate report of suicide and suicidal behaviour.

2. Coding of ADR reports
Reports of suspected adverse drug reactions are coded as precisely as possible. Wherever possible, coding accurately reflects reporters’ observations to minimise the potential for assumptions which may lead to an inconsistent approach and skew the data. All relevant details of the report are classified and an electronic image of original reports is available on the system to enable the totality of the data to be reviewed at any time. Standardised terms are used to code based on the terminology that has become the internationally adopted standard in drug regulation: MedDRA (Medical Dictionary for Regulatory Activities). Using this terminology, similar reactions used to describe essentially the same reaction by different reporters will tend to be grouped under similar terms. In this way, it is possible to analys& similar collections of terms being consistently reported.

Examples of ‘miscoding’ of suicidal reactions associated with paroxetine the database have been reviewed. Further information on some of the specific examples referred to are as follows.

ABPI reports
Yellow Cards are not submitted by Association of the British Pharmaceutical Industry (ABPI). There are several different formats of the Yellow Card form in circulation and the type of the form used to submit the report is recorded on ADROIT. ‘ABPI’ refers to the Yellow Cards provided to health professionals in the Data Sheet Compendium published by the ABPI.

Onset time
Onset of the suspected adverse reaction in the structured onset time field is coded if full dates are provided in the report. If statements about onset are provided (e.g. ‘one day after starting’) the information is recorded in the ‘free text’ area of the database which is included in any subsequent analysis. Information which associates the suspected reaction with increasing or decreasing the dose, where available, is reflected in the text of the report. Where time to onset of reaction does not correspond to the information cited in the text of the report, the database has automatically calculated onset from the dates the drug was started and reaction began. In some cases this automatic calculation may not be appropriate and during transposition this field should be manually removed. This has not occurred in all cases.

Non-accidental overdose
Reports may be coded as non-accidental overdose or suicidal behaviour. In undertaking risk assessment, the MHRA recognise that non-accidental overdose and suicidal behaviour may reflect the same suspected reaction and take into account this relationship during assessment. Of 32 reports classified as overdose and 35 classified as non-accidental overdose, 5 mention suicidal thoughts. In these cases, the suicidal behaviour term has been correctly classified as a suspected adverse reaction.

Electric Shock Sensation
Five reports of the first 78 Yellow Cards received mentioned "electric shock sensation" in the context of withdrawal reactions. Reports of "electric shock sensation" were mapped to the term "paraesthesia". We agree that the coding of this term does not accurately reflect the reported reaction and therefore will be writing to the MedDRA Maintenance Organisation for an alternative mapping. The reports identified as being coded as ‘Electric Shock’ are incorrect and these have been corrected to the term ‘electric shock sensation’.

Information on patient history and outcomes of reactions are coded as ‘unknown’ only when the reporter does not provide this information on the Yellow Card.

Motion sickness
At the time of the reports, the ADROIT database mapped "motion sickness" to "travel sickness". These reports were not, therefore, miscoded. Motion sickness is now a recognised term on the MedDRA dictionary.

3. Follow up of Yellow Card Reports
Follow up procedures are designed to ensure that relevant information is sought if this is missing from reports of serious reactions which could potentially be new signals. There are some half a million UK reports on the database with around 60,000 new reports received each year. Pressure of time on potential reporters has been identified as one of the reasons for under-reporting adverse reactions through the Yellow Card Scheme and this also applies to requests for follow up information. There is therefore a need to be selective and focussed about which reports require followed up to enable effective risk assessment. Standard operating procedures are in place to guide staff as to which reports should be followed up. Any report which contains a serious reaction and is incomplete in key areas is considered for follow up. The nature of the information considered necessary to complete a report will vary depending on the nature of the suspected adverse reaction. The decision to seek additional information is often a matter for the professional judgment of MHRA staff.

Of 91 reports of suspected suicidal behaviour associated with paroxetine submitted over a period of 12 years, 38 (42%) were followed up for additional information. This rate of follow up is in line with general expectation. Whether the remaining cases fitted the criteria in the Agency’s standard operating procedure at that time is a matter of judgement.

4. Limitations of the data
All spontaneous reporting schemes have limitations in that the data they provide cannot be used to determine the frequency of Adverse Drug Reactions (ADRs) as not all reactions are reported. It is generally accepted that the primary purpose of a spontaneous reporting scheme is ‘signal’ detection (that is to say the detection of a possible causal link between a drug and suspected adverse reaction, which requires further investigation to establish whether a link exists). This has also been recognised in the recent Review of Access to the Yellow Card Scheme. For the purposes of signal detection the minimum information required is a suspected drug and reaction and a ‘identifiable’ reporter and patient. Any report containing this information is considered ‘valid’ and added to the ADROIT database.

5. Signal detection and risk assessment
Taking into account that medical diagnosis is not always precise, when the database is interrogated for detection and analysis of a possible risk, all the coding terms are included which could be relevant to the risk under investigation and the totality of the data, including the image of the original Yellow Card, is examined.

6. Suicidal behaviour and changes in SSRI concentration
A possible ‘signal’ of suicidal behaviour with SSRIs was identified in the early 1990’s. The data from spontaneous reporting cannot generally be used to establish a causal association between a medicine and a suspected adverse reaction, particularly if the suspected adverse reaction is similar to or consistent with underlying disease pathology. The ‘signal’ needs to be confirmed or quantified by other sources of evidence such as clinical trials or epidemiological studies. For this reason the MHRA has instituted a rigorous review of all evidence, including epidemiological, by the CSM Expert Working Group on Safety of SSRIs (EWG).

September 2004