Underreporting of suspected adverse drug reactions to newly marketed (“black triangle”) drugs in general practice: observational study

BMJ 1998; 317 doi: (Published 11 July 1998) Cite this as: BMJ 1998;317:119
  1. Richard M Martin (drmann{at}, clinical research fellowa,
  2. Karan V Kapoor, research assistanta,
  3. Lynda V Wilton, visiting honorary lecturerb,
  4. Ronald D Mann, senior professorial fellowb
  1. aDrug Safety Research Unit, Southampton SO31 1AA
  2. bSchool of Medicine, Faculty of Medicine, Health and Biological Sciences, University of Southampton, Southampton S016 7PX
  1. Correspondence to: Dr Martin
  • Accepted 12 February 1998

Data on side effects of newly launched drugs are limited,1 highlighting the need for effective postmarketing surveillance. An inverted black triangle (▾) on product literature identifies new products. Suspected adverse reactions to these drugs, however minor, should be reported to the Committee on Safety of Medicines through the yellow card scheme.2 Adverse reactions are underreported,3 and few doctors in the United Kingdom know the meaning of the “black triangle” symbol.4 We assessed the degree of underreporting of suspected adverse reactions to new drugs in general practice and determined if reporting varied when reactions were severe or previously unrecognised.

Patients, methods, and results

The Drug Safety Research Unit performs observational cohort studies (prescription event monitoring) on selected newly marketed drugs in general practice. All patients in England who have been dispensed selected new drugs are identified for these studies by the Prescription Pricing Authority. Questionnaires (“green forms”) are subsequently sent to prescribers asking about clinical events, suspected adverse drug reactions, and events reported to the Committee on Safety of Medicines as suspected adverse reactions. For the 10 drugs we examined (acarbose, risperidone, fluvastatin, tramadol, gabapentin, famciclovir, lansoprazole, zolpidem, venlafaxine, and losartan) median exposure was 46 435 (interquartile range 24 524 to 55 735) patient months. Events recorded by general practitioners as suspected adverse reactions, and those stated as having been reported to the Committee on Safety of Medicines, were classified as serious or non-serious, using the definition published in the British National Formulary.2 We determined whether the event was listed (“labelled”) in the summary of product characteristics at the time of the study; events not listed were classified as unlabelled. Reports stating “non-specific side effects” or intolerance were not classified. By calculating a risk ratio, using non-serious labelled events as the reference group, we determined the likelihood of each category of adverse reaction being reported to the Committee on Safety of Medicines.

There were 3045 events (in 2034 patients) reported as suspected adverse reactions on the green forms during the 10 studies. General practitioners indicated that they had reported 275 (9.0%; 95% confidence interval 8.0% to10.0%) of these reactions to the Committee on Safety of Medicines: reporting was highest for serious unlabelled reactions (26/81; 32.1%) and lowest for non-serious labelled reactions (94/1443; 6.5%) (table). Serious unlabelled and non-serious unlabelled reactions were significantly more likely to be reported than were non-serious labelled reactions. According to general practitioners' responses, the proportion of serious labelled reactions also reported on yellow cards (7/64; 10.9%) was only slightly greater than that of non-serious labelled reactions.

Suspected adverse drug reactions reported by general practitioners on green forms for 10 newly marketed “black triangle” drugs during prescription event monitoring studies 1994-7

View this table:


These findings show a selective reporting bias to the Committee on Safety of Medicines, with general practitioners notifying a greater proportion of adverse reactions that are of greatest clinical concern. Our estimates are subject to potential reporting and recall biases. Some doctors who had submitted a yellow card may not have completed the green form. We would have underestimated the proportion of yellow cards submitted if green form responders were less likely to complete yellow cards than green form non-responders. It seems more plausible that green form responders would be at least as likely to report yellow cards as green form non-responders. Doctors may not have indicated that a yellow card was submitted. As the number of yellow cards reported per doctor is low,5 the impact of recall bias on our estimates is probably limited. Our overall estimate of underreporting corresponds to previous estimates.5 The message that doctors should submit yellow cards for all suspected adverse drug reactions to “black triangle” drugs should be reinforced.


We thank the general practitioners who take part in prescription-event monitoring studies. We also thank Shayne Freemantle and Gillian Pearce for technical help with accessing the data.

Contributors: RMM was involved in formulating the study hypothesis, executing and coordinating the study, study design, analysis and interpretation of data, and writing the paper. KVK was involved in interrogating the prescription-event monitoring database and preparing the data for analysis and contributed to the writing of the paper. LVW was involved in study design, discussion of core ideas, quality control, and interpretation of data and contributed to the writing of the paper. RDM initiated the research project, discussed core ideas, helped formulate the study hypothesis, participated in study design, was involved in interpretation of results, and edited the paper. RDM will act as guarantor.

Funding: None.

Conflict of interest: None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
View Abstract