Informing patients effectively about the level of risk of side effects is important, as the patient's perception of such effects appears an important influence on decisions about taking� medicines¹. Patient information leaflets currently list side effects without frequency data. However, the EU now recommends the use of 5 verbal descriptors to indicate frequency� (very common, common, uncommon, rare & very rare). In a series of experimental studies, we found people given the verbal frequency descriptors recommended by the EU over-estimated the side effect risk². Unexpectedly, we also found that people given numerical frequency information (e.g.15%) rarely gave this same percentage, when asked to estimate their personal risk. The aim here is to summarise estimates of side effect risk among those given the numerical information, in order to understand why people tend to over-estimate.

Method

403 adults received written information about a medicine in which the side effect frequency was given numerically (e.g. "This side effect occurs in 2% of people who take this medicine"). The stated frequency for the side-effects studied ranged from 0.02% to 15%. 125 of these participants were given information about a medicine they were themselves taking (statins, inhalers or oral antibiotics). The remaining 278 participants were given information about a fictional antibiotic 'Epidoxin' (part of a hypothetical scenario about consulting their GP with a sore throat). All participants were asked "What do you think is the probability that you will experience the side effect with this medicine?"

Results

Only 158 people (39%) gave the correct estimate i.e. stating that the probability was the same as given on the information sheet. Over-estimates of side effect risk were more common (184; 45%) than under-estimates (57; 14%). Most of those giving an underestimate (32 of 63) gave the estimated probability as 0%. An estimate of 0% was more likely when participants were given information about low frequency side effects (i.e. <1% rather than >1%).

Among over-estimaters, the probability estimate was almost always given as a round number (e.g. 10%, 25% and 50% were common choices). Those given low frequency information often rounded their overestimate to a comparatively low figure (e.g. told 0.04%, common overestimates were 1% and 2%). Some participants told the rate was 0.02% gave back their estimate as 2%, although there was little other evidence to suggest that participants misunderstood the given numerical information. Participants in the 'real medicines' studies were less likely than those in the fictitious medicines studies to give a correct answer.

Discussion

Almost half those given numerical side effect information overestimated the probability of experiencing the side effect themselves. There was little evidence to suggest that calculation errors or misunderstanding percentages were the cause (although poor numerical literacy cannot be discounted). Alternative explanations (e.g. perceived personal susceptibility or mistrust of health information) need to be explored in further research.

References

1. Knapp P, Berry DC & Raynor DK. �Testing two methods of presenting side effect information about common medicines�. International Journal of Pharmacy Practice 2001: 9: R6
2. Berry DC, Knapp P, Raynor DK. Informing medicine users about side effect risk. Lancet. In press.