Background
The perceived risk of a medicine is said to be the most important aspect when a person considers whether to take it(1). Recent studies found both verbal and numerical risk descriptors are associated with a large overestimation of side effect risk(2). One explanation is that people have high baseline estimates of risk, i.e. in the absence of provided information, the public associate medicines with a higher risk than is the case. There are few data on people's baseline estimates of the likelihood of the harm and benefits of medicines, made in the absence of provided information. Here we looked at peoples' estimates for two common medicines: penicillin and ibuprofen.

Method
We used a randomised controlled design with two independent variables (medicine; order of questions). 104 adults were recruited through opportunity sampling in an urban setting. They self-completed a questionnaire, which used a scenario of seeing a doctor and being prescribed a medicine. Half the participants received a scenario of penicillin for respiratory infection and the other half a scenario of ibuprofen for back pain. They completed questions on the likelihood of getting any side effect with the medicine and of it having a benefit (curing the infection or stopping the pain). To control for any framing effects, half the participants received the benefits question before the side effect question (and the others, vice versa). Participants also estimated the likelihood of 2 individual side effects associated with each medicine.

Results
Participants gave highly variable estimates of the likelihood of harm and benefit. With penicillin the mean estimate of benefit was 67.5% (SD 21.5), while the mean overall estimate of harm was 26.4% (23.1). With ibuprofen the mean estimate of benefit was 55.6% (28.7), while the mean estimate of harm was 30.0% (22.9). There was no significant correlation between participants' estimates of harm and benefit of with a medicine. The order of questions made no meaningful difference to estimates of harm or benefit associated with either medicine. Individual side effects elicited mean estimates of: (penicillin) diarrhoea 19.8% (17.3), blood disorders 12.9% (13.1); (ibuprofen) stomach upset 24.7% (18.8), kidney problems 15.0% (18.5).

Conclusions
People think that these common medicines will work in more than half of cases, much more effective than the evidence for them. They also estimate that any side effect would occur in more than a quarter of cases. Individual side effects were estimated in about a fifth of cases, much higher than known rates. Participants rarely used low numbers in estimates (e.g.<1%). The low correlation between harm and benefit estimates suggests we cannot easily categorise people as negative or positive about medicines. The large overestimation of risk of side effects seen in descriptor studies may be partly due to high baseline estimates.

References
1. DC Berry, IC Michas, T Gillie, M Forster. What do patients want to know about their medicines and what do doctors want to tell them. Psychology & Health 1997, 12: 467-80.
2. DC Berry, P Knapp, DK Raynor. Provision of information about drug side-effects to patients. Lancet 2002, 359: 853-854.