Communicating the harmful effects of medicines

In 2004 the US Food and Drug Administration warned that antidepressants could increase suicidality in children and adolescents.1 Three years later it ordered that all antidepressants should carry an expanded black box warning, incorporating information about an increased risk of suicidal symptoms in young adults aged 18-24 years.2 These warnings were associated with widespread media coverage.3 4 5

Warnings from regulatory agencies about harms from drugs are known to reduce prescribing rates. In a linked paper Lu and colleagues (doi:10.1136/bmj.g3596) show that not only did prescribing rates decrease in this age group but there was a modest but significant increase in the rate of self poisoning, an important suicidal behavior.6 The net effect of the warning was probably counterproductive and led to more harm.

Completed suicide is such a rare event that even this large observational study lacked the power to investigate this outcome. Nevertheless, self poisoning is a major event for patients, their family, and health services. Lu and colleagues’ study is a good example of the value of a quasiexperimental pharmacoepidemiological design following a timed intervention in investigating important but uncommon adverse events.6 The findings are not only relevant to the ongoing debate about the risk of prescribing antidepressants for young people but also raise important questions about the impact of warnings about drug safety on treatment related behaviours and health outcomes.

The issue around antidepressants may be a special case, with no shortage of strong opinions that often seem driven more by ideology than by science.7 Regardless, the negative effect of warnings about drugs should be considered in general. Scientific findings associated with harmful events probably attract more interest and therefore more media coverage than do the results of positive effects.8 As regulatory authorities have more to lose from not providing warnings than from providing them, and with society becoming increasingly risk averse, might they be more likely to issue warnings even if the evidence of net harm is uncertain?

The publicity around medical research needs to be managed carefully. The idea that treatments cause harms as well as benefits is well accepted. In practice, however, it sometimes seems that warnings are driven by a desire to avoid any potential harm even if this also prevents benefits—a too simplistic application of “first, do no harm.” An unduly negative approach to medical treatments in the media can be harmful if doctors and patients are put off using well established drugs. The World Health Organization estimates that about half of medicines prescribed for long term conditions are not taken as prescribed.9

Studies of patients’ beliefs about medicines show that concerns about potential harms from drugs may be more prevalent than the experience of harmful effects.10 Moreover, concerns about specific prescribed medicines are related to more general, negative views about drugs as a class of treatment held by many people.11 12 13 As well as impeding adherence, concerns about drugs may result in increased reporting of side effects. In a longitudinal follow-up study of patients with rheumatoid arthritis concerns about drug treatment at baseline predicted reports of side effects at six month follow-up, after controlling for disease and treatment variables and previous experience of side effects.14

A better understanding is needed of how warnings and reports of potential harm influence patients’ concerns about their medicines, adherence, and doctors’ prescribing behaviour. Negative background attitudes towards medicines may mean that statements warning that a drug “may be linked” to harm are readily interpreted as “are linked” to harm. Ironically, if such perceptions result in non-adherence to an essential drug or an increase in reported adverse effects, the warning of potential harm may cause an unintended and paradoxical increase in harm.

Concern has been justified about the biased presentation of the effects of treatments, including antidepressants, as overly beneficial through publication bias and excessive marketing. We need to ensure that clinical trials are not designed to obscure harm through patient selection, drug use before randomisation, and exclusion of patients who experience adverse effects. Concealing harm through selective reporting of trial results can lead to the loss of effective treatments, which could have been retained had the harms been openly reported and clearly explained to patients (for example, rofecoxib). A lack of trust in drug manufacturers to be honest about adverse effects adds to the problem and needs to be tackled.

However, we should also be concerned about the way harms are communicated, given the risks associated with untreated illness. As Lu and colleagues’ show, reports of harm and warnings from regulators, in common with other healthcare interventions, can cause harm as well as benefit.6 We need a better understanding of how patients, doctors, and others respond to warnings to ensure that attempts to reduce drug related harm do not inadvertently achieve the opposite.

This sort of study has implications for premarketing regulatory requirements. Randomised trials are always going to be too small and underpowered to detect important but uncommon adverse events. Excessive premarketing safety requirements inflate the cost of drug development and are one reason why many pharmaceutical companies have left the area of mental health. Innovation is needed in regulatory approaches to new medicines. One approach would be to use the rich data in individual electronic health records to conduct extensive and powerful observational evaluations of the effects of treatments in the real world. Then, fewer and shorter randomised preregistration studies may be needed, perhaps with an initial period of marketing authorisation limited to settings participating in pharmacovigilance programmes.