Prescription drugs are high-benefit but also high-risk for patients. However, the harms of treatment are often poorly quantified by the randomised controlled trials that define benefit and these trials often exclude the majority of real-world patients with target conditions because of higher risk of harm. This creates the need for observational research to quantify the risks of drugs in real-world populations (pharmaco-epidemiology). Even for drugs where the risks are well-known, we have shown that high-risk prescribing is common. Such prescribing is not always inappropriate, because prescribers and patients are often trying to balance benefit and risk in conditions of considerable uncertainty where all options are problematic in some way. Improving primary care prescribing safety to reduce risk and harm while avoiding under-treatment is a complex task. We have current grant and PhD funding across the spectrum from the basic science of pharmacoepidemiogy to the applied science of development and randomised trial evaluation of interventions to improve safety, and can offer PhDs anywhere in this spectrum.