The adverse consequences of stable coronary artery disease (CAD) are death, myocardial infarction (MI) and angina. Trials in stable CAD show that percutaneous coronary intervention (PCI) does not reduce mortality. PCI does appear to reduce spontaneous MI rates but at the expense of causing some periprocedural MI. Therefore, the main purpose of PCI is to relieve angina. Indeed, patients and physicians often choose PCI rather than first attempting to control symptoms with anti-anginal medications as recommended by guidelines. Nevertheless, it is unclear how effective PCI is at relieving angina. This is because, whereas anti-anginal medications are universally required to be tested against placebo, there is no such requirement for procedural interventions such as PCI. The first placebo-controlled trial of PCI showed a surprisingly small effect size. This may be because it is overly simplistic to assume that the presence of a stenosis and inducible ischaemia in a patient means that the clinical chest pain they report is caused by ischaemia. In this article, we review the evidence base and argue that if we as a medical specialty wish to lead the science of procedures for symptom control, we should recognise the special merit of placebo-controlled experiments.