The third section of the symposium will describe how clinical prediction rules (CPRs) have been developed in the UK and what progress has been made in practical use of them.

There are two main families of cancer CPRs for primary care use: Risk assessment tools (RATs) and Q-Cancer. They have many similarities, and a few differences. Both families of CPRs have been derived from study of GP records. For RATs, this was initially from paper records, then from the CPRD, the largest electronic primary care database in the world. RATs used case-control designs, with the outputs being the risk of cancer, expressed as a PPV. These risks were for single symptoms and pairs of symptoms, plus abnormal laboratory results. Q-Cancer used a different large electronic database, and a quasi-prospective design. Read-coded symptoms plus risk factors such as family history are included in a multivariable equation, which can give a PPV as an output.

CPRs have been tested in the UK. Initial testing was of a desk-easel/mousemat design. 614 GPs in 165 English practices used colorectal or lung RATs 2,593 times in 6 months. Compared to the previous six months, there were 292 additional chest X-rays, 104 extra two-week appointments with a chest physician, and 47 more lung cancer diagnoses. For patients with colorectal symptoms, there were 304 more urgent gastroenterological referrals, 270 more colonoscopies and 10 more cancers.

The next phase of development has been to integrate both RATs and Q-cancer into a GP clinical system. This has been done in collaboration between Macmillan and BMJ Informatica, with CR-UK leading the evaluation. A large study is underway in 2013. At the symposium we hope to demonstrate the system live, and to give some early results.