A synthetic control arm is a comparison group for a clinical trial that is drawn from real patients' records, instead of being enrolled in the trial. It is also called an external control arm. In an ordinary trial, the patients are divided into groups called arms. One arm receives the new medicine. The other, the control arm, receives a placebo, which is a dummy treatment, or the usual treatment. Comparing the two arms shows whether the medicine works.
Sometimes a control arm is very hard to enrol. A disease may be so rare that only a few dozen patients can be found. Or it may be so serious that patients and doctors will not accept a dummy treatment. In such cases, a trial can give the new medicine to every patient. The comparison group then comes from the records of past patients with the same disease, who received the usual care.
The difficulty is making the two groups truly comparable. The patients from records must match the trial patients in age, in how severe their disease was and in earlier treatments. Statisticians use methods such as propensity scores. A propensity score is a number that says how closely each patient from the records resembles the trial patients. The statisticians also check that the result was measured in the same way in both groups.
Say a drug company is testing a medicine for a rare muscle-wasting disease in children. It finds 60 children for the trial, and all 60 receive the medicine. A real-world evidence (RWE) platform searches the records of hospitals in several countries. It finds 900 children with the disease who were treated in the past. It selects the 180 who best match the trial children in age and in how far they could walk in six minutes at the start. After 12 months, the trial children walk 15 metres less than at the start, on average. The matched children from the records lost 45 metres over the same period. The difference of 30 metres is the evidence that the medicine slows the disease.
The trial's own results are recorded in electronic data capture (EDC). The design, including how many patients from records to match, is planned first in clinical trial design software, which simulates the trial before it starts.
A synthetic control arm has a known weakness. Patients from records may differ in ways that the records do not show, such as how carefully their families followed the treatment. So regulators accept it mainly for rare diseases and some cancers, where an ordinary control arm is not possible. The US Food and Drug Administration (FDA) and the European Medicines Agency have accepted such evidence in some approvals. It is one of the ways in which clinical trials are changing, and it depends on de-identified data.