Therac-25 Radiation Overdoses and Software Safety

The claim and its context

Therac-25, a computer-controlled radiation therapy machine, delivered massive unintended doses to patients in several incidents in the 1980s. The central documented cause was a combination of software defects, unsafe design assumptions, and inadequate incident response—not an intentional experiment or evidence of a plot against patients.

Historical chronology

Therac-25 entered clinical use in 1983. Between 1985 and 1987, treatment accidents were reported in the United States and Canada. Patients reported sensations such as burning; operators initially received confusing error messages. Investigations identified race conditions and shortcomings in the machine’s safety design, testing, and reporting.

What the record establishes

Nancy Leveson and Clark Turner’s investigation reconstructs six known accidents, with severe injuries and at least three deaths attributed in the literature. Exact dose reconstructions were difficult because the system did not reliably record delivered dose. The case became a landmark in software safety precisely because ordinary software failure could have physical consequences.

Clinical and scientific context

Radiotherapy intentionally delivers ionizing radiation to destroy tumors while limiting exposure to healthy tissue. A machine must independently prevent unsafe beam settings and provide reliable dose records. Therac-25 removed hardware interlocks used in earlier models and relied more heavily on software; a rapid operator response could expose a patient to a dangerous beam.

Counterevidence, uncertainty, and calibration

Investigators disagreed about details of individual exposures because machine logs and dosimetry were inadequate. Not all reported patient injuries can be causally assigned with equal certainty. The evidence supports preventable device safety failures and organizational shortcomings, not a claim that radiation therapy as a field or modern devices share the same defect.

What could change the assessment

Original service logs, patient records, source code, and independent dosimetry could refine individual dose and causation estimates. A claim of deliberate action would require evidence of intent or instructions beyond the technical failure record.

Source notes and limitations

1. Leveson and Turner — An Investigation of the Therac-25 Accidents, IEEE Computer (1993)Foundational peer-reviewed reconstruction using incident reports and technical analysis; dose records were incomplete.
2. Leveson — Medical Devices: The Therac-25Researcher’s technical case-study archive with chronology and design analysis; it is a retrospective engineering source.
3. University of Virginia — A History of the Introduction and Shut Down of Therac-25Ethics case-study history of the accidents and shutdown; a secondary account complementing the technical investigation.