The Therac-25 was a linear accelerator built in the 1980s by Atomic Energy of Canada Ltd. It was an accelerator physicist’s dream, and it was a marvel of the radiotherapy world – until it wasn’t.
In the 1970s, the ideal photon beam energy for radiotherapy was still not known, and several investigators were doing research in this area. R.S. Bush and W.E. Alt conducted retrospective studies looking at actuarial survival for gynecological malignancies in eras where megavoltage beam energies had been developed and were implemented clinically, or for previous eras, before the technology was available in the clinic. Harold Johns and Jack Cunningham, in their analysis of treatment planning point to an ideal energy or 25 MeV (see pages 449 and 716 of Johns and Cunningham, 4th Ed), which was felt to deliver optimal dose distributions with the equipment available at that time.
Manufacturers were responding by creating radiotherapy units with ever increasing beam energies. The Betatron was developed to produce high electron beam energies, however it suffered from low dose rates, as explained by Rawlinson. Other manufacturers responded with medical linacs that were capable of both high energy as well as high dose rates. Varian introduced the Clinac 35, which was capable of 25 MV, and Thompson-CSF, a French company, built the Sagittaire, which was capable of 25 MV photons and 35 MeV electrons. In Canada, Atomic Energy of Canada Limited Commercial Products Division (AECL) created the Therac-25, or Reflexotron, as it was also known.
The Therac-25 was and is still be considered one of the most advanced accelerator designs ever to be used clinically. Linear accelerator waveguides transfer energy from oscillating microwaves to electron beams. As electrons travel down the waveguide, the speed of the electron is synchronized with the oscillating electric fields of the microwaves, and so they are continuously accelerated to higher energies. Because the microwaves oscillate, there are accelerating gradients pointing in both the forward and reverse directions. The idea of the Therac-25 was to use both the forward and reverse acceleration properties to create a high energy beam with much less microwave power.
The design was based on a theoretical papery Veksler, published in the Journal of Physics USSR in 1945 (Volume 9, 153). To implement the idea, the Therac-25 employed a reflecting magnet at the distal end of the accelerator. By varying the position of the reflecting magnet, an electron beam energy from zero to twice the single pass energy gain of the accelerator could be achieved. For 25 MeV, a single pass energy or 12.5 MeV was needed, and this could be supplied by a magnetron which was mounted on the rotating gantry. Magnetrons operate at lower pulsed voltages than klystrons, and so this offered considerable advantages over other higher energy machines that were klystron based. However, there were technical challenges for this machine that also had to be solved. Since the electron beam returned to the gun end before being deflected towards the patient by the bending magnet, the gun had to be specially designed with a hole in it so that returning electrons would not damage the thermionic cathode. As well, the reflecting magnet generated considerable amounts of radiation, since the electron beam is essentially stopped and reflected backwards, and so considerable lead shielding was needed to reduce the leakage radiation. Finally, the magnetron was mounted on the gantry, and rotated as the beam angles for treatments changed. The magnetron cathode is mounted by small electric leads, and early in magnetron designs, the cathode could sag with gantry rotation, which decreased the machine stability.
As a particle accelerator, this was a fantastically advanced machine. However, as a medical device, it was also one of the first to be computer controlled. While advanced in terms of technology, it was this feature that ultimately led to its downfall. It’s hard to know exactly when computer control of medical devices was first introduced, but it is safe to say that the Therac-25 implementation was one of the world’s first. It’s system control design was recycled from the Therac-6, which was a medical linear accelerator also produced by AECL in collaboration with CGR-MeV, the makers of the Sagittaire. However, unlike the Therac-6, the Therac-25 relied more extensively on software for its safety features, as opposed to hardware interlocks. Through an unexpected sequence of operator keystrokes, the unit could be configured with the incorrect electron beam for the target/carousel position, causing very high dose rates. This was a latent design flaw that was not discovered until it caused overdoses for six patients, three of which were fatal. The regulatory control of medical devices was different in the 1980s than it is today, and the FDA did not respond in the same way as it would now. Eventually, after the sixth incident, the FDA and AECL agreed to shut down operation of all Therac-25 units until a permanent fix was implemented.
The fault was extensively studied, especially by Nancy Leveson. The experience helped to improve software design considerably for medical devices and for safety systems generally. Today, our complex computer-controlled radiotherapy systems have a fantastic safety record, thanks in part to the lessons learned from the Therac 25.