ABSTRACT
After Mary Smith,1 a technician at the East Texas Treatment Center in 1986, had welcomed the can cer patient who was scheduled for a 25 MeV elec tron beam treatment of 180 rads, she helped him onto the treatment table where he positioned him self comfortably After situating the patient, the experienced technician left the treatment room and entered the treatment data on the console of the Therac-25 (T-25) dual-mode accelerator. As she entered the data she mistakenly typed an “X ” for X-ray treatment instead of an “X ” for electron treatment. This slip had happened before, because the majority of treatments were X-ray treatments. Being familiar with the necessary steps, it took her only a few seconds to edit the incorrect entry by moving up the cursor to the entry field and cor recting it. The interface of the accelerator was a full-screen interface on a DEC VT100 terminal. Mary proceeded by hitting the enter key several times to verify the previous entries and last hit “B” to begin the treatment. Almost immediately the machine went into treatment pause mode and dis played the error message “Malfunction 54.” This happened fairly frequently on a daily basis, so Mary was not worried. The accelerator had two malfunc tion modes: If the condition was serious, the treat ment was suspended and the accelerator required a complete reset; if it was a less serious error, a treat ment pause was initiated. Mary knew that “Malfunction 54” initiated a treatment pause, so
she pressed “P” to proceed and resume the treat ment. Because of the cryptic nature of the error messages, she did not know what the error meant and assumed that it was not a serious problem because it could be fixed easily by pressing “P.” Mary hit “B” again to start the treatment, and again the accelerator went into the treatment pause mode. Meanwhile, the patient in the treatment room had experienced two extremely painful shocks when Mary had hit the “B” key. He could not notify her because the video and audio moni toring equipment was malfunctioning. During this treatment, the patient had received a massive over dose (possible doses of up to 25,000 rads in less than 1 second over an area of about 1 cm2). He died several months after the incident. The error investigation indicated that the accident was caused by the technician who entered, too rapidly, an unex pected sequence of commands, and not by the software developer, who did not expect operators to be this proficient. The operator’s rapid entry tripped the computer, which retracted the metal plate used during X-ray mode to absorb radiation but left the power setting on maximum. The termi nal display showed the beam mode as active, while it was set to deliver X-ray blasts of 25,000 rads. Over several months, six accidents were reported in which patients received massive overdoses of rads, and several of them died as a consequence.
