Information Technology Reference
In-Depth Information
AECL also decided to reuse some of the Therac-6 and Therac-20 software in the
Therac-25. Code reuse saves time and money. Theoretically, “tried-and-true” software
is more reliable than newly written code, but as we shall see, that assumption was invalid
in this case.
AECL shipped its first Therac-25 in 1983. In all, it delivered 11 systems in Canada
and the United States. The Therac-25 was a large machine that was placed in its own
room. Shielding in the walls, ceiling, and floor of the room prevented outsiders from
being exposed to radiation. A television camera, microphone, and speaker in the room
allowed the technician in an adjoining room to view and communicate with the patient
undergoing treatment.
We now chronicle six major incidents involving the Therac-25 that occurred between
June 1985 and January 1987, along with the AECL responses to them.
MARIETTA, GEORGIA, JUNE 1985
A 61-year-old breast cancer patient was being treated at the Kennestone Regional On-
cology Center. After radiation was administered to the area of her collarbone, she com-
plained that she had been burned.
The Kennestone physicist contacted AECL and asked if it was possible that the
Therac-25 had failed to diffuse the electron beam. Engineers at AECL replied that this
could not happen.
The patient suffered crippling injuries as a result of the overdose, which the physi-
cist later estimated was 75 to 100 times too large. She sued AECL and the hospital in
October 1985.
HAMILTON, ONTARIO, JULY 1985
A 40-year-old woman was being treated for cervical cancer at the Ontario Cancer Foun-
dation. When the operator tried to administer the treatment, the machine shut down
after five seconds with an error message. According to the display, the linear accelerator
had not yet delivered any radiation to the patient. Following standard operating pro-
cedure, the operator typed “P” for “proceed.” The system shut down in the same way,
indicating that the patient had not yet received a dose of radiation. (Recall it was not
unusual for the machine to malfunction several dozen times a day.) The operator typed
“P” three more times, always with the same result, until the system entered “treatment
suspend” mode.
The operator went into the room where the patient was. The patient complained
that she had been burned. The lab called in a service technician, who could find nothing
wrong with the machine. The clinic reported the malfunction to AECL.
When the patient returned for further treatment three days later, she was hospital-
ized for a radiation overdose. It was later estimated that she had received between 65 and
85 times the normal dose of radiation. The patient died of cancer in November 1985.
