When evaluating the risks of radiotherapy, it should always be taken into account that the patient is subjected to a potential benefit from the radiotherapy. Any patient, who enters into the process of receiving radiation treatment, will be exposed to many potential sources of harm throughout the medical procedures. While the probability of harm occurring is low, the consequences of that harm can be grave for the individual patient. The combination of these two elements, probability of harm occurring and consequence when harm occurs, constitutes the risk the patient is subjected to. Major cases of accidental exposure in radiotherapy have only been reported a few times over all the years of radiotherapy activity throughout the world, but these events have mostly been very well investigated and it is important that the lessons to learn from these cases are not lost as time passes.
Presented below are very brief case histories of nine major cases of accidental exposures in radiotherapy, together with lessons to learn directly from these cases, and resources for developing a deeper understanding of each of the events.
A cobalt unit for teletherapy at a hospital in Ohio was initially calibrated correctly. During the period 1974-1976 the physicist failed to perform regular measurements (calibrations and QA), but relied on estimations of the decay of the source to predict the dose rate and calculate the treatment time. Rather than calculating the decay, the physicist plotted the dose rate on a graph paper and extrapolated the decaying dose rate over time. Decay was determined from a straight-line plot on a semi-log graph paper with a calendar ordinate. However, the physicist continued the plot on a page that had linear scales on both axes. This created two problems: (1) the linear Y-axis did not correspond to the original log Y-axis, so straight line extrapolation resulted in ever-more incorrect output values; and (2) the linear X-axis did not correspond to the original calendar axis, so extrapolation led to incorrect date values.
These errors in predicting the dose-rate were made by the physicist in the time period 1974-1976. The errors resulted in the dose-rate being under-estimated by 10% to 45%, which translates to the patients receiving corresponding overdoses of 10% to 55% (where the magnitude of error increased almost linearly with time).
The incident came to light because patients started exhibiting symptoms of overexposure. When the accident was investigated, the physicist produced ten calibration documents showing the correct machine output, which subsequently were discovered to have been fabricated. The output of the cobalt unit had, in fact, not been checked for 22 months. 426 patients received significant overdoses as a result.
Until 1982, a hospital in the United Kingdom relied on manual calculations for the correct dose to be delivered to the tumour. Treatments were generally performed at a standard source-to-skin distance (SSD) of 100 cm. Isocentric treatments were rarely performed, because manual calculations of these treatments were cumbersome. Some non-standard SSD treatments were performed, and SSD-correction was then applied. A non-written procedure was in effect for treatments at non-standard SSD (including the few isocentric treatments). The technologists calculated a correction factor in these cases, based on the actual SSD used.
A computerized treatment planning system (TPS) was acquired in 1981, and after some preliminary testing brought into clinical use in the autumn of 1982. Partly because the new TPS simplified the calculation procedures, the hospital began treating with isocentric techniques more frequently. When the first isocentric TPS plan was ready, it was assumed by the technologists that correction factors for non-standard SSD should be applied, and a hospital physicist approved this procedure. It was not recognized that the TPS already correctly applied an inverse-square correction for isocentric treatments. The technologists continued to apply the distance correction factor to all subsequent non-standard SSD calculations. Consequently, distance correction factor was applied twice for all patients treated isocentrically, or at non-standard SSD. This error caused patients to receive doses lower than prescribed.
In 1991 a new computer planning system was installed and a discrepancy was discovered between the new plans and those from the previous system. Further investigation revealed that the original TPS already contained within it the correction for calculations at non-standard SSD. A formal investigation was initiated, and the incorrect procedures were found to have been in place until 1991, or for approximately nine years. During the 9-year period, 6% of patients treated in the department were treated with isocentric technique; for many of these patients it formed only part of their treatment. Evaluation showed that of 1045 patients whose calculations were affected by the incorrect procedures, 492 developed local recurrences that could be attributed to the error. Underdose varied between 5 and 35%.
ASH, D., BATES, T., Report on the clinical effects of inadvertent radiation underdosage in 1045 patients, Clin. Oncol. 6 (1994) 214-225.
Six events of accidental exposure, relating to the same type of accelerator and involving massive overdoses, are known to have happened in the 1980's in USA and Canada. This type of accelerator relied on software for safety, while older models had mechanical and electrical safety interlocks. Several of these events involved unintended carousel positioning prior to treatment.
A typical event could look like this: the operator had in error selected "X" for X rays instead of "E" for electrons and thus moved the cursor up on the screen to correct the entry, followed by hitting the return key several times to skip to the bottom of the screen, and press "B" for "beam-on". After a moment, the console would display 'Malfunction 54' and 'treatment pause'. To start irradiation again, the operator would press "P" for "proceed".
It was later shown that the speed of entering parameters meant a malfunction was permitted through the design of the software, causing the accelerator carousel to be improperly set up for the radiation modality. The result was extremely high electron energy fluence, being repeated several times (by attempting to "proceed" several times), and directed towards the patient. Patients would react immediately, e.g. complaining about feeling a burning sensation, but the malfunctions intermittent characteristics (relying on speed of keyboard input) made the problem difficult to isolate.
An old Co-60 source was exchanged for a new one in a hospital in Maryland, USA, in 1987. All corresponding data for treatment time calculations was updated by a consulting medical physicist, except data for treatment with cobalt beam trimmer bars. The oncologist stated to the consulting physicist that trimmer bars would not be used for treatment anymore, so the data file / program for this type of treatment was not updated by the physicist. Approximately half a year later, treatment with trimmer bars for whole brain treatment was initiated. The staff used the old computer file for calculation of treatment time with trimmer bars, but because the file contained the outdated source activity, patient treatment times were too long.
Therapy staff observed skin erythema on several patients during the time period September 1987 until October 1988. They occasionally expressed their concern to the hospital oncologist, but the reactions were judged as normal during radiotherapy. In October 1988 the consulting medical physicist was notified, and found a non-updated computer data file / program. Patients had been receiving doses 75% greater than prescribed. 33 patients were treated with this particular technique during a time period of 13 months and 20 patients were dead (either during the course of treatment or after conclusion of treatment) at the time of notification to the authorities.
NUCLEAR REGULATORY COMMISSION, Report to Congress on Abnormal Occurrences, NUREG-0090, Volume 11, No. 4. USNRC, Washington DC (1988).
In December 1990 at a hospital in Spain, there was a breakdown of a linear accelerator. A technician from a company was nearby, maintaining a cobalt unit, and was called over to the accelerator. The technician started repair-work the following day, recovering the beam, but an instrument on the control panel always indicated the maximum electron energy (36 MeV), regardless of the selected electron energy value e.g. 7, 10, 13 MeV. Treatments resumed on the following Monday. The technologists observed the discrepancy between the energy selected and the one indicated on the instrument on the control panel. The interpretation was that the indicator must have got stuck at 36 MeV, while the selected energy was the one indicated on the energy selection keyboard.
In reality, a transistor had short-circuited, and independent of the voltage at the base a full current was always fed to the magnet system, making it possible to get a beam only when maximum electron energy was used. To get a beam for each of the electron energies, they had all been adjusted to maximum energy. The design of this accelerator meant that a homogenous field was achieved through scanning of the electron beam, where the current of the scanning magnet had to match the selected electron energy. As the electron energy was at the maximum, the deflection in the scanning magnets was too small and the field thus became concentrated in the centre. This increased the energy fluence and therefore the dose. For 7 MeV, the absorbed dose was about 9 times the intended. This resulting increase in dose was smaller for higher energies and it became nearly unity when the selected energy coincided with the actual energy. During the 10 days of faulty operations, 27 patients were treated using electrons with the equipment. Of the 27 patients, 15 died as a consequence of the overexposure (most of them within 1 year). Two more died with radiation as a major contributing factor.
Include in the Quality Assurance programme:
SOCIEDAD ESPAÑOLA DE FÍSICA MÉDICA, The Accident of the Linear Accelerator in the "Hospital Clínico de Zaragoza", SEFM, Madrid (1991).
A cobalt source was exchanged for a new one in 1996 in a hospital in Costa Rica. At the subsequent calibration, the medical physicist incorrectly interpreted 0.3 minutes as being 30 seconds (instead of the correct interpretation of 18 seconds). As a consequence, the absorbed dose rate of the new source was underestimated, resulting in treatment times being overestimated by 66%.
A radiation oncologist from another hospital, whose patients were treated in the hospital where the event occurred, noticed some unusually severe reactions in some of the patients treated on the cobalt unit. These reactions were related to the skin and lower intestinal track, e.g., diarrhoea and abdominal pain. When another physicist crosschecked the dose rate, the error was found. 115 patients were affected, and two years after the event at least 17 patients had died from the overexposure.
A hospital in Panama, in the year 2000, was working to a busy schedule, treating 70 to 80 patients per day on a single cobalt unit. Many patients were treated in the evening with only a single therapist present, using multiple fields with an SSD set up technique including blocks and wedges. The treatment planning system (TPS) that was used in the hospital, allowed for a maximum of four shielding blocks to be entered in any field in order to calculate dose distribution and treatment time. In April 2000, one of the oncologists required one additional (fifth) block for some treatments in the pelvic region. To overcome the limitation of four blocks imposed by the TPS, a new way of entering data was tried (August 2000): entering several blocks "at once". The TPS accepted the data entry, without giving a warning, but calculated incorrect treatment times.
In November 2000, radiation oncologists observed unusual reactions in some patients (unusually prolonged diarrhoea). The physicists checked the patient charts but did not find any abnormality. However, the computer calculations were not questioned. The treatment time incorrectly calculated by the TPS was approximately twice the required for the correct dose. In total, 28 patients were affected by these incorrect calculations. A few months after discovery, at least five patients had died related to the overdose of radiation.
Five patients were affected by accidental exposure in a hospital in Poland in 2001. Following a power failure at the department, an accelerator was automatically shut down. When electrical power was restored, the accelerator was restarted and some tests were completed without any indication of problem, except a low dose rate indication, which led to the filament current limitation being increased to a high level by staff. The remaining treatments were completed. Two of the patients indicated that they sensed a burning sensation during treatment. The accelerator was taken out of clinical use after the last patient had been treated, and a physicist measured the absorbed dose on the unit. The reading was extremely high.
Further investigations revealed that there had been a double fault: (1) a fault in a fuse of the power supply to the beam monitoring system lead to a high dose rate, even though the display indicated a lower value than normal; and (2) a diode was broken in the safety interlock chain, which should be indicating problems in the dosimetry system. These faults combined, through the design of the system, meant that no problem was indicated, while dose rate was many times higher than intended. All five patients received substantial overdoses and developed local radiation injuries of varying severities.
In 1992, a patient was being treated for anal carcinoma in Indiana, USA, using a high dose rate brachytherapy afterloader, where the source is attached to a wire that can be extended under remote control through one or more catheters in succession into the patient. Five catheters had been placed into the target volume, and a pre-treatment check using a dummy wire testing the catheters had been completed without problems. When the source was being introduced into the catheters, this went well for the first four of these. Upon attempting to direct the source into the fifth catheter, the control console reported an error. After several attempts, the treatment was abandoned.
When the treatment was terminated, the staff entered the treatment room, disconnected the HDR unit from the implanted catheters and removed the patient. An area radiation alarm indicated high radiation levels, but this was ignored. The staff later reported that the alarm "often malfunctioned" and they were used to ignore it. A radiation survey meter was available but was not used to confirm or rule out the area alarm's signal. The HDR console reported that the source was "safe" and the patient was transported back to her nursing home. The hospital staff did not recognize that the source had broken loose from the guide wire, and had remained inside the catheter.
The catheters remained in the patient, with the HDR source, as the patient was transported back to the nursing home. The catheter containing the source fell out four days later and was placed in a "medical biohazard" trash bag. A radiation detector identified radiation emissions from the trash in a trailer sometime later, and the source could be traced back to its origin. The patient had received a massive overdose, and died shortly after the source fell out.
NUCLEAR REGULATORY COMMISSION, Report to Congress on Abnormal Occurrences. 92-18. Loss of Iridium-192 Source and Medical Therapy Misadministration at Indiana Regional Cancer Center in Indiana, Pennsylvania, NUREG-0090, Volume 15, No. 4. USNRC, Washington DC (1992).