Unintended overexposure of a patient during radiotherapy treatment at the Edinburgh Cancer Centre, in September 2015

4. The nature of the error

4.1 Background to radiotherapy planning and treatment

Radiotherapy involves a series of stages, beginning with clear identification of the size, shape, and position of the tumour (or other region of tissue) to be treated, followed by planning of how best to direct the radiation at the treatment site while minimising damage to healthy surrounding tissue. The resulting plan is then used to direct the treatment machine (the 'Linear Accelerator' or 'Linac') in delivering the prescribed radiation dose.

All ionising radiation, even at the relatively low levels used in the diagnosis of disease or injury can increase the risk of cancer, and it is imperative that all exposures are optimised to minimize this risk. The type of radiation used in radiotherapy treatments is essentially the same as is used for diagnostic X-rays, but the doses delivered in radiotherapy treatments are generally much higher.

At higher doses, X-rays can cause acute tissue damage, and in radiotherapy the intent is to carefully target high doses of radiation so as to inflict such damage on the intended area of treatment. In practice this inevitably means that surrounding, healthy tissue will also receive a high dose of radiation, and this must be kept within tolerable limits. All clinical errors in the use of radiation are taken seriously, but the potential for serious harm to the patient is much greater in radiotherapy than in diagnostic radiology, and special precautions in the planning and delivery of radiotherapy treatments are essential.

Following pre-treatment imaging of the treatment area, the oncologist directing the treatment will prescribe the dose of radiation and the method of treatment delivery. This will normally be recorded on a 'Radiotherapy Prescription Sheet'.

The dose of radiation is measured in units called 'Grays', so, for example, the oncologist might prescribe a total dose to the treatment area of 30 Grays and, for optimal delivery with minimal damage to surrounding tissue, might require that this is delivered in 10 'fractions' of 3 Grays, one per weekday, over a period of two weeks.

The prescribed method of treatment delivery might be by a single beam to one point on the skin, for example, a single beam to the back of the neck, or in multiple beams to more than one point on the skin. In the latter case, by varying the point of delivery, damage to healthy tissue can be mitigated.

The Radiotherapy Prescription Sheet is then passed to the operators responsible for treatment planning, who will undertake the (often complex) task of calculating precisely how the prescribed dose of radiation is to be delivered by the patient by the Linac. This includes positioning of the patient relative to the source of radiation and calculation of the required radiation output from the Linac in terms of 'Monitor Units''

'Monitor Units' ( MU) is the term used for the reading that arises from the monitor on the Linac that indicates the total amount of radiation delivered during an exposure. Once the individually calculated number of MUs for the treatment field have been delivered by the Linac, it terminates automatically. The MU setting is therefore critical in achieving the correct dose.

Complex treatment plans involve the use of sophisticated treatment planning software, but for less complex treatments, MU calculation might be carried out manually. However, manual MU calculations will always be checked independently using a different methodology which usually involves appropriate computer software.

The treatment plan is then provided to the radiographers who deliver the treatment, and data from the treatment planning software is exported electronically to the control systems for the Linac, where it is used for treatment of the patient.

4.2 The prescribed treatment for this patient

Myeloma is a cancer that develops in the plasma cells found in bone marrow. These malignant myeloma cells produce abnormal proteins that can have a number of serious health consequences, including weakened bones and an increased risk of fractures. This typically occurs in the most active bone marrow, which includes the marrow in the spine, pelvic bones, and hips, and because numerous sites can be affected, the disease is often referred to a 'multiple myeloma'.

In September of 2015, a patient diagnosed with multiple myeloma was prescribed palliative radiotherapy treatment at the ECC involving irradiation of the vertebrae of the neck to address pain and disability being caused by a bone fragment from a collapsed 'C3' vertebra ( Figure 1).

This form of treatment is often delivered in a single shot (or 'beam') from behind the neck, but because of concerns that the emerging beam might damage tissues of the mouth, the treatment prescribed for this patient involved the use of two beams, one from each side of the neck. This is generally known as a 'parallel opposed pair' treatment.

The prescription called for a total dose of 20 Grays (usually written as 'Gy') of X-ray radiation to be delivered in 5 fractions, each of 400 centiGrays (hundredths of a Gray, usually written as cGy) over a period of 5 consecutive days. Each of the 400cGy fractions was to be divided into two 200cGy beams, one to be delivered from the left side of the neck, and one from the right. The oncologist wrote this information in the patient's 'Radiotherapy Planning Sheet'.

The oncologist identified the volume for treatment, which included the collapsed cervical vertebra, in a virtual simulation software package ('Tumor LOC'). An appropriate beam to deliver the required treatment from the right side of the neck was then defined by the oncologist in Tumor LOC. This beam was then transferred by the treatment planners to the 'Aria' electronic treatment planning system, where it was mirrored to produce the left lateral beam for treatment. Both the right and left lateral treatment beams were then available in Aria for transfer to other systems as required.

4.3 Positioning the patient for a parallel opposed pair treatment

For a 'parallel opposed pair' treatment, the 'head' of the Linac from where the X-rays emerge is rotated through 180 degrees between the two treatments ( Figure 2). The spacing between the point at which the X-rays are produced within the Linac head (usually called the 'focus' or the 'source') in its left and right position is precisely 200 centimetres.

For this treatment, the patient can be positioned in two different ways. In one of these, the point of treatment within the body of the patient is positioned at the midpoint between the two Linac head positions, i.e. at 100cm from the 'focus', and the patient stays at this point throughout. This is known as an 'isocentric' treatment.

The alternative method (as used here) is to set the patient such that the skin on the right side of the patient is at a fixed distance from the focus for the 'right lateral' beam, and then, when the Linac head is rotated, move the patient on the table sideways so that the skin on the left side is at either the same or a different fixed distance from the focus for the 'left lateral' beam. In this particular case, the patient was moved in such a way that the 'focus to skin distance' ( FSD*) was the same for both positions of the Linac head, and was 100cm. This is known therefore as a 'lateral parallel opposed pair to equal FSD' or as a 'lateral parallel opposed pair to 100cm FSD'.

[*A frequently used alternative term for the 'focus to skin distance' ( FSD) is the 'source to skin distance' ( SSD).]

4.4. Manual calculation of the 'depth dose'

The radiation dose that is prescribed by the oncologist is the dose to the point of treatment. However, since radiation is absorbed by the skin and deeper tissues before reaching the point of treatment, the dose at the surface of the skin needs to be higher than the prescribed dose to allow for this attenuation. Therefore, to calculate manually the dose at the skin surface, which will be the dose delivered to this point by the Linac, the treatment planner uses 'depth dose' tables ( Appendix 1) that define the level of attenuation according to the depth of the point of treatment below the skin surface. This aspect of the dose calculation is frequently referred to as the 'depth dose' calculation.

Consider, for example, treatment of a tumour the centre of which is 10cm below the skin surface using a single beam, for which the oncologist has prescribed a total of 20Gy to be delivered in 5 equal fractions, on 5 consecutive days.

1. The total dose to be delivered is 20Gy.
2. The level of attenuation (from tables) for 10 cm depth is 30%, which means that only 70% of the dose to the surface of the skin will reach the tumour.
3. The required total dose to be delivered to the surface of the skin (referred to as the 'given dose'), as calculated by the treatment planner, is therefore
(20/0.7)Gy = 28.6Gy.
4. Therefore, on each of the 5 days of treatment, the Linac will be set to deliver a 'given dose' of (28.6/5)Gy = 5.71Gy to the surface of the skin.

4.5. 'Depth dose' calculation for a 'lateral parallel opposed pair to 100cm FSD'

For this myeloma patient, the point of treatment at the centre of the neck was taken to be 5.5cm below the skin surface, and the prescription was for 20Gy to be delivered in 5 fractions by 'parallel opposed' beams, as in Figure 2. In this case there are two alternative means of manual calculation whereby the treatment planner can arrive at the same correct answer. These are as follows:

Method 1
1. The total dose to be delivered is 20Gy.
2. The level of attenuation (from tables) for 5.5 cm depth is 17.4%, which means that only 82.6% of the dose to the surface of the skin will reach the area of treatment.
3. The required total dose to be delivered to the surface of the skin on each side of the neck, as calculated by the treatment planner, is therefore, (20/(0.826+0.826))Gy = (2000/(1.652)) =12.1Gy.
4. Therefore, on each of the 5 days of treatment, the Linac will be set to deliver a 'given dose' of (12.1/5)Gy = 2.42Gy to the surface of the skin on both sides of the neck.

Method 2
1. The total dose to be delivered is 20Gy.
2. The level of attenuation (from tables) for 5.5 cm depth is 17.4%, which means that only 82.6% of the dose to the surface of the skin will reach the area of treatment.
3. The required total dose to be delivered to the surface of the skin on each side of the neck, as calculated by the treatment planner, is therefore, (10/0.826)Gy = 12.1Gy.
4. Therefore, on each of the 5 days of treatment, the Linac will be set to deliver a 'given dose' of (12.1/5) Gy = 2.42Gy to the surface of the skin on both sides of the neck.

Clearly, the only difference between Method 1 and Method 2 is that, in Method 1, double the attenuation has been applied to the total of the doses (20Gy) to both sides of the neck, whereas in Method 2, the dose to be delivered to the treatment point from each side of the neck (10Gy) has been divided by 0.826. The answer is the same.

Critically, therefore, for calculations using Method 1, the treatment planner must ensure that the attenuation is doubled (in this case from 0.826 to 1.652), and for calculations using Method 2, the treatment planner must ensure that the dose is halved (in this case from 20 to 10Gy). Failure to do so will, in either case, result in a doubling of the calculated 'given dose' to the skin surface.

The terminology used to describe the proportion of the 'given dose' that reaches the target is the 'depth dose', and in the relevant ECC Employer's Written Procedure the related calculations expressed in terms of the 'percentage depth dose'. So, for example, the 'percentage depth dose' for the calculation above would be 82.6%.

The relevant ECC Employer's Written Procedure for the manual 'depth dose' calculations involved here is EP2\ ECC\3402 'Calculating and Checking Monitor Units of Photon Beam Treatments-Manual Calculations'. The relevant extract from EP2\ ECC\3402 is :

Parallel opposed fields (equal weightings)

Prescribed to central

The dose at central will be 1/2 of that of one field. Find the percentage depth dose from one field for half-separation, double it and equate it the prescribed dose. The given dose will equal the 100% for that field

Example

9 x 9 field separation 20 cm FSD 90 cm
Percentage depth dose at centre from one field = 64.8%
Percentage depth dose at centre from two fields = 129.6%
129.6% = 2000 cGy
100% = 2000 x 100 = 1543 = given dose.
129.6

4.6 How did the error in the manual calculation happen?

Figure 3 is a copy of the relevant section from the actual Radiotherapy Prescription Sheet for this patient, wherein the radiation dose in expressed in centiGrays. These entries were made by Radiographer A.

The first row in this table 'RtLat' and 'LtLat' separates the calculations for the right lateral and left lateral beams ( Figure 2).

The numbers in the second row relate to the calibration of the Linac, and give the relationship between the number that the radiographer enters into the Linac at the start of treatment, the 'monitor units' or ' MU', and the Linac output in centiGrays. So in this case, the linac will deliver a 'given dose' of 0.983cGy to the skin at 100cm FSD for each monitor unit to which the machine is set.

The 'correction factor' in the third row would apply where, for example, the beam of radiation needed to pass through the Linac table before reaching the patient, and the figure in the fourth row would be modified accordingly. In this case there was no such interruption of the beam, so the 'Final output' in the fourth row remains at 0.983cGy for each monitor unit to which the machine is set.

In the fifth row, Radiographer A has, in accordance with the relevant ECC Protocol ( EP2\ ECC\3402), used Method 1 above. The entry here is, '82.6% = 2000cGy in 5# at 5.5.cm', wherein '82.6%' is the percentage (assessed from tables) of the dose at the surface of the skin that remains at the treatment site which is 5.5cm below the skin surface, '2000cGy' is the total prescribed dose, and '5#' is a shorthand term, widely used, and meaning '5 fractions'. However, the critical error here is that, in accordance with this ECC Protocol, the entry should have been '165.2% = 2000cGy in 5# at 5.5.cm'. Because of this, the total dose was divided by 0.826, rather than the correct figure of 1.652. Hence the 'Given dose' in row six, for each side of the neck has been calculated as 2422cGy instead of the intended dose of 1211cGy.

In the seventh row, the erroneous figure of 2422cGy is divided by the number in the first row (0.983) to give the 'Total monitor units' of 2464, one fifth of which (Row 8) is 493.

The practice within the ECC is for the second treatment planner to use the same method of calculation, but to do so independently. Accordingly, Radiographer B has stated that Method 1 was again used but with precisely the same error, hence giving the same erroneous result.

In accordance with normal ECC practice, this second calculation was done on a random sheet of paper which was immediately discarded. There is, therefore, no way in which documentation of this calculation can be revisited.

4.7. Independent MU Calculation using 'RadCalc'

Although both of the manual calculations described in Sub-section 4.6 had similarly arrived at the wrong answer, the next step in the process, whereby the monitor units are calculated electronically from the source data, should have made the planners aware of the error and prompted a re-evaluation of the manual calculations. Clearly this did not happen, and this Sub-section seeks to establish why this was the case.

Figure 4 describes the workflow for the various computer programmes used at the ECC for treatment planning and delivery, including RadCalc, the specialist software used for independent MU calculation.

The process begins with manual entry of data from the Radiotherapy Prescription Sheet into the Aria module called RTChart and External Beam Planning. This data includes the prescribed dose, fractionation, the method of delivery, and the result of the manual calculations.

Data input to RTChart and External Beam Planning was carried out correctly by Radiographer C, and this included the (wrong) value for the MU per dose fraction for each of the two parallel opposed fields (493) from the manual calculation.

Within Aria there is an 'export wizard' by which data can be transferred electronically to RadCalc, where the MU are calculated independently using the prescribed dose, fractionation, and method of delivery. The result is then compared to the figure that has been entered into Aria from the manual calculations.

RadCalc was accessed by Radiographer C, and patient data imported from Aria on the basis of patient identifiers, including hospital number and date. This was done correctly, but, again, with inclusion of the erroneous MU from the manual calculation of the 'given dose' per fraction to each side of the neck.

Within RadCalc, there are a number of 'screens' that are available for selection by the user. The first of these to be viewed is the 'Prescriptions' screen, which displays the 'Prescribed dose', entered as 2000cGy, the 'Dose per Fraction', entered as 400cGy , and the number of fractions, entered as 5. All of these numbers are correct for this patient.

The next screen selected is 'Points and off axis assistance'. The version of this screen that would have appeared for this patient, which has been obtained for the purposes of this description by repeating the data transfer from Aria at a later date, is included here as Figure 5. This has two sections, one of which, labelled as 'Calculation Points', includes data that has been imported from Aria for the relevant points on skin at the right and the left side of the neck, labelled respectively as 'IsoCenter_1' and 'IsoCenter_2'. This data includes the Radiotherapy Prescription Dose labelled as ' RTP Dose (cGy)', which is the contribution to the prescribed dose at the site of treatment from the relevant (right lateral or left lateral) beam, in this case 200cGy from each of the two beams.

ECC Employer's Written Procedure EP2\ ECC\3422 'RadCalc Instructions', ( Appendix 2 to this report) then requires the creation of a 'reference point' which in this case is at the centre of the neck, i.e. midway between IsoCenter_1 and IsoCenter_2. This new point is labelled RadCalc as either 'IsoCenter_1_Copy', or 'IsoCenter_3'.

The final bullet point in the 'RadCalc Instructions' ( Appendix 2) instructs the user to open the 'Photon Beams' screen in RadCalc (recreated here as Figure 6) and change the entries for 'Select Calc Pt' in the 'Beam Setup' section to 'IsoCenter_3' for both beams.

The 'Points and Off-Axis Assistance' screen that should have appeared at this point in the procedure (recreated for this report) is shown here as Figure 7.

The requirement here is that the MU calculated by RadCalc for each of the two beam entry points (IsoCenter_1 and IsoCenter_2) should agreed with the manually calculated MU to within 2.5%. Clearly, as shown in the top right of the screen in Figure 7, this is not the case.

This should have led the radiographers concerned to question their manual calculations. However, as noted in the aforementioned RIG report, there was a perception among the Radiographers that RadCalc ' did not work well for parallel opposed fields at 100cm FSD'. Therefore, instead of questioning the manual calculations, the Radiographers sought to determine why RadCalc had (in their minds) 'failed', and a fourth radiographer, Radiographer D tried to resolve this discrepancy.

Despite these attempts, the difference remained unresolved, so, in accordance with normal ECC practice, they referred their concerns to the appropriate member of the Treatment Planning Section, who then attended the 'Aria' room and took control of the RacCalc calculation.

On questioning, none of the operators involved could clearly recall the precise entries that were in the various columns in the 'Points and off axis assistance' screen at this point in time. Both Radiographer C and Radiographer D have accepted that it is possible that, in seeking this resolution, some manual changes to the parameters on this screen could have been made. However neither Radiographer could recall having made any such changes, or could offer any reason for having done so. Physicist A has stated with some certainty that he did not change any of this data.

Precisely what happened next regarding the various entries in the relevant RadCalc screen remains unresolved, but actual versions of the 'Points and off axis assistance' screens that were saved finally for this patient are shown here as Figures 8 and 9.

Referring firstly to Figure 8, and comparing the entries in the 'Beam Data for IsoCenter_1', section with those in Figure 5, it is clear that the ' RTP Dose' entries in Figure 8 have been changed from those that would have appeared initially. In particular, the figure for the '1 RT Lat' beam has been changed from '200' to '400', and figures have been entered for the '2 LT Lat beam' where there should be none. (The data here should be for the (selected) 1 RT Lat beam only.)

Figure 9 shows that (cf. Figure 7) these same changes in the ' RTP Dose' entries from '200' to '400' have been made for the 'Beam Data for IsoCenter_3'.

For the purposes of this report, Figure 10 is a recreation of Figure 7, the intention of which is to illustrate the immediate effect of making this change from '200' to '400'. As demonstrated, the immediate result of changing the ' RTP Dose' entries from '200' to '400' is to trigger the appearance of an alert indicating that ' The cumulative dose per fraction for the beams associated with this prescription [now 800cGy] exceeds the prescribed dose per fraction [400cGy] by 100%'. In this regard, the producers of RadCalc have commented that ' this warning is intended to aid the user in diagnosing the error they made in manually entering information'.

[The fact that this alert does not appear in either of Figures 8 or 9 could have arisen because the user has ticked the ' Do not show this warning again' box. However, any change that is made by the user, for example, in moving to and back from the 'Photon Beams' screen also causes this alert to disappear from the screen.]

Regarding this alert, Physicist A stated at interview that :

'This is something that sometimes comes up if you have more than one isocentre and more than one reference point. So on that basis it didn't ring alarm bells.'

Whatever the scenario, this alert appears to have been ignored.

Referring again to Figure 9 (cf. Figure 7) the effect of changing the ' RTP Dose' entries in the from '200' to '400' has been to change the entries in the top right hand corner of Figures 9 (for both 1 RT Lat and 2 LT Lat beams) which are now ' MU = 488, Plan MU =493, % Diff = -1.0%', indicating apparent agreement within tolerance, and this agreement appears to have been taken as confirmation of the correctness of the manual calculations.

In the (recreated) 'Photon Beams' in Figure 6 the 'Isodose Line at Calc Pt (%)' entry is the total of the entries in the 'Dose at Calc Point (cGy)' for the RLat and LLat breams divided by the total prescribed dose per fraction and express as a percentage. In this case, therefore, the figure is:

((200 + 200)/400)x100% = 100%.

This is the figure expected.

Figure 11 is the saved version of the 'Photon Beams' screen for this patient. In this case, the 'Dose at Calc Point (cGy)' copied from the entry for ' RTP Dose' in the 'Beam Data for IsoCenter_3' section of the 'Points and off axis assistance' screen is now 400cGy for both beams. Therefore, the 'Isodose Line at Calc Pt (%)'

((400 + 400)/400)x100% = 200%.

Had the fact that this entry was 200% rather than the expected 100% been noted and had the implication been understood, then this should have given the user a further clear indication that the dose data being used by RadCalc in calculating the Monitor Units for comparison with those calculated manually was 100% too high.

In summary, at this point in the process, as a result of what appears to have been inappropriate manipulation of the on-screen data, the figures in the top right had box in Figure 9 indicated that the difference between the manually calculated monitor units as entered into RTChart and the figure calculated by RadCalc was within the required tolerance for both the 'Right lateral' and 'Left lateral' beams. It appears, therefore, that the other indicators within RadCalc that the manually calculated MU figure was 100% too high were dismissed because of a perception by the users that RadCalc ' did not work well for parallel opposed fields at 100cm FSD'. The treatment planning and delivery process therefore proceeded accordingly, using the wrong MU figure.