5. Investigation of the circumstances of the incident
5.1 Question arising
Having described the nature of the errors in Section 4 of this report, this Section gives more detailed consideration to the circumstances surrounding these errors.
Of particular concern here are the following questions:
- Why did Radiographer A make the error described in Sub-section 4.6 of this report?
- Why did Radiographer B arrive at the same erroneous answer?
- Why did the warning from RadCalc that the manual calculations were 100% too high not lead to a fundamental re-evaluation of the manual calculations?
- How did Physicist A derive figures from RadCalc that apparently agreed with the erroneous manual calculations?
- Is RadCalc, as it is used at the ECC, suitable for calculations of this type
- Why did the experienced treatment radiographers fail to notice that the monitor units on the plan seemed unusually high?
In seeking answers to these questions, formal interviews were held with the duty holders who were involved directly in the treatment planning process. The content of the following sub-sections draws on the information obtained in these interviews.
5.2 Why did Radiographer A make the error in the manual calculation?
Simply stated, the mistake made by Radiographer A was that, with regard to Sub-section 4.6 and to the method of manual calculation quoted therein from ECC Employer's Written Procedure ' EP2\ ECC\3402 Calculating and Checking Monitor Units of Photon Beam Treatments-Manual Calculations', he failed to double the 'percentage depth dose'.
Figures obtained from the ECC indicate that for the four year period 2012 to 2015, a total of 181 cervical spine radiotherapy procedures were planned and delivered, with an even spread over the four years. Of these, 121 were planned and delivered using a single posterior field, 53 were planned and delivered using (as in this case) lateral parallel opposed fields at 100cm FSD, and seven were planned and delivered using lateral isocentric fields (see Sub-section 4.3).
At interview, Radiographer A estimated that, over the past four years, he had successfully carried out around 5 cervical spine plans for lateral parallel opposed pair treatments to 100cm FSD.
The concern here is that learning that is not reinforced by regular practice can be forgotten, and the likelihood of this will depend on both the complexity of the process, and the degree to which the steps involved are intuitive.
In this case, the process is relatively simple, in the sense that, starting with the prescribed dose to the target point, the task is to make a relatively straightforward calculation of the (equal) doses to each side of the neck. However, this level of simplicity may have contributed to the error, in that both Radiographer A and Radiographer B stated at interview that they recognized that this was a straightforward calculation that they were confident of undertaking correctly, so neither felt the need to refer to ECC Employer's Written Procedure ' EP2\ ECC\3402 Calculating and Checking Monitor Units of Photon Beam Treatments-Manual Calculations'.
For a 'Parallel opposed pair with unequal FSD', both the beam dispersion and (usually) the distances between the skin and the target area are different. It is intuitively obvious therefore (as taught in operator training) that division of the dose into separate beams is required, prior to a different 'depth dose' correction being applied to each (as in 'Method 2 in Sub-section 4.5 of this report).
For a 'Parallel opposed pair with equal FSD', the same intuitive method could be used, but with the same 'depth dose' correction being applied to each of the two beams.
In this regard, it could reasonably be argued that the ECC practice, whereby the treatment planner must ensure that the 'depth dose' correction is doubled and applied to the total dose in both beams, is not intuitively obvious. Further, it could also be argued that the means by which this is described in the ECC Employer's Written Procedure EP2\ ECC\3402 (Sub-section 4.5 of this report) and the way in which the calculation is commonly written, as in this case (Figure 3 but corrected):
'165.2% = 2000cGy in 5# @ 5.5cm'
does not help to make this method of calculation any more intuitive or robust.
At interview, both Radiographer A and Radiographer B volunteered that there were aware of these different methods whereby this calculation could be done. They also agreed that if the method of calculation had required that the beam be divided between the two opposed beams such that no doubling of the 'depth dose' figure would be required, then they were less likely to have made this mistake. Neither of these Radiographers could offer any other explanation of why both made the same error.
In summary, it clearly is the case that manual calculations for 'Parallel opposed pair with equal FSD' had been infrequent, and far less in number than single posterior field treatments where no doubling of the depth dose is required. Also, the current practice whereby the 'depth dose' correction is doubled and applied to the total dose in both beams, rather than dividing the dose between the two beams, together with the means by which this calculation is expressed, does not help to make the change required from the single posterior field calculation an intuitive one.
A reasonable conclusion would be, therefore, that in making an infrequent change in method to one where the process involved is not intuitively obvious, special precautions should be applied to ensure that any critical changes are applied correctly.
5.3 Why did Radiographer B make the same error?
Clearly, the fact that both Radiographer A and Radiographer B made precisely the same error calls into question the degree of independence between the two calculations.
In this regard, Radiographer A and Radiographer B have stated that, in accordance with normal practice, they took the prescribed treatment parameters from the Radiotherapy Prescription Sheet independently of each other, and with no related discussion. They then seated themselves at different desks within the same room and, again with no discussion, used similar 'depth dose' data tables, and the same method of calculation, to calculate the monitor units per treatment fraction for each of the two opposing beams.
It was only on completion of these calculations that they compared answers.
Nothing has emerged during this investigation to cast doubt on these accounts.
At interview, Radiographer B estimated that, over the past four years, he had successfully carried out 'between 3 and 8' cervical spine plans for lateral parallel opposed pair treatments to 100cm FSD.
It seems reasonable to conclude, therefore, that the error made by Radiographer B occurred also as a consequence of the infrequency of the use of this method of calculation and the non-intuitive nature of the change required from the more common posterior single beam treatment.
5.4. Why did the RadCalc result not lead to a re-evaluation of the manual calculations?
At interview, Radiographer A, Radiographer B, and Physicist A all agreed that the difference between the MU as calculated manually and by RadCalc should have prompted a review of the source data and manual calculations. This begs the question of why such a re-evaluation was not pursued.
Historically, in the ECC, the method of treatment for all areas of the body where two lateral beams were involved was to define the required focus to skin distance for each of the two beams and position the patient accordingly ( Sub-section 4.3). However, over the years, this had changed to 'isocentric' treatment for all except palliative treatment of the cervical spine and treatment of the 'whole brain'. Therefore, the use of RadCalc by the Radiographers for 'lateral parallel opposed pair at 100cm FSD' treatments was much less frequent than its use for isocentric treatments.
[There is no particular reason why these two procedures could not have been changed to isocentric treatments, and ECC staff have stated that the reason why these two procedures had not been changed over was simply for 'historical reasons'.]
In defining the RadCalc procedure for 'lateral parallel opposed pair at 100cm FSD' rather than for isocentric treatments, the relevant ECC Employer's Written Procedure EP\ ECC\3422 'RadCalc Instructions', requires some additional steps involving manipulation of the on-screen parameters by the operator. The relevant extract from this Procedure is annexed here as Appendix 2.
The report from the ECC Radiotherapy Incident Working Group noted that the ' perception within radiography staff is that it [RadCalc] does not work well for parallel opposed fields at 100cm FSD'. Further to this, at interview Radiographer A commented that:
' with RadCalc with this particular type of calculation it seems like ... it gets muddled up.'
'.. it feels more like you have to make the RadCalc fit the calculation than making the calculation be checked by RadCalc'.
Reference to Figures 8 and 9, and to the discussion of the content of these screens in Sub-section 4.7, suggests also that, whereas all of the operators involved might have known how to run RadCalc, they lacked the fundamental understanding of the RadCalc process and of the significance of the entries appearing in the various fields that would have pointed to the nature of the error. It might be supposed that this lack of understanding would have contributed further to their lack of confidence in the computed results.
In contrast to their lack of confidence in RadCalc, as noted in Sub-section 5.2, Radiographers A and B stated at interview that they considered that the foregoing manual calculation was a relatively simple one that they were confident of undertaking correctly.
It must be concluded, therefore, that it was this high level of confidence in the correctness of their manual calculations among the radiographers involved, together with their lack of confidence in the use of RadCalc for this infrequent method of treatment, that led to a shared assumption that their manual calculations were correct and the RadCalc MU calculation was wrong. Their interaction following completion of the manual calculation might have reinforced this confidence.
This belief then appears to have permeated all subsequent attempts to reconcile the different means of calculations, and it was therefore the RadCalc rather than the manual calculation process that became the subject of re-evaluation and manipulation until satisfactory final agreement between the computed and manually calculated MUs appeared to have been achieved. However, as illustrated in Sub-section 4.7, had the significance of the entries in the various RadCalc fields been properly understood, there were sufficient indications that this was not the case.
5.5 How did Physicist A achieve apparent agreement between RadCalc and the manual calculations?
As discussed in Sub-section 4.7, the best assessment of what happened is that someone changed the values in the ' RTP Dose' column of both the 'Beam Data for IsoCenter_1' and the 'Beam Data for IsoCenter_3' section 'Points and Off Axis Assistance' screen from the correct value of '200' to '400' for both the RLat and LLat beams, resulting in apparent agreement between the computed and manually calculated MU.
However, it should be reiterated that nothing has emerged during this investigation to confirm when this change was made or to clarify who made it.
The question that arises, therefore, is why would the entries in the 'Beam Data for IsoCenter_1' and in the in the 'Beam Data for IsoCenter_3' sections of Figures 8 and 9 have been changed as described. Again, no clear answer has emerged during this investigation.
However, one plausible explanation relates to the penultimate bullet point in Appendix 2 which requires that ' With Isocentre_3 selected, set the SSD for both beams to 100cm and then enter the prescribed dose per fraction in RTP Dose'. Whereas the intention of this instruction is that the operator should ' enter the prescribed dose per fraction in RTP Dose' for the beam in question, it does not state this explicitly, and it might have been taken by the operator as relating instead to the total prescribed dose per fraction (400cGy).
Further evidence of operator confusion at this point is the fact the changes to referred to in bullet point 5 of Appendix 2 appear to have been made not only, as intended, to the beam data for 'IsoCenter _3', but also the that for 'IsoCenter_1', for which no such changes should be made.
At interview Physicist A commented that having first of all checked that the correct patient had been selected ' I started to do the process and the process that I first of all used was the one that I had used myself when I had been doing these RadCalc sort of investigation, but a few days before I had gone on holiday the work instructions had been updated to include a way of doing it. So I realised that by doing it my own way I had done something wrong and I stopped what I was doing at that point and then went on to the work instructions.'
In this regard, comparing the earlier version of ECC's Employer's Written Procedure EP2\ ECC\3422 'RadCalc Instructions' with the extant version, it is clear that the instructions for ' POP calculation at 100cm FSD' ( Appendix 2) had undergone significant change. However, there is no evidence of any resulting retraining of any of the operators involved with this incident.
Physicist A had considerable previous experience in the use of RadCalc, and would therefore have been expected to identify the evident inconsistencies in the on-screen data. It therefore seems likely that the recent changes to EP2\ ECC\3422 'RadCalc Instructions' contributed to his acceptance of this apparent agreement between the calculated monitor units and the 'plan monitor units' when is should have been clear that this was, in fact, a result of some inappropriate changes to the input data.
5.6 The suitability of RadCalc for calculations of this type.
The operators' lack of confidence in the efficacy of RadCalc for parallel opposed fields at 100cm FSD raises the broader question of the suitability of RadCalc for calculations of this type.
With regard to this perception, the producers of RadCalc, 'Lifeline Software Inc.', have commented that ' RadCalc is fully capable and well suited for these types of treatments'. This, therefore, leads to the further question of whether the issues underlying this perception could have arisen because the procedure used at the ECC for RadCalc calculations for parallel opposed fields at 100cm FSD ( Appendix 2) was incorrect, or was not optimal.
Data input to RadCalc is from two different sources, data transfer from the associated treatment planning software, and by direct manual input. When operated in accordance with EP\ ECC\3422 'RadCalc Instructions', the data that is transferred from RTChart was the prescribed dose, fractionation, the method of delivery, and the result of the manual calculations.
Data that required subsequent manual input were the 'depth' of this calculation point, the 'focus to skin distance' (referred to in RadCalc and in EP\ ECC\3422 'RadCalc Instructions', as the 'source to skin distance' ( SSD)). This manual manipulation of RadCalc also included the creation of the 'calculation point' in the centre of the neck.
The view of the producers of RadCalc was sought, therefore, on whether this was considered to be the recommended or optimal method for non-isocentric treatments such as the one under consideration here. Their view was that whereas this method correctly applied would produce the answer required, it is not the method recommended, and a simpler, less error-prone method was available.
In particular, if, instead of copying one of the 'Isocentres' so the coordinates could be changed so as to identify the 'calculation point' ('IsoCenter 3'), this point could simply be identified in the treatment planning system, and imported into the software system, thus avoiding this step. In this way the rest of the process defined in the six bullet points under 'For patients with POP calculation at 100cm FSD' in EP\ ECC\3422 'Radcalc Instructions' would become unnecessary, and the likelihood of alerts associated with the sequencing of manual data input would be reduced.
It was also noted by the producers of RadCalc that the terminology used, in particular "IsoCenter_3" or 'IsoCenter _1_Copy' was not helpful and that an alternative label such "Calc Point" or "Mid-Plane" would make this point more easily distinguished from the other points when the "Photon Beams" screen is selected.
5.7 Why did the treatment radiographers fail to notice that the MU seemed unusually high?
Even after completion of treatment planning, there remained one final stage at which it might be expected that the error would be identified. Experienced treatment radiographers who operate the Linacs usually have a feel for the number of MU used for manually calculated treatments, even for treatments that are as infrequent as that considered here.
When questioned on why the error was not identified at this stage, the treatment radiographers indicated that their focus was likely to have been on the difficulties in patient set-up that arose from the condition of the patient, and on the associated daily imaging required to verify correct positioning of the treatment area. It seems likely, therefore, that these were contributory factors in their failure to notice that the MU calculated by the treatment planners were unusually high for this particular treatment.