A Multicentre Timing Study of Intensity-modulated Radiotherapy Planning and Delivery
Article Outline
Abstract
Aims
The aim of the study was to measure how long the intensity-modulated radiotherapy (IMRT) process takes, both for planning and delivery, using several IMRT techniques and departments.
Materials and methods
Timings were measured at three radiotherapy centres for each step of the process of outlining, planning and delivering IMRT for head and neck cancers. Times were measured for a total of 63 patients; 27 with helical tomotherapy, 37 with dynamic sliding window (26 in one centre, 11 in another) and nine with step-and-shoot.
Results
The mean time to outline a patient was 108
min, to produce and check the plan 7.9h, to carry out and analyse patient-specific quality assurance 1.9h. The mean treatment time (including on-treatment verification imaging where carried out), measured gate to gate, was 28min 10s for first fractions and 20min 20s for subsequent fractions.
Conclusion
An analysis of subgroups showed some differences in times between techniques, and some differences between departments with the same techniques. For all four techniques, the median time from the end of outlining to the start of treatment was under 3 weeks.
Key words: IMRT, resources, timing
Introduction
There is considerable evidence that intensity-modulated radiotherapy (IMRT) benefits patients, especially for head and neck radiotherapy [1]. After the recommendations of the National Radiotherapy Advisory Group [2], there is pressure to implement IMRT more widely in England than is currently achieved. One of the potential difficulties is a belief that the resources in terms of staff time to create and check an IMRT plan are large, and that IMRT leads to low throughput on linear accelerators. There is, therefore, considerable national interest in knowing the level of resources necessary to implement IMRT.
Overview
A SCOPUS search on ‘IMRT resource’ yielded 42 results. Of these, only three papers contained measurements of the time taken for either the treatment planning or the treatment delivery process for IMRT [3], [4], [5]. The references from Miles et al. [3] yielded another relevant paper [6]. A PubMed search on ‘radiotherapy, intensity-modulated/economics’ gave four results, of which one contained timing data [7] and a PubMed search on ‘tomotherapy workload time’ added one extra paper [8], giving a total of six papers containing timing resource information for IMRT.
The literature has reported timing data for target volume outlining, the creation of treatment plans, quality assurance measurements and treatment delivery (including the time taken to verify the geometric placement).
However, care should be taken comparing these data, as there are many variables within these data sets; the outlining and treatment planning may be undertaken by different staff groups, the treatment imaging protocols may be different between centres and treatment delivery may be recorded as different aspects of the process. In particular, some references [8] include only beam-on time and others the total gate–gate time (time from the patient entering to exiting the treatment room). Most of these data are from single-centre studies. The one study with planning times from multiple centres was questionnaire-based rather than a timing study.
We describe the results of a multicentre timing study, carried out in three departments in East Anglia, looking at times for volume definition, treatment planning, quality assurance and delivery of IMRT of head and neck cancer.
Methods
Data were collected prospectively, from three radiotherapy departments, all using IMRT for the treatment of head and neck cancer, between July 2008 and May 2009. In two of these centres (Norwich and Ipswich) the technique was dynamic (sliding window) IMRT on a Varian linac. The third centre (Cambridge) had two techniques: step-and-shoot IMRT on a Siemens linac and helical IMRT on a TomoTherapy HiArt™ unit. The treatment planning systems used were Eclipse in Ipswich and Norwich, TomoTherapy, for tomotherapy in Cambridge and XiO for step-and-shoot in Cambridge. The delineation of anatomical volumes (outlining) in Cambridge was carried out using ProSoma™ virtual simulator software for both IMRT techniques. All centres taking part had well-established (>2 years) IMRT programmes.
The process followed at each centre was broadly similar: patient planning images were acquired using as a minimum computed tomography (and in some cases magnetic resonance imaging or positron emission tomography); target volumes were outlined; organs at risk were outlined and data were prepared for treatment plan creation; an IMRT treatment plan was created using the chosen inverse-planning techniques; plans created were independently checked; an IMRT quality assurance plan was created, applying the patient’s plan to a phantom; some measurements were carried out on the machine using the quality assurance plan; the results of the quality assurance were analysed. A slight variation existed at Ipswich, where the pre-planning (including organs at risk outlining) was carried out before the target volumes were outlined, and the checking of the plan took place after the quality assurance measurements.
On the treatment machine, the process was as follows: the patient entered the room, was set up on the couch according to local protocols; where local imaging protocols specified daily imaging, imaging was carried out to local protocols (planar imaging or Megavoltage Computed Tomography (MV CT) volumetric imaging); local correction protocols were followed to act on the imaging, and the treatment was then delivered.
The level of on-treatment imaging involved with the treatment delivery varied between techniques. For patients on tomotherapy, daily volumetric imaging was carried out, with on-line image analysis and daily set-up correction. At Ipswich, daily portal imaging with on-line image analysis and daily set-up correction was carried out using orthogonal pairs of planar images. At Norwich, and for the step-and-shoot in Cambridge, portal imaging was carried out using orthogonal pairs of planar images for the first three to five fractions, with off-line image analysis for calculation and correction of systematic set-up errors.
The different process tasks are undertaken by different staff groups at each centre. All target volume outlining in this study was undertaken by consultant oncologists, but plan preparation, where normal structures are outlined, was carried out either by oncologists, radiographers, physicists or dosimetrists. Plan creation, checking and quality assurance procedures were undertaken by radiotherapy physicists or dosimetrists, but quality assurance analysis was undertaken exclusively by physicists. Treatment delivery and verification was carried out by radiographers in all the evaluation centres. The tasks timed were carried out by the standard staff members performing these tasks clinically.
Times were recorded for each stage of the process described above, apart from the initial computed tomography scanning. Data were therefore gathered for outlining by oncologists, pre-planning (including additional outlining), plan creation, plan checking, preparation of quality assurance plan, measurement of quality assurance plan and analysis of quality assurance measurements. On the treatment machines, times were recorded from the patient entering the room, the start of the first imaging beam, the start and the end of treatment delivery, and exit from the room.
The dates on which each task took place were also recorded, to enable the analysis of process times, a measure of the length of time taken in the local system between finishing successive tasks in the pathway. Data were also collected for each task as to the experience of the person carrying out the task, whether they had carried out the task less than five times, between five and 10 times or more than 10 times.
Results
Data were collected for 37 dynamic IMRT patients (26 at Norwich, 11 at Ipswich), 27 tomotherapy patients and nine step-and-shoot IMRT patients.
Outlining Times
The overall mean time by the oncologists to delineate the volumes was 108min.
Figure 1 shows the time taken, subdivided by technique and centre. The P value, calculated using a two-tailed Student’s t-test, shows that none of the four sets of measurements had a significant difference from the overall mean.
Fig. 1
Outlining times (minutes) by intensity-modulated radiotherapy technique/department.
Figure 2 shows the results subdivided by oncologist. Calculating P values for each oncologist relative to Dr 3, all P values were below 0.03.
Fig. 2
Outlining times (minutes) by oncologist.
Process Times
The number of calendar days for each stage of the planning and quality assurance process is shown in Fig. 3. The order of the processes varied slightly between departments, hence the use of the category ‘final process’. In all departments except Ipswich, the check of the treatment plan was carried out before the quality assurance measurements, with the production of the quality assurance plan happening before or in parallel with the check of the plan. For these departments the ‘final process’ was the analysis of the quality assurance measurements. In Ipswich, the ‘final process’ was the check of the treatment plan. The end of this process indicates when the plan is ready to be treated; the time from ‘final process’ to treatment indicates either linac availability or a wish to start on a particular day of the week.
Fig. 3
Process times (days). See text for an explanation of processes.
The spread of process times is indicated with the lower and upper deciles.
Planning and Checking Times
The overall mean time for the planning and checking was 477min (7.9h). Figure 4 shows the times for each subtask of the process. ‘Pre-plan’ includes the outlining of any structures not outlined by the clinician (including any dummy structures required for optimisation). In the case of tomotherapy, it also includes starting the beamlet calculation; this calculation runs as a batch job overnight, with the planning task being completed the next day. The actual time taken by the beamlet calculation is not included in these figures, as no staff involvement is required.
Fig. 4
Task times (hours) for planning and checking.
Quality Assurance Analysis and Measurement
The overall average time for the measurement and analysis of patient-specific quality assurance was 122min (2.0h). Figure 5 shows the times for each subtask of the process.
Fig. 5
Task times (hours) for quality assurance measurements and analysis.
To determine whether the differences apparent in Fig. 4, Fig. 5 were significant, P values were calculated. Table 1 shows the means, standard deviation and P values (calculated relative to the overall mean time).
Table 1. Time in minutes for planning and checking, and for quality assurance and analysis
| n | Planning and checking | Quality assurance measurement and analysis | |||||
|---|---|---|---|---|---|---|---|
| Mean | Standard deviation | P value | Mean | Standard deviation | P value | ||
| Tomotherapy | 27 | 409 | 128 | 0.00 | 92 | 50 | 0.02 |
| Cambridge step-and-shoot | 9 | 908 | 300 | 0.00 | 252 | 108 | 0.01 |
| Norfolk and Norwich | 26 | 484 | 173 | 0.94 | 109 | 51 | 0.58 |
| Ipswich | 11 | 339 | 217 | 0.05 | 73 | 12 | 0.00 |
It can be seen that Norwich had no significant difference from the overall mean. Ipswich and tomotherapy were both significantly faster than the mean, whereas step-and-shoot IMRT in Cambridge was significantly slower, both to plan and to quality assure.
The quality assurance process for tomotherapy involves a single delivery of the plan to a ‘cheese phantom’ containing a film and two ionisation chambers. The Cambridge process for step-and-shoot involves delivering the plan a minimum of three times to a phantom, plus the acquisition of fluence maps. The Ipswich quality assurance process involves the measurement by ionisation chamber of one or more dose points in a water equivalent slab phantom together with portal dosimetry on each individual beam using the Varian portal dosimetry system Norwich changed their quality assurance process during the course of this study; for most of the patients (n=18) the quality assurance involved the use of a two-dimensional array as well as portal dosimetry. For a minority of patients (n=7), the two-dimensional array was omitted. The ones with the two-dimensional array had measurement times of 54±26min, with analysis times of 70±30min. Without the two-dimensional array, these reduced to a measurement time of 19±6min (P value<0.01), with analysis times of 54±23min (P value=0.19). So omitting the two-dimensional array made a significant reduction in measurement time, but not in analysis time. Norwich were also the only centre using in vivo diodes for IMRT treatments and the quality assurance process included the time taken to select an appropriate position for the diode within the beam.
Treatment Times
The total treatment session time (measured from the patient entering to exiting the room) averaged 28min 10s for the first fraction of a course of radiotherapy, and 20min 20s for subsequent fractions. Table 2 shows these values for each department/technique. Fig. 6, Fig. 7 show the times for the individual subtasks of the treatment process. It is impossible to extract the imaging times from the beam times for Norwich, as the protocol for off-line imaging results in the imaging beams being intermingled with the treatment fields.
Table 2. Overall treatment times (minutes:seconds)
| First fraction | Subsequent fractions | |||||||
|---|---|---|---|---|---|---|---|---|
| n | Mean | Standard deviation | P value | n | Mean | Standard deviation | P value | |
| Tomotherapy | 14 | 24:42 | 4:18 | 0.01 | 22 | 20:22 | 4:41 | 0.97 |
| Cambridge step-and-shoot | 5 | 31:46 | 11:48 | 0.53 | 8 | 23:45 | 2:04 | 0.00 |
| Norfolk and Norwich | 13 | 32:54 | 10:19 | 0.12 | 21 | 18:31 | 2:51 | 0.01 |
| Ipswich | 7 | 23:46 | 4:41 | 0.05 | 11 | 21:13 | 2:45 | 0.31 |
Fig. 6
Times for each subtask of the process for first fractions. Imaging times at Norfolk and Norwich (N&N) are included in the ‘beam’ times.
Fig. 7
Times for each subtask of the process for subsequent fractions. Imaging times for Cambridge step-and-shoot (Cam S&S) are averaged over the course of treatment, including fractions without imaging. Imaging times at Norfolk and Norwich (N&N) are included in the ‘beam’ times.
Staff Experience of Intensity-modulated Radiotherapy
To test the hypothesis that staff became quicker with experience, we analysed the data subdivided into whether or not 10 procedures had been completed by that member of staff. (We also collected data on whether less than five or between five and 10 procedures had been completed, but these gave groups too small, which we pooled.) Outlining was not included, as all five oncologists had previously outlined more than 10 cases. The data in Table 3 give borderline significant support to this hypothesis for the production of plans, but not for checking. Unexpectedly, the analyses of quality assurance results became slower with experience.
Table 3. Times (minutes) subdivided by whether staff have completed 10 of these procedures
| Outline | Transfer | Plan | Produce quality assurance plan | Check | Quality assurance measure | Quality assurance analyse | ||
|---|---|---|---|---|---|---|---|---|
| ≤10 | n | 4 | 28 | 33 | 23 | 19 | 16 | 12 |
| Mean | 86.3 | 87.7 | 299.5 | 106.3 | 86.1 | 67.2 | 32.5 | |
| Standard deviation | 39.4 | 48.6 | 166.2 | 67.7 | 82.3 | 30.9 | 19.2 | |
| >10 | n | 49 | 33 | 35 | 47 | 52 | 50 | 31 |
| Mean | 112.6 | 74.1 | 237.6 | 41.5 | 81.6 | 57.1 | 53.2 | |
| Standard deviation | 60.9 | 60.1 | 179.4 | 34.8 | 89.2 | 51.3 | 41.5 | |
| P value | 0.29 | 0.33 | 0.14 | 0.00 | 0.85 | 0.35 | 0.03 |
Discussion
Process Times
Figure 3 shows that in all departments the median time from outlining to treatment was between 2 and 3 weeks, and in all but one department the upper decile was also below 3 weeks. A possible reason for this is that we are all trying to avoid breaches of the 31 day target for the time from decision to treatment. Because the stages of the process before outlining, which include mould room and computed tomography imaging, will usually take at least 1 week, and often 2 weeks, we are usually left with only 3 weeks for the rest of the planning process. Where other delays have occurred before the outlining stage, this can be even less; in these cases, as shown in the lower deciles in Fig. 3, it is possible to go through the planning process in just over a week. However, to do this for all patients would require considerable excess capacity to deal with peaks and troughs of demand [9].
Clinician Outlining
On average, the oncologist time required to outline head and neck IMRT was 108min. The time taken to outline target volumes varied between oncologists, even between two at the same centre. Owing to the small numbers, we are unable to show that there was a statistically significant difference between centres; however, the difference between two oncologists at the same centre was statistically significant. The issue of whether it is better to be fast, or whether taking longer produces better results and fewer errors, is beyond the scope of this study. When allocating resources for IMRT, the fact that some oncologists will take longer than others may need to be taken into account.
Planning, Checking and Quality Assurance
The staff time (dosimetrists or physicists) for planning and checking averaged 7.9h per patient. The one system that was significantly greater than this was the planning of step-and-shoot IMRT using CMS XiO, where the process took 15h. Staff at Addenbrooke’s are looking at processes to see how this can be shortened.
The times to produce plans for individual patients showed high standard deviations; this makes it hard to plan work efficiently, if plans are to be booked into tight schedules.
All three departments carry out patient-specific quality assurance measurements on all head and neck IMRT patient treatment plans. Figure 5 shows that for Varian and tomotherapy, the average time required on the treatment machine was 48min per patient. For step-and-shoot IMRT it took 149min; this is partly due to the fact that at the time of the study Addenbrooke’s were using a phantom that did not enable dose points and films to be measured simultaneously, hence the need to deliver the plan several times. This combined with a long beam-on time to give a long measurement time. Addenbrooke’s are working on ways to shorten this, and have recently purchased a commercial phantom of the design described by van Esch et al. [10].
Changes in the quality assurance method, introduced by Norwich in the course of this study, were shown to have a significant effect in reducing the time required on the treatment machine.
The average time to analyse the results was just under 1h. This does not get faster with staff experience.
The variety of quality assurance methods used in the three departments are in broad agreement with the recommendations of IPEM report 96 [11]. This recommends that at the early stage in a programme, measurements might include point doses in the high-dose region and organs at risk, fluence maps and combined dose distributions in multiple planes. This can later be scaled back depending on the evidence acquired from previous measurements, following the method outlined by Budgel et al. [12]. The quality assurance times reported in this study will probably reduce in the future, at all three centres.
Treatment Times
The in-room time required for a single IMRT treatment session averaged 28min 10s for the first fraction of a course of radiotherapy, and 20min 20s for subsequent fractions. Over a typical 34 fraction course of radiotherapy, this will average 20min 33s per fraction, or 2.92 fractions per hour. This is clearly less than the 4.0–4.5 fractions per hour suggested by the National Radiotherapy Advisory Group, but these recommendations were based on a standard case mix including palliative and conformal radiotherapy. For a machine dedicated to head and neck IMRT, throughput will clearly be less than the National Radiotherapy Advisory Group recommendations. However, these patients tend to take longer than the average patient to treat even without IMRT, so the average throughput of a non-IMRT machine will be increased by removing these patients, so overall throughput is not necessarily reduced.
The in-room times were very similar across all forms of IMRT studied, although step-and-shoot was slightly slower than the mean. It should be noted that the only centre using on-line volumetric daily imaging and correction (Cambridge, tomotherapy), did not have an overall treatment time greater than the mean. This is because the increased imaging and correction time was offset by a shorter beam-on time.
Staff Experience of Intensity-modulated Radiotherapy
It is generally believed that staff carrying out a new task will initially take a longer time over it, and will then get faster with experience. The data in Table 3 give borderline significant support to this hypothesis for the production of plans, but not for checking. Unexpectedly, the analyses of quality assurance results became slower with experience; there is no obvious reason for this to be so. It may be the case that quality assurance checks that are expected for some reason to be more complicated than normal are less likely to be assigned to inexperienced staff. However, we suspect that it may well be a statistical artefact; if six things with no difference are measured, the chance that none of them will show a P value<0.05 is (0.95)^6=0.74, so there is a >25% chance of one having a ‘significant’ difference even when there is no difference.
There are two possible reasons for this lack of difference. One is that the learning curve for an individual may be relatively steep, with full competency reached after only a few procedures. Another is that the increase in speed with experience applies to a department, rather than to an individual, with improvements being embedded in departmental protocols.
Comparison with Other Studies
Very few publications give timing data for head and neck IMRT. Miles et al. [3] published results in 2005 on times required for IMRT in a single institution. For head and neck cancer, the oncologists’ outlining time had a median time of 2.3h compared with a mean of 1.8h in this study. The ‘physician-related time’ in the survey of Das et al. [5] averaged 18h. This was not a timing study, but was based on responses to a questionnaire.
In Miles et al. [3], the median radiographer+physics planning time was 11.0h compared with our mean of 8.1h. Both of these are similar to the average of ‘about 11h’ reported by Das et al. [5], which is also the only study to look at times across multiple planning systems. Their results did not include tomotherapy, but did include Eclipse (9h) and XiO (8h). This suggests that the difference we observed between XiO and Eclipse may result from local practices (especially related to the checking of plans) rather than to underlying speed differences between the two systems. Murthy et al. [4] recently published planning times for 10 IMRT patients in a single department in India. The median planning time was 5.4h, using Plato as the planning system.
The median quality assurance time in Miles et al. [3] was 2.5h compared with our mean of 2.0h.
Miles et al. [3] reported treatment times of 22min for the first fraction and 11.2min for subsequent fractions, which are shorter than those reported here; however, their times were on-couch to off-couch times, whereas we report gate to gate times, which will be longer. Munter et al. [6] reported a mean beam-on time of 12.6min for step-and-shoot IMRT to the head and neck; this is similar to the 11.8min for the same technique in Fig. 7. Van de Werf et al. [7] reported mean and median times for IMRT of 15.0 and 12.0min, respectively. This did not distinguish between head and neck and other forms of IMRT, hence may have a large proportion of prostate patients making comparisons hard. Their time for a first fraction was longer than subsequent fractions by 8.0min (mean) or 9.0min (median), which agrees with our results. Bijdekerke et al. [8] reported overall treatment times averaging 21.3min for ‘head’ patients on tomotherapy; they did not distinguish between first and subsequent fractions. Murthy et al. [4] recorded a gate-to-gate time of 27.8min for step-and-shoot IMRT. These times are similar to our results.
Limitations of this Study
Most of the data for a given technique were from a single centre, with the exception of the dynamic IMRT data, which were from two centres. Ideally this study should be extended to include several other centres at each technique, to rule out whether issues specific to one department cause problems. These data are a snapshot of current practice at the three centres; developments in quality assurance and in treatment techniques, such as RapidArc or Volumetric Modulated Arc Therapy (VMAT), will need similar studies to be carried out. Similarly, the only forms of image-guided radiotherapy included in this study were planar imaging and MV CT. The effect of other forms of image-guided radiotherapy on the time required also needs to be studied. Time is just one of the resources required for IMRT; other resources need to be considered, such as equipment, training and commissioning. Quantification of these issues is beyond the scope of this study. As a timing study, our results should be less susceptible to bias than a questionnaire-based study.
Conclusions
We have described the results of a multicentre timing study, giving times for the planning and treatment of IMRT of the head and neck. The times will be useful for those making business cases for the resources that may be required when setting up or extending an IMRT service. A linac can treat just under three fractions an hour of head and neck IMRT The total planning, checking and quality assurance time averages 10.1h of physics/planning staff time per patient. When calculating staffing needs for an IMRT service, it should be borne in mind that these figures are only the increment per patient for an established service, and do not include the considerable staff time in setting up a service and carrying out system-specific quality assurance checks.
Acknowledgements
A study of this nature requires input from a very large number of people recording the time they take on each step of the process. The following is an incomplete list of those who contributed in this way: Outlining: Tom Roques, Craig Martin, Sarah Jefferies, Richard Benson, Chris Scrase. Planning, checking and quality assurance: many physicists and dosimetrists, too numerous to list. On-treatment timings: many radiographers, too numerous to list. Special thanks to June Dean and Stuart Archer for co-ordinating data from Cambridge, and Ros Perry for co-ordinating data returns from Ipswich. Help with transcribing from written forms to Excel: Katie Eyre, Janet Backer.
References
- Staffurth JN on behalf of the Radiotherapy Development Board. Clinical evidence for IMRT. Clin Oncol, 2010;22(8):643–657.
- . Radiotherapy: developing a world class service for England. London: Department of Health; 2007;
- The impact of introducing intensity modulated radiotherapy into routine clinical practice. Radiother Oncol. 2005;77(3):241–246
- Time trial: a prospective comparative study of the time-resource burden for three-dimensional conformal radiotherapy and intensity modulated radiotherapy in head and neck cancers. J Cancer Res Ther. 2009;5(2):107–112
- . Analysis of treatment planning time among systems and planners for IMRT. J Am Coll Radiol. 2009;6(7):514–517
- Stereotactic intensity modulated radiation therapy and inverse planning for tumours of the head and neck region: clinical implementation of the step and shoot method and first clinical results. Radiother Oncol. 2003;66:313–321
- . Time and motion study of radiotherapy delivery: economic burden of increased quality assurance and IMRT. Radiother Oncol. 2009;93:137–140
- TomoTherapy: implications on daily workload and scheduling patients. Radiother Oncol. 2008;86:224–230
- . How much surplus capacity is required to maintain low waiting times?. Clin Oncol. 2001;13:24–28
- . On-line quality assurance of rotational radiotherapy treatment delivery by means of a 2D ion chamber array and the Octavius phantom. Med Phys. 2007;34(10):3825–3837
- Guidance for the Clinical Implementation of Intensity Modulated Radiation Therapy. Institute of Physics and Engineering in Medicine Report 96. York: IPEM; 2008;
- . Quantitative analysis of patient-specific dosimetric IMRT verification. Phys Med Biol. 2005;50:103–119
PII: S0936-6555(10)00208-6
doi:10.1016/j.clon.2010.06.011
© 2010 The Royal College of Radiologists. Published by Elsevier Inc. All rights reserved.
