- Email: [email protected]
Cost Minimisation Analysis: Kilovoltage Imaging with Automated Repositioning Versus Electronic Portal Imaging in Image-guided Radiotherapy for Prostate Cancer
Clinical Oncology 24 (2012) e93ee99 Contents lists available at SciVerse ScienceDirect
Clinical Oncology journal homepage: www.clinicaloncologyonline.net
Original Article
Cost Minimisation Analysis: Kilovoltage Imaging with Automated Repositioning Versus Electronic Portal Imaging in Image-guided Radiotherapy for Prostate Cancer S. Gill *, S. Younie y, A. Rolfo z, J. Thomas *, S. Siva *, C. Fox x, T. Kron x, D. Phillips *, K.H. Tai *jj, F. Foroudi *jj * Department
of Radiation Oncology, Peter MacCallum Cancer Centre, Melbourne, Australia Faculty of Health, Medicine, Nursing and Behavioural Sciences, Deakin University, Melbourne, Australia z Radiation Therapy Services, Peter MacCallum Cancer Centre, Melbourne, Australia x Department of Physical Sciences, Peter MacCallum Cancer Centre, Melbourne, Australia jj Department of Pathology, University of Melbourne, Melbourne, Australia y
Received 10 January 2011; received in revised form 23 March 2012; accepted 20 April 2012
Abstract Aims: To compare the treatment time and cost of prostate cancer fiducial marker image-guided radiotherapy (IGRT) using orthogonal kilovoltage imaging (KVI) and automated couch shifts and orthogonal electronic portal imaging (EPI) and manual couch shifts. Materials and methods: IGRT treatment delivery times were recorded automatically on either unit. Costing was calculated from real costs derived from the implementation of a new radiotherapy centre. To derive cost per minute for EPI and KVI units the total annual setting up and running costs were divided by the total annual working time. The cost per IGRT fraction was calculated by multiplying the cost per minute by the duration of treatment. A sensitivity analysis was conducted to test the robustness of our analysis. Treatment times without couch shift were compared. Results: Time data were analysed for 8648 fractions, 6057 from KVI treatment and 2591 from EPI treatment from a total of 294 patients. The median time for KVI treatment was 6.0 min (interquartile range 5.1e7.4 min) and for EPI treatment it was 10.0 min (interquartile range 8.3e11.8 min) (P value < 0.0001). The cost per fraction for KVI was A$258.79 and for EPI was A$345.50. The cost saving per fraction for KVI varied between A$66.09 and A$101.64 by sensitivity analysis. In patients where no couch shift was made, the median treatment delivery time for EPI was 8.8 min and for KVI was 5.1 min. Conclusions: Treatment time is less on KVI units compared with EPI units. This is probably due to automation of couch shift and faster evaluation of imaging on KVI units. Annual running costs greatly outweigh initial setting up costs and therefore the cost per fraction was less with KVI, despite higher initial costs. The selection of appropriate IGRT equipment can make IGRT practical within radiotherapy departments. Crown Copyright Ó 2012 Published by Elsevier Ltd on behalf of The Royal College of Radiologists. All rights reserved. Key words: Cost minimisation analysis; fiducial markers; IGRT; prostate cancer; treatment time
Introduction Dose escalation in prostate cancer reduces biochemical failure [1]. Image-guided radiotherapy (IGRT) may allow safe dose escalation by avoidance of inadvertent irradiation of normal tissue [2,3]. There are numerous methods for conducting IGRT in prostate cancer [4,5]. In our experience, IGRT using implanted fiducial markers and orthogonal imaging is emerging as a practical, reliable and effective method to ensure target localisation before each fraction of radiotherapy [6,7]. Author for correspondence: S. Gill, Division of Radiation Oncology, Peter MacCallum Cancer Centre, Locked Bag 1, A’Beckett Street, Victoria 8006, Australia. Tel: þ61-396561111; Fax: þ61-396561424. E-mail address: [email protected] (S. Gill).
Previous studies have shown that online prostate fiducial marker IGRT with orthogonal electronic portal imaging (EPI) takes longer to deliver than conventional prostate radiotherapy [8]. For many departments, the longer treatment time with IGRT may affect waiting lists. Technology that reduces treatment time could make IGRT more practical for routine implementation in busy radiotherapy departments. Several recent linear accelerator engineering innovations may allow quicker IGRT treatment. Gold seed fiducial markers show up with more clarity on kilovoltage imaging (KVI) compared with megavoltage imaging (MVI; see Figure 1) [9e11]. Another recently introduced technology is automated couch repositioning, which refers to the adjustment of the treatment couch made directly from the treatment console without the radiation therapist having to re-enter the treatment room [12].
0936-6555/$36.00 Crown Copyright Ó 2012 Published by Elsevier Ltd on behalf of The Royal College of Radiologists. All rights reserved. doi:10.1016/j.clon.2012.04.004
e94
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
Fig 1. Comparison of electronic portal imaging (EPI) (on the left) and kilovoltage imaging (KVI) in a prostate cancer patient with three gold seed fiducial markers implanted in the prostate gland illustrating the difference in image quality and visualisation of the fiducial markers.
A 2007 survey of UK National Health Service radiotherapy centres showed that nearly half the centres surveyed did not plan on implementing IGRT mostly due to cost issues [13]. The purpose of this study was to compare IGRT treatment time using EPI with manual shift and KVI with automated shift to determine if costs can be saved by using new technology in IGRT through any potential reduction in treatment time. The purchase costs of new technology are greater, and it is unknown if any potential reduction in treatment time could offset this. To explore this, a cost minimisation analysis was conducted, whereby we assumed that the two units have equivalent clinical outcomes and only financial costs were explored [14].
Materials and Methods IGRT was introduced in 2007 to a functioning radiotherapy centre that had six EPI and four KVI units installed.
Patients were assigned to either unit based on availability. Details of the patient group, gold seed implantation procedure, simulation and planning and quality assurance processes have been published previously [6]. Between March 2007 and January 2009 treatment delivery time data were collected prospectively from 294 prostate cancer patients treated with IGRT, of which 6% were treated with intensity-modulated radiotherapy and 94% with conformal radiotherapy. All patients were treated on Varian linear accelerators (Varian Medical Systems, Palo Alto, USA). The protocol in our department is to treat all radical IGRT prostate patients to a total dose 78 Gy in 39 fractions 5 days per week. Image registration was carried out using Impac Software (IMPAC Medical Systems Inc., Sunnyvale, CA, USA) on the EPI linear accelerators, and directly on the treatment console on KVI linear accelerators. On both units, only translational shifts were corrected, there was no rotational or tilt correction. On EPI, after manual couch shift but before the fraction was delivered, a repeat set of verification images
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
was taken to check that the shift was made in the right direction. Automated shifts on the KVI console were not verified with a second set of imaging, as shifts were made by computer. After the fraction was delivered, orthogonal imaging was carried out on all patients on EPI and KVI to assess interfraction motion. Due to the increased treatment time associated with manual couch shifts and re-verification on the EPI units, a couch shift would only be made if the fiducial marker displacement was equal to or more than 3 mm. On KVI units, the action threshold was 0 mm. Treatment delivery time was collected automatically. As part of IGRT, all patients had a set of images performed pre-treatment. As part of a separate study looking at intrafraction prostate motion, all patients had a second set of verification images at the end of treatment [15]. Time stamps were recorded by the ImpacÔ record and verify system at the time of the first and last verification image. On EPI units, the post-treatment image was taken during the last treatment field using a few monitor units from the treatment beam. On KVI, post-treatment images were taken after the completion of treatment. We estimate that the actual treatment will probably be up to 30 s longer for EPI and up to 30 s shorter for KVI because of the difference in the way we have measured treatment delivery time on the two units. Set-up time was estimated in 10 patients using manually operated digital stopwatches from the time the patient was called in until the first pre-treatment image was taken. There is an extra component for quality assurance at the start of the day to check for accuracy of couch shift on the KVI machine with automated couch shift. The radiation therapist would image a phantom, make a couch shift and reimage the phantom. The phantom has radio-opaque markers that are visible on KVI, and the distance between markers confirms that the amount of couch shift and the direction are correct. However, at the start of the day, about half an hour is allocated to daily quality assurance on the EPI unit, and half an hour is allocated for the KVI unit, and therefore there is no extra time lost in quality assurance of the KVI units compared with the standard daily quality assurance on EPI units. Data Analysis and Statistical Methods Reports were created from the ImpacÔ database using Crystal Reports software (San Jose, CA, USA). The reports were processed using in-house software onto a Microsoft Access Database [16]. The estimated time taken to treat an individual patient was calculated by adding the set-up time to the treatment delivery time. When analysing the data, to remove aberrant data we only included treatment delivery times between 3 and 30 min. Time data were logged to better meet the assumptions of the statistical technique. A two-sided z-test from a mixed effects model with random intercepts for patients was used to compare the treatment delivery times. Costs Costs were estimated from the viewpoint of the health service provider and only costs incurred in the delivery of the
e95
service were included. All costs are using 2010 Australian dollar (A$) estimates. Indicative prices of the KVI and EPI linear accelerators and planning stations were provided by Varian in US dollars and were converted to Australian dollars at the exchange rate of A$1 ¼ US$ 0.90 (average January to October 2010). Institutional costs were grouped into capital costs representing a one-off investment and recurrent costs based on a two linear accelerator department model. Building costs were taken from that of a new radiotherapy centre being set up, with about 2500 m2 of work area þ 30% circulation area. The figure quoted includes fixtures and fittings. This new centre has four radiotherapy bunkers, two of which will be kept for future expansion. Land value was not taken into consideration. Four planning stations were thought to be required as a minimum for a two linear accelerator department. The cost of purchasing the automated couch is included in the total cost of the KVI linear accelerator. Capital costs are shown in Table 1. Recurrent costs were obtained from an internal analysis benchmarked across a number of sites and services operated by our organisation. The full list of recurrent costs taken into account is shown in Table 2. Spare parts per annum cost more on the KVI than the EPI units and average figures were obtained from our engineering department. The cost per patient of implantation of three gold seeds for IGRT was estimated at A$130.70, although this cost was not added to the cost per fraction, as this cost was shared as a standard component of both imaging arms. The lifetime of equipment was taken from federal government policy to fund for replacement after the stated period. Where no recommendation was available from the federal government, lifetimes of equipment were estimated by our physicist. For simplification, we assumed that at the Table 1 Capital costs (A$) in 2010 dollars Purchase price
Lifetime
Building 31,000,000 40 5 years Varian EPI iX 2,222,222 10 years 120 MLC Varian On-Board 2,777,778 10 years Imager iX 120 MLC Varian Eclipse 188,889 5 years planning station Three further 216,667 5 years planning stations Computed tomography 818,820 5 years simulator Record and verify 550,000 5 years system Computers, servers 120,000 5 years and software Annual capital cost for a two EPI centre Annual capital cost for a two KVI centre
Estimated cost per annum 2,060,307 301,928 377,410 44,841 51,436 194,383 130,567 28,487 A$3,113,877 A$3,264,841
EPI, electronic portal imaging; MLC, multileaf collimator; KVI, kilovoltage imaging.
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
A$4,175,948 A$4,217,948
PSA, prostate-specific antigen; IT, information technology; EPI, electronic portal imaging; KVI, kilovoltage imaging.
5
end of its lifetime, capital would have been amortised. Capital costs were converted into equivalent annual cost by dividing the net present value by the present value of annuity of $1 in arrears [14]. The discount rate was taken as per the 10 year Australian Government Bond rate (risk free rate), which is presently 6%. Total linear accelerator work time was calculated based on the number of workdays per annum minus machine down time in days, from figures that were obtained from our engineering department. The cost minimisation analysis for this study was carried out similarly to that described by other authors [17]. A multiway sensitivity analysis was carried out by varying the assumptions made about cost, lifetime and treatment time to work out the maximal and minimal costs of both EPI and KVI, such that the difference in cost between the two groups was the smallest or largest possible. For the sensitivity analysis, capital costs, spare parts costs and lifetime were varied by 10%, wages were varied by 5%, and service costs by 2.5%, reflecting possible fluctuations in cost. Time was varied between the interquartile ranges (IQR) for treatment time for both units. Because the IQR was wide, this was also assumed to incorporate any errors in time measurement, including the difference in method of time measurement as described above. The discount rate was varied from 5 to 7%.
25
120,000 275,563 123,888 1,243,784 258,103 275,000 55,000 78,301 50,000 66,000 48,667 5000 26,000 70,000 46,000 75,000 30,000 7000
20
Annual recurrent costs for a two EPI centre Annual recurrent costs for a two KVI centre
1,343,642
15
Medical director (1), radiation oncologists (3), registrar (1) Deputy site director (1) Physicists (3) Engineers (1.5) Radiation therapists (8 senior, 10 base grade) Nurses (4) Receptionist (5) Office manager (1) PSA (2) IT (0.5) Service contract for record and verify system Service contract for treatment planning system Spare parts for EPI (2 units) Spare parts for KVI (2 units) Building and maintenance Utilities Cleaning Linen Waste removal
lateral images were taken after treatment. Thus, treatment delivery times were available for 75% of the total fractions for these patients. The difference in action threshold resulted in a couch shift being made about 98% of the time on KVI linear accelerators and 50% of the time on EPI linear accelerators. The median time for KVI treatment was 6.0 min (IQR 5.1e7.4 min), and for EPI treatment it was 10.0 min (IQR 8.3e11.8 min) (P value < 0.0001) (see Figure 1). The average treatment delivery time for KVI is between 35 and 40% (point estimate 38%) shorter than that for EPI. The distribution of treatment delivery times taken is shown in Figure 2. In patients where no couch shift was made, the median treatment delivery time for EPI was 8.75 min and the median time for KVI was 5.13 min. For this subgroup analysis we analysed 1213 fractions where no couch shift was made (1124 EPI, 89 KVI) from 154 patients (88 EPI, 66 KVI). We found that using KVI over EPI reduced treatment time by 39% (P < 0.0001 95% confidence interval 35e43%), i.e. proportionally similar to the time reduction in the whole group. The mean set-up time for 10 prostate IGRT patients was 4.8 min (range 3.0e6.2 min; standard deviation 59 s). With the addition of set-up time and treatment delivery time, the IQR for the total time taken to treat varied from 13.1 to 16.6 min per fraction for EPI and 9.9 to 12.2 min per fraction for KVI. These figures were used for the sensitivity analysis shown below. Of six EPI units at our centre, the average downtime per year for two EPI units was 7.4 days (range 2e11.4). Of four KVI units at our centre, the average downtime per year for KVI was 5.4 days (range 1.8e8). The difference was not significant and for the purposes of our calculations we will assume both EPI and KVI units have an approximately equal total breakdown time of 6.4 days. The total number of treatment minutes per day refers to the total number of minutes that each linear accelerator is actually being used to physically treat patients, and excludes machine warm up time. The typical patient appointments start at 8 am and run until 5.45 pm or about
Percent of Group
Table 2 Recurrent costs (A$) in 2010 dollars
10
e96
Time data were available for 8863 fractions. After the removal of aberrant data, time data were analysed for 8648 fractions, 6057 from KVI treatment and 2591 from EPI treatment from a total of 294 patients. The data extraction programme only included data if both anterioreposterior and
0
Results 3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Treatment Time (mins)
Fig 2. Treatment delivery time for electronic portal imaging (EPI; solid square) and kilovoltage imaging (KVI; shaded square).
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
585 min per day. For two linear accelerators, the total number of treatment minutes is 1170. Our centre runs Monday to Friday and also for 4 h each Saturday morning (0.4 of a day). The total number of treatment minutes per year ¼ [weekdays (261) public holidays (9) þ weekends worked (21.3) downtime (6.4)] total minutes patients treated per day (1170) ¼ 312 273 min of treatment time per year in total for a two linear accelerator department. For Kilovoltage Imaging Total annual costs for a two KVI linear accelerator cancer centre ¼ $7,482,789 Estimated cost per minute¼$7,482,789 O 312,273 ¼ $23.96 Estimated cost per fraction with KVI ¼ $23.96 (6 þ 4.8) ¼ $258.79 Estimated cost of a 39 fraction course of IGRT ¼ $10,092.93
For Electronic Portal Imaging Total annual costs for a two EPI linear accelerator cancer centre ¼ $7,289,825 Estimated cost per minute ¼ $7,289,825 O 312,273 ¼ $23.34 Estimated cost per fraction with EPI ¼ $23.34 (10 þ 4.8) ¼ $345.50 Estimated cost of a 39 fraction course of IGRT ¼ $13,474.39 The cost per fraction is A$86.71 or 25% less for KVI compared with EPI. The incremental cost difference for a 39 fraction course is A$3381. In this study, wage accounted for about 52% of the total department cost per annum, and equipment (excluding building cost) amounted to around 15% of the total department cost per annum. Table 3 shows the results of the sensitivity analysis. In the absence of statistical testing, a sensitivity analysis is an effective way of testing the robustness of this analysis. For a given time period, the discount rate would be identical for either unit and therefore the same discount rates are compared for both units. The analysis showed that the difference in cost per fraction for the cheapest and most expensive EPI and KVI units ranged from A$66.09 to A$101.64.
e97
Discussion The use of KVI with automated couch shift resulted in a calculated cost saving compared with the use of EPI with manual couch shift, due to a reduction in required treatment time. Figures from a typical two linear accelerator satellite centre (Peter MacCallum, Box Hill from 2007 to 2009) show that 14.5% of all radiotherapy fractions per year were delivered to treat prostate cancer patients with IGRT. The proportion of prostate cancer patients treated with radical radiotherapy is similar to that reported by a UK department [8]. In that study, the time taken to treat six patients with EPI IGRT was compared with matched controls treated with conformal radiotherapy without IGRT. IGRT was found to take 3.1 min longer than conformal radiotherapy without IGRT (13.1 min versus 10.0 min). It was estimated that to treat all prostate cancer patients radically with IGRT would increase machine workload by 2.2 h per day. This study shows that by using KVI with automated couch shift, treatment time is reduced by a median of 4 min compared with EPI. The selection of appropriate IGRT equipment can make routine IGRT practical within busy radiotherapy departments. The total number of couch shifts was 48% more for KVI compared with EPI (98% versus 50%). This was because of the difference in the action threshold, which was 0 mm for KVI and 3 mm for EPI. A study from the Princess Margaret Hospital looking at EPI IGRT for prostate cancer showed that treatment delivery times were in the order of 8.7 min for patients requiring isocentre adjustment and 6.1 min for those who did not [18]. The author stated that the action level of 3 mm was specified for EPI treatment at the Princess Margaret Hospital because it was felt that an action level of 3 mm would be more practical within the time frame of a standard appointment. In this study, despite the difference in action threshold, treatment time was still less on the KVI units compared with EPI units. Automated couch shift probably contributes significantly to reducing treatment time. However, the difference between automated couch shift and KVI results in a median 4 min faster treatment (10 to 6 min) compared with EPI units. In patients who did not have couch shifts, treatment delivery time was faster on KVI units by a median of 3.62 min (8.75 to 5.13) compared with EPI units. Both these figures lead us to deduce that the time saved by using automate couch correction is in the vicinity of only 0.38 min
Table 3 Sensitivity analysis. Least costly scenario (discount rate 5%) Capital þ recurrent cost per annum (A$) Cost per minute (A$) Duration per fraction IQR þ set-up time (4.8 min) Cost per fraction (A$) Cost difference (EPI e KVI) (A$)
Most costly scenario (discount rate 7%)
EPI
KVI
EPI
KVI
6,929,706 22.19 13.1 min 290.70
7,084,987 22.69 9.9 min 224.62
7,715,008 24.71 16.6 min 410.12
7,895,966 25.29 12.2 min 308.48
66.09
EPI, electronic portal imaging; KVI, kilovoltage imaging; IQR, interquartile range.
101.64
e98
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
(4.00 to 3.62), leading us to conclude that other factors are responsible for the faster workflow on KVI. The possible other factors resulting in a faster workflow include faster interpretation of imaging and easier use of the software used for matching (IMPAC for EPI, Varian console for KVI). As both KV and MV linacs have a similar monitor unit output rate, presumably treatment time on KVI units was less because of faster detection of fiducial markers by radiation therapy staff with KVI compared with MVI. Reducing treatment delivery time may hold additional clinical benefit as it seems that intrafraction prostate displacement, i.e. the probability of prostate displacement larger than a given distance, increases with increasing duration of treatment [15,19]. Therefore, it should be a goal to reduce treatment time as much as possible, as a longer treatment not only leads to an increased cost of treatment, but may also compromise coverage of the target due to intrafraction motion if inadequate clinical target volume to planning target volume margins are used. Intrafraction motion should be accounted for when determining appropriate clinical target volume to planning target volume margins, and reducing treatment time may have an effect on margins. Although we have not done this in our study, one method to further reduce treatment time may be to use a mix of KVI from the lateral direction, and MVI from the anterior, which will result in less time taken to rotate the gantry, thereby reducing the time taken to image the patient. However, the time for gantry rotation (about 15 seconds) is not significantly longer than the time taken to retract the KV arm before MVI. The limitations of this study should be discussed. Spare parts data in this study were taken as estimates from our engineering department. Over the last 10 years, our engineering department has replaced only two MVI panels. Since installation, no KVI panels have needed replacing. The EPI panel on a KVI linac is seldom used and therefore has a longer life expectancy. Although these were real figures obtained from our engineering department, other departments may have a different experience. The approximate cost of an MVI and a KVI panel are A$70 000 and A$150 000, respectively. However, for the first year the panels are covered by warranty. Therefore, the costing figures should be considered as the minimum, and may increase if wear and tear of equipment is greater. Delaney et al. [20] analysed treatment time for 4316 radiotherapy fractions and found 17 factors that influenced treatment time significantly. Slight differences between KVI and EPI have not been considered in detail, for example the quality assurance time for both have been considered the same. However, KVI quality assurance includes a component for cone beam computed tomography that is hard to separate out. There were more patients treated on KVI than EPI, which could imply that KVI radiation therapists were more experienced in prostate IGRT, although we think this is unlikely, as prostate cancer radiotherapy is very common, and it would be unusual for radiation therapists on either unit not to be regularly exposed to IGRT. A limitation of the cost minimisation analysis is that we have not considered any potential clinical benefits of one imaging modality over another. A cost minimisation analysis has the considerable appeal of keeping the comparison simple, which may be appropriate for
comparing new radiotherapy equipment as late effects and survival data take many years to be borne out and are probably similar between KVI and MVI. This model gives us a measurable input (resources used) and output, i.e. course of radiotherapy. The cost of all steps in radiotherapy treatment, such as simulation, contouring and planning, are accounted for by adding up their set-up and running costs. On the whole, cancer care worldwide is challenged by resource issues [21]. Similar to other studies, staffing costs in our study accounted for more than three times the annual cost compared with equipment costs (52% versus 15%) [22]. Because running costs outweigh the initial purchase costs, a time saving technology can result in an overall cost saving.
Conclusion There are several reasons why treatment time reduction is preferable in radiotherapy, i.e. it is more convenient for the patient, it results in a cost saving for the department, and there is a theoretical reduction in geographical miss of the tumour due to a reduction in intrafraction motion. The finding in this study is not unexpected, as we have two new time saving technologies being assessed, and we would expect to find that KVI with automated couch shift is slightly quicker than MVI with manual couch shift. This time analysis was conducted to quantify the difference, and it was unexpected that we found that patients who did not have a couch shift at all also had a treatment time reduction by 3.62 min, which leads us to conclude that other factors in the workflow, such as interpretation of imaging, is quicker on the KVI units. Many centres still use MVI for IGRT, and manual couch shift, because of cost issues. This study supports the use of new technology in IGRT equipment, which could result in an overall cost saving through faster treatment time.
Acknowledgments Presented in part at the Annual Scientific Meeting of Royal Australian and New Zealand College of Radiologists, Adelaide, Australia 2008. Peter MacCallum Cancer Centre has a research agreement with Varian Medical Systems. The authors thank Brad Shilling for engineering cost figures.
References [1] Viani GA, Stefano EJ, Afonso SL. Higher-than-conventional radiation doses in localized prostate cancer treatment: a meta-analysis of randomized, controlled trials. Int J Radiat Oncol Biol Phys 2009;74:1405e1418. [2] Gauthier I, Carrier JF, Beliveau-Nadeau D, Fortin B, Taussky D. Dosimetric impact and theoretical clinical benefits of fiducial markers for dose escalated prostate cancer radiation treatment. Int J Radiat Oncol Biol Phys 2009;74:1128e1133. [3] Jarad M, Andrew B, Robert B, et al. Image guided dose escalated prostate radiotherapy: still room to improve. Radiat Oncol 2009;4. doi:10.1186/1748-717X-4-50. [4] Kupelian PA, Langen KM, Willoughby TR, Zeidan OA, Meeks SL. Image-guided radiotherapy for localized prostate cancer: treating a moving target. Semin Radiat Oncol 2008;18(1):58e66.
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99 [5] Button MR, Staffurth JN. Clinical application of image-guided radiotherapy in bladder and prostate cancer. Clin Oncol (R Coll Radiol) 2010;22:698e706. [6] Thompson A, Fox C, Foroudi F, et al. Planning and implementing an implanted fiducial programme for prostate cancer radiation therapy. J Med Imaging Radiat Oncol 2008;52:419e424. [7] Gill S, Thomas J, Fox C, et al. Acute toxicity in prostate cancer patients treated with and without image-guided radiotherapy. Radiat Oncol 2011;6:145. [8] Rimmer YL, Burnet NG, Routsis DS, et al. Practical issues in the implementation of image-guided radiotherapy for the treatment of prostate cancer within a UK department. Clin Oncol (R Coll Radiol) 2008;20:22e30. [9] Jaffray DA, Chawla K, Yu C, Wong JW. Dual-beam imaging for online verification of radiotherapy field placement. Int J Radiat Oncol Biol Phys 1995;33:1273e1280. [10] Pisani L, Lockman D, Jaffray D, Yan D, Martinez A, Wong J. Setup error in radiotherapy: on-line correction using electronic kilovoltage and megavoltage radiographs. Int J Radiat Oncol Biol Phys 2000;47:825. [11] Gill S, Thomas J, Fox C, et al. Electronic portal imaging vs kilovoltage imaging in fiducial marker image guided radiotherapy for prostate cancer: an analysis of set-up uncertainties. Br J Radiol 2012;85:176e182. [12] Bel A, Petrascu O, Van de Vondel I, et al. A computerized remote table control for fast on-line patient repositioning: implementation and clinical feasibility. Med Phys 2000;27:354. [13] Jefferies S, Taylor A, Reznek R. Results of a national survey of radiotherapy planning and delivery in the UK in 2007. Clin Oncol (R Coll Radiol) 2009;21:204e217.
e99
[14] Drummond MF, Sculpher MJ, Torrance GW, O’Brien BJ, Stoddart GL. Methods for the economic evaluation of health care programmes. New York: Oxford University Press; 2005. [15] Kron T, Thomas J, Fox C, et al. Intra-fraction prostate displacement in radiotherapy estimated from pre- and posttreatment imaging of patients with implanted fiducial markers. Radiother Oncol 2010;95:191e197. [16] Fox C, Fisher R, Kron T, et al. Extraction of data for margin calculations in prostate radiotherapy from a commercial record and verify system. J Med Imaging Radiat Oncol 2010;54:161e170. [17] Foroudi F, Lapsley H, Manderson C, Yeghiaian-Alvandi R. Cost-minimization analysis: radiation treatment with and without a multi-leaf collimator. Int J Radiat Oncol Biol Phys 2000;47:1443e1448. [18] Chung PWM, Haycocks T, Brown T, et al. On-line aSi portal imaging of implanted fiducial markers for the reduction of interfraction error during conformal radiotherapy of prostate carcinoma. Int J Radiat Oncol Biol Phys 2004;60:329e334. [19] Langen KM, Willoughby TR, Meeks SL, et al. Observations on real-time prostate gland motion using electromagnetic tracking. Int J Radiat Oncol Biol Phys 2008;71:1084e1090. [20] Delaney GP, Shafiq RJ, Jalaludin BB, Barton MB. The development of a new basic treatment equivalent model to assess linear accelerator throughput. Clin Oncol (R Coll Radiol) 2005;17:311e318. [21] Baume P. cROI. A vision for radiotherapy. Canberra: Department of Health; 2002. [22] Lievens Y, Van den Bogaert W, Kesteloot K. Activity-based costing: a practical model for cost calculation in radiotherapy. Int J Radiat Oncol Biol Phys 2003;57:522e535.
Clinical Oncology journal homepage: www.clinicaloncologyonline.net
Original Article
Cost Minimisation Analysis: Kilovoltage Imaging with Automated Repositioning Versus Electronic Portal Imaging in Image-guided Radiotherapy for Prostate Cancer S. Gill *, S. Younie y, A. Rolfo z, J. Thomas *, S. Siva *, C. Fox x, T. Kron x, D. Phillips *, K.H. Tai *jj, F. Foroudi *jj * Department
of Radiation Oncology, Peter MacCallum Cancer Centre, Melbourne, Australia Faculty of Health, Medicine, Nursing and Behavioural Sciences, Deakin University, Melbourne, Australia z Radiation Therapy Services, Peter MacCallum Cancer Centre, Melbourne, Australia x Department of Physical Sciences, Peter MacCallum Cancer Centre, Melbourne, Australia jj Department of Pathology, University of Melbourne, Melbourne, Australia y
Received 10 January 2011; received in revised form 23 March 2012; accepted 20 April 2012
Abstract Aims: To compare the treatment time and cost of prostate cancer fiducial marker image-guided radiotherapy (IGRT) using orthogonal kilovoltage imaging (KVI) and automated couch shifts and orthogonal electronic portal imaging (EPI) and manual couch shifts. Materials and methods: IGRT treatment delivery times were recorded automatically on either unit. Costing was calculated from real costs derived from the implementation of a new radiotherapy centre. To derive cost per minute for EPI and KVI units the total annual setting up and running costs were divided by the total annual working time. The cost per IGRT fraction was calculated by multiplying the cost per minute by the duration of treatment. A sensitivity analysis was conducted to test the robustness of our analysis. Treatment times without couch shift were compared. Results: Time data were analysed for 8648 fractions, 6057 from KVI treatment and 2591 from EPI treatment from a total of 294 patients. The median time for KVI treatment was 6.0 min (interquartile range 5.1e7.4 min) and for EPI treatment it was 10.0 min (interquartile range 8.3e11.8 min) (P value < 0.0001). The cost per fraction for KVI was A$258.79 and for EPI was A$345.50. The cost saving per fraction for KVI varied between A$66.09 and A$101.64 by sensitivity analysis. In patients where no couch shift was made, the median treatment delivery time for EPI was 8.8 min and for KVI was 5.1 min. Conclusions: Treatment time is less on KVI units compared with EPI units. This is probably due to automation of couch shift and faster evaluation of imaging on KVI units. Annual running costs greatly outweigh initial setting up costs and therefore the cost per fraction was less with KVI, despite higher initial costs. The selection of appropriate IGRT equipment can make IGRT practical within radiotherapy departments. Crown Copyright Ó 2012 Published by Elsevier Ltd on behalf of The Royal College of Radiologists. All rights reserved. Key words: Cost minimisation analysis; fiducial markers; IGRT; prostate cancer; treatment time
Introduction Dose escalation in prostate cancer reduces biochemical failure [1]. Image-guided radiotherapy (IGRT) may allow safe dose escalation by avoidance of inadvertent irradiation of normal tissue [2,3]. There are numerous methods for conducting IGRT in prostate cancer [4,5]. In our experience, IGRT using implanted fiducial markers and orthogonal imaging is emerging as a practical, reliable and effective method to ensure target localisation before each fraction of radiotherapy [6,7]. Author for correspondence: S. Gill, Division of Radiation Oncology, Peter MacCallum Cancer Centre, Locked Bag 1, A’Beckett Street, Victoria 8006, Australia. Tel: þ61-396561111; Fax: þ61-396561424. E-mail address: [email protected] (S. Gill).
Previous studies have shown that online prostate fiducial marker IGRT with orthogonal electronic portal imaging (EPI) takes longer to deliver than conventional prostate radiotherapy [8]. For many departments, the longer treatment time with IGRT may affect waiting lists. Technology that reduces treatment time could make IGRT more practical for routine implementation in busy radiotherapy departments. Several recent linear accelerator engineering innovations may allow quicker IGRT treatment. Gold seed fiducial markers show up with more clarity on kilovoltage imaging (KVI) compared with megavoltage imaging (MVI; see Figure 1) [9e11]. Another recently introduced technology is automated couch repositioning, which refers to the adjustment of the treatment couch made directly from the treatment console without the radiation therapist having to re-enter the treatment room [12].
0936-6555/$36.00 Crown Copyright Ó 2012 Published by Elsevier Ltd on behalf of The Royal College of Radiologists. All rights reserved. doi:10.1016/j.clon.2012.04.004
e94
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
Fig 1. Comparison of electronic portal imaging (EPI) (on the left) and kilovoltage imaging (KVI) in a prostate cancer patient with three gold seed fiducial markers implanted in the prostate gland illustrating the difference in image quality and visualisation of the fiducial markers.
A 2007 survey of UK National Health Service radiotherapy centres showed that nearly half the centres surveyed did not plan on implementing IGRT mostly due to cost issues [13]. The purpose of this study was to compare IGRT treatment time using EPI with manual shift and KVI with automated shift to determine if costs can be saved by using new technology in IGRT through any potential reduction in treatment time. The purchase costs of new technology are greater, and it is unknown if any potential reduction in treatment time could offset this. To explore this, a cost minimisation analysis was conducted, whereby we assumed that the two units have equivalent clinical outcomes and only financial costs were explored [14].
Materials and Methods IGRT was introduced in 2007 to a functioning radiotherapy centre that had six EPI and four KVI units installed.
Patients were assigned to either unit based on availability. Details of the patient group, gold seed implantation procedure, simulation and planning and quality assurance processes have been published previously [6]. Between March 2007 and January 2009 treatment delivery time data were collected prospectively from 294 prostate cancer patients treated with IGRT, of which 6% were treated with intensity-modulated radiotherapy and 94% with conformal radiotherapy. All patients were treated on Varian linear accelerators (Varian Medical Systems, Palo Alto, USA). The protocol in our department is to treat all radical IGRT prostate patients to a total dose 78 Gy in 39 fractions 5 days per week. Image registration was carried out using Impac Software (IMPAC Medical Systems Inc., Sunnyvale, CA, USA) on the EPI linear accelerators, and directly on the treatment console on KVI linear accelerators. On both units, only translational shifts were corrected, there was no rotational or tilt correction. On EPI, after manual couch shift but before the fraction was delivered, a repeat set of verification images
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
was taken to check that the shift was made in the right direction. Automated shifts on the KVI console were not verified with a second set of imaging, as shifts were made by computer. After the fraction was delivered, orthogonal imaging was carried out on all patients on EPI and KVI to assess interfraction motion. Due to the increased treatment time associated with manual couch shifts and re-verification on the EPI units, a couch shift would only be made if the fiducial marker displacement was equal to or more than 3 mm. On KVI units, the action threshold was 0 mm. Treatment delivery time was collected automatically. As part of IGRT, all patients had a set of images performed pre-treatment. As part of a separate study looking at intrafraction prostate motion, all patients had a second set of verification images at the end of treatment [15]. Time stamps were recorded by the ImpacÔ record and verify system at the time of the first and last verification image. On EPI units, the post-treatment image was taken during the last treatment field using a few monitor units from the treatment beam. On KVI, post-treatment images were taken after the completion of treatment. We estimate that the actual treatment will probably be up to 30 s longer for EPI and up to 30 s shorter for KVI because of the difference in the way we have measured treatment delivery time on the two units. Set-up time was estimated in 10 patients using manually operated digital stopwatches from the time the patient was called in until the first pre-treatment image was taken. There is an extra component for quality assurance at the start of the day to check for accuracy of couch shift on the KVI machine with automated couch shift. The radiation therapist would image a phantom, make a couch shift and reimage the phantom. The phantom has radio-opaque markers that are visible on KVI, and the distance between markers confirms that the amount of couch shift and the direction are correct. However, at the start of the day, about half an hour is allocated to daily quality assurance on the EPI unit, and half an hour is allocated for the KVI unit, and therefore there is no extra time lost in quality assurance of the KVI units compared with the standard daily quality assurance on EPI units. Data Analysis and Statistical Methods Reports were created from the ImpacÔ database using Crystal Reports software (San Jose, CA, USA). The reports were processed using in-house software onto a Microsoft Access Database [16]. The estimated time taken to treat an individual patient was calculated by adding the set-up time to the treatment delivery time. When analysing the data, to remove aberrant data we only included treatment delivery times between 3 and 30 min. Time data were logged to better meet the assumptions of the statistical technique. A two-sided z-test from a mixed effects model with random intercepts for patients was used to compare the treatment delivery times. Costs Costs were estimated from the viewpoint of the health service provider and only costs incurred in the delivery of the
e95
service were included. All costs are using 2010 Australian dollar (A$) estimates. Indicative prices of the KVI and EPI linear accelerators and planning stations were provided by Varian in US dollars and were converted to Australian dollars at the exchange rate of A$1 ¼ US$ 0.90 (average January to October 2010). Institutional costs were grouped into capital costs representing a one-off investment and recurrent costs based on a two linear accelerator department model. Building costs were taken from that of a new radiotherapy centre being set up, with about 2500 m2 of work area þ 30% circulation area. The figure quoted includes fixtures and fittings. This new centre has four radiotherapy bunkers, two of which will be kept for future expansion. Land value was not taken into consideration. Four planning stations were thought to be required as a minimum for a two linear accelerator department. The cost of purchasing the automated couch is included in the total cost of the KVI linear accelerator. Capital costs are shown in Table 1. Recurrent costs were obtained from an internal analysis benchmarked across a number of sites and services operated by our organisation. The full list of recurrent costs taken into account is shown in Table 2. Spare parts per annum cost more on the KVI than the EPI units and average figures were obtained from our engineering department. The cost per patient of implantation of three gold seeds for IGRT was estimated at A$130.70, although this cost was not added to the cost per fraction, as this cost was shared as a standard component of both imaging arms. The lifetime of equipment was taken from federal government policy to fund for replacement after the stated period. Where no recommendation was available from the federal government, lifetimes of equipment were estimated by our physicist. For simplification, we assumed that at the Table 1 Capital costs (A$) in 2010 dollars Purchase price
Lifetime
Building 31,000,000 40 5 years Varian EPI iX 2,222,222 10 years 120 MLC Varian On-Board 2,777,778 10 years Imager iX 120 MLC Varian Eclipse 188,889 5 years planning station Three further 216,667 5 years planning stations Computed tomography 818,820 5 years simulator Record and verify 550,000 5 years system Computers, servers 120,000 5 years and software Annual capital cost for a two EPI centre Annual capital cost for a two KVI centre
Estimated cost per annum 2,060,307 301,928 377,410 44,841 51,436 194,383 130,567 28,487 A$3,113,877 A$3,264,841
EPI, electronic portal imaging; MLC, multileaf collimator; KVI, kilovoltage imaging.
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
A$4,175,948 A$4,217,948
PSA, prostate-specific antigen; IT, information technology; EPI, electronic portal imaging; KVI, kilovoltage imaging.
5
end of its lifetime, capital would have been amortised. Capital costs were converted into equivalent annual cost by dividing the net present value by the present value of annuity of $1 in arrears [14]. The discount rate was taken as per the 10 year Australian Government Bond rate (risk free rate), which is presently 6%. Total linear accelerator work time was calculated based on the number of workdays per annum minus machine down time in days, from figures that were obtained from our engineering department. The cost minimisation analysis for this study was carried out similarly to that described by other authors [17]. A multiway sensitivity analysis was carried out by varying the assumptions made about cost, lifetime and treatment time to work out the maximal and minimal costs of both EPI and KVI, such that the difference in cost between the two groups was the smallest or largest possible. For the sensitivity analysis, capital costs, spare parts costs and lifetime were varied by 10%, wages were varied by 5%, and service costs by 2.5%, reflecting possible fluctuations in cost. Time was varied between the interquartile ranges (IQR) for treatment time for both units. Because the IQR was wide, this was also assumed to incorporate any errors in time measurement, including the difference in method of time measurement as described above. The discount rate was varied from 5 to 7%.
25
120,000 275,563 123,888 1,243,784 258,103 275,000 55,000 78,301 50,000 66,000 48,667 5000 26,000 70,000 46,000 75,000 30,000 7000
20
Annual recurrent costs for a two EPI centre Annual recurrent costs for a two KVI centre
1,343,642
15
Medical director (1), radiation oncologists (3), registrar (1) Deputy site director (1) Physicists (3) Engineers (1.5) Radiation therapists (8 senior, 10 base grade) Nurses (4) Receptionist (5) Office manager (1) PSA (2) IT (0.5) Service contract for record and verify system Service contract for treatment planning system Spare parts for EPI (2 units) Spare parts for KVI (2 units) Building and maintenance Utilities Cleaning Linen Waste removal
lateral images were taken after treatment. Thus, treatment delivery times were available for 75% of the total fractions for these patients. The difference in action threshold resulted in a couch shift being made about 98% of the time on KVI linear accelerators and 50% of the time on EPI linear accelerators. The median time for KVI treatment was 6.0 min (IQR 5.1e7.4 min), and for EPI treatment it was 10.0 min (IQR 8.3e11.8 min) (P value < 0.0001) (see Figure 1). The average treatment delivery time for KVI is between 35 and 40% (point estimate 38%) shorter than that for EPI. The distribution of treatment delivery times taken is shown in Figure 2. In patients where no couch shift was made, the median treatment delivery time for EPI was 8.75 min and the median time for KVI was 5.13 min. For this subgroup analysis we analysed 1213 fractions where no couch shift was made (1124 EPI, 89 KVI) from 154 patients (88 EPI, 66 KVI). We found that using KVI over EPI reduced treatment time by 39% (P < 0.0001 95% confidence interval 35e43%), i.e. proportionally similar to the time reduction in the whole group. The mean set-up time for 10 prostate IGRT patients was 4.8 min (range 3.0e6.2 min; standard deviation 59 s). With the addition of set-up time and treatment delivery time, the IQR for the total time taken to treat varied from 13.1 to 16.6 min per fraction for EPI and 9.9 to 12.2 min per fraction for KVI. These figures were used for the sensitivity analysis shown below. Of six EPI units at our centre, the average downtime per year for two EPI units was 7.4 days (range 2e11.4). Of four KVI units at our centre, the average downtime per year for KVI was 5.4 days (range 1.8e8). The difference was not significant and for the purposes of our calculations we will assume both EPI and KVI units have an approximately equal total breakdown time of 6.4 days. The total number of treatment minutes per day refers to the total number of minutes that each linear accelerator is actually being used to physically treat patients, and excludes machine warm up time. The typical patient appointments start at 8 am and run until 5.45 pm or about
Percent of Group
Table 2 Recurrent costs (A$) in 2010 dollars
10
e96
Time data were available for 8863 fractions. After the removal of aberrant data, time data were analysed for 8648 fractions, 6057 from KVI treatment and 2591 from EPI treatment from a total of 294 patients. The data extraction programme only included data if both anterioreposterior and
0
Results 3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Treatment Time (mins)
Fig 2. Treatment delivery time for electronic portal imaging (EPI; solid square) and kilovoltage imaging (KVI; shaded square).
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
585 min per day. For two linear accelerators, the total number of treatment minutes is 1170. Our centre runs Monday to Friday and also for 4 h each Saturday morning (0.4 of a day). The total number of treatment minutes per year ¼ [weekdays (261) public holidays (9) þ weekends worked (21.3) downtime (6.4)] total minutes patients treated per day (1170) ¼ 312 273 min of treatment time per year in total for a two linear accelerator department. For Kilovoltage Imaging Total annual costs for a two KVI linear accelerator cancer centre ¼ $7,482,789 Estimated cost per minute¼$7,482,789 O 312,273 ¼ $23.96 Estimated cost per fraction with KVI ¼ $23.96 (6 þ 4.8) ¼ $258.79 Estimated cost of a 39 fraction course of IGRT ¼ $10,092.93
For Electronic Portal Imaging Total annual costs for a two EPI linear accelerator cancer centre ¼ $7,289,825 Estimated cost per minute ¼ $7,289,825 O 312,273 ¼ $23.34 Estimated cost per fraction with EPI ¼ $23.34 (10 þ 4.8) ¼ $345.50 Estimated cost of a 39 fraction course of IGRT ¼ $13,474.39 The cost per fraction is A$86.71 or 25% less for KVI compared with EPI. The incremental cost difference for a 39 fraction course is A$3381. In this study, wage accounted for about 52% of the total department cost per annum, and equipment (excluding building cost) amounted to around 15% of the total department cost per annum. Table 3 shows the results of the sensitivity analysis. In the absence of statistical testing, a sensitivity analysis is an effective way of testing the robustness of this analysis. For a given time period, the discount rate would be identical for either unit and therefore the same discount rates are compared for both units. The analysis showed that the difference in cost per fraction for the cheapest and most expensive EPI and KVI units ranged from A$66.09 to A$101.64.
e97
Discussion The use of KVI with automated couch shift resulted in a calculated cost saving compared with the use of EPI with manual couch shift, due to a reduction in required treatment time. Figures from a typical two linear accelerator satellite centre (Peter MacCallum, Box Hill from 2007 to 2009) show that 14.5% of all radiotherapy fractions per year were delivered to treat prostate cancer patients with IGRT. The proportion of prostate cancer patients treated with radical radiotherapy is similar to that reported by a UK department [8]. In that study, the time taken to treat six patients with EPI IGRT was compared with matched controls treated with conformal radiotherapy without IGRT. IGRT was found to take 3.1 min longer than conformal radiotherapy without IGRT (13.1 min versus 10.0 min). It was estimated that to treat all prostate cancer patients radically with IGRT would increase machine workload by 2.2 h per day. This study shows that by using KVI with automated couch shift, treatment time is reduced by a median of 4 min compared with EPI. The selection of appropriate IGRT equipment can make routine IGRT practical within busy radiotherapy departments. The total number of couch shifts was 48% more for KVI compared with EPI (98% versus 50%). This was because of the difference in the action threshold, which was 0 mm for KVI and 3 mm for EPI. A study from the Princess Margaret Hospital looking at EPI IGRT for prostate cancer showed that treatment delivery times were in the order of 8.7 min for patients requiring isocentre adjustment and 6.1 min for those who did not [18]. The author stated that the action level of 3 mm was specified for EPI treatment at the Princess Margaret Hospital because it was felt that an action level of 3 mm would be more practical within the time frame of a standard appointment. In this study, despite the difference in action threshold, treatment time was still less on the KVI units compared with EPI units. Automated couch shift probably contributes significantly to reducing treatment time. However, the difference between automated couch shift and KVI results in a median 4 min faster treatment (10 to 6 min) compared with EPI units. In patients who did not have couch shifts, treatment delivery time was faster on KVI units by a median of 3.62 min (8.75 to 5.13) compared with EPI units. Both these figures lead us to deduce that the time saved by using automate couch correction is in the vicinity of only 0.38 min
Table 3 Sensitivity analysis. Least costly scenario (discount rate 5%) Capital þ recurrent cost per annum (A$) Cost per minute (A$) Duration per fraction IQR þ set-up time (4.8 min) Cost per fraction (A$) Cost difference (EPI e KVI) (A$)
Most costly scenario (discount rate 7%)
EPI
KVI
EPI
KVI
6,929,706 22.19 13.1 min 290.70
7,084,987 22.69 9.9 min 224.62
7,715,008 24.71 16.6 min 410.12
7,895,966 25.29 12.2 min 308.48
66.09
EPI, electronic portal imaging; KVI, kilovoltage imaging; IQR, interquartile range.
101.64
e98
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99
(4.00 to 3.62), leading us to conclude that other factors are responsible for the faster workflow on KVI. The possible other factors resulting in a faster workflow include faster interpretation of imaging and easier use of the software used for matching (IMPAC for EPI, Varian console for KVI). As both KV and MV linacs have a similar monitor unit output rate, presumably treatment time on KVI units was less because of faster detection of fiducial markers by radiation therapy staff with KVI compared with MVI. Reducing treatment delivery time may hold additional clinical benefit as it seems that intrafraction prostate displacement, i.e. the probability of prostate displacement larger than a given distance, increases with increasing duration of treatment [15,19]. Therefore, it should be a goal to reduce treatment time as much as possible, as a longer treatment not only leads to an increased cost of treatment, but may also compromise coverage of the target due to intrafraction motion if inadequate clinical target volume to planning target volume margins are used. Intrafraction motion should be accounted for when determining appropriate clinical target volume to planning target volume margins, and reducing treatment time may have an effect on margins. Although we have not done this in our study, one method to further reduce treatment time may be to use a mix of KVI from the lateral direction, and MVI from the anterior, which will result in less time taken to rotate the gantry, thereby reducing the time taken to image the patient. However, the time for gantry rotation (about 15 seconds) is not significantly longer than the time taken to retract the KV arm before MVI. The limitations of this study should be discussed. Spare parts data in this study were taken as estimates from our engineering department. Over the last 10 years, our engineering department has replaced only two MVI panels. Since installation, no KVI panels have needed replacing. The EPI panel on a KVI linac is seldom used and therefore has a longer life expectancy. Although these were real figures obtained from our engineering department, other departments may have a different experience. The approximate cost of an MVI and a KVI panel are A$70 000 and A$150 000, respectively. However, for the first year the panels are covered by warranty. Therefore, the costing figures should be considered as the minimum, and may increase if wear and tear of equipment is greater. Delaney et al. [20] analysed treatment time for 4316 radiotherapy fractions and found 17 factors that influenced treatment time significantly. Slight differences between KVI and EPI have not been considered in detail, for example the quality assurance time for both have been considered the same. However, KVI quality assurance includes a component for cone beam computed tomography that is hard to separate out. There were more patients treated on KVI than EPI, which could imply that KVI radiation therapists were more experienced in prostate IGRT, although we think this is unlikely, as prostate cancer radiotherapy is very common, and it would be unusual for radiation therapists on either unit not to be regularly exposed to IGRT. A limitation of the cost minimisation analysis is that we have not considered any potential clinical benefits of one imaging modality over another. A cost minimisation analysis has the considerable appeal of keeping the comparison simple, which may be appropriate for
comparing new radiotherapy equipment as late effects and survival data take many years to be borne out and are probably similar between KVI and MVI. This model gives us a measurable input (resources used) and output, i.e. course of radiotherapy. The cost of all steps in radiotherapy treatment, such as simulation, contouring and planning, are accounted for by adding up their set-up and running costs. On the whole, cancer care worldwide is challenged by resource issues [21]. Similar to other studies, staffing costs in our study accounted for more than three times the annual cost compared with equipment costs (52% versus 15%) [22]. Because running costs outweigh the initial purchase costs, a time saving technology can result in an overall cost saving.
Conclusion There are several reasons why treatment time reduction is preferable in radiotherapy, i.e. it is more convenient for the patient, it results in a cost saving for the department, and there is a theoretical reduction in geographical miss of the tumour due to a reduction in intrafraction motion. The finding in this study is not unexpected, as we have two new time saving technologies being assessed, and we would expect to find that KVI with automated couch shift is slightly quicker than MVI with manual couch shift. This time analysis was conducted to quantify the difference, and it was unexpected that we found that patients who did not have a couch shift at all also had a treatment time reduction by 3.62 min, which leads us to conclude that other factors in the workflow, such as interpretation of imaging, is quicker on the KVI units. Many centres still use MVI for IGRT, and manual couch shift, because of cost issues. This study supports the use of new technology in IGRT equipment, which could result in an overall cost saving through faster treatment time.
Acknowledgments Presented in part at the Annual Scientific Meeting of Royal Australian and New Zealand College of Radiologists, Adelaide, Australia 2008. Peter MacCallum Cancer Centre has a research agreement with Varian Medical Systems. The authors thank Brad Shilling for engineering cost figures.
References [1] Viani GA, Stefano EJ, Afonso SL. Higher-than-conventional radiation doses in localized prostate cancer treatment: a meta-analysis of randomized, controlled trials. Int J Radiat Oncol Biol Phys 2009;74:1405e1418. [2] Gauthier I, Carrier JF, Beliveau-Nadeau D, Fortin B, Taussky D. Dosimetric impact and theoretical clinical benefits of fiducial markers for dose escalated prostate cancer radiation treatment. Int J Radiat Oncol Biol Phys 2009;74:1128e1133. [3] Jarad M, Andrew B, Robert B, et al. Image guided dose escalated prostate radiotherapy: still room to improve. Radiat Oncol 2009;4. doi:10.1186/1748-717X-4-50. [4] Kupelian PA, Langen KM, Willoughby TR, Zeidan OA, Meeks SL. Image-guided radiotherapy for localized prostate cancer: treating a moving target. Semin Radiat Oncol 2008;18(1):58e66.
S. Gill et al. / Clinical Oncology 24 (2012) e93ee99 [5] Button MR, Staffurth JN. Clinical application of image-guided radiotherapy in bladder and prostate cancer. Clin Oncol (R Coll Radiol) 2010;22:698e706. [6] Thompson A, Fox C, Foroudi F, et al. Planning and implementing an implanted fiducial programme for prostate cancer radiation therapy. J Med Imaging Radiat Oncol 2008;52:419e424. [7] Gill S, Thomas J, Fox C, et al. Acute toxicity in prostate cancer patients treated with and without image-guided radiotherapy. Radiat Oncol 2011;6:145. [8] Rimmer YL, Burnet NG, Routsis DS, et al. Practical issues in the implementation of image-guided radiotherapy for the treatment of prostate cancer within a UK department. Clin Oncol (R Coll Radiol) 2008;20:22e30. [9] Jaffray DA, Chawla K, Yu C, Wong JW. Dual-beam imaging for online verification of radiotherapy field placement. Int J Radiat Oncol Biol Phys 1995;33:1273e1280. [10] Pisani L, Lockman D, Jaffray D, Yan D, Martinez A, Wong J. Setup error in radiotherapy: on-line correction using electronic kilovoltage and megavoltage radiographs. Int J Radiat Oncol Biol Phys 2000;47:825. [11] Gill S, Thomas J, Fox C, et al. Electronic portal imaging vs kilovoltage imaging in fiducial marker image guided radiotherapy for prostate cancer: an analysis of set-up uncertainties. Br J Radiol 2012;85:176e182. [12] Bel A, Petrascu O, Van de Vondel I, et al. A computerized remote table control for fast on-line patient repositioning: implementation and clinical feasibility. Med Phys 2000;27:354. [13] Jefferies S, Taylor A, Reznek R. Results of a national survey of radiotherapy planning and delivery in the UK in 2007. Clin Oncol (R Coll Radiol) 2009;21:204e217.
e99
[14] Drummond MF, Sculpher MJ, Torrance GW, O’Brien BJ, Stoddart GL. Methods for the economic evaluation of health care programmes. New York: Oxford University Press; 2005. [15] Kron T, Thomas J, Fox C, et al. Intra-fraction prostate displacement in radiotherapy estimated from pre- and posttreatment imaging of patients with implanted fiducial markers. Radiother Oncol 2010;95:191e197. [16] Fox C, Fisher R, Kron T, et al. Extraction of data for margin calculations in prostate radiotherapy from a commercial record and verify system. J Med Imaging Radiat Oncol 2010;54:161e170. [17] Foroudi F, Lapsley H, Manderson C, Yeghiaian-Alvandi R. Cost-minimization analysis: radiation treatment with and without a multi-leaf collimator. Int J Radiat Oncol Biol Phys 2000;47:1443e1448. [18] Chung PWM, Haycocks T, Brown T, et al. On-line aSi portal imaging of implanted fiducial markers for the reduction of interfraction error during conformal radiotherapy of prostate carcinoma. Int J Radiat Oncol Biol Phys 2004;60:329e334. [19] Langen KM, Willoughby TR, Meeks SL, et al. Observations on real-time prostate gland motion using electromagnetic tracking. Int J Radiat Oncol Biol Phys 2008;71:1084e1090. [20] Delaney GP, Shafiq RJ, Jalaludin BB, Barton MB. The development of a new basic treatment equivalent model to assess linear accelerator throughput. Clin Oncol (R Coll Radiol) 2005;17:311e318. [21] Baume P. cROI. A vision for radiotherapy. Canberra: Department of Health; 2002. [22] Lievens Y, Van den Bogaert W, Kesteloot K. Activity-based costing: a practical model for cost calculation in radiotherapy. Int J Radiat Oncol Biol Phys 2003;57:522e535.