- Email: [email protected]
Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand
Radiotherapy and Oncology xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Original article
Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand Karen Wong a,b,⇑, Geoff P. Delaney a,b,c, Michael B. Barton a,b a Collaboration for Cancer Outcomes Research and Evaluation (CCORE), Ingham Institute for Applied Medical Research, Liverpool Hospital, UNSW Australia; b Department of Radiation Oncology, Cancer Therapy Centre, Liverpool Hospital, Sydney; and c University of Western Sydney, Australia
a r t i c l e
i n f o
Article history: Received 19 August 2015 Received in revised form 8 December 2015 Accepted 13 December 2015 Available online xxxx Keywords: Optimal Radiotherapy Fractions Cancer
a b s t r a c t Background and purpose: The recently updated optimal radiotherapy utilisation model estimated that 48.3% of all cancer patients should receive external beam radiotherapy at least once during their disease course. Adapting this model, we constructed an evidence-based model to estimate the optimal number of fractions for notifiable cancers in Australia to determine equipment and workload implications. Materials and methods: The optimal number of fractions was calculated based on the frequency of specific clinical conditions where radiotherapy is indicated and the evidence-based recommended number of fractions for each condition. Sensitivity analysis was performed to assess the impact of variables on the model. Results: Of the 27 cancer sites, the optimal number of fractions for the first course of radiotherapy ranged from 0 to 23.3 per cancer patient, and 1.5 to 29.1 per treatment course. Brain, prostate and head and neck cancers had the highest average number of fractions per course. Overall, the optimal number of fractions was 9.4 per cancer patient (range 8.7–10.0) and 19.4 per course (range 18.0–20.7). Conclusions: These results provide valuable data for radiotherapy services planning and comparison with actual practice. The model can be easily adapted by inserting population-specific epidemiological data thus making it applicable to other jurisdictions. Ó 2015 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology xxx (2015) xxx–xxx
Radiotherapy services planning requires a reliable estimate of radiotherapy demand. Actual radiotherapy utilisation (RTU) rates vary substantially throughout Australia and internationally [1]. Delaney et al. [2] constructed an evidence-based optimal RTU model which estimated that 52.3% of all cancer patients should be treated with external beam radiotherapy at least once during their disease course. Recent update of the model estimated an optimal RTU rate of 48.3% due to changes in epidemiological data and radiotherapy indications, and refinements of the model structure [3]. Further work is required to determine equipment and workload implications of the model. A treatment fraction is a fundamental unit of radiotherapy productivity. The average number of fractions per radiotherapy course in a department will depend on the proportion of patients receiving radical versus palliative treatment. Number of fractions has been used for radiotherapy services planning. Morgan et al. [4] estimated that an extra 50 linear accelerators were required in Australia and New Zealand in 2009 to achieve a 52.3% RTU rate, based on 19 fractions per treatment course which reflected actual ⇑ Corresponding author at: Department of Radiation Oncology, Cancer Therapy Centre, Liverpool Hospital, Locked Bag 7103, Liverpool BC, NSW 1871, Australia. E-mail address: [email protected] (K. Wong).
practice. Williams et al. [5] modelled the radiotherapy activity required to deliver an evidence-based radiotherapy service and compared with actual radiotherapy activity in the UK in 2005. A 33% increase in activity was required to achieve a 52% RTU rate. A further increase of 37% in activity was required when guideline-recommended evidence-based dose-fractionation schedules [6] were taken into consideration. Substantial variation in radiotherapy fractionation practices has also been observed. The average number of fractions per treatment course ranged from 9.1 to 23.5 in the 24 Radiation Oncology centres in New South Wales (NSW) in 2013 [7]. The overall average number of fractions per course in NSW was 19. In comparison, the average number of fractions per course was 13.7 in Scotland in 2003 [8]. Variation was also observed in the five Radiation Oncology departments in Scotland (ranging from 11.7 to 17.3 fractions per course) [8] and in the UK (ranging from 13.0 in England to 17.8 in Ireland) [9]. Casemix alone does not account for all fractionation variations. An evidence-based optimal radiotherapy fractionation (RTF) model was constructed to estimate for the first course of radiotherapy: (i) the optimal number of fractions per cancer patient and per treatment course, (ii) the proportion of patients that should receive radical versus palliative radiotherapy, (iii) the optimal number of
http://dx.doi.org/10.1016/j.radonc.2015.12.001 0167-8140/Ó 2015 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
2
Optimal number of fractions for cancer
fractions per radical and per palliative radiotherapy course if all cases were treated according to national and international guidelines.
Materials and methods The updated RTU model [3] was used as the basis of this study. As the purpose of the RTU model was to determine the proportion of cancer patients who have at least one indication for radiotherapy at some time in their disease course, patients requiring radiotherapy were counted once, even if they had multiple indications at different stages in their illness. The current study was limited to the first course of radiotherapy. We recently published the breast cancer RTF model based on the original RTU model, and the methodology was discussed in detail [10]. The same methodology has been applied to all cancer sites of the updated RTU model. The recommended dose-fractionation schedules were derived from evidence-based treatment guidelines published between January 2000 and December 2014. An indication for radiotherapy was defined as a clinical situation for which radiotherapy was recommended as the treatment of choice on the basis of published evidence that radiotherapy has a superior clinical outcome compared to alternative treatment modalities (including no treatment), and where the patient was suitable to undergo radiotherapy based on an assessment of performance status and co-morbidities. Radiotherapy indications were derived from evidence-based treatment guidelines issued by reputed national and international organisations. In the RTU model, patient and tumour-related attributes were used to define specific radiotherapy indications, so each branch point represented a particular radiotherapy indication. For this study, some of the RTU model branches were split to model more specific clinical situations where the fractionation schemes vary between branches. Proportions of patients with the different attributes associated with additional branches were obtained by performing Medline searches, manual bibliographic searches and examination of review articles. Australian data were used if available as the primary purpose was to apply this model to the Australian population. TreeAge Pro Suite 2009TM was used to construct the RTF model as a decision tree. ‘‘Chance nodes” (represented by circles) were used to depict different clinical scenarios. A node’s branches represented the outcomes or alternatives associated with each clinical scenario, the number underneath each branch representing the proportion of patients with that attribute. The recommended number of fractions, derived from evidence-based treatment guidelines, was added at each ‘‘terminal node” (represented by a triangle) as the final outcome, referred to as the ‘‘payoff”. If the guidelines did not adequately address dose-fractionation schedules, other sources including meta-analyses and randomised controlled trials were identified. The quality of evidence was assessed according to the National Health and Medical Research Council hierarchy of levels of evidence [11]. When a range of fractions was recommended, the number of fractions best supported by evidence was used in the calculations. If there were a number of sources of equal quality that recommended different fractionation schedules, the Australian guidelines recommendation was used if available, as the primary purpose was to make recommendations for radiotherapy services in Australia. When fractionation recommendations were not in the Australian guidelines, the lowest of the range of fractions recommended in the other guidelines was used. The effect of higher fraction numbers on the model was tested by sensitivity analysis. The decision tree was analysed from right to left. For each cancer site, the optimal number of fractions per patient, depicted by
the number at the left most node, represented a weighted mean taking into account all the payoffs (recommended numbers of fractions at the terminal nodes) and the probability of each clinical scenario. By dividing this number by the proportion of patients with that particular cancer recommended to have radiotherapy, the optimal number of fractions per treatment course was calculated. The overall optimal number of fractions was the weighted average of the optimal number of fractions for all cancer sites, taking into account the different proportions of these cancers. For patients with an indication of radiotherapy, further analysis was performed to determine the proportion of patients recommended to have radical versus palliative radiotherapy, and the optimal number of fractions per radical and per palliative course. For each branch with a range of recommended number of fractions or a range of epidemiological data, one-way sensitivity analysis was conducted to assess the effect on the optimal estimate for each cancer site and for the entire tree. Results There are 27 cancer sites in the RTF model (Table 1). For each cancer site, the optimal number of fractions ranged from 0 to 23.3 per cancer patient. As an example, the rectal cancer RTF model is shown in Fig. 1, with the optimal number of fractions per patient being 14.4. For the entire cancer population, the optimal number of fractions was 9.4 per cancer patient. This was a weighted average of all indications including radical, palliative and cases in which radiotherapy was not recommended. The last column of Table 1 shows the optimal number of fractions per radiotherapy course, considering only the patient population for whom radiotherapy was recommended. This ranged from 1.5 to 29.1, with the highest being brain, prostate and head and
Table 1 Optimal radiotherapy utilisation rate and number of fractions. Cancer site
Bladder Brain Breast Cervix Colon Gallbladder Head and neck Kidney Leukaemia Liver Lung Lymphoma Melanoma Myeloma Oesophagus Ovary Pancreas Prostate Rectum Stomach Testis Thyroid Unknown primary Uterus Vagina Vulva Other Total
Optimal radiotherapy utilisation (%)
Optimal number of fractions per cancer patient
Optimal number of fractions per treatment course
2.0 1.4 12.2 1.0 8.4 0.6 3.3
47 80 87 71 4 17 74
4.9 23.3 14.3 15 0.1 4.1 20
10.4 29.1 16.4 21.1 2.5 24.1 27.0
2.3 2.3 1.2 9.0 4.2 9.9 1.2 1.2 1.1 2.1 18.4 4.2 1.8 0.8 1.8 2.4
15 4 0 78 73 21 45 71 4 49 58 60 27 15 4 61
0.3 0.3 0 12.1 10.4 3.9 1.6 10 0.3 10.3 16.3 14.4 5.0 2.2 0.5 0.9
2.0 7.5 – 15.5 14.2 18.6 3.6 14.1 7.5 21.0 28.1 24.0 18.5 14.7 12.5 1.5
32 94 39 19 48.4
7.1 20.7 9.4 3.5 9.4
22.2 22.0 24.1 18.4 19.4
Proportion of all cancers in Australia (%)
1.8 0.1 0.3 5.0 100.0
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
3
K. Wong et al. / Radiotherapy and Oncology xxx (2015) xxx–xxx
Fig. 1. Optimal radiotherapy fractionation model for rectal cancer.
neck cancers. Overall, the optimal number of fractions was 19.4 per course. For the first course of radiotherapy, 77% of patients should optimally be treated with radical intent and 23% with palliative intent. The optimal number of fractions per treatment course was 24.3 for radical radiotherapy and 3.0 for palliative radiotherapy. One-way sensitivity analysis was performed for each variable by setting upper and lower data limits and modelling the RTF model using these extreme values. For the entire cancer population, the optimal number of fractions varied from 8.7 to 10.0 per cancer patient. With an optimal RTU rate of 48.4%, this translated to 18.0 to 20.7 fractions per treatment course for the first course of radiotherapy. The optimal number of fractions per course ranged from 22.4 to 25.9 for radical radiotherapy, and 2.8 to 4.0 for palliative radiotherapy.
Table 2 Estimated optimal number of radiotherapy fractions per 1000 new cancer cases. New cancer cases
Number 1000
Number of first radiotherapy courses (=number of new cancer cases optimal radiotherapy utilisation rate) Number of fractions for first course of radiotherapy (=number of first radiotherapy courses optimal number of fractions per first course of radiotherapy) Number of retreatment courses (=number of first radiotherapy courses actual retreatment rate) Number of fractions for retreatment (=number of retreatment courses optimal number of fractions per retreatment course) Total number of courses (=number of first radiotherapy courses + number of retreatment courses) Total number of fractions (=number of fractions for first radiotherapy courses + number of fractions for retreatment courses)
1000 0.484 = 484 484 19.4 = 9390
484 0.26 = 126 126 3.0 = 378
484 + 126 = 610 9390 + 378 = 9768
Discussion The RTF model estimates that the optimal number of fractions for the first course of radiotherapy is 9.4 per cancer patient and 19.4 per treatment course. Sensitivity analysis showed a small range of number of fractions suggesting robustness of the model. The RTF model provides a benchmark for radiotherapy services planning on a population basis. For every 1000 new cases of notifiable cancer, the radiotherapy requirements may be calculated as shown in Table 2. As retreatment is outside of the scope of this study, the actual retreatment rate in NSW in 2013 (26%) [7] is used to estimate the proportion of patients requiring retreatment, and our estimated optimal number of fractions per palliative course (3.0 fractions per course) is applied to estimate the optimal number of fractions for retreatment as most retreatment is palliative. For every 1000 new cancer cases, 610 treatment courses and 9768 fractions will be required, which is an average of 16.0 frac-
tions per course when both first and retreatment courses are considered. Applying the range of optimal number of fractions per treatment course (18.0–20.7 per first course and 2.8–4.0 per retreatment course), the optimal number of fractions per treatment course (first and retreatment courses included) is estimated to range from 14.9 to 17.2. Our estimate is slightly below the evidence-based estimates of 17.4 per treatment course for 2006 and 17.7 for 2011 from the Malthus model [12], probably because there are several differences between the models including epidemiological data used, structure of the model, sources of recommended number of fractions, and actual retreatment rates applied to estimate retreatment demand. Estimating the optimal number of fractions has been recognised as valuable in radiotherapy services planning. The Global Taskforce on Radiotherapy for Cancer Control, set up by the Union for
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
4
Optimal number of fractions for cancer
Table 3 Comparison of optimal with actual number of fractions per treatment course. Treatment
Optimal number of fractions per course (range)
Actual number of fractions per course
Ontario, Canada 1996–97 [14] First course All (first and retreatment courses) Radical Palliative
19.4 (18.0–20.7) 16 (14.9–17.2)
– 15.3
France 1999 [15] – 20
24.3 (22.4–25.9) 3.0 (2.8–4.0)
24.9 5.8
– –
International Cancer Control, estimated that 119 million fractions were necessary for all cancer patients worldwide to receive full access to guideline-based radiotherapy in 2012, increasing to over 204 million in 2035, based on evidence-based optimal RTU rates and number of fractions per treatment course [13]. An investment of US $184 billion would be required to achieve global equity in radiotherapy access by 2035. The RTF model provides a benchmark for service delivery and allows comparison with actual practice. Population-based studies showed that the average number of fractions per course varied between 13.4 and 20 when all treatment courses were considered [7–9,14–17] (Table 3). The number of fractions per course in Ontario and Sweden was close to our evidence-based estimate, but higher than our estimate in France and NSW, and lower in Scotland, the UK and England. The average number of fractions per radical course approximated our estimate except it was lower in the UK [9], suggesting hypofractionation was more commonly used. The average number of fractions per palliative course was higher than our estimate except in the UK. In Scotland and England [8,17], the average number of fractions per treatment course was below our estimate, however the number of fractions per radical course approximated our estimate, and that per palliative course was above our estimate, suggesting perhaps a higher than optimal proportion of patients were treated with palliative intent. Data from NSW showed that our optimal estimates of proportions of patients who should receive radical versus palliative radiotherapy are potentially achievable. In 2013, 69.5% of first radiotherapy courses were radical in intent [7]. This model has a number of limitations. Firstly, the model only considered the first course of radiotherapy. Retreatment demand needs to be considered in radiotherapy services planning. Patterns of relapse data over time could be incorporated into this model to estimate the optimal number of fractions for retreatment. This is a complex process that requires additional epidemiological data not currently available from the literature. The modelling needs to include data on the natural history and frequency of the development of symptoms with an indication for radiotherapy, the time course between episodes and the overall history of the disease. Future research addressing these issues will be helpful to more accurately predict radiotherapy demand. Secondly, the model included only notifiable cancers in Australia, excluding non-melanomatous skin cancers and benign tumours which are not notifiable. It is difficult to estimate the optimal number of fractions for these conditions as their incidence is unknown. Additional workload needs to be considered when this model is used in radiotherapy services planning. Thirdly, there was a lack of high quality epidemiological data for some clinical situations, particularly performance status and comorbidity data. Because the lack of these data applied mainly to
Sweden 2001 [16]
Scotland 2003 [8]
– 14.6
– 13.7
23.9 7.0
24 7 (primary tumour) 4 (metastasis)
UK 2007 [9] 15.4 13.4
England 2011–2012 [17] – 13.8
New South Wales, Australia 2013 [7]
20.6 4.0
23.0 4.6
– –
– 19
the terminal branches of the model, the impact on the overall estimate was small. Furthermore, sensitivity analysis where high quality data were lacking showed a minor impact of these uncertainties on the overall estimate. Lastly, there are differing recommendations on radiotherapy indication and dose-fractionation schedule for many clinical situations in the treatment guidelines. For some clinical situations, the radiotherapy indications are poorly defined. There are also clinical situations with widely disparate treatment options, one of which could be radiotherapy such as early prostate cancer. Sensitivity analysis showed that the variables had a minor impact on the overall estimate. Our study expanded on the evidence-based RTU model to aid in radiotherapy services planning. RTU and fractionation benchmarks can be evidence-based or criterion-based. RTU criterion-based benchmarking is an empirical method derived from real life observations. The benchmarks are demonstrably achievable, however, the validity of the benchmarks hinges on the validity of the criteria chosen to identify the benchmark population [18]. Evidence-based methods have the advantage of avoiding potential biases associated with criterion-based methods. Furthermore, the RTF model can be readily modified should there be future changes in cancer incidence, stage distribution, radiotherapy indication and dosefractionation schedule, and can be easily adapted for different jurisdictions. Conflict of interest statement There is no conflict of interest. References [1] Borras JM, Lievens Y, Dunscombe P, Coffey M, Malicki J, Corral J, et al. The optimal utilization proportion of external beam radiotherapy in European countries: an ESTRO-HERO analysis. Radiother Oncol 2015. [2] Delaney G, Jacob S, Featherstone C, Barton M. The role of radiotherapy in cancer treatment: estimating optimal utilization from a review of evidencebased clinical guidelines. Cancer 2005;104:1129–37. [3] Barton MB, Jacob S, Shafiq J, Wong K, Thompson SR, Hanna TP, et al. Estimating the demand for radiotherapy from the evidence: a review of changes from 2003 to 2012. Radiother Oncol 2014;112:140–4. [4] Morgan G, Barton M, Crossing S, Bull C, Penman A. A ’Catch Up’ plan for radiotherapy in New South Wales to 2012. J Med Imaging Radiat Oncol 2009;53:419–30. [5] Williams MV, Summers ET, Drinkwater K, Barrett A. Radiotherapy dose fractionation, access and waiting times in the countries of the UK in 2005. Clin Oncol 2007;19:273–86. [6] The Royal College of Radiologists. Radiotherapy dose-fractionation. https:// www.rcr.ac.uk/sites/default/files/publication/Dose-Fractionation_Final.pdf; 2006. Accessed 7/8/2015. [7] Health System Planning and Investment Branch. Radiotherapy treatment services to NSW residents 2013 annual report. Sydney: NSW Ministry of Health; 2014.
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
K. Wong et al. / Radiotherapy and Oncology xxx (2015) xxx–xxx [8] Scottish Executive Health Department. Radiotherapy Activity Planning for Scotland 2011-2015. http://www.gov.scot/Resource/Doc/90297/0021749.pdf; 2006. Accessed 21/7/2015. [9] Williams MV, Drinkwater KJ. Geographical variation in radiotherapy services across the UK in 2007 and the effect of deprivation. Clin Oncol 2009;21:431–40. [10] Wong K, Delaney GP, Barton MB. Estimation of the optimal number of radiotherapy fractions for breast cancer: a review of the evidence. Radiother Oncol 2015. [11] National Health and Medical Research Council. A guide to the development, implementation and evaluation of clinical practice guidelines. Canberra: Commonwealth of Australia; 1999. [12] Round CE, Williams MV, Mee T, Kirkby NF, Cooper T, Hoskin P. Radiotherapy demand and activity in England 2006-2020. Clin Oncol 2013;25: 522–30. [13] Atun R, Jaffray DA, Barton MB, Bray F, Baumann M, Vikram B, et al. Expanding global access to radiotherapy. Lancet Oncol 2015;16:1153–86.
5
[14] Dixon P, Mackillop W. Could changes in clinical practice reduce waiting lists for radiotherapy? J Health Serv Res Policy 2001;6:70–7. [15] Ruggieri-Pignon S, Pignon T, Marty M, Rodde-Dunet MH, Destembert B, Fritsch B. Infrastructure of radiation oncology in France: a large survey of evolution of external beam radiotherapy practice. Int J Radiat Oncol Biol Phys 2005;61:507–16. [16] Moller TR, Brorsson B, Ceberg J, Frodin JE, Lindholm C, Nylen U, et al. A prospective survey of radiotherapy practice 2001 in Sweden. Acta Oncol 2003;42:387–410. [17] Department of Health. Radiotherapy Services in England 2012. https:// www.gov.uk/government/uploads/system/uploads/attachment_data/file/ 213151/Radiotherapy-Services-in-England-2012.pdf; 2012. Accessed 21/7/ 2015. [18] Mackillop WJ, Kong W, Brundage M, Hanna TP, Zhang-Salomons J, McLaughlin PY, et al. A comparison of evidence-based estimates and empirical benchmarks of the appropriate rate of use of radiation therapy in Ontario. Int J Radiat Oncol Biol Phys 2015;91:1099–107.
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
Contents lists available at ScienceDirect
Radiotherapy and Oncology journal homepage: www.thegreenjournal.com
Original article
Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand Karen Wong a,b,⇑, Geoff P. Delaney a,b,c, Michael B. Barton a,b a Collaboration for Cancer Outcomes Research and Evaluation (CCORE), Ingham Institute for Applied Medical Research, Liverpool Hospital, UNSW Australia; b Department of Radiation Oncology, Cancer Therapy Centre, Liverpool Hospital, Sydney; and c University of Western Sydney, Australia
a r t i c l e
i n f o
Article history: Received 19 August 2015 Received in revised form 8 December 2015 Accepted 13 December 2015 Available online xxxx Keywords: Optimal Radiotherapy Fractions Cancer
a b s t r a c t Background and purpose: The recently updated optimal radiotherapy utilisation model estimated that 48.3% of all cancer patients should receive external beam radiotherapy at least once during their disease course. Adapting this model, we constructed an evidence-based model to estimate the optimal number of fractions for notifiable cancers in Australia to determine equipment and workload implications. Materials and methods: The optimal number of fractions was calculated based on the frequency of specific clinical conditions where radiotherapy is indicated and the evidence-based recommended number of fractions for each condition. Sensitivity analysis was performed to assess the impact of variables on the model. Results: Of the 27 cancer sites, the optimal number of fractions for the first course of radiotherapy ranged from 0 to 23.3 per cancer patient, and 1.5 to 29.1 per treatment course. Brain, prostate and head and neck cancers had the highest average number of fractions per course. Overall, the optimal number of fractions was 9.4 per cancer patient (range 8.7–10.0) and 19.4 per course (range 18.0–20.7). Conclusions: These results provide valuable data for radiotherapy services planning and comparison with actual practice. The model can be easily adapted by inserting population-specific epidemiological data thus making it applicable to other jurisdictions. Ó 2015 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology xxx (2015) xxx–xxx
Radiotherapy services planning requires a reliable estimate of radiotherapy demand. Actual radiotherapy utilisation (RTU) rates vary substantially throughout Australia and internationally [1]. Delaney et al. [2] constructed an evidence-based optimal RTU model which estimated that 52.3% of all cancer patients should be treated with external beam radiotherapy at least once during their disease course. Recent update of the model estimated an optimal RTU rate of 48.3% due to changes in epidemiological data and radiotherapy indications, and refinements of the model structure [3]. Further work is required to determine equipment and workload implications of the model. A treatment fraction is a fundamental unit of radiotherapy productivity. The average number of fractions per radiotherapy course in a department will depend on the proportion of patients receiving radical versus palliative treatment. Number of fractions has been used for radiotherapy services planning. Morgan et al. [4] estimated that an extra 50 linear accelerators were required in Australia and New Zealand in 2009 to achieve a 52.3% RTU rate, based on 19 fractions per treatment course which reflected actual ⇑ Corresponding author at: Department of Radiation Oncology, Cancer Therapy Centre, Liverpool Hospital, Locked Bag 7103, Liverpool BC, NSW 1871, Australia. E-mail address: [email protected] (K. Wong).
practice. Williams et al. [5] modelled the radiotherapy activity required to deliver an evidence-based radiotherapy service and compared with actual radiotherapy activity in the UK in 2005. A 33% increase in activity was required to achieve a 52% RTU rate. A further increase of 37% in activity was required when guideline-recommended evidence-based dose-fractionation schedules [6] were taken into consideration. Substantial variation in radiotherapy fractionation practices has also been observed. The average number of fractions per treatment course ranged from 9.1 to 23.5 in the 24 Radiation Oncology centres in New South Wales (NSW) in 2013 [7]. The overall average number of fractions per course in NSW was 19. In comparison, the average number of fractions per course was 13.7 in Scotland in 2003 [8]. Variation was also observed in the five Radiation Oncology departments in Scotland (ranging from 11.7 to 17.3 fractions per course) [8] and in the UK (ranging from 13.0 in England to 17.8 in Ireland) [9]. Casemix alone does not account for all fractionation variations. An evidence-based optimal radiotherapy fractionation (RTF) model was constructed to estimate for the first course of radiotherapy: (i) the optimal number of fractions per cancer patient and per treatment course, (ii) the proportion of patients that should receive radical versus palliative radiotherapy, (iii) the optimal number of
http://dx.doi.org/10.1016/j.radonc.2015.12.001 0167-8140/Ó 2015 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
2
Optimal number of fractions for cancer
fractions per radical and per palliative radiotherapy course if all cases were treated according to national and international guidelines.
Materials and methods The updated RTU model [3] was used as the basis of this study. As the purpose of the RTU model was to determine the proportion of cancer patients who have at least one indication for radiotherapy at some time in their disease course, patients requiring radiotherapy were counted once, even if they had multiple indications at different stages in their illness. The current study was limited to the first course of radiotherapy. We recently published the breast cancer RTF model based on the original RTU model, and the methodology was discussed in detail [10]. The same methodology has been applied to all cancer sites of the updated RTU model. The recommended dose-fractionation schedules were derived from evidence-based treatment guidelines published between January 2000 and December 2014. An indication for radiotherapy was defined as a clinical situation for which radiotherapy was recommended as the treatment of choice on the basis of published evidence that radiotherapy has a superior clinical outcome compared to alternative treatment modalities (including no treatment), and where the patient was suitable to undergo radiotherapy based on an assessment of performance status and co-morbidities. Radiotherapy indications were derived from evidence-based treatment guidelines issued by reputed national and international organisations. In the RTU model, patient and tumour-related attributes were used to define specific radiotherapy indications, so each branch point represented a particular radiotherapy indication. For this study, some of the RTU model branches were split to model more specific clinical situations where the fractionation schemes vary between branches. Proportions of patients with the different attributes associated with additional branches were obtained by performing Medline searches, manual bibliographic searches and examination of review articles. Australian data were used if available as the primary purpose was to apply this model to the Australian population. TreeAge Pro Suite 2009TM was used to construct the RTF model as a decision tree. ‘‘Chance nodes” (represented by circles) were used to depict different clinical scenarios. A node’s branches represented the outcomes or alternatives associated with each clinical scenario, the number underneath each branch representing the proportion of patients with that attribute. The recommended number of fractions, derived from evidence-based treatment guidelines, was added at each ‘‘terminal node” (represented by a triangle) as the final outcome, referred to as the ‘‘payoff”. If the guidelines did not adequately address dose-fractionation schedules, other sources including meta-analyses and randomised controlled trials were identified. The quality of evidence was assessed according to the National Health and Medical Research Council hierarchy of levels of evidence [11]. When a range of fractions was recommended, the number of fractions best supported by evidence was used in the calculations. If there were a number of sources of equal quality that recommended different fractionation schedules, the Australian guidelines recommendation was used if available, as the primary purpose was to make recommendations for radiotherapy services in Australia. When fractionation recommendations were not in the Australian guidelines, the lowest of the range of fractions recommended in the other guidelines was used. The effect of higher fraction numbers on the model was tested by sensitivity analysis. The decision tree was analysed from right to left. For each cancer site, the optimal number of fractions per patient, depicted by
the number at the left most node, represented a weighted mean taking into account all the payoffs (recommended numbers of fractions at the terminal nodes) and the probability of each clinical scenario. By dividing this number by the proportion of patients with that particular cancer recommended to have radiotherapy, the optimal number of fractions per treatment course was calculated. The overall optimal number of fractions was the weighted average of the optimal number of fractions for all cancer sites, taking into account the different proportions of these cancers. For patients with an indication of radiotherapy, further analysis was performed to determine the proportion of patients recommended to have radical versus palliative radiotherapy, and the optimal number of fractions per radical and per palliative course. For each branch with a range of recommended number of fractions or a range of epidemiological data, one-way sensitivity analysis was conducted to assess the effect on the optimal estimate for each cancer site and for the entire tree. Results There are 27 cancer sites in the RTF model (Table 1). For each cancer site, the optimal number of fractions ranged from 0 to 23.3 per cancer patient. As an example, the rectal cancer RTF model is shown in Fig. 1, with the optimal number of fractions per patient being 14.4. For the entire cancer population, the optimal number of fractions was 9.4 per cancer patient. This was a weighted average of all indications including radical, palliative and cases in which radiotherapy was not recommended. The last column of Table 1 shows the optimal number of fractions per radiotherapy course, considering only the patient population for whom radiotherapy was recommended. This ranged from 1.5 to 29.1, with the highest being brain, prostate and head and
Table 1 Optimal radiotherapy utilisation rate and number of fractions. Cancer site
Bladder Brain Breast Cervix Colon Gallbladder Head and neck Kidney Leukaemia Liver Lung Lymphoma Melanoma Myeloma Oesophagus Ovary Pancreas Prostate Rectum Stomach Testis Thyroid Unknown primary Uterus Vagina Vulva Other Total
Optimal radiotherapy utilisation (%)
Optimal number of fractions per cancer patient
Optimal number of fractions per treatment course
2.0 1.4 12.2 1.0 8.4 0.6 3.3
47 80 87 71 4 17 74
4.9 23.3 14.3 15 0.1 4.1 20
10.4 29.1 16.4 21.1 2.5 24.1 27.0
2.3 2.3 1.2 9.0 4.2 9.9 1.2 1.2 1.1 2.1 18.4 4.2 1.8 0.8 1.8 2.4
15 4 0 78 73 21 45 71 4 49 58 60 27 15 4 61
0.3 0.3 0 12.1 10.4 3.9 1.6 10 0.3 10.3 16.3 14.4 5.0 2.2 0.5 0.9
2.0 7.5 – 15.5 14.2 18.6 3.6 14.1 7.5 21.0 28.1 24.0 18.5 14.7 12.5 1.5
32 94 39 19 48.4
7.1 20.7 9.4 3.5 9.4
22.2 22.0 24.1 18.4 19.4
Proportion of all cancers in Australia (%)
1.8 0.1 0.3 5.0 100.0
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
3
K. Wong et al. / Radiotherapy and Oncology xxx (2015) xxx–xxx
Fig. 1. Optimal radiotherapy fractionation model for rectal cancer.
neck cancers. Overall, the optimal number of fractions was 19.4 per course. For the first course of radiotherapy, 77% of patients should optimally be treated with radical intent and 23% with palliative intent. The optimal number of fractions per treatment course was 24.3 for radical radiotherapy and 3.0 for palliative radiotherapy. One-way sensitivity analysis was performed for each variable by setting upper and lower data limits and modelling the RTF model using these extreme values. For the entire cancer population, the optimal number of fractions varied from 8.7 to 10.0 per cancer patient. With an optimal RTU rate of 48.4%, this translated to 18.0 to 20.7 fractions per treatment course for the first course of radiotherapy. The optimal number of fractions per course ranged from 22.4 to 25.9 for radical radiotherapy, and 2.8 to 4.0 for palliative radiotherapy.
Table 2 Estimated optimal number of radiotherapy fractions per 1000 new cancer cases. New cancer cases
Number 1000
Number of first radiotherapy courses (=number of new cancer cases optimal radiotherapy utilisation rate) Number of fractions for first course of radiotherapy (=number of first radiotherapy courses optimal number of fractions per first course of radiotherapy) Number of retreatment courses (=number of first radiotherapy courses actual retreatment rate) Number of fractions for retreatment (=number of retreatment courses optimal number of fractions per retreatment course) Total number of courses (=number of first radiotherapy courses + number of retreatment courses) Total number of fractions (=number of fractions for first radiotherapy courses + number of fractions for retreatment courses)
1000 0.484 = 484 484 19.4 = 9390
484 0.26 = 126 126 3.0 = 378
484 + 126 = 610 9390 + 378 = 9768
Discussion The RTF model estimates that the optimal number of fractions for the first course of radiotherapy is 9.4 per cancer patient and 19.4 per treatment course. Sensitivity analysis showed a small range of number of fractions suggesting robustness of the model. The RTF model provides a benchmark for radiotherapy services planning on a population basis. For every 1000 new cases of notifiable cancer, the radiotherapy requirements may be calculated as shown in Table 2. As retreatment is outside of the scope of this study, the actual retreatment rate in NSW in 2013 (26%) [7] is used to estimate the proportion of patients requiring retreatment, and our estimated optimal number of fractions per palliative course (3.0 fractions per course) is applied to estimate the optimal number of fractions for retreatment as most retreatment is palliative. For every 1000 new cancer cases, 610 treatment courses and 9768 fractions will be required, which is an average of 16.0 frac-
tions per course when both first and retreatment courses are considered. Applying the range of optimal number of fractions per treatment course (18.0–20.7 per first course and 2.8–4.0 per retreatment course), the optimal number of fractions per treatment course (first and retreatment courses included) is estimated to range from 14.9 to 17.2. Our estimate is slightly below the evidence-based estimates of 17.4 per treatment course for 2006 and 17.7 for 2011 from the Malthus model [12], probably because there are several differences between the models including epidemiological data used, structure of the model, sources of recommended number of fractions, and actual retreatment rates applied to estimate retreatment demand. Estimating the optimal number of fractions has been recognised as valuable in radiotherapy services planning. The Global Taskforce on Radiotherapy for Cancer Control, set up by the Union for
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
4
Optimal number of fractions for cancer
Table 3 Comparison of optimal with actual number of fractions per treatment course. Treatment
Optimal number of fractions per course (range)
Actual number of fractions per course
Ontario, Canada 1996–97 [14] First course All (first and retreatment courses) Radical Palliative
19.4 (18.0–20.7) 16 (14.9–17.2)
– 15.3
France 1999 [15] – 20
24.3 (22.4–25.9) 3.0 (2.8–4.0)
24.9 5.8
– –
International Cancer Control, estimated that 119 million fractions were necessary for all cancer patients worldwide to receive full access to guideline-based radiotherapy in 2012, increasing to over 204 million in 2035, based on evidence-based optimal RTU rates and number of fractions per treatment course [13]. An investment of US $184 billion would be required to achieve global equity in radiotherapy access by 2035. The RTF model provides a benchmark for service delivery and allows comparison with actual practice. Population-based studies showed that the average number of fractions per course varied between 13.4 and 20 when all treatment courses were considered [7–9,14–17] (Table 3). The number of fractions per course in Ontario and Sweden was close to our evidence-based estimate, but higher than our estimate in France and NSW, and lower in Scotland, the UK and England. The average number of fractions per radical course approximated our estimate except it was lower in the UK [9], suggesting hypofractionation was more commonly used. The average number of fractions per palliative course was higher than our estimate except in the UK. In Scotland and England [8,17], the average number of fractions per treatment course was below our estimate, however the number of fractions per radical course approximated our estimate, and that per palliative course was above our estimate, suggesting perhaps a higher than optimal proportion of patients were treated with palliative intent. Data from NSW showed that our optimal estimates of proportions of patients who should receive radical versus palliative radiotherapy are potentially achievable. In 2013, 69.5% of first radiotherapy courses were radical in intent [7]. This model has a number of limitations. Firstly, the model only considered the first course of radiotherapy. Retreatment demand needs to be considered in radiotherapy services planning. Patterns of relapse data over time could be incorporated into this model to estimate the optimal number of fractions for retreatment. This is a complex process that requires additional epidemiological data not currently available from the literature. The modelling needs to include data on the natural history and frequency of the development of symptoms with an indication for radiotherapy, the time course between episodes and the overall history of the disease. Future research addressing these issues will be helpful to more accurately predict radiotherapy demand. Secondly, the model included only notifiable cancers in Australia, excluding non-melanomatous skin cancers and benign tumours which are not notifiable. It is difficult to estimate the optimal number of fractions for these conditions as their incidence is unknown. Additional workload needs to be considered when this model is used in radiotherapy services planning. Thirdly, there was a lack of high quality epidemiological data for some clinical situations, particularly performance status and comorbidity data. Because the lack of these data applied mainly to
Sweden 2001 [16]
Scotland 2003 [8]
– 14.6
– 13.7
23.9 7.0
24 7 (primary tumour) 4 (metastasis)
UK 2007 [9] 15.4 13.4
England 2011–2012 [17] – 13.8
New South Wales, Australia 2013 [7]
20.6 4.0
23.0 4.6
– –
– 19
the terminal branches of the model, the impact on the overall estimate was small. Furthermore, sensitivity analysis where high quality data were lacking showed a minor impact of these uncertainties on the overall estimate. Lastly, there are differing recommendations on radiotherapy indication and dose-fractionation schedule for many clinical situations in the treatment guidelines. For some clinical situations, the radiotherapy indications are poorly defined. There are also clinical situations with widely disparate treatment options, one of which could be radiotherapy such as early prostate cancer. Sensitivity analysis showed that the variables had a minor impact on the overall estimate. Our study expanded on the evidence-based RTU model to aid in radiotherapy services planning. RTU and fractionation benchmarks can be evidence-based or criterion-based. RTU criterion-based benchmarking is an empirical method derived from real life observations. The benchmarks are demonstrably achievable, however, the validity of the benchmarks hinges on the validity of the criteria chosen to identify the benchmark population [18]. Evidence-based methods have the advantage of avoiding potential biases associated with criterion-based methods. Furthermore, the RTF model can be readily modified should there be future changes in cancer incidence, stage distribution, radiotherapy indication and dosefractionation schedule, and can be easily adapted for different jurisdictions. Conflict of interest statement There is no conflict of interest. References [1] Borras JM, Lievens Y, Dunscombe P, Coffey M, Malicki J, Corral J, et al. The optimal utilization proportion of external beam radiotherapy in European countries: an ESTRO-HERO analysis. Radiother Oncol 2015. [2] Delaney G, Jacob S, Featherstone C, Barton M. The role of radiotherapy in cancer treatment: estimating optimal utilization from a review of evidencebased clinical guidelines. Cancer 2005;104:1129–37. [3] Barton MB, Jacob S, Shafiq J, Wong K, Thompson SR, Hanna TP, et al. Estimating the demand for radiotherapy from the evidence: a review of changes from 2003 to 2012. Radiother Oncol 2014;112:140–4. [4] Morgan G, Barton M, Crossing S, Bull C, Penman A. A ’Catch Up’ plan for radiotherapy in New South Wales to 2012. J Med Imaging Radiat Oncol 2009;53:419–30. [5] Williams MV, Summers ET, Drinkwater K, Barrett A. Radiotherapy dose fractionation, access and waiting times in the countries of the UK in 2005. Clin Oncol 2007;19:273–86. [6] The Royal College of Radiologists. Radiotherapy dose-fractionation. https:// www.rcr.ac.uk/sites/default/files/publication/Dose-Fractionation_Final.pdf; 2006. Accessed 7/8/2015. [7] Health System Planning and Investment Branch. Radiotherapy treatment services to NSW residents 2013 annual report. Sydney: NSW Ministry of Health; 2014.
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001
K. Wong et al. / Radiotherapy and Oncology xxx (2015) xxx–xxx [8] Scottish Executive Health Department. Radiotherapy Activity Planning for Scotland 2011-2015. http://www.gov.scot/Resource/Doc/90297/0021749.pdf; 2006. Accessed 21/7/2015. [9] Williams MV, Drinkwater KJ. Geographical variation in radiotherapy services across the UK in 2007 and the effect of deprivation. Clin Oncol 2009;21:431–40. [10] Wong K, Delaney GP, Barton MB. Estimation of the optimal number of radiotherapy fractions for breast cancer: a review of the evidence. Radiother Oncol 2015. [11] National Health and Medical Research Council. A guide to the development, implementation and evaluation of clinical practice guidelines. Canberra: Commonwealth of Australia; 1999. [12] Round CE, Williams MV, Mee T, Kirkby NF, Cooper T, Hoskin P. Radiotherapy demand and activity in England 2006-2020. Clin Oncol 2013;25: 522–30. [13] Atun R, Jaffray DA, Barton MB, Bray F, Baumann M, Vikram B, et al. Expanding global access to radiotherapy. Lancet Oncol 2015;16:1153–86.
5
[14] Dixon P, Mackillop W. Could changes in clinical practice reduce waiting lists for radiotherapy? J Health Serv Res Policy 2001;6:70–7. [15] Ruggieri-Pignon S, Pignon T, Marty M, Rodde-Dunet MH, Destembert B, Fritsch B. Infrastructure of radiation oncology in France: a large survey of evolution of external beam radiotherapy practice. Int J Radiat Oncol Biol Phys 2005;61:507–16. [16] Moller TR, Brorsson B, Ceberg J, Frodin JE, Lindholm C, Nylen U, et al. A prospective survey of radiotherapy practice 2001 in Sweden. Acta Oncol 2003;42:387–410. [17] Department of Health. Radiotherapy Services in England 2012. https:// www.gov.uk/government/uploads/system/uploads/attachment_data/file/ 213151/Radiotherapy-Services-in-England-2012.pdf; 2012. Accessed 21/7/ 2015. [18] Mackillop WJ, Kong W, Brundage M, Hanna TP, Zhang-Salomons J, McLaughlin PY, et al. A comparison of evidence-based estimates and empirical benchmarks of the appropriate rate of use of radiation therapy in Ontario. Int J Radiat Oncol Biol Phys 2015;91:1099–107.
Please cite this article in press as: Wong K et al. Evidence-based optimal number of radiotherapy fractions for cancer: A useful tool to estimate radiotherapy demand. Radiother Oncol (2015), http://dx.doi.org/10.1016/j.radonc.2015.12.001