Estimating Optimal Utilization from a Review of Evidence-Based Clinical Guidelines
Geoff Delaney, M.B.B.S, M.D. 1,2 Susannah Jacob, M.B.B.S., M.D., M.H.A. 1 Carolyn Featherstone, M.B.Ch.B. 1 Michael Barton, M.B.B.S. 1,2 1 Collaboration for Cancer Outcomes Research and Evaluation (CCORE), Liverpool Hospital, Sydney, Australia. 2 University of New South Wales, Sydney, Austra- lia. This project was funded by a grant from the Australian Commonwealth Department of Health and Ageing. The authors thank the members of the steering committee of the Australian National Cancer Con- trol Initiative and the forty-two reviewers involved in this project for their comments on the study design and decision trees. Carolyn Featherstones current address: Beatson Oncology Centre, Glasgow, Scotland, United King- dom. Address for reprints: Geoff Delaney, M.B.B.S., M.D., Collaboration for Cancer Outcomes, Re- search and Evaluation, Liverpool Hospital, Locked Bag 7103, Liverpool BC, NSW 1871, Australia; Fax: (011) 61 2 9828 5299; E-mail: Geoff.Delaney@swsahs.nsw.gov.au Received February 14, 2005; revision received April 25, 2005; accepted May 25, 2005. Radiotherapy utilization rates for cancer vary widely internationally. It has previ- ously been suggested that approximately 50% of all cancer patients should receive radiation. However, this estimate was not evidence-based. The aim of this study was to estimate the ideal proportion of new cases of cancer that should receive radiotherapy at least once during the course of their illness based on the best available evidence. An optimal radiotherapy utilization tree was constructed for each cancer based upon indications for radiotherapy taken from evidence-based treatment guidelines. The proportion of patients with clinical attributes that indi- cated a possible benet from radiotherapy was obtained by adding epidemiologic data to the radiotherapy utilization tree. The optimal proportion of patients with cancer that should receive radiotherapy was then calculated using TreeAge (Tree- Age Software, Williamstown, MA) software. Sensitivity analyses using univariate analysis and Monte Carlo simulations were performed. The proportion of patients with cancer in whom external beam radiotherapy is indicated according to the best available evidence was calculated to be 52%. Monte Carlo analysis indicated that the 95% condence limits were from 51.7% to 53.1%. The tightness of the con- dence interval suggests that the overall estimate is robust. Comparison with actual radiotherapy utilization data suggests a shortfall in actual radiotherapy delivery. This methodology allows comparison of optimal rates with actual rates to identify areas where improvements in the evidence-based use of radiotherapy can be made. It provides valuable data for radiotherapy service planning. Actual rates need to be addressed to ensure better radiotherapy utilization. Cancer 2005;104: 112937. 2005 American Cancer Society. KEYWORDS: radiotherapy utilization, cancer, evidence-based. T he planning of efcient and equitable treatment services for a population requires a rational and defensible estimate of demand. This has particular relevance for planning services that require sig- nicant capital expenditure such as radiotherapy. Radiotherapy is an essential mode of cancer treatment and contributes to the cure or palliation of many cancer patients. Radiotherapy facilities have high capital costs and their operation is staff intensive. In this project, we have undertaken to calculate a rational estimate of need for radio- therapy, based on the occurrence of each type of cancer, the evi- dence-based indication for radiotherapy in the treatment of each type of cancer, and the probability that radiotherapy will be chosen as a form of treatment. The radiotherapy utilization rate is dened as the proportion of a dened population of patients with cancer that receives at least one course of radiotherapy during their lifetime. Previous reports from Australian Commonwealth and State agencies have proposed that 50% of all new cases of notiable cancer in Australia should be treated 1129 2005 American Cancer Society DOI 10.1002/cncr.21324 Published online 3 August 2005 in Wiley InterScience (www.interscience.wiley.com). with external beam radiotherapy. 15 (Notiable can- cers are cancers for which registry data are available.) Although this gure is based almost entirely on expert opinion, it is currently accepted as the guide for esti- mating utilization and is used to plan for the distribu- tion and number of linear accelerators. However, its validity is questionable, and it is not responsive to changing clinical indications. There are signicant variations in actual radiotherapy utilization rates re- ported in Australia, the United States, Canada, and the Nordic countries, where utilization ranges from 20 55% of all new cancer cases. 611 These variations stress the importance of using rigorous evidence-based methods to estimate an optimal radiotherapy utiliza- tion rate that can act as a benchmark against which actual utilization rates can be compared. This report estimates an ideal rate of radiotherapy utilization for cancer in Australia based on the inci- dence of each type of cancer, the evidence-based in- dication for radiotherapy in the treatment of that can- cer, and the proportion of cancer patients included in that indication for radiotherapy. The authors of the current study have previously published optimal radiotherapy utilization rates for breast carcinoma, 12 lung carcinoma, 13 melanoma, 14 gastrointestinal cancers, 15 genitourinary cancers, 16 head and neck cancers, 17 gynecologic cancers, 18,19 he- matologic malignancies, 20,21 central nervous system tumors, unknown primary cancers, and thyroid carci- nomas. 22 This article reports the estimated overall op- timal radiotherapy utilization rate for all registered cancers in Australia and compares the optimal rate with known actual rates of radiotherapy utilization. Objectives The objectives of this study were: To estimate the ideal proportion of new cases of no- tiable cancer that should receive megavoltage external-beam radiotherapy at some time during the course of their illness using the best available evidence. To develop a model of radiotherapy utilization that can be used to estimate the effect of future changes in the relative distribution of tumor sites, changes in stage at presentation, and changes in indications for radiotherapy on the optimal radio- therapy utilization rate. To compare the estimated optimal rates with actual rates of radiotherapy use. MATERIALS AND METHODS In this study, an indication for radiotherapy was dened as a clinical situation in 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 radiother- apy based on an assessment of performance status indicators and the presence or absence of comorbidi- ties. The superiority of radiotherapy over other treat- ment options could be because of better survival, local control, or toxicity proles. The study was limited to all notiable cancers with an incidence of 1% of the Australian cancer population. Notiable cancers are cancers for which registry data are available. In Aus- tralia this includes ductal carcinoma in situ of the breast but does not include nonmelanomatous skin cancers and benign tumors. The indications for radiotherapy for each cancer site were derived from treatment guidelines issued by national and international institutions or specialist groups and published (including on the Internet) be- fore December 2003. If guidelines did not exist for particular cancer types and tumor sites, or where the guidelines did not adequately address radiotherapy use, other sources of evidence were identied. These included treatment reviews, randomized controlled trials, population-based studies of care, and single- institution studies. The evidence for indications for radiotherapy was classied using the Australian National Health and Medical Research Council (NHMRC) hierarchy of lev- els of evidence (Table 1), with only the highest level of evidence being used for each indication for radiother- apy. 23 As our purpose was to make recommendations for radiotherapy services in Australia, the highest pri- ority was given to Australian evidence-based clinical practice guidelines issued by national institutions such as the NHMRC or the National Breast Cancer Centre. If these did not exist, then guidelines from other countries were used wherever possible. Radiotherapy utilization trees for individual can- cer sites were constructed based upon the treatment TABLE 1 Levels of Evidence for Indications for Radiotherapy 23 Level of evidence Description I Systematic review of all relevant randomized studies II At least 1 properly conducted randomized trial III Well designed controlled trials without randomization a IV Case series a These include trials with pseudo-randomization where a awed randomization method occurred (e.g., alternate allocation of treatments) or comparative studies with either comparative or historical controls. 1130 CANCER September 15, 2005 / Volume 104 / Number 6 recommendations obtained from evidence-based treatment guidelines. We used decision analysis soft- ware (TreeAge Data version 3.5, TreeAge Software, Williamstown, MA) to illustrate the indications for ra- diotherapy in a diagrammatic form (as a tree), to per- form basic calculations such as multiplication of fac- tors and summation of the results, and to perform sensitivity analyses of variability. Parameters can be readily adjusted in the tree if indications for radiother- apy or epidemiologic data distributions change in the future and the software can then rapidly calculate the adjusted utilization rates. The utilization trees depict the clinical conditions for which radiotherapy is indicated. Each terminal branch of the tree shows whether or not radiotherapy is recommended for a particular type of cancer in individuals with specic clinical attributes. In some circumstances, the indication for radiotherapy oc- curred in the initial stages of management. In other circumstances, radiotherapy was given later in the disease course (for instance, in patients who devel- oped a local recurrence and who had not previously had an indication for treatment with radiotherapy). Similar methodology has been used by others. 6,2426 This is the rst published report of an analysis of all cancers. The purpose of our project was to determine the proportion of all cancer patients who have at least one indication for radiotherapy at some time in the course of their illness. Patients requiring radiotherapy were therefore counted only once, even if they had multiple indications at different stages in their illness. The radiotherapy utilization trees also depict the proportion of patients represented by each branch point of the tree. The relative quality of epidemiologic data from various sources was ranked according to a scoring system that gave greatest importance to Aus- tralian population-based data. Population-based datasets from other countries were also used. Popula- tion-based databases were preferred because they were considered less likely to be affected by referral or selection bias (compared with hospital-based data- bases) and, therefore, were more likely to be represen- tative of the entire population of patients with cancer. Table 2 shows the hierarchy of quality of epidemio- logic data used. The proportion of patients for whom radiotherapy would be recommended was calculated for each can- cer site by calculating the frequency of each indication for radiotherapy and then summing the frequencies to give the total optimal rate of use. The overall optimal radiotherapy utilization rate was calculated by sum- ming the optimal utilization rates derived for each cancer site, calculated as a proportion of all cancers. As this project involved determining estimates for optimal radiotherapy utilization for all notiable can- cers with an incidence of 1%, the remaining cancers that have an incidence of 1% have been called other cancers in the radiotherapy utilization tree and comprise 2% of the entire cancer population ac- cording to the Australian Institute of Health and Wel- fare report. 27 These cancers include pediatric cancers, sarcomas of soft tissue and bone, cancers of the me- diastinum, orbit, peritoneum, retroperitoneum, penis, and pleura as well as other rare malignancies. Some of these malignancies are commonly treated with radio- therapy (such as soft tissue sarcomas), and others are rarely treated with radiation (e.g., peritoneal and pleu- ral tumors). For the purpose of the current study, specic radiotherapy utilization trees were not con- structed for each of these uncommon cancers. We assumed that the requirement for radiotherapy for this miscellaneous group was 50% and then per- formed sensitivity analysis where the use of radiother- apy for other cancers ranges between 0% and 100%. This is included in the sensitivity analysis performed for the entire radiotherapy decision tree and is de- scribed later. For some branches of the trees, there was a rela- tive lack of high quality epidemiologic data, and, for some other branches, epidemiologic data differed sig- nicantly across different data sources of equal qual- ity. Monte Carlo simulations were performed to assess the impact on the radiotherapy utilization rate that would result from variations in epidemiologic data, different probabilities of benet from treatment, or uncertainty in the indication for radiotherapy. Monte Carlo simulations are based upon random sampling of variables from discrete and continuous distributions using individual trial data. Observing the statistical TABLE 2 Hierarchy of Epidemiologic Data a Quality of source Source Type Australian National Epidemiological data Australian State Cancer Registry Epidemiologic databases from other large international groups (e.g. SEER) Results from reports of a random sample from a population Comprehensive multiinstitution database Comprehensive single-institution database Multiinstitution reports on selected groups (e.g. multiinstitution clinical trials) Single-institution reports on selected groups of cases Expert opinion a Modied from Tyledsley et al. 6 Radiotherapy Utilization in Cancer/Delaney et al. 1131 properties of many trials using random sampled val- ues allows additional insight into performance of a model. The main weakness of the Monte Carlo anal- ysis, in the current study, is that the relative impor- tance of all of data used is weighted by study size and may not necessarily be ranked by study quality, which was impossible to assess for each dataset. Funding and Peer Review The project was funded by the Department of Health and Ageing of the Australian Government and super- vised by the National Cancer Control Initiative (NCCI). An expert steering committee was convened for this project by the NCCI with representation from major nongovernmental cancer organizations, consumers, epidemiologists, radiation and medical oncologists, surgeons, palliative care specialists, and experts in evidence and treatment guidelines. A multidisciplinary panel of expert reviewers was established, comprising ninety-one nationally recog- nized oncology experts from the elds of medical, surgical, and radiation oncology, palliative care, and oncology nursing. Forty-two of these reviewers pro- vided comments, and 43% of reviewers were from nonradiation oncology specialties also commented. Comparison with Actual Radiotherapy Utilization Rates Actual radiotherapy utilization rates were obtained from published and unpublished sources covering the years 1990 to 2001. These actual rates were tabulated and compared with estimated optimal radiotherapy utilization rates. RESULTS Recommended optimal radiotherapy utilization rates and optimal radiotherapy utilization trees for breast, 12 lung, 13 skin (melanoma), 14 genitourinary, 16 gastroin- testinal, 15 gynecologic, 18,19 and head and neck can- cers, 17 hematologic malignancies, 20,21 and central ner- vous system, thyroid, and unknown primary site tumors 22 have been reported in detail elsewhere. A summary of the calculated ideal radiotherapy utiliza- tion rates for the various tumor sites are presented in Table 3 along with the proportion of cancer that each tumor site composes in Australia. Overall Optimal Radiotherapy Utilization Rate The optimal radiotherapy utilization rates in Table 3 varied from a low recommended rate of 0% for liver cancer patients to a high rate of 92% for patients with central nervous system tumors. The recommended TABLE 3 Optimal Radiotherapy Utilization Rate by Cancer Type Tumor type Proportion of all cancers Proportion of patients receiving radiotherapy Patients receiving radiotherapy (% of all cancers) Reference Breast 0.13 83 10.8 Delaney et al. 12 Lung 0.10 76 7.6 Delaney et al. 13 Melanoma 0.11 23 2.5 Delaney et al. 14 Prostate 0.12 60 7.2 Delaney et al. 16 Gynecologic 0.05 35 1.8 Delaney et al. 18,19 Colon 0.09 14 1.3 Delaney et al. 15 Rectum 0.05 61 3.1 Delaney et al. 15 Head and neck 0.04 78 3.1 Delaney et al. 17 Gall bladder 0.01 13 0.1 Delaney et al. 15 Liver 0.01 0 0.0 Delaney et al. 15 Esophageal 0.01 80 0.8 Delaney et al. 15 Stomach 0.02 68 1.4 Delaney et al. 15 Pancreas 0.02 57 1.1 Delaney et al. 15 Lymphoma 0.04 65 2.6 Featherstone et al. 20 Leukemia 0.03 4 0.1 Featherstone et al. 21 Myeloma 0.01 38 0.4 Featherstone et al. 21 Central nervous system 0.02 92 1.8 Delaney et al. 22 Renal 0.03 27 0.8 Delaney et al. 16 Bladder 0.03 58 1.7 Delaney et al. 16 Testis 0.01 49 0.5 Delaney et al. 16 Thyroid 0.01 10 0.1 Delaney et al. 22 Unknown primary 0.04 61 2.4 Delaney et al. 22 Other 0.02 50 1.0 See citations in text Total 1.00 - 52.3 1132 CANCER September 15, 2005 / Volume 104 / Number 6 overall optimal radiotherapy utilization rate was cal- culated to be 52.3%. Sensitivity Analysis Some variables in the utilization tree were associated with signicant uncertainties, which can be catego- rized as follows: 1. Uncertainty in the data where the values of epi- demiologic data obtained from multiple sources differed signicantly. Typically these were near the terminal ends of the tree where large studies on incidence rates were lacking. 2. Uncertainty in the indication for radiotherapy where guidelines had no specic criteria, or con- icting criteria, for consideration of radiotherapy. For example, one guideline for breast cancer rec- ommended radiotherapy for postmastectomy pa- tients with 3 axillary nodes involved, but also advocated consideration of radiotherapy in all patients with nodal involvement. 28 Other guide- lines either mention that radiotherapy in patients with involvement of less than 4 axillary lymph nodes is controversial 29 or avoid the issue com- pletely. 30 3. Uncertainty in the choice between radiotherapy and other treatment options of equal efcacy, such as surgery, observation, or radiotherapy for localized prostate adenocarcinoma. The actual branchpoints where these uncertain- ties existed have been described in reports of each specic cancer sites radiotherapy utilization trees. Sensitivity analysis allows an assessment of the effect that data uncertainty may have on the overall radio- therapy utilization estimate. Two different types of sensitivity analyses were performed. One-way sensi- tivity analyses allowed assessment of the effect of varying the value of each variable on the overall model in a univariate fashion. One-way sensitivities were presented for each of the decision trees where uncer- tainty existed and are not repeated here. 1217,1922 For a more global multivariate-type assessment, Monte Carlo simulations can be performed to assess the ef- fect of multiple uncertainties on the overall radiother- apy utilization rate. Monte Carlo simulations are based upon random sampling of variables from dis- crete and continuous distributions using individual trial data. Multivariate sensitivity analysis using Monte Carlo analysis on 10 4 simulations indicates that the 95% condence limits for our optimal radiotherapy estimate were 51.7% and 53.1%. Comparison with Actual Radiotherapy Utilization Rates Table 4 shows the actual rates of radiotherapy utiliza- tion from population-based reports from Sweden, the United Kingdom, the United States and some national and state patterns of care studies in Australia. 5,3139 DISCUSSION We have used an evidence-based technique to calcu- late an overall estimate of optimal radiotherapy utili- zation of 52.3% for all notiable cancer in Australia. This nal estimate is remarkably precise (as measured by the tight condence limits) despite uncertainty ex- isting in relation to data for some indications for ra- diotherapy and occasional uncertainty between treat- ment options of approximately equal efcacy. The tight condence interval may be explained by the fact that good quality data existed for the initial branches of the tree (for example, data such as tumor type and stage at presentation). Most of the uncertainty existed in the distal or near-terminal branches of the tree and, therefore, affected only very small proportions of the cancer population and had little effect on the overall estimate. In addition, the effect of these variations was such that some would increase the overall utilization rate whereas others would reduce it, so that, to a large extent, they cancelled out each other. The model of radiotherapy utilization developed in this project has many benets. 1. It provides a benchmark for planning radiotherapy services on a population basis.The results from this study can be useful in the planning of appropriate radiotherapy services for a given population using the following calculations. For every 1000 cancer cases in a population, 523 patients would need radiation as an optimal part of their management based upon the results of this project (calculated optimal radiotherapy utilization rate of 52.3%). A further 120 patients, of the above 523 patients, will require retreatment (based upon an ac- tual retreatment rate of 23%). 40 This means that an estimated 643 courses of treatment will be required for every 1000 cancer patients diagnosed with a registered cancer. These calculations are summarized in Table 5. This will allow population-based estimates for the number of possible treatment courses (and, therefore, the amount of resources) that should be provided for any particular area. This study was performed for the Australian government and was viewed as a study that would provide an evidence-based planning target. Radiotherapy Utilization in Cancer/Delaney et al. 1133 2. Modeling the effect that changes to a particular cancer incidence or changes in stage distribution have on the overall recommended radiotherapy utilization rate is another benet. It is possible to easily modify the model should there be changes in the relative incidence of certain cancers, a change in the stage distribution, or a change in treatment recommendations. The model can also be used to estimate the optimal radiotherapy utilization rate for other countries that may have dif- TABLE 4 Comparison of Optimal with Actual Radiotherapy Utilization Rates Cancer site % Optimal radiotherapy utilization rate Actual radiotherapy utilization rates % Sweden National 2001 31 % USA % UK (NYCRIS) 1999 34 % Australia SEER 1995 2000 32 ACS a 2001 33 National 1995 35 2000 36 NSW 2000 5 VIC 2000 37 1993 38 SA a 1990 1994 39 Breast cancer 83 81 42 44 54 41 71 24 40 Lung cancer 76 71 39 36 - - 49 44 38 Melanoma 23 23 2 1 - - 13 - 2 Prostate 60 51 27 41 16 - - - 44 Kidney 27 63 8 4 9 - - - 11 Urinary bladder 58 17 4 3 26 - - - 26 Testis 49 48 40 - NR - - - 43 Esophagus 80 73 54 - 31 - - - 47 Stomach 68 7 15 - 4 - - - 6 Pancreas 57 6 16 - 4 - - - 4 Liver 0 - 3 - 3 - - - 3 Gall bladder 13 9 14 - 9 - - - 5 Colon 14 6 2 1 2 3 - - 3 Rectum 61 56 40 41 33 38 - - 17 Oral cavity 74 94 b NR - - - - - 44 c Lip 20 22 8 - - - - - 2 Larynx 100 100 75 - - - - - 80 Oropharynx 100 100 70 - - - - - - Salivary gland 87 60 55 - - - - - - Hypopharynx 100 39 74 - - - - - - Paranasal sinuses 100 100 NR - - - - - - Nasopharynx 100 100 84 - - - - - - Unknown primary (head & neck) 90 NR - - - - - - - Uterus 46 64 22 25 - - - - 26 Cervix 58 83 44 33 - - - - 41 Central nervous system 92 37 59 - - - - - 52 Lymphoma 65 40 - - - - - - 24 Leukemia 4 8 - - - - - - 6 Myeloma 38 82 - - - - - - 34 All cancers 52 43 24 - - - - - 25 NR: Not reported; ACS: American College of Surgeons; SEER: Surveillance, Epidemiology and End Results database (National Cancer Institute); NYCRIS: Northern and Yorkshire Cancer Registry and Information Service; NSW: the state of New South Wales; VIC: the state of Victoria; SA: the state of South Australia. a First treatment only. b Includes brachytherapy. c Includes salivary glands. TABLE 5 Estimated Optimal Number of Courses of Treatment per 1000 Registered Cancers Percentage Total no. New registered cancers N/A 1000 Patients requiring radiation 52.3 523 Retreatments 23 120 Total number of courses of radiotherapy required 643 1134 CANCER September 15, 2005 / Volume 104 / Number 6 fering cancer-specic proportions. For example, if an- other country with a very different cancer incidence prole were to use the model, then the only require- ment to recalculate the optimal radiotherapy utiliza- tion rate would be to alter the incidence of each of the cancers in the tree and recalculate. Similarly, a change in stage distribution of cancer due to development of superior staging investigations (such as positron emis- sion tomography in nonsmall cell lung cancer), or following the introduction of a screening program could easily be incorporated into the model. 3. This model provides a benchmark for service deliv- ery. The radiotherapy utilization trees that have been developed for each of the tumor sites are a diagram- matic representation of optimal evidence-based can- cer care from a radiotherapy perspective. Epidemio- logic data from patterns of care studies will allow comparisons to be made between the actual rates of radiotherapy delivery and the evidence-based ideal rate. Analysis of the distributions of tumor stage, his- tology, age, performance status, and other factors will better dene any discrepancy between the actual and ideal utilization rates. Table 4 compares optimal radiotherapy utilization rates with available rates of actual radiotherapy utili- zation obtained from population-based data. The ta- ble highlights the paucity of the data on actual radio- therapy utilization, the high variability of the actual radiotherapy rates across different regions, and the general shortfall in radiotherapy use for most major tumor sites including the common tumor sites that have well known evidence-based treatment guidelines (e.g., breast cancer). These data are not subdivided by the various stages or other clinical attributes that would make a direct comparison between the optimal trees and the actual practice, although future studies may be designed to identify subgroups of patients so that, when shortfalls in radiotherapy are identied, specic details as to types of patients where the short- falls are greatest may help direct quality improvement programs to the areas of most gain. 4. This model can determine optimal rates and re- sources for other treatment modalities. The methodology used here could be readily adapted to consider other treatments (such as surgery, chemotherapy, or palliative care) for cancer. It could also be used to plan other services if criteria for se- lecting appropriate patients were known for that par- ticular service. For instance, if we knew the factors that predict the need for palliative care referral, or cancer genetics services, then resource planning could be assisted by calculating the optimal utilization rate in a similar fashion to that described here for radio- therapy. 5. This model may be used to predict future radio- therapy workload. The radiotherapy utilization tree predicts whether patients should receive any radiotherapy but does not assess whether the treatment intent would be pallia- tive or radical, and the tree predicts neither the num- ber of fractions of treatment required nor the com- plexity of the patients care. Various models of complexity have been reported in the literature that may be used in future studies so that even more ac- curate predictions of radiotherapy workload could be determined by calculating the actual number of treat- ment fractions that may be expected for a given pop- ulation. Some Limitations of the Study Were Identied Quality of data The current study has identied areas where good quality epidemiologic data (based on stage, perfor- mance status, etc.) were lacking. We have overcome the problem by performing modeling and sensitivity analyses to indicate the relatively minor effect that any of these uncertainties could have on overall utilization rate. 1217,1922 Skin cancer and benign diseases provide workload for radiation oncology departments but are not included as registered cancers and, therefore, have not been factored into the model Notiable cancers are cancers for which statutory re- quirements exist to notify a state cancer registry. Stat- utory notication in Australia excludes nonmelano- matous skin cancers and benign tumors but includes ductal carcinoma in situ of the breast. A limitation of the study is that there are other uses for radiotherapy that are not included in this estimate and that will need consideration when planning radiotherapy re- sources. Radiotherapy has an established role in man- agement of nonmalignant conditions (benign tumors and noncancerous conditions) as well as a role in the management of nonregistered cancers such as non- melanomatous skin cancers. The overall need for ra- diotherapy resources is difcult to estimate for these nonregistered conditions, as the overall incidence of these conditions is unknown, and evidence-based treatment guidelines do not exist for most of these conditions. Data obtained from selected hospitals in Australia show that around 11% of patients who re- ceive external beam radiotherapy are treated for non- Radiotherapy Utilization in Cancer/Delaney et al. 1135 notiable conditions. 41 It remains important to con- sider this additional workload in resource planning. Other forms of radiotherapy have not been considered Inclusion of other forms of radiotherapy such as brachytherapy (interstitial and intracavitary) and/or with radioactive isotopes (iodine, yttrium, samarium, strontium, etc.) are beyond the scope of this article. However, these other forms of radiotherapy should be considered when planning radiotherapy resources and could be the subject of further study. Controversies in the recommended use of radiotherapy Despite using treatment guidelines to determine indi- cations for radiotherapy, there are many areas where the role of radiotherapy remains poorly dened or where the indications for the use of radiotherapy re- main vague. This is mainly due to poor evidence and the lack of good quality trials. We have identied some areas where future research would be useful. The model is easily amended should new evidence for or against the use of radiotherapy for a specic clinical situation emerge. The effect of patient choice considerations We did not consider the effect of patient choice be- cause of the risk that the studies reporting patient preference might have been confounded by availabil- ity of radiotherapy to the study population. Little or no data are presented in these studies to judge whether access to resources was factored into the decision for or against radiotherapy when alternative treatment options were available. Rare indications for radiotherapy have not been included in the overall estimate In many cancers there will be a small proportion of patients who may appropriately receive radiotherapy for rare indications, usually for metastases such as symptomatic lung, soft tissue, or subcutaneous metas- tases. They were not included because incidence data were not available and the proportions of patients with symptoms that required radiotherapy could not be estimated. Although only of small overall impact in their own right, the cumulative total of these indica- tions could increase the overall radiotherapy utiliza- tion estimate by 12% at the most. Conclusions The overall estimate for radiotherapy utilization is 52.3% based upon the best available evidence. Al- though the scope of this study is conned to exploring the optimal utilization of external beam megavoltage radiotherapy for notiable cancers, the overall esti- mate provides a useful tool for assisting in planning adequate radiotherapy resources. Population-based data from the United States, the United Kingdom, Sweden, and Australia suggest that there is a signi- cant shortfall between the optimal rate and the pro- portion of patients currently treated with radiotherapy that warrants further research and action. REFERENCES 1. National Health and Medical Research Council. Beam and isotope radiotherapyA report of the Australian Health Technology Advisory Committee. Publication no. 2036. Can- berra: Commonwealth Department of Health and Family Services, 1996. 2. Statewide Services Development Branch. Radiotherapy management information system. State health publication no.(SSDB) 980139. Sydney: NSW Health Department, 1997. 3. Statewide Services Development Branch. Radiotherapy management information system. State health publication no. (SSDB) 970069. Sydney: NSW Health Department, 1996. 4. Statewide Services Development Branch. Radiotherapy management information system. State health publication no. (SSDB) 980139. Sydney: NSW Health Department, 1998. 5. Statewide Services Development Branch. NSW radiotherapy management information system report 2000. Sydney: NSW Health Department, 2001. 6. Tyldesley S, Boyd C, Shulze K, Walker H, Mackillop WJ. Estimating the need for radiotherapy for lung cancer: an evidence-based, epidemiologic approach. Int J Radiat Oncol Biol Phys. 2001;49:973985. 7. Mackillop WJ, Dixon P, Zhou Y, et al. Variations in the management and outcome of non-small cell lung cancer in Ontario. Radiother Oncol. 1994;32:105115. 8. Mackillop WJ, Groome PA, Zhang-Solomons J, et al. Does a centralized radiotherapy system provide adequate access in care? J Clin Oncol. 1997;15:12611271. 9. Lote K, Moller T, Nordman E, et al. Resources and produc- tivity in radiation oncology in Denmark, Finland, Iceland, Norway and Sweden during 1987. Acta Oncol. 1991;30:555 561. 10. Denham JW. How do we bring an acceptable level of radio- therapy services to a dispersed population? Australas Radiol. 1995;39:171173. 11. Barton MB. Radiotherapy utilisation in New South Wales from 1996 to 1998. Australas Radiol. 2000;44:483484. 12. Delaney G, Barton B, Jacob S. Estimation of an optimal radiotherapy utilization rate for breast carcinoma: a review of the evidence. Cancer. 2003;98:19771986. 13. Delaney G, Barton M, Jacob S, Jalaludin B. A model for decision making for the use of radiotherapy in lung cancer. Lancet Oncol. 2003;4:120128. 14. Delaney G, Barton M, Jacob S. Estimation of an optimal radiotherapy utilization rate for melanoma. A review of the evidence. Cancer. 2004;100:12931301. 15. Delaney G, Barton M, Jacob S. Estimation of an optimal radiotherapy utilization rate for gastrointestinal cancer: A review of the evidence. Cancer. 2004;101:657670. 16. Delaney G, Jacob S, Barton M. Estimating the optimal ex- ternal beam radiotherapy utilization rate for genitourinary malignancies. Cancer. 2005;103:462473. 17. Delaney G, Jacob S, Barton M. Estimation of an optimal external beam radiotherapy utilization rate for head and neck carcinoma. Cancer. 2005;103:22162227. 1136 CANCER September 15, 2005 / Volume 104 / Number 6 18. Delaney G, Jacob S, Barton M. Estimation of an optimal radiotherapy utilization rate for gynecologic cancer: part I-malignancies of the cervix, ovary, vagina, and vulva. Can- cer. 2004;101:671681. 19. Delaney G, Jacob S, Barton M. Estimation of an optimal radiotherapy utilization rate for gynecologic cancer: part II-carcinoma of the endometrium. Cancer. 2004;101:682 692. 20. Featherstone C, Delaney G, Jacob S, Barton M. Estimating the optimal utilization rates of radiotherapy for hematologic malignancies from a review of the evidence: part I-lym- phoma. Cancer. 2005;103:383392. 21. Featherstone C, Delaney G, Jacob S, Barton M. Estimating the optimal utilization rates of radiotherapy for hematologic malignancies from a review of the evidence: part II-leuke- mia and myeloma. Cancer. 2005;103:393401. 22. Delaney G, Jacob S, Barton M. Estimating the optimal ra- diotherapy utilization for cancer of the central nervous sys- tem, thyroid cancer, and cancer of unknown primary origin from evidence-based clinical guidelines. Cancer. in press. 23. National Health and Medical Research Council. Guide to the development, implementation and evaluation of clinical practice guidelines. Appendix B, 56. Canberra: National Health and Medical Research Council, 1998. 24. Foroudi F, Tyldesley S, Walker H, Mackillop WJ. An evi- dence-based estimate of appropriate radiotherapy utiliza- tion rate for breast cancer. Int J Radiat Oncol Biol Phys. 2002;53:12401253. 25. Foroudi F, Tyldesley S, Barbera L, Huang J, Mackillop WJ. An evidence-based estimate of the appropriate radiotherapy utilization rate for colorectal cancer. Int J Radiat Oncol Biol Phys. 2003;56:12951307. 26. Foroudi F, Tyldesley S, Barbera L, Huang J, Mackillop WJ. Evidence-based estimate of appropriate radiotherapy utili- zation rate for prostate cancer. Int J Radiat Oncol Biol Phys. 2003;55:5163. 27. Australian Institute of Health and Welfare (AIHW) and Aus- tralasian Association of Cancer Registries (AACR). Cancer in Australia 1998. CAN 12. 2001. Cancer Series No 17. Can- berra: AIHW and AACR, 2001. 28. National Comprehensive Cancer Network. National practice guidelines in oncology-breast cancer. Version 2. Available from URL: www.nccn.org. 2002. [Accessed February 21, 2003]. 29. National Cancer Institute. PDQ cancer information summa- ries: Treatment of breast cancer. Available from URL: www.nci.nih.gov. 2003. [Accessed February21, 2003]. 30. NHMRC National Breast Cancer Centre. Clinical practice guidelines for the management of advanced breast cancer. National Health and Medical Research Council. Kings Cross: NHMRC National Breast Cancer Centre, 2001. 31. Moller TR, Brorsson B, Ceberg J, et al. A prospective survey of radiotherapy practice 2001 in Sweden. Acta Oncol. 2003; 42:387410. 32. National Cancer Institute (Cancer Statistics Branch). SEER Stat 5.0. Surveillance, Epidemiology and End Results cancer incidence public-use database, 19732000. Bethesda: US Department of Health and Human Services, 2002. 33. American College of Surgeons. National Cancer Database. Available from URL: http://www.facs.org.dept/cancer/ ncdb/whatsncdb.html [accessed on November 25, 2003]. 34. Northern and Yorkshire Cancer Registry and Information Service (NYCRIS). Northern and Yorkshire Cancer Networks. A report on incidence and management for the main sites of cancer 1999. Available from URL: http://www.nycris. org.uk/. 2002. [Accessed on November 25, 2003]. 35. Hill D, Jamrozik K, White V, Collins J, et al. Surgical Man- agement of Breast Cancer in Australia in 1995. Kings Cross, NHMRC National Breast Cancer Centre, 1999. 36. Clinical Governance Unit. The National Colorectal Cancer Care Survey. Australian clinical practice in 2000. Melbourne: National Cancer Control Initiative, 2002. 37. Hill DJ, White VM, Giles GG, Collins JP, Kitchen PR. Changes in the investigation and management of primary operable breast cancer in Victoria. Med J Aust. 1994;161:110118. 38. Richardson GE, Thurseld VJ, Giles GG. Reported manage- ment of lung cancer in Victoria in 1993: comparison with best practice. MJA. 2000;172:321324. 39. Luke C, Chapman P, Priest K, Roder D. Use of radiotherapy in the primary treatment of cancer in South Australia. Aus- tralas Radiol. 2003;47:161167. 40. Statewide Services Development Branch. Radiotherapy management information system report 2003. Sydney: NSW Health Department, 2004. 41. Barton M, Frommer M, Olver I, Cox C, Crowe P, et al. A cancer services framework for Victoria and future directions for the Peter MacCallum Cancer Institute. Available from URL: http://www.health.vic.gov.au/cancer/docs/vcsnalre- port.pdf. 2003. [Accessed January 10, 2005]. Radiotherapy Utilization in Cancer/Delaney et al. 1137