Table 2. Studies comparing cost of IMRT and 3D-CRT.

Study	Patients, no.	Type of study	Data source	Comparison	Cost definition	Costs included (direct vs indirect)	Perspective	Main findings including comment on risk of bias
Studies from the payer’s perspective
Konski (2005) [41]	NA	Cost modeling (10-yr timeline)	Hospital billing data, patient questionnaires	IMRT versus 3D-CRT	Costs modeled in decision tree analyses (Markov model)	Direct costs (Medicare reimbursements); did not look at copays or deductibles	Payer (Medicare)	70 yr old with intermediate risk Mean IMRT cost: $33 837 (2004) Mean 3D-CRT cost: $21 377 (2004) Difference: $12 460 (2004) ICER: $16 182/QALY 70 yr old with good risk Mean IMRT cost: $31 950 (2004) Mean 3D-CRT cost: $19 213 (2004) Difference: $12 737 (2004) ICER: $17 448/QALY Sensitivity analysis found that a longer time horizon and younger age favorably impacted the cost-effectiveness ratio. Probabilistic sensitivity analysis was performed to address uncertainty concerning cost, transition probabilities, and utilities using a second-order Monte Carlo simulation. Risk of bias moderate due to uncertainty of the data abstracted from the literature.
Konski (2006) [83]	NA	Cost modeling	Literature review, patient questionnaires, or preference at institution	IMRT versus 3D-CRT	Costs modeled in decision tree analyses (Markov model)	Direct costs	Payer (Medicare)	70 yr old with intermediate-risk prostate cancer Mean IMRT cost: $47 931 (2004) Mean 3D-CRT cost: $21 865 (2004) Difference: $26 066 (2004) ICER: $40 101/QALY IMRT has ~53% probability of being cost effective if the ICER for deciding cost effectiveness is $50 000/QALY. The results, however, are dependent on the assumptions of improved biochemical disease-free survival with fewer patients undergoing salvage therapy and improved QOL after treatment. The results are sensitive to the time horizon of the analysis and the utilities used to inform the model. Probabilistic sensitivity analysis was performed to address uncertainty concerning cost, transition probabilities, and utilities using a second-order Monte Carlo simulation. Risk of bias moderate due to uncertainty of the data abstracted from the literature.
Nguyen (2011) [39]	N not given for IMRT and 3D-CRT	Retrospective cohort study	SEER-Medicare 2002–2005	IMRT versus 3D-CRT	Medicare payments in the year after diagnosis minus those in the year prior to diagnosis	Medicare payments (did not look at indirect costs)	Payer (Medicare)	Mean IMRT cost in 2002:$37 125 (2008) Mean 3D-CRT cost in 2002: $22 384 (2008) Difference: $14 741 (2008) Mean IMRT cost in 2005: $31 574 (2008) Mean 3D-CRT cost in 2005: $20 588 (2008) Difference: $10 986 (2008) Risk of bias high. No adjusted models were constructed. IMRT patients were younger, healthier, more often nonwhite; live in higher educated and more urban areas; live in areas with higher median incomes; do not live in the south or midwest; and are less likely to have higher stage disease.
Hummel (2012) [84]	NA	Cost modeling (lifetime perspective)	Data from systematic review	IMRT versus 3D-CRT	Costs modeled in decision tree analyses	Direct and indirect costs	UK National Health Service perspective	Baseline age 70 yr Mean IMRT cost: £5921 (2009) Mean 3D-CRT cost: £4799 (2009) Difference: £1122 (2009) In scenarios where survival is greater with IMRT, IMRT was cost effective (ICER <£20 000). When only a difference in late GI toxicity was assumed, the ICER was highly sensitive to uncertain model parameters. Univariate sensitivity analyses performed on key parameters, such as age, incremental cost of IMRT in comparison with 3D-CRT, and duration of late GI toxicity. Limitation: There are limited clinical data comparing IMRT and 3D-CRT. Risk of bias moderate due to uncertainty of the data abstracted from the literature.
Yong (2012) [40]	NA	Cost modeling (lifetime time horizon)	Data from systematic review and survey from radiation oncologists and physicists for estimated pretreatment preparation time	IMRT versus 3D-CRT	Costs estimated through activity-based costing with input from radiation oncologists, physicists, and treatment planners (Markov model)	Direct costs plus cost of equipment (capital, construction, maintenance and operating costs), cost of supplies, personnel, and overhead costs	Payer (Canadian health care system)	Baseline age 70 yr Mean IMRT cost: C$14 520 (2009) Mean 3D-CRT cost: C$13 509 (2009) Difference: C$1019 (2009) ICER: C$26 768/QALY Assumes equal biochemical survival between IMRT and 3D-CRT, but lower frequency of GI toxicity for IMRT Assumes a treatment dose of >70 Gy. Cost estimate is for a mature program. Sensitivity analyses were conducted to assess the robustness of the model results and to evaluate the cost effectiveness of IMRT in different scenarios. Limitations: Assumes the same dose. Some feel that IMRT’s advantage lies not in its technique but in the ability to give a higher dose; cost of GI toxicity based on a survey of a few radiation oncologists. Risk of bias moderate due to uncertainty of the data abstracted from the literature.
Carter (2014) [44]	NA	Cost modeling (20-yr timeline)	Data from systematic review, a series of component studies, and a workshop	IMRT versus 3D-CRT in the postprostatecto my setting	Markov decision model	Direct health care costs (did not include capital costs)	Australian health care system	Average patient age 65 yr Mean IMRT cost: $32 816 Mean 3D-CRT cost: $33 917 Difference: $−1101 ICER: $41 572/QALY (no year for inflation adjustment) Performed a series of one-way and probabilistic sensitivity analysis. IMRT was found to be more effective and less costly than 3D-CRT. However, the differences are small, and cost utility analyses over a long follow-up period are highly dependent on assumptions about toxicity and costs of treating these. Limitation: Heavy reliance on clinical judgment to generate inputs for the model due to a lack of published evidence. Risk of bias moderate due to uncertainty of the data abstracted from the literature.
Cooperbe rg (2013) [35]	NA	Cost utility analyses	Review of the published literature to estimate outcomes	IMRT versus 3D-CRT	Decision-analytic Markov model to evaluate QALYs, and lifetime costs. Costs for visits and procedures derived from the 2009 National Medicare Fee schedule	Direct costs and sensitivity analyses incorporating cost to patient (time of work)	Payer (Medicare)	Average patient age 65 yr Direct medical IMRT cost: $27 084 (2009) Direct medical 3D-CRT cost: $13 013 (2009) Difference: $14 071 (2009) IMRT was found to be significantly more effective than 3D-CRT among patients with low-risk disease (0.5 QALYs gained). No ICER calculated. Limitation: Multiple assumptions underlie the model. Risk of bias low, as sensitivity analyses changing many assumptions did not change the outcome of the analyses.
Yong (2016) [42]	NA	Cost modeling	Data from literature review and consulting with radiation oncologists, physicists, and radiation therapists	IMRT versus 3D-CRT	Costs estimated through activity-based costing	Direct costs plus cost of equipment (capital, construction, maintenance and operating costs), cost of supplies, personnel, and overhead costs	Payer (Canadian health care system)	Mean IMRT cost: C$12 834 (2009) Mean 3D-CRT cost: C$12 453 (2009) Difference: C$381 (2009) Tested various scenarios by varying the program maturity and the use of volumetric modulated arc therapy alongside IMRT, longer dosimetry time for IMRT, longer quality assurance for IMRT. In the start-up scenario, the incremental cost of IMRT increased by a factor of 11 ($4268 vs $381), which highlights the importance of evaluating timing. Risk of bias moderate due to cost estimates derived from two centers in Ontario.
Zemplényi (2016) [43]	NA	Cost modeling (10-yr time line)	Data from systematic reviews and interviews with radiation oncologists and physicists	High-dose IMRT and hypofractionated IMRT versus 3D-CRT	Costs modeled in decision tree analyses (Markov model)	Direct and indirect costs	Public payer perspective (Hungary)	Average age 70 yr Mean high-dose IMRT cost: €6831 Mean 3D-CRT cost: €7160 Difference high-dose IMRT: €−329 ICER high-dose IMRT: €−1624/QALY (no year for inflation adjustment) Significant differences in the cost structure and cost levels of hospitals in central and eastern European countries are noted compared with the USA and Canada. In Hungary, every type of EBRT is categorized into the same DRG and so has the same reimbursement. As such, they applied a microcosting method to calculate the real cost of 3D-CRT and IMRT. A series of one-way sensitivity analyses were performed for key parameters; a probabilistic sensitivity analysis was performed to assess the uncertainty in the cost-effectiveness analysis by varying model assumptions simultaneously. Limitation: The study was conducted under the assumption that higher doses can only safely be delivered with IMRT; thus, more favorable tumor control with IMRT is not linked to the therapeutic technique, but the escalated doses. Risk of bias is high as treatment costs were determined based on the data from one institution and uncertainty surrounds the assumptions about treatment effectiveness.

GI = gastrointestinal; ICER = incremental cost effectiveness ratio; IMRT = intensity-modulated radiotherapy; QALY = quality-adjusted life year; QOL = quality of life; GI = gastrointestinal; DRG = diagnosis-related group; 3D-CRT = three-dimensional conformal radiotherapy; NA = not applicable; SEER = Surveillance, Epidemiology, and End Results.

Note: For any cost data, the year to which inflation adjustment was made is indicated in parenthesis. IMRT studies are sorted by perspective taken and listed in chronological order based on the years included in the study.