Intensity-modulated radiotherapy: current status and issues of interest

IJROBP 2001;51:880    The IMRT Collaborative Working Group (CWG)


Radiotherapy planning and delivery are in the process of changing dramatically. This change is being driven in large part by continuing advances in computer hardware and software that has led to the development of sophisticated three-dimensional radiation treatment planning (3D-RTP) and computer-controlled radiation therapy (CCRT) delivery systems. Such planning and delivery systems have made practical the implementation of three-dimensional conformal radiation therapy (3D-CRT). The goal of 3D-CRT is to conform the spatial distribution of the prescribed dose to the 3D target volume (cancerous cells plus a margin for spatial uncertainties) and at the same time minimize the dose to the surrounding normal structures. Typically, the delivery of 3D-CRT is accomplished with a set of fixed radiation beams, which are shaped using the projection of the target volume. The radiation beams normally have a uniform intensity across the field, or, where appropriate, have this intensity modified by simple beam fluence-modifying devices, such as wedges or compensating filters.

However, even before this form of 3D-CRT (henceforth referred to as conventional 3D-CRT) has been implemented throughout the radiation oncology community, a new type of conformal planning and delivery technology is evolving. This new type of 3D-CRT, intensity-modulated radiation therapy (IMRT), is based on the use of optimized non-uniform radiation beam intensities incident on the patient. IMRT treatment plans are often generated using inverse planning or automated optimization 3D-RTP systems, which use computer optimization techniques to help determine the distribution of intensities across the target volume.

In any new area of technology, new words and new uses of old words rapidly come into being. Although this is necessary and desirable, a poorly defined term can lead to a misunderstanding in reporting the clinical results and also in research and development. For example, various other descriptors have been used in the past in reference to IMRT, including generalized 3D-CRT, unconstrained 3D-CRT, and computer-controlled conformal RT. The IMRT Collaborative Working Group (CWG) supports the establishment of a consistent and clear nomenclature for use in IMRT. To this end, a glossary of words and phrases currently used in IMRT is given. Where clarification is needed, recommendations for new terminology are given.

As emphasized throughout this report, IMRT techniques are significantly more complex than many other traditional forms of RT, including conventional 3D-CRT. However, as discussed in later sections of this report, IMRT has the potential to achieve a much higher degree of target conformity and/or normal tissue sparing than most other treatment techniques, especially for target volumes and/or organs at risk with complex shapes and/or concave regions.

It is important for the reader to fully appreciate that modern IMRT is more than just the use of non-uniform intensities in radiation fields. Beam modifiers such as wedges and compensators have been used for many years to accommodate missing tissue and in some instances to shape dose distributions. However, as previously stated, modern IMRT is generally designed using inverse planning (or other methods) to optimize the shape of the dose distribution, with the capability of generating concave dose distributions and providing specific sparing of sensitive normal structures within complex treatment geometries. Thus, determining the optimum beam fluence is an integral component of IMRT. In fact, the central planning problem for IMRT is to determine the physically deliverable modulated beam fluence profiles that result in a dose distribution that most closely matches the desired one.

The clinical use of IMRT is in its beginning phase and has been implemented in only a few centers around the world. Much research and developmental work remains to be done to help make the application of this new technology straightforward and easy to perform. To date, only a few thousand patients have been treated using commercial and university-developed  IMRT systems. The potential advantages of IMRT and inverse planning are relatively easy to demonstrate qualitatively in treatment planning exercises (see the section “Clinical Experience”), but careful comparative studies and clinical trials are needed to show that IMRT leads to improved outcomes. It is also possible that IMRT and inverse planning offer practical advantages that may not yet be fully appreciated by the radiation oncology community. That is, when IMRT is fully developed, the potential is significant for this integrated 3D planning and delivery technology to result in lower cost treatment machines and improved efficiencies in planning, delivery, and treatment verification, all of which will may make a valuable contribution to lowering the overall costs of RT while improving the therapeutic results.

This report is intended to create a snapshot in time of IMRT technology and its use. The intended audience is practicing physicians and medical physicists. We also believe that many of the recommendations and suggestions may be of interest to IMRT equipment manufacturers and research funding agencies. We have tried to present a balanced summary that gives some historical perspective, addresses important IMRT issues, and highlights the most relevant publications. In some sections (e.g., “Facility Planning and Radiation Safety”), the reader will find that the depth of discussion and detail presented is much more than in others. This was required to support specific recommendations but made for some unevenness in the writing.

IMRT Historical review

The main technological precursors for the development of IMRT were the development of image-based 3D-RTP systems and the development of computer-controlled delivery systems.

3D treatment planning systems

Computerized RT planning was first reported >40 years ago. Early dedicated RTP systems depended on two-dimensional (2D) contour information and calculated doses based on relatively simple 2D dose models. This type of planning was (and continues to be) widely used throughout the RT community. The first 3D approach to treatment planning dose calculation and display is credited to Sterling, who demonstrated a computer-generated film loop that gave the illusion of a 3D view of the anatomy and the calculated isodose distribution (2D color washes) throughout a treatment volume. van de Geijn also performed early work in 3D dose-calculation models. Much of this work was eventually integrated into commercial RTP systems, but the full potential of image-based 3D treatment planning was not available to these early systems.

Reinstein  took the first real step toward clinically usable 3D-RTP in 1978 with the development of the beam’s-eye view display. The beam’s-eye view display provides the planner with a view from the perspective of the source of the radiation beam, looking down the rays of the divergent beam, and results in a view of the anatomy similar to a simulator radiograph. At the same time, the introduction of CT scanning and its use for RT significantly improved the way patient anatomy could be specified in treatment planning. In 1983, Goitein and Abrams demonstrated how CT data made possible high-quality color beam’s-eye view displays and simulated radiographs computed from CT data (referred to as digitally reconstructed radiographs). Finally, between 1986 and 1989, several robust university-developed 3D-RTP systems began to be implemented in clinical use.

The additional development of 3D-RTP systems throughout the past 10 years, most importantly, including the commercial availability of 3D-RTP systems, has led to widespread adoption of 3D planning in many clinics. One of the keys to the acceptance of 3D-RTP throughout the community was a series of research contracts funded by the National Cancer Institute in the 1980s and early 1990s to evaluate the potential of 3D-RTP and to make recommendations to the National Cancer Institute for future research in this area. Each of these contracts funded a CWG to evaluate various aspects of 3D-RTP. Important developments and refinements in 3D planning technology came from these contracts, particularly plan evaluation software tools, such as dose-volume histograms (DVHs), and biologic effect models, such as tumor control probability (TCP) and normal tissue complication probability (NTCP) models, as well as efforts to stimulate and document the current state of knowledge about these effects. Many of these features are crucial parts of plan optimization, which is critical to IMRT. Similar collaborative groups elsewhere in the world, for example, the Computer Aided Radiotherapy project in the Nordic countries, also contributed significantly to the development of 3D treatment planning.

Target volume and dose specification and reporting

The importance of providing a clear and unambiguous description of the RT when specifying a treatment regimen and reporting clinical results is obvious. Difficulties associated with dose and volume specification for conventional RT have been pointed out by several authors. For example, is the reported dose the minimal dose to the target volume? Or is it the dose at or near the center of the target volume? The International Commission on Radiation Units and Measurements (ICRU) has addressed the issue of consistent volume and dose specification in RT, publishing ICRU Report 29 in 1978, ICRU Report 50 in 1993, and ICRU Report 62 in 1999.

One of the important factors that has contributed to the success of the current 3D-RTP process is the standardization of nomenclature published in ICRU Report 50. This report has given the radiation oncology community a language and method for image-based 3D planning for defining the volumes of known tumor (gross tumor volume [GTV]), suspected microscopic spread (clinical target volume [CTV]), and marginal volumes necessary to account for setup variations and organ and patient motion (PTV).

The ICRU recently updated the recommendations of ICRU Report 50, but did not address the specific needs peculiar to IMRT. There appears to be a need for some modification in the ICRU recommendations.

Target volume specification

The clinical use of IMRT is generally motivated by the desire to conform the high-dose region to the target without inflicting unacceptable normal tissue complications. In general, the high-dose region is shaped to conform to the geometry of the target in three dimensions with rapid fall-off in all directions outside the target volume. Thus, the dose delivered to tissue outside the target volume can be significant if setup error or tumor motions are greater than the allowed treatment planning margins. In addition, because each IMRT segment treats only a portion of the target volume at a time, there may be significant dosimetric consequences if the patient and/or the target moves during treatment. Hence, it is clear that IMRT imposes a more stringent requirement than conventional RT in terms of accounting for patient position-related organ motion, interfraction organ motion, and intrafraction organ motion. All technical and clinical aspects of this part of the treatment planning process must be re-evaluated in light of this requirement.


Treatment planning, whether forward or inverse, can be a futile endeavor if the tumor volume is not correctly identified. As previously indicated, ICRU Report 50 identifies three different volumes that should be delineated. The GTV describes the part of the cancer that can be directly imaged or palpated. CT alone often fails to identify the GTV adequately or, more accurately, does not identify the same GTV as determined using other imaging studies. MRI, various nuclear medicine studies, magnetic resonance spectroscopy, and even ultrasonography are in routine, but sporadic, use within the radiation oncology community. However, no guidelines exist to aid the clinician in knowing the conditions under which specific imaging modalities would be best used. This is an important area of research.

The delineation of the CTV depends heavily on a priori knowledge of the behavior of a given tumor. For a given GTV, tumor histologic features, and patient type, a set of probabilities exists that the tumor will, or will not, extend into a given regional organ or lymph node. However, these specific data are usually not available to the radiation oncologist—only general principles are known. More quantitative, consistent definition of CTVs is an important need.

Planning target volume

A critical point in the planning and delivery of IMRT is the prescription of meaningful PTVs for the patient. The PTVs must ensure proper coverage of the CTVs in the presence of the interfraction and intrafraction variation of treatment setup and organ motion. An inadequate PTV will typically lead to under dosing of the CTV and/or overdosing of the surrounding organs at risk. The conventional approach of creating a PTV by assigning a uniform margin around the CTV is no longer adequate for IMRT.

The complexity of IMRT necessitates the careful examination of whether a computer-optimized plan can be faithfully delivered to the patient. Conversely, one can ask whether a particular patient is a suitable candidate for IMRT treatment. Recent in-depth studies based on daily electronic portal imaging and repeated CT scanning clearly demonstrate that uniform margin reduction, to the level required for dose escalation, cannot accommodate the variation of treatment setup and organ motion. On the other hand, such studies improve our understanding of treatment uncertainty and allow the development of new approaches for more appropriate PTV prescription.

For each patient, there are two components to the geometric uncertainty: setup variation and organ motion. A first-order approach is to treat them independently, although in several instances, they have compounding effects on each other. Both setup and motion must be considered to accommodate the inter- and intrafraction treatment variation. Efforts have also been made to further model the variation into its systematic and random components.

For the daily setup variation, the systematic component is often larger than the random component. It follows that with the conventional approach of prescribing the PTV, according to an institutional standard, much of the PTV margin used is needed to accommodate interpatient variation. A substantial margin reduction can be attained by correcting the systematic component such that only random setup variation needs to be accounted for by the PTV. This inherently individualized approach requires more frequent portal imaging for determining the systematic and random components of the setup variation. Although the approach does require increased efforts from all treatment personnel, several clinical models have been successfully and efficiently implemented. Furthermore, recent experience suggests that, with a properly implemented network infrastructure that accommodates electronic portal imaging, the process imposes minor, if not a smaller, burden on the personnel than the present practice with weekly port film imaging. As for the intrafraction setup stability, very few systematic studies have been done to evaluate its magnitude. It is often assumed to be insignificant in conventional treatment. This assumption needs to be re-examined for IMRT because of the extended treatment time.

The study of PTV-organ motion has been mainly directed to the problems of interfraction variation in the treatment of prostate cancer and the intrafraction variation of breathing motion for disease in the thoracic and upper abdominal regions. Several radiographic and CT studies have shown that the prostate position can vary by >10 mm between treatment fractions and that the variation in rectal position exhibits a time-trend dependence with the course of treatment).

Some uses of IMRT are predicated on the desire to escalate the tumor dose delivered to the patient. To accomplish this goal, it may be helpful if the margin for the PTV can be reduced further with more direct treatment intervention. The Adaptive Radiation Therapy paradigm described by Yan  uses early measurements during patient treatments to more appropriately prescribe the required margins for later treatments based on the localization data for the individual patient.

A second method for improving the geometrical accuracy is a “target of the day” approach, which relies on image guidance in which the target position is identified daily. This can be performed with various imaging procedures such as ultrasound imaging, radiographic imaging of implanted radiopaque markers, or tomographic (CT) imaging. The general principle is to adjust the field to the daily position of the target as detected by each imaging procedure. Radiographic-guided delivery has been implemented for the head-and-neck region without implanted radiopaque markers, where the rigid body model of treatment variation may be valid. The overall PTV margin can be as small as a few millimeters. Ultrasound imaging has been adapted for IMRT prostate treatment localization at several institutes. It should be noted that with radiographic or ultrasound guidance, a new patient reference point in relation to the treatment machine isocenter is calculated daily. Because the adjustment of the field position does not account for possible shape changes, and there are more residual errors in the process, a residual margin needs to be prescribed. Early reports on ultrasound-guided delivery suggest that the margin for PTV-organ motion can be reduced to about 5 mm. The tomographic guidance method is theoretically the most comprehensive of the three approaches, because the correction would account for both nonrigid soft tissue variation and daily setup error.

Both adaptive and image-guided delivery of IMRT may substantially increase the demand on the resources of the clinic, but may be necessary when the dose is escalated beyond normal ranges. The adaptive approach transfers the necessary effort off-line to preserve daily treatment efficiency. The image-guided approach is conceptually more powerful, but at the cost of additional daily effort. A third approach to reduce organ motion is through improved immobilization. For example, this has been accomplished for prostate cancer treatment by inflating a balloon in the rectum during the delivery of the IMRT.

A modest reduction in the size of the margin allowed for organ motion has been attained by using CT scans acquired near the end of the normal breathing cycle for planning. More recently, active intervention in ventilatory motion has been investigated, including the use of ventilatory-based gating, breath-holding, and active breathing control. The different methods include various tradeoffs, ranging from machine control, which is not dependent on the patient, to systems that are completely dependent on the patient; all require continued research before they are routinely available for clinical use. The choice of approaches to reduce the PTV for breathing motion is dependent (at least partly) on the IMRT delivery technique. The discrete nature of the SMLC method may be amenable to all three methods listed above. However, for delivery with the DMLC or tomotherapy methods, the breath-hold methods might be more suitable because gating requires additional control of the coupled mechanical motion.

Dose specification

Specification of the doses used for both prescription and reporting is difficult for the fast-changing and non-uniform dose distributions often found when IMRT is used, and the specification is a problem that needs much new work. The ICRU recommendations regarding dose reporting for traditional 3D-CRT include the dose at or near the center of the PTV, as well as the maximal and minimal dose to the PTV. ICRU also recommends that any additional information such as the mean dose and the DVHs be reported when available. No firm recommendations regarding dose prescription have been provided.

The Nordic Association of Clinical Physics has proposed that for relatively small dose non-uniformity, the mean dose and its standard deviation to the CTV (with margin for internal motion) be used for both treatment prescription and reporting. When the relative standard deviation of the dose distribution is larger than the tolerance range (for steeply responding tumors and normal tissues, a relative standard deviation <2.5%, and for more shallow responding tumors and normal tissues, a relative standard deviation of no more than 5%), the Nordic Association of Clinical Physics recommends that the minimal dose to the CTV and mean dose delivered to the hot and cold volumes within the CTV be reported. Note, the Nordic Association of Clinical Physics defines the hot volume as a volume that receives a dose larger than the prescribed dose by an amount larger than the tolerance limit. A cold volume is defined as a volume inside the CTV that receives a dose lower than the prescribed dose to the CTV by an amount larger than the tolerance limit.

It should be pointed out that significant problems are associated with dose reporting of the maximal and/or minimal doses. For example, the minimal dose is highly uncertain because of uncertainties in the placement of the region of interest near high-gradient regions. The maximal dose is also unreliable, because it corresponds to the high-dose tail of the DVH. In both cases, the values depend on the voxel size. Also, with the advent in the future of Monte Carlo-based treatment planning, the maximal and minimal doses in a region of interest will be, by definition, several standard deviations away from the true maximal or minimal doses.

It is likely that to make possible quantitative use of clinical results involving IMRT, the entire DVH for each of the pertinent volumes (PTV, CTV, and the organs at risk) will need to be reported. Therefore, the CWG believes that the dose-volume data available directly from the DVHs generated by IMRT planning systems are more suitable for correlating with clinical outcomes. The CWG suggests that, as a minimum, the dose that covers 95% (D95) and 100% (D100) of both the CTV and the PTV and the percentage of the CTV and PTV receiving the prescribed dose (V100) be obtained from a DVH and reported. The mean, minimal, and maximal doses (averaged over the nearest neighbor voxels) to each CTV and PTV should also be reported. Similarly, for the organs at risk, the mean, minimal, and maximal doses and other relevant dose-volume data should be reported.

Recommendations: target volume and dose specification

The following list summarizes the CWG’s recommendations regarding target volume and dose specification for IMRT for the purposes of correlating them with the clinical outcome:

1.Clinicians should specify the target volume(s) following the recommendations of ICRU Reports 50 and 62 198, 199.
2.The PTV must (at least attempt to) ensure proper coverage of the CTV in the presence of the inter- and intrafraction variation of treatment setup and organ motion. The conventional approach of assigning a uniform margin around the CTV is generally no longer adequate when IMRT plans are considered.
3.Application of IMRT to sites that are susceptible to breathing motion should be limited until proper accommodation of motion uncertainties is included.
4.Important research issues in the area of target volume specification include the following:
Development of guidelines defining which specific imaging modalities should be used for GTV delineation for specific sites.
Development of more automatic and robust methods of image registration suitable for routine use in RT treatment planning.
Development of quantitative methods and rules for CTV delineatio
Development of methods and/or technology to better account for and reduce spatial uncertainties.
5.As a minimum, the following information should be reported for the purpose of correlating the dose with the clinical outcome:
Prescribed (intended) dose, as well as the point or volume to which it is prescribed; a fractionation prescription should also be included.
Dose that covers 95% (D95) of the PTV and CTV.
Dose that covers 100% (D100) of the PTV and CTV (i.e., the minimal dose).
Mean and maximal doses within the PTV and CTV.
Percentage of the PTV and CTV that received the prescribed dose (V100).
For each organ at risk, the maximal, minimal, and mean doses, the volume of the organ receiving that dose, and other relevant dose-volume data.

Clinical summary

Thus, we see that no definitive studies have conclusively demonstrated the impact of IMRT on improved tumor control and decreased long-term morbidity, nor have any studies demonstrated the superiority of one particular IMRT technique—at least on a clinical basis. Ultimately, the value of IMRT needs to be tested to show that the use of IMRT will further the 3D hypothesis, as advocated by Lichter; that is, that 3D-CRT will allow higher doses of radiation to be delivered with equal or less morbidity than standard techniques. The IMRT CWG does not advocate the direct testing of IMRT vs. no IMRT—that would be too reductionist—nor, even, necessarily 3D vs. 2D planning. Rather, we advocate the testing of the high doses made possible by IMRT vs. conventional doses. The areas of interest include cranial, head and neck, lung, breast, pancreas, prostate, and gynecologic applications.


IMRT is an advanced form of external beam irradiation often used to perform 3D-CRT. It represents one of the most important technical advances in RT since the advent of the medical linear accelerator. Currently, most IMRT approaches increase the time and effort required by physicians and physicists, because optimization systems are not yet robust enough to provide automated solutions for all disease sites, and routine QA testing is still quite time intensive. Considerable research is needed to model clinical outcomes to allow truly automated solutions.

Current IMRT delivery systems are essentially first-generation systems, and no single method stands out as the ultimate solution. In addition, IMRT techniques appear to place greater stress on the treatment machines. Currently, no articles about the effects of IMRT use on machine reliability, downtime, and failure rate have been published. This could become a potential issue in the future.

The instrumentation and methods used for IMRT QA procedures and testing are not yet well established. In addition, many fundamental questions regarding IMRT are still unanswered, including the radiobiologic consequences of altered time-dose-fractionation and that the dose heterogeneity for both the target and normal tissues may be much greater with IMRT compared with traditional RT techniques.

All that said, this new process of planning and treatment delivery shows significant potential for improving the therapeutic ratio. Also, although inefficient today, it is expected that IMRT, when fully developed, will improve the efficiency with which external beam RT can be planned and delivered, and thus potentially lower costs.

The recommendations contained in this report are intended to be both technical and advisory, but the ultimate responsibility for the clinical decisions pertaining to the implementation and use of IMRT rests with the radiation oncologist and radiation oncology physicist. It should be well understood that this is an evolving field, and the CWG expects modifications of these recommendations as new technology and data become available.