Answers are grounded in this paper and cite their sources.
It is standard practice to evaluate dose distributions in different tissues, organs, and tumors through visualization of dose-volume histograms (DVHs) and/or analysis of quantities derived from DVHs. The DVH does not contain spatial information and has several other shortcomings, however, and some have suggested that evaluation of treatment plans using DVHs is no longer appropriate. This is the claim debated in this month's Point/Counterpoint. The fundamental tenet of radiation therapy is to deliver a tumoricidal dose to the target while limiting the dose to the normal tissues and organs at risk (OARs), and the goal of treatment planning is to optimize these objectives. The DVH was introduced in the 1980s as a simple way to evaluate plan quality by condensing a 3D dose distribution into a 2D graphical representation of the dose distribution throughout the target volume and each of the OARs.1 In today's era of modern radiation therapy with competing techniques such as IMRT, VMAT, and proton therapy, it is my opinion that DVHs are no longer the best tool for plan quality evaluation. Let us start by identifying some of the shortcomings of the DVH. By the very nature of the DVH, there is a loss of spatial information of the dose in the volume under consideration.1,2 When comparing competing plans, such information about the location of hot and cold spots can be critical to determination of the effectiveness of that treatment.3 The DVH can only be calculated for the defined volume of interest and its shape may be misleading. The accuracy with which the DVH is estimated depends highly on the accuracy with which the target or OARs is delineated.1 Accurate target and OAR delineation is a serious problem in radiation therapy.4 DVH accuracy is also dependent on the accuracy of the dose distributions computed by the treatment-planning system. This is affected by several sources of uncertainty, such as algorithm limitations, measurement uncertainty in the data used to model the beam, and residual differences between measured and computed dose.5 Furthermore, when delivery uncertainties are present, the DVH may not always accurately describe the dose distribution actually delivered to the patient.6,7 It has long been recognized that for some normal tissues such as rectum and bladder, the dose distribution of importance is that to the surface rather than the whole volume.8–10 Radiation doses delivered to the contents of the rectum or bladder are of no consequence to the normal tissue complication probability (NTCP). Another limitation of DVHs is that the interpretation of the plot is rather subjective. There is a need to better understand the implication of small differences between DVHs. This is more urgent in today's practice of RT with competing techniques to treat the same target. It is clear that dose distributions in targets and OARs are just surrogates for tumor control probability (TCP) and NTCP. DVHs do not provide any information about the biological response of the tumor and OARs. A truly objective score of a treatment plan should be linked to the potential clinical outcome. Thus, one needs tools that can objectively quantify the probability of an end point of interest such as TCP or NTCP as a function of delivered dose distribution.11 On the basis of the above stated shortcomings of DVHs, it is my contention that the DVH in today's highly conformal radiation therapy environment is no longer an adequate tool for plan quality evaluation. Since its introduction more than 30 years ago, the DVH has become a staple of plan evaluation in radiation oncology.12 So why put it out to pasture now? There is no question that the traditional planning DVH has its limitations: representation of a single time point; no biological or functional information; cannot account for tissue motion or deformation; and contains no spatial dose distribution information. But these limitations will apply to any metric that is based on a plan at a single point in time. I argue that we can address many of the shortcomings of the DVH as opposed to eliminating it. Recent and current topics of clinical research have served to address some of these issues. Image guided radiation therapy (IGRT) has been shown to reduce setup error prior to initiation of treatment.13 Gated delivery has the potential to minimize anatomic deviations during treatment compared to the initial plan geometry.14 Tumor tracking can potentially account for intrafraction motion of the target.15 Deformable image registration has the potential to account for changes in the shape of the target and surrounding normal tissues and relate the delivered dose distribution back to the original structure DVHs.16 Functional information can be used to modify the DVH to evaluate only tissue that may be adversely impacted by the delivered radiation.17 In theory, combining all of these developments can make the treatment planning DVHs more meaningful than ever. As physicists, we seek to develop objective and quantitative metrics to make plan optimization and evaluation more efficient and meaningful. For example, we have moved from visually evaluating dose distributions, to DVH curves, and now to biological models (TCP, NTCP, EUD) to evaluate and rank plan quality. In each step of the process some information is lost, but additional information is gained. However, we should be mindful that every metric has associated uncertainties that must be acknowledged in their use. For example, studies have shown that the specific biological and treatment delivery parameters used can significantly influence plan optimization, evaluation, and ranking.18,19 Even as we have moved through the various methods of plan evaluation, we have not completely eliminated older methods. Our physician colleagues still evaluate dose distributions relative to the patient anatomy in order to evaluate plan quality. We cannot ignore the fact that there is still a subjective component to plan evaluation using metrics that are well-established and familiar. In conclusion, while the DVH is not the appropriate choice for plan evaluation, it is still an appropriate choice. While it should not be the sole method for evaluating treatment plans, I predict that it will continue to be used with past, present, and future evaluation methods to help us generate the best possible treatment plans for our patients. Dr. Parker has articulated strongly his support for DVH but I disagree with his opening argument that we should accept the status quo. The field of medical physics has advanced to where it is today because those who came before us questioned, revised, and improved upon the status quo. Hence, we should strive towards a better way of evaluating plans so that competing techniques can be properly assessed. Dr. Parker and I both agree that the DVH in its present form has significant limitations. He goes further to present some of the recent developments that could potentially improve the accuracy of DVHs. However, if any new tool is a derivative of DVH, then the inaccuracies inherent in DVH will be propagated along with it. Hence, the plethora of indices or metrics20 that have been used as quantitative indices of DVHs are bound by the same uncertainties. This makes it imperative to find a tool that is less sensitive to the uncertainties in dose computations and volume delineation. Some of the shortcomings of the DVH can never be totally eliminated. It is not surprising that over the years, different researchers have proposed alternative approaches. For example, to address the loss in spatial information in the DVH, the zDVH,1 the 2D dose–volume scoring-function histogram (DVSH),21 and slice base plan evaluation,22 have been proposed. The biologically effective dose DVH has also been proposed23 to account for the effect of dose on the given tissue taking into consideration the fraction size, overall treatment time, etc. However, as stated in my opening statement, with increased knowledge of tumor and normal tissue radiation responses and advances in molecular imaging, biologically based evaluation must be the best way forward.24 In conclusion, if we agree that DVH is flawed, then we must also agree that a better tool is required because these flaws might be impacting our decision on what we consider the optimal plan for our patients. We need an appropriate tool to guide us to make the right decisions and thus achieve our goal of improved tumor control. If my colleague is going to argue that the DVH is no longer appropriate for plan evaluation, then he should identify how his suggested replacement metrics address their shortcomings. I agree that there is a loss of spatial information in the DVH, and I addressed that issue in my opening statement. I also agree that some organs should be evaluated by dose-surface or dose-wall histograms. But that is not an inherent limitation of the DVH itself, only an indicator that we need to use a different evaluation tool that more accurately reflects the structure of the tissue of interest.8 As for the other limitations identified by my colleague (structure delineation, dose distribution accuracy, algorithm limitations, delivery uncertainties, measurement uncertainty in commissioning data, etc.), their impacts are not limited solely to the DVH. Errors or uncertainties in any of these quantities will manifest in any metric that is used to evaluate competing treatment plans. He suggests that we should use end-point-of-interest metrics such as TCP and NTCP. Studies have shown that systematic errors in calculated dose (e.g., inaccurate dose algorithms, and inaccurate commissioning data) can significantly impact the calculated values of these parameters.25,26 Additionally, the choice of biological model, as well as uncertainties in the model itself (e.g., tissue sensitivity), will also impact the calculated values.27 The dose evaluation metric is only as good as the data used for its calculation. As I stated in my opening argument, all metrics have some associated uncertainty that must be acknowledged in their use. During a patient's course of treatment, our goal is to make the delivered dose (magnitude and location) match that of the treatment plan. The more we can reduce the uncertainties in the treatment planning and delivery process, the more accurate and relevant the DVH (or any evaluation metric) becomes. Given that DVHs remain a primary method of plan quality evaluation, we should seek to correlate our DVH experiences with new metrics as they develop and mature so as to ensure clinical experience continuity throughout the transition. In any case, the sun has still not set on the DVH in radiation oncology.
No takes yet. Share an insight, caveat, or question.
Njeh et al. (2015) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: