Dosimetry is Alive and Well

Dear Editors:

We read with much interest the article—and associated Letter to the Editor—from Siegel, Stabin, and Sharkey in the current issue of Cancer, Biotherapy and radiopharmaceuticals. Scientific issues raised by this article are important, and deserve an indepth analysis. In this letter, we present early comments that we hope can contribute to the start of future scientific discussions. However, the scientific issues raised by this article should—in our opinion—be put in perspective within the broader frame of patient-specific dosimetry in molecular radiotherapy.

The evidence of an absorbed dose–effect relationship is the underlying principle for any treatment with ionizing radiation, even though no guidance is provided as to evaluate the efficacy or toxicity of a given treatment.

For targeted radionuclide therapy (TRT) Sgouros emphasized the fact that: “[t]he objective of dosimetry in targeted radionuclide therapy is to provide information that will help improve patient care. With this objective, estimated absorbed dose is useful to the extent that it relates to response.” ¹ This connection to the clinical usefulness of dosimetry was also advocated by Stabin and Brill in the Journal of Nuclear Medicine. These researchers stated: “[P]atients given radiopharmaceuticals for therapy deserve the same individualized attention and optimization of their radiation therapy as do patients treated with external sources of radiation, which has been undergoing constant improvement for decades.” ²

However, patient-specific dosimetry is seldom performed by nonresearch centers, because of the difficulties involved and the lack of standardized protocols. Currently, the majority of treatments are administered without the performance of prospective or retrospective absorbed dose calculations. This can result in under- or overtreatment of individual patients and precludes a direct comparison of results and strategies among centers. ³ The fact is that the implementation of a dosimetric protocol in clinical practice is a complex task, involving multiple steps including quantitative imaging (to derive the accurate localization of activity in the patient), pharmacokinetics modeling (to obtain the total number of disintegrations—the cumulated activity—within a given region of interest from a necessarily restricted number of timepoint measurements), and absorbed-dose calculation.

For decades, the MIRD committee has provided the scientific and medical community with dosimetric models and data regarding that last aspect. ⁴ The S factors that give the mean absorbed dose per unit cumulated activity (Gy/Bq.s) for reference organs and tissues from anthropomorphic phantoms (originally meant for radiation protection estimates within a context of diagnostic nuclear medicine procedures) have been widely used in TRT dosimetry, mostly because to the lack of computing tools allowing for patient-specific absorbed-dose calculation. The situation is changing fast with the advent of patient-specific absorbed dose-calculation packages, but the current situation in the clinical domain is that “one size fits all” S factors are still widely used, even though a simple mass-adjustment of standards S factors may sometimes provide acceptable results. ⁵

Conversely, in the clinical environment, most of the resources were oriented toward trying to increase the accuracy of activity quantification, by implementing acquisition and processing protocols that would provide for accurate quantitative imaging at the patient level (thus providing, at least, an accurate patient-specific activity determination). However, despite all such efforts and associated improvements in image activity quantification, no protocol has emerged to date as an absolute reference for use in clinical practice.

This, and the apparent failure of dosimetry to predict efficacy or toxicity ^6,7 led the European Association of Nuclear Medicine (EANM) therapy committee to conclude: “Although dosimetry has been of enormous value in the pre-clinical phase of radiopharmaceutical development, its clinical use to optimize administered activity on an individual patient basis has been less evident.” ⁷

In fact, now, the situation has evolved, with the publication of articles demonstrating that absorbed dose–effect relationships could be observed in clinical practice. These article cover bone-marrow toxicity for non–Hodgkin lymphoma radioimmunotherapy, ⁸ mIBG treatment of neuroblastoma, ⁹ and treatment efficacy in radioiodine ablation in differentiated thyroid cancer. ¹⁰ It must be emphasized that these approaches use a methodology that is not extremely refined and do not include radiobiology. This somehow highlights the fact that even “basic” dosimetric methodology can still provide clinically useful data. The current trend in absorbed dose–effect relationship assessment, however, is to implement radiobiological modeling into the dosimetric framework: Radiobiological modeling is introduced to convert the spatial distribution of absorbed dose into biologically effective dose (BED) and equivalent uniform absorbed dose parameters. ¹¹ Clinical evidence of the relevance of these approaches has been published, essentially in the domain of critical organ toxicity (most often bone marrow, but also kidneys for peptide therapy). Of particular interest, is the study reported by Pauwels et al. on peptide receptor radionuclide therapy with ⁹⁰Y-labeled somatostatin analogues. ¹² In a related publication, researchers came to the conclusion: “The use of a refined absorbed dose methodology led to the finding of a clear kidney dose–response relationship in patients treated with ⁹⁰Y-DOTATOC.” ¹³ Indeed, observation of the results presented by these researchers highlights both the complexity and evolving conceptions regarding the implementation of dosimetry in clinical practice:

The clinical settings used a very complex and refined activity determination protocol based on quantitative positron emission tomography (PET) imaging of ⁸⁶Y-labeled DOTATOC. ¹³ However, the first dosimetric results obtained using MIRDOSE 3.1, ¹⁴ “one size fits all” S factors did not yield any significant dose–toxicity correlation.

Kidney dose computation using MIRD Pamphlet 19 ¹⁵ and CT-measured kidney cortex volumes instead of MIRDOSE 3.1 and standard volumes resulted in a wider range of absorbed-dose estimates but still provided a poor absorbed dose–toxicity relationship (r = 0.54; p = 0.02).

Introducing radiobiological parameters (by taking into account the dose–rate effect) and representing the toxicity as a function of the BED provided for a much better correlation (r = 0.93; p < 0.0001), despite the fact that the parameters applied (α/ß and DNA repair half-time) were taken from the literature.

From a methodological point of view, some conclusions can be drawn from this study: Clinical dosimetric protocols should not only focus on as accurate as possible activity determination but should also consider specific absorbed-dose calculation approaches, for which accurate assessment of volumes of interests is a prerequisite. ¹⁶ Yet, even that may not be sufficient, and radiobiological input may also be necessary in order to put in evidence a radiation dose–effect relationship.

The presentation of MIRD pamphlet 20 ¹⁷ is the logic sequel of the previously cited article. The conclusion was: “Absorbed dose computed from a uniform distribution of activity in the kidney using standard reference phantoms without regard to the dose-rate and spatial nonuniformities will not provide meaningful dose–response correlations.”

In their contribution, Siegel et al. have evaluated the predictive value of absorbed dose, BED, and time–dose-fractionation TDF (TDF) factors for use in patients who have undergone peptide radionuclide receptor therapy (PRRT). These researchers concluded that TDF is as good as/better than BED for predicting kidney toxicity, suggesting the use of models not only derived from external beam radiotherapy (XBRT).

The need for improvement in the radiobiological models currently in use, including BED, is unquestionable; however, before calling the TDF back into the group of viable models, much more evidence than just the studies critiqued in the article should be provided. In fact, the two models are not mathematically equivalent and do not yield exactly the same results, although these models could produce results quite similar over the clinically relevant range. If absorbed doses and dose rates vary enough, differences between clinical results and predictions from either model should become noticeable.

Initially, the TDF factor was devised in the context of XBRT/brachytherapy. In a recent publication entitled “21 years of Biologically Effective Dose,” Fowler recalled:

In 1989, the British Journal of Radiology (BJR) published an article that introduced the term BED, biologically effective dose, as a linear quadratic (LQ)–based formula with an overall time factor included, to replace Dr. Frank Ellis's (1969) nominal standard dose (NSD) and the Orton and Ellis (1973) time–dose factor (TDF) tables… . ¹⁸

As was acknowledged by Siegel et al., TDF was found faulty for several reasons. In particular, Fowler stated: “Nominal standard dose, TDF and CRE [cumulative radiation effect] overestimated the effect of overall time on late reactions and underestimated it on early reactions…. They should no longer be used.” ¹⁹ Another objection to the use of TDF is that the issue of sparing kidney tissue by fractionation is not addressed with the TDF representation, unless one introduces some repair-correction factor. So, to prove the point further, more details about the advantages and disadvantages of TDF in PRRT use (and beyond) should be reported—via the availability of clinical studies involving a large enough number of patients.

In a talk given at the 1st International Symposium on Radionuclide Therapy and Radiopharmaceutical Dosimetry ²⁰ in 2004 (Helsinki) M. Stabin sorted clinical dosimetry into 4 categories:

No dosimetry (most common practice)

Bad dosimetry (state of the art)

Good dosimetry (sometimes observed)

Excellent dosimetry (still an area of research).

Since then, impressive methodological improvements have been reported, both in quantitative imaging and absorbed dose calculations. In addition, further evidence of the absorbed dose–effect relationship has been published in clinical studies. It is important to note that clinical dosimetry studies may or may not be based on refined methodological approaches and still be of a high clinical value (i.e., impact patient's management and help in adjusting injected activity). If dosimetry is well-reported, ²¹ then other groups can re-analyze the data, as Siegel et al. did.

The scientific discussion initiated in this Cancer Biotherapy and Radiopharmaceuticals issue goes beyond the widespread belief that two physicists trying to address a given problem are likely to end up with three different opinions. Dosimetry is now reaching a mature age in terms of methodology. With an increasing number of groups working in the field, scientific expertise is more often available and allows the generation of more reliable dosimetric data. It is important to notice that the current scientific discussion was made possible by the availability of clinical data that was shared and analyzed by different groups. Converging toward a new consensus approach to radiobiological modeling in our field will improve the robustness of clinical findings derived from future trials, thereby increasing the value of molecular radiotherapy treatments.