Keywords

6.1 The Gretchenfrage

How is it now with thy religion, say?

I know thou art a dear good man,

But fear thy thoughts do not run much that way.

Faust; a Tragedy. Johann Wolfgang von Goethe

The lines above are uttered by Gretchen (diminutive of the given name Margarete) in Faust, Part one, the most known work of Johann Wolfgang von Goethe, who is considered by many the greatest German poet of all time. Gretchen’s question (the “Gretchenfrage” in German) became a dictum for any issue which is both crucial as well as delicate. Gretchen, an innocent and faithful character asks Dr. Faust about his piety, with the respondent being secretly in league with the devil. Answering puts him in a dilemma. Honesty most probably would make his love interest leave and deceit always is the cornerstone of downfall and tragedy, so Faust evades the question. Ironically enough, he must have some kind of religious faith, since it would not make any sense for a categorical atheist to bond with a godlike, though sinister entity as Mephistopheles.

What does this have to do with dosimetry in Radionuclide Therapy, except the obvious geographical analogy, i.e., the cultural center Weimar as domain of Goethe and Schiller and the nearby Zentralklinik in Bad Berka as the scientific spearhead in the diagnosis and treatment with unsealed, radiolabeled compounds? Just like the Gretchenfrage, the question about the necessity, usefulness, and accuracy of dosimetry calculations in Radionuclide Therapy is both crucial as well as delicate. Crucial, because Radionuclide Therapy has proved its effectiveness in many studies and with regard to different nuclides [1,2,3,4,5]. Moreover, there is no doubt that the reduction of tumor mass stems from the effects of ionizing radiation on malignant cells rather than any other property of the radiolabeled compound.

And yet, in the field of Radionuclide Therapy, this radiobiological knowledge, developed and ensured by decades of radiation research, somehow cannot be transformed smoothly to an elegant correlation function with the calculated, absorbed radiation dose on one side, and the reduction of metastatic tissue on the other. This is the delicate aspect when it comes to dosimetry calculations in Radionuclide Therapy and not uncommonly physicians and medical physicists are tempted, just as Dr. Faust, to evade the question about its necessity, usefulness and accuracy, as well as the associated effort. An ongoing debate among scientists on the value of dosimetry beyond the preclinical phase of radiopharmaceutical development, such as the clinical implementation of personalized, image-based dosimetry, has been held in conferences and journals for years [6,7,8,9]. The purpose of this article is to give a structured and unbiased overview on the issue of dosimetry in Radionuclide Therapy, in the hope to spare the reader from Faust’s lament:

And here, at last, I stand, poor fool!

As wise as when I entered school;

6.2 The Forms of Dosimetry in a Nutshell

The basis for all dosimetry in Nuclear Medicine is of course the MIRD (Medical Internal Radiation Dose) methodology. This article refrains from unrolling the mathematical and physical details, since there is plenty of excellent literature doing this, with [10] and [11] being just two examples.

Quantitative molecular imaging such as planar scintigraphy, SPECT/CT, and PET/CT is used to get hold of the tracer’s bio-distribution within the patient whereas the radiation transport and the resulting dose deposition pattern solely depends on the radionuclide’s emission spectrum and the geometry, i.e., the anatomy. The latter can be represented by phantoms, i.e., computational body models that have developed from being composed of simple geometric forms [12, 13] to the voxel-based ICRP (International Commission on Radiological Protection) reference phantoms [14]. Among the software packages that include these necessary data matrices were the original MIRDOSE [15] which then became OLINDA/EXM [16], as well as IDAC [17]. All the above offer so-called organ dosimetry, i.e., they calculate the mean absorbed radiation dose (AD) to an organ with both, the activity distribution in the source organ as well as the AD in the target organ being homogenously distributed over the organ’s volume, with the option for adjustment to the weight of the patients’ organs. In this simplified form of dosimetry, the AD to lesions can be approximated by the geometrical assumption of the lesion being a sphere (aka spherical model) [17].

Phantom-based organ dosimetry, as described in the last paragraph, is the most widely applied form of dosimetry. However, it is certainly not the most sophisticated form of personalized medicine. For this one needs to perform three-dimensional (3D), voxel-based dosimetry [18], where neither the source nor the target is represented by an a-priory phantom geometry, but by the patients’ imaging data, i.e., PET/CT or SPECT/CT. There the heterogenous activity distribution is given by sequential, nuclear scans and the likewise heterogenous dose distribution is calculated on the voxel base (see Fig. 6.1) by means of Monte Carlo simulations [20] or some forms of convolutions with dose kernels [21,22,23,24]. Naturally this form of dosimetry is both, much more complex as well as prone to errors and only a few commercial software packages offer a workflow for it.

Fig. 6.1
4 absorbed dose images of axial and coronal sections PET C T for tumors 3 and 11. The region of the tumor indicates fluorescence.

Transverse and coronal cross sections depicting voxel-based absorbed dose images of two tumors (thyroid cancer). Adapted from [19]

Full size image

The expression theragnostics (also spelled theranostics) is a fusion of the word therapy and diagnosis. In nuclear medicine, this term describes the use of one or multiple tracers to either predict the absorbed dose in the course of treatment planning and with the help of a diagnostic scan, or to calculate dosimetry via accompanying imaging during therapy. Thereby, often pairs of radionuclides are used that either are isotopes of the same element or can be labeled to the same carrier molecule, thus assuming the biokinetics to be identical. Table 6.1 gives a short overview of the most common radionuclides used in theranostics. Both, phantom based as well as individual voxel-based dosimetry, can be used for thernostics and the following sections shall give you a review of the respective studies.

Table 6.1 Combination of radionuclides used for theranostics in radionuclide therapy. Bold print indicates imaging data used for dosimetry
Full size table

6.3 Dose Quantities and Dose-Response in EBRT

In very general terms natural sciences is all about finding the right quantity to describe or model the phenomena observed. The basic quantity in radiation dosimetry is the mean absorbed dose (AD), which equals the energy deposited by ionizing radiation per unit mass of an anatomic structure (e.g., organ, tumor). This quantity is averaged for the mass and the irradiation time. In external beam radiotherapy (EBRT) multiple studies have demonstrated dose-response relationships between the lesions’ AD and its response to the therapy, i.e., its shrinkage, often in the form of a sigmoidal dose-response curve [40, 41], as depicted in Fig. 6.2. When radiation oncologists perform treatment planning in EBRT they not only use the AD, but also other dosimetric concepts, such as dose volume histograms (DVH), the equivalent dose in 2 Gy fractions (EQD2) [43], or the biologic effective dose (BED). The latter takes into account the differences in dose rate, the repair half-time for sublethal tissue damage, the average doubling time for tumor clonogenic cells, as well as the assumed intrinsic radiosensitivity which is based on the linear-quadratic model [10, 44,45,46].

Fig. 6.2
A line graph on local control versus BED. 2 curves for P T V margin and isocenter begins at the origin and combines at 175 grays.

The object of desire regarding the dosimetry of radionuclide therapy: a sigmoidal-shaped curve describing local tumor control dependent on BED (a/bZ10) of different EBRT irradiation regimes applied in Wuerzburg. The doses were calculated for the PTV-margin and the isocenter. The number of local failures compared to the total number of targets treated by the different fractionation regimes is shown in brackets. (From Wulf et al. [42])

Full size image

Just as for the AD, dose-response curves can be found in EBRT with respect to the BED [42]. When pondering about the dose-response relationship of tumor tissue, the unofficial supreme discipline of radiation oncology, one may not forget the dose-response relationships of healthy tissue, i.e., radiotoxicities, which is equally important for the patients’ wellbeing. Here too, EBRT [47] as well as brachytherapy [43] have produced a remarkable evidence base.

Finally, the equivalent uniform (biological effective) dose (EUD) models the impact of the spatial dose distribution on the response [48]. The BED of each voxel is used to generate an EUD value for a specified volume (e.g., organ, tumor). Mathematically, in the case of dose nonuniformity the value of the EUD is always lower than the AD, which is why some studies suggested the EUD to be a better predictor of tumor response [19, 48,49,50]. Other quantities derived from voxel-based dosimetry include threshold approaches such as, e.g., D70 (the minimum dose to 70% of the voxels constituting the tumor volume) which in one study reliably predicted response or local failure in the treatment of hepatocellular carcinoma (HCC) with 90Y resin microspheres [51]. Yet another study on this particular therapeutic field showed that the coverage of a lesion in terms of the DVH is much more predictive for progression-free survival (PFS) and overall survival (OS) than the mean absorbed dose [52].

6.4 Dose-Response I: Radiotoxicities in Radionuclide Therapy

The concept of BED was implemented in a multiregional kidney dosimetry model in MIRD Pamphlet No. 20 [46].

Bodei et al. [53] used this model and retrospective patient data to calculate a threshold for kidney toxicity at a BED of 40 Gy in patients without risk factors. Given the normally accepted tolerable dose for healthy kidneys of 23 Gy as known from EBRT [54], this value can now be exceeded since, e.g., the threshold BED of 40 Gy for 177Lu-DOTATATE corresponds to an AD of 28 Gy. If fractionation is taken into account, i.e., multiple therapy cycles of radionuclide therapy, this value can even be increased to 35 Gy, thereby also increasing the potential dose to malignant tissue and demonstrating the practical value of dosimetry in radionuclide therapy, where the kidney is always the organ with the highest radiation burden.

Nevertheless, renal impairment is by far means not the only dose-effect of healthy tissue. For a comprehensive review it is referred to the formidable meta-analysis of Strigari and co-workers [55], who found 79 studies investigating dosimetry, of which 48 studies found an absorbed dose-effect correlation. Apart from renal toxicity, radiotoxicities due to radionuclide therapy mainly affect blood, marrow, and liver as listed in Table 6.2. A closer look reveals that liver toxicity only occurs in selective internal radionuclide therapy (SIRT), a treatment modality explicitly used to irradiate liver malignancies, thus explaining its high radiation burden. In researching the published literature on 177Lu-PSMA therapy of metastatic castration-resistant prostate cancer, which became the rising star in radionuclide therapy within the last years, toxicities regarding kidney, blood and salivary gland were found to be minimal [3].

Table 6.2 Radiotoxicities reported in Radionuclide Therapy (except kidney which has a BED threshold dose of 40, see text above). NET stands for neuroendocrine tumors
Full size table

Altogether it can be said that dosimetry did a fine job in quantifying possible radiotoxicities, calculating threshold doses, and establishing safety protocols, in some cases with the help of more sophisticated, radiobiological dose models as for the kidneys [46, 53], thereby making it possible to keep damage to healthy tissue in check and developing radionuclide therapy toward a cancer treatment modality with relatively mild side effects.

6.5 Dose-Response II: Tumor Response in Radionuclide Therapy

As alluded above, data from studies becomes sparse when the focus is on the dose-response of lesions, i.e., if one wants to correlate the tumor control with the radiation dose in the form of a dose-response curve as in Fig. 6.2. Table 6.2 gives an overview and concentrates on studies that actually reported dose-response curves, rather than mere threshold doses.

Pioneering work on this field includes the study of Koral and co-workers, who produced a sigmoidal curve relating AD and tumor volume for untreated patients suffering from low-grade follicular lymphoma and receiving 131I-tositumomab [64]. Another work often cited is from Pauwels et al., who presented the first correlation between AD and tumor reduction for gastroenteropancreatic neuroendocrine tumors (NET), treated with 90Y-DOTATOC [32]. The most relevant study for peptide receptor radionuclide therapy (PRRT) however is a recent one by Ilan et al. [65] where the correlation between AD and tumor volume reduction of patients treated with 177Lu-DOTATATE yielded Pearson coefficients R2 never seen before, with 0.64 for tumors of diameter >2.2 cm and 0.91 (!) for tumors of diameter >4 cm (Fig. 6.3). Furthermore, Dewajara et al. [49] demonstrated a correlation between tumor reduction and EUD for refractory B-cell lymphoma treated with 131I-tositumomab, albeit with rather low correlation coefficients (see Table 6.3).

Fig. 6.3
A line graph on response rate versus absorbed dose. The curves begin at the origin and increase progressively with each absorbed dose.

The scientific community would like to see more graphs like this: Tumor response in relation to tumor absorbed dose for all lesions evaluated with a diameter >2.2 cm (blue circles) and for lesions with a diameter >4.0 cm (red triangles). Taken from Cremonesi et al. [29], who adapted it from the original source, Ilan et al. [65]

Full size image
Table 6.3 Reported dose-response curves in radionuclide therapy
Full size table

When studying the results of Table 6.3 one should not forget all studies, that didn’t report a dose-response relationship for lesions in radionuclide therapy. Substitutionally, two examples are given here: Jahn et al. [66] were unable to relate tumor shrinkage or biochemical response to the AD for small intestinal NET. Similarly, Barna et al. [67] looked into the dose-effect relationships in 177Lu-PSMA I&T radionuclide therapy for metastatic castration-resistant prostate cancer by investigating 217 possible correlations between dosimetric quantities, biomarkers, and tumor shrinkage and only found 37 of them to be statically significant, none of them related to tumor reduction. For the sake of completeness, it has to be added that this study relied on planar scintigraphy rather than SPECT/CT.

In discussing the correlation between radiation dose and tumor reduction one should not forget the circumstances of radionuclide therapy, in particular the quantity and availability of data which of course has a huge impact on the likelihood to observe certain phenomena. Table 6.3 illustrates this issue by the fact that only one study actually performed voxel-based, patient-specific dosimetry [49] which inevitably is necessary for the calculation of a voxel-based quantity such as the EUD. This way the question remains whether the EUD is not reported more frequently because (a) the associated prediction of tumor response is poor or (b) simply because of the lack of studies that take the trouble and effort to perform this more complex and laborious form of dose calculations.

Likewise, when comparing the number of cases in EBRT with the ones in Radionuclide Therapy, it is not realistic to expect the same degree of consolidation regarding dose-response relationships for tumor tissue. Moreover, patients undergoing radionuclide therapy almost always had several previous treatments, like hormonal therapy and/or chemotherapy, all of which affect the immune system as well as the individual state of health and subsequently also the tumor response. Certainly, the radio-oncologic lessons learned from EBRT cannot be extrapolated straightforward to Radionuclide Therapy, which in its application form rather resembles a systemic therapy than a local one.

6.6 The Answer to the Gretchenfrage

Facing the load of scientific studies cited above we return to the beginning, the Gretchenfrage:

How do we feel about dosimetry in Radionuclide Therapy?

In the scientific context one might reformulate it to:

What is the benefit of dosimetry in Radionuclide Therapy?

The term benefit in this discussion shall not only relate to scientific exploration, i.e., the investigation of dose-response relationships, but also to advantages for the individual patient and the clinical workflow. In EBRT the often-used keyword “Personalized Medicine” is realized by tailoring the spatial dose distribution to the individual case by means of several, adjustable parameters, such as angular distribution, frequency, and intensity modulation. Radionuclide Therapy is far away from this luxury and basically only has one adjusting screw, the administered activity. Here, personalized medicine implies the tailoring of the injected activity for the individual patient in a way, that malignant tissues receive the highest possible dose without the occurrence of radiotoxicities, thus avoiding under- as well as overtreatment.

Dosimetry calculations can do exactly that and in case of the combinations 124I/131I as well as 111In/86Y/90Y enable an a-priory treatment planning. The diagnostic scan yields the tracer’s biokinetics and uptake in lesions which are then used in the computation and prediction of the dose distribution of the therapeutic nuclide [19, 25,26,27, 31,32,33,34, 48]. In doing this, either phantom-based dose calculations including the spherical model for lesions or voxel-based dosimetry based on patient-specific imaging data can be applied. Both procedures, in smaller or greater detail, provide the possibility of tailoring the administered activity for the sake of the best, individualized treatment. Reminding Table 6.3, voxel-based dosimetry does not necessarily produce a higher correlation coefficient. A voxel also constitutes a finite volume and there is no guarantee that it provides the absorbed dose at the biologically relevant scale, since dosimetry at a microscopic level remains inaccessible [9]. Nonetheless, it has to be kept in mind that the a-priory knowledge of the dose to malignant tissue in combination with published threshold doses for tumor response [55] and/or dose-response curves as in Table 6.3 enable the assessment of the expected course of the disease and therefore allow for prospective treatment planning.

This ideal scenario does of course not work for all forms of radionuclide therapy as outlined in Table 6.1. 68Ga-PET/CT scans may be feasible for the correlation of lesions’ SUV in PET with the absorbed dose from the later therapy or for the selection of appropriate candidates for PRRT (see [8] for a review), but cannot be used for a real dose calculation due to the short half-life of 68Ga which is unable to produce the necessary biokinetics of the later phase. Still, dosimetry based on the accompanying 177Lu scans is very useful, since the evaluation of the calculated organ and lesion doses of a cycle provides the necessary information to adjust the administered activity for all futures cycles and therewith enable a patient-specific optimization of radionuclide therapy. The incorporation of multiple cycles into the BED concept for kidney toxicity [53] is a good example for the value of optimization.

There is of course no guarantee that subsequent cycles will show the same relation between administered activity and absorbed dose, since uptake and biokinetics of irritated tissue, in particular lesions, will vary. Both, studies showing big differences [68] as well as minimal ones [69, 70], can be found in the literature. Another important result reported by Garkavij et al. [69] is that patients evaluated with planar-based dosimetry may have been undertreated compared to other methods. This is confirmed by Zechman et al. [71] who showed in their review that the absorbed doses to the kidneys are systematically overestimated when using planar imaging.

Last but not least, in order to systematically investigate dosimetry in Radionuclide Therapy and dose-response effects, one has to do dosimetric studies. Solid and quantitative data which has way higher R2 than in Table 6.3 are a prerequisite for personalized medicine and subsequently for the patient’s welfare. Dosimetry might not be the only predictor for this, but certainly is an essential one. It’s neither a magic flute nor will it generically explain all the effects in Radionuclide Therapy. But it can be used for personalized treatment planning as well as optimization, and it is getting better and more accurate with every new study performed.

Dr. Faust, in his despair to gain knowledge, even gives in to magic. Luckily as scientist we can rely on reason and evidence to face the same challenge:

That I may know what the world contains

In its innermost heart and finer veins.

Faust; a Tragedy. Johann Wolfgang von Goethe