Letter to the Editor

In a recent publication, Ricke et al. [1] described the safe 90Y radioembolization of lung tumors using resin microspheres. There also are recent reports of 90Y radioembolization being safely performed to the kidney [2] and spleen [3]. However, as early as 1965, Blanchard et al. [4] showed that 90Y radioembolization to sites other than the liver is risky, with potential for serious nontarget radiation toxicity. We are recently reminded that safe and effective 90Y radioembolization is a combination of “technical success” and “clinical success” [5], the latter being a radiobiological function of tissue dose-response.

The aim of this Letter is to provide a dosimetric roadmap on predictive dosimetry for safe and effective 90Y radioembolization to sites other than the liver. The concepts presented here are of a general nature, and user discretion is advised in accordance to the specific nature of each case, tissue radiobiology, dosimetric limitations, institutional expertise, equipment, and ethical standards.

Predictive dosimetry requires 90Y microsphere biodistribution simulation using a suitable inert surrogate, e.g., 99mTc macroaggregated albumin [6, 7], assessed by scintigraphy of the target arterial territory and relevant regions-at-risk. Conceptually, simulation scintigraphy should be regarded as a dosimetric map depicting the predicted final positions of all injected microspheres, where all possible flow-shunt-implantation distributions have irreversibly occurred.

Dosimetric Assumptions

(1) Only a single arterial territory is undergoing 90Y radioembolization; (2) 90Y microspheres are injected under identical technical and physiological circumstances as its surrogate (e.g., catheter position and type, hemodynamic status), as far as possible; (3) there are no more than three tissue compartments for dosimetry, i.e., tumor, nontumorous tissue, and lung (Fig. 1); (4) vascular stasis and microsphere reflux has not occurred at the conclusion of 90Y radioembolization; (5) implanted tissue activity is proportional to scintigraphic counts, and its quantification is accurate; (6) there is no postimplantation redistribution of microspheres nor leeching of 90Y activity; (7) there are no arteriovenous shunts from nontumorous tissue contributing to lung shunting of microspheres. If the user suspects that arteriovenous shunting may be occurring within nontumorous tissue (e.g., congenital vascular malformations, previous surgery, or trauma), the entire predictive dosimetry will revert in principle to the original “partition model” [8] as was developed for 90Y radioembolization to the liver. This is because the original partition model accounts for the possibility of arteriovenous shunts occurring from nontumorous cirrhotic liver parenchyma.

Fig. 1
figure 1

Concept of tricompartmental MIRD predictive dosimetry involving the tumor, nontumorous tissue, and lung

Full size image

General Points

(1) Minimum technology for simulation scintigraphy should be the single photon emission computed tomography with integrated CT (SPECT/CT); (2) no recommendations on desired absorbed doses are made here, as these depend on case-specific tissue radiobiology, technical factors and treatment intent (e.g., palliative, downstage, or cure); (3) dosimetric implications of the parallel-serial model in radiobiology, and the different radiobiological effects of 90Y resin versus glass microspheres are not discussed here; (4) 90Y glass microspheres, which have higher specific activity and lower microembolic potential, may have theoretical advantages in safety and efficacy over 90Y resin microspheres in situations where vascular capacitance is anticipated to be low (e.g., intratumoral injection).

General Dosimetric Theory

Within a closed system where all injected activity is conserved, the desired total activity of 90Y microspheres to be injected, A Total (GBq) is:

$$A_{\text{Total}} = A_{\text{T}} + A_{\text{N}} + A_{\text{L}}$$
(1)

A T is the total activity implanted within all tumor tissue, after shunting to nontumorous tissue and lung have occurred; A N is the total activity implanted in all nontumorous tissue within the target arterial territory; A L is the total activity implanted in lung due to tumor-to-lung shunting.

The tumor-to-lung shunt fraction (TLSF), is defined as:

$${\text{TLSF}} = \frac{{A_{\text{L}} }}{{A_{\text{T}} + A_{\text{L}} }}$$
(2)

Equation 2 is approximated by \(C_{\text{L}} /(C_{\text{T}} \; + \;C_{\text{L}} )\), where C L and C T are the total counts in lung and all tumor tissue respectively, quantified by scintigraphy. The TLSF assumes no lung shunt contribution from nontumorous tissue, and is therefore conceptually different from the “liver-to-lung shunt fraction” defined by the original partition model [8].

The ratio of total activity in tumor to nontumorous tissue, R, is defined as:

$$R = \frac{{A_{\text{T}} }}{{A_{\text{N}} }}$$
(3)

Equation 3 is approximated by\(C_{\text{T}} /C_{\text{N}}\), where C N is the total counts in all nontumorous tissue quantified by scintigraphy. R is a ratio of activities, not of mean radioconcentrations and, therefore, is conceptually different from the “tumor-to-normal liver ratio” defined by the original partition model [8].

The critical step in predictive dosimetry requires the user to decide an appropriate intended tissue mean absorbed dose to be delivered to any one of the three dosimetric compartments, i.e., tumor, nontumorous tissue, or lung. In this example, let us give a value to the intended tumor mean absorbed dose, D T (Gy), defined as:

D T = 50 \(\left( {A_{\text{T}} /m_{\text{T}} } \right)\) by Medical Internal Radiation Dose (MIRD) schema, where “50” is the approximate absorbed dose coefficient of 1GBq of 90Y uniformly distributed throughout 1 kg of tissue [9], and m T (kg) is the total mass of all tumors within the target arterial territory. Tissue mass (m) may be estimated by CT volumetry and its mean radiographic density on a calibrated system, with attention to accurate delineation of volumes-of-interest. Therefore, A T may be expressed as:

$$A_{\text{T}} \; = \;\frac{{D_{\text{T}} m_{\text{T}} }}{50}$$
(4)

From Eqs. 3 and 4, the intended mean absorbed dose to non-tumorous tissue, D N (Gy), of mass m N (kg), is therefore:

$$D_{\text{N}} \; = \;\frac{{D_{\text{T}} m_{\text{T}} }}{{Rm_{\text{N}} }}$$

From Eqs. 2 and 4, the intended mean absorbed dose to lung, D L (Gy), of mass m L (kg), is therefore:

$$D_{\text{L}} \; = \;\left( {\frac{{D_{\text{T}} m_{\text{T}} }}{{m_{\text{L}} }}} \right)\left( {\frac{\text{TLSF}}{{1\; - \;{\text{TLSF}}}}} \right)$$

Applying Eqs. 2, 3 and 4 into Eq. 1:

$$A_{\text{Total}} \; = \;\frac{{D_{\text{T}} m_{\text{T}} }}{50}\; + \;\frac{{D_{\text{T}} m_{\text{T}} }}{50R}\; + \;\left( {\frac{{D_{\text{T}} m_{\text{T}} }}{50}} \right)\left( {\frac{\text{TLSF}}{{1\; - \;{\text{TLSF}}}}} \right)$$

In the simplest dosimetric scenario, all injected 90Y microspheres are completely implanted within tumor, i.e., both A N and TLSF are zero (Fig. 2). Therefore, Eq. 1 is reduced to simply:

$$A_{\text{Total}} = A_{\text{T}} = \frac{{D_{\text{T}} m_{\text{T}} }}{50}$$
(5)
Fig. 2
figure 2

The simplest dosimetric scenario where all injected 90Y microspheres are completely implanted within tumor

Full size image

If nontumorous tissue is present within the target arterial territory (i.e., A N >0), but there is no tumor-to-lung shunting (i.e., TLSF = 0), then Eq. 1 is reduced to:

$$A_{\text{Total}} \; = \;A_{\text{T}} \; + \;A_{\text{N}} \; = \;\frac{{D_{\text{T}} m_{\text{T}} }}{50\,}\; + \;\frac{{D_{\text{T}} m_{\text{T}} }}{50R}$$

If tumor-to-lung shunting is present (i.e., TLSF >0), but there is no nontumorous tissue within the targeted arterial territory (i.e., A N  = 0), then Eq. 1 is reduced to:

$$A_{\text{Total}} \; = \;A_{\text{T}} \; + \;A_{\text{L}} \; = \;\frac{{D_{\text{T}} m_{\text{T}} }}{50}\; + \;\left( {\frac{{D_{\text{T}} m_{\text{T}} }}{50}} \right)\left( {\frac{\text{TLSF}}{{ 1\; - \;{\text{TLSF}}}}} \right)$$
(6)

In cases of multiple small or ill-defined tumor lesions which cannot be reliably quantified or delineated on imaging (e.g., diffuse lymphomatous infiltration), the user may choose to conceptually regard the entire target arterial territory as “tumor” (Fig. 3) and plan a single intended mean absorbed dose across the entire target arterial territory. Its dosimetry will be analogous to either Eqs. 5 or 6, depending on the TLSF.

Fig. 3
figure 3

For multiple, small or ill-defined tumor lesions, the entire target arterial territory may be conceptually regarded as “tumor” and treated by a single intended mean absorbed dose

Full size image

Technology and innovations that may improve the safety and effectiveness of 90Y radioembolization to sites other than the liver are: (1) hybrid tomographic scintigraphy with integrated CT, i.e., SPECT/CT [7, 8] or PET/CT [10]; (2) antireflux catheters to minimize reflux of 90Y microspheres into nontarget tissue [11]; (3) Voxel dosimetry and dose-volume histograms to overcome inherent limitations of dosimetric planning based on mean absorbed doses [12]; (4) multi-isotope surrogates to qualitatively and quantitatively assess microsphere biodistribution within multiple target arterial territories [13, 14]; (5) future possibility of combined external beam radiotherapy with 90Y radioembolization.