Radiation Dose Distribution in Human Kidneys by Octreotides in Peptide Receptor Radionuclide Therapy

MATERIALS AND METHODS

Autoradiogram data of kidney samples from 3 patients with renal tumors were used. The methods used in gathering the ex vivo autoradiographs were described earlier (13). In brief: the male patients were injected with 220 MBq ¹¹¹In DTPA-octreotide 2–4 d before nephrectomy. Disease-free kidney tissue sections of 10-μm thickness were exposed for 1 d to a phosphor imaging screen (Packard Instruments). The screens were analyzed with a Cyclone phosphor imager (Packard) with a spatial resolution of 500 μm for ¹¹¹In.

The radioactivity distributions from the autoradiography data files were averaged over 4 × 4 pixel elements for the kidney samples obtained at 96 h (Fig. 1A) and at 72 h (Fig. 1B). The finer structure in the activity distribution of the sample at 48 h (Fig. 1C) made averaging over 3 × 3 pixels necessary. Pixel size was 0.092 mm/pixel. The averaging was performed to reduce the amount of input data and computation time. Furthermore, the resulting sizes of the voxels (0.37 and 0.28 mm, respectively) are more in proportion to the spatial resolution of the imager. Boundary surfaces were fitted to the autoradiographs, both on the outside border of the kidney and also between the cortex and medulla. The averaging and fitting was performed with the Matlab code (The Mathworks, Inc.). The equations for these surfaces were used to define the kidney segments in the Monte Carlo code MCNP5 (15) (Oak Ridge National Laboratory, Los Alamos, NM). The segments were made 1-cm thick to generate a 3D structure. The segments were superimposed by a rectangular grid structure corresponding to the averaging voxels (0.37 × 0.37 × 10 mm³ for Figs. 1A and 1B and 0.28 × 0.28 × 10 mm³ for Fig. 1C) for both the source definition and the dose calculation.

A stylized model of the 3D activity distribution was made for the MCNP5 code inside the MIRD19 multiregion kidney phantom (8) and is shown in Figure 2. The volumes of cortex, medulla, and pelvis were superimposed with 3-mm-sized voxels. Inside the voxels in the cortex volume cylinders of 1.7-mm diameter were positioned in the center to represent the streaky uptake pattern seen on the autoradiograph. The activity distribution in the 3D model was according to the average distributions on the 3 autoradiographs indicated in Table 1 (13). The cortex high radioactivity is assigned to the central cylinders in the voxels and to a 1.8-mm-thick region in the cortex aligning the medulla (JM region). The cortex low radioactivity is assigned to the remaining voxels in the cortex. The activity distribution in the medulla was assumed to be homogeneous, as the autoradiography results do not show the level of heterogeneity found in the cortex. This model is referred to as 3D voxel.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

Adult-size MIRD19 multiregion kidney model in sagittal view (x−z) on left and in coronal view (y−z) on right, with overlying 3-mm-voxel grid in medullae (green), pelvis (yellow), and cortex (blue/red pattern). Voxels in cortex were given a 1.7-mm cylindric core representing medullary rays (MR in dark blue); together with a 1.8-mm-thick JM (in light blue) region they form the high-activity-uptake volume. Low-uptake cortex volume (non-MR) is shown in red. Papillae were not voxelized and are colored purple.

As the reference condition, a homogeneous radioactivity distribution of 70.8% in the cortex and 29.2% in the medullae was considered. Two methods were applied to calculate the kidney dose for this situation. The data from MIRD19 were used for method 1 (named as MIRD19). In the second method the same 3-mm voxel structure was superimposed over the MIRD19 kidney model, without the central cylinder inhomogeneity, however (named as M19 voxel).

The Monte Carlo calculations with MCNP5 were performed to calculate the radiation transport inside the geometries for the high-energy β-emitter ⁹⁰Y; the γ-ray, low-energy β-ray, and internal conversion electron emitter ¹⁷⁷Lu; and the γ-ray and Auger electron emitter ¹¹¹In. The emission data were obtained from the RADAR Web site (16). The typical number of particle histories necessary to constrain statistical fluctuations in the total energy absorbed for each voxel below 5% was 10 million for the electrons and 3 million for the γ-rays. MCNP computing times were approximately 50 h for the β-rays and approximately 100 h for the γ-rays on a 3-GHz personal computer.

A substantial part in the Auger electron spectrum (13% of energy emitted) of ¹¹¹In is below the lowest energy threshold for MCNP of 1 keV. As no electron tracks are generated below this threshold, MCNP discards the energy emitted by these low-energy Auger electrons. The missing energy was added to the source voxels in the postprocessing phase.

The output data were analyzed with Matlab and Excel (Microsoft Corp.) software to produce isodose curves and DVHs. Summation of the different radiation components of the deposited energy for ¹¹¹In and ¹⁷⁷Lu was also performed in Matlab. The dose per decay to each voxel was calculated by dividing the deposited energy by its voxel mass. The relative biological effectiveness was equal to one for each component (β-rays, γ-rays, or low-energy (Auger) electrons). Both the dose distributions and the average dose to each kidney segment in the MIRD19 model were calculated to enable comparison with the published data from MIRD19.

RESULTS

DVHs

DVHs for the 3 radionuclides are shown in Figures 3, 4, and 5 for the homogeneous and inhomogeneous radioactivity distributions. The normalization of the dose was based on the MIRD19 dose to the cortex, which was set to a mean dose of 27 Gy. The DVHs for the homogeneous distribution show an almost block form around the average dose with deviations being caused by the penetrative nature of the radiation; ¹¹¹In with its γ-rays and ⁹⁰Y with its high-energy β-particles show a more shallow curve in comparison with the low-energy β-emitter ¹⁷⁷Lu. The DVH for the kidney volume resembles that for the cortex volume, apart from the low dose part for ¹⁷⁷Lu and, to a lesser extent, for ⁹⁰Y, where the zero to low dose to the pelvis (4.3% of the kidney volume) manifests itself. In the dose region between 20 and 30 Gy, the dose to the medulla shows up as a shoulder in the reduction of the kidney volume. Although the irradiation differs quite strongly from inhomogeneous irradiation by external x-ray beams, the same methods (17) can be used to extract effective volumes from these DVHs. Effective volumes together with other DVH characteristics are summarized in Table 4.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

DVHs for kidney (A) and cortex (B) for ⁹⁰Y source activity: 71% in cortex and 29% in medullae in either a homogenous distribution or a distribution following the mean autoradiograph results. Average dose to cortex was taken as 27 Gy, calculated according to MIRD19 (8).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4. 

DVHs for kidney (A) and cortex (B) for ¹¹¹In source activity: 71% in the cortex and 29% in medullae in either a homogenous distribution or a distribution following the mean autoradiograph results. Average dose to cortex was taken as 27 Gy, calculated according to MIRD19 (8).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5. 

DVHs for kidney (A) and cortex (B) for ¹⁷⁷Lu source activity: 71% in cortex and 29% in medullae in either a homogenous distribution or a distribution following the mean autoradiograph results. Average dose to cortex was taken as 27 Gy, calculated according to MIRD19 (8).

View this table:

• View inline
• View popup

TABLE 4

Summary DVH Data for Homogeneous and Heterogeneous Activity Distribution in Kidney Based on a 27-Gy Dose to Cortex Following MIRD19 (8)

When the radioactivity is assumed to be homogeneously distributed, 5% of the kidney volume would get a dose above 31–32 Gy (D₀₅) for all nuclides. The location of this maximum was outside the cortex, as the cortex D₀₅ ranges between 27 and 30 Gy. The relative volumes receiving 70% of the mean dose of (70% x 27 = ) 18.9 Gy (V₇₀) are for the homogeneous distribution around 75% of the total renal volume.

With the heterogeneous radioactivity distribution D₀₅ rises considerably for ¹¹¹In and ¹⁷⁷Lu to 39 and 46 Gy, respectively. The maximum dose is then found within the cortex. V₇₀ drops by 30% with the heterogeneous distribution and for the cortex even by 40%. The effective volume (17,18) shows the lowest values for the heterogeneous distribution of ¹⁷⁷Lu.

DISCUSSION

In nuclear medicine traditionally doses are based on volume-averaged uptake values over the region of interest (ROI) using the planar imaging conjugate view method. With this method the only possible outcome is a mean dose to the organ by fitting a time–activity curve to the ROI data and using the resulting cumulated activity as input for a dosimetry code, such as OLINDA (16). With quantitative PET and in future quantitative SPECT imaging, activity uptake patterns can be measured 3-dimensionally. Ideally, also the dosimetry will have to go to 3 dimensions. In PRRT with, for example, ⁹⁰Y-DOTA-Tyr³-octreotide it is of great importance to know the constraints in using average kidney dose as a predictor of end-stage renal disease. Replacing average dose by DVHs may be necessary for short-range β-emitters and electron emitters when the activity is heterogeneously distributed.

There is little knowledge on the effect of heterogeneous dose distributions in the kidneys. In rats the effects of regional distribution of the renal radioactivity uptake has been studied as a function of the size of the antibody to which the radionuclide is linked (19). More efficient filtering of smaller antibodies by the glomeruli relatively increases the cortical uptake. Peptide reabsorption occurs in the proximal tubules. Isodose curves were calculated and showed that the combination of short-ranged radiation and small-size antibodies (or peptides) made separate dose calculations for the rat cortex and medulla necessary.

The heterogeneous uptake pattern of ¹¹¹In-DTPA-octreotide in the human kidney has a large influence on the dose distribution inside the kidney, especially for short-range β-emitters and electron emitters. The calculated isodose curves for ¹¹¹In and ¹⁷⁷Lu show steep gradients between the regions with high and low radionuclide uptake. DVHs are a better instrument for evaluating the differences between the 3 radionuclides considered in this study. Especially ¹⁷⁷Lu seems promising in tolerating higher kidney doses in comparison with ¹¹¹In and ⁹⁰Y. With an average dose of 27 Gy to the cortex, 60% of the cortex volume (D₄₀) receives a dose below 7.1 Gy with ¹⁷⁷Lu. ¹¹¹In gives a D₄₀ below 17 Gy, and ⁹⁰Y shows a D₄₀ of 17.5 Gy. The maximum dose to the cortex is, however, the highest for ¹⁷⁷Lu: 5% of the cortex volume (D₀₅) gets a dose above 47 Gy, whereas ¹¹¹In gives a D₀₅ of 40 Gy and ⁹⁰Y gives a D₀₅ of 25 Gy.

Differences between the dose conversion factors derived from MIRD19 and the values for homogeneous voxelized distribution are evident for ⁹⁰Y and ¹⁷⁷Lu. This could be due to the use of different Monte Carlo codes; for MIRD19 the EGS code was used, whereas in this work MCNP was used. Also the voxel structure may cause boundary problems, such as partial-volume effects in voxels crossing the kidney−region boundaries.

By focusing on the dose to the cortex we assume that this region is responsible for the radiation-induced end-stage renal disease observed after radionuclide therapy in some patients. The experience in external beam radiotherapy (XRT) kidney dosimetry is quite contrary compared with PRRT: The dose to the outer regions of the cortex is lower than that to the regions bordering the medulla. Also the heterogeneity in the radioactivity uptake in the kidneys for PRRT could very well show us the functional subunits (FSUs) of the kidney: the nephrons. Unfortunately, it has not been possible to get good histology of the tissue responsible for the heterogeneous renal radioactivity uptake pattern of ¹¹¹In-DTPA-octreotide. It is then questionable whether the radiation dose–response of the kidneys still can be considered to follow the behavior typical for parallel structures; in external beam therapy this was concluded from the fact that the critical fraction of FSUs needed to maintain kidney function was determined to be 0.77 (20). Adjacent unirradiated FSUs will take over the function of the irradiated FSUs, and cell migration from the unirradiated part to the irradiated elements may restore the functional status of the kidney.

The structure of the radiation damage with PRRT is, however, very different; it is expected to be maximal in the JM region with all tubuli leading into the medulla. When all tubuli in this region get doses well above the threshold for damage, no neighboring elements will remain to take over their function. Whether cell migration from up- or downstream unirradiated parts of the affected tubuli is fast enough to repair its function seems unlikely. However, the observed clinical experience is that high-dose ¹¹¹In-DTPA-octreotide and ⁹⁰Y-DOTA-Tyr³-octreotide therapy, with average kidney doses of 27 Gy, leading to JM doses of 39 and 32 Gy, respectively, for the heterogeneous distribution was given without subsequent major renal problems (1). More specifically, the tubular function showed no clinically relevant deterioration, either shortly after PRRT or during long follow-up. Therapy with ¹⁷⁷Lu-DOTA-Tyr³-octreotate at an average kidney dose of 23 Gy, which still leads to a JM dose of 40 Gy, is reported to be even less toxic to the kidneys. The distribution and the size of the glomeruli in the cortex are not homogeneous (21,22), but it is unclear whether this also reflects their contribution to the total kidney function. High doses to the JM regions in the cortex could very well devastate the glomeruli in that region, so the renal function must be maintained only by the glomeruli in the outer cortex. Still, both in the dose–response of the kidney (6) as in the clinical pattern of radiation-induced renal disease by PRRT, no overt unexplained differences were observed from the situation after kidney radiation exposure by XRT.

A commonly applied method in XRT for reducing the DVH data to a single dose parameter is the effective volume method (17,18). Each voxel for a parallel structured organ obeys the same dose–volume relationship as the whole organ. The tolerance dose TD(v) after uniform radiation to a partial volume v depends on the tolerance dose for uniform irradiation of the whole organ (v = 1) by the Lyman equation: TD(v) = TD(1) · v⁻ⁿ .The volume dependence parameter n = 0.7 (23) was obtained by fitting to the kidney tolerance data of Emami et al. (24). Each step in the kidney DVH of volume ΔV_i with dose D_i is then transformed to an effective volume in the adjacent dose bin with higher dose, which is iterated until each volume section of the DVH is transformed to the effective volume with a maximum dose D_m by:

The effective volumes for ⁹⁰Y, ¹¹¹In, and ¹⁷⁷Lu are given in Table 4. Homogeneous radioactivity distribution in the cortex creates a dose distribution as if just 56% (¹¹¹In) to 72% (¹⁷⁷Lu) of the cortex volume is homogeneously irradiated with the maximum dose of 30 Gy and 27 Gy, respectively. When the activity is distributed according to the autoradiography data the effective volume drops especially for ¹⁷⁷Lu to 24% with a higher maximum at 47 Gy. In the case of ⁹⁰Y the effective volume shows a lower decrease when introducing heterogeneity, but also the maximum drops from 29 to 25 Gy, as the MIRD19-based dosimetry overestimates the dose to the cortex for inhomogeneous ⁹⁰Y activity distribution.

In PRRT there is hardly any possibility of actively shaping its DVH to a safer profile, unlike in XRT. The only possibility would be to change from ⁹⁰Y to ¹⁷⁷Lu, when the activity is distributed heterogeneously. Again, it must be stressed that partial irradiation in XRT fields is very different from the dose heterogeneity observed with PRRT. In XRT either a single kidney or the upper poles of both kidneys are irradiated. It is, however, logical to expect that the threshold for late kidney radiation damage after therapy with ¹⁷⁷Lu-DOTA-Tyr³-octreotate will show an even higher shift from the external beam threshold than that observed with ⁹⁰Y-DOTA-Tyr³-octreotide, both due to the heterogeneous dose distribution as well as due to the longer half-life. Taking the Lyman equation literally, this shift for the effective volume of 24% for ¹⁷⁷Lu will be a factor of (v⁻ⁿ = 0.24^−0.7 = ) 2.7. In reality this shift will be lower, as the observed heterogeneity is undoubtedly a transient situation, but even with the homogeneous activity distribution V_eff = 72% leads to a shift in tolerance dose with a factor of 1.26. This would indicate a tolerance dose for ¹⁷⁷Lu of 29 Gy instead of 23 Gy.

CONCLUSION

Heterogeneous radioactivity distribution in the kidneys has only consequences to the dose distribution for low-energy β-emitters and electron emitters such as ¹¹¹In and ¹⁷⁷Lu. Large volumes of the cortex get smaller doses than average with ¹¹¹In and ¹⁷⁷Lu. The higher doses to the kidney regions with high radioactivity uptake do not create a higher incidence of renal damage in clinical practice compared with therapy with ⁹⁰Y. The use of DVHs instead of average doses is strongly advocated for all radionuclides considered, irrespective of any heterogeneity in activity distribution over the kidney. In the case of the short- and medium-range electron emitters, such as ¹¹¹In and ¹⁷⁷Lu, DVHs give a better indication of relative volumes that will exceed tolerance doses. PRRT based on ⁹⁰Y, emitting longer-ranged β-rays, will create more dose averaging inside the kidney regions. The consequences of heterogeneous dose distribution in the kidney for radiation damage to its functional subunits are not clear. However, if it follows the Lyman relationship, an additional sparing of the kidney of minimally a factor 1.25 can be expected. It is, however, rarely possible to obtain activity distribution data on the tissue level.