Reducing the number of CTs performed to monitor personalized dosimetry during peptide receptor radionuclide therapy (PRRT)

¹⁷⁷Lutetium (¹⁷⁷Lu) is a useful radionuclide in several targeted peptide receptor radionuclide therapies (PRRTs) due to its favorable decay characteristics and the possibility of reliable labeling of biomolecules used for tumor targeting. Being both a beta and gamma emitter, the therapeutic effect achieved thanks to beta emission can be monitored by a gamma camera with planar or SPECT imaging. In order to maximize the treatment efficiency of [¹⁷⁷Lu]-DOTA-TATE [1–3], shown to be an effective therapy of neuroendocrine tumors (NETs), the amount of ¹⁷⁷Lu radioactivity to be administered has to achieve an optimal therapeutic effect of radioligand therapy, i.e., to lead to a maximal absorbed dose in the tumors with limited side effects to the most radiosensitive non-pathological tissues, namely the kidneys and bone marrow [4–7]. For this purpose, individual dosimetry after each cycle of treatment is necessary to evaluate the cumulative absorbed dose to the organs at risk and to decide whether the patient can continue to receive further treatments. Recommended schedule of treatment with [¹⁷⁷Lu]-DOTA-TATE [8] consists of a fractionated therapy of four 7.4 GBq (200 mCi) injections spaced by 6–12 weeks, provided that the cumulative dose will not exceed the safety limits of 23 Gy [9–11] for the kidneys and 2 Gy [9, 12] for the bone marrow. It has previously been shown that in most cases, the kidneys are the dose-limiting organs in PRRT treatments with [¹⁷⁷Lu]-DOTA-TATE [9]. However, a recent study by Bergsma et al. [13] showed low nephrotoxicity after PRRT with a cumulative activity of 29.6 GBq (800 mCi) of [¹⁷⁷Lu]-DOTA-TATE and a mean calculated radiation dose to the kidneys of 20.1 ± 4.9 Gy. Three of their 323 patients developed renal toxicity grade 2, and no grade 3 or 4 renal toxicity occurred. Other authors [7, 14] have argued that higher safe absorbed doses to the kidneys (28–29 Gy) could probably be adopted due to the inhomogeneous distribution at a microscopic level of PRRT with [¹⁷⁷Lu]-DOTA-TATE and to the short range of its β-particles in the tissue.

In order to estimate the pharmacokinetics of [¹⁷⁷Lu]-DOTA-TATE and to calculate the radiation doses absorbed by the patient’s organs, quantitative single photon emission computed tomography (SPECT) images corrected for photon attenuation (from computed tomography (CT) attenuation maps) and blurring (resolution recovery) are acquired after each cycle of treatment [15, 16]. Following the EANM/MIRD guidelines [17], full dosimetry is performed after the first treatment with three SPECT/CT acquisitions at 24, 96, and 168 h after the administration of [¹⁷⁷Lu]-DOTA-TATE. For the subsequent treatment cycles, only a single SPECT/CT study approximately 24 h after the treatment is performed assuming an unchanged effective half-life of [¹⁷⁷Lu]-DOTA-TATE between treatments as proved and proposed by Garske et al. [12]. In our study, the radioactivity distribution and pharmacokinetics of [¹⁷⁷Lu]-DOTA-TATE in the kidneys, liver, spleen, remainder of the body, and tumors have been estimated by drawing volumes of interests (VOIs) either in the anatomical (CT) or functional (SPECT) images. The kinetics of radioligand in the bone marrow has been estimated from the radioactivity concentration measured in blood samples taken after the treatment. Indeed, Forrer et al. [18] showed that the activity concentration in the bone marrow after the treatment is equal to that in the blood.

A typical patient, completing 4 cycles of treatment, would have done six SPECT/CTs across his whole PRRT treatment. Across the different cycles of treatment, the acquired CT may show changes that can be caused by the changes in the patient position under the gamma camera, response to treatment or progressive disease, change in patient weight, or gastric emptiness or fullness. However, due to the short time interval between courses of treatment and the usually slow response to treatment (or alternatively disease progression) in a population of patients receiving PRRT, our impression was that in most cases, no visually blatant changes were observed on the CTs between treatments. Therefore, for a given patient, attenuation maps based on subsequent CTs across the cycles of treatment should be very similar. Consequently, we hypothesize that the use of the attenuation map generated from the CT of the first SPECT/CT acquisition provides a sufficiently good match to be used for subsequent SPECTs.

The main aim of this work was to compare, over the whole series of treatment, the dosimetry results for the kidneys and bone marrow based on the “classical” protocol to those based on a new protocol. In the “classical” one, dosimetry is calculated based on a SPECT and CT after each cycle. In the new one, the dosimetry is calculated based on a SPECT after each cycle of treatment but on one single CT for all the cycles. Individual patient dosimetry along the treatment is used to design a patient-tailored treatment schedule. At our institution, before giving a subsequent PRRT treatment, the absorbed doses to the kidneys and bone marrow during the previous p treatments are examined and an expected cumulative dose after the following (p + 1) cycle is determined. The latter is calculated as the mean dose over the previous p treatments to which is added the cumulative dose absorbed over these p treatments. In our institution, we withheld PRRT if the expected cumulated dose after the next treatment will exceed by 10% the safe dose limits, i.e., 25.3 Gy for the kidneys and 2.2 Gy for the bone marrow (except for very specific clinical setups where cost/benefit ratio led us to continue treatment beyond these limits). In this work, we compared the actual patient management (whether or not treatment was stopped due to dosimetry) to the hypothetical management of the patient, had we used the new “single CT” protocol.

Percentage differences in the absorbed doses calculated using the two protocols will be shown on Bland and Altman [19] plots for the kidneys and bone marrow. Also, Pearson’s correlation coefficient between data obtained from one method versus another has been computed and will be presented for the two risk organs. Finally, the intra- and inter-observer reproducibility will be also calculated.

Figure 3 shows the expected cumulative doses for kidneys calculated using either the “classical” or the “single CT” protocol. For all the 24 patients, the hypothetical management based on the new “single CT” protocol would have been similar to the “classical” management based on the whole series of SPECT/CTs as shown in this figure. Indeed, all the patients discontinued/not discontinued with the “classical” protocol would have been discontinued/not discontinued by using the new “single CT” protocol. Our null hypothesis was that there was no difference in the treatment management of the patients between the two protocols. A sample size of 24 patients has a power of 92% to detect a true difference of at least 10% at the 0.05 level of significance, using a one-sided binomial test. We used StatXact, Cytel Inc., Cambridge MA, version 10 20.

Figure 4 shows the differences in the absorbed doses to the kidneys and bone marrow calculated from the two different protocols for all the 83 treatments included in this study. The Bland and Altman plots, showing the relative difference between the two calculation methods, and the correlation curves, testifying to the degree of relationship between them, are plotted in this figure. Similarly, Fig. 5 presents the results for the cumulative dose absorbed to these organs at risk for the 24 patients included in this work. It is noteworthy that the agreement between the two methods was excellent over all series of therapy cycles (cf. Fig. 4). Indeed, for the kidneys, mean relative differences of 5.94 ± 9.38% over the 83 therapy cycles and 5.30 ± 6.20% over the 24 cumulative doses has been obtained with a Pearson’s correlation coefficient r = 0.97 and r = 0.95, respectively (all P < 0.0001). For the bone marrow, the absorbed doses have been obtained with mean relative differences of 0.91 ± 7.24% over all the therapy cycles and 0.48 ± 4.88% for the cumulative doses with r = 0.99 (all P < 0.0001). It is noteworthy from these results that for the bone marrow, the error due to the “single CT” protocol is, as expected, very low since the largest contribution is derived from the locally absorbed dose [9] that is calculated from the radioactivity measured in the blood samples [18]. However, the bone marrow is by far a lower risk organ than the kidneys. Indeed, in Reference [9], the bone marrow was the dose-limiting organ in only 1.5% of the patients.

The intra-observer reproducibility for absorbed dose measurements was excellent for the kidneys and bone marrow (all r = 0.99 with P < 0.0001). The mean paired difference was − 0.98 ± 3.40% for the kidneys and 0.84 ± 6.45% for the bone marrow (cf. Fig. 6). Similarly, a good inter-observer agreement for the kidneys was observed (r = 0.96, P < 0.0001) with a mean difference of − 6.50 ± 6.76%. This bias between results calculated from SPECT/CT acquisitions inherently aligned has as sole source differences in VOI size and placement between users.

Finally, since in the “single CT” protocol the registration has to be done manually on a SPECT and a CT is not performed simultaneously, we studied the influence of misregistration on a subset of patients. The error in the calculated absorbed dose coming from misregistration depends on the VOI delineation approach used, i.e., manual CT-based delineation or semi-automated SPECT-based segmentation. We chose five SPECT/CTs of different patients for whom the delineation of the kidneys was easily done using the semi-automated SPECT-based segmentation. Dosimetry results obtained with SPECT/CT “inherently aligned” were used as a reference. We repeated the dosimetry calculations while introducing manually a misregistration of 2, 6, 10, and 20 mm in the axial direction between the SPECT and the CT. The relative errors to the kidney dose were 1.8 ± 3.6% (mean ± SD), 5.9 ± 3.4%, 13.0 ± 4.0%, and 27.8 ± 5.3%, respectively, using manual delineation. Using semi-automated SPECT-based segmentation, relative errors were 0.9 ± 1.3%, 0.4 ± 4.9%, 2.8 ± 6.1%, and 3.6 ± 3.3%, respectively. It can be concluded that for small (< 5 mm) misregistration values, its influence on dosimetry values is not significant. Larger values of misregistration are easily detectable by the observer and would prompt manual correction of the registration. In addition, whenever possible, the semi-automated SPECT-based segmentation should be preferred.

The results presented above show that the kidney and bone marrow dosimetry can be reliably done by registration of all but the first post-treatment SPECTs with the CT of the first post-treatment scan. From an operative standpoint, it means that the vast majority of the post-treatment scans can be done on a non-hybrid γ-camera. This result allows for more flexibility in the management of γ-camera time, appreciable in a busy NM center with several cameras, some of which may not be hybrid. Moreover, it opens the possibility of doing these scans in NM centers devoid of hybrid cameras. Although by now most centers performing PRRT are large regional centers equipped with SPECT/CT device, there is no good reason for it since it does not require any specialized equipment. It would be much more desirable and easier to ship the vial of PRRT treatment to small rural centers, similarly to what is done for conventional chemotherapy, than to put the burden of the commutations on the patient and its accompanying person. In this perspective, our work demonstrates that patient’ access to a hybrid SPECT/CT machine should not be the limiting factor to propose him PRRT treatment. An interesting further study would be to assess the reliability of the dosimetry calculations using a single external CT done several months beforehand the onset of treatment that would alleviate completely the need for a hybrid camera in this clinical setup.

From a radiation exposure perspective, this work allows the reduction up to 5/6 = 83% in the non-therapeutic patient exposure. However, the CT radiation is negligible compared to radiation of the treatment. Typically, the radiation dose absorbed by the kidneys following a [¹⁷⁷Lu]-DOTA-TATE treatment is around 5 Sv [9, 13] versus few mSv [27] for a CT over the abdomen.

Our dosimetry data are consistent with previously published works. Figure 7a, b respectively shows the distribution of the dose to the kidneys and bone marrow for the 83 treatments given to the 24 patients included in this work. The mean kidney absorbed dose after a single PRRT treatment of 7.4 GBq was about 5.4 Gy with an absorbed dose ranging from 2 to 14 Gy. This is in agreement with the mean calculated radiation doses in References [7, 13, 28] and with the kidney doses range observed by Sandström et al. [9]. For the bone marrow, the mean kidney absorbed dose per therapy cycle was around 71 mGy which is well below the 100 mGy value found by Sandström et al. [9]. However, the histogram presented in the latter reference presents the absorbed doses for the first therapy cycle only and not for subsequent cycle showing a decrease in the absorbed dose sometimes by more than 50% [9]. Moreover, as said before, here, the mono-exponential fit for the bone marrow residence time calculation was based on radioactivity in the blood at two close time points only, 18 and 25 h post-injection. This certainly led to an underestimation of cumulated activity and therefore of the absorbed dose to this organ.

Finally, it is noteworthy that the influence of the “single CT” protocol on the tumor dosimetry is higher than on the kidney doses. Indeed, a mean relative error of 7.1% between doses calculated with the two protocols with a 95% confidence interval (CI) of − 48.6 and 62.8% has been obtained for tumors over the 83 cycles of treatment. This indicates that differences in tumor dose calculations are spread out over a wider range of values than the kidneys (CI [− 6.9%, 17.5%]). Such deviations are certainly due to lesions misregistration when aligning the first CT with a subsequent SPECT using the kidneys as a reference. However, the clinical relevance of the tumor dosimetry is limited, since it is not part of the clinical management of the patient.

Absorbed doses to the kidneys and bone marrow were calculated using the “classical” protocol and the new “single CT” protocol over 83 treatments for 24 patients. An excellent agreement between the results calculated from the two protocols was obtained with Pearson’s correlation coefficients r = 0.95 and r = 0.99 (all P < 0.0001) and mean relative differences of 5.30 ± 6.20% and 0.48 ± 4.88% on the cumulative dose for the kidneys and bone marrow, respectively. The management of the patients, i.e., whether or not to stop the treatment because of unsafe absorbed dose to the risk organs using the new protocol, is the same with the two protocols. These results show that dosimetry calculations to critical organs can be done using a single CT registered to serial SPECTs, reducing the need for a hybrid camera.