Background
Coincidence imaging of low-abundance yttrium-90 (
90Y) internal pair production by
90Y positron emission tomography with integrated computed tomography (PET/CT) achieves high-resolution imaging of post-radioembolization microsphere biodistribution. In part 1, we reviewed the recent literature supporting the use of post-radioembolization
90Y PET/CT, described our scan protocol, patient cohort, diagnostic reporting guidelines, and results of qualitative analysis [
1]. In brief, we showed that with proper diagnostic reporting technique and emphasis on continuity of care, the presence of background noise did not pose a problem and
90Y PET/CT consistently out-performed
90Y bremsstrahlung single-photon emission computed tomography with integrated CT (SPECT/CT) in all aspects of qualitative analysis [
1].
The terms ‘predictive dosimetry’, ‘target’, ‘non-target’, ‘technical success’, and the ‘planning-therapy continuum’ were also defined in part 1 [
1]. It is important for the understanding of part 2 to reiterate that our definition of technical success, an adaptation from conventional reporting standards [
2], refers to the
qualitative assessment of a satisfactory
90Y activity biodistribution in accordance with radiation planning expectations and should not be confused with ‘clinical success’ [
2] which is
quantitatively related to dose-response radiobiology.
A knowledge gap exists today between institutions which practice predictive dosimetry and others which rely on semi-empirical methods [
3]. Readers who are accustomed to semi-empirical therapy planning may struggle to understand the dosimetric concepts discussed in this paper or its relevance to clinical practice. These readers are encouraged to refer to our recent series of publications explaining concepts in modern predictive dosimetry for
90Y resin microspheres and common misconceptions [
3‐
6].
Where prognostication is of concern, qualitative analysis alone will not suffice [
3]. The scientific language of dose-response radiobiology is the radiation absorbed dose expressed in grays (Gy), not the prescribed activity expressed in becquerels (Bq). To realize the full potential of post-radioembolization
90Y PET/CT, absorbed dose quantification should be performed in relevant clinical settings. As
90Y radioembolization is a brachytherapy delivered by β
--emitting microspheres, any tissue response is expected to follow predictable dose-response radiobiology. In principle, accurate knowledge of radiation thresholds in both target and non-target tissue facilitates the optimization of predictive dosimetry and enables the prognostication of technically unsuccessful cases to guide adjuvant therapy or mitigative action to minimize non-target radiation toxicity.
However, accurate tissue radiation thresholds for
90Y resin microspheres have remained elusive despite more than two decades of clinical use. Current radiation planning limits are broad and quote mean absorbed doses which falsely assume a uniform dose distribution: tumor > 120 Gy, non-tumorous liver < 50 to 70 Gy, lungs < 20 to 30 Gy [
7‐
9]. As a consequence of this uncertainty, dosimetric dilemmas may be encountered in patients who may benefit from radiation planning up to the limits of normal tissue radiation tolerance. Furthermore, normal tissue radiation thresholds for
90Y resin microspheres shunted to non-target viscera such as the stomach or duodenum are largely unknown, precluding informed decision making for appropriate mitigative action. Historically, research into dose-response data had been technically challenging: intraoperative beta probes or histological examinations are invasive [
10,
11], quantification by
90Y bremsstrahlung scintigraphy is problematic and largely inaccurate [
1], and predictive dosimetry simulated by
99mTc macroaggregated albumin (MAA) is subject to variable accuracy due to the physical limitations of MAA [
12].
Dosimetric studies have shown
99mTc MAA to be feasible for simulating the post-radioembolization biodistribution of
90Y resin microspheres [
13‐
16]. However,
99mTc MAA is an imperfect surrogate for
90Y resin microspheres. Due to biophysical and technical differences such as particle size, specific gravity, injected particle load, microembolization, and catheter placement,
99mTc MAA can never exactly replicate the post-radioembolization biodistribution of
90Y resin microspheres [
12,
17]. Therefore, predictive dosimetry simulated by
99mTc MAA provides only an
estimate of the tissue absorbed doses
intended by the nuclear medicine physician [
6]. Traditionally based on planar scintigraphy [
13,
14], modern predictive dosimetry employs SPECT/CT to tomographically assess the biodistribution of
99mTc MAA and to improve its quantitative accuracy [
6]. So far, the accuracy of
99mTc MAA SPECT/CT predictive dosimetry has only been indirectly validated by inference from follow-up response [
6,
18,
19]. For technically successful cases without visually significant discordant biodistribution between
99mTc MAA and
90Y resin microspheres, a direct ‘Gy-to-Gy’ comparison of intended doses by
99mTc MAA SPECT/CT predictive dosimetry versus post-radioembolization doses by microsphere biodistribution analysis has not yet been performed to date.
Recent experimental and clinical studies have shown post-radioembolization
90Y PET quantification to be accurate and feasible [
1,
20,
21].
90Y PET/CT thus presents a new opportunity to study the radiobiology of
90Y resin microspheres in a rapid, convenient, and noninvasive manner, with the ability to tomographically evaluate the dose distribution of an entire target organ in high resolution by a single scan. Moreover, quantitative
90Y PET data can be translated into dose-volume histograms (DVHs) to dosimetrically account for the heterogeneous nature of microsphere biodistribution.
In part 2 of our two-part retrospective report, we focus on post-radioembolization
90Y PET quantification on the same patient cohort as was reported in part 1 [
1]. We analyze dose-responses in tumor and non-target tissue using
90Y PET-based voxel dosimetry or Medical Internal Radiation Dose (MIRD) macrodosimetry [
7,
22]. We describe the potential of
90Y DVHs to guide predictive dosimetry and discuss how the quantification of non-target absorbed doses may impact post-radioembolization care. Finally, we evaluate the accuracy of tumor
99mTc MAA SPECT/CT predictive dosimetry in direct comparison to post-radioembolization doses by
90Y PET.
Discussion
With its superior image resolution and quantitative capability,
90Y PET/CT is a powerful tool to be integrated into the
90Y radioembolization workflow. Quantitatively, as bland microembolization by
90Y resin microspheres has negligible biological effect [
33], clinical outcomes or toxicities in both target and non-target tissue may be predicted from dose-response radiobiology quantified by
90Y PET, guiding appropriate adjuvant or mitigative action.
We have reported some of the earliest
90Y dose-response results for resin microspheres expressed in the form of DVHs. Our threshold of
D
70 > 100 Gy for HCC complete response is consistent with post-radioembolization DVH case examples of HCC and melanoma liver metastasis shown by Strigari and D’Arienzo et al., respectively [
34,
35].
90Y DVHs are able to account for the heterogeneous nature of microsphere biodistribution and are scientifically superior to MIRD macrodosimetry which falsely assumes a uniform dose distribution. dose-response data from
90Y DVHs have the potential to guide future improvements in predictive dosimetry such as
99mTc MAA SPECT/CT-based DVHs with isodose maps to plan safer and more effective
90Y radioembolization [
36]. For example, our data suggest that HCC tumor outcomes may be optimized by escalating the injected
90Y activity to achieve an intended tumor dose of
D
70 > 100 Gy simulated by
99mTc MAA SPECT/CT DVH, within safety limitations to the lungs and non-tumorous liver.
For non-target activity, we observed that a mean absorbed dose of 18 Gy to the stomach was asymptomatic, while ≥49 Gy to the stomach or duodenum led to CTCAE grade 3 toxicity. Quantification of non-target absorbed doses by
90Y PET effectively translates subjective, qualitative information into objective and radiobiologically meaningful data to guide post-radioembolization care. This may involve close follow-up, extended use of proton pump inhibitors, surveillance endoscopy, or a trial of radioprotecting agents. For the kidney, we observed that a combined bilateral mean dose of 9 Gy (
V
20 8%) did not result in any clinically relevant nephrotoxicity. This finding has clinical implications for the planning of safe kidney-directed
90Y radioembolization for renal malignancies [
37,
38]. We were unable to define the exact normal tissue radiation thresholds based on our limited case series. However, greater clarity on this issue shall be gained in the years ahead as post-radioembolization
90Y PET/CT gains popularity around the world.
In this retrospective report, we have shown tumor 99mTc MAA SPECT/CT predictive dosimetry to be feasible within a small, highly select data set. Prospective validation trials on tumor predictive dosimetry are now needed, using 90Y PET-based absorbed doses as a reference. Our tumor dose-response results are only orientating and should be further validated by greater patient numbers and by using more adequate tumor response end points. It should also be noted that our dose-response results for both target and non-target tissue may not necessarily be valid for 90Y glass microspheres due to different physical characteristics.
We did not perform absorbed dose quantification of the non-tumorous liver or lungs as these are regions of much lower
90Y radioconcentration which may require further research and technical considerations on its quantitative accuracy by
90Y PET [
39]. However, Elschot et al. have recently shown that
90Y PET-based DVHs of the non-tumorous liver may be feasible [
40]. To date, there is no published data on the use of
90Y PET for post-radioembolization lung dose quantification. Nevertheless, we recognize the importance of tissue radiobiology in non-tumorous liver and lungs for the safe and effective clinical application of predictive dosimetry in
90Y radioembolization. This should be explored in future
90Y PET-based dosimetry research.
The specific aim of our tumor analysis was to evaluate the best possible performance of tumor predictive dosimetry planned by 99mTc MAA SPECT/CT. We are therefore limited by our study design to restrict our discussion only to the tumor aspect of predictive dosimetry; our study cannot draw conclusions on predictive dosimetry applied to non-tumorous liver or lungs. To this end, we have shown in our highly select tumor data set that the intended tumor mean doses by 99mTc MAA SPECT/CT predictive dosimetry correlated well with post-radioembolization doses by 90Y PET. From a clinical perspective, this means that if tumor predictive dosimetry had been applied to a technically successful 90Y radioembolization under near-ideal dosimetric conditions, then the intended tumor mean doses may be assumed valid and retrospective tumor dose quantification is usually unnecessary, unless for research or quality assurance.
In less ideal dosimetric conditions, tumor predictive dosimetry may be of variable accuracy, and retrospective tumor dose quantification by 90Y PET may be an option at the discretion of the nuclear medicine physician, e.g., MIRD-based predictive dosimetry applied to massive tumors with large heterogeneous areas of relative hypovascularity, where its assumption of a uniform activity biodistribution risks dosimetric uncertainty; or MIRD-based tumor predictive dosimetry applied to intended radiomicrosphere lobectomy or segmentectomy, where potential vascular stasis and reflux introduces dosimetric uncertainty. However, in technically unsuccessful90Y radioembolization, we feel that retrospective tumor dose quantification by 90Y PET may be routinely indicated because the intended tumor doses by predictive dosimetry may have become invalid. Furthermore, non-target activity may also be present. In such situations, the prognosis remains uncertain unless retrospective dose quantification is performed within appropriate regions of interest.
90Y PET/CT is but one component near the end of the entire planning-therapy continuum.
90Y PET/CT is a descriptive modality to assess post-implantation microsphere biodistribution, where all inadvertent or adverse findings have already irreversibly occurred. To the nuclear medicine physician, the component of the planning-therapy continuum with the greatest prognostic impact, but also more complex, is predictive dosimetry. In principle,
90Y radioembolization planned by predictive dosimetry is not subject to prognostic uncertainties experienced by semi-empirical
90Y activity prescription, a common practice within the medical oncology paradigm. In the modern era of personalized medicine, predictive dosimetry planned by modern tomographical techniques give nuclear medicine physicians unprecedented insight into the achievable outcomes for every
90Y radioembolization [
6]. Technical success does not necessarily imply clinical success: the former is a qualitative assessment of microsphere biodistribution; the latter is a quantitative function of dose-response radiobiology. While interventional radiologists do their utmost to achieve technical success, it is the responsibility of the nuclear medicine physician to ensure that their efforts can be translated into clinical success by the rigorous application of predictive dosimetry.
Acknowledgements
We thank the following individuals for their contributions: Dr Glenn P. Sy, M.D., Clinical Application Specialist, GE Healthcare ASEAN, for assisting with the 90Y PET/CT optimization; Miss Toh Ying, Research Coordinator, Department of Nuclear Medicine and PET, Singapore General Hospital, for the administrative support; Dr Terence Teo Kiat Beng, Dr Yeow Tow Non, and Dr Apoorva Gogna, Interventional Radiologists, Department of Diagnostic Radiology, Singapore General Hospital, for performing the 90Y radioembolization of the patients in this report; Miss Doreen Lau Ai Hui and Dr Kei Pin Lin, Department of Diagnostic Radiology, Singapore General Hospital, for providing the CT images; and Dr Choo Su Pin, Medical Oncologist, Department of Medical Oncology, National Cancer Centre Singapore, for patient referrals and reviewing the manuscript.
Competing interests
YHK, ASWG, KHT, and PKHC received research funding from Sirtex Medical Singapore. ASWG and PKHC received honoraria from Sirtex Medical Singapore. The other authors declare that they have no competing interests.
Authors’ contributions
YHK, YST, GKYL, SS, AEHT, DCEN, and ASWG were involved in the study design, implementation, analysis, and manuscript preparation. JDS, JY, and DWT were involved in the study design, scan optimization, and manuscript preparation. CB, JAB, RJF, and TSTC were involved in the data analysis and manuscript preparation. PKHC was involved in the clinical care and manuscript preparation. MCB, FGI, RHGL, KHT, and BST were involved in the radioembolization, angiographic analysis, and manuscript preparation. All authors read and approved the final manuscript.