Background
Peptide receptor radionuclide therapy (PRRT) with [
177Lu]Lu-DOTA-TATE for treatment of somatostatin receptor (SSTR) expressing neuroendocrine tumours (NETs) [
1,
2] is typically preceded by SSTR-PET imaging using [
68Ga]Ga-DOTA-TATE or -TOC to ensure adequate receptor expression [
3,
4]. The use of a theragnostic approach with the same or similar peptides for imaging and therapy offers opportunities for therapy stratification, but there is today no consensus on the predictive value of
68Ga-SSTR-PET/CT imaging with respect to response, absorbed doses, or activity uptakes in tumours and normal organs for therapy. A number of studies have investigated the relationship between SSTR expression quantified from
68Ga-SSTR-PET/CT and the outcome of [
177Lu]Lu-DOTA-TATE or -TOC therapy of NETs [
5‐
10]. When examining such relationships, it is often implicitly assumed that a high tumour uptake in pre-therapeutic
68Ga-SSTR-PET/CT images also infers high tumour uptake and absorbed dose during
177Lu therapy.
To the best of our knowledge, there is to date only one study that made a direct, quantitative comparison of results from [
68Ga]Ga-DOTA-TOC PET and absorbed doses delivered during [
177Lu]Lu-DOTA-TOC therapy for NET patients [
11]. In that study, tumour dosimetry was performed for 21 patients based on serial planar
177Lu imaging, and a statistically significant correlation (
r = 0.7) was found between the
68Ga-SUV (SUV
mean or SUV
max) and the
177Lu absorbed dose [
11]. Furthermore, a few reports on similar radiopharmaceuticals or indications are available. Krebs et al. [
12] reported on the treatment of 20 NET patients using a SSTR antagonist (
177Lu-satoreotide tetraxetan) with pre-therapeutic
68Ga-imaging and
177Lu dosimetry based on a hybrid SPECT–planar method. Various quantitative parameters were analysed, including tumour-to-normal-tissue SUV ratios, and the highest correlation (
r = 0.5) was found between
68Ga-SUV
peak and the
177Lu absorbed dose to lesions [
12]. Hänscheid et al. [
13] reported data from 11 patients treated for meningioma, where
177Lu dosimetry was performed with a hybrid SPECT–planar method. They found that the
68Ga-SUV
max correlated well with the
177Lu activity concentration 1 h after administration (
r = 0.95), whilst the correlation to
177Lu absorbed dose was moderate (
r = 0.76). For [
177Lu]Lu-PSMA, pre-therapeutic
68Ga-PET/CT and
177Lu absorbed doses have also been compared, e.g. by Peters et al. [
14].
Investigation of possible relationships between uptakes of 68Ga-SSTR-PET and absorbed doses in 177Lu PRRT can be made from different perspectives. In the above-mentioned studies, the relationship was approached on a population level, reflecting the overall relationship across patients. For metastatic disease, analyses can also be made across the tumours within individual patients, addressing the distribution of uptakes and absorbed doses, i.e. whether a higher uptake of [68Ga]Ga-DOTA-TATE in one tumour than another generally means that the absorbed dose is higher for that tumour in [177Lu]Lu-DOTA-TATE therapy. Thirdly, the question can be posed as an estimation problem, to understand whether and how well absorbed doses in 177Lu PRRT can be predicted from a pre-therapeutic 68Ga-SSTR-PET. This perspective is relevant with regards to personalized dose planning, where both tumours and normal organs need to be considered. The various perspectives need to be considered separately, as they require different methods for evaluation.
Studies that compared the activity uptakes in
68Ga-SSTR-PET/CT with the uptakes and absorbed doses in
177Lu-PRRT have mainly used different variants of SUV for evaluation of the
68Ga images. Besides SUV, different tumour-to-tissue ratios have been proposed, where reference tissues include the liver parenchyma, spleen, or blood [
5,
15,
16]. Using SUV ratios is partly methodologically motivated, as this may partly mitigate the SUV dependence on factors such as reconstruction settings, the PET/CT system, and the accumulation time [
17]. Another motivation is the pharmacokinetics, as demonstrated for 10 patients examined by dynamic [
68Ga]Ga-DOTA-TATE and -TOC PET/CT, leading to the suggestion of using the tumour-to-blood SUV ratio [
16,
18]. However, a simpler, and more fundamental parameter than SUV is the activity concentration. Although SUV is well established as a metric in diagnostics and patient selection from
68Ga-SSTR-PET/CT [
3,
4], the reasons for using SUV are less evident when attempting to find a relationship to the therapeutic absorbed dose from
177Lu. Specifically, the inclusion of the patient's weight can be questioned (SUV = activity concentration × weight/injected activity), as the weight does not enter the calculation of the absorbed dose to tumours and organs.
The aim of this study was to investigate whether and how parameters derived from [68Ga]Ga-DOTA-TATE PET/CT relate to the uptake and absorbed doses delivered during [177Lu]Lu-DOTA-TATE therapy in NET patients. As basic property for [68Ga]Ga-DOTA-TATE quantification the activity concentration per administered activity is calculated, which is then complemented by different SUV-based metrices. For [177Lu]Lu-DOTA-TATE both the activity concentration and the absorbed dose per administered activity are considered. Furthermore, the possibility to predict 177Lu absorbed doses for tumours based on quantitative [68Ga]Ga-DOTA-TATE PET/CT images combined with population mean effective half-lives for [177Lu]Lu-DOTA-TATE, separated on grade-1 and grade-2 NETs, is studied. This study thus aims to complement and expand on earlier studies, using modern quantification methods, and analysing data for both organs and tumours, considering correlations as well as the ability of absorbed dose prediction.
Discussion
In this study, we have investigated the relationship between uptakes of [68Ga]Ga-DOTA-TATE quantified in PET images, and uptakes and absorbed doses to tumours and organs during subsequent treatment with [177Lu]Lu-DOTA-TATE for NETs. In summary, for tumours we see a significant (p < 0.05), moderately strong (r = 0.71), relationship across patients between the activity concentration from 68Ga-PET images and the absorbed dose from 177Lu-PRRT. A stronger relationship is seen with respect to the 177Lu activity concentration from SPECT images 24 h after injection. On an individual level, the ability to predict the 177Lu absorbed dose to tumours based solely on a 68Ga-PET image is limited, with a 95% coverage interval of − 65% to 248%.
The use of
68Ga-SSTR-PET for correlation with outcome and prognosis of NETs has been investigated both in general and with respect to
177Lu-PRRT [
5‐
10]. However, the connection between
68Ga-SSTR-PET uptakes and absorbed doses during therapy is less studied [
11]. Even if absorbed dose is not a direct measure of treatment outcome and toxicity, it is an established parameter in other forms of radiotherapy and is being gradually better established also for radionuclide therapy [
35,
42,
43]. Hence, we believe that an increased understanding of relationships between
68Ga-SSTR imaging and absorbed doses in
177Lu-PRRT fills an important gap.
A fundamental difficulty for quantitative interpretation of pre-therapeutic
68Ga-SSTR-PET with respect to the absorbed doses delivered during
177Lu-PRRT lies in the different half-lives of
68Ga and
177Lu [
44] (6.6 d versus 68 min [
28,
29]).
68Ga-SSTR-PET is typically performed 1 h p.i. [
4] while therapy with [
177Lu]Lu-DOTA-TATE extends over several days or weeks [
39]. So although the ligand is identical, the time scales of the processes exploited with
68Ga imaging and
177Lu therapy are markedly different, limiting the accuracy for prediction of the time-integrated activity and absorbed dose [
45]. There are also other factors that differ between the
68Ga-SSTR-PET and
177Lu-PRRT, such as the method of administration (bolus versus extended infusion), and the fraction of the peptides that are radiolabelled which differs by nearly three orders of magnitude. At the same time,
68Ga-SSTR-PET imaging is today clinically used as part of the patient-selection process for
177Lu-PRRT, and hence, to some extent, a correlation is implicitly assumed.
For tumours, the strengths of the obtained correlations between uptakes of [
68Ga]Ga-DOTA-TATE and absorbed doses in
177Lu-PRRT are on par with those reported previously for NET and meningioma [
11,
13], and higher than those reported for satoreotide tetraxetan [
12]. Comparison between uptakes in [
68Ga]Ga-PSMA-11-PET and absorbed dose in therapy with
177Lu-PSMA-617 have also shown similar correlations [
14]. Importantly however, from such correlations on a group level, it cannot be directly inferred that the therapeutic absorbed doses can be predicted for the individual patient. Based on the presented approach for prediction, using the
68Ga-PET activity concentration combined with population-based effective half-lives for [
177Lu]Lu-DOTA-TATE for NETs, only rough estimates of the absorbed doses in the upcoming therapy are obtained (Fig.
5). Personalized treatment planning based on
68Ga-PET imaging will thus require more elaborate approaches, such as the inclusion of pharmacokinetic modelling [
46].
The poor agreement between absorbed dose estimates (Fig.
5) can partly be theoretically explained by the combination of a protracted therapeutic delivery and a measurement at 1 h p.i. [
45]. As such, considerable dispersion is expected. However, in principle, the accuracy of a measurement method needs to be considered in relation to the requirements for the application, and the results in Fig.
5 could then still be informative in cases when only a rough estimate is necessary. Apart from mathematical and biological considerations, different absorbed dose calculation methods are also used for the PET-based estimation compared to the peri-therapeutic dosimetry. However, the benefit of full Monte Carlo simulations compared with using local energy-deposition is typically small for
177Lu [
39,
47] and is not expected to be the major reason for the disagreement between the estimated values.
Among the organs, only spleen exhibits a significant correlation between the uptake of [
68Ga]Ga-DOTA-TATE and the absorbed dose in
177Lu-PRRT (Fig.
2). For kidneys, considered the primary organ-at-risk for
177Lu-PRRT, we see no significant relationship, one possible reason being the co-administration of renal protective amino acids for [
177Lu]Lu-DOTA-TATE. For liver parenchyma, the estimation of the activity concentration suffers from practical challenges for VOI definition. Although small VOIs have been applied there is a risk that tumour may have been included, both due to spillover from adjacent tumours in the images and due to microscopic disease. Whether or not a patient is on treatment with long-acting SSA has, in previous publications, been observed to affect the liver uptake of [
68Ga]Ga-DOTA-TATE and only to a lesser degree the tumour uptake [
20]. According to the same authors variable time intervals from the last SSA injection did not affect uptake. It is therefore unlikely that this factor contributed to the dispersion in data for the liver and the tumour-to-liver ratio.
The stronger correlations obtained between the
68Ga and
177Lu activity concentrations, compared to the
177Lu absorbed dose (Table
1) were expected. Absorbed dose depends on a combination of initial activity uptake and excretion, while the activity concentration measured in
68Ga-PET at 1 h almost exclusively reflects the initial activity uptake. The uptake measured in
177Lu-SPECT at 24 h is less affected by the excretion than the absorbed dose is, which reduces the variability relative to the activity concentration at 1 h, measured in
68Ga-PET.
Of interest, our results provide no support for using different types of normalization of the 68Ga activity concentration to improve the relationship to absorbed dose in 177Lu-PRRT, neither with respect to normalization to body weight, i.e. calculation of SUV, nor with respect to a reference tissue. In this study, SUVs were calculated according to clinical practice, with no PVC applied, which may in part affect the correlations obtained. However, in relation to the 177Lu absorbed dose, there is no theoretical reason to normalize the activity concentration to body weight. Even if the body size, as an indirect measure of the plasma volume, may affect the activity uptake, this will act the same for diagnostics and therapy. Normalization to a reference tissue can in principle be motivated to cancel differences between receptor-bound activity and activity in blood in different patients. However, in our data such normalizations only increase the dispersion. The practical difficulties associated with the estimation of activity concentration or SUV in blood or liver parenchyma from 68Ga-PET images need to be emphasized. In a static 68Ga-PET image, blood SUV is associated with large uncertainties as it requires the measurement of low activity concentrations, which, in addition to the associated statistical variation, puts great demands on the accuracy of compensations for scattered and random coincidences. Thus, we believe that from both a theoretical and a practical point of view, it is preferable to study the AC/IA directly rather than normalized variants thereof.
For the correlation analyses for kidneys, the sensitivity to individual data points, as revealed by the leave-one-out analysis, should be noted (Fig.
2 and Table
1). The correlations obtained are largely governed by one or two data points rather than reflecting a general trend, and the significant correlations should hence be interpreted cautiously. Similar instability was not found for tumours.
The analysis of tumour data is more complex than for organs because of the varying number of tumours per patient, for which independence cannot be assumed. For this reason, the problem of finding a relationship between the uptakes in
68Ga-PET and the therapy is separated into two questions: 1) whether there is a relationship between patients when regarding the mean values for the tumours within each patient and 2) whether there is a relationship for the separate tumours within patients, following the methodology presented by Bland and Altman [
40,
41]. Regarding the inter-patient analysis, a moderate correlation is obtained for the
177Lu-AD/IA as a function of
68Ga-AC/IA, while a stronger correlation is obtained for the
177Lu-AC/IA. This indicates that there is a group-level relationship between the uptake in
68Ga-PET and the
177Lu absorbed dose. The intra-patient analysis shows similar results, where the relationship is weaker for
177Lu-AD/IA than for
177Lu-AC/IA. This indicates that there is a correlation also within individual patients, i.e. on average a high
68Ga uptake for a separate tumour also corresponds to a high absorbed dose in subsequent
177Lu-PRRT. The two analyses are complementary, and it is concluded that there are statistically significant, but moderately strong, correlations both intra- and inter-patient.
Two important limitations of this study are the low number of included patients and the relatively permissive inclusion criterion of a [
68Ga]Ga-DOTA-TATE PET performed up to 20 weeks prior to PRRT. The patient population was, however, one of well-differentiated NET with a low Ki67-index, i.e. the likelihood of significant change in tumour volume over the given time interval is small. Furthermore, the actual median time from PET imaging to PRRT was 5 weeks, further reducing such a potential confounder. The dosimetry methods used in this study have been extensively validated in previous papers [
27,
30,
31]. In principle, however, image-based dosimetry based on SPECT-only imaging would be preferable to the hybrid method used. Furthermore, the employed cutoff volume of 5 mL is of concern, partly because it reduces the number of included tumours, but also because it systematically excludes tumours with a certain characteristic which could theoretically lead to biased results. However, the uncertainties associated with estimated volumes (Fig.
1) and activity concentrations of structures with dimensions close to the system spatial resolution are well known [
31]. Thus, excluding the smallest tumours was considered necessary not to contaminate the results.
In summary, we find that there is a statistical relationship between tumour uptake at [68Ga]Ga-DOTA-TATE PET and absorbed dose to tumours in subsequent 177Lu-PRRT, but that this association is moderate at best. Given that previous studies have shown correlations of approximately the same strength, methodological differences notwithstanding, we believe that these moderate correlations reflect the actual strength of the relationship, rather than being a result of measurement uncertainties. Furthermore, we find no, or unstable, relationships for organs, except for spleen. Thus, at the group level there are relevant relationships between the uptake in [68Ga]Ga-DOTA-TATE PET and the upcoming 177Lu-PRRT. However, to be able to practically use [68Ga]Ga-DOTA-TATE PET for absorbed dose planning at the individual level, more complex models are needed that take patient-specific factors into account, beyond simple univariate analyses.