¹⁷⁷Lu-DOTATATE Theranostics

Abstract

Purpose:

Pretreatment predictions of absorbed doses can be especially valuable for patient selection and dosimetry-guided individualization of radiopharmaceutical therapy. Our goal was to build regression models using pretherapy ⁶⁸Ga-DOTATATE PET uptake data and other baseline clinical factors/biomarkers to predict renal absorbed dose delivered by ¹⁷⁷Lu-DOTATATE peptide receptor radionuclide therapy (¹⁷⁷Lu-PRRT) for neuroendocrine tumors. We explore the combination of biomarkers and ⁶⁸Ga PET uptake metrics, hypothesizing that they will improve predictive power over univariable regression.

Patients and Methods:

Pretherapy ⁶⁸Ga-DOTATATE PET/CTs were analyzed for 25 patients (50 kidneys) who also underwent quantitative ¹⁷⁷Lu SPECT/CT imaging at approximately 4, 24, 96, and 168 hours after cycle 1 of ¹⁷⁷Lu-PRRT. Kidneys were contoured on the CT of the PET/CT and SPECT/CT using validated deep learning–based tools. Dosimetry was performed by coupling the multi–time point SPECT/CT images with an in-house Monte Carlo code. Pretherapy renal PET SUV metrics, activity concentration per injected activity (Bq/mL/MBq), and other baseline clinical factors/biomarkers were investigated as predictors of the ¹⁷⁷Lu SPECT/CT-derived mean absorbed dose per injected activity to the kidneys using univariable and bivariable models. Leave-one-out cross-validation (LOOCV) was used to estimate model performance using root mean squared error and absolute percent error in predicted renal absorbed dose including mean absolute percent error (MAPE) and associated standard deviation (SD).

Results:

The median therapy-delivered renal dose was 0.5 Gy/GBq (range, 0.2–1.0 Gy/GBq). In LOOCV of univariable models, PET uptake (Bq/mL/MBq) performs best with MAPE of 18.0% (SD = 13.3%), and estimated glomerular filtration rate (eGFR) gives an MAPE of 28.5% (SD = 19.2%). Bivariable regression with both PET uptake and eGFR gives LOOCV MAPE of 17.3% (SD = 11.8%), indicating minimal improvement over univariable models.

Conclusions:

Pretherapy ⁶⁸Ga-DOTATATE PET renal uptake can be used to predict post-¹⁷⁷Lu-PRRT SPECT-derived mean absorbed dose to the kidneys with accuracy within 18%, on average. Compared with PET uptake alone, including eGFR in the same model to account for patient-specific kinetics did not improve predictive power. Following further validation of these preliminary findings in an independent cohort, predictions using renal PET uptake can be used in the clinic for patient selection and individualization of treatment before initiating the first cycle of PRRT.

Pretherapy imaging for treatment planning has the potential to be an effective tool for dosimetry-guided radiopharmaceutical therapy (RPT). Dosimetry-guided RPT can be used to tailor ¹⁷⁷Lu-DOTATATE peptide receptor radionuclide therapy (PRRT) on a patient-specific basis as demonstrated by clinical trials, which aim to modulate the number of treatment cycles¹ or the injected activity per cycle² until the patient reaches a specified cumulative renal absorbed dose or biologically effective dose limit. Furthermore, Del Prete et al² use a pretherapy prediction of renal absorbed dose to target a first cycle dose to the kidneys of 5 Gy. Improvements to the accuracy of these pretherapy predictions would allow for further escalations of first cycle injected activity, which could lead to larger absorbed doses to tumoral tissue while still considering safety. Such an approach is particularly relevant in light of a recent study reporting a reduction in tumor-absorbed doses as the cycles proceed.³

In the theranostics approach to RPT, diagnostic imaging agents such as ¹²⁴I or ⁶⁸Ga are coupled with therapeutic agents such as ¹³¹I or ¹⁷⁷Lu, and the resulting theranostic “pairs” (¹²⁴I/¹³¹I or ⁶⁸Ga/¹⁷⁷Lu) are used to assess patient status and predict pharmaco-kinetics to inform RPT.^{4 68}Ga-DOTATATE PET uptake is routinely used to determine eligibility of patients for ¹⁷⁷Lu-DOTATATE PRRT, and the activity distribution of the 2 radiotracers has been compared previously.^5,6 Although the feasibility of using ¹²⁴I PET as a surrogate for performing dosimetry in ¹³¹I therapies has been well-demonstrated,⁷ the use of short-lived ⁶⁸Ga for predicting delivered absorbed doses from therapies such as ¹⁷⁷Lu-DOTATATE has not been well-established. However, in relatively small cohorts, the correlation between pretherapy PET uptake with ⁶⁸Ga-labeled somatostatin analogs and absorbed doses delivered by PRRT has been demonstrated for tumors.^8,9 These studies do not investigate normal tissues and include suboptimal posttherapy imaging techniques such as serial planar imaging⁹ and serial planar imaging with a single SPECT/CT⁸ to perform absorbed dose estimates. In addition, correlations have been established between glomerular filtration rate and renal absorbed dose after ¹⁷⁷Lu-PRRT.^10,11

Recently, ⁶⁸Ga-PSMA-11 PET/CT uptake data and average population kinetics were used to predict lesion and organ absorbed doses determined by posttherapy imaging after ¹⁷⁷Lu-PSMA-617 radioligand therapy for treatment of metastatic castration-resistant prostate cancer.¹² This study found moderate correlation between the pretherapy dose prediction and the therapy-delivered dose for tumors but reports stronger relationships for normal tissues. Correlations between pretherapy ⁶⁸Ga-PSMA-11 PET uptake and post-¹⁷⁷Lu-PSMA-617 therapy absorbed dose in tumors and normal tissues have also been reported elsewhere.¹³

Considering the paucity of studies investigating pretherapy predictions of renal absorbed dose in ¹⁷⁷Lu-DOTATATE PRRT, our study aimed to develop such models using not only a single pretherapy ⁶⁸Ga-DOTATATE PET/CT scan but also widely available baseline biomarkers and clinical factors (Fig. 1). We hypothesized that the inclusion of biomarkers into our models may be able to account for patient-specific kinetics that impact absorbed dose but are challenging to extract from imaging a short-lived surrogate, thus improving predictive power over ⁶⁸Ga PET uptake alone. We describe the construction of regression models using data from 25 patients with testing by leave-one-out cross-validation (LOOCV).

FIGURE 1.

Graphical overview depicting pretherapy predictive dosimetry performed with ⁶⁸Ga PET uptake and biomarkers (top) and therapy-delivered dosimetry performed using serial ¹⁷⁷Lu SPECT/CT imaging (bottom).

PATIENTS AND METHODS

Patients

Patients with gastroenteropancreatic neuroendocrine tumors who received at least 1 cycle of PRRT with standard (7.4 GBq) ¹⁷⁷Lu-labeled DOTATATE (LUTATHERA)¹⁴ at the University of Michigan Hospital and provided written informed consent for posttherapy SPECT/CT imaging as part of an ongoing institutional review board–approved research protocol were eligible for this study. We limited patients to those who had ⁶⁸Ga-DOTATATE PET/CT performed within 1 year of treatment and further excluded 1 patient who had technical problems with their PET/CT acquisition. Patient characteristics are included in Supplemental Table S1 (Supplemental Digital Content, http://links.lww.com/CNM/A417).

Kidney Delineation

Left and right kidneys were segmented using deep learning–based autosegmentation¹⁵ on the CT of the PET/CT and the reference SPECT/CT. Kidney segmentations were verified by a radiologist and adjusted as needed to correct for anatomical anomalies such as cysts and to match the uptake intensity on the functional image in cases of misregistration between the PET/SPECT and CT.

Pretherapy ⁶⁸Ga-DOTATATE PET/CT Imaging-Based Factors

Imaging and Reconstruction

According to standard practice, long-acting somatostatin analog therapy is held 3 to 4 weeks before imaging with ⁶⁸Ga-labeled DOTATATE (NETSPOT).^{16 68}Ga-DOTATATE PET/CTs were performed on a Biograph mCT (Siemens Healthineers) (22/25), Biograph TruePoint (Siemens Healthineers) (1/25), Discovery MI (GE Healthcare) (1/25), and Discovery STE (GE Healthcare) (1/25) at approximately 60 minutes postinjection of ⁶⁸Ga-DOTATATE (median, 158.7 MBq; range, 144.3–196.5 MBq). Data were reconstructed using vendor-supplied software with recommended reconstruction parameters to obtain quantitative images in units of Bq/mL. No partial volume correction was applied.

Uptake Metrics

The SUV_mean, SUV_peak, and mean activity concentration for the left and right kidney of each patient were determined separately using an automated workflow. SUV_peak, which is defined as the average SUV of the hottest 1 cm³ spherical subregion per PERCIST 1.0,¹⁷ was calculated for each kidney. Renal ⁶⁸Ga PET uptake (Bq/mL/MBq) is defined as the mean activity concentration in each kidney at PET acquisition normalized by the administered activity at time of injection.

Tumor Burden

To investigate the possible influence that activity sequestration in tumoral tissues has on renal dose (tumor sink effect¹⁸), tumor burden was determined on ⁶⁸Ga-DOTATATE PET using a semiautomated workflow with manual adjustment (Supplemental Figure S1, Supplemental Digital Content, http://links.lww.com/CNM/A417). Final tumor burden volumes were approved by a nuclear medicine physician. Total lesion volume (TLV), defined as the volume of the entire tumor burden volume of interest, in mL, and total lesion somatostatin receptor expression (SSTRe), also in mL, calculated as SUV_mean,tot × TLV, were included for regression analysis. Here, SUV_mean,tot is the average SUV of the entire tumor burden volume.

Surrogate Dosimetry

Although the primary focus of the study was to construct regression models for absorbed dose prediction, we also investigated using ⁶⁸Ga-DOTATATE PET uptake for surrogate dosimetry of ¹⁷⁷Lu-DOTATATE, which relies on the assumption that the biological half-life of the 2 agents is equal. Under this assumption, only a correction for differences in the physical half-life of the agents is needed to predict the therapeutic agent uptake from the surrogate agent uptake. Given that $A_{{}^{68}Ga,i}(t_{meas})$ is the ⁶⁸Ga activity concentration in voxel, $i$, at time after administration, $t_{meas}$, normalized by the activity administered, then the corresponding ¹⁷⁷Lu activity concentration normalized by ¹⁷⁷Lu activity administered can be expressed as:

$$\begin{gathered} A_{{}^{{}_{{}^{177}}}Lu,i}(t_{meas})=A_{{}^{{}_{{}^{68}}}Ga,i}(t_{meas})\times \text{exp}\left({\lambda }_{{}^{{}_{{}^{68}}}Ga}\times t_{meas}\right) \\ \times \text{exp}\left(-{\lambda }_{{}^{{}_{{}^{177}}}Lu}\times t_{meas}\right) \end{gathered}$$ (1)

where ${\lambda }_{{}^{68}Ga}$ and ${\lambda }_{{}^{177}Lu}$ are the physical decay constants for ⁶⁸Ga and ¹⁷⁷Lu, respectively. Because of the short half-life of ⁶⁸Ga, multi–time point ⁶⁸Ga PET is not feasible for determination of ¹⁷⁷Lu-DOTATATE kinetics. Therefore, the ¹⁷⁷Lu-DOTATATE time activity curve was assumed to be a monoexponential function with population average effective half-life for kidneys. Then, the time-integrated activity coefficient (TIAC) of each voxel can be calculated by:

$$\begin{gathered} TIA{C}_{{}_i}=A_{{}^{{}_{{}^{177}}}Lu,i}(t_{meas})=\text{exp}\left(\widehat{\lambda }\times t_{meas}\right)\underset{0}{\overset{\infty }{\int }}\text{exp}\left(-\widehat{\lambda }\times t\right)dt \\ =\frac{A_{{}^{{}_{{}^{177}}}Lu,i}(t_{meas})}{\widehat{\lambda }}\times \text{exp}\left(\widehat{\lambda }\times t_{meas}\right) \end{gathered}$$ (2)

where $A_{{}^{177}Lu,i}(t_{meas})$ is given by Equation 1, and $\widehat{\lambda }$ is the population mean of the effective decay constant. We adopt a $\widehat{\lambda }$ corresponding to an effective half-life of 51.6 hours from Sundlöv et al,¹ which was derived from a similar population of patients. Each voxel in the PET image is converted, using Equations 1 and 2, to produce a surrogate voxelized TIAC map. This TIAC map and a CT-derived density map are given as inputs to an in-house Monte Carlo code to generate the surrogate absorbed dose map and the corresponding predicted renal mean absorbed doses.

Baseline Biological Markers and Clinical Factors for Regression Analysis

In addition to the previously described 5 PET-based metrics, we sought readily available clinical factors and biomarkers, focusing on those expected to be related to renal function or tumor burden, both of which have been reported to impact absorbed dose to kidney.^10,18 Values of 7 such parameters obtained from the notes of the treating nuclear medicine physician at the time of consult for PRRT are presented in Table 1. Creatinine is an established measure of kidney function,¹⁹ and estimated glomerular filtration rate (eGFR) has been previously demonstrated to have a relationship with renal absorbed dose.^10,11 We included alkaline phosphatase, a reported prognostic marker of progression-free survival in patients with NETs,²⁰ to explore potential effects of tumor burden. Association of anemia, pancytopenia, bone marrow hypocellularity, and dysfunction of bone marrow–derived stromal cells with chronic kidney disease motivate the remaining factors.^21-23 We chose not to include factors such as age, sex, body size, hypertension, or diabetes as these should be captured by the creatinine level or are included in the eGFR calculation. Our relatively small sample size was a consideration for limiting the number of factors from all available.

TABLE 1.

Baseline Factors Consisting of 7 Biomarkers and 5 ⁶⁸Ga PET/CT-Derived Metrics Considered for Univariable Prediction of the Therapy-Delivered Renal Absorbed Dose Listed in the Last Row

Metric	Values
White blood cells (× 103/μL)	6.9 (2.6–13.6)
Absolute neutrophils (× 103/μL)	4.6 (1.4–7.5)
Hemoglobin (g/dL)	13.4 (9.5–16.0)
Platelets (× 103/μL)	219 (138–466)
eGFR (mL/min)	81.1 (25.4–102.7)
Creatinine (mg/dL)	0.9 (0.5–2.0)
Alkaline phosphatase (IU/L)	118 (48–234)
Renal PET uptake (Bq/mL/MBq)	64.0 (30.4–148.9)
Renal PET SUVmean	10.5 (5.0–19.4)
Renal PET SUVpeak	16.0 (8.6–60.1)
TLV from PET (mL)	286.9 (33.2–2778.1)
SSTRe from PET (mL)	2998.3 (116.6–30,884.3)
Delivered absorbed dose estimated from 177Lu SPECT/CT* (Gy/GBq)	0.5 (0.2–1.0)

Values are reported as median (range) and include all n = 50 kidneys.

^*Considered to be the reference absorbed dose.

Posttherapy ¹⁷⁷Lu-DOTATATE SPECT/CT Imaging and Dosimetry

¹⁷⁷Lu SPECT/CT imaging was performed on a Siemens Intevo at approximately 4, 24, 96, and 168 hours after cycle 1 of ¹⁷⁷Lu-DOTATATE administration with 25 minute acquisitions using manufacturer recommended settings. Data were reconstructed with Siemens xSPECT Quant to obtain images in units of Bq/mL. Activity quantification accuracy in xSPECT has been previously reported, ²⁴ and we used the reconstruction parameters from Dewaraja et al.¹⁵ Partial volume correction was not applied to the kidney activity quantification.

As detailed in Dewaraja et al,¹⁵ a contour intensity–based SPECT-SPECT alignment was used to coregister the time points, and these images were used as inputs to an in-house Monte Carlo code along with a CT-derived density map to generate dose rate maps. Dose rate maps were fit with monoexponential or biexponential functions based on Akaike Information Criteria as proposed by Sarrut et al,²⁵ and the resulting fits were integrated to generate an absorbed dose map.¹⁵ Subsequently, the mean dose within the left and right kidney was determined (Fig. 1, bottom row).

Regression Analysis for Absorbed Dose Prediction

Univariable regression models were generated to predict renal absorbed dose using baseline imaging and nonimaging data. Univariable regression with ordinary least squares (OLS) minimization was performed using Python 3.9.6 (Statsmodels 0.12.2). Goodness-of-fit is reported as the coefficient of determination (R²), and significance for univariable regression is denoted by P < 0.05. Because we hypothesized that eGFR would enhance PET-based predictions, we also built a bivariable regression model with renal PET uptake and baseline eGFR using OLS in Python 3.9.6. Leave-one-out cross-validation was used to estimate model performance using R², root mean squared error (RMSE), absolute percent error (APE), and mean absolute percent error (MAPE) across all patients as follows:

$$\begin{gathered} RMSE=\sqrt{\frac{1}{n}\sum _{i=1}^n(A{D}_{pred,i}-A{D}_{meas,i}{)}^2} \\ AP{E}_i=\mid \frac{A{D}_{pred,i}-A{D}_{meas,i}}{A{D}_{meas,i}}\mid \times 100\% \\ MAPE=\frac{1}{n}\sum _{i=1}^{n}AP{E}_i \end{gathered}$$

Patients’ left and right kidneys are included in the same model in our analysis. This decision was justified by tests using Akaike Information Criteria that determined the location (left/right) of the kidney did not significantly influence absorbed dose predictions for any biomarkers.

RESULTS

Baseline Data and Absorbed Dose Values

A total of 25 patients with a combined 50 kidneys met the study criteria. The values of data used for regression including 7 baseline factors obtained from medical records and 5 factors obtained from ⁶⁸Ga PET/CT imaging are summarized in Table 1. The therapy-delivered absorbed dose estimated from the post–cycle 1 multi–time point ¹⁷⁷Lu SPECT/CT imaging (reference absorbed dose) has a median value of 0.5 Gy/GBq with range of 0.2–1.0 Gy/GBq. Correlations among all 12 baseline factors considered are presented in Supplemental Figure S2 (Supplemental Digital Content, http://links.lww.com/CNM/A417) as a heat map of Pearson r values.

Absorbed Dose Prediction Using Regression Models

Univariable Models

Figure 2 shows regression results for ⁶⁸Ga PET-based factors, and Figure 3 shows biomarkers that were significant for predicting the delivered absorbed dose. Leave-one-out cross-validation for these models yields the R², RMSE, MAPE, and max APE with associated standard deviation (SD) reported in Table 2. Regression results for all factors considered (Table 1) are presented in Supplemental Figure S3 (Supplemental Digital Content, http://links.lww.com/CNM/A417). Upon visual inspection of Figures 2 and 3, it is evident that the patient with highest absorbed dose (1.0 Gy/GBq to left and right kidney), who also had the lowest eGFR, may be an outlier in many cases including, notably, in the models with PET uptake and eGFR. Accordingly, unless otherwise noted, all values are reported for correlations with this patient’s kidneys excluded. Note that this patient did not initially meet the inclusion criteria for treatment based on eGFR <30 mL/min¹⁴; however, the patient’s eGFR recovered to >30 mL/min on the day of therapy and qualified for treatment. The value used in our analysis was the eGFR at time of consult to be consistent for all patients.

FIGURE 2.

Univariable regressions for ⁶⁸Ga PET-based baseline factors, which were significant (P < 0.05) for predicting therapy-delivered renal absorbed dose. Regression lines in red are fit to all n = 50 kidneys. Regression lines in black are fit to n = 48 data points that exclude the patient (indicated by the red circles) with an unusually low eGFR.

FIGURE 3.

Univariable regressions for baseline factors retrieved from medical records, which were significant (P < 0.05) for predicting therapy-delivered renal absorbed dose. Regression lines in red are fit to all n = 50 kidneys. Regression lines in black are fit to n = 48 data points that exclude the patient (indicated by the red circles) with an unusually low eGFR.

TABLE 2.

Leave-One-Out Cross-Validation R², RMSE, MAPE (SD), and Maximum APE for Significant Univariable Predictors of Therapy-Delivered Renal Absorbed Dose (n = 48 Kidneys)

Regressor	R 2	RMSE (Gy/GBq)	MAPE (SD)	Max APE
SSTRe (mL)	0.02	0.167	29.7% (20.9%)	78.2%
Absolute neutrophils (× 103/μL)	0.24	0.147	28.5% (18.9%)	78.7%
eGFR (mL/min)	0.19	0.152	28.5% (19.2%)	107.1%
Creatinine (mg/dL)	0.05	0.164	31.8% (20.4%)	108.6%
Renal PET uptake (Bq/mL/MBq)	0.60	0.107	18.0% (13.3%)	49.5%
Renal PET SUVmean	0.40	0.131	24.2% (19.3%)	69.2%
Renal PET SUVpeak	0.12	0.158	27.9% (22.9%)	94.2%

Bland-Altman analysis plots (Fig. 4) compare pretherapy predictions of absorbed dose from the univariable models with PET uptake and eGFR against the therapy-delivered absorbed dose. We find the mean (95% confidence interval) relative percent error is 4.8% (−38.3% to 48.0%) for the model with PET uptake and 9.1% (−56.2% to 74.5%) for the model with eGFR.

FIGURE 4.

Bland-Altman plots of relative percent error in LOOCV model prediction versus therapy-delivered absorbed dose with model predictions provided by (A) univariable model with PET uptake and (B) univariable model with eGFR. Horizontal axis is the therapy-delivered renal absorbed dose, and the vertical axis is the relative percent difference between LOOCV model predictions and delivered dose $\left(\frac{A{D}_{pred,i}-A{D}_{meas,i}}{A{D}_{meas,i}}\times 100\%\right)$. Mean is indicated by solid horizontal line and the 95% confidence interval (mean ± 1.96 × SD) is indicated by dashed lines (n = 48 kidneys).

Bivariable Model

Bivariable regression when eGFR and renal PET uptake are used as predictors yields LOOCV results: R² = 0.64, RMSE = 0.102 Gy/GBq, MAPE = 17.3% (SD = 11.8%).

Surrogate Dosimetry Predictions

Predictions of therapy-delivered renal absorbed dose using Equations 1 and 2 have median value 0.8 Gy/GBq (range, 0.4–1.8 Gy/GBq). When compared with the renal absorbed doses provided by multi–time point SPECT/CT imaging, the surrogate dosimetry predictions overestimate the therapy-delivered absorbed dose by a median factor of 1.74 (range, 1.00–2.82).

DISCUSSION

To our knowledge, this is the first study investigating regression models for predicting renal absorbed doses in ¹⁷⁷Lu-DOTATATE PRRT using baseline ⁶⁸Ga PET imaging and nonimaging biomarkers. In LOOCV, RMSE and MAPE (SD) for the univariable model with PET uptake is 0.107 Gy/GBq and 18.0% (13.3%) (Table 2). Notably, predictions for 65% of kidneys have an absolute percent error of <20% (Fig. 4). Beyond univariable models, we were interested in pursuing bivariable models that might provide improved performance. PET uptake and eGFR performed well in univariable analysis (Table 2). These findings, coupled with the biological rationale that decreased renal function indicated by decreased eGFR would lead to slower rates of clearance potentially resulting in an increased renal absorbed dose that may not be captured by PET uptake alone, motivated our bivariable model utilizing PET uptake and eGFR. The LOOCV RMSE and MAPE (SD) for this model are 0.102 Gy/GBq and 17.3% (11.8%) and only demonstrates marginal improvement over the univariable models and indicates that PET uptake alone is robust for posttherapy renal absorbed dose prediction.

Other predictors aside from PET uptake and eGFR were significant for prediction of therapy-delivered absorbed dose in univariable analysis (Table 2). Renal PET SUV_mean and PET SUV_peak are significant and correlated in the expected direction with delivered absorbed dose, but slightly underperform renal PET uptake. Total lesion SSTRe also showed a weak correlation with therapy-delivered absorbed dose in the direction expected based on the tumor sink effect.¹⁸ Among non-PET biomarkers, eGFR and creatinine are both significant, and once again, the direction of the correlations matches expectations from biology and relationships observed by other groups^10,11 with delivered absorbed dose increasing with increased creatinine and decreasing with increased eGFR. Absolute neutrophils are also significantly correlated with delivered renal absorbed dose with a possible explanation provided by increased neutrophil count being associated with severity of chronic kidney disease.²⁶

The surrogate renal doses calculated using Equations 1 and 2 overestimate the reference absorbed dose by a median factor of 1.74. This result is consistent with the fact that the amino acid solution that is coadministered with ¹⁷⁷Lu-DOTATATE but is not given with the ⁶⁸Ga-DOTATATE is expected to reduce the absorbed dose to kidneys by 47% (range, 34%–59%).¹⁴ This indirect confirmation of the nephroprotection offered by the amino acid infusion further supports the continuation of the practice, especially if dose escalation is to be considered.

The tumor sink effect, whereby the sequestration of radiotracer by tumor masses leads to less availability for healthy tissues, has been reported for ⁶⁸Ga-DOTATATE PET¹⁸ and ⁶⁸Ga-PSMA-11 PET.^13,27,28 Yet, the study by Werner et al²⁹ for ⁶⁸Ga-DOTATOC PET did not find evidence of a tumor sink effect. We find a significant correlation for total lesion SSTRe with renal PET uptake (R² = 0.23, P < 0.001 in Supplemental Figure S2, Supplemental Digital Content, http://links.lww.com/CNM/A417) and a significant, but weaker, correlation with therapy-delivered renal absorbed dose in univariable analysis (R² = 0.09, P = 0.04) (Supplemental Figure S3, Supplemental Digital Content, http://links.lww.com/CNM/A417).

A clinical trial modifying injected ¹⁷⁷Lu-DOTATATE activity² used a prediction of renal absorbed dose before the first cycle that is based on eGFR and body surface area, but not PET uptake.¹¹ These predictions were reported to have a maximum error of 105.8% based on a retrospective analysis of 33 patients. The improved accuracy demonstrated by our regression models (Table 2) indicates that PET uptake information could be used to further escalate the activity for the first cycle while maintaining the same level of patient safety. A retrospective study of 500 patients, which showed most patients (~2/3), can receive more than the 4 cycles while adhering to a 23-Gy renal absorbed dose limit³⁰ and highlights the interpatient variability and the role of individualized treatment planning in ¹⁷⁷Lu-DOTATATE PRRT. It is worth noting that, in our study, the predicted absorbed doses tend to underestimate the therapy-delivered absorbed dose when the therapy-delivered absorbed dose was high and overestimate when it was low (Fig. 4). This effect could indicate the need for further investigation of multivariable regression models to compensate for the biasing phenomenon.

The primary limitations of our study are the small sample size and the lack of validation on an independent cohort of patients, which could affect the generalizability of our models. Although we showed the proof-of-concept of combining pretherapy ⁶⁸Ga PET uptake with eGFR, this concept could be further expanded to additional biomarkers and clinical factors when adequate sample sizes for appropriate statistical power are available. Furthermore, our study was limited to PET uptake and biomarkers that are available before the start of the first cycle of treatment and was only tested for predicted renal absorbed doses after the first cycle. In addition, we note that some of the clinical factors we investigated can be highly variable. For example, eGFR can be affected by the patient’s hydration level at the time of testing. Another limitation of our study is that we only investigated prediction of renal absorbed doses, although ideally, dosimetry-guided treatment should consider both the absorbed dose to normal organs and tumors to evaluate the efficacy-toxicity tradeoff. Even if validated, the models set forth in the current work are meant to be supplemental to and are not designed to be a replacement for posttherapy dosimetry. Recent efforts to simplify posttherapy dosimetry with single time point imaging^31,32 will couple well with our pretherapy dosimetry models to reduce the risk of renal toxicity in ¹⁷⁷Lu-DOTATATE PRRT and potentially enable patient-specific treatment planning beyond the standard 4-cycle administration as needed to enhance efficacy.

CONCLUSIONS

We present regression models that predict mean absorbed dose to the kidneys after the first cycle of ¹⁷⁷Lu-DOTATATE PRRT using only baseline information routinely available before treatment. We show that univariable models of ⁶⁸Ga-DOTATATE PET renal uptake can be a reliable predictor of renal absorbed dose delivered by ¹⁷⁷Lu-DOTATATE PRRT. Furthermore, combining the renal PET uptake with eGFR does not improve these estimates. We expect to validate our models on an independent cohort, thus providing clinicians with the ability to select patients and plan ¹⁷⁷Lu-DOTATATE PRRT with a high level of confidence before the first cycle of treatment.