In this manuscript, we report on the development of a dosimetry biomarker management approach to administer precision RAI therapy to patients with mDTC. The 124I PET imaging biomarker provides the treating physician with a tool to select the amount of radioactivity (mCi or MBq) expected to achieve a prescribed radiation-absorbed dose (cGy) to all lesions with SUVs above the selected value likely to achieve a therapeutic response. This manuscript focuses on the biomarker method development, while validation studies with patient outcomes and dose response findings in patients treated in an ongoing study will be discussed in subsequent manuscripts.
Like external beam radiotherapy, it is widely believed that the treatment effectiveness of RAI at the individual lesion level is dependent on the radiation-absorbed dose to the individual lesion. Maxon et al. were among the first to make technically adequate quantitative dose estimates (14). These measurements showed complete responses at 8,500 cGy per lesion in 75% of metastatic thyroid cancer lesions to lymph nodes, and a treatment response threshold in a majority of lesions was observed at > 2000 cGy. Based on prior work by Maxon et al., we made an operational definition that a patient with any lesion with a predicted dose of > 2000 cGy would likely respond to treatment (14). Therefore, in this study, we used an actionable threshold of 2,000 cGy as the minimum radiation-absorbed dose for the patient to proceed with 131-I RAI therapy, although other thresholds could be used. Consequently, mDTC patients are administered 131I RAI treatment only if they are likely to benefit from it.
Using the proposed approach, we confront a major problem of RAI therapy: the considerable heterogeneity of radiation-absorbed dose to lesions within a given patient, and between patients with mDTC, at a given amount of MBq 131I administered. Variation in measured cGy dose from lesion to lesion may be both technical and biologic in nature. Although the technical features such as difficulty in imaging small tumors quantitatively may play a role in inaccurate dosimetry, it is likely that the observed differences in cGy from lesion to lesion is predominantly biological in nature. This hypothesis is being actively explored.
When investigating single timepoint predictor, the 48-hour timepoint was found to be the best single predictor of the average integrated AUC uptake (directly proportional to cGy dose) for individual lesions. Further research is warranted to explore the impact of characteristics such as clearance, as it can vary greatly from one patient to another, and to encapsulate outlier radioiodine kinetic profiles into the prediction model. This will include extending the regression model to the possibility of adding a second timepoint for the prediction. The current research incorporates useful information about the variability in lesion uptake by considering all lesions from all subjects in the calculation of a prediction interval, in order to best determine the predicted prescribed radioactivity to achieve a radiation-absorbed dose that exceeds the desired threshold for therapeutic efficacy with a stipulated precision, typically 90% or 95% probability. Table 5 provides a useful statistical tool that allows treating clinicians to select a lesion within a patient that they wish to prescribe a radiation dose of at least 2000 cGy. The table columns provide the radioactivity amount that should be administered to have a 50%, 90%, 95%, and 97.5% probability of achieving a 2000 cGy target dose. In patients with multiple lesions, those with higher SUVs at 48h would be expected to receive proportionally higher doses, and those with lower SUVs lower doses. Based on the statistical model derived prediction interval (PI), Table 5 allows physicians to estimate the fraction of a patient’s lesion burden that will receive a given radiation dose such as 20000 cGy, which is expected to produce some therapeutic benefit, thereby assisting the physician in determining whether a patient will benefit from radioiodine therapy in all or some of the lesions.
Our approach shows promising results in demonstrating a correlation between integrated AUC and a single timepoint in our learning set of 21 patients, but to further improve the precision of our predictor, recruitment of a larger patient cohort is in process. A simulation study estimated that an increase in sample size from 21 to 60 patients would increase the precision (as measured by the half-width of the 95% PI on the log-scale) from 1.38 to 1.33, but beyond this number, the gain is very small (1.31 and 1.30 with 120 and 1,000 patients, respectively).
Dosimetry for both tumor and normal tissues during RAI and other targeted radiotherapy must be carefully considered if we are to maximize potential benefits for individual patient management. Furthermore, we recognize the unmet need for more optimal dosimetry for both normal tissues as well as tumors. Other investigators have also recognized this need for a practical single-timepoint imaging method, particularly to assure the patient safety of those undergoing theranostic treatments (15–29). Hänscheid et al investigated the accuracy of a single imaging timepoint to predict the dosimetry for key normal tissues and tumor vs. clearance fitting from serial gamma camera images from 177Lu-DOTATATE or 177Lu-DOTATOC treatments (27). In that study, they looked at the dose to kidney, liver, spleen, and 30 NET lesions following the administration of 177Lu-DOTATATE or 177Lu-DOTATOC. They studied different timepoints post-administration and found the lowest maximum errors at 96h and reported deviations from the time integral of median of + 5% (range, -9% to + 17%) for kidneys, + 6% (range, -7% to + 12%) for livers, + 8% (range, + 2% to + 20%) for spleens, and + 6% (range, -11% to + 16%) for NET lesions (18). Willowson et al performed a similar study with a focus on kidney dosimetry to anticipate renal toxicity (17). They reported an average deviation from doses obtained from complete image data on cycle 1 of 13% and 2% when using 4 h data only and 24 h data only. A recent study by Hou et al (23) examined different theranostic agents and suggested that simplified single-timepoint dosimetry approaches may work well for 177Lu-DOTATATE, but the generalizability of single-timepoint imaging for dosimetry for certain targeting agents such as 177Lu-PSMA targeted bone metastases may be less successful.
Finally, personalized radioiodine dosimetry in RAI focused on estimating the MTA would ensure that treatment would not result in a blood and whole-body dose that would exceed the threshold for serious bone suppression or radiation lung fibrosis for patients with extensive lung metastases (10). The shift in dosimetry emphasis being proposed here is toward the rational selection of treatment activity based on a population-averaged statistical model relating single-timepoint 124I lesion SUV measurements with dose expectation and subsequently response prognosis, consistent with the normal tissue-limiting MTA.
Clinically, we recognize that quantitative SPECT imaging is an alternative approach to lesional dosimetry. Since 131I is clinically approved and widely used, potentially developing a single time point approach to lesional dosimetry based on 131I is certainly appealing. However, the 124I PET has major technical advantages mainly related to a combination of higher sensitivity (80–100 times) and better resolution, which translates into significantly better quantitative performance, especially for small metastatic lesions. For these reasons we chose PET and 124I for the proof-of-principal phase of biomarker development and in the setting of clinical research.
In summary, we have provided initial validation of a single time point lesional dosimetry biomarker utilizing 124 I PET scanning. When coupled with a knowledge of the MTA determined by blood and whole-body clearance, clinicians can utilize the relationship between administered activity and lesion al dosimetry to optimize a RAI treatment strategy that maximizes therapeutic effectiveness while minimizing the risk of serious adverse events.