Abstract
PET/CT with 68Ga-DOTA-somatostatin analogs has been tested for therapy monitoring in patients with neuroendocrine tumors (NETs). However, SUVs in tumors do not correlate with the net influx rate (Ki), as a representation of the somatostatin receptor expression. In this study, tumor-to-blood ratio (TBR) was evaluated as an alternative tool for semiquantitative assessment of 68Ga-DOTATOC and 68Ga-DOTATATE tumor uptake and as a therapy monitoring tool for patients with NETs. Methods: Twenty-two NET patients underwent a 45-min dynamic PET/CT scan after injection of 68Ga-DOTATOC or 68Ga-DOTATATE. Ki was determined using the Patlak method, and TBR was calculated for the 40- to 45-min interval. Results: A linear relation was found between Ki and TBR, with a square of Pearson correlation of 0.98 and 0.93 for 68Ga-DOTATOC and 68Ga-DOTATATE, respectively. Conclusion: A high correlation was found between Ki and TBR. Hence, TBR reflects somatostatin receptor density more accurately than SUV and is suggested as the preferred metric for semiquantitative assessment of 68Ga-DOTATOC and 68Ga-DOTATATE tumor uptake.
Neuroendocrine tumors (NETs) are rare neoplasms that arise from endocrine cells distributed throughout the body and have diverse biologic and clinical characteristics (1). The feature of high cellular expression of somatostatin receptors (SSTRs) in NETs enables the use of radiolabeled somatostatin analogs for imaging and therapy. During the past decade, PET using 68Ga-labeled somatostatin analogs, such as 68Ga-DOTATOC, 68Ga-DOTANOC, and 68Ga-DOTATATE, has gradually replaced SSTR scintigraphy with 111In-DTPA-octreotide (OctreoScan; Mallinckrodt) (2,3) and become the standard method for SSTR imaging of NETs (4).
PET/CT with 68Ga-DOTATOC and 68Ga-DOTATATE has also been suggested as a tool for evaluation of therapy response in patients with NETs (5–8). For metabolic tracers such as 18F-FDG, it can be assumed that the tracer’s distribution volume is the whole body since glucose is consumed by all tissues, which means that the SUV can be used as a reasonable measure of metabolism. A challenge with PET/CT using receptor ligands, such as 68Ga-DOTATATE and 68Ga-DOTATOC, is that the distribution volume instead is confined to those tissues that are, in fact, taking up the tracer, which may affect the SUV quantification. In one study (5), it was found that the changes in tumor SUV between baseline and follow-up 68Ga-DOTATOC PET/CT did not correlate with the therapy outcome of peptide receptor radionuclide therapy. The same finding was also reported for another study (6), although changes in the tumor-to-spleen SUV ratio between baseline and follow-up 68Ga-DOTATOC were shown to be more accurate than changes in tumor SUVmax to evaluate the response to peptide receptor radionuclide therapy. The difficulties of applying static tumor uptake measurements in these 2 therapy monitoring studies may at least partly be explained by the results in a study (7) on tracer kinetics of 68Ga-DOTATOC and 68Ga-DOTATATE. In that work, net influx rate (Ki), assumed to more accurately reflect SSTR density than SUV, was estimated on the basis of dynamic PET imaging, and it was found that SUV saturated (SUV > 20–25) at a static value for high Ki values (Ki > 0.2). Hence, SUV does not appear to reflect SSTR density for tumors with high SSTR expression. The hypothesis of the present work is that saturation in SUV for high Ki values may be explained by low availability of 68Ga-DOTATOC and 68Ga-DOTATATE in the blood at some time after administration due to the substantial amounts of SSTR in these patients. Hence, the tumor-to-blood ratio (TBR) may be a better metric than SUV to quantify changes in SSTR expression to assess NET therapy response. The aim of this study was to evaluate the correlation between Ki and TBR for patients undergoing PET/CT with 68Ga-DOTATOC and 68Ga-DOTATATE.
MATERIALS AND METHODS
Patients
The data in this work were collected from 3 different studies that were approved by the Regional Ethics Review Board in Uppsala. All patients gave written informed consent before inclusion in each study.
The study included 22 patients (11 men and 11 women; mean age, 63 y; range, 47–75 y) with histologically confirmed disseminated NETs (10 small-intestinal, 6 pancreatic, 2 rectal, 1 duodenal, 1 lung, and 2 pancreatic neuroendocrine cancers; 4 grade 1 [Ki-67 ≤ 2%], 15 grade 2 [Ki-67 = 3–20%], 2 grade 3 [Ki-67 ≥ 20%], and 1 without biopsy). The clinical patient data are presented in Table 1. Some patients were examined with both 68Ga-DOTATOC and 68Ga-DOTATATE on consecutive days (n = 6), whereas the remainder were examined with either 68Ga-DOTATOC or 68Ga-DOTATATE. One patient was examined with both 68Ga-DOTATOC and 68Ga-DOTATATE with 1 y apart. Sixteen patients underwent a 68Ga-DOTATOC PET/CT examination after a bolus injection of 131 ± 47 MBq, 25 ± 8 μg (range, 62–198 MBq, 15–47 μg), and 13 patients underwent a 68Ga-DOTATATE PET/CT examination after a bolus injection of 107 ± 31 MBq, 23 ± 9 μg (range, 76–197 MBq, 8–39 μg).
Demographics and Clinical Characteristics of Study Patients
Image Acquisition and Reconstruction
The patients were examined on a Discovery ST, Discovery IQ or Discovery MI PET/CT scanner (GE Healthcare). They underwent a low-dose CT scan (140 kV; auto mA, 20–80 mA) followed by a 45-min dynamic PET examination of the abdomen. The dynamic PET examination started simultaneously with the intravenous injection of 68Ga-DOTATOC or 68Ga-DOTATATE and consisted of 22 time frames of increasing duration (6 × 10 s, 3 × 20 s, 3 × 60 s, 5 × 180 s, and 5 × 300 s). All appropriate corrections were applied to the PET data, and the reconstruction settings are specified in Table 2.
Reconstruction Settings for the 3 Different Scanners
Image-Derived Input Functions
The total radioactivity concentration in the arterial plasma was used as the input function. Volumes of interest were drawn using a 70% isocontour over the descending thoracic aorta in 10 consecutive image planes in the time frame in which the first passage of the bolus was best visualized (frames 1–10) and then projected onto all time frames in the dynamic examination, generating an arterial time–activity concentration curve (NEDPAS software; VU University Medical Centre, Amsterdam (9)). The image-derived input functions were calculated by multiplying the arterial time–activity concentration curve with a fixed plasma–to–whole-blood ratio of 1.6 based on data from previous work (7) and plasma–to–whole-blood ratio (unpublished data) (mean, 1.6 for both tracers; range, 1.45–1.73). Blood SUV at 40–45 min was determined using the isocontour volume of interest (70%) in the descending aorta at the last frame of the dynamic scan.
Kinetic Analysis
Tumors with a diameter larger than 1 cm and with high tracer uptake (determined visually) were included for evaluation. Isocontour tumor volumes of interest (50%) were drawn in the 20- to 45-min (frames 18–22) summation image of the dynamic examination and were projected onto all time frames to generate tumor time–activity concentration curves. Ki was determined using the Patlak method (10) as previously described (11). SUV and TBR were computed for the last frame of the dynamic scan (i.e., 40–45 min after injection).
Statistical Analysis
The difference in blood SUV between high (>0.2) and low (<0.2) Ki values was determined using a Mann–Whitney test, with the significance level set to a P value of less than 0.05 (Prism, version 6.07; GraphPad Software, Inc.). In this test, only 1 tumor per patient was included (the tumor with the highest Ki) because some patients had several tumors whereas others had only one, and inclusion of all tumors would bias the results toward patients with more tumors.
The relation between Ki and TBR was evaluated using linear regression and Pearson correlation and compared with the relation between Ki and SUV. In this test, all tumors were included.
Ki, SUVtumor, and TBR were also compared between 68Ga-DOTATOC and 68Ga-DOTATATE using Deming regression, Pearson correlation, and Wilcoxon matched-pairs tests (significance level set to P < 0.05).
RESULTS
In total, 71 tumors were included in the study: 38 for 68Ga-DOTATOC (6 patients with 1 tumor, 3 with 2 tumors, 3 with 3 tumors, 3 with 4 tumors, and 1 with 5 tumors) and 33 for 68Ga-DOTATATE (4 patients with 1 tumor, 4 with 2 tumors, 2 with 3 tumors, 1 with 4 tumors, 1 with 5 tumors, and 1 with 6 tumors). For 68Ga-DOTATOC, SUV in aortal blood at 45 min after injection was significantly lower in patients with high Ki values than in those with low Ki values (P = 0.017, Mann–Whitney test; Fig. 1A). The difference was smaller for 68Ga-DOTATATE (P = 0.127, Mann–Whitney test; Fig. 1B). The relation between SUV in blood and Ki is presented in Figures 1C and 1D for 68Ga-DOTATOC and 68Ga-DOTATATE, respectively.
(A and B) Box plots of SUV in blood at 40–45 min after injection for 68Ga-DOTATOC (A) and 68Ga-DOTATATE (B) for high and low Ki values. One tumor per patient is included in plots. Boxes are median and interquartile range, and whiskers are full range of data. Significant differences (P < 0.05) were found in SUVblood between high and low Ki for 68Ga-DOTATOC, however, not for 68Ga-DOTATATE (P > 0.05). (C and D) Relation between Ki and SUV in blood for 68Ga-DOTATOC (C) and 68Ga-DOTATATE (D). Solid line represents exponential fit (y = a/x) for visual illustration.
A linear relation was found between Ki and TBR (all tumors included), with a square of Pearson correlation of 0.98 and 0.93 for 68Ga-DOTATOC and 68Ga-DOTATATE, respectively (Fig. 2). Ki was compared with SUV for the same tumors, and the relation is illustrated in Figures 2C and 2D. The square of Pearson correlation between Ki and tumor SUV using a hyperbolic fit was 0.81 and 0.78 for 68Ga-DOTATOC and 68Ga-DOTATATE, respectively. Tumor SUV, blood SUV, TBR, and Ki for each patient are also presented in Supplemental Tables 1 and 2 for both tracers (supplemental materials are available at http://jnm.snmjournals.org).
Correlation between Ki and TBR for 68Ga-DOTATOC (A) and 68Ga-DOTATATE (B) and between Ki and SUV in tumors for 68Ga-DOTATOC (C) and 68Ga-DOTATATE (D). Solid lines represent linear regression fits (A and B) and fits to hyperbolic line (C and D), and dashed lines are 95% confidence band of these fits.
A significant difference in TBR was found between 68Ga-DOTATOC and 68Ga- DOTATATE (P = 0.019, Wilcoxon matched-pairs test). However, for SUVtumor and Ki, there were no significant differences (SUVtumor, P = 0.413; Ki, P = 0.083; Wilcoxon matched-pairs test). A linear relation between 68Ga-DOTATOC and 68Ga-DOTATATE was found for Ki, SUVtumor and TBR, with a square of Pearson correlation of 0.81, 0.76, and 0.88, respectively (Fig. 3). The slopes of the Deming regression line were 1.2, 1.1, and 1.4 for Ki, SUVtumor, and TBR, respectively.
Comparison of Ki, SUVtumor, and TBR between 68Ga-DOTATOC and 68Ga-DOTATATE. Significant difference was found between 68Ga-DOTATOC and 68Ga-DOTATATE for TBR (P = 0.019, Wilcoxon matched-pairs test) but not for Ki or SUVtumor (P = 0.083 and 0.413, respectively).
DISCUSSION
Early prediction of treatment response is essential to guide tumor therapy and avoid unnecessary side effects and costs from ineffective treatments. SUV has been proposed as a measure of SSTR density in NETs, but changes of the tumor SUV in 68Ga-DOTATOC PET/CT during peptide receptor radionuclide therapy have not been found to reliably correlate with the treatment outcome (5,6,8). Ki is likely to reflect the tumor SSTR density more adequately than SUV. In a previous study (7) comparing 68Ga-DOTATOC and 68Ga-DOTATATE, it was shown that Ki and SUV did not correlate linearly for NETs, especially at high SUVs (>20–25). The present work partly used data from the same subjects, and as seen in Figures 2C and 2D, the addition of more subjects did not alter this conclusion.
This study suggests that the nonlinear relation between Ki and SUV for high Ki values can be attributed to faster blood clearance in patients with a high tumor receptor expression. For 68Ga-DOTATOC, SUV in blood at 45 min after injection was significantly lower in patients with high Ki values than in those with low Ki values (Fig. 1A). For 68Ga-DOTATATE, this difference was not significant (Fig. 1B). The low blood SUV in patients with high Ki values may be overestimated because of spill-in from surrounding tissues and a positive bias in low-activity areas as commonly seen in PET, whereas the high blood SUV may be underestimated because of the partial-volume effect. Taking these factors into consideration, the difference in blood SUV between the 2 groups would increase even further.
Figures 2A and 2B clearly illustrate that, contrary to the nonlinear relation between Ki and SUV, there is a linear relation between Ki and TBR and that the nonlinear correlation between Ki and SUV can be attributed to low availability of tracer in blood. Since plasma concentrations during the course of the scan are implicitly considered when estimating Ki, differences in plasma concentration of the tracer do not affect accuracy. However, since the low blood activity concentrations will limit the absolute amount of tracer available for uptake in tissue, SUV will be affected by low plasma concentrations and will not always follow Ki. Most probably, the total amount of SSTR in some patients is so large that nearly all peptide is cleared from the plasma during the initial part of the examination, leading to the apparent saturation of tumor SUVs. The clearance rate for the individual patient also depends on, for example, kidney function, uptake in kidneys, and spleen. However, the amount of tracer in blood is probably one of the most important factors; moreover, the amount of tracer in blood is a factor that may be influenced by how much peptide is administered, whereas the SSTR expression in each tumor, the combined total SSTR expression, and the patient’s renal function cannot be affected.
The activity concentrations in blood were determined by delineating the aorta using a 70% isocontour volume of interest. In this case, the underestimation of the activity concentration in the aorta is theoretically 6%–7% when assuming an aorta diameter of 2.5 cm and a spatial resolution of 5 mm. The results in this paper were not corrected for this underestimation. However, since this underestimation will affect Ki and TBR equally, the conclusions of the study would not change.
The patients in this study were from 3 different NET studies and thus underwent examinations on different scanners with varied reconstruction settings. As known, reconstruction parameters affect Ki and SUV (12). However, since the reconstruction will affect Ki and SUV similarly, the variations in reconstruction settings between scanners will not affect the conclusions of the present work. For example, the partial-volume effect will similarly affect Ki and SUV and the results will consequently be the same regardless of whether the reconstruction includes correction for partial-volume effect or not.
CONCLUSION
A linear relation with a high correlation was found between Ki and TBR for both 68Ga-DOTATOC and 68Ga DOTATATE. Hence, TBR reflects SSTR density better than SUV and would be the preferred measurement tool for semiquantitative assessment of 68Ga-DOTATOC and 68Ga-DOTATATE tumor uptake and as a means for NET therapy monitoring.
DISCLOSURE
No potential conflict of interest relevant to this article was reported.
KEY POINTS
QUESTION: The purpose of this study was to evaluate the use of TBR as an alternative tool for semiquantitative assessment of 68Ga-DOTATOC and 68Ga-DOTATATE tumor uptake and as a therapy monitoring tool for patients with NETs by evaluating the relation between Ki and TBR for 68Ga-DOTATOC and 68Ga-DOTATATE.
PERTINENT FINDINGS: For both 68Ga-DOTATOC and 68Ga-DOTATATE, a linear relation with a high correlation was found between Ki and TBR. Hence, TBR can be used as a tool for semiquantitative assessment of 68Ga-DOTATOC and 68Ga-DOTATATE tumor uptake and as a means for NET therapy monitoring.
IMPLICATIONS FOR PATIENT CARE: The finding offers a new tool for assessing tumor uptake of 68Ga-DOTATOC and 68Ga-DOTATATE and a new therapy monitoring tool for patients with NETs.
Acknowledgments
We thank the staff of the PET Centre for their assistance with the PET/CT examinations.
Footnotes
↵* Contributed equally to this work.
Published online Jul. 13, 2019.
- © 2020 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication March 11, 2019.
- Accepted for publication June 18, 2019.
