Tumour-to-normal tissue (T/N) dosimetry ratios role in assessment of ⁹⁰Y selective internal radiation therapy (SIRT)

Abstract

Objective:

The purpose of our work is to assess the role of tumour-to-normal tissue (T/N) dosimetry ratios for predicting response in patients undergoing locoregional therapy to the liver with ⁹⁰Y microspheres.

Methods:

A total of 39 patients (7 female:32 male, mean age 68.3 ± 7.6 years), underwent positron emission tomography (PET)/CT imaging after treatment with ⁹⁰Y microspheres. For attenuation correction and localization of the ⁹⁰Y microspheres, the low-dose, non-diagnostic CT images from PET/CT were used. The acquisition took 15 min and the reconstruction matrix size was 200 × 200 × 75 mm and voxel size of 4.07 × 4.07 × 3.00 mm. For dosimetry calculations, the local deposition method with known activity of ⁹⁰Y was used. For each patient, regions of interest for tumour(s) and whole liver were manually created; the normal tissue region of interest was created automatically. mRECIST criteria on MRI done at 1 month post-treatment and subsequently every 3 months after ⁹⁰Y treatment, were used to assess response.

Results:

For 39 patients, the mean liver, tumour and normal tissue doses (mean ± SD) were, 55.17 ± 26.04 Gy, 911.87 ± 866.54 Gy and 47.79 ± 20.47 Gy, respectively. Among these patients, 31 (79%) showed complete response (CR) and 8 (21%) showed progression of disease (PD). For patients with CR, the mean T/N dose ratio obtained was 24.91 (range 3.09–80.12) and for patients with PD, the mean T/N dose ratio was significantly lower, at 6.69 (range 0.36–14.75).

Conclusion:

Our data show that patients with CR have a statistically higher T/N dose ratio than those with PD. Because, the number of PD cases was limited and partial volume effect was not considered, further investigation is warranted.

Advances in knowledge:

T/N dosimetry ratios can be used for assessing response in patients undergoing locoregional therapy to the liver with ⁹⁰Y microspheres.

Introduction

Recently, there has been a growing interest in the use of ⁹⁰Y selective internal radiation therapy (SIRT) in the treatment of unresectable hepatocellular carcinoma (HCC), because of its good time-to-progression, overall survival and toxicity profile.¹ In addition, the intra arterial injection of ⁹⁰Y microspheres, if properly performed, delivers high-radiation doses to the tumour while sparing liver parenchyma. Prior to the ⁹⁰Y SIRT procedure itself, the administration and imaging of ^99mTc-MAA (microaggregated albumin) particles, an analogue of ⁹⁰Y microspheres, is performed. Its purpose is to assess lung shunting, assess intrahepatic distribution and exclude extrahepatic deposition. A lung shunt of 20% or higher, or any extrahepatic deposition can be contraindications for treatment. ^99mTc-MAA scans can also be used for predictive personalized dosimetry calculations. In this case, perfect matching between ^99mTc- MAA and ⁹⁰Y distribution is assumed.² However, several studies have shown that the correlation is not always accurate due to differences in particle characteristics, as well as differences in catheter positioning between ^99mTc-MAA and ⁹⁰Y studies.^3,4 Therefore, the validation of the post-SIRT ⁹⁰Y microsphere distribution is necessary. First, it is used to detect possible extrahepatic activity, which can cause ulceration, gastrointestinal bleeds and other serious complications.^5–7 Second, it can be used to perform accurate measures of the absorbed radiation dose delivered to normal liver tissue and liver tumours. These data can help determine whether treatment successes, or treatment failures, and adverse events, can be attributed to the dose that the normal liver or tumour received. The post-therapy dosimetry is also an important predictor of treatment efficacy.⁸ The true post-therapy ⁹⁰Y distribution can be obtained by positron emission tomography (PET) or bremsstrahlung single photon emission computed tomography (bSPECT) imaging.⁹ However, quantitative bSPECT imaging is challenging due to the continuous nature of the bremsstrahlung energy spectrum, inefficient bremsstrahlung production, scatter and septal penetration.¹⁰ Post-therapy PET ⁹⁰Y images are the gold-standard as they are far superior, both qualitatively and quantitatively, to bSPECT ⁹⁰Y images.¹¹

In this retrospective study, the purpose was to assess the role of tumour-to-normal tissue (T/N) dosimetry ratios for evaluation of response in patients undergoing liver treatment with ⁹⁰Y microspheres.

Methods and materials

Study population

From June 2013 to May 2018, 46 patients with unresectable liver metastases of different primary tumours were enrolled. Seven of the cases were classified as partial response (PR) and stable disease (SD), according to mRECIST therapy response criteria, and were omitted because of the small number of cases. The 39 remaining patients (7 female:32 male, mean age 68.3 ± 7.6 years, (mean ± SD)), were included in the study. For these patients, the size of the tumour was 12.45 cm³ ± 4.76 (median, 5.56 cm³; range, 79.40–1.86 cm³), and the mean total radiation activity infused was 2.07 GBq ± 1.03 (median, 1.90 GBq; range, 0.5–4.7 GBq). In all cases, there was no extrahepatic leakage and lung shunting was less than 5%. The institutional review board (IRB) approved the imaging protocols, and written informed consent was obtained for each subject enrolled. 34 patients were treated with TheraSphere ^® (glass microspheres; Boston Scientific, Marlborough, MA) and 5 with SIR-Sphere ^® (resin microspheres; Sirtex Medical, Sydney, Australia). Exclusion criteria were a small number of cases according to the mRECIST evaluation, which in this study excluded PR and SD cases.

Imaging protocol

All patients had a mapping scan using ^99mTc-MAA SPECT/CT performed on a dual-head Infinia SPECT/CT gamma camera (GE Medical Systems, Milwaukee, WI), 2 weeks prior to SIRT with ⁹⁰Y microspheres. An activity of 185 MBq ± 6% and a LEHR collimator were used in all cases. 60 views at 30 s per view were acquired for the SPECT/CT studies. An ordered subset expectation maximization algorithm with 2 iterations and 10 subsets (2i10s) and a Hann window with a cut-off frequency of 0.48 cycles/pixel was used to reconstruct images. The reconstruction matrix size was 128 × 128 and the pixel size was 4.4 mm. The non-diagnostic CT scan was used for attenuation correction. The CT was performed with 120 kV, 5.8 mAs, 10 mm axial sampling and a 256 × 256 matrix size. After SIRT with ⁹⁰Y microspheres, each patient was imaged on a time-of-flight, four-ring PET/CT system Biograph mCT (Siemens Medical Systems, Erlangen, Germany). Due to the relatively long axial field of view, only one-bed position acquisitions were used. Attenuation correction and localization of the ⁹⁰Y microspheres was performed using the low-dose CT images. The acquisition time was 15 min. The reconstruction matrix size was 200 × 200 x 75, and the voxel size was 4.07 × 4.07 x 3.00 mm³. ⁹⁰Y PET imaging is possible because there are 32 pair productions per million ⁹⁰Y decays. However, the number of true events is low because the LSO crystal and bremsstrahlung random coincidences are often a large fraction of the prompt coincidences. In order to improve activity recovery and the accuracy of quantification, the iterative 3D Poisson-ordered subset expectation maximization algorithm with point spread function and time-of-flight correction was used for the reconstruction of PET data, with 2 iterations, 21 subsets (2i21s) and a 5 mm Gaussian post-reconstruction filter.¹²

Region of interest (ROI) creation and dosimetry calculation

Reconstructed PET/CT images were transferred to a processing platform and region of interest (ROI) of the liver and tumour was created to calculate ⁹⁰Y dosimetry using MIM v. 6.8 software (MIM Software Inc., Beachwood, OH). In order to precisely determine tumour location, each patient had a contrast-enhanced MRI, CT or cone-beam CT imaging study, done 4–12 days before ^99mTc-MAA scan. These images, which were available on the hospital PACS system, were used as reference images in creation of tumour and liver ROIs in the CT component of the post-therapy ⁹⁰Y PET/CT study. In Figure 1, post-therapy ⁹⁰Y PET/CT fused images are shown with liver and tumour ROIs created using MRI VIBE corresponding images as references. MRI images, whether or not they are contrast enhanced, have much better soft tissue contrast than those from CT, allowing for superior delineation between tumour and normal liver parenchyma. In the MIM v. 6.8 software, the automatic process of transferring ROIs from MRI to CT did not always work, mainly because the scaling function was not available, and deformable transformations could not always correctly handle the differences in matrix and voxel size. However, as seen in the Figure 1, overlapping liver MRI and CT ROIs are close, showing good match between the image modalities. For all patients, tumour and liver ROIs were manually finalized on the CT images from PET/CT studies.

Figure 1.

In the first row, there are ⁹⁰Y post-therapy fused PET/CT liver images with tumour and liver ROIs. In the second row, there are corresponding MRI VIBE sequence images with the same ROIs. Tumour ROIs were first derived manually using MRI images and then transferred manually to the PET/CT images, although an attempt was made to transfer tumour ROIs automatically. PET, positron emission tomography; ROI, region of interest; VIBE, volumetric interpolated breath-hold examination.

For the dosimetry calculations, local deposition method with known activity of ⁹⁰Y was used. Since ⁹⁰Y decays almost entirely with β^- emission (0.93 MeV mean energy, half-life 64.1 h, 2.5 mm mean tissue penetration),¹³ the local deposition method is a practical alternative dose–point–kernel convolution approach.¹⁴

Statistical analysis

Welch’s t-test, or unequal variances t-test was used for statistical analysis because of unequal sample size of variables and unequal sample distribution variance. p-Values less than 0.05 were considered significant. Statistical calculations were performed using MedCalc Statistical Software v. 19.4.1.¹⁵ Variables were reported as mean ± SD and/or as percentages.

Results

For 39 patients, the mean liver, tumour and normal tissue doses (mean ± SD) were, 55.17 ± 26.04 Gy, 911.87 ± 866.54 Gy and 47.79 ± 20.47 Gy, respectively. Among these patients, 31 (79%) showed complete response (CR) and 8 (21%) showed progression of disease (PD). For patients with CR, the mean T/N dose ratio obtained was 24.91 (range 3.09–80.12) and for patients with PD, the mean T/N dose ratio was significantly lower, at 6.69 (range 0.36–14.75). A two-tailed probability from the Welch test, assuming unequal variance, showed a p-value of 0.0001. Therefore, the difference between the mean T/N ratios for CR and PD subjects was significantly different. In Figure 2, the T/N ratios are shown, including their mean values and error bars with 95% confidence intervals. In Figure 3, a combined PET/CT image shows a case with a high T/N ratio of 44.18. The tumour area and relatively high ⁹⁰Y activity are nicely correlated. Figure 4 shows a case with a low T/N of 0.65, where the tumour ROI and high⁹⁰Y activity area are not overlapping.

Figure 2.

The CR and PD T/N ratios. For CR patients, the mean T/N dose ratio was 24.91 (range 3.09–80.12) and for PD patients, the mean T/N dose ratio was 6.69 (range 0.36–14.75). CR, complete response; PD, progression of disease; T/N, tumour-to-normal tissue.

Figure 3.

The case of a high T/N = 44.18 ratio. The tumour and high ⁹⁰Y activity areas are nicely overlapping. T/N, tumour-to-normal tissue.

Figure 4.

The case of a low T/N = 0.65 ratio. The tumour and high ⁹⁰Y activity areas are not overlapping. T/N, tumour-to-normal tissue.

Discussion

One of the main goals in ⁹⁰Y SIRT treatment of unresectable HCC, is to maximize tumour dose and minimize normal tissue dose, thus preserving liver parenchyma. In addition, in this process, one has to deliver to tumour at least recommended tumour mean absorbed dose and keep normal liver tissue below or equal to the safety threshold of absorbed dose of 40 Gy.¹⁶ However, the definition of standardized tumour mean dose thresholds is still debatable. Recently published studies recommended a minimum mean target absorbed dose to tumour of 100–120 Gy for HCC, liver metastatic colorectal cancer and cholangiocarcinoma,¹⁶ second study recommended a 200 Gy threshold for HCC¹⁷ and third study recommends a target mean dose of 100–250 Gy for HCC and 40–60 Gy for colorectal cancer metastases.¹⁸ For neuroendocrine tumour liver metastases, using ^99mTc-MAA and SPECT/CT system authors in a recent work concluded that an estimated dose greater than 191.3 Gy predicted treatment response with high sensitivity and specificity.¹⁹ Predicting patient response to ⁹⁰Y SIRT is still a challenging task and recently, more sophisticated methods have been proposed based on radiobiologic dose metrics, which included lesion absorbed dose metrics, biological effective dose metrics, equivalent uniform dose and equivalent uniform biological effective dose,²⁰ and combining ⁹⁰Y PET-derived radiomics and absorbed dose.²¹

Ever since the first implementation of ⁹⁰Y SIRT in clinical studies, T/N tissue ratios have been used to estimate optimal absorbed doses, but primarily using predictive dosimetry based on ^99mTc-MAA to mimic ⁹⁰Y distribution.² However, in some cases, there can be large discrepancies between these distributions and using ^99mTc-MAA to predict ⁹⁰Y distribution is still an approximation.³ Some groups have concluded that ^99mTc-MAA is not a good predictor of ⁹⁰Y distribution at all.^22,23 If carefully performed, we believed, that ^99mTc-MAA is useful in predicting ⁹⁰Y distribution. However, the final ⁹⁰Y distribution can only be confirmed by post-therapy imaging such as bSPECT, or the superior PET imaging. In our experience,²⁴ the main source of ^99mTc-MAA and ⁹⁰Y distribution mismatch is attributed to differences in catheter positioning. The role of the interventional radiologist is essential in catheter positioning and to avoid stealing artery branches at the critical bifurcations. If the ⁹⁰Y activity is not delivered to the correct area, such as the case shown in Figure 4, the T/N ratio is close to 1.0, or even lower. In such situations, thorough follow-up is required, and repeated treatment may be necessary. According to our records, the patient whose images were shown in Figure 4 responded to a second ⁹⁰Y treatment. However, the results of the second treatment were excluded from the study in order to omit interference between⁹⁰Y treatments.

Here, we are using a personalized dosimetry approach based on quantifiable imaging, rather than using semi-quantitative, vendor-designed dosimetry methods. In addition, post-therapy quantitative PET/CT imaging, showing true ⁹⁰Y distribution, was used, providing accurate dosimetry values for total liver, tumours, and normal tissue. This method allows for more accurate calculations of T/N dosimetry ratios than those obtained using ^99mTc-MAA images and/or semi-quantitative dosimetry approaches.

The limitation of our approach includes the exclusion of PR and SD cases, due to their limited number. Few PR and SD cases prevented any meaningful statistical analysis. Therefore, we only compared the two ends of the spectrum, i.e. CR and PD cases, omitting intermediate ones. The total number of PD cases, which was eight, was also a limiting factor. Lastly, the predictive value of T/N in ⁹⁰Y treatment of unresectable HCC can be improved by applying partial volume corrections for lesions smaller than 2.5 cm¹³ **and respiratory motion corrections for lesions in superior hepatic lobes.²⁵

Conclusion

T/N is a parameter that enables the assessment of the ⁹⁰Y SIRT procedure, providing that the tumour and normal tissue received the recommended dose. According to our results, we conclude that patients with PD have a statistically lower T/N dose ratio than those with CR. For patients with low T/N ratios, close to or less than 1.0, thorough follow-up is required, and repeated treatment may be necessary. Further investigation is warranted because the number of PD cases was limited, partial volume effect was not considered and motion correction was not applied for lesions in superior hepatic lobes.