ABSTRACT

PURPOSE

We aimed to evaluate whether the perfusion pattern from pretreatment hepatic artery perfusion scintigraphy (HAPS) in patients with liver malignancies can predict response to selective internal radiation therapy (SIRT).

METHODS

This retrospective study analyzed 152 consecutive patients treated with yttrium-90 (90Y) resin microspheres between April 2015 and July 2017. HAPS using single-photon emission computed tomography/computed tomography (SPECT/CT) with 99mtechnetium macroaggregated albumin (99mTc-MAA) was performed before SIRT. Investigators visually classified perfusion patterns of tumors as heterogeneous or diffuse in HAPS. Between diffuse and heterogeneous pattern group, positron emission tomography/computed tomography (PET/CT) and magnetic resonance imaging (MRI) were performed in third and sixth month after SIRT, and tumor response assessed and compared by using RECIST 1.1 or mRECIST. Overall survival (OS) and progression-free survival (PFS) were also compared with Kaplan-Meier/log-rank analyses.

RESULTS

Of 216 SIRT procedures, 172 were classified as heterogeneous and 44 as diffuse. Diffuse 99mTc- MAA uptake was associated with longer median OS than heterogeneous (22.2 vs. 14.4 months, respectively; P = .047). Subsegmental infusion was associated with longer OS than either lobar or segmental infusion (P = .090). Mean estimated OS was longer in patients with hepatocellular carcinoma (HCC) (34.2 months) than with colorectal carcinoma (CRC) (16.4 months) (P = .044). Patients with both diffuse and heterogeneous patterns were able to show complete response after SIRT. No statistically significant differences were observed between perfusion patterns and PFS or response rates to SIRT.

CONCLUSION

Although tumor perfusion patterns from preplanning HAPS analyses are useful for estimating tumor uptake of 90Y, they may not reliably predict hepatic treatment response, as patients with different perfusion patterns can show clinical response to SIRT.

Selective internal radiation therapy (SIRT) with yttrium-90 (⁹⁰Y)–labeled microspheres has shown promising results in the treatment of patients with unresectable primary and metastatic liver cancer.^{1, 2} SIRT takes advantage of the dual blood supply of liver; while the normal liver mostly receives blood from the portal vein, liver tumor neovascularization mainly originates from the hepatic artery. Targeted administration of ⁹⁰Y microspheres through the hepatic artery can deliver high doses of radiation to hepatic tumors with minimal exposure to healthy liver parenchyma.³ Hemodynamic flow within the hepatic arterial system and tumor vasculature affect microsphere distribution, as well as activity and delivered dose of ⁹⁰Y.⁴

Evaluating intratumoral heterogeneity with functional molecular imaging can predict the behavior of malignant tumors and eventual treatment response. Heterogeneity of uptake among metastatic lesions is influenced by factors such as cellular hypoxia, necrosis, vascularization, and proliferation—all which affect treatment response.^{5, 6} Assessment of tumor perfusion is an essential preparatory step in SIRT to predict ⁹⁰Y microsphere distribution. A pretreatment planning angiography maps the hepatic and tumor vasculature and identifies any potential off-target perfusion so that offtarget ⁹⁰Y uptake can be prevented with prophylactic vessel embolization.⁷ Subsequent hepatic artery perfusion scintigraphy (HAPS) with ^99mtechnetium macroaggregated albumin (^99mTc-MAA) single-photon emission computed tomography/computed tomography (SPECT/CT) imaging is performed to simulate ⁹⁰Y microsphere injection. This imaging 1) detects and quantifies pulmonary shunting, 2) identifies extrahepatic uptake, and 3) determines the tumor-toliver uptake ratio (TLR).^8-11 TLR is defined as the blood flow distribution ratio between the tumor and normal liver compartments and is an index of tumor-targeting effectiveness.¹² Because the size of ^99mTcMAA particles is comparable to that of ⁹⁰Y microspheres (25-35 microns), they should be distributed similarly when injected into the hepatic arterial vasculature and liver tissue, though differences can exist. This association can indicate the relative density of the hepatic microvascular structure and identify the associated areas of deposition.^{3, 9} Therefore, distribution of 99m Tc-MAA activity is a measure of expected 90Y microsphere activity distribution and thus estimated radiation dose received.

Studies investigating the value of assessing pre-SIRT ^99mTc-MAA perfusion patterns for predicting ⁹⁰Y distribution/dosimetry and tumor response had mixed results.^13–15 One study described a semiquantitative imaging approach involving positron emission tomography/computed tomography (PET/CT), SPECT, and CT scans in patients with liver metastases from colorectal cancer (CRC). Using a cutoff value of 1 for the ^99mTc-MAA TLR, the authors predicted the post-SIRT metabolic response with a sensitivity of 89%, specificity of 65%, positive predictive value of 71%, and negative predictive value of 87%.¹³ Another study in patients with metastatic CRC used magnetic resonance imaging (MRI) to evaluate the association between 99m Tc-MAA uptake and response to SIRT and found no significant correlation. Because patients with low ^99mTc-MAA uptake showed significant response to therapy, the authors advised against excluding patients from SIRT who lacked scintigraphic ^99mTc-MAA accumulation.¹⁴ A retrospective analysis that assessed ^99mTc-MAA uptake with HAPS in 80 patients with mixed tumor diagnoses reported a correlation between type of perfusion pattern and tumor response, although response was the sole measure of SIRT efficacy.¹⁵

Considering that tumor vascularization and the tumor-to-normal arterial blood flow ratio may vary considerably among patients and tumor types, heterogeneous treatment responses and perfusion patterns are expected. Varied tumor vasculature may result in heterogeneous or diffuse HAPS perfusion patterns.¹⁶ Hypervascular tumors are expected to concentrate ⁹⁰Y microspheres more efficiently and respond better to SIRT than hypovascular tumors. However, high tumor vascularization indicated by HAPS may not necessarily be a prerequisite for successful SIRT, as an inverse correlation between ^99mTc-MAA accumulation and patient survival has been reported. Extensive tumor vascularization may be accompanied by particularly aggressive tumor biology that manifests regardless of tumor dose.¹⁶ Thus, a reliable model to predict SIRT efficacy remains elusive. We hypothesized that heterogeneous perfusion of ^99mTc-MAA may reflect tumor heterogeneity, which might lead to different treatment response than diffuse perfusion reflecting tumor homogeneity. This study investigated the associations between tumor perfusion visualized with preparatory HAPS SPECT/CT imaging, patient survival, and tumor response to SIRT assessed using fluorine-18 fluorodeoxyglucose (¹⁸F-FDG) PET/CT and abdominal MRI in a large population (152 patients and 216 SIRT procedures). We also considered the role of tumor vascularity and method of SIRT administration.

Methods

Patients from a single institution treated with ⁹⁰Y resin microspheres (SIR-Spheres^®, Sirtex Medical Limited) between April 2015 and July 2017 were retrospectively analyzed. Eligible patients had unresectable liver-dominant disease, adequate hepatic reserve (albumin > 3.0 g/dL, total bilirubin < 2mg/ dL), and Eastern Cooperative Oncology Group (ECOG) performance status of 0-2. All patients provided written informed consent for SIRT. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. For this type of study, formal consent is waived. This study was approved by Ethical Committee of our instution (31.08.2018/80).

All patient cases were reviewed prior to treatment by an interdisciplinary tumor board that included a medical oncologist, gastroenterologist, interventional radiologist, radiation oncologist, surgeon, and nuclear medicine expert. Before SIRT, all patients underwent routine laboratory analyses, including a complete blood count and tests for liver and renal function and coagulation. Levels of disease-specific tumor markers were measured for primary liver malignancies; markers included alphafetoprotein (AFP) for hepatocellular carcinoma (HCC) and carbohydrate antigen 199, AFP, and carcinoembryonic antigen for cholangiocellular carcinoma (CCC).

Pretreatment planning angiography and HAPS SPECT/CT

HAPS SPECT/CT was performed in all patients following the standard pretreatment hepatomesenteric angiography and C-arm cone-beam CT to assess tumor vasculature.⁸ The ^99mTc-MAA (Pulmocis^®, IBA Molecular) (10-90 μm particles) was prepared immediately before the scan and injected into a hepatic artery branch feeding the target region through a percutaneous catheter. The injection activity, concentration, and final volume were 148 MBq (4 mCi), 100,000 particles/mL, and 4 mL, respectively. SPECT/CT imaging acquisition was started within 45 minutes of the ^99mTc-MAA injection. Oral sodium perchlorate was administered before the procedure to avoid formation of ^99mTc-MAA degradation–related artifacts that could potentially generate false activity.¹⁷ Uptake of ^99mTc-MAA was captured with a dual-head gamma camera equipped with low-energy, ultra-highresolution collimators (Symbia T16, Siemens Healthcare). SPECT acquisition parameters were peak energy 140 keV (window: 20%), step-and-shoot protocol, and 25 s/projection at 256 × 256 matrix. CT acquisition parameters were 10 mm axial sampling, 140 kpV, 2.5 mA, and 256 × 256 matrix size.

Tumor and treatment area volumes were calculated from angiogram and SPECT/CT fusion images using OsiriX MD software (Pixmeo SARL), which allowed investigators to manually select regions of interest (ROI). Total tumor volume for each patient was calculated as the sum of the volumes of all selected lesions. Treatment area volume corresponded to ^99mTc-MAA distribution area, defined as the sum of the tumor area and the normal parenchyma area with 99m Tc-MAA uptake. In necrotic tumors, volume of the area of low uptake was not analyzed.

HAPS images were evaluated visually by one of the authors, and tumor perfusion pattern was classified as heterogeneous or diffuse (Figure 1). In diffuse patterns, MAA particles are distributed almost uniformly within the treatment area, including the tumor, with no hypoactive areas. In contrast, heterogeneous patterns show a lessuniform distribution, often with hypoactive areas.

SIRT and postprocedure treatment response evaluation

The partition model was used to calculate ⁹⁰Y activity and the dose to administer to each patient.^{18, 19} SIRT was performed in accordance with standard methods.⁸

In accordance with procedure standards, 90Y microspheres were infused through the hepatic artery using three different approaches: lobar, egmental, or subsegmental (via direct tumor-feeding vessels).²⁰ Subsegmental therapy was performed in patients with solitary lesions to allow ⁹⁰Y microsphere delivery. Segmental infusion was performed in patients with a tumor (either primary or metastatic) localized in a distinct liver egment supplied by a single segmented branch of the hepatic artery, which was identified and catheterized. In patients with bilobar disease,  ach lobe was treated sequentially at 4- to 6-week intervals to maintain a degree of functional capacity in the contralateral lobe. These patients ere assessed 4 weeks after the first ⁹⁰Y microsphere administration to decide whether to proceed with the second treatment.

Postprocedure ⁹⁰Y-bremsstrahlung SPECT/ CT scan was performed with medium-energy all-purpose collimators (energy window, 160 keV ± 30%; matrix, 128 × 128; 6°steps; 40 s/ frame) and reconstructed using an iterative method. Attenuation correction was applied to both ^99mTc-MAA and ⁹⁰Y SPECT/CT.

After SIRT, all patients were prescribed a proton pump inhibitor and steroid, analgesic, and antinausea drugs. Because ⁹⁰Y is a pure beta emitter, isolation for radioprotection was not necessary after treatment. Although the radiation hazard presented by patients to other people is very low, traces of 90Y can be detected in urine; all patients received complete instructions about radiation safety precautions.

Patient follow-up

To evaluate therapy response 3 and 6 months after SIRT, patients underwent hepatocyte-specific, contrast-enhanced MRI (Ingenia 3.0T MR System, Philips; and Signa HDi EchoSpeed 1.5T, GE Healthcare) and ¹⁸F-FDG PET/CT (Biograph 6, Siemens Medical Systems). The findings acquired with third and sixth months imagings were compared with initial ¹⁸F-FDG PET/ CT and contrast-enhanced MRI according to therapy response criteria.

The European Organization for Research and Treatment of Cancer criteria (EORTC) and Response Evaluation Criteria in Solid Tumors (RECIST) were used to assess tumor response to SIRT in ¹⁸F-FDG PET/CT images based on baseline-chosen, lesion-specific ROIs observed on each scan.²¹ The RECIST version 1.1 or modified RECIST criteria were used to evaluate tumor response in MRI images of hypovascular and hypervascular tumors. Hypervascular tumors have increased intensity on arterial- or portalphase images on dynamic contrastenhanced MRI.

Because distant metastases can occur during follow-up despite a primary tumor response, multiple response endpoints were considered. Objective response was evaluated only in treated lesions, whereas overall response was analyzed in both hepatic and extrahepatic lesions. Hepatic response was evaluated in all lesions in both liver lobes, regardless of treatment with SIRT.

Statistical analysis

Analyses were performed using the IBM SPSS Statistics Solution, version 20.0 (IBM) software package. Progression-free survival (PFS) and overall survival (OS) were calculated from the first HAPS procedure date for all patients, considered as the baseline date. Categorical variables were expressed as numbers and percentages, and continuous variables as mean and standard deviation or median and range, where appropriate. The chi-square test and MannWhitney U test were used to compare categorical and continuous variables, respectively. For comparison of categorical variables, Fisher exact test was applied if the expected cell count was small. For survival analysis, OS was calculated using the Kaplan-Meier method, and the Log rank test was used to compare Kaplan-Meier curves. The threshold for statistical significance for all tests was P< .05.

Results

Overall, 216 HAPS/SIRT procedures were performed in 152 patients. Table 1 represents the characteristics of patient group according to the liver perfusion patterns in HAPS (Table 1). Patients had an ECOG performance status of 0 (n = 127, 83.6%) or 1 (n = 25, 16.4%). Twenty-four SIRT procedures were performed in patients with HCC and portal vein tumor thrombosis (PVTT); follow-up data were available for 18/24 patients. The HAPS pretreatment planning measures were determined and stratified by perfusion pattern (Table 2).

Four patients were not treated because they had LSF > 20% and were excluded from analyses. Overall, 14 lobar, 92 segmental, and 110 subsegmental ⁹⁰Y infusion SIRT procedures were performed. To increase the tumor-to-liver dose ratio, 12/92 patients who received radiation segmentectomy underwent protective coil embolization of the subsegmental branches supplying the normal parenchyma before SIRT. No patients showed any extrahepatic leakage during SIRT. We also performed right lobar treatment in 6 patients in whom the tumor covered almost the entire right lobe.

The mean interval between HAPS and SIRT was 11.7 ± 6 days (0-34), and mean follow-up time after SIRT was 12.1 ± 7.8 months (0.5-32.8). Twelve (7.9%) patients died before 3 months, and 32 (21%) patients died between the 3-month and 6-month follow-ups.

Heterogeneous and diffuse ^99mTc-MAA uptake patterns were observed in 172 (79.6%) and 44 (20.4%) procedures, respectively. Estimated median OS for the heterogeneous pattern subgroup was 14.4 months (95% confidence interval [CI], 10.48-18.32) and for the diffuse pattern subgroup was 22.2 months (95% CI, 13.96-30.44) (P = .047) (Figure 2a). No statistically significant differences were observed between perfusion pattern and PFS (P = .743) (Figure 2b).

Median OS estimated for the three different infusion methods was 12.6 months (95% CI, 6.5-18.8) for lobar, 12.7 months (95% CI, 9.8-15.6) for segmental, and 19.1 months (95% CI, 14.3-23.9) for subsegmental (P = .090) (Figure 2c). When the OS of the two most frequently encountered tumor types was analyzed, median OS could not be estimated for CRC because fewer than half of patients with CRC had died before the end of the follow-up period. We therefore compared the mean OS between the two groups, which was 34.2 months (95% CI, 26.9-41.5) in patients with HCC and 16.4 months (95% CI, 13.1-19.7) in patients with CRC (P = .044) (Figure 2d).

Tumor vascularity analysis of MRI images of all 216 procedures revealed 138 (63.9%) hypervascular and 78 (36.1%) hypovascular perfusion patterns (Table 3).

The majority of procedures (172/216) were associated with heterogeneous perfusion patterns. Diffuse pattern was least common with lobar treatment and most common with subsegmental treatment (Table 3). The percentage of solitary lesions (rather than multifocal) was higher in patients who underwent subsegmental (45.5%) and lobar (42.9%) procedures compared to segmental procedures (18.5%).

Of the 152 patients, 137 (90%) and 88 (56%) had imaging data available at the 3and 6-month follow-ups, respectively. Diffuse and heterogeneous perfusion patterns were associated with different response rates. The diffuse pattern was associated with a significantly higher complete response (CR) rate than the heterogeneous pattern with both PET/CT and MRI, and at both the 3- and 6-month assessments (P < .05 for all comparisons; Figure 3). MRI evaluations at 3 months showed a higher partial response (PR) rate in patients with a heterogeneous pattern than in those with a diffuse pattern (58.6% vs. 40.0%; P = .038; Figure 3c).

Overall, no statistically significant differences were observed between heterogeneous and diffuse perfusion patterns for objective, overall, and hepatic responses rates (CR + PR + stable disease [SD]) (Figure 3). Despite receiving higher ⁹⁰Y radiation activity (33.0 ± 15.4 vs. 28.5 ± 13.3 mCi), patients showing heterogeneous HAPS perfusion patterns had lower tumor-absorbed doses than patients with diffuse patterns (191.1 ± 70.4 vs. 250.6 ± 78.8 Gy, respectively; Table 2). This difference in lower absorbed dose was statistically significant (P < .001).

Of 18 analyzed patients with HCC and PVTT, 1 had CR (based on both PET/CT and MRI results), 13 had PR, 2 had SD, and 2 had progressive disease (not shown).

Hypervascular tumors were significantly associated with higher objective, overall, and hepatic response rates than hypovascular tumors, as observed in both 3- and 6-month PET/CT and MRI follow-up results (P < .05 for all but one comparison; Table 4).

Figure 4 shows an exemplary case of a patient with CRC who showed a heterogeneous perfusion pattern with a complete metabolic response at 3 months after SIRT. A bremsstrahlung image acquired after SIRT (Figure 4d) demonstrates a discrepancy with the HAPS image acquired before SIRT (Figure 4c).

Fatigue, pain, nausea, vomiting, and fever were the most common adverse effects after treatment and were experienced in most patients to varying degrees. No grade 3 and 4 toxicities were observed at the 3-month follow-up. In addition, no patients demonstrated any features of radiation-induced liver disease. No pulmonary toxicity was reported.

Discussion

In this retrospective analysis, we evaluated the association between ^99mTc-MAA uptake patterns observed in pre-SIRT HAPS scans and posttreatment OS, PFS, and tumor response assessed with MRI and ¹⁸F-FDG PET/ CT scans. We examined the correlation between perfusion pattern and factors such as 90Y infusion method and tumor vascularity.

We found that a diffuse perfusion pattern was significantly associated with longer median OS than a heterogeneous pattern (P = .047). However, no other statistically significant association between perfusion and treatment efficacy (ie, PFS and objective, overall, and hepatic response rates) was observed. This suggests that although perfusion of ⁹⁰Y may influence treatment outcome, it is not the only factor determining treatment response. Our results are consistent with a previous retrospective analysis in 80 patients with primary and metastatic liver tumors,¹⁵ which found that diffuse or heterogeneous ^99mTc-MAA HAPS perfusion patterns were a good predictor of early response to SIRT. The authors reported that a diffuse pattern was associated with better tumor response.¹⁵ However, OS and PFS were not evaluated, and the body surface area (BSA) formula was used to calculate ⁹⁰Y activity and administered dose. The commonly used BSA formula postulates that BSA correlates with the size of each patient’s liver and tumor burden, but it neglects variabilities in tumor vascularity and tumor-to-normal ratio among patients.^{18, 19} In this study, we calculated dose and activity of ⁹⁰Y using the partition model, a more accurate and personalized method based on Medical Internal Radiation Dose principles. This model hypothesizes that three distinct vascular compartments (lungs, tumor, and uninvolved liver parenchyma) can be partitioned during ⁹⁰Y microsphere infusion.^{18, 19} We considered the partition model a more appropriate method to calculate ⁹⁰Y activity and dose according to the predicted microsphere distribution based on ^99mTc-MAA HAPS scans.

Although heterogeneity of tumors has been demonstrated to predict treatment failure and drug resistance,⁵ our results included many instances in which patients with heterogeneous perfusion patterns, such as patient case 1 (Figure 4), showed positive responses to SIRT. We believe that patients should not be disqualified for SIRT solely based on perfusion pattern shown during preplanning HAPS analyses.

The largest subpopulation in the study had HCC. The benefits of SIRT in patients with HCC and PVTT have been demonstrated in two randomized controlled trials and one large retrospective analysis, in which higher tumor response rates (CR or PR), PFS and time to progression, and OS were compared in patients treated with SIRT or sorafenib.^22–24 In our study, we showed that SIRT is an effective treatment option in patients with HCC and tumorrelated PVTT, classified as advanced HCC Barcelona Clinic Liver Cancer stage C. Irrespective of tumor HAPS perfusion pattern, 16/18 patients with HCC and PVTT responded to SIRT.

Cumulative OS was almost twice as long in patients with HCC versus those with CRC (Figure 2d). This may be because patients with HCC who are candidates for superselective embolization naturally have better prognoses than those with metastatic refractory CRC who have already undergone several previous cycles of chemotherapy. Likewise, subsegmental treatment resulted in longer median OS than lobar or segmental treatment, although the difference was not statistically significant (P = .090) (Figure 2c). Livers undergoing lobar infusion might be more likely to show heterogeneous distribution because they are more likely to contain multifocal tumors, whereas a subsegmental injection would most likely generate a diffuse pattern. Patients who undergo subsegmental infusion likely had a solitary lesion at baseline, whereas a larger number of segments were likely involved when lobar or segmental infusions were required. Indeed, the percentage of solitary lesions was higher in patients who underwent subsegmental treatment versus segmental treatment (45.5% vs. 18.5%). As a result, it is natural to detect diffuse perfusion pattern and to obtain better prognosis rates with subsegmantal infusions due to lower stage disease and smaller tumor dimension. Therefore, this can be evaluated as selection bias for study group.

Despite receiving a higher ⁹⁰Y radiation dose than patients with a diffuse perfusion pattern, patients with a heterogeneous perfusion pattern showed a lower tumorabsorbed dose (Table 2). This may be because lesions that showed a heterogeneous pattern were more multifocal than those with a diffuse pattern and were treated more favorably using lobar or segmental (≥1) ⁹⁰Y SIRT instead of superselective subsegmental treatment. The rest of the dose may be absorbed by the normal parenchyma between multiple lesions.

The type of HAPS perfusion pattern is not the only parameter that affects clinical outcomes of SIRT; other factors to consider include the relationship between tumor biology and vascularity, functional integrity of uninvolved hepatic parenchyma, and relative radiosensitivities of the tumor and normal liver. The radiation dose administered, tumor hemodynamics, and vascularity are among the main factors that affect tumor response to SIRT.⁴ Although tumor vascularity and degree of ^99mTc-MAA uptake in pre-SIRT HAPS images cannot entirely explain SIRT efficacy, we reported significantly higher response rates in hypervascular tumors than in hypovascular tumors at both 3- and 6-month posttreatment time points and with both PET/CT and MRI imaging modalities.

Radiation segmentectomy is often used to treat liver tumors with unfavorable location or large size that prevents resection or ablation.²⁵ Radiation segmentectomy minimizes the volume of normal tissue exposed to radiation and maximizes the tumordelivered ⁹⁰Y dose. A dose restriction for normal liver parenchyma is not required when using this method. There is little published evidence on ⁹⁰Y microspheres and radiation segmentectomy. In our study, 93 patients underwent radiation segmentectomy, and all showed good treatment responses, even in patients with heterogeneous perfusion.

SIRT has been shown to effectively downstage patients with different primary tumors or liver metastases for subsequent transplantation²⁶ or surgical resection.²⁷ In our study, SIRT was successfully used as a bridge to transplantation in three patients with HCC, and as a bridge to resection in two patients with HCC. SIRT was effective in downstaging tumors for resection and transplantation in patients with diverse tumor perfusion patterns.

We evaluated intratumoral heterogeneity visually from MAA perfusion patterns and classified each pattern as heterogeneous or diffuse. Some other interesting quantitative approaches are available to measure heterogeneity in metastatic tumors. For example, FDG metabolism can be used to generate a heterogeneity index by measuring the maximum or average standardized uptake value (SUVmax or SUVmean).²⁸ In our study, ¹⁸F-FDG PET/CT facilitated the selection of patients for surgical resection or liver transplantation, though it was only used after SIRT. In addition to radiological images of the abdomen that evaluate treatment response after locoregional therapy of liver tumors, whole-body scans with ¹⁸F-FDG PET/CT can also detect extrahepatic metastases and inform treatment decisions, such as addition of systemic therapy. Distant metastases can occur during follow-up despite a primary tumor response, even in patients with HCC, which has a low incidence of extrahepatic metastasis.²⁹ This occurred in 3 of 68 patients with HCC. Additionally, two patients with liver tumor progression after SIRT developed extrahepatic metastases. 18F-FDG PET/CT uptake may be able to identify patients with HCC who are more likely to have a poor prognosis.

Post-therapeutic dosimetry with bremsstrahlung SPECT/CT is a good complementary evaluation to confirm MAA findings. Because it corresponds to the real distribution of the therapeutic agent, it may more accurately reflect the true dose and can be used to define tumoral threshold doses and liver maximal tolerated dose. However, both the heterogeneous distribution of microspheres within the intrahepatic vasculature and the anatomic attenuation of the broad energy spectrum of secondary bremsstrahlung x-ray radiation could limit the value of post-therapy bremsstrahlung scans in evaluating microsphere localization. Resulting HAPS images and bremsstrahlung images may be discordant, and dosimetry based on pretreatment images may not reflect the true exposure after SIRT. In our study, a discrepancy between imaging methods was observed in several patients, as demonstrated in patient case 1 (Figures 4c and 4d). The slight difference in particle sizes between ^99mTc-MAA and ⁹⁰Y microspheres, differences in catheter tip position and radiopharmaceutical reflux to peripheral vascular structures causing imaging artifacts during pretreatment planning angiography and SIRT might also contribute to calculation inconsistencies.

There are some limitations in the study. First, it is a nonrandomized retrospective study and there is no internal control group. Second, it is single-center study. Although our hospital is a tertiary care research hospital, it reflects the results of only one center. Third, our study group consists of patients of various disease stages and pathological types. Therefore, it has a heterogeneous cohort. Lastly, due to being a tertiary care hospital, the medical data of some patients who were referred from other institutions were absent.

In conclusion, tumoral lesions characterized by diffuse ^99mTc-MAA uptake in HAPS SPECT/CT images were associated with longer median OS after ⁹⁰Y SIRT than tumors with heterogeneous uptake, though no significant differences were observed in tumor response rates between subgroups with diffuse and heterogeneous perfusion. Despite a statistically significant difference in OS between patients with diffuse and heterogeneous patterns, the number of patients with diffuse pattern was relatively small (44/216, 20.4%). Thus, SIRT remains an option regardless of MAA perfusion pattern. Nevertheless, assessment of tumor perfusion may still be useful to evaluate response to therapy. Further analyses are needed to better understand the full relation between these parameters and tumor biology.