Radiation-associated hepatic injury has been the major hindrance in the treatment of liver malignancies using external beam radiotherapy. The incidence of potentially lethal radiation hepatitis is approximately 75% with doses in excess of 40 Gy to the whole liver using conventional external beam radiation treatment (RT) . Therefore the tumoricidal dose is difficult to achieve using external beam RT. Conformal and stereotactic radiation therapy techniques can be used to deliver much higher radiation doses for treatment of focal disease; however, since hepatic metastases are most often multifocal and irregular in shape and may replace large parts of the liver volume, only a minority of patients are optimal candidates for such therapies.
The therapeutic ratio with SIRT, compared to external beam RT, is significantly improved, and the tumor absorbed doses from SIRT are typically 4 to 6 times higher than those to the liver tissue . The tumoricidal dose with SIRT has not been clearly determined, and the liver absorbed doses, clinically judged to be tolerated well, are mostly based on coarse dosimetric estimates. Although the clinical efficacy of SIRT in producing objective responses, with a relatively wide margin of safety, has been clearly demonstrated in numerous studies [4, 5, 13–16, 25], a safe liver absorbed dose has not been defined. The liver absorbed doses reported in the literature range from 34 to 181 Gy [25–29]. This variability is primarily due to the differences in the dosimetric methodologies adopted. It is further accentuated with the heterogeneity in the available retrospective data that involves different disease types with different underlying functional liver abnormalities and previous treatments.
Microspheres are deposited in the liver as a number of discrete clusters, rather than point sources that are homogeneously distributed throughout the tumor. This irregular but non-random clustering of microspheres produces a highly heterogeneous radiation dose distribution pattern. Microsphere distribution in and around tumors is complex. Microspheres lodge preferentially in the growing rim of the tumor. The central portion of lesions, often times hypoperfused and sometimes necrotic, receive a much smaller number of microspheres, if any. The variability in the dose distribution with Y-90 microsphere treatment has been emphasized in a number of studies [24, 30, 31]. The uncertainty of dose deposition is very unsettling for both conventional brachytherapy, and the MIRD methodology of nuclear medicine.
Microspheres distribute within tumor and liver compartments. The differential deposition of microspheres favoring the tumor compartment is expressed as the TLR. The "selectivity" of the Y-90 microsphere treatment is determined primarily by the TLR, the biologic determinants of which include the angiogenic potential and pattern of the tumors, and the growth rate-dependent tumor perfusion. The TLR, in our study, did not correlate with tumor volume, liver volume, or percent tumor involvement. There was no correlation between the TLR and tumor type within the observed range of 2.8 to 15.4. The TLR had a linear relationship with the tumor absorbed dose with an equation of y = 15.02x + 29.06 and an R-squared of 0.24.
Currently, MAA scans are performed routinely prior to SIRT for the assessment of lung shunting and extrahepatic uptake. This is widely recognized to be important for minimizing the risk of potentially debilitating adverse events such as radiation pneumonitis or radiation gastritis/duodenitis. MAA particles are of similar size and density as resin microspheres (MAA particles have diameters in the range 15 to 30 microns and a density of 1.3 g/cc and SIR-Sphere microspheres have diameters of 35 ± 5 microns and a density of 1.6 g/cc). Although the number of particles for MAA and resin microsphere suspensions (in a unit of administered volume) is significantly different (microspheres are concentrated 10-fold) and the potential effect of this difference on the distribution kinetics has never been studied, it is reasonable to conclude that the two particles will be distributed similarly within the liver and the tumor compartments. Several studies have demonstrated the value of the MAA scan in planning of a successful SIRT [32–34]. The role of the MAA scan was challenged by Dhabuwala et al. in a recent paper which aimed to determine whether the pattern or degree of MAA uptake by liver tumors after hepatic arterial injection might be a predictor of response to subsequent SIRT with resin microspheres . The authors were not able to demonstrate a correlation between MAA uptake in the CRC liver metastases prior to SIRT and response to the treatment determined by CEA, CT, or survival time. However, this study suffered serious methodological problems. The MAA was injected 1 to 2 hours prior to the SIRT. The SIRT was administered to the patients after injecting angiotensin 2 into the hepatic artery, whereas the MAA perfusion scans were performed without angiotensin 2. The effects of the incomplete clearance of MAA on subsequent Y-90 microsphere injection and the application of angiotensin 2 pharmacological manipulation to the therapy session but not to the MAA administration confounded the results. The use of simple planar images for the quantitative element of the study brought a significant degree of imprecision to the measurement of the MAA uptake as well.
The theory of SIRT is based on a favorable TLR. MAA imaging is a good measure of TLR, despite its significant technical limitations. A high resolution SPECT acquisition and processing is essential in MAA image interpretation and quantitation. The avascular core of the lesions needs to be excluded from the activity calculation during quantitative image processing for improved accuracy. The MAA directly shows the site and the amount of anticipated radiation dose deposition. The vascularity of different tumor types has been studied extensively. A vascularity grade as assessed by an arteriogram may not necessarily predict the TLR measured from an MAA scan. The wide range of TLR for many types of tumors also indicate that, based on the histologic type, the targeting efficiency of Y-90 microspheres (or MAA) may not be correctly estimated. The individual TLR is determined by the stage and the pattern of angiogenesis, and also by synchronous events of apoptosis and necrosis. There are no scientific grounds for precluding any particular type of liver malignancy for SIRT, nor is there a credible way of predicting favorable tumor targeting without MAA imaging. The correlation between the TLR and clinical outcome may be more difficult to establish as the clinical response is determined by many other factors. The radiosensitivity of the tumor (intrinsic or microenviromental), therapeutic threshold, and the tumor growth rate-to-necrosis/apoptosis ratio could play significant role in producing objective tumor responses.
The second important observation in our study was related to the relationships between tumor and liver absorbed doses and administered activity. The tumor absorbed dose did not correlate with the administered activity. However, there was a linear correlation between the liver absorbed dose and the administered activity.
The estimation of the tumor absorbed dose with Y-90 microspheres using MIRD macrodosimetry approach is problematic. The non-uniformity of microsphere distribution is not only a function of irregular tumor angio-architecture, but also is complicated with flow-bound distribution physics. Although the total dose delivered to the tumor compartment can be determined reasonably accurate, the dose deposition pattern in the different zones within a given tumor mass does challenge the limits of macrodosimetry. The angiogenic belt at the tumor-normal tissue boundary has the highest and relatively uniform microsphere distribution. A full radiation dose deposition occurs in this zone while an inner zone receives an attenuated dose from the far range beta emission of the former. The radiation deposition within deeper tumor regions is drastically lower than that at the periphery.
The inhomogeneity in the distribution of the microspheres in the liver tissue is to a lesser scale when compared to the distribution of microspheres in the tumor. Radiation dose to healthy liver parenchyma is determined by the number of microspheres administered, the specific activity of the microspheres, and the distance between the microspheres implanted. The liver tissue receiving the highest dose is that which is immediately surrounding the tumor. The radiation injury to this area of parenchyma is more pronounced. There is an abrupt drop in the radiation dose towards the surrounding liver a few millimeters from the high dose zone around the tumor . The remainder of the liver receives less radiation than would be predicted from assuming a homogeneous distribution of radiation dose throughout the parenchyma. Fox et al. have demonstrated that almost 90% of the liver tissue received less than the dose predicted by assuming uniform distribution, and a third of the tissue received less than one-third of the predicted dose . This partially explains the lack of classical clinical radiation hepatitis at much higher dose levels as estimated by MIRD methodology. The linear relationship between administered activity and liver dose observed in our study indicates that MIRD estimates could be clinically useful in defining safe liver dose limits. The application of microdosimetry techniques will improve the reproducibility of the radiation dose calculations. This clearly requires higher resolution functional diagnostic imaging.
The third set of observations in our study was related to the dose response in tumor and liver. Objective tumor response was seen in more than 50% of all patients. There was no difference in mean absorbed tumor dose estimates amongst the disease types. The differences in response could perhaps be explained by variations in radiosensitivity. This result could also be partly due to the differences in the techniques used to assess the treatment response. FDG-PET response usually is observed before a volume change is appreciated on CT imaging . Similarly, tumor markers and/or octreoscan changes might be more sensitive indices of objective response for NET compared to CT imaging alone. A concrete conclusion regarding the differences in objective response amongst disease categories may not be drawn with a restricted sample size.
There was no classical clinical radiation hepatitis or therapy-induced liver failure at liver absorbed doses up to 99 Gy. In 63% of the patients, there were no significant LFT changes and in 18%, the initial LFT abnormalities were improved in 2 to 4 weeks. A grade 1 to 2 rise in LFT values was not uncommon following treatments associated with angiographic evidence of stasis at the end of the microsphere administration.
The liver response (toxicity) to radiation appears to be different in SIRT than in external beam RT. This is mainly due to the lower dose rates maintained by virtue of the selectivity of Y-90 microsphere treatment. The differences in the severity and pattern of liver toxicity could be related to the radio-patho-biologic differences in both techniques. The pathogenesis of radiation damage from external beam RT to the liver is dominated by vascular injury in the centri-lobular region. External beam radiation causes alterations in the central veins such as eccentric wall thickening and subsequent obliteration of the lumen leading to veno-occlusive disease . Radiation from microspheres, however, is deposited primarily in the portal triads. The relative segregation of the central vein from the radiation source was suggested to be important in minimizing the toxicity due to venoocclusive disease . However, the toxicity related to the radiation effects on the biliary tract and distal portal tributaries should be approached with due caution.