The actual dose of radiation absorbed by tumor tissue cannot be calculated after 90Y-RE. It depends on the activity administered, which can be measured, but also on different parameters that cannot be determined, from the exact positioning of the catheter to the evolving arterial hemodynamics over the injection period or the vessel density inside the different tumors of the same patient. All these factors provide differences in the biodistribution of microspheres and in the absorbed doses of radiation in different areas of the treated tissue. In practice, dosimetry is based on a simulation with macroaggregated albumin labeled with technetium 99 (^99mTc-MAA). It is performed in order to calculate lung shunting from a planar scintigraphy, and may be used to estimate the average dose that will be delivered to the non-tumoral liver and to the tumor from planar or SPECT images, calculating the tumor/non-tumor ratio. Although the ability of ^99mTc-MAA images to predict Y90 activity distribution and dosimetry is far from ideal, the majority of series that explore retrospectively the dosimetry based on ^99mTc-MAA images have shown a dose–response relationship. Lau et al. (1994) first reported that an objective response by combined WHO and alpha-fetoprotein criteria occurred in higher proportion (7/8) of patients in which every tumor absorbed more than 120 Gy when compared to those (1/8) in which at least one tumor received an estimated absorbed dose of less than 120 Gy. Survival was also better in the former group (median survival: 55.9 vs. 26.2 weeks). More recently, in a study of 36 patients treated with 90Y-RE, a higher estimated dose delivered to the tumor was associated with the appearance of radiological response according to EASL criteria and with overall survival, and a threshold for tumor absorbed dose of 205 Gy enabled response prediction with a high accuracy (Garin et al. 2012). Finally, in a dosimetric analysis of 65 tumor lesions, EASL response correlated with the dose absorbed by target lesions (Spearman’s r = 0.60, 95 % CI: 0.41–0.74, p < 0.001). Lesions lacking objective response received a median dose of 275 Gy, while responding tumors were found to absorb a median of 490 Gy. An efficacy threshold of 500 Gy significantly predicted the observed objective response and limited to 20 % the rate of nonresponders (AUC = 0.78) (Mazzaferro et al. 2013). Nevertheless, it should be emphasized that a consistent cut-off value that ensures a tumor response has not been reported, and that no prospective study has confirmed these observations. All these data support the current search for innovative treatment planning based on tumor/non-tumor dosimetry methods applied to ^99mTc-MAA SPECT as pretreatment prediction of efficacy.

It has been recently proposed that changes in levels of alpha-fetoprotein after locoregional therapy can be considered a good method for assessing tumor response and survival, as well as an early objective screening tool for progression by imaging (Sherman 2010). Alpha-fetoprotein response has also been explored after 90Y-RE, as it is described in Table 1. With different cut-off values for elevated alpha-fetoprotein (40 and 400 ng/ml), alpha-fetoprotein response defined as more than 50 % decrease from baseline was observed in 30–89 % of patients (Lau et al. 1998; Gaba et al. 2009; Riaz et al. 2009; Iñarrairaegui et al. 2010b; Mazzaferro et al. 2013). Time to alpha-fetoprotein response was 2.7 months (CI 95 % 2.3–4 months) and alpha-fetoprotein response correlated with overall survival (p < 0.001) (Riaz et al. 2009).

The predominantly arterial vascularization of liver tumors provides the basis for the aim of delivering tumoricidal doses of radiation to liver tumors while sparing the non-tumoral tissue. However, in 90Y-RE the non-tumoral liver tissue also absorbs a certain dose of radiation. The effects that this radiation produces in the liver are largely unknown but may translate into a variety of changes in liver function tests, morphological changes such as liver atrophy, and clinical syndromes including radioembolization-induced liver disease (REILD) that is covered in another chapter of this book.

Having described the effects of 90Y-RE in the tumor and the liver, we will review in the next pages the available evidence supporting the role of 90Y-RE in the different stages of HCC using the Barcelona Clinic Liver Cancer (BCLC) classification, that has been endorsed by the two main Western Hepatology societies, EASL (European Association For The Study Of The Liver; European Organisation For Research And Treatment Of Cancer 2012) and the American Association for the Study of Liver Diseases (AASLD) (Bruix and Sherman 2011). All the evidence comes from retrospective series or non-controlled prospective studies (Lau et al. 1994; Geschwind et al. 2004; Salem et al. 2005, 2006; Lewandowski et al. 2009; Kulik et al. 2006, 2008; Iñarrairaegui et al. 2010a, b), without any randomized controlled trials comparing 90Y-RE with other available therapies. However, in the last 3 years four large series have provided valuable information in the different stages of HCC, and may allow to put the outcomes in the perspective of other treatments, mainly in terms of survival (Salem et al. 2010; Hilgard et al. 2010; Sangro et al. 2011; Mazzaferro et al. 2013) (Table 2). Besides, some of these series provide with data on toxicity (Table 3) and on surrogate endpoints that are likely to predict clinical benefit, as objective tumor response rates and time to progression. Despite this growing body of evidence and due to the lack of randomized controlled trials, 90Y-RE is not recommended in the proposed algorithms of treatment in the above mentioned EASL and AASLD guidelines (European Association For The Study Of The Liver; European Organisation For Research And Treatment Of Cancer 2012; Bruix and Sherman 2011). However, its use is considered in others guidelines, as the European Society for Medical Oncology (ESMO) (Jelic et al. 2010) and the National Comprehensive Cancer Network (NCCN) (Benson et al. 2009) and has been a matter of debate and editorials in one of the most notable journal of the specialty lately (Kulik 2010; Bolondi and Piscaglia 2013).