A Clinical Dosimetric Perspective Uncovers New Evidence and Offers New Insight in Favor of ^99mTc-Macroaggregated Albumin for Predictive Dosimetry in ⁹⁰Y Resin Microsphere Radioembolization

TO THE EDITOR: At first glance, the results of a recent study by Wondergem et al. (1) may appear discouraging for the evolving science of personalized predictive dosimetry for ⁹⁰Y radioembolization, especially to less experienced readers. However, the dosimetric implications of their data may be interpreted more favorably in support of the use of ^99mTc-macroaggregated albumin (MAA) predictive dosimetry in clinical practice.

Based on 28 procedures among 22 patients deemed to have optimal agreement on catheter tip positions between ^99mTc-MAA and ⁹⁰Y-resin microsphere injections, Wondergem et al. found the mean difference in liver segment volume-of-interest radioconcentration to be −0.026 MBq/cm³, with an SD of the differences of 0.2837 MBq/cm³ (1). Their data showed wide 95% limits of agreement that, at the outset, seemed to suggest ^99mTc-MAA to be a poor surrogate to simulate the postradioembolization biodistribution of ⁹⁰Y-resin microspheres. This may be too stringent a requirement. For a procedure as technically complex as ⁹⁰Y radioembolization, it may instead be more practical and clinically meaningful to consider the dosimetric implications within ±1 SD of the differences, that is, 68% limits of agreement.

To illustrate this point, let us take a typical patient from the authors’ dataset: a patient with inoperable chemorefractory colorectal liver metastasis without chronic hepatitis, less than 25% liver involvement by tumor, undergoing whole-liver ⁹⁰Y-resin microsphere radioembolization (1). We assign the following typical parameters for this patient: tumor mass of 200 g, nontumorous liver mass of 1,500 g, and a modestly favorable mean tumor-to-normal liver (T/N) ratio of 2. Central to this dosimetric example is the partition model formula for calculating the mean T/N ratio (2), which is mathematically independent of the extent of hepatopulmonary shunting. The tumor mean absorbed dose may be expressed as Equation 1, [D_mean × (m_T + m_L)]/[m_T + (m_L/TNR)], where D_mean is the whole-liver mean absorbed dose averaged across tumorous and nontumorous liver, m_T is the tumor mass, m_L is the nontumorous liver mass, and TNR is the mean T/N ratio.

By partition modeling, let us aim to deliver intended mean absorbed doses to tumor and nontumorous liver of 120 Gy and 60 Gy, respectively, in keeping with current radiation planning guidelines (3). From Equation 1, this translates into an intended D_mean of 67 Gy for this patient. Assuming a normal distribution of data and using a ⁹⁰Y mean absorbed dose conversion factor of 49.7 Gy per MBq/cm³ (1), we now apply the results provided by Wondergem et al.: mean difference in segmental volume-of-interest radioconcentration, −0.026 MBq/cm³; SD of the differences, 0.2837 MBq/cm³ (1). The actual D_mean is now corrected to 65.7 Gy, with its lower and upper 68% limits of agreement at 51.6 and 79.8 Gy, respectively. Applying the latter 2 figures back into Equation 1, we can expect 84% of patients to receive an actual tumor mean absorbed dose of more than 92 Gy, sufficient to achieve at least stable disease for several months or possibly a slight response (4). Similarly, we can expect 84% of patients to not exceed an actual nontumorous liver mean absorbed dose of 71 Gy, within recommended limits for the avoidance of radiomicrosphere hepatotoxicity (3). The converse is true: only a minority, that is, 16%, of patients may be at risk of significant tumor under-dosing or inadvertent radiomicrosphere hepatotoxicity. Considering the general complexity of ⁹⁰Y radioembolization, most physicians will find these treatment odds favorable and consistent with best practice in the modern era of personalized medicine.

A conservative mean T/N ratio of 2 was used in this example of colorectal liver metastasis. Most patients have higher, more favorable mean T/N ratios (4), which enable deliberate escalation of the intended tumor mean absorbed dose beyond 120 Gy when within safety limitations to the nontumorous liver and lung. Hence, many patients can achieve better dosimetric results than presented in this example. Equation 1 has an infinite number of possible dosimetric scenarios, which the reader is encouraged to explore. A thorough understanding of the interplay between mean T/N ratios, intended mean absorbed doses, tissue masses, and hepatopulmonary shunting is paramount for safe and effective predictive dosimetry by partition modeling (2,4).

It has been common knowledge for years that ^99mTc-MAA is an imperfect surrogate for ⁹⁰Y-resin microspheres (5), and no study has claimed otherwise. ^99mTc-MAA should be regarded as a tool, and the usefulness of any tool is only as good as its user and the complexity of the task at hand. For basic predictive dosimetry, partition modeling can be performed using a pocket calculator (2), and SPECT/CT (4) is now widely available to replace planar ^99mTc-MAA scintigraphy. For advanced predictive dosimetry, affordable and increasingly powerful computers and software can rapidly generate dose–volume histograms from ^99mTc-MAA SPECT/CT data (6). Correlation of ⁹⁰Y SPECT or ⁹⁰Y PET dose distributions with ^99mTc-MAA and newer microspheres as they appear will add further confidence in the utility of the treatment planning procedure, but for the moment, it is reasonable to proceed with ^99mTc-MAA. Today, the major barrier to the routine application of predictive dosimetry for ⁹⁰Y radioembolization is no longer the state of the art but rather the state of our hearts.

• © 2013 by the Society of Nuclear Medicine and Molecular Imaging, Inc.