The Myth of Multimodality Diagnostic Agents

Abstract

I F BUZZWORDS WERE MEASURED in terms of grant dollars awarded, ‘‘multimodality’’ would rank right up there with ‘‘nanotechnology.’’ There are two primary motivations for multimodality imaging. One is to exploit the best features of two (or more) different imaging technologies and to use coregistration to enhance image interpretation or quantitation. Positron emission tomography–computed tomography (PET-CT) and PET–magnetic resonance imaging (PETMRI) are great examples. The second is to create diagnostic agents composed of two ‘‘effector domains,’’ each providing contrast for a different modality. By doing so, the reasoning goes, one creates an ‘‘internal gold standard’’ by which one effector domain, for example, a fluorophore, is validated using a quantifiable second effector domain, for example, a radiotracer. The former motivation for multimodality imaging makes sense. The second does not. Let me start by stating that there is no such thing as a multimodality diagnostic agent. Why? The problem is one of relative scale or, rather, ‘‘effect size.’’ MRI requires tens to hundreds of micromolar of agent to change the relaxivity of surrounding water molecules (eg, Gd), to induce a chemical shift (eg, PARACEST agents), or to generate signal above background (eg, hyperpolarized agents). Fluorescence-based optical imaging, even in the nearinfrared (NIR), requires nanomolar to micromolar concentration of agent. And PET and single-photon computed tomography (SPECT) require picomolar concentration of radiotracers for detectability. Thus, it is impossible to create a single agent with a molar equivalent of each domain. Consider, for example, a small molecule–targeting ligand conjugated covalently to an NIR fluorophore and a radiometal-chelated 1,4,7,10-tetraazacyclododecane1,4,7,10-tetraacetic acid (DOTA). If one tries to create a single molecule with an equimolar concentration of each effector molecule, which requires a final concentration equal to the least sensitive imaging modality, the radioactivity would be enormous. So what is typically done is to use a mixture of agents, one with a radioactive metal and one with a cold metal. However, this violates the tracer principle and dilutes the specific activity of the agent to levels that are typically used in a blocking experiment! Consider, then, a Gd-based MRI agent combined with an NIR fluorophore. The problem is the same, only worse. When used at the concentration required for MRI contrast, the fluorophore will be quenched, that is, possibly undetectable fluorescence emission. Nor do multimodality diagnostic agents make any sense from the standpoints of biodistribution and clearance and clinical translation. Adding a second effector domain to a targeted diagnostic agent creates a ‘‘new chemical entity,’’ which requires independent regulatory approval. And, depending on the overall size of the final molecule, the second domain will often dominate biodistribution and clearance, making the final image more a reflection of the second effector domain than the first effector domain being validated. Doesn’t the size of the diagnostic agent matter? Not really. Violation of the tracer principle and ‘‘self-blocking’’ occur whenever PET or SPECT radiotracers are combined with optical imaging or MRI, and quenching of fluorescence occurs whenever optical imaging is combined with MRI, regardless of whether the diagnostic agent is based on a nanoparticle, an antibody, or a small molecule. Multimodality nanoparticles are especially problematic, in fact, owing to the narrow range of hydrodynamic diameters ($ <5.5 nm) that permit clearance from the body. Finally, the same problems hold true for newer, single-chemical entities that are composed of multiple effector domains. From the Department of Radiology, Harvard Medical School, and the Center for Molecular Imaging, Beth Israel Deaconess Medical Center, Boston, MA.