CT contrast reveals where injected cancer drugs actually go inside tissue

Injecting drugs directly into tumors sounds deceptively simple: place a needle inside the cancer, push in the therapeutic, and let a highly concentrated dose do its work while sparing the rest of the body. In practice, the fate of that injected fluid is anything but predictable. Pressurized liquid can tear open microscopic channels in tissue, leak into blood vessels or lymphatics, and spread far beyond the intended target. A team of researchers at the National Institutes of Health has now devised a way to watch this process unfold, using a common CT contrast agent as a visible stand-in for a drug that would otherwise be invisible to clinical imaging. Their findings, published in CVIR Oncology, offer some of the clearest evidence yet that how a drug is injected can matter as much as the drug itself.

The study addressed a fundamental blind spot in interventional oncology. Most therapeutic molecules cannot be seen on CT, MRI, or ultrasound, so clinicians injecting drugs directly into tumors have no way of knowing whether the agent stayed in the lesion or escaped into surrounding tissue. This uncertainty is compounded by the physics of interstitial injection: fluid pushed into tissue at clinically practical rates generates pressures that can exceed tumor interstitial pressure by an order of magnitude, driving convective flow away from the needle tip. When that pressure becomes great enough, it can fracture the tissue architecture, creating preferential pathways along which drug escapes rather than distributing uniformly. Until now, there have been no standardized technical guidelines governing needle selection, injection volume, or infusion rate for intratumoral delivery.

The researchers’ strategy was elegant in its simplicity. They co-injected iodixanol, a widely used iodinated CT contrast agent, together with fluorescently labeled albumin, which served as a surrogate for a real drug. Fluorescent albumin was chosen specifically because its size approximates that of LMB-100, an antibody-toxin conjugate being tested against mesothelioma in a clinical trial. Because iodixanol is radiopaque, its distribution could be mapped in three dimensions by CT within two minutes of injection, while the fluorescent albumin could be visualized afterward with fluorescence microscopy of thinly sectioned frozen tissue. Comparing the two distributions side by side would reveal whether a contrast agent visible on a clinical scanner can reliably predict where a therapeutic protein actually travels.

The physical chemistry of the two agents, however, turned out to be strikingly different. Dynamic light scattering measurements showed that iodixanol has a hydrodynamic diameter of just 2.7 nanometers, compared with 17.0 nanometers for the fluorescent albumin, a statistically significant difference. Their surface charges diverged as well: iodixanol carried a nearly neutral zeta potential of 2.3 millivolts, while the albumin was strongly negative at minus 17.7 millivolts. These differences matter enormously in the confined geometry of the interstitial space, where a molecule’s size determines whether it can weave between cells and collagen fibers, and its charge determines whether it binds to or repels components of the extracellular matrix.

When the team injected mixtures of the two agents into blocks of ex vivo bovine liver at volumes of 1, 2, or 4 milliliters, delivered at 1 milliliter per minute, the consequences of those physicochemical differences became immediately apparent. Iodixanol spread through the tissue over a volume roughly seven times larger than that occupied by the fluorescent albumin. The contrast agent’s concentration was highest at the center of the injectate, matching the concentration loaded into the syringe, and then declined gradually with distance, a profile suggestive of diffusive permeation through the interstitial gel. The albumin behaved differently: it accumulated in highly concentrated pockets, sometimes exceeding the injected concentration at the distribution’s core, and then dropped off sharply at a well-defined boundary, as though the large, negatively charged protein had been physically restrained by the tissue matrix and by binding to collagen and positively charged proteins.

Despite these differences, the two distributions were strongly correlated. The distribution volumes measured by CT and by fluorescence microscopy tracked each other with a coefficient of determination of 0.89, and a leave-one-out analysis confirmed that this relationship was robust, with correlation values ranging from 0.65 to 0.98 across all subsets of the data. In other words, while the contrast agent cannot be treated as an exact replica of the drug, its CT footprint provides a meaningful, quantifiable estimate of where a similarly sized therapeutic will end up. That is a powerful tool for preclinical development and for planning clinical injection protocols, because it converts an invisible process into something a radiologist can measure in real time.

Perhaps the most clinically consequential finding concerned injection volume. When the researchers normalized the distribution volume by the injected volume, the average value stayed roughly constant across all three injection sizes, but the variance ballooned with larger volumes. One-milliliter injections produced tightly clustered, predictable distributions, whereas 4-milliliter injections showed wildly variable spread, with contrast visibly leaking out of the tissue blocks in some cases. The fluorescent albumin, meanwhile, stayed remarkably localized at all volumes, with normalized distribution volumes averaging just 0.12. The implication is clear: dividing a total dose into multiple small-volume injections at different sites, a strategy known as dose fractionation, minimizes local fluid pressure, reduces the risk of tissue fracture and leakage, and yields far more predictable drug localization than a single large bolus.

This benchtop insight translated directly into the clinic. The same team applied a fractionation strategy in a Phase 1 trial testing intratumoral injection of LMB-100, an anti-mesothelin immunotoxin, combined with the immunotherapy ipilimumab in patients with mesothelioma. Rather than delivering the full 4-milliliter dose at one site, clinicians used a multi-pronged needle with three curved injection tines that deploy from the sides of the main cannula, extending up to a 5-centimeter array diameter. The tines were advanced in incremental stages, partially retracted, and the base needle was rotated 60 degrees before redeployment, allowing small aliquots to be deposited across the entire tumor volume. Ultrasound provided real-time visualization of tine position and could detect extravasation the moment it occurred, prompting the operator to halt injection at that site. PET-CT fusion imaging identified metabolically active tumor regions to target.

The broader significance of this work extends well beyond one immunotoxin. The imageable surrogate methodology could be used to optimize needle selection, injection spacing, infusion rates, and drug concentrations for any locally delivered therapy, from oncolytic viruses to nanoparticles to gelling polymer depots. It could also underpin the development of the standardized technical guidelines that intratumoral injection currently lacks, replacing intuition with quantitative data about how injectate properties and injection parameters interact. The authors caution that their experiments were performed in normal ex vivo liver, which lacks the vascular clearance, lymphatic drainage, and elevated interstitial pressures of living tumors, and that differences in spatial resolution between CT and fluorescence microscopy introduce some uncertainty. Larger, more charged drugs such as doxorubicin or oncolytic viruses would likely deviate even further from the contrast agent’s distribution than albumin did.

Even with those caveats, the study marks a turning point in how direct injection into tumors can be understood and controlled. It demonstrates that contrast co-injection can serve as an imageable proxy for drug distribution, that physicochemical properties like size and charge govern how far a molecule travels through tissue, and that small, multifocal injections outperform large single boluses in predictability. For a field where treatment success has long depended on where a needle tip happens to deposit its payload, making the invisible visible may prove to be one of the most practical advances in image-guided drug delivery in years.

Subject of Research: CT-based mapping of drug distribution after direct interstitial injection in ex vivo liver tissue

Article Title: Mapping drug distribution using CT imaging following direct tissue injection in ex vivo liver: informing clinical implementation

Article References: Morhard, R., Mauda-Havakuk, M., Delgado, J. F., Kassin, M. T., Ghafoor, A., Pastan, I., Hassan, R., Karanian, J. W., Pritchard, W. F., Wood, B. J., & Mikhail, A. S. (2025). Mapping drug distribution using CT imaging following direct tissue injection in ex vivo liver: informing clinical implementation. CVIR Oncology, 1(1), Article 27. https://doi.org/10.1007/s44343-025-00027-x

Image Credits: AI Generated

DOI: 10.1007/s44343-025-00027-x

Keywords: intratumoral injection, CT contrast, iodixanol, drug distribution, image-guided delivery, dose fractionation, fluorescent albumin, immunotoxin LMB-100, mesothelioma, interventional oncology, tissue fracture, ex vivo liver