PET-based compartmental modeling of (124)I-A33 antibody: quantitative characterization of patient-specific tumor targeting in colorectal cancer

Abstract

Purpose: The molecular specificity of monoclonal antibodies (mAbs) directed against tumor antigens has proven effective for targeted therapy of human cancers, as shown by a growing list of successful antibody-based drug products. We describe a novel, nonlinear compartmental model using PET-derived data to determine the "best-fit" parameters and model-derived quantities for optimizing biodistribution of intravenously injected (124)I-labeled antitumor antibodies.

Methods: As an example of this paradigm, quantitative image and kinetic analyses of anti-A33 humanized mAb (also known as "A33") were performed in 11 colorectal cancer patients. Serial whole-body PET scans of (124)I-labeled A33 and blood samples were acquired and the resulting tissue time-activity data for each patient were fit to a nonlinear compartmental model using the SAAM II computer code.

Results: Excellent agreement was observed between fitted and measured parameters of tumor uptake, "off-target" uptake in bowel mucosa, blood clearance, tumor antigen levels, and percent antigen occupancy.

Conclusion: This approach should be generally applicable to antibody-antigen systems in human tumors for which the masses of antigen-expressing tumor and of normal tissues can be estimated and for which antibody kinetics can be measured with PET. Ultimately, based on each patient's resulting "best-fit" nonlinear model, a patient-specific optimum mAb dose (in micromoles, for example) may be derived.

Figure 1

Coronal PET images of a patient with colorectal cancer metastatic to liver at 2 and 7 days after injection of ¹²⁴I-huA33 mAb. The images shown are maximum-instensity projections (MIPs). The 2-day image shows two high-uptake foci in the liver (single arrow) and splenic flexture (double arrow) corresponding to known metastic lesions and some uptake in bowel. At 7 days, there is persistent uptake in the liver metastases and prominent uptake in normal bowel (triple arrow). For this patient, the model-derived A33 concentrations in tumor and normal bowel were 500 and 20 nM, respectively, and the maximum A33 occupancies 10 and 1%, respectively. These occupancy values are at the low end of the range for our patient cohort (See Table 1.)

Figure 2

Non-linear compartmental model of systemically administered anti-A33 mAb in patient with A33-expressing tumors. The square brackets indicate concentrations (in M), the indices 1, 2, and 3 correspond to plasma, normal bowel, and tumor, respectively (as indicated), k(i,j) is the fractional exchange rate of anti-A33 mAb to compartment i from compartment j (in /h), Q _i is the amount of anti-A33 mAb in compartment i (in mole), V_i is the volume of compartment i (in l) (equivalent to the mass of tissue i in kg), k(0,i) is the rate of elimination of ¹²⁴I-mAb from compartment i (in /h), Flux (i,j) is the absolute exchange rate of anti-A33 mAb to compartment i from compartment j (in mole/h), and k_a is the association rate constant for the binding of the anti-A33 mAb to A33 (in /M/h).

Figure 3

Time-activity data %ID/g versus time post-injection) for plasma, A33-expressing colorectal tumor, and normal colon following intravenous injection of ¹²⁴I-A33 into a colorectal cancer patient. The points represent the measured data and the curves the “best-fit” model-derived data obtained using the model in Figure 2. For this case, the best-fit model-derived tumor and normal-colon A33 concentrations were 2.0 and 4.5 nM, respectively, and the antibody-antigen association rate constant 6.5×10⁻⁹ /M /hr.