Comments on Assays Used to Measure VEGF in the Presence of Anti-VEGF Therapeutics

Many clinical studies have sought to assess the systemic inhibition of vascular endothelial growth factor-A (VEGF-A) due to systemic exposure to anti-VEGF therapeutics following intravitreal administration and ocular clearance, due to its potential safety implications.¹ These anti-VEGF molecules, the majority of which are not in complex with ocular VEGF, can bind to and neutralize systemic VEGF. Typically, VEGF in either serum or plasma samples from anti-VEGF–treated patients is measured with a sandwich enzyme-linked immunosorbent assay, with the R&D Systems kit (Minneapolis, MN, USA) being the most commonly used. This assay uses reagents that bind to the anti-VEGF binding sites, and thus cannot measure VEGF when it is bound to anti-VEGFs. This measured “unbound” or “free” VEGF is the biologically active species and therefore the most physiologically relevant for evaluating systemic safety risks of anti-VEGFs. It is for this reason that investigators were most interested in measuring “free” VEGF when assessing the impact of systemic anti-VEGF derived from ocular clearance. Results from these studies have shown decreased systemic VEGF levels for some of the anti-VEGFs that correlated with the levels of anti-VEGF.^2–4 

The validity of results from the R&D Systems kit VEGF assay and of another commonly used VEGF assay has been questioned in a recent publication in IOVS.⁵ The main criticism is that it can be readily demonstrated, as we ourselves have done and as noted earlier, that anti-VEGF therapeutics interfere with VEGF measurements in samples where both are present. Takahashi et al.⁵ concluded that caution must be exercised when interpreting these data, a concern we share, but not for the same reasons. For us, the caution is 2-fold. Investigators must (1) understand which form of VEGF (free, drug-bound, or both) they are interested in measuring and why, and (2) must ensure that the chosen assay can deliver the desired type of measurement. 

Additionally, we have concerns regarding some of the methodology used by Takahashi et al.⁵ In particular, these authors conclude that measured VEGF concentrations using the assay kits were lower than calculated theoretical values, and thus suspect. The calculated values are determined from the concentration of VEGF and anti-VEGF (known values in their experiments) and literature-derived dissociation constant (K_D) values, using standard mass action equations. Therein lies the problem. The authors failed to cite a highly relevant paper⁶ that clearly demonstrated errors in the methodology used by one source of K_D values, and the method and format of such experiments dramatically impact the results obtained. Therefore, whichever value one chooses to use, it must be recognized that it may or may not be appropriate for the calculations done by Takahashi et al.,⁵ who then use those calculated results as a benchmark against an actual measured concentration. This K_D-related concern is perhaps best illustrated by the data Takahashi et al. describe when comparing measured VEGF in the presence of similar concentrations of ranibizumab and aflibercept. These two anti-VEGFs showed similar free VEGF levels despite the K_D values being roughly 100-fold different based on their literature source, concluding their assay measurements were in error. In contrast, we feel this supports the conclusions in the Yang et al.⁶ paper that affinity values experimentally determined are method dependent and may not reflect what is occurring in vivo or even in an in vitro assay system different from that used to determine K_D.⁶ 

In fact, we feel that, if anything, these assays will tend to overestimate free VEGF levels, for the following reason. VEGF/anti-VEGF complexes in samples that are diluted into an assay will tend to dissociate in proportion to the amount of sample dilution, and are dependent on the kinetics of complex dissociation and assay duration. This is a consequence of the law of mass action, which describes all reversible binding events. Thus, a diluted sample will tend to have more free VEGF than present in undiluted (“neat”) samples, the latter being a truer estimate of in vivo free VEGF. For the anti-VEGFs, this effect is fairly negligible, because all are very tight binding with slow off-rates relative to the assay duration. 

In conclusion, we believe that this anti-VEGF interference in VEGF measurement is a desired feature of the commonly used VEGF assays (e.g., R&D Systems Quantikine kit)—not a flaw. Our reasoning is that the unbound form of VEGF is the only biologically relevant form to consider when assessing either exposure–response or exposure–safety relationships. It is also a valid approach for assessing the relative effectiveness of VEGF inhibition as a surrogate marker of efficacy.⁷