Matts Kågedal, Cheikh Diack, Katie Maass, Angelica Quartino
Genentech INC
Objectives: Ranibizumab is indicated for treatment of neovascular Age related Macular Degeneration (nAMD). According to the US Prescribing Information (USPI), ranibizumab 0.5 mg is recommended to be administered by intravitreal injection once a month. In practice however most physicians adjust the dosing interval based on the observed response in patients. Pro re nata (PRN, “as needed”, monthly evaluations with treatment only if disease activity is present) or treat and extend (T&E, progressive extension of treatment intervals up to 12 weeks depending on the clinical findings) are commonly used. The real world evidence suggests that the result is not always optimal [1]. Ideally individualization of treatment should be based on factors known at baseline. This would allow for more patients receiving optimal dose interval faster. A PKPD model for best corrected visual acuity (BCVA) has previously been developed that assesses the impact of baseline covariates age, leakage size, central retinal leakage thickness, presence of cysts and choroidal neovascularization (CNV) type on disease progression and drug response [2]. Currently, there is no easy way to understand how these covariate effects could be used to individualize dosing interval.
A simulation study based on the previously developed model was performed to predict the potential benefit of individualized dosing and to identify patients that may tolerate longer dosing intervals with maintained vision.
Methods: Simulations were performed to predict vision loss with increasing dose interval and to relate this loss to baseline covariates in the model. To this end, virtual patients (N ~1000) were generated based on baseline covariates from a previously performed study in nAMD. Parameter values for each virtual patient were derived based on covariates and simulated random effects. Longitudinal BCVA time curves (IPREDs) were generated for different dosing intervals for each virtual patient. Similarly the expected BCVA-time curves based on fixed effects only (PREDs) were generated for each virtual patient. The loss of vision when reducing dose frequency from monthly to every four months (Q4M) was evaluated at month eight. A baseline meta-covariate corresponding to the loss of BCVA at month 8 with Q4M versus QM dosing was derived based on the PRED-curve. This meta-covariate hence reflected the expected joint effect of baseline covariates in the model. Similarly for each virtual patient the individual loss of vision with Q4M versus QM dosing was determined from the IPRED curve. The correlation between the meta-covariate and the individually determined loss of vision was then assessed. The impact of reduced dosing frequency was assessed based on the risk for ≥4 letters lower BCVA versus monthly dosing.
Results: For patients with the highest expected loss in BCVA with reduced dosing frequency (lowest quartile of meta-covariate), 53% of the patients were predicted to have ≥4 letters lower BCVA with Q4M treatment relative to QM. The corresponding risk for patients predicted to tolerate less frequent dosing (highest quartile of meta-covariate) was 27%. Baseline leakage size (BLEA) was the individual covariate that was most correlated with the meta-covariate.
Conclusions: The meta-covariate reflecting the joint effect of all baseline covariates in the model or BLEA may be used to predict which patients that are expected to benefit from more frequent dosing and which patients that can go longer between treatments. Further validation of the model with respect to predicting durability of treatment response is needed.
References:
[1] Holz FG, et al. Multi-country real-life experience of anti-vascular endothelial growth factor therapy for wet age-related macular degeneration Br J Ophthalmol 2015;99:220–226.
[2] Diack C, Mazer N A, Schwab D. An individually matched virtual ranibizumab treatment arm in neovascular age-related macular degeneration. Poster, ARVO 2019.
Reference: PAGE () Abstr 9279 [www.page-meeting.org/?abstract=9279]
Poster: Methodology - Covariate/Variability Models