Pharmacokinetics of Ranibizumab in Patients with Neovascular Age-Related Macular Degeneration: A Population Approach

Abstract

Purpose.: To characterize ranibizumab pharmacokinetics in patients with AMD.

Methods.: A population approach of nonlinear mixed-effect pharmacokinetic modeling based on concentration-time data from 2993 serum samples from 674 AMD patients enrolled in 5 phase 1 to 3 clinical trials of single or multiple intravitreal (ITV) doses of ranibizumab (0.3–2.0 mg/eye) administered biweekly or monthly for up to 24 months.

Results.: A total of 696 concentration-time records from 229 subjects with one or more measurable total serum ranibizumab concentrations were analyzed. The systemic concentration-time data for ranibizumab were best described by a one-compartment model with first-order absorption into and first-order elimination from the systemic circulation. Vitreous elimination half-life (t_1/2) was calculated to be 9 days and the intrinsic systemic elimination t_1/2 was calculated to be approximately 2 hours. Following ITV administration, ranibizumab egresses slowly into the systemic circulation, resulting in an apparent serum t_1/2 of 9 days. Systemic-to-vitreous exposure ratio was estimated to be 1:90,000. With monthly and quarterly ITV regimens, the serum concentrations of ranibizumab at steady-state for both the 0.3 and 0.5 mg/eye dose levels were estimated to be below the range needed to inhibit VEGF-A–induced endothelial cell proliferation in vitro by 50% at all times.

Conclusions.: Systemic exposure to ranibizumab after ITV injection was very low due to elimination on reaching systemic circulation from the vitreous. Population pharmacokinetic analysis of data from a representative sample of AMD patients did not identify clinically significant sources or correlates of variability in ranibizumab exposure. (ClinicalTrials.gov numbers, NCT00056836, NCT00056823.)

Introduction

Ranibizumab (Lucentis, Genentech, Inc., South San Francisco, CA) is a humanized monoclonal antibody antigen-binding fragment (Fab) engineered to bind with high affinity and to potently inhibit all known biologically active isoforms of VEGF-A. ¹ Intravitreal (ITV) administration of ranibizumab on a monthly or pro re nata basis is associated with significant stabilization and improvement of visual acuity, resolution of macular edema, as measured by retinal thickness, and with associated low rates of serious ocular or systemic adverse events. ^2–7 The Food and Drug Administration (FDA) has approved ranibizumab for the treatment of neovascular (wet) AMD (initial approval: 2006), macular edema following retinal vein occlusion, and diabetic macular edema. ⁸  

Ranibizumab is administered to patients by ITV injection at 30-day intervals or longer. In cynomolgus monkeys and rabbits, the concentration of ranibizumab in the ocular compartment (vitreous, retina, and aqueous humor) decays by 50% approximately every 3 days. ^9–11 Radioactivity from labeled ranibizumab molecules has been detected in all retinal layers, including the retinal pigment epithelium, at 24-hour post-ITV injection. ⁹ The distribution of ranibizumab into the retina and the aqueous humor following ITV injection, as well as the pattern of flow of water between aqueous and vitreous compartments, suggest that this Fab can reach the systemic circulation by either entering the choroid vessels or via the aqueous humor outflow. ^9,10,12 In cynomolgus monkeys, the serum elimination half-life (t_1/2) of ranibizumab (0.5 mg) after ITV administration was 3.59 days. ¹⁰ The elimination was significantly faster (15 hours) when the dose was administered intravenously. ¹⁰ This difference suggests a depot effect in the eye. 

Ranibizumab is injected in the vitreous, which is stagnant and would slow the drug passage. ¹² The slow apparent serum half-life in cynomolgus monkeys after ITV administration (3.59 days) is the result of the rate-limiting slow drug passage in the vitreous and follows flip-flop pharmacokinetic principles ^13–15 such that the apparent slower serum half-life following ITV administration is actually the result of the slow egress from the ocular compartment into the systemic circulation. Additionally, there is no evidence that the drug is metabolized in the intraocular compartment. ¹⁰ Consequently, over a period of several days, the ITV-injected ranibizumab slowly egresses into the systemic circulation where it is subject to systemic catabolism due to absence of FcRn (the neonatal Fc receptor for IgG) protection of a full-length antibody, ¹⁶ and renal filtration and catabolism. ^17,18 This disposition pattern minimizes the systemic exposure over the entire course of treatment. In rabbits and cynomolgus monkeys, the serum concentration of ranibizumab after ITV administration was more than 10,000- and 1500-fold lower than the corresponding vitreous concentration, respectively. ^9,10  

Here, we present the first comprehensive report on the pharmacokinetics (PK) of ranibizumab in patients with AMD based on data from a registrational phase 3 trial and several smaller studies. 

Materials and Methods

This PK analysis in human subjects was conducted in accordance with the Declaration of Helsinki, FDA regulations, and the Health Insurance Portability and Accountability Act. The respective institutional review boards approved the study protocol before the clinical trial was initiated. All enrolled subjects provided informed consent in writing. Compliance information for the clinical trials during which the PK sampling was performed has been published previously. ^6,19–22  

Clinical Trials, PK Sampling, and Bioanalytic Determinations

Serum samples for PK analysis were collected from subjects enrolled in two phase 1, ^21,22 two phase 1/2, ^19,20 and one phase 3⁶ clinical trials of ranibizumab for the treatment of neovascular AMD. In all trials, one eye per subject was selected as a “study eye” to receive ITV ranibizumab as a bolus. Information on the subtypes of choroidal neovascularization (CNV), ranibizumab dosing, and PK sampling schedules are summarized for each study in Table 1. The total concentration of ranibizumab in serum was determined by an electrochemiluminescent assay. ²³ Briefly, ranibizumab molecules were captured by biotinylated recombinant humanized VEGF-A bound to streptavidin-coated paramagnetic beads. A ruthenium-base tag (ORI−TAG) conjugated rabbit anti−rhuFab V2 antibody was used for detection. The intra-assay and interassay precision across concentration range were less than 10% and 20%, respectively. Ranibizumab concentrations lower than 0.3 ng/mL were considered lower-than-reportable (LTR). 

Table 1. 

 

Clinical Studies Included in the Population PK Analysis

Study	Study Details
	CNV Type	Ranibizumab Dose, mg/eye	Dosing Frequency	Concomitant Therapy	Blood Sampling Scheme
Phase 1
Rosenfeld et al.22	All	0.05, 0.15, 0.3, 0.5, and 1.0	Single dose	None	1 hour postdose and on days 1, 7, 14, 42, and 90
Rosenfeld et al.21	All	0.3, 0.5, 0.75, 1.0, 1.25, 1.5, 1.75, and 2.0	Intrasubject escalation (q2w, q4w)	None	1 hour before and after all doses
Phase 1/2
Heier et al.19	All	0.3 and 0.5	q4w	None	1 hour before and after all doses and at weeks 2, 6, and 14
Heier et al.20 (FOCUS)	Predominantly classic	0.5	Monthly for 24 months*	Verteporfin PDT (7 days before treatment)	Days 7 and 14 post first dose and 1 hour predose at 6 months
Phase 3
Rosenfeld et al.6 (MARINA)	Minimally classic or occult	0.3 and 0.5	Monthly for 24 months*	None	On day 30 during months 6 and 12

 FOCUS = RhuFab V2 Ocular Treatment Combining the Use of Visudyne to Evaluate Safety; q2w = once every 2 weeks; q4w = once every 4 weeks.

*  Only data from the first 12 months were included in this analysis.

Population Pharmacokinetic Analysis

The PK data were analyzed using the nonlinear mixed-effects modeling program (NONMEM, Version VII; GloboMax LLC, Hanover, MD) together with the R program for goodness-of-fit assessment and covariate model building (Version 2.10; provided in the public domain by The R Project for Statistical Computing, http://www.r-project.org/). The first-order conditional estimation with interaction method in NONMEM was used throughout the model development. Outline of the population modeling approach, which has been described in detail, ²⁴ is provided in Figure 1. The initial step was to select a structural base PK model that best described the data. Next, demographic, physiologic, and disease-specific covariates, such as age, creatinine clearance (CrCL), CNV type, and concomitant photodynamic therapy (PDT) therapy, were screened to determine if they can explain some of the variability in the observed data (Fig. 1). All covariates adopted their values measured at baseline (i.e., pretreatment). No time-varying covariate was included in this population PK analysis. The missing values were imputed for the covariates to the median of the available values. All covariates that were determined to have statistically significant correlation with ranibizumab PK were incorporated into a final model and a final estimation was conducted. Only observed (measurable) concentration data were used to develop the base and covariate PK models. The model was evaluated by visual predictive check. A nonparametric bootstrap approach was used to estimate the precision of the model parameter estimates and assess the stability of the model (results not reported in this article). ²⁵  

Figure 1. 

 

PK modeling approach. ALK, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase.

 

*Measured by fluorescein angiogram with unit of disc area (DA), where 1 DA was 1500 to 1700 μm.

 

†Estimated using the Cockcroft-Gault equation. ⁴⁰

The effect of LTR samples on the estimate of PK parameters in the base model was assessed by the maximum likelihood (ML) method. ²⁶ LTR serum concentrations immediately preceding or following a measurable serum concentration within 1 dosing period were included in the data set by acknowledging that the true concentration values were somewhere between zero and the lower limit of quantitation (349 samples). The rest of the LTR serum concentrations, including those from subjects with no measurable sample, were treated as missing. 

To understand the population median exposure profiles in serum and vitreous humor after ITV administration, as well as the population spread (i.e., the 5th and 95th percentiles of the exposure profiles), population parameters obtained from the final model were used to simulate 500 serum and vitreous concentration–time profiles of ranibizumab after monthly and quarterly ITV injections of 0.3 and 0.5 mg per eye. The distribution of ranibizumab in the vitreous humor was assumed to be homogeneous, with a vitreous volume of 4 mL. ²⁷ Covariates were obtained for 500 subjects with random sampling (with replacement) from the data set used for developing the final model. The predicted population medians and 5th/95th percentiles of maximum concentration (C_max), minimum concentration (C_min), and area under the concentration-time curve (AUC) at steady state were reported for both vitreous and serum concentration–time profiles. The ratio between simulated vitreous and serum AUC at steady state was calculated. Population simulations of ranibizumab serum and vitreous concentration–time profiles were conducted for the two dosing regimens under investigation: (1) 0.3 and 0.5 mg per eye monthly ITV administration (hereafter referred to as the monthly regimen); and (2) 0.3 and 0.5 mg per eye monthly ITV administration for 3 months followed by quarterly ITV administration (hereafter referred to as the quarterly regimen). 

Results

In total, 696 ranibizumab serum concentration–time records from 229 subjects with at least one measurable serum ranibizumab concentration were included in the analysis (Table 2). In the analysis population, Caucasians accounted for 95%, females were 56%, median age was 78 years, and median weight 76 kg. The CNV type distribution was 45.8% predominantly classic, 35.7% minimally classic, and 18.5% occult. During the treatment period assessed in this analysis, 42% of the subjects received concomitant PDT and 22% received local or systemic IOP-lowering medication. The percentage of observations with imputed values for continuous laboratory assessment covariates were low (<3%) except for CrCL (16%). The median CrCL was 65.22 mL/min. The median values for lesion size, area of CNV, and area of leakage with retinal pigment epithelium staining were 2.9 mm², 2.45 mm², and 3.05 mm², respectively. 

Table 2. 

 

PK Sampling Information

Category	Study
	Phase 1		Phase 1/2		Phase 3	All Studies
	Rosenfeld et al.22	Rosenfeld et al.21	Heier et al.19	Heier et al.20 (FOCUS)	Rosenfeld et al.6 (MARINA)
No. subjects treated with ranibizumab	27	29	62	105	477	700
No. subjects with PK samples collected	27	29	62	105	451	674
No. subjects with ≥1 evaluable concentration	24	29	53	97	26	229
No. total samples from subjects with ≥1 evaluable concentration	140	336	862	830	51	2219
No. evaluable samples* from subjects with ≥1 evaluable concentration	56	244	192	178	26	696†
Evaluable samples/subject‡
Mean	2.3	8.4	3.6	1.8	1.0	3.0
Minimum	1	1	1	1	1	1
Maximum	4	16	11	4	1	16
Percentage of evaluable samples per sampling time point§
1 hr (0–0.5 d)	31.8	78.4	33.8	15.3	83.3	48.0
Day 1 (0.5–4 d)	100	n/a	n/a	92.3	100	97.5
Day 7 (4–10 d)	62.5	n/a	57.1	96.5	100	85.4
Day 14 (11–17 d)	40.9	83.8	33.9	63.2	100	54.8
Day 30 (27–33 d)	n/a	33.3	4.2	2.1	29.4	5.5
Day 42 (39–45 d)	0	0	0	0	0	0
Day 90 (87–93 d)	0	n/a	0	n/a	n/a	0

 n/a, not applicable.

*  Evaluable samples are the samples with measurable concentration. ²⁴

†  Twenty-six records collected from 23 subjects 1 hour after dosing, which was expected to be 10 to 100 times higher than the normal concentration level, were kept in the dataset. The sporadic high concentrations at 1 hour after dosing were limited to a small fraction of subjects/occasions, occurred randomly over the course of treatment, and were not associated with any systemic adverse event. These 26 records could be well described by including the minor elimination pathway in the model. In total, 22 non-LTR records (22/718 or 3.1%) were excluded from the analysis: two subjects (six records) with high concentrations at 1 hour occurred for more than one occasion were excluded before fitting data to the base model, because of the large interoccasion variability that was not considered in the current model. In addition, 16 records from 16 subjects with weighted residuals in NONMEM analysis (|WRES|) > 5) were excluded after fitting data to the base model.

‡  Subjects with at least one measurable serum sample.

§  Time after dose. A time range was assigned to each scheduled sampling time point to account for variability in actual sampling time.

Base Model Structure of Ranibizumab PK after ITV Administration

A one-compartment model with first-order absorption into and first-order elimination from the systemic circulation was chosen as the best base model to describe the systemic concentration–time data for ranibizumab following ITV administration (Fig. 2). An additional minor pathway from vitreous to systemic circulation was built in (not shown in Fig. 2 for simplicity purposes) to account for the rare (∼2%) and sporadic incidences of higher than expected serum concentration immediately after ITV injection. Similar phenomenon has been observed in animal studies with ITV injection. ²⁸ Although the reason is currently unclear, it is conceivable that on rare occasions, a small fraction of the injected drug may have had transient direct access to blood circulation through the needle wound. The fraction was estimated from the data to be 2.35% on average with large interindividual variability. 

Figure 2. 

 

Schematic representation of one-compartment model of the disposition of ranibizumab following intravitreal administration with first-order absorption into and first-order elimination from the systemic circulation. A_vit , amount of ranibizumab in the vitreous compartment; A_sys , amount of ranibizumab in the systemic compartment; CL, clearance; K _a, rate of systemic absorption/rate of vitreous elimination; V _c, apparent volume of the central compartment.

In animal models, the rate of vitreal to systemic absorption of ranibizumab is slower than the rate of systemic elimination. ^9,10 In the absence of human vitreous PK data, it was assumed that the same pattern of faster system clearance and slower systemic absorption applies to humans. The ranibizumab population mean PK parameters from the base model using measurable data thus were estimated to be the following, with the estimated interindividual variability in parentheses: 

•  
  System clearance where “CL” is the clearance of ranibizumab and “F” is bioavailability:

 

 

 

•  
  Volume of distribution where “Vc” is the apparent volume of the central compartment:

 

 

 

•  
  Rate of systemic absorption or rate of vitreous elimination

 

 

 

where Ka is the rate of systemic absorption or rate of vitreous elimination. 

Owing to data censoring in this analysis (i.e., LTR samples were considered as missing), underestimation of CL/F and higher uncertainty in Ka parameter estimations were anticipated. Although the magnitude of CL/F underestimation cannot be fully estimated with current data, efforts were made to include as many of the LTR samples as possible with a statistical method widely adopted in analysis of censored data ²⁹ to evaluate the impact of LTR samples on the parameter estimates. 

Assessment of Base Model Sensitivity to LTR Samples

The population mean systemic clearance estimated from the measurable concentration data versus measurable data plus LTRs was 25.5 L/d vs. 27.3 L/d, respectively (<10% difference). Predicted individual sample concentration and steady-state AUC also showed good correlation between predictions based on observed data only and predictions made by using the ML approach with LTR samples (Fig. 3). Therefore, covariate analysis was conducted using observed data only. 

Figure 3. 

 

Comparison of base model predictions based on observed data only and using the maximum likelihood approach with lower-than-reportable samples. Solid lines are the lines of linear regression with the correlation coefficient r ² reported; dashed lines are the lines of unity.

Final Model and Covariate Effects

The population parameter estimates of the final model and covariate effects are summarized in Table 3. Vitreous elimination half-life is thus calculated to be 9 days (ln2/Ka = 8.6 days); and systemic elimination half life is calculated to be 2 hours (ln2/(CL/F/[Vc/F]) = 0.09 days). 

Table 3. 

 

Population Parameter Estimates for the Final Model

Parameter	Final Estimate (% RSE)
Evaluable subjects, n	229
Evaluable data points	696
Objective function	−457.442
θ1, Typical CL/F, L/d	24.1 (4.52)
θ2, Typical Vc/F, L	3.01 (13.3)
θ3, Typical Ka, d−1	0.0806 (7.33)
θ4, Typical fraction	0.0251 (23.2)
Covariate exponent for CL/F
θ5, for CrCL	0.303 (32.0)
Covariate multiplier for Ka
θ6, for PDT	−0.353 (20.7)
ωCL/F , %	31.4 (22.0)
ω Vc/F , %	68.6 (27.2)
ω Ka, %	21.8 (96.2)
ωFraction, %	88.7 (34.8)
Residual variability
θ7, additive, ng/mL	0.145 (42.8)
θ8, proportional, %	33.4 (8.44)

 Specifically in this case, PDT is verteporfin for injection. RSE, relative SE of the estimates; θ, model parameter/covariate effect/residual variability; ω, intersubject variability as CV%.

CrCL was the most significant covariate for explaining the CL/F interindividual variability (approximately ±15% change at the 5th and 95th percentiles of CrCL), although it explained only 10.5% of the interindividual variance. Reduction in ranibizumab clearance was deemed to be minimal in subjects with renal impairment (Table 4) and was considered clinically insignificant. Age was not statistically significantly correlated with CL/F after correction of CrCL, as the age effect was already reflected in CrCL. 

Table 4. 

 

Individual Post Hoc Clearance Per Renal Function Groups

Renal Function*	Estimated Creatinine Clearance, mL/min	No. Subjects (% of Total)	Ranibizumab Clearance Mean ± SD, L/d
Normal	>80	64 (32.0)	27.0 ± 5.2
Mild impairment	50 to 80	93 (46.5)	24.2 ± 5.0
Moderate impairment	30 to 50	40 (20.0)	22.3 ± 5.7
Severe impairment	<30	3 (1.5)	15.7 ± 1.6

*  Renal functions were grouped per FDA guidance. ²⁵

Subjects with one or more concomitant PDT procedures had 35.3% lower Ka than those who had no PDT while being treated with ranibizumab. The covariate effects in the final model explained 45.8% of the interindividual variance for Ka. Consequently, vitreous elimination half-life was estimated to be slightly longer, approximately 13.3 days, in patients treated with PDT (mean from 100 patients), compared with 9 days in the entire population. 

Goodness-of-fit plots for the final model are presented in Supplemental Figure S1 (see Supplementary Material and Supplementary Fig. S1). The population prediction from NONMEM analysis (PRED) versus dependent variable (DV) plot indicated reasonable population model fits to the observed data. A moderate degree of improvement was observed in the individual prediction from nonlinear mixed-effect model analysis (IPRED) versus DV plot, which indicated that the residual errors in the current models were likely inflated by interoccasion variability. Estimation of interoccasion variability was not practical based on the limited data available (on average, three evaluable samples per subject) with a large number of occasions (up to 12 monthly injections). A bias of underprediction of low measured concentrations was observed in the population conditional weighted residuals for population prediction in NONMEM analysis (CWRES) versus PRED plot. ²⁸ Further investigation suggested that interoccasion variability in the serum concentrations immediately after dosing may play a role in such bias. The fit of the model predictions to observed data of six representative subjects is illustrated in Figure 4. A higher than expected concentration data point, presumably due to fast leakage of small fraction of drug into the blood circulation, is present in one of the subjects, and described well by the model. 

Figure 4. 

 

Comparison between the observed and model-predicted serum time–concentration curve in six representative subjects. Obs, observed data.

The final model was fit repeatedly to 500 replicates of the data set obtained by bootstrapping the original database. Mean values of the bootstrapped parameters and their 95% confidence intervals (CIs) were consistent with the point estimates and the corresponding CI obtained from the final model, which demonstrates a good and stable estimation of the parameters (data not shown). 

Vitreous and Serum Steady-State Ranibizumab Concentrations

Population simulations were conducted with the final model with covariate effect to calculate the central tendency and spread of vitreous and serum exposure of ranibizumab with the monthly and quarterly dosing regimens 0.3- and 0.5-mg dose levels (Fig. 5). Based on the model predictions, ranibizumab (0.3 mg or 0.5 mg) reached maximum serum concentration approximately 0.5 days after ITV administration. The exposure to ranibizumab in the vitreous humor was estimated at approximately 90,000 times that in systemic circulation (Table 5). With monthly dosing of 0.5 mg per eye, the median steady-state serum trough concentration of ranibizumab was estimated to be 0.22 ng/mL (Table 5). The serum concentration of ranibizumab 30 days after dosing was predicted to be higher than the lower limit of quantitation (0.3 ng/mL) in approximately 20% of the samples, which is higher than the 4.8% observed in the phase 3 study of 0.5 mg per eye. This difference is consistent with the overprediction when comparing the observed data–only approach with the ML approach with LTRs (Fig. 3). 

Figure 5. 

 

Population prediction of serum and vitreous concentration–time profiles with 0.5 mg per eye dosing.

Table 5. 

 

Model-Predicted Vitreous and Serum Steady-State Ranibizumab Concentrations and AUCs

Regimen	Steady-State Median (5th; 95th Percentiles) Ranibizumab Concentration
	Cmax, ng/mL	Cmin, ng/mL	AUC, d × ng/mL
0.3 mg/eye monthly for 12 doses
Serum	0.91 (0.47; 1.7)	0.13 (0.042; 0.29)	13 (6.6; 23)
Vitreous	87,000 (77,000; 120,000)	12,000 (2300; 41,000)	1,200,000 (670,000; 2,200,000)
0.5 mg/eye monthly for 12 doses
Serum	1.5 (0.79; 2.9)	0.22 (0.069; 0.49)	21 (11; 38)
Vitreous	140,000 (130,000; 190,000)	20,000 (3800; 68,000)	1,900,000 (1,100,000; 3,600,000)
0.3 mg/eye monthly for 3 doses, then quarterly for 3 doses
Serum	0.77 (0.34; 1.7)	0.0021 (0.000022; 0.022)	n/a
Vitreous	75,000 (75,000; 79,000)	200 (1; 4000)	n/a
0.5 mg/eye monthly for 3 doses, then quarterly for 3 doses
Serum	1.3 (0.57; 2.8)	0.0035 (0.000037; 0.037)	n/a
Vitreous	120,000 (120,000; 130,000)	330 (1.7; 6600)	n/a

 Simulation values at month 11 pre- and postdose were summarized as steady-state C_min and C_max, respectively, with a vitreous elimination half-life of 9 days. The AUC is the area under the concentration-time curve for one dose, or one monthly interval at steady state. n/a, not applicable.

With monthly and quarterly ITV regimens, the serum concentrations of ranibizumab at steady state for both the 0.3 and 0.5 mg per eye dose levels were estimated to be below the range needed to reduce VEGF-A–induced endothelial cell proliferation in vitro by 50% at all times (half maximal inhibitory concentration [IC₅₀] = 11–27 ng/mL). ^8,30 With monthly dosing, the steady-state vitreous concentrations of ranibizumab (0.3 and 0.5 mg per eye) were estimated to be well above the in vitro IC₅₀ range at all times. With the quarterly regimen, the median values of C_min in vitreous humor were estimated to be just above the in vitro IC₅₀ range, with large variability. At the end of the quarterly regimen, approximately 25% and 10% of simulated vitreous concentration–time profiles were estimated to fall below the in vitro IC₅₀ range for the 0.3 and 0.5 mg per eye dose levels, respectively. 

Discussion

This is the first comprehensive report on the PK of ranibizumab in subjects with neovascular AMD based on data from hundreds of samples collected during the registrational phase 3 trial MARINA (Minimally Classic/Occult Trial of the Anti-VEGF Antibody Ranibizumab in the Treatment of Neovascular Age-Related Macular Degeneration) and several smaller studies. The vitreous t_1/2 of ranibizumab in patients with AMD was estimated at 9 days. On reaching the systemic circulation, ranibizumab is eliminated with intrinsic t_1/2 of 2 hours. However, because the ranibizumab deposited in the stagnant vitreous egresses slowly into the systemic circulation, the apparent serum t_1/2 following ITV administration was 9 days. The systemic-to-vitreous exposure ratio for ranibizumab was 1:90,000. The steady-state serum concentrations of total ranibizumab were at all times below the concentrations needed to reduce VEGF-A–induced endothelial cell proliferation in vitro by 50%. 

The estimated systemic Vc/F was comparable with the physiologic volume of serum for a typical human, which suggests nearly 100% bioavailability of ranibizumab after ITV injection. There were no clinically significant sources or correlates of variability in ranibizumab exposure. 

Current analysis used serum sampling to estimate the intravitreal half-life of ranibizumab. Such technique has been widely adopted in the estimation of in vivo release half-life of a drug depot or a controlled release drug formulation in human subjects. ³¹ In the case of ranibizumab, the eye acts as a depot that slowly releases the drug into the systemic circulation. The observed/apparent systemic half-life of ranibizumab should be driven by the slowest kinetic process among intravitreal elimination, systemic tissue distribution, and renal filtration. Current analysis adopted the assumption that ITV elimination is the slowest process of the three for ranibizumab, based on both findings from animal PK data of ranibizumab ^9,10 and PK characterization of other Fab molecules in humans. ^17,18  

The vitreous t_1/2 of ranibizumab in patients with AMD was estimated to be 9 days. It has been acknowledged that the t_1/2 of ranibizumab in humans is difficult to predict from animal models data and may be longer than in animals because the volume of the human vitreous is three times that in rabbits and monkeys. ^9,10 Our findings suggest that the vitreous t_1/2 of the Fab ranibizumab is similar to the vitreous t_1/2 of the full-length monoclonal antibody bevacizumab estimated at 8 to 11 days. ^32,33 Durairaj et al. ³⁴ have shown that with molecular characteristics being equal, molecular size was the most important predictor of intravitreal t_1/2; however there were no Fabs among the tested molecules. Further, the primary elimination mechanism of high molecular weight molecules from the vitreous is reported to be through the physical fluid flow of the aqueous humor, where the role of molecular weight may be significantly diminished. ³⁵  

The present study estimates that following ITV administration in humans, the systemic t_1/2 of ranibizumab is 2 hours once the molecule reaches the serum. The systemic t_1/2 of bevacizumab has been reported to be approximately 20 days, following intravenous injections of 10 to 15 mg/kg in cancer patients. ²⁴ Reports of Fab t_1/2 in human are limited. Digoxin-Fab is reported to have a distribution t_1/2 of 1 hour and elimination t_1/2 of 20 hours following intravenous injection ¹⁷ ; a polyclonal anti-TNFα Fab was reported to have elimination t_1/2 of 20 hours. ³⁶ Radiolabeled Fab molecules, used as imaging agents, have elimination t_1/2 on the order of a couple of hours. ³⁷ Full-length antibodies like bevacizumab are subject to Fc-mediated recycling via the FcRn receptor, which results in slower clearance and prolonged systemic t_1/2. ^38,39 Antibody fragments, such as ranibizumab are not recycled due to the lack of an Fc region, and ranibizumab is small enough to be filtered by the kidney, thus subjected to renal catabolism. ^16–18 The estimated intrinsic t_1/2 of 2 hours for ranibizumab is an underestimation of the true elimination t_1/2, as the 2-hour t_1/2 is a hybrid of the distribution t_1/2 and the elimination t_1/2. Limited systemic data in the first days following ITV injection does not allow for the independent estimation the two t_1/2 estimates. 

The estimated systemic Vc/F was comparable with the physiologic volume of serum for a typical human, which suggests nearly 100% bioavailability of ranibizumab after ITV injection. The estimated rate of systemic clearance of ranibizumab was in good agreement with the fast catabolism and renal filtration mechanisms of Fab elimination. ^16–18 The systemic-to-vitreous exposure ratio for ranibizumab was 1:90,000. Our estimates suggest that with both monthly and quarterly dosing, the steady-state serum concentrations of ranibizumab were at all times below the concentrations needed to reduce VEGF-A–induced endothelial cell proliferation in vitro by 50% (IC₅₀ = 11–27 ng/mL). The systemic exposure of ranibizumab in the clinical studies was so low that approximately 70% of samples collected were below the assay detection limit of 0.3 ng/mL (Table 2). 

The estimated rate of vitreous elimination of ranibizumab was slower in patients treated concomitantly with verteporfin PDT. It is conceivable that the slower rate of absorption from the vitreous humor to the systemic circulation was a direct result of the vessel occlusion and soft tissue scarring. Based on the model structure, this change of Ka will lower the systemic C_max and will not affect the systemic AUC of ranibizumab, as systemic AUC is a function of dose and clearance only, given the assumed 100% bioavailability. 

Serum CrCL was the most significant covariate for ranibizumab CL/F. However, when compared with the large intersubject variability of CL/F, the effect of CrCL on CL/F was considered to have no clinical significance. No covariate was found to be significant for Vc/F. 

In summary, the PK analysis results presented here suggest that biologically meaningful steady-state vitreous concentrations of ranibizumab are achieved with monthly ITV dosing. The t_1/2 of ranibizumab was estimated at approximately 9 days in the vitreous followed by an observed serum t_1/2 of ranibizumab of approximately 9 days. Because the rate of vitreous elimination is the rate-limiting step, the apparent serum t_1/2 of ranibizumab after ITV administration would be equivalent to its vitreous t_1/2 of 9 days. Our results do not support adjustment of the ranibizumab dose based on subject characteristics. 

Supplementary Materials

pdf S1 

Acknowledgments

The authors thank Robert Y. Kim, MD (formerly of Genentech, Inc.), for his clinical guidance and input; Steven Butler, PhD (formerly of Genentech, Inc.), for his strategic guidance and statistical input; Rene Bruno, PhD (Pharsight), for his review and modeling guidance; Anne Kearns, BS (formerly of Genentech, Inc.), and her team for assay support; and Lisa Price, MS (Genentech, Inc.), for clinical data assembly. Third-party writing and formatting assistance for this manuscript was provided by Ivo Stoilov, MD, CMPP. 

Footnotes

 Supported by Genentech, Inc., South San Francisco, California, member of the Roche group.

Footnotes

 Presented in part at the annual meetings of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 1–5, 2005; the American Association of Pharmaceutical Scientists, San Antonio, Texas, October 29–November 2, 2006; and the American Academy of Ophthalmology, Orlando, Florida, October 22–25, 2011.

Footnotes

 Disclosure: L. Xu, Genentech, Inc. (E); T. Lu, Genentech, Inc. (E); L. Tuomi, Genentech, Inc. (I, E); N. Jumbe, Genentech, Inc. (E); J. Lu, Genentech, Inc. (E); S. Eppler, Genentech, Inc. (E); P. Kuebler, Genentech, Inc. (E); L.A. Damico-Beyer, Genentech, Inc. (I, E); A. Joshi, Genentech, Inc. (I, E)