Abstract
Purpose:
Surgically implanted intraocular lenses (IOLs) may be used as drug-delivery devices, but their effectiveness is not well defined. Computational fluid dynamics models were developed to investigate the capability of IOLs to release drugs at therapeutic concentrations.
Methods:
Models were generated using COMSOL Multiphysics. Primary open-angle glaucoma (POAG) and wet age-related macular degeneration (AMD) were simulated by reducing aqueous vein and choroidal blood flow, respectively. Release of dexamethasone, ganciclovir, or dextran was studied using common IOL materials, polydimethylsiloxane (PDMS) and poly(2-hydroxyethyl methacrylate) (PHEMA).
Results:
Drug clearance proceeds mainly through choroidal blood flow. When fully constricted, maximum concentration at the choroid (Cmax) values increased by 32.4% to 39,800%. Compared to dexamethasone, Cmax in different tissues decreased by 6.07% to 96.0% for ganciclovir and dextran, and clearance rates decreased by 16% to 69% for ganciclovir and by 92% to 100% for dextran. Using PDMS as the IOL reduced clearance rates by 91.3% to 94.6% compared to PHEMA.
Conclusions:
In diseased eyes, drugs accumulate mainly in posterior tissue; thus, choroidal drug toxicity must be assessed prior to IOL implantation in POAG and AMD patients. Moreover, drug properties modulated concentration profiles in all ocular segments. The hydrophobic small-molecule dexamethasone attained the highest concentrations and cleared the fastest, whereas hydrophilic macromolecular dextran attained the lowest concentrations and cleared the slowest. Furthermore, high concentrations were achieved quickly following release from PHEMA, whereas PDMS allowed for sustained release.
Translational Relevance:
In silico results can guide scientists and clinicians regarding important physiological and chemical factors that modulate tissue drug concentrations from drug-eluting IOLs.
Cataracts are the leading cause of blindness worldwide and affect approximately 21 million American adults.
1 A cataractous lens is removed through phacoemulsification and replaced by an intraocular lens (IOL).
1 Despite the positive outcomes of cataract surgery, complications may arise, including posterior capsule opacification (PCO),
2 especially for patients with secondary diseases such as uveitis. PCO is typically treated by removing the posterior capsule by neodymium-doped yttrium aluminum garnet (Nd:YAG) laser capsulotomy, which is inconvenient and costly.
3 Additionally, uveitis may arise post-surgery, leading to vision reduction or loss.
4 However, drug treatments such as corticosteroids can reduce inflammation caused by uveitis, and recent evidence suggests that similar drugs could reduce the occurrence of PCO.
5–7
To treat PCO and other ailments in the eye, drugs must be delivered at effective concentrations over the proper period. Conventional administration techniques include topical eye drops and periocular/intravitreal injections.
8 Low bioavailability is observed following topical administration and periocular injections, and intravitreal injections have low patient compliance and may lead to endophthalmitis, retinal detachment, and/or hemorrhage.
8 Such challenges prompted the development of alternative approaches. Drug-loaded implants are an attractive option because they can be placed near damaged tissue and be designed for sustained release.
8
Because IOLs are implanted during cataract surgery, they may be used as drug-delivery devices to prevent, alleviate, and treat post-surgery complications.
3 Drug can be preloaded into IOLs and released following implantation. Before an IOL may be used in this manner, tests must be performed to determine whether it can release drugs at therapeutic yet nontoxic concentrations to appropriate eye segments. True concentration measurements require human trials, but these cannot be conducted until estimates are obtained and laboratory tests are conducted.
Such estimates may be obtained through in silico modeling. Ocular drug delivery studies have been performed using COMSOL (Stockholm, Sweden) Multiphysics,
9 a computational fluid dynamics (CFD) tool that offers full simulation capability incorporating fluid mechanics, heat transfer, and transport phenomena. Studies have explored topical drug administration
10,11 and release from episcleral and subconjunctival implants,
12 but no existing models simulate release from an IOL. Also, previous models do not simulate diseased conditions known to alter the physical properties of ocular tissues. By altering choroidal and aqueous blood flow rates, the effects of conditions such as primary open-angle glaucoma (POAG) and wet age-related macular degeneration (AMD) on drug clearance or accumulation can be predicted.
In the present research, CFD models were generated to simulate and quantify physiologically representative drug release from an IOL in both healthy and diseased eyes. Three drugs with varying properties were assessed: dexamethasone (hydrophobic small-molecule drug), ganciclovir (hydrophilic small-molecule drug), and dextran (hydrophilic macromolecular model drug). CFD models were used to obtain estimates of drug concentration ranges within ocular tissue. Estimates aid in determining whether an IOL can effectively release a given drug to treat post-cataract surgery complications. Specifically, modeling can characterize drug concentration ranges, accumulation, and clearance rates in various eye segments, as well as release rates from different IOL materials.
Time-Dependent Studies Exploring Changes in Physiological and Transport Parameters
CFD models were generated as a tool to characterize the effects of disease, drug properties, and IOL material on drug concentration ranges in ocular tissue following IOL release. First, steady-state ocular temperature and flow profiles were generated because these phenomena affect drug transport and clearance and thus are necessary to obtain accurate drug concentration profiles. Time-dependent studies were then performed to analyze the effects of diseased conditions, drug properties, and IOL material on drug concentration ranges within ocular tissue.
Studies reducing (but not eliminating) choroidal blood flow were conducted to model wet AMD cases of varying severity. Inlet velocities are listed and
Cmax,
tmax, and
th for each ocular segment are reported in the
Supplementary Materials, Section L. Percent changes in
Cmax and
th compared to the base case are reported in
Table 5. Percent changes for blood flow removal (100% reduction) are included in
Table 5 for comparison.
Table 5. Percent Change in Cmax and th for Dexamethasone With Reduced Choroidal Flow Compared to the Base Case
Table 5. Percent Change in Cmax and th for Dexamethasone With Reduced Choroidal Flow Compared to the Base Case
Similar to results obtained from eliminating blood flow, increases in Cmax in the aqueous and cornea did not exceed the 5% threshold when blood flow was reduced to a lower degree. Cmax surpassed the threshold in the choroid at 50% reduction, but increases greater than 5% in all posterior segments were only observed when flow had been reduced by 99%. The first increases greater than 5% in th were apparent at 75% reduction (for the sclera and choroid), but 99% reduction was required for increases in th to exceed the threshold in all segments. Additionally, at 99% flow reduction, all increases were class 1, other than the increase in Cmax in the choroid (which was class 3), whereas at 100% reduction only increases in th in the anterior segment were class 1, and remaining increases were classes 2 or 3 (for Cmax in the choroid). This finding implies that choroidal blood flow greatly contributes to drug clearance even at low velocities.
Hydrophilicity/Hydrophobicity and Molecular Size Modulate Maximum Drug Concentration and Clearance
Physiologically representative in silico models were generated to characterize IOL drug release to treat post-cataract surgery complications. The ocular temperature profiles from the models matched physiological values, including normal corneal surface and corneal limbus temperatures (
Fig. 4;
Supplementary Materials, Section J).
42 The aqueous flow velocity exceeded that of the vitreous because the fluid does not enter the vitreous directly but rather is a result of posterior-directed aqueous flow.
17,18 The aqueous vein and choroid were modeled like rigid pipes, which implies continuous drug clearance. However, the literature suggests that flow into aqueous veins is pulsatile
30 and that a portion of choroidal blood flow is pulsatile,
45 meaning that clearance may not be continuous. The pulsatile nature of such flows is not well characterized, but models can be updated when physiological or ex vivo verified parameters (frequencies, amplitudes) are obtained.
Restricting flow at different interfaces mimics inflammation and disease. Eliminating flow from the aqueous vein showed that patients with POAG would experience elevated drug concentrations in the sclera and choroid, whereas eliminating flow from the choroid showed that patients with AMD would experience increased drug concentrations in all posterior segments. Adverse effects could occur in both cases. For example, toxic hydroxychloroquine levels may lead to choriocapillaris degeneration,
46 and elevated amiodarone levels may result in choroidal neovascularization.
47 Other drugs may also cause adverse side effects in ocular tissue. Choroidal and scleral drug toxicities must be assessed in patients with POAG and posterior tissue toxicity must be analyzed in patients with AMD before an IOL is used for drug delivery. Also, results suggest that, unless AMD is severe (flow reduction greater than 50%), only the drug concentration at the choroid will be altered. The extent of this effect can be fully characterized when a physiological or ex vivo–verified blood flow rate estimate is obtained.
The size and charge of drugs influence their concentration and diffusion through ocular media. In the models, hydrophobic drugs attained higher concentrations than hydrophilic drugs, except where hydrophilic drug molecular size hindered drug diffusion. Restricted diffusion resulted in accumulation and increased concentration, as was observed for dextran in the retina. Clearance depended on molecular size and hydrophilicity/hydrophobicity in all ocular segments; clearance was fastest for hydrophobic small molecules and slowest for hydrophilic macromolecules, trends that are in agreement with results from previous studies.
48 Dextran thus accumulated in the retina because its large size (40,000 Da) resulted in low retinal pigment epithelium permeability, which is the rate-limiting step.
37
Delivery profiles are also material specific, illustrating that the material of the IOL can be designed to control delivery. PHEMA released elevated concentrations within a short time frame, whereas PDMS offered sustained delivery for when long delivery times are required. However, to sustain release from PHEMA, loading by supercritical fluid impregnation or external coatings can help to reduce diffusion.
24 Release times may be further altered by incorporating drugs into the IOL during synthesis at levels above their solubility limits or loading drugs into reservoirs attached to the IOL.
24 As such, it may be more efficient to control the delivery profile by altering the loading strategy rather than the IOL material. Different forms of loading can be analyzed in models when appropriate diffusion coefficients have been obtained experimentally.
In summary, our results show that diseased conditions lead to drug accumulation in posterior ocular tissue, most notably in the choroid. Accordingly, choroidal drug toxicity should be analyzed prior to use of a drug-eluting IOL in patients with POAG or AMD. As well, drug properties affected maximum concentration and clearance; the highest concentrations and fastest clearance rates were attained by the hydrophobic small molecule, whereas the opposite was true for the hydrophilic macromolecule. Also, the IOL material altered concentration profiles; high concentrations were achieved quickly following release from PHEMA, whereas sustained release was achieved following release from PDMS.
The authors thank Jordan Pernari for helping with the development of methods.
Supported by funding from Queen's University and the Natural Sciences and Engineering Research Council of Canada.
Disclosure: D.P. Clasky, None; L. Meunier, None; L.A. Wells, None