Physical Parameters Influencing the Rate of Subretinal Injection: A Microfluidic Analysis Authors: Aaron Nagiel, Lamberto Rondoni Published: 2026 ABSTRACT PurposeSubretinal injection during vitrectomy has become an important technique owing to the growing number of cell and gene therapy trials for retinal disease. Reducing the injection rate can mitigate trauma or shear forces on the retinal pigment epithelium, yet many surgeons alter the subretinal injection system in ways that could influence this rate.MethodsWe created a theoretical model of a standard subretinal injection system by applying the Hagen–Poiseuille equation, assuming steady, incompressible, laminar flow of a water-like Newtonian fluid. The system consisted of the vitrectomy console, plastic tubing, the MedOne Microdose kit, and the MedOne 38G/41G subretinal cannula set.ResultsThis model generated an open-access web app allowing user modification of parameters including cannula length and diameter, beveling, viscosity, console injection pressure, and intraocular pressure. The internal diameter of the cannula had the most prominent effect on flow, with small increases producing large increases in flow. Increasing console pressure linearly increased tip flow. Trimming the tip by 3 mm modestly increased flow by 18%, whereas beveling the tip or altering intraocular pressure produced negligible changes.ConclusionsWith this theoretical model, we were able to identify key parameters that influence the rate of subretinal injection, thus providing important guidance for surgeons.Translational RelevanceThis open-access web application enables vitreoretinal surgeons and gene- and cell-therapy trialists to assess how modifications to the subretinal injection system alter flow at the cannula tip, supporting standardized and reproducible delivery that minimizes flow-related trauma to the outer retina. Introduction Over the past decade, pars plana vitrectomy with subretinal injection has become a routine surgical technique for delivering cell and gene therapies to the outer retina. The 2017 approval of voretigene neparvovec-rzyl (Luxturna)—the first in vivo gene therapy licensed by the US Food and Drug Administration—was based on a randomized, controlled phase 3 trial in which the vector was administered by bilateral subretinal injection, with durable functional benefit demonstrated through at least four years of follow-up.1,2 This milestone catalyzed rapid growth of the field, and since then dozens of subretinal gene therapy trials have been registered worldwide, with the majority using adeno-associated viral vectors to treat inherited retinal degenerations.3,4 Delivering vector directly into the subretinal space places the therapeutic agent in immediate contact with the photoreceptors and retinal pigment epithelium (RPE), generating high local transduction at a comparatively low viral dose while bypassing the internal limiting membrane and limiting systemic and intraocular immune exposure.5 Despite the growing use of the technique, the execution remains highly variable, and the delivery parameters that govern how vector reaches the subretinal space are neither standardized nor well defined across surgeons.6,7 Operators differ in the console injection pressure they select, the intraocular (infusion) pressure they maintain, the length to which the cannula is trimmed (if any), and whether to bevel the cannula tip.8 Because the cannulas used for subretinal delivery—typically 38-gauge external/41-gauge internal—place these injections firmly within the microfluidic domain, even small modifications to these parameters can produce disproportionately large changes in the flow rate at the cannula tip.7,9 This is clinically important because controlled experimental studies have shown that higher flow rates produce greater disruption of the RPE, whereas slower, pressure-limited injection mitigates this damage.9–11 Excessive flow and the resulting mechanical stress on the neurosensory retina have likewise been implicated in foveal stretch, dehiscence, and macular hole formation.7 In the present study, we develop a theoretical microfluidic model of the subretinal injection system—comprising the vitrectomy console, connecting tubing, and the 38-gauge/41-gauge subretinal cannula—by applying the Hagen–Poiseuille relation for laminar flow of a Newtonian fluid. The model isolates the contribution of individual, surgeon-modifiable parameters and demonstrates that console injection pressure scales tip flow rate linearly, that trimming or beveling the cannula exerts only modest effects, and that cannula diameter is by far the dominant determinant of flow. By quantifying these relationships, the model offers a generalizable framework to help surgeons and clinical trialists achieve precise and reproducible control over subretinal injection speed. Methods The subretinal injection apparatus was modeled as a series of rigid cylindrical conduits conveying an incompressible, Newtonian fluid under steady, laminar, fully developed flow with a no-slip boundary condition. This assumption is consistent with experimental measurements showing that flow through a 41-gauge cannula remains laminar up to approximately 32 PSI, above the range of console pressures modeled here.9 For such flow through a straight channel of circular cross-section, the volumetric flow rate and the axial pressure drop are related by the Hagen–Poiseuille equation,12(1)ΔP=8μLQ/πR4where ΔP is the pressure drop along the channel, µ is the dynamic viscosity of the fluid, L is the channel length, Q is the volumetric flow rate, and R is the internal radius. Rearranging for the flow rate gives (2)Q=πR4ΔP/8μLwhich exposes the strong fourth-power dependence of flow on the internal radius, such that small changes in lumen diameter exert a disproportionate effect on the delivered rate. Each component of the system can be assigned a hydraulic resistance RH = 8µL/πR4, so that ΔP = Q · RH. Because the segments are connected in series and the fluid is treated as incompressible, the volumetric flow rate is conserved along the system and the total pressure drop equals the sum of the contributions of the individual segments, (3)ΔPtotal=ΔPtubing+ΔPcannula Fluid is driven into the subretinal space only when the pressure delivered at the cannula tip exceeds the intraocular pressure. When the mean flow velocity v¯ is known in place of Q, the two are related through the cross-sectional area as Q = v¯πR2. We applied this framework to the subretinal injection system used at our institution. Pressurized air was supplied by an Alcon Constellation Vision System (Alcon, Fort Worth, TX, USA), for which the console injection pressure is operator-selectable—typically a maximum of 12–16 PSI—and is modulated in real time through a linear foot pedal. The console drives the plunger of a 1-mL syringe containing the therapeutic agent (Microdose injection kit; MedOne Surgical, Sarasota, FL, USA) through flexible polymer tubing approximately 218 cm in length, with an outer diameter of 0.5 cm and an internal lumen diameter of approximately 0.2 cm. The syringe couples directly to a MedOne 25-gauge subretinal injection cannula with a 38-gauge (outer)/41-gauge (inner) polyimide tip, whose tip has an internal lumen diameter of 0.10 mm and a total length of 20 mm. Subretinal delivery proceeds by advancing this tip to the retinal surface and raising a localized bleb via retinotomy.5,6 Baseline parameters are summarized in Table 1. The working fluid was assumed to approximate water at body temperature (37°C; µ ≈ 0.691 × 10−3 Pa · s)13 and to be incompressible, so that the flow entering the tubing equals the flow exiting the tip. With these assumptions the model computes the delivered tip flow for any specified combination of input parameters and was constructed to allow each parameter—console pressure, cannula length, cannula internal radius, tip bevel angle, fluid viscosity, and intraocular pressure—to be varied independently about this baseline. Table 1.Standard (Baseline) Operating Parameters Used in the Microfluidic Model of Subretinal InjectionParameterStandard ValueSubretinal polyimide cannula length20 mmCannula internal diameter (41G)0.10 mmCannula bevel angle (uncut)0°Fluid temperature37°CFluid viscosity (water at 37°C)0.691 mPa · sConsole injection pressure8 PSIIntraocular pressure10 mm Hg Standard (Baseline) Operating Parameters Used in the Microfluidic Model of Subretinal Injection For the accompanying web application (10.17632/bpw5st58h9.1),14 the following assumptions were used: The flow is in a steady state, and has low Reynolds number Re = ρ · v · d/µ, with ρ fixed at 1000 kg/m³ which is the density of water at room temperature and pressure. This allows the flow to be treated as a laminar, fully-developed flow of a Newtonian fluid in a straight, rigid, uniform circular lumen. Consequently, the use of the Hagen–Poiseuille formula is justified. The flow-regime indicator simply confirms whether the Poiseuille equation is valid for the settings entered; it is not a safety threshold for the retina or the vector. By default, the application makes no correction for entrance effects (the short developing region at the start of a channel) or for height differences within the system, unless the minor-loss option is selected. When that option is selected, only small corrections are applied. Results A biophysical model was developed as described in the Methods based on the following example system: Alcon Constellation vitrectomy console, Viscous Fluid Control tubing, MedOne Microdose kit, and MedOne 38G/41G tip subretinal cannula. This allowed independent modification of several physical parameters governing subretinal delivery, including fluid temperature, cannula length and internal diameter, cannula bevel angle, fluid viscosity, console injection pressure, and intraocular pressure (IOP). Note that this required a few assumptions and simplifications: (1) the injected fluid is incompressible; (2) the fluid instantly heats to body temperature as it enters the eye; and (3) the viscosity of the fluid is equivalent to water. The baseline values used for the model, chosen to reflect routine clinical practice, are summarized in Table 1. Note that although these represent the major modifiable parameters, there remain other complex and difficult to quantify parameters such as the degree of photoreceptor-RPE adhesion (resistance to detachment), viscoelasticity of the neurosensory retina, and tubing compliance/pneumatic lag. Console injection pressure had an appreciable effect on delivered rate, scaling flow in linear proportion as predicted by Equation 2. At the baseline 8 PSI, the model delivered approximately 9.6 µL/s at the cannula tip, corresponding to a mean tip velocity of 1.22 m/s. Doubling the console pressure to 16 PSI doubled the flow to 19.3 µL/s (+103%; 2.5 m/s), and increasing it to 24 PSI tripled the flow to 29.1 µL/s (+205%; 3.7 m/s) (Table 2). To verify the modeled flow rate, we analyzed a representative, real-world subretinal injection procedure for voretigene neparvovec-rzyl (Supplementary Video S1). In this video, a total volume of 300 µL is delivered at a console pressure of approximately 12 PSI uniformly over 20 seconds thus representing a flow rate of 15 µL/s. This aligns with the model's predicted flow of 14.4 µL/s under these conditions. Table 2.Representative Effects of Individual Parameters on Subretinal Injection Flow and Tip VelocityModeled ParameterModified Value (Fold Change)Fluid Flow at Tip (µL/s)Fold Change in FlowFluid Speed at Tip (m/s)Fold Change in Speed vs. Baseline (%)Baseline—9.61.00 (—)1.221.00 (—)Console injection pressure16 PSI (2×)19.32.03 (+103%)2.462.03 (+103%)24 PSI (3×)29.13.05 (+205%)3.713.05 (+205%)Cannula length19 mm (−1 mm)10.11.05 (+5%)1.281.05 (+5%)17 mm (−3 mm)11.21.18 (+18%)1.431.18 (+18%)Cannula internal diameter0.08 mm3.90.41 (−59%)0.780.64 (−36%)0.09 mm6.30.66 (−34%)0.990.81 (−19%)0.11 mm14.01.46 (+46%)1.471.21 (+21%)0.12 mm19.82.07 (+107%)1.751.44 (+44%)Tip bevel45°≈9.6≈1.00 (<1%)≈1.22≈1.00 (<1%)Intraocular pressure5 mm Hg (0.5×)9.71.01 (+1%)1.231.01 (+1%)25 mm Hg (2.5×)9.20.96 (−4%)1.170.96 (−4%)35 mm Hg (3.5×)9.00.94 (−6%)1.140.94 (−6%)Each parameter was varied independently about the baseline configuration (Table 1), holding all others constant. Flow rate and mean tip velocity were computed from the Hagen–Poiseuille model (Equation 2). Fold changes are expressed relative to baseline. Because tip velocity equals flow divided by the lumen cross-sectional area, the flow and tip-speed fold changes coincide for every parameter except cannula internal diameter, where flow scales with the fourth power and tip velocity with the square of the lumen radius. Representative Effects of Individual Parameters on Subretinal Injection Flow and Tip Velocity Each parameter was varied independently about the baseline configuration (Table 1), holding all others constant. Flow rate and mean tip velocity were computed from the Hagen–Poiseuille model (Equation 2). Fold changes are expressed relative to baseline. Because tip velocity equals flow divided by the lumen cross-sectional area, the flow and tip-speed fold changes coincide for every parameter except cannula internal diameter, where flow scales with the fourth power and tip velocity with the square of the lumen radius. Cannula length exerted a comparatively modest, inverse effect on flow. Trimming the 20 mm tip by 1 mm increased the delivered rate by approximately 5% (to 10.1 µL/s), and trimming by 3 mm increased it by approximately 18% (to 11.2 µL/s) (Table 2). Although shortening the cannula reproducibly accelerated injection, the magnitude of this effect was small relative to changes in console pressure or lumen diameter. By far the dominant determinant of flow was the internal diameter of the cannula, reflecting the fourth-power dependence of flow on lumen radius (Q ∝ R⁴). Holding all other parameters at baseline, widening the lumen from 0.10 mm to 0.11 and 0.12 mm increased flow by 46% and 107% (to 14.0 and 19.8 µL/s, respectively), whereas narrowing it to 0.09 and 0.08 mm reduced flow by 34% and 59% (to 6.3 and 3.9 µL/s, respectively) (Table 2). A 20% increase in diameter (0.10 to 0.12 mm) thus more than doubled the delivered rate, underscoring that even small differences in cannula bore dominate all other surgeon-modifiable parameters. In contrast, tip beveling and intraocular pressure produced only minor changes in flow. Beveling the tip at 45° shortens the effective cannula length by approximately √2 times the diameter (≈0.14 mm, less than 1% of the 20 mm length) and accordingly altered flow by less than 1%. Varying IOP across the clinically encountered range likewise had little effect: relative to the 10 mmHg baseline, flow changed by +1% at 5 mm Hg, −4% at 25 mm Hg, and −6% at 35 mm Hg (Table 2), because intraocular pressure is small compared with the console driving pressure. We developed a web application (http://doi.org/10.17632/bpw5st58h9.1)14 that allows the user to modify key parameters to generate customized outputs for flow and velocity for the above system. The modifiable parameters include cannula length, inner diameter, injection pressure, intraocular pressure, and fluid viscosity. Discussion In this study we developed a theoretical microfluidic model of the subretinal injection system, grounded in the Hagen–Poiseuille relationship, to quantify how individual, surgeon-modifiable parameters govern the rate and velocity of fluid delivery at the cannula tip. The model yielded four principal findings (examples in Table 2). First, console injection pressure scaled tip flow linearly, such that raising the console from 8 to 16 PSI doubled the delivered rate and raising it to 24 PSI tripled it. Second, the internal diameter of the cannula was by far the dominant determinant of flow, reflecting its fourth-power dependence on lumen radius: widening the bore from 0.10 to 0.12 mm more than doubled the delivered rate, whereas a modestly narrower lumen sharply reduced it. Third, trimming the cannula produced only a modest, inverse change in flow, raising the rate by approximately 18% when 3 mm was removed from a 20 mm tip. Finally, tip beveling and intraocular pressure exerted negligible effects across the clinically encountered range. Together these results provide a quantitative framework with which a surgeon can anticipate how a given change to the injection system will alter delivery speed. These considerations are increasingly relevant as subretinal delivery has become the route of choice for an expanding range of retinal gene and cell therapies.4,15 Across these diverse therapeutics the delivered flow rate is a shared and consequential variable: controlled experimental studies have shown that higher flow rates and injection pressures produce greater loss and structural disruption of the RPE, whereas slower, pressure-limited delivery mitigates this damage.9–11 Because the cannula and console settings that determine flow differ widely among surgeons and trials, standardizing the parameters that govern injection speed is essential both to protect the outer retina and to ensure reproducible dosing. Reassuringly, the model approximated empirical observation: for a representative clinical injection of voretigene neparvovec analyzed frame-by-frame (Supplementary Video S1), the delivered rate of approximately 15 µL/s closely matched the 14.4 µL/s predicted by the model at the 12 PSI console pressure recorded in the video. Nevertheless, several limitations to this work must be highlighted. First, the injectate was modeled as water, whereas real therapeutics may behave as non-Newtonian fluids when they contain surfactants, high-titer viral particles, or suspended cells, all of which can raise the effective viscosity and alter flow.7 Second, the model treats the tubing and pneumatic drive as rigid and instantaneous, yet in practice the compliance of the polymer tubing and the compressibility of the pressurized air column introduce a capacitance that produces lag between the console signal and the flow actually delivered at the tip. Third, reflux of fluid back through the retinotomy was assumed to be zero, so that the entire delivered volume enters the subretinal space. Finally, intraocular pressure was held constant, although it rises transiently as the bleb forms; as shown in Table 2, however, IOP was among the least influential parameters, so this simplification is unlikely to materially affect the modeled rate. This study has additional limitations. Foremost, the model is purely theoretical and was not validated against a direct, instrumented measurement of tip flow; our comparison rests on a single representative surgical video (Supplementary Video S1) rather than a controlled flow-metrology experiment. Therefore the results presented here and in the web application should be considered unverified and require clinical judgement. Prospective bench measurement of the delivered volume across a range of cannulas, pressures, and fluids will be needed to confirm the predicted relationships. A further practical limitation is that the true console pressure applied during a given injection is seldom known: surgeons typically record only the maximum pressure set on the console, whereas the instantaneous pressure is modulated continuously through the foot pedal. This uncertainty in the real-world input pressure constrains how precisely the model can be applied retrospectively to an individual case. In summary, we present a generalizable biophysical model that quantifies how the physical parameters of the subretinal injection system determine the rate and velocity of fluid delivery, identifying cannula internal diameter and console injection pressure as the dominant levers and confirming that cannula length, tip bevel, and intraocular pressure are comparatively minor (Table 2). By making these relationships explicit, the model offers surgeons a rational basis for selecting and adjusting their injection setup to achieve a target delivery speed and thereby minimize flow-related trauma to the outer retina. Future work can extend the framework to additional cannula and tip geometries and, most importantly, to the non-Newtonian, cell- and vector-laden fluids used in contemporary gene and cell therapies, coupled with direct experimental validation of the predicted flow rates. Supplementary Material Supplement 1