Research ArticleAPPLIED SCIENCES AND ENGINEERING

Modeling, design, and machine learning-based framework for optimal injectability of microparticle-based drug formulations

See allHide authors and affiliations

Science Advances  08 Jul 2020:
Vol. 6, no. 28, eabb6594
DOI: 10.1126/sciadv.abb6594
  • Fig. 1 Microparticle clog formation as a major barrier to administration of particulate biopharmaceutical formulations.

    Schematic illustration of a typical syringe attached to a hypodermic needle containing a homogenous mixture of polymeric microparticles (A) before injection, (B) during injection, and (C) after injection. The remaining volume of injection solution after full course of plunger in the syringe needle is called dead space or dead volume. Depending on different design elements, particles are not entirely transferred into the patient through the needle and they are prone to accumulate in the syringe or clog the needle. An experimental case of microparticle accumulation in the syringe and needle inlet in (D) empty syringe and (E) after the full course of plunger displacement. The system illustrated herein is a 3-ml-sized BD Luer-Lok syringe attached to an 18G BD hypodermic needle filled with a mixture of water/poly(lactic-co-glycolic acid) (PLGA). The microparticles are PLGA microspheres with average diameter of 325 μm. Photo credit: Morteza Sarmadi, Massachusetts Institute of Technology.

  • Fig. 2 Numerical modeling of clog formation in hypodermic syringe needles.

    (A) Illustration of the multiphysics numerical model used for simulations corresponding to geometry of a 3-ml-sized BD syringe containing 2 ml of particle-solution mixture. (B) Meshed geometry and velocity contours within different locations including (C) syringe body (barrel), (D) syringe tip (adapter), and (E) needle. (F) Initial position of particles in syringe barrel following random spatial distribution. (G) Critical location in the syringe needle defined before the needle inlet to incorporate clog formation domain in the numerical model. (H) A comparison between numerical and experimental values of injectability. (I) Injectability decreased over time as the microparticles clogged the needle. (J) Illustration of particles clogging the needle in the model for different needle gauges and injection solutions. Microspheres with an average diameter of 202 ± 6 μm and various concentrations were used in these simulations.

  • Fig. 3 Numerical results demonstrating the effect of design parameters important in drug delivery on microparticle injectability.

    (A) The effect of particle concentration (200 μm in diameter, spherical) on microparticle injectability. (B) Effect of particle sphericity (shape) on injectability. (C) Injectability of microspheres (1000 particles) as a function of particle size. Effect of initial particle offset (10,000 particles) from (D) the plunger and (E) the syringe inner wall on injectability. Particles are illustrated as green dots in the syringe.

  • Fig. 4 Experimental study on the effect of different parameters on injectability.

    Mean response of injectability obtained from Taguchi analysis indicating the effect of the design parameters including (A) viscosity, (B) needle gauge (DOE1), (C) needle gauge (DOE2), (D) particle shape, (E) particle size, and (F) particle concentration. ANOVA results demonstrating the comparative significance and ranking of each of the design parameters studied for 1-ml and 3-ml syringes are shown (****P < 0.0001 and *P < 0.05). (G) From DOE1, solution viscosity was identified as the most important factor, and (H) in DOE2, needle size and particle size were the two dominant design parameters.

  • Fig. 5 Prediction of injectability for various syringe-needle systems.

    (A) The plot contour demonstrates predicted injectability as a function of the two nondimensional parameters calculated using Eq. 4. (B) The relationship between actual microparticle injectability from the experiments and the predicted injectability calculated by the formula and ANN. High injectability region was assumed where the lowest bound of actual injectability (average subtracted by standard deviation) was greater than 50%. Error bars show SD. (C) Proposed flowchart demonstrating the potential application of the proposed predictive tools for design of high injectability drug delivery microparticles.

  • Fig. 6 Design, optimization, manufacturing, and in vivo testing of a customized syringe made for high injectability applications.

    (A) Design parameters considered in the numerical modeling and optimization of the syringe tip (adapter) based on two interconnected nozzles. (B) Meshed geometry of the syringe tip. (C) Velocity magnitude contour in some of the designs as the criterion for optimization. (D) Examples of detailed design of the syringe tip profile in SolidWorks based on different design parameters such as θ and L. Scale bars, 5 mm. (E) Different 3-ml-sized syringes (top five syringes) were manufactured by stereolithography (SLA) 3D printing to optimize needle attachment, decrease dead volume, and improve injectability in vitro. A 1-ml version of the optimum design (θ ~ 25°) based on in vitro injections was further manufactured for in vivo injections (two bottom designs). (F) Comparison between 1-ml version of the proposed syringe and a comparable commercial syringe. Unlike in the proposed design, particles were observed to accumulate in sharp corners in the barrel of the commercial syringe. The proposed design also demonstrated less dead volume accompanied by less particle waste. More details can be found in the Supplementary Materials. (G) Scanning electron microscopy (SEM) images of PLGA1-fabricated core-shell microparticles, a recently developed platform for single-injection vaccination. Different core geometries but the same exterior size (400 μm by 400 μm by 300 μm) can be achieved by modifying the manufacturing steps. Different configurations of the sealed and unsealed base layer in high and low magnifications are demonstrated. Scale bars, 500 μm. A full description of the manufacturing steps can be found in (5). (H to J) In vivo subcutaneous injection results (n = 5) using different strategies to enhance microparticle injectability (*P < 0.05, ***P < 0.001, and ****P < 0.0001). Particles were initially loaded into the syringe either randomly distributed (noted as “distributed”) or with a certain offset from the plunger (noted as “with offset”). The term inside the brackets indicates type of the syringe, either the commercial or the proposed syringe. PLGA1, PLGA2, and PLGA3 refer to cubic particles with a dimension of 400 μm by 400 μm by 300 μm, 350 μm by 350 μm by 330 μm, and 162 μm by 162 μm by 162 μm made from PLGA(Resomer 502H), respectively. PLGA1 has a cubic internal core of 200 μm by 200 μm by 100 μm, while PLGA 2 and PLGA 3 are solid (i.e., nonhollow). Photo credit: Morteza Sarmadi, Massachusetts Institute of Technology.

Supplementary Materials

  • Supplementary Materials

    Modeling, design, and machine learning-based framework for optimal injectability of microparticle-based drug formulations

    Morteza Sarmadi, Adam M. Behrens, Kevin J. McHugh, Hannah T. M. Contreras, Zachary L. Tochka, Xueguang Lu, Robert Langer, Ana Jaklenec

    Download Supplement

    This PDF file includes:

    • Supplementary Materials and Methods
    • Figs. S1 to S5
    • Tables S1 to S5
    • References

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article