(260c) Real-Time Monitoring and Model-Based Prediction of Product Properties in a Dropwise Additive Manufacturing Process for Pharmaceuticals

Radcliffe, A. J., Purdue University
Nagy, Z. K., Purdue University
Reklaitis, G. V., Purdue University
In recent years additive manufacturing (AM) technology has attracted considerable interest for its potential applications to on-demand manufacturing of dosage forms with individually adjusted doses, which is a necessity for implementation of precision medicine at the societal level, and for development of robust manufacturing platforms for early drug development or continuous manufacturing. Commensurate with the level of interest is the variety of different technologies based on two-dimensional (2D) or three-dimensional printing (3D) techniques, all of which present the common advantages of precision and flexibility in dosing inherent to AM processes. However, differences exist in the printing material requirements and in the achievable production rates for the process; both must be considered when proposing implementations in the various contexts (compounding in clinical or pharmacy setting; full-scale production). In the former, stringent requirements imply that the system would be suited to a narrow range of dosage forms, while the latter is an eminently practical consideration – if the production time for a single dose is on the order of minutes, then the system may struggle to meet demand.

Of the AM techniques, the system based on the drop-on-demand method (2D printing) for fluid deposition/dispensing has demonstrated the capacity to serve as flexible manufacturing platform for pharmaceutical dosage forms, with doses in amounts of 1-100 mg active ingredient produced from printing materials which ranged from drug solutions to drug-polymer melts and concentrated suspensions of micronized particles [1, 2, 3]. Most recently demonstrated was the development of the system for precision dispensing of micronized powders by processing them as particle suspensions, which, as a fluid-based method, avoids powder flow issues and may readily be adapted to new particulate materials using dimensional analysis [3, 4]. For dosage units which contain 100mg of active drug, production rates of up to 2000 doses per hour, with product RSD < 2%, have been achieved with suspension processing (3, 4). This system uses the drop-on-demand (DoD) printing method to deposit drops on to a prepared substrate, wherein the content of each dose is determined by the number, volume and composition of drops; consequently, quick adjustments to dosage unit properties may be made by changing the deposited fluid volume, and/or composition.

From a process perspective, the DoD method presents a unique opportunity for monitoring of dose uniformity through the online measurement of each drop added to a given substrate; this enables model-based estimation of the drug product properties (e.g. active ingredient and/or excipient amount), which when combined with statistical modeling, can be used for real-time quality control purposes. To achieve this, one must have real-time measurements of drop volume and concentration in the printing fluid just upstream of the drop ejection point. As demonstrated in previous works, drop volume measurements may be obtained using an online image acquisition system and image analysis [1, 2], with corrections necessitated by particle effects made using a Bayesian statistical model [5]. In-line concentration measurement is complicated by the small scale of the process, which precludes the use of some sensors, and by the variety of process fluids for which the sensor must have the capacity to accurately measure. For pure fluids there are few limitations imposed by physical constraints, but for more interesting printing materials such as concentrated particulate suspensions, choices for sensing technology are limited to ultrasound velocity (correlation with particle fraction through modified Urick equation) or Raman spectroscopy.

In this work we investigate these potential methods for in-line concentration monitoring for particle suspensions within the dropwise additive manufacturing system. Combining the real-time drop volume and inline concentration measurements, we are able to make model-based predictions for properties of dosage units produced during manufacturing runs for doses of variable potency; comparison of model predictions with offline analysis of the dosage units is then presented. Through the use of a statistical approach for model-based prediction, and, consequently, quantification of uncertainty in the dose content, additional modeling work enables making a decision as to whether each dose should be released to the patient, formulated based on minimization of the expected loss in the Bayesian context. In this way, an important step towards the realization of precision medicine is made, namely, model-based real-time release of doses (at least, clinical/compounding settings).


  1. Hirshfield, L., Giridhar, A., Taylor, L. S., Harris, M. T., & Reklaitis, G. V. (2014). Dropwise additive manufacturing of pharmaceutical products for solvent-based dosage forms. Journal of pharmaceutical sciences, 103(2), 496-506.
  2. Içten, E., Giridhar, A., Taylor, L. S., Nagy, Z. K., & Reklaitis, G. V. (2015). Dropwise additive manufacturing of pharmaceutical products for melt-based dosage forms. Journal of pharmaceutical sciences, 104(5), 1641-1649.
  3. Radcliffe, A. J., Hilden, J. L., Nagy, Z. K., & Reklaitis, G. V. (2019). Dropwise Additive Manufacturing of Pharmaceutical Products Using Particle Suspensions. Journal of pharmaceutical sciences, 108(2), 914-928.
  4. Radcliffe, A. J., & Reklaitis, G. V. (2017). Dropwise additive manufacturing using particulate suspensions: feasible operating space and throughput rates. In Computer Aided Chemical Engineering (Vol. 40, pp. 1207-1212). Elsevier.
  5. Radcliffe, A. J., & Reklaitis, G. V. (2018). Bayesian estimation of product attributes from on-line measurements in a dropwise additive manufacturing system. In Computer Aided Chemical Engineering (Vol. 43, pp. 1243-1248). Elsevier.