1. Introduction
Chemotherapy (CTX) drugs are a vital tool for the treatment of cancers (Reference Arruebo, Vilaboa, Sáez-Gutierrez, Lambea, Tres, Valladares and González-FernándezArruebo et al., 2011). However, due to their inherently cytotoxic nature, handling the drugs and preparing prescriptions for patients carries innate risks for pharmacists and technicians (Reference Graeve, McGovern, Alexander, Church, Ryan and PolovichGraeve et al., 2017). Addressing this issue becomes increasingly important as cancer incidences continue to rise globally due in part to improved screening techniques (Reference Soerjomataram and BraySoerjomataram & Bray, 2021), an aging global population (Reference Ju, Zheng, Wang, Zhang, Zeng, Chen, Sun, Li and WeiJu et al., 2023), and the rapid growth of ultra-processed food consumption combined with an increase in obesity (Reference Kliemann, Al Nahas, Vamos, Touvier, Kesse-Guyot, Gunter, Millett and HuybrechtsKliemann et al., 2022). Handling CTX drugs is necessary as part of the compounding process, whereby the CTX drug is diluted and dosed to the prescribed amount before being placed within a dispensing device (e.g.: an IV bag). Factors such as patient age, weight, and height all influence the dilution and dosage and are highly patient specific (Reference Chiumente, Russi, Todino, Mengato, Coppola, Rivano, Palozzo and JommiChiumente et al., 2021). Automated Drug Compounding (ADC) addresses many of the issues present in manual compounding, but is still in its infancy, with many ADC systems being bulky, expensive, and disruptive to existing hospital procedures and processes.
The aim of this paper is to assess the challenges surrounding ADC adoption thereby answering the following research question:
“What challenges arise in developing a syringe-compatible robotic end effector for automated drug compounding, and which design features are critical for successful implementation?”
The paper begins by evaluating the existing literature for manual and automated drug compounding in order to contextualise the research and highlight the existing design challenges that must be addressed (Section 2). The paper continues by distilling these design challenges into a list of features and requirements that an end effector must adhere to in order to address the research question (Section 3.1). Using these requirements, a design that addresses the challenges is identified (Section 3.2). An evaluation of this design is documented (Section 4). Finally, a conclusion (Section 5) is provided outlining the main findings and highlighting future work streams.
2. Related work on (automated) drug compounding
ADC has the ability to increase dose accuracy, protect technicians from exposure, needle wounds, and repetitive strain injury (Reference Cerutti, Ledoux, Vantard, Cerfon, Kimbidima, Larbre, Herledan, Lattard, Baudouin, Caffin, Schwiertz, Ranchon and RioufolCerutti et al., 2023; Reference McLeod, Zochowska, Leonard, Crow, Jacklin and FranklinMcLeod et al., 2012; Reference Ratner, Spinelli, Beking, Lorenzi, Chow, Teschke, Le, Gallagher and Dimich-WardRatner et al., 2010). ADCs can also reduce the operational costs of pharmacies by reducing waste (Reference Baan, Geersing, Crul, Franssen and KlousBaan et al., 2022) which has the added benefit of reducing environmental impacts (Reference Baan, Geersing, Crul, Franssen and KlousBaan et al., 2022). Furthermore, by reducing reliance on human intervention in the drug compounding process, ADCs unlock new opportunities, such as automated quality assurance and control as well as improved inventory management (Reference Bond, Raehl and FrankeBond et al., 2002; Reference Gilbert, Kozak, Dobish, Bourrier, Koke, Kukreti, Logan, Easty and TrbovichGilbert et al., 2018; Reference Poppe, Savage and EckelPoppe et al., 2016). Indeed, devices such as the APOTECA chemo can cover up to 70.4% pharmacy workload for inpatient treatment (Reference Iwamoto, Morikawa, Hioki, Sudo, Paolucci and OkudaIwamoto et al., 2017). ADCs have been able to achieve this while simultaneously streamlining production, increasing throughput (Reference Capilli, Enrico, Federici and ComandoneCapilli et al., 2022; Reference Geersing, Pourahmad, Lodewijk, Franssen, Knibbe and CrulGeersing et al., 2023), and reducing both material and labour costs (Reference Yang, Ni, Zhang and PengYang et al., 2023).
These advantages are particularly important for CTX drug preparation, as the cost of these drugs can be considerable, and is only expected to rise (Reference Prasad, De Jesús and MailankodyPrasad et al., 2017). Indeed, preparations for treating cancer and any associated symptoms arising from chemotherapy treatment can range between 50 to 1500 USD per dose, with particularly novel, aggressive, or bespoke treatments costing as much as 30,000 USD per dose (Reference Gupta, Nshuti, Grewal, Sedhom, Check, Parsons, Blaes, Virnig, Lustberg, Subbiah, Nipp, Dy and DusetzinaGupta et al., 2022; Reference Siddiqui and RajkumarSiddiqui & Rajkumar, 2012; Reference Sohi, Levy, Delibasic, Davis, Mahar, Amirazodi, Earle, Hallet, Hammad, Shah, Mittmann and CoburnSohi et al., 2021).
Despite the advantages of ADC systems, there are considerable challenges to their implementation and wider adoption. These are driven either by their limitations, lack of sophistication, or both. The adoption cost for large, self-contained machines (which have their own built in isolator hood for clean uncontaminated air provision) can be prohibitive for many hospitals (Reference Masini, Nanni, Antaridi, Gallegati, Marri, Paolucci, Minguzzi and AltiniMasini et al., 2014). These costs are not limited to the purchase price of the ADC but also staff training, installation, and maintenance. Additionally, ADCs disrupt existing workflows, take up valuable floorspace in aseptic units, and cause disruption to established processes and procedures during the initial adoption period (Reference Cerutti, Ledoux, Vantard, Cerfon, Kimbidima, Larbre, Herledan, Lattard, Baudouin, Caffin, Schwiertz, Ranchon and RioufolCerutti et al., 2023). Many ADC systems require dedicated technical staff for supervision and, while able to handle the compounding, need continuous manual supervision to load drugs, syringes, IV bags, and other consumables inside them (Reference Soumoy and HecqSoumoy & Hecq, 2019; Reference Yoon, Kim and BangYoon et al., 2024). A particular challenge for ADC systems is the end effector used by the robotic arms to grip syringes used for drug dispensing. Indeed, many of the challenges faced by ADC systems are linked to the complexity of including automation in a process that was not originally designed with automation in mind. Vials, syringes, IV bags, etc. have not been designed with robotic operators in mind, nor have the processes to interface with them. For example, ensuring all the drug has been removed from a vial to help reduce wasted product, can be complex to address for ADCs. This is because the vials are covered in human readable text, rather than just a QR or bar code, which greatly complicates computer vision approaches.
2.1. Problem context and design challenge
CTX drugs are commonly shipped in the form of pre-sealed vials with a pierceable rubber septum containing either a lyophilised powder that must be reconstructed or a ready to use liquid solution with a predefined concentration. This process presents a number of challenges (Reference Ness and MartinsNess & Martins, 2022):
-
• Withdrawing material from vials creates a negative pressure, which can affect the efficiency with which the drug is removed.
-
• In sufficiently large vials, the negative pressure can be enough to cause damage to equipment.
-
• Adding liquid to lyophilised vials can over pressurise them, leading to the drug escaping as an aerosol as the needle is removed.
-
• Repeatedly piercing the septum weakens its structural integrity compromising the seal.
In traditional compounding, an operator will use a needle to pierce the septum and extract the drug. This extraction is generally achieved either via peristaltic pumps, usually reserved for larger volumes, or syringes (Reference Soumoy and HecqSoumoy & Hecq, 2019). An analysis of existing ADC review papers (Reference Batson, Mitchell, Lau, Canobbio, de Goede, Singh and LoeschBatson et al., 2020; Reference Cerutti, Ledoux, Vantard, Cerfon, Kimbidima, Larbre, Herledan, Lattard, Baudouin, Caffin, Schwiertz, Ranchon and RioufolCerutti et al., 2023; Reference Soumoy and HecqSoumoy & Hecq, 2019; Reference Yang, Ni, Zhang and PengYang et al., 2023) highlighted that the majority of ADC systems use a syringe operated by a linear actuator. This is true for both self-contained and benchtop ADCs, examples of which are shown in Figure 1. This is linked to the drawbacks inherent to peristaltic pumps such as their lower accuracy at low flow rates (Reference Gibson, Knudsen, Arney, Deng, Sims and PeterfreundGibson et al., 2022). This makes them poor candidates for applications where low volumes of liquid need to be moved at high accuracy whilst minimising the waste of expensive drugs.
Clockwise from top left: Mundus HD (Equashield, 2025), APOTECA Chemo (Cone Health, 2018), SmartCompounders Chemo (SmartCompounders, 2025), MIBMIX C1 Mini (Hemedis, 2025)
Figure 1 Long description
The image contains four separate photos of automated drug compounding systems. Panel A: Mundus HD. A robotic arm is positioned next to a cylindrical device with a label Mundus HD. Panel B: APOTECA Chemo. A large circular machine with a robotic arm attached. Panel C: SmartCompounders Chemo. A device with a yellow bag on top, connected to multiple containers and a screen. Panel D: MIBMIX C1 Mini. A compact machine with several cylindrical containers and a robotic arm.
Furthermore, all peristaltic pumps require tubing. This introduces two challenges (Reference MandelMandel, 2018):
-
1. Any tubing must be primed to remove any air to guarantee the correct dosage is dispensed
-
2. All tubing must be either disposed of or fully flushed to remove any contaminants
In contrast, a syringe is fully disposable after use, and this is, in fact, its intended use. Some liquid remains trapped within the needle once the drug has been dispensed. This dead volume is a function of the needle gauge and length, but is generally minute (ca. 7.255 µL for a 20 gauge 25.4mm long needle ( Needle Gauge Chart, 2026 )) due to the notably reduced length and diameter compared to tubing. Indeed, of the four devices shown in Figure 1, only the MIBMIX C1 makes use of any tubing to dispense drugs. Much like peristaltic pumps, syringe drivers, like that used by the MIBMIX C1, suffer from significant design drawbacks that make them less than ideal for compounding, especially when smaller batches are required.
One additional advantage of using a syringe to withdraw directly through the vial by means of a needle is that syringes can be actuated and primed/purged while connected to the vial. This means that a more precise dose can be withdrawn. This is achieved by tilting the vial and syringe in such a way that any air trapped within the syringe is allowed to float to the top, near the needle, and can then be purged out by depressing the plunger (Reference Ness and MartinsNess & Martins, 2022). The process is described in detail in Figure 2. As can be seen, the manual process is somewhat involved and requires the use of both hands to hold the vial and syringe steady as they are tapped.
Process for manual withdrawal of liquid from sealed vial. From left to right: the syringe is primed with air, the air is injected into the vial to replace the liquid that will be removed, the vial and syringe are inverted and the desired amount of liquid is withdrawn, the syringe gently flicked to force air bubbles to rise to the top, the bubbles are expelled, additional liquid is withdrawn to make up for any losses in the purging process (Reference Hada, Na, Jeong, Choi, Kim and JeongHada et al., 2023)
However, latent errors are present in all compounding processes and can be exacerbated by insufficient quality control measures. As shown in Figure 2, an operator, without realising, may not purge all the trapped air bubbles. As such, ADCs need to be able to guarantee that the drug compounding process is within the allowable margin of error for a specific prescription. Doing so requires a robust monitoring process with a closed feedback loop and continuous data capture to provide an auditable trail (Reference Yaniv, Orsborn, Bonkowski, Chew, Krämer, Calabrese, De La Paz Pacheco Ramos, Palombi, Lim, Tabbara, Masini, Schierl, Bufarini, Peaty and PaolucciYaniv et al., 2017; Reference Yaniv and KnoerYaniv & Knoer, 2013). Cross checking volumetric doses against gravimetric sensor readouts can achieve this goal.
2.2. Observational analysis
In addition to the literature presented, and to better contextualise and understand how aseptic CTX drug manufacturing is conducted in the real world, an observational monocentric study was conducted at a hospital in the United Kingdom. Unlike similar studies (Reference Calvo-Haro, Pascau, Asencio-Pascual, Calvo-Manuel, Cancho-Gil, Del Cañizo López, Fanjul-Gómez, García-Leal, González-Casaurrán, González-Leyte, León-Luis, Mediavilla-Santos, Ochandiano-Caicoya, Pérez-Caballero, Ribed-Sánchez, Río-Gómez, Sánchez-Pérez, Serrano-Andreu, Tousidonis-Rial and Perez-MañanesCalvo-Haro et al., 2021; Reference Eddous, Lamé, Decante, Yannou, Agathon, Aubrège, Talon and Dacosta-NobleEddous et al., 2023) the observational analysis focused on gathering data on drug demand as well as quantifying the person hour cost involved for the aseptic unit, rather than capturing behavioural data. This was driven primarily by the complications of data gathering within aseptic units, where only strictly required personnel are permitted, and the need to minimise disruption.
The analysis was conducted over a three-month period, from September to November 2025. A total of 33 participants across all departments involved in the aseptic compounding unit of the hospital were timed and tracked to understand how their time was allocated each workday. Staff members were divided into three categories based on job title and role. These were: Pharmacists, Technicians, and Assistants.
Compounding tasks were rounded to the nearest 5 minutes with an average time for each drug being derived from the time taken across all the technicians and pharmacists. Similarly, ancillary tasks were tracked and measured using the same approach. These tasks included scrubbing and changing into and out of PPE before and after entering the aseptic unit, cleaning surfaces, waste disposal, stock taking, ordering drugs, and completing paperwork for QA and tracing.
During the three-month period a team consisting of pharmacists and technicians spent a cumulative 570 person hours over the course of the 13-week evaluation period exclusively focused on manufacturing CTX drugs - an average of 44hr/week. In addition to the actual compounding time, the staff spent 2,605 person hours on ancillary and preparatory work related to CTX compounding. This discrepancy between the time actually spent preparing the drugs and all the ancillary tasks that must be performed before and after, highlight the case for ADC systems as a tool to reduce workload requirements. Indeed, of the 2,605 person hours spent on ancillary task over the course of the three months, 580 person hours alone were spent on scrubbing and changing into and out of PPE - almost the same amount of time spent compounding drugs.
Based on the aforementioned findings of Reference Iwamoto, Morikawa, Hioki, Sudo, Paolucci and OkudaIwamoto et al. (2017), that ca. 70% of pharmacy workload can be covered by an ADC system, the hospital would stand to free up approximately 30 person hours per week. This saving is further magnified when taking the considerable amount of ancillary time that accompanies the actual compounding time. Reducing the scrubbing and changing time by 70% alone would greatly free up personnel, enabling them to focus on higher-value tasks.
However, ADC adoption remains low (Reference Ganio and JerryGanio & Jerry, 2022). The inability of current ADC systems to act unsupervised, with minimal human intervention, is a primary barrier. According to a survey of pharmacy professionals (Reference Yaniv, Orsborn, Bonkowski, Chew, Krämer, Calabrese, De La Paz Pacheco Ramos, Palombi, Lim, Tabbara, Masini, Schierl, Bufarini, Peaty and PaolucciYaniv et al., 2017), one of the pain points identified with existing ADCs is their inability to work with a wide variety of syringes limiting their utility and requiring human oversight, which in turn greatly reduces the benefits of adoption.
3. Design development, prototype,and evaluation
Building on the challenges and requirements outlined in Sections 3.2 and 2.2 this section details the development and evaluation of a robotic end effector capable of supporting ADC systems. Several design options were explored at the outset, with three initial concepts developed to address the functional requirements identified. These concepts were systematically iterated and evaluated to arrive at a final concept. The methodology used for this process was in line with the double-diamond approach. This concept is presented in Section 3.2.
3.1. Overview of design challenges
A robotic arm can be used to address many of the challenges identified, indeed, many ADC systems currently on the market make use of robotic arms to some capacity (Reference Batson, Mitchell, Lau, Canobbio, de Goede, Singh and LoeschBatson et al., 2020; Reference Cerutti, Ledoux, Vantard, Cerfon, Kimbidima, Larbre, Herledan, Lattard, Baudouin, Caffin, Schwiertz, Ranchon and RioufolCerutti et al., 2023; Reference Soumoy and HecqSoumoy & Hecq, 2019; Reference Yang, Ni, Zhang and PengYang et al., 2023). However, as previously noted, many ADC systems currently lack an end effector with which to manipulate the syringe directly. For example, the APOTECA Chemo, can be seen in Figure 1 using a separate syringe priming mechanism that uses a linear actuator. The robot arm in this device serves merely to load the syringe and vials into the correct receptacles so the compounding process can be undertaken in isolation from human operators. This approach adds complexity, cost, and increases manufacturing time by introducing additional steps. The reliance on linear actuators seems to be predominantly driven by the lack of alternatives suitable for adapting to a robotic end effector. Such an end effector would need, based on the previously analysed literature, the following capabilities:
-
1. Be capable of withdrawing enough liquid to complete a full batch, reducing the number of times the septum on the vials needs to be pierced.
-
2. Prevent cross-contamination between batches, preferably through single use syringes, which are in line with current compounding practice.
-
3. Reduce wastage by having a sufficiently high degree of accuracy when withdrawing, minimising waste caused by overfilling.
-
4. The ability for the robotic arm to quickly switch between batches using different drugs quickly without risking cross-contamination, thereby supporting higher manufacturing rates.
-
5. Based on the identified drawbacks, there should be as little tubing between the syringe and the vial and/or the destination container.
-
6. A compact form factor to allow for use in smaller installations such as within existing aseptic isolators which also decreases the risk of collisions with other items.
-
7. Centre of mass close to robot’s “wrist” in order to improve positional accuracy and reduce the risk of exceeding the robot’s maximum load rating when the arm is at full extension.
-
8. Accurate, analogue adjacent movement of the syringe plunger to give granular control both during withdrawal from the vial and dispensing into the destination container.
-
9. The end effector must be sufficiently robust enough to withstand the forces of repeated insertion into vial septa, IV bag ports, etc.
-
10. Lastly, the end effector should not materially slow down production compared to manual compounding approaches.
3.2. Proposed design
Using the existing linear actuators as a baseline, seven early-stage end effector concepts were developed using the S.C.A.M.P.E.R. technique. These were evaluated, merged, and edited to arrive at three concepts that were deemed capable of handling the design challenges enumerated in Section 2.2. On evaluating the three concepts, a single design based on the principle of a lead screw was selected for further development. The selection decision was based predominantly on the design’s ability to provide accurate dosages and the small form factor which could bring the centre of mass closer to the robot’s wrist.
Figure 3 shows the final design of the end effector. As can be seen from the figure, the design is based on an interrupted screw mechanism driven by a larger plunger gear which fits around the screw thread cut into the plunger actuator. This gear is in turn connected to a motor gear which actuates the whole mechanism. The main body of the end effector holds the motor in place as well as holding the plunger gear in place by means of a ring sitting on top of the gear.
The syringe is inserted into the end effector by fully extending the plunger, placing the plunger actuator’s latch around the base of the syringe’s plunger flange and aligning the interrupted screw thread on the actuator with the threads on the plunger gear. The motor can then be triggered to slowly lower the plunger allowing the syringe’s barrel to be lowered into the recess below the plunger gear. This allows the syringe to be twisted and locked by means of two locking tabs that fit around the barrel’s flange. Once this is done, the end effector can be calibrated by allowing the plunger to go fully down and the end effector is ready for autonomous use by the robot.
CAD assembly of final design concept showing both left and right side of the robotic end effector with labels highlighting the principal components
In addition to the end effector’s ability to accurately control a syringe, it is also equipped with a quick disconnect feature. The wrist magnet and locator, as well as the spring-loaded magnetic contact pins, allow the end effector to be quickly docked and removed by the robotic arm itself without human intervention. The spring-loaded pins provide power and control commands to the motor while the wrist magnet ensures a secure fit while the end effector is attached to the robot’s wrist. The design was developed with additive manufacturing in mind. This was driven by the accessibility of the technology, its ability to manufacture complex geometries and the relatively low costs involved, especially for small volume production.
The proposed end effector was designed to use the Dynamixel XL330-M288-T motor. This motor is uniquely well suited to this application due to a combination of capabilities (RoboSavvy, 2025): The relatively high 0.52Nm stall torque allows the motor to operate the plunger despite the unlubricated mechanism. The gear ratio between motor and plunger gear is 1:2, and the plunger actuator thread pitch is 3mm. Given that a 5mL syringe has a total plunger travel of ca. 58mm, it would take approximately 19 revolutions of the plunger gear to fully fill or empty the syringe.
As the Dynamixel has a no load RPM of 104, this gives a theoretical maximum speed of 22s to fully fill or empty the syringe. Furthermore, the built in encoder has a resolution of 0.0879°. as such the minimum theoretical resolution for the plunger movement is in steps of 0.37μm, which for a 5mL syringe results in 0.032μL of liquid being moved in each step. The flexibility and compressibility of the syringe plunger, plunger actuator and the backlash between plunger and motor gears will significantly impact this value.
In addition to the design’s ability to dispense with high accuracy (0.032μL) and suitable speed (0.227mL/s), the complete assembly, without the syringe, weighs ca. 65g and ca. 70g with both syringe and needle, well below the maximum payload capacity of even small robotic arms. One additional advantage of this design is the ease with which it can be adjusted to accept larger or smaller syringes. By using a parametric design approach to generate the gears and motor position, it becomes a simple matter of changing the gear ratios between motor and plunger gear, within the CAD model, which in turn allows for a larger diameter syringe to be fitted. The thread pitch on the plunger actuator can also be adjusted to maintain the same plunger speed. This parametric approach makes the design flexible and easy to adjust to the specific needs of individual compounding pharmacies.
One final benefit of this design is the high baud rate of the Dynamixel (max. 4,500,000 bits per second). This high data transfer speed gives it excellent responsiveness meaning it is ideally suited to integration within a closed control loop. A closed control loop with a gravimetric sensor would allow the motor to be quickly slowed or stopped once the correct amount of drug has been dispensed.
4. Design evaluation
Figure 4 shows a prototype end effector installed on a modified Niryo Ned 2. The Ned 2’s wrist was modified, adding a collet to house a set of spring-loaded contact pins to interface with those on the end effector, allowing power and communication with the Dynamixel motor. This setup allows the Ned 2 to quickly swap out the end effector. The Ned 2’s payload capacity at full extension is 300g, well above the end effector’s weight even if larger syringes were used. No detrimental effects to the robot’s movement were observed. This was tested by moving the robot through its degrees of freedom to identify if the end effector would restrict movement more than the standard gripper attachment. In addition, speed and repeatability tests were run where the robot moved the needle between multiple points to test whether the end effector impacted the robot’s tracking. To test the speed and repeatability of the robot arm, each joint was manipulated manually, moving them to random positions. The position of the filling port on the IV bag was fixed (as shown in Figure 4). Once the robot arm was moved to a random position, it was ordered to move the syringe in line with the filling port on the IV bag and insert it in there. Again, no negative impact was observed when running this test nor was the robot appreciably slowed down. The test was run with 50 different random positions for the robot joints and in all cases it found the filling port and correctly aligned the needle.
In addition to showing the end effector mounted on the robot arm, Figure 4 shows the end effector being used to fill an IV bag. Using water as a test liquid it was observed that the end effector could be fully emptied in ca. 25s using a blunt 20mm needle and 5mL syringe. This is only slightly slower than the theoretical maximum speed of 22s, though the viscosity of the liquid will impact this speed.
A full test of the robot was performed to evaluate its capability compared to human operators. The robot took approximately 3min 50sec to: 1. Grab the syringe end effector from the tool post; 2. Pierce the septum on a vial; 3. Fill the 5mL syringe completely with liquid; 4. Locate and pierce the septum on an IV bag; 5. Fill the IV bag with the full 5mL contained within the syringe; 6. Replace the syringe end effector in the tool post; 7. Grab a gripper end effector from the tool post; 8. Grab the IV bag and hand it to a human; 9. Replace the gripper in the tool post and return to standby to restart the cycle. This time taken is highly comparable to that of human operators as described in Section 2.2, where the shortest amount of time for compounding was approximately 5 minutes.
Three photos showing a prototype of the design presented in Section 3.2. From left to right the photos show the left side of the end effector, the right side, and the end effector in use filling an IV bag
Based on initial lab tests and observations, the end effector design has proven to be successful. It meets the requirements described in Section 3.1 as enumerated below.
-
1. With a 5mL syringe the end effector is only capable of handling small batches of a few IV bags; however, as the design can easily be adjusted to accommodate larger syringes, larger batches should be possible. This is in line with the recommendations of Reference Yaniv, Orsborn, Bonkowski, Chew, Krämer, Calabrese, De La Paz Pacheco Ramos, Palombi, Lim, Tabbara, Masini, Schierl, Bufarini, Peaty and PaolucciYaniv et al. (2017) who highlighted the need for greater syringe compatibility in ADC systems.
-
2. The end effector does rely on single use syringes and needles eliminating the risk of cross contamination present with other approaches that rely on tubing (Reference MandelMandel, 2018).
-
3. With a theoretical accuracy of 0.37μm the end effector more than meets this expectation (Reference Cerutti, Ledoux, Vantard, Cerfon, Kimbidima, Larbre, Herledan, Lattard, Baudouin, Caffin, Schwiertz, Ranchon and RioufolCerutti et al., 2023). Indeed, further testing may reveal that decreasing the accuracy to increase filling/emptying speed could be beneficial.
-
4. By implementing a quick release in the robotic end effector, it is possible for the robot to quickly switch between different syringes, each designated to a specific drug and thus quickly shift between batches without the risk of cross contamination (Reference MandelMandel, 2018; Reference Yaniv, Orsborn, Bonkowski, Chew, Krämer, Calabrese, De La Paz Pacheco Ramos, Palombi, Lim, Tabbara, Masini, Schierl, Bufarini, Peaty and PaolucciYaniv et al., 2017).
-
5. The end effector makes no use of tubing and does not interfere with the robot’s free movement. As a result, it can use a shorter 20mm needle further reducing any “dead volume”. This helps maintain drug withdrawal accuracy, reducing the risk of trapped air (Reference Hada, Na, Jeong, Choi, Kim and JeongHada et al., 2023).
-
6. The end effector’s low weight of 70g with a syringe, compact form factor (110mm height, 80mm depth, and 50mm width) means it can easily fit on small benchtop robotic arms and within most isolators this helps address cost concerns, improving future access to ADCs (Reference Masini, Nanni, Antaridi, Gallegati, Marri, Paolucci, Minguzzi and AltiniMasini et al., 2014).
-
7. The majority of the weight of the end effector comes from the magnet and motor which are strategically positioned close to the robot’s wrist. The syringe is, however farther from the wrist but the low total weight means there is no real impact on the robot’s ability to operate.
-
8. While not analogue as the Dynamixel works in set steps, the 0.032μL dispensing accuracy is likely far higher than any drug’s margin of error (Reference Yaniv, Orsborn, Bonkowski, Chew, Krämer, Calabrese, De La Paz Pacheco Ramos, Palombi, Lim, Tabbara, Masini, Schierl, Bufarini, Peaty and PaolucciYaniv et al., 2017; Reference Yaniv and KnoerYaniv & Knoer, 2013). Even if the actual accuracy is two orders of magnitude lower it would still be within allowable parameters.
-
9. Stress testing on the end effector revealed no negative effects either to the robotic arm of the end effector itself even after repeated insertions into IV bags and vial septa.
-
10. The end effector does not appear to be a significant bottleneck to the automated compounding process, with the robotic arm taking approximately 3min 50sec, compared to a human operator’s 5 minutes for a comparable operation based on a comparison of observations of the end effector and human pharmacists and technicians (as detailed in Section 2.2).
Despite the design’s clear promise, there are several drawbacks to the design. The most glaring of which is the difficulty with which syringes are installed in the end effector. At present, this is an involved process that requires two hands and direct control over the robot to feed the plunger actuator in order to lock the syringe in place.
This limitation does not negatively impact the robot’s ability to operate but will undoubtedly become a hinderance to adoption. Developing the design in such a way that routine maintenance of the syringes is easier, or even automated, would greatly increase the value of the end effector. This is particularly true for compounding pharmacies that wish to switch between multiple drugs repeatedly, as having one end effector for each drug may prove to not be practical in those circumstances.
Additionally, while the design is flexible and able to adapt to different syringe sizes, each syringe requires a bespoke end effector to be designed for that syringe diameter as well as desired accuracy and feed rate. A design that could accept a range of syringes would, again, greatly benefit adoption.
5. Conclusion
This paper presents a novel design for a robotic end effector capable of accurately, safely, and efficiently handling CTX drugs in aseptic compounding pharmacies. The design presented is a small (110mm height, 80mm depth, and 50mm width), lightweight (ca. 70g), and accurate (0.032μL) end effector that can be readily mounted on even small desktop robotic arms. The end effector can be manufactured using material extrusion additive manufacturing and can be adjusted to work with syringes of different diameters.
Though no tests to failure were conducted the testing highlighted that it is capable of repeatedly piercing vial septa and IV bag diaphragms over the course of multiple days. The end effector also did not interfere in any appreciable way with the control, speed, accuracy, or repeatability of the robotic arm it was installed on.
Future work should attempt to address the design limitations highlighted, namely the need to adjust the end effector to accept larger diameter syringes rather than natively being able to accommodate a range of sizes. Additionally, the process for installing and removing of the syringe from the end effector is relatively involved. It requires the use of both hands to hold all the components together and the balancing of the plunger actuator over the plunger gear. Furthermore, it necessitates a second person to trigger the robot to begin screwing the plunger actuator into the plunger gear while the first holds everything aligned. As such, the end effector would greatly benefit from redesign to streamline and simplify the syringe installation and removal process. This would have the benefit of greatly improving user friendliness and maintainability of the end effector over time.
Acknowledgement
The work presented within this paper has been undertaken as part of UKRI grant EP/X025470/ and EP/X525650/1.
Open access
