A three-year study of a first-generation chemotherapy ...

PRACTICE REPORT Chemotherapy-compounding robot

PRACTICE REPORT

A three-year study of a first-generation chemotherapy-compounding robot Zubeir Nurgat, Dima Faris, Maher Mominah, Arris Vibar, Abdulrazaq Al-Jazairi, Sheena Ewing, Mohammed Ashour, Shrouq Kamel Qaisi, Sakra Balhareth, and Ahmed Al-Jedai

P

harmacy has been at the forefront in adopting new technology to enhance medication safety and improve the efficiency of patient care services.1 Electronic medication management and the use of automated dispensing cabinets and carousels are now practice norms in hospital pharmacies of all sizes due to technological advancements and developments. Recently, a major focus of attention on automation technology in pharmacy has been the use of i.v. compounding robots, in particular those used for the preparation of i.v. antineoplastic medications.2-5 One feature of robotic systems that offers an advantage over manual (i.e., volumetric measurement–based) preparation is the use of gravimetric measurements to precisely weigh the quantities of antineoplastic medications and other products needed to prepare a dose. The robot’s accuracy is predetermined (a typical margin of error is ±5%); any preparation

Purpose. Results of a performance evaluation of an automated system for compounding antineoplastic preparations are reported. Methods. Three years after the pharmacy department of a hospital in Saudi Arabia installed an i.v.-compounding robot (CytoCare, Health Robotics), data captured by the pharmacy information system and the machine’s integrated software were analyzed to assess the performance of the robot in terms of compounding accuracy, days of operation, and downtime. Results. The robot was used to prepare 3.82%, 10.80%, and 13.79% of selected antineoplastics compounded in 2010, 2011, and 2012, respectively. The robot failed to meet the specified dose accuracy range of ±5% in compounding 3 of 337 chemotherapy preparations (0.9%) in 2010, 349 of 1516 preparations (23%) in 2011, and 460 of 2993 preparations (15%) in 2012. The robot was operational on 40%, 39%, and

that exceeds the margin is rejected, or “failed.” This level of accuracy and re-

Zubeir Nurgat, B.Pharm., M.Sc., BCOP, is Manager; Dima Faris, B.Sc.Phm., is Pharmacist; Maher Mominah, B.Sc.Phm., is Specialist, Pharmacy Automation and Support Services; Arris Vibar, B.Sc.Phm., is Pharmacist; and Abdulrazaq Al-Jazairi, Pharm.D., M.B.A., BCPS, is Head, Medical/Critical Care, Pharmaceutical Care Division, King Faisal Specialist Hospital & Research Centre (KFSH&RC), Riyadh, Saudi Arabia. S heena E wing , B.S c ., M.S c ., RGN, OCN, is Program Director, Nursing, KFSH&RC. M ohammed Ashour, B.Sc., M.Sc., is Manager, Ambulatory Care Oncology Pharmacy; Shrouq Kamel Qaisi, Pharm.D., is Clinical Pharmacy Specialist, Ambulatory Care; and Sakra Balhareth, Pharm.D., BCPS, BCACP, is Clinical Pharmacy Specialist, Drug Information Center, Pharmaceutical Care Division, KFSH&RC. Ahmed Al-Jedai, Pharm.D., M.B.A., BCPS, is Director, Pharmaceutical Care

1036

Am J Health-Syst Pharm—Vol 72 Jun 15, 2015

61% of available workdays in 2010, 2011, and 2012, respectively. Robot throughput relative to the pharmacy’s manual compounding process was low, with substantial medication waste resulting from dose preparation failures. Implementation challenges included workflow disruptions due to robot downtime, mechanical issues (e.g., robot arm-clamping failures), difficulty obtaining gravimetric data for some drugs, and the need to recalibrate the device to accept i.v. bags, syringes, and medication vials incompatible with manufacturer specifications. Conclusion. The introduction of a chemotherapy-compounding robot for preparation of patient-specific i.v. antineoplastic drugs had a limited efficiency impact in practice. This solution, with its numerous limitations and technical difficulties, is not yet mature enough for universal adoption. Am J Health-Syst Pharm. 2015; 72:103645

producibility is more stringent than that imposed by the United States

Division, and Associate Professor, College of Medicine, Alfaisal University, KFSH&RC. Address correspondence to Dr. Al-Jedai ([email protected]). The following are acknowledged: the hospital administration at King Faisal Specialist Hospital & Research Centre (KFSH&RC) for approving the purchase of the robotic compounding system; the KFSH&RC oncology pharmacy and i.v. admixture room staff; and the KFSH&RC pharmacy informatics and automation staff, in particular manager Amar Hijazi. The authors have declared no potential conflicts of interest. Copyright © 2015, American Society of Health-System Pharmacists, Inc. All rights reserved. 1079-2082/15/0602-1036. DOI 10.2146/ajhp140256


Pharmacopeia (USP) chapter 795 standards for compounded products, which allow a variance of ±10%.6 Under the USP standards, a deviation from the targeted concentration of greater than 10% is considered an imprecision or dilution error. The manual compounding of i.v. preparations for anesthesiology and chemotherapy has been reported to involve higher rates of discrepancy from designated ranges of acceptability.3,7 A study that assessed the accuracy of volumetric technique in the preparation of chemotherapy doses found that 71.7% of the doses prepared were within ±5% and 87.4% were within ±10% of the ordered dose.8 The study authors suggested that “the process of volumetric technique alone is not sufficient to accurately prepare chemotherapy doses, and a focus of improving the process of preparing i.v. medications needs to be undertaken.”8 The robot ensures the sterility of antineoplastic preparations because it operates in an aseptic (i.e., International Organization for Standarization [ISO] class 5) environment and prevents drug cross-contamination by using sterile syringes for each transfer. In addition, the compounding process involves numerous safety verification checks. The robot photographs the barcode on the medication label and compares it with barcode data in its database. The robot labels every preparation it makes, minimizing the potential for errors.9,10 The data captured during the compounding process are retrievable and provide an important tool for quality control purposes. The robotic i.v. compounding environment also offers the advantage of minimizing the occupational exposure of healthcare professionals to antineoplastic drugs that are hazardous and require careful, safe handling.11-13 Despite the numerous theoretical advantages of an automated process over manual preparation of i.v. antineoplastic medications, the adoption

of robots for chemotherapy compounding remains low. According to the 2011 ASHP national survey of pharmacy practice in hospital settings, robotic chemotherapycompounding devices were used in only 0.1% of the hospitals surveyed.1 An even smaller percentage of centers around the world have adopted chemotherapy-compounding devices in hospital pharmacy practice settings.5 There may be many reasons for the very low rate of adoption of robotic chemotherapy-compounding devices for the preparation of i.v. antineoplastic medications. These include the prohibitive cost of initial system implementation, doubts about operational efficiency (i.e., how many doses can be prepared in a given period of time by a robotic versus a manual process), a lack of knowledge regarding the long-term benefits of robotic compounding, and the unknown risks of adopting new robotic technology.2,3 This report describes the experience of implementing and utilizing a first-generation chemotherapycompounding robot for the preparation of i.v. antineoplastic agents over a period of three years at King Faisal Specialist Hospital & Research Centre (KFSH&RC) in Riyadh, Saudi Arabia. Background KFSH&RC is a 985-bed tertiary care hospital that provides comprehensive healthcare in all specialties, including hematology and oncology. The pharmaceutical care division operates 24 hours a day, seven days a week, and supports the oncology infusion clinic, which is open five days a week (the clinic is closed on weekends and public holidays). “New-start” chemotherapy preparations are performed on weekends only in the instance of complex cases and combined-modality treatment protocols. The increased workload associated with the preparation of

antineoplastic agents is a persistent problem. During the five-year period of 2004–08, the number of antineoplastic medications prepared increased by 30% to an average of approximately 275 chemotherapy doses per day. In 2008, a work measurement study was conducted at KFSH&RC by an independent industrial engineering specialist, who looked at various tasks performed by the pharmacy staff. The pharmacy staff was rated on their performance and activities by the engineer, who considered the staff ’s pace of work and assigned a performance rating factor (RF). The “normal” RF is 100%. The normal time to perform a manual preparation was calculated by multiplying the observed time by the RF. The staff was briefed by the supervisor about the nature of the study and took part voluntarily. Eight pharmacy technical staff members were randomly selected to participate in the study; two had three years and six had five years of experience in the aseptic preparation of compounded sterile products. All had completed annual aseptic training courses. Observations took place at various times on six randomly selected days without advance notification of the selected staff members. Antineoplastic medication doses prepared in i.v. syringes, chemotherapy i.v. bags with a volume greater than 250 mL, and glass bottles were included in the evaluation. The antineoplastic medication preparation process was observed and timed continuously from the moment the technician started compounding until the finished product was handed to the designated product checker. The work measurement data were used to calculate the full-time equivalent (FTE) requirements to handle the current volume of work and reengineer the i.v. chemotherapy service for automation use. We began our search for a suitable robot for the preparation of i.v. antineoplastic


1037


agents in June 2008. Our search for the most suitable robot was hindered by a lack of published data and narrative relating to the “real-life” experience of implementing robotics for i.v. antineoplastic preparation. In July 2009, we acquired a CytoCare robot (Health Robotics, Bolzano, Italy) through a local vendor contracted by the manufacturer of the robot. The Health Robotics engineers carried out system installation, implementation, staff training, and preliminary validation, which have been described by others.2,9,10 In March 2010, we went live with robotic preparation of a limited number of antineoplastic medications in i.v. bags (Baxter, Deerfield, IL) with volumes of 25–1000 mL. The list of antineoplastic medications included solutions of cisplatin (50 mg/50 mL), cyclophosphamide (500 mg/20 mL), cytarabine (2000 mg/40 mL), etoposide (200 mg/10 mL), ganciclovir (500 mg/10 mL), ifosfamide (2000 mg/20 mL), and vincristine (1 mg/1 mL and 5 mg/5 mL). This list was determined based on the availability of antineoplastic medications in vials, robot specification requirements, workload priorities, and the availability of specific-gravity data. Methods Computer-generated data obtained from the pharmacy information system and the CytoCare system’s integrated software (CytoPlan, Health Robotics) over a three-year period (2010–12) were analyzed. Pharmacy staff used computerized physician order entry (CPOE) to collate all chemotherapy orders entered. At the time of the study, chemotherapy orders were not entered by the physician through CPOE; instead, they were entered manually by a pharmacist after verification of the order. We assessed the accuracy of the robotically prepared antineoplastic medication doses by looking at pass and fail rates (i.e., how often the robot prepared doses within or 1038

outside the designated quality parameters of ±5%). We also quantified system uptime, assessed reasons for downtime, and evaluated the FTE staffing requirements to run the robot. We compared the productivity of robotic and manual processes by comparing the number of preparations performed by the robot during its operational time and the number of manual preparations. We classified reasons for nonoperation of the robot as either technical or nontechnical. Technical problems were defined as those related to the function of the robot, including mechanical or software-associated difficulties such as needle-search problems (i.e., the camera’s failure to locate the needle before injecting liquid into or aspirating liquid from the bag), small dose-quantity errors (i.e., errors of 250 mL (n = 49)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49

S.D.

Mean

Normal time

Figure 2. Normal times for manual preparation of 49 chemotherapy doses in infusion bags with a volume of greater than 250 mL at the study site. Normal times were calculated by multiplying the actual preparation time by a performance rating factor (RF) based on observations of the pharmacy staff’s work pace (normal RF, 100%).



ing safety, accuracy, and efficiency. In implementing the first-generation robotic chemotherapy-compounding device for the preparation of patientspecific i.v. antineoplastic doses, we experienced numerous challenges, especially with regard to time efficiency. The compounding robot’s lack of speed remains its Achilles’ heel. During the three-year study, the number of patient-specific doses prepared by the robot in an approximately seven-hour shift rarely exceeded 50, which is comparable to the reported experiences of other users of robotic systems.4,5 The average number of doses prepared by the robot in a seven-hour shift was comparable to the number manually produced in two to three hours by trained and experienced pharmacy staff. Other research indicates that the mean drug preparation time is increased at least twofold with the use of a robot versus a manual preparation process.3

In busy infusion centers, such delays can create a backlog of patients waiting to be treated, which is not acceptable to either patients or hospital administrators. In the inpatient environment that served as the study site, the use of the robot required workflow adjustments to maximize its functionality and avoid periods of idleness. Rescheduling of workflow in 2012 to allow for the manual reconstitution of lyophilized powdered medications prior to loading into the robot reduced failure rates associated with the shaker mechanism and increased the number of preparations by the robot. Interestingly, similar workflow adjustments to enable advance reconstitution of lyophilized powders were described by other researchers using a different robot for the preparation of patientspecific chemotherapy.4 At KFSH&RC, preparing antineoplastic medications with relatively long stability times (i.e., vincristine

and ganciclovir), as well as standard doses, in a batching process maximized the utility and throughput of the robot. At the beginning of the study, vincristine was dispensed as a final product in 2-mL syringes instead of in small-volume parenteral containers (minibags), which is now the recommended practice. After reengineering our practice to permit dispensing of vincristine in minibags,14 we were able to utilize the robot for the preparation of vincristine in batch mode in the latter part of 2012. Batch preparation of i.v. antineoplastic medications in minibags rather than syringes appears to be a promising approach to improving the efficiency of the robot. Approximately 23% and 15% of the robot’s preparations failed to meet the more stringent accuracy standard (±5%) in 2011 and 2012, respectively. The secondary data analysis using the ±10% accuracy limit indicated improved failure rates

Figure 3. The blue line shows the trend in daily number of chemotherapy doses prepared by the compounding robot on 62 consecutive workdays in 2012; each corresponding red bar depicts the number of doses that did not meet the specified accuracy standard (±5%). 50 45

No. Doses Prepared

40 35 30 25 20 15 10 5 0 1

5

10

15

20

25

30

35

40

45

50

55

60

Day


1041


(6.1% and 3.5% in 2011 and 2012, respectively); the associated reduction in waste could justify a shift to the more liberalized accuracy standard. A 10% variance is still considered an acceptable error range, and such variances are rated as inconsequential deviations with little or no potential for patient harm.3,4 It has been reported that standards for robotic preparation of i.v. antineoplastic medications encompassing an accuracy variance of ±10% have been adopted in practice.3,4 Failed preparations created a financial impact because these particular doses were not manually adjusted if they failed to meet the robot’s accuracy range of ±5%, even if the variance was below ±1% or was due to a minor mechanical malfunction of the robot. Such doses were discarded in the robot’s cytotoxic bin, and replacement doses were

prepared manually in the class II biological safety cabinet (BSC), which created additional work. Although our failure rate at the accuracy limit of ±5% was high, we had selected less-expensive and mostly generic medications for robotic preparation; this enabled us to mitigate losses due to waste. Furthermore, we did not use closed-system transfer devices (CSTDs) for the preparation of i.v. antineoplastic medications, to protect staff from exposure as recommended.11-13 The robot can be considered a closed system since it operates in an ISO class 5 environment, which eliminates the need for CSTDs. The reduction in the usage of CSTDs in the preparation of i.v. antineoplastic medications is an important factor when considering overall cost efficiency. In addition, the use of CSTDs to extend beyond-use dating of drugs

as part of vial optimization programs can further reduce the cost of the preparations by the robot.1 We prepared the doses in i.v. bags of 25–1000 mL rather than syringes due to a variety of reasons. The first reason was the persistent issue of air bubbles in syringes, which was related specifically to small volumes of 0.2 and 0.3 mL. This problem was due to the robot attempting to withdraw the drug from the vial on multiple occasions, resulting in the formation of air bubbles in the final product; in such cases, adjusting the syringe manually would have exposed the staff to antineoplastic drugs through aerosol exposure and spills. Second, the current version of the robot does not cap syringes during preparations; these syringes had to be capped manually, increasing the risks of staff needle-stick injuries and antineoplastic exposure. Third, the robot

Figure 4. Monthly numbers of chemotherapy doses prepared by the coumpounding robot over three years. An asterisk indicates that no doses were robotically prepared in a given month. 600

500

No. Doses Prepared

400

300

2010 2011 2012

200

100

0

* Jan

* Feb

* Mar

Apr

May

Jun

Jul

Month

1042


Aug

Sep

Oct

Nov

Dec


uses 3-, 10-, and 60-mL syringes only; this was the main disadvantage, as the majority of our antineoplastic doses required syringe sizes of 20 or 30 mL, and nursing staff were reluctant to administer chemotherapy in 60-mL syringes due to the increased risk of repetitive strain injury. The downtime of the robot was attributed to a variety of technical and nontechnical issues. In 2010, during the installation and testing phase, we encountered difficulties with calibration of the robot’s sensi-

tivity due to problems with external exhaust facilities resulting from installation of the robot and the BSC in the same negative-pressure ISO class 7 cleanroom (the robot was initially located adjacent to the BSC to address workflow issues caused by space limitations and to facilitate planning for robot downtime). It was soon realized that the capacity of the exhaust fan to maintain negative pressure was insufficient. To correct the problem, the combined requirements of the BSC and the robot (in

Table 1.

Number (%) of Robot-Compounded Chemotherapy Doses Rejected, by Year and Variance Limit Variance Limit Year

n

2010 2011 2012

337 1516 2993

≤5%

>5% but