The automation of the compounding of mixtures for intravenous infusion from approved drugs (e. g. mixtures for parenteral nutrition [PN], chemotherapy) can improve the quality of pharmaceutical preparation compared to manual processing (1). To some extent, it can increase the productivity especially when it’s coupled to the centralization of preparations (2, 3). These automated compounding devices [ACD] Henri Becquerel Centre are usually classified:
Depending on the method used for calculating the quantities to be delivered (e. g. gravimetric or volumetric method);
Depending on the type of the pumping system (e. g. peristaltic pump or piston pump);
Depending on the range of quantities that may be pumped accurately (e. g. macrocompounders for volumes of 2 mL–4 L, microcompounders for volumes of between 0.1 mL and 500 mL and micromacrocompounders for volumes from 0.1 mL to 4 L);
Or by the number of solutes that can be used by the ACD for compounding.
These ACD receive a factory design qualification (DQ) to ensure they meet the specifications established by the manufacturer. After installation on the production site and in addition to the installation qualification (IQ) and operational qualification (OQ), it is recommended performing performance qualification (PQ) to check that the ACD is in practice able to function effectively and reproducibly, based on its specifications and for specific operations as assigned.
For PN ACD, these performance qualifications are listed in the US Pharmacopoeia Chapter “Pharmaceutical Compounding – Sterile preparations” under “verification of automated compounding devices for parenteral nutrition compounding” (4). It states that verification of the accuracy, precision and reproducibility of the volumes delivered by the controller could be performed for this purpose.
The aim of this study is to qualify two new ACD for the compounding of PN mixtures.
Materials and methods
A first single pump ACD MediMixmini® R 4010 (Impromediform GmbH) with a 4-way kit MF 4014 (Impromediform GmbH) consisting of four perfusors with droppers and a 50 mL piston device was tested. This controller is designed for pumping volumes ranging from 5 to 9,999 mL with a 1 mL step.
A 12-pumps ACD MediMixmulti® R (Impromediform GmbH) with MF 41204 (central tube), MF 41205 (tubes and fittings), and MF 41207 kits (syringes) (Impromediform GmbH) was also tested. This controller is adapted to deliver volumes ranging from 0.5 to 9,999 mL with a 0.1 mL step.
Solutions and emulsions used during the single pump ACD setup and qualification were:
A 50 % dextrose solution (DE) in 500 mL glass bottles (B Braun) with a density ρG=1.18 g/mL and a viscosity ηG ≈ 5.14 mPa•s at 25 °C (manufacturer data);
A 20 % lipid emulsion (Medialipide®, B Braun; LP) in 500 mL glass bottle with a density ρL=0.99 g/mL and a viscosity ηL ≈ 2.36 mPa•s at 25 °C (manufacturer data);
Water for injections (WA) in 1 L glass bottle (Lavoisier) with a density ρL=1.00 g/mL and a viscosity ηL ≈ 1 mPa•s at 24 °C (manufacturer data);
Solutions and devices used for the qualifications of the 12-pumps ACD were listed on Table 1.
Single pump ACD
Setting of the single pump ACD
The pumping speeds, the delivery speeds, and the pause time between the pumping and the delivery steps were set in order to obtain the best precision and the best accuracy in quantities delivered for each of the three components. For this purpose, each of the three parameters was adjusted independently of the two others that were left in a fixed value judged the least unfavourable. The pump speed was set between 50 and 100 mL/min (leaving the delivery rate at 50 mL/min and the pause time at 2,000 ms) and the delivery rate was then set between 50 and 200 mL/min (leaving the pumping speeds at 50 mL/min and the pause time at 2,000 ms). The pause time was set between 1,200 and 300 ms (leaving the pumping and delivery speeds to 50 mL/min), only for the 50 % dextrose. For each speed or each pause time explored, a volume of 50 mL was requested three times.
The pumping speed, the speed of delivery and the pause time that allowed obtaining the best accuracy and better accuracy for three pharmaceutical products were selected for the performance qualifications of the single pump controller (speeds cannot be changed during production on this controller).
Performance qualification of the single pump ACD
Once the single pump ACD has been set, PQ was performed to assess the repeatability and the reproducibility of pumping operations with the three drugs. They consisted of a series of three pumping at three levels of volume (5; 50; 500 mL) for three products (DE, WA, LP), over 3 days. Precision, intermediate precision, and accuracy were then calculated.
Performance qualification of the 12-pumps ACD
Factory speed settings have been kept for performance qualification of the 12 pumps ACD. The manufacturer of the ACD performed a calibration during operational qualification (i. e. before performance qualification). Performance qualifications consisted of a series of three pumping at two volume levels (0.5, 5.0 mL) for eight products (PH, PO, NA, MG, ZN, OE, VE, VI), and three volume-levels (0.5, 5.0, 50 mL) for four products (WA, CA, AA, DE), over 3 days. Requested volumes were recovered at syringe outlets just before a non-return valve. Precision, accuracy and intermediate precision were then calculated.
Three batches (one per day) of five standardized bags for neonatal parenteral nutrition (5) containing 50.4 mL of DE, 52 mL of AA, 3.8 mL of NA, 4.3 mL of PO, 8 mL of CA, 6 mL of PH, 1.6 mL of OE, 2 mL of MG and 121.9 mL of WA has been produced to check ACD performance. A routine control of weight, osmolality, sodium concentration, potassium concentration has been performed on each bag. Precision, accuracy and intermediate precision were then calculated for these parameters.
Measurements, calculations and statistical analyses
All masses were measured on a calibrated precision scale (XS20025, Mettler Toledo). Each product density was measured or confirmed experimentally by repeated measurements (n=3) of the mass of 10 mL of liquid exactly measured with a volumetric flask. Theoretical masses were equal to requested volumes multiplied by densities of products. A corresponding experimental mass value was measured for each requested volume and the difference between experimental and theoretical values were used for PQ.
Osmolality was measured with an osmometer (Model 2020, Advanced Instrument). Theoretical osmolality was calculated from the equation:
was equal to the concentration of the anhydrous solute (Kg/L) of each product used. Osmolarity were equal to those mentioned on products.
Sodium and potassium concentrations were measured with an atomic absorption spectrophotometer (IL943, Instrument laboratory) and theoretical concentrations were calculated from concentrations of both electrolytes mentioned on products (there was no sodium or potassium amount that cannot be quantified e. g. sodium quantities related to pH adjustment of commercial products with sodium hydroxide).
Accuracy, precision and intermediate precision were then calculated based on the recommendations of the French Society of Pharmaceutical Science and Technology (6, 7). Accuracy was assessed using the relative bias. Precision and intermediate precision was assessed using relative standard deviations associated with the intra-series variance or the sum of the intra- and the inter-serie variances (7). Descriptive statistics (e. g. relative standard deviation calculation) and figures were made using “R” (8).
Single pump ACD
Modifications of pumping or delivery speeds had little impact on the accuracy and precision of delivered quantities (Figure 1). Relative biases were in the range -0.55 % to +0.00 % and relative standard deviations were in the range +0.11 % to +0.00 % (three identical measures). A pumping speed of 50 mL/min has been selected as this speed provided a means relative bias of 0 % with the most viscous products (DE). The pause time had little impact on the precision and accuracy of pumped volumes (relative biases <0.01 % and relative standard deviation <0.03 %) and has been set up to 300 ms to optimize the pumping time. Finally, the delivery speed also had little impact on the precision and the accuracy of delivered volumes. Relative biases were in the range +0.55 % to +0.00 %, and relative standard deviations were in the range +0.07 % to +0.00 %. The delivery speed of 200 mL/min was also selected to optimize pumping time.
Performance qualification of the single pump ACD
Accuracy, assessed by calculating mean relative biases, ranged between –0.05 % and –1.71 % for the various tested products and the different volumes requested. The lowest level (5 mL) gave the highest relative biases for the three products, however, they remained between and 0.84 % and –1.71 % (Table 2).
Precision, assessed by calculating relative standard deviation of the mean mass delivered, was less than 0.5 % for all products and all level (Figure 2) and intermediate precision ranged between 0.01 % and 0.53 %, reflecting good repeatability and reproducibility of the masses delivered (Table 2).
Performance qualification of the twelve pumps ACD
Accuracy ranged between +0.85 % and –10.85 % for products pumped with the 50 mL pump, +0.55 and –6.35 % for products pumped with the 20 mL pump, and +0.93 % and –2.53 % for products pumped with the 10 mL pump. The lowest level (0.5 mL) gave the highest relative bias (Table 3). Relative standard deviation was less than 6.5 % for all products and all level (Figure 3) and intermediate precision ranged between +0.06 % and +7.45 % for the various tested products pumped with the 50 mL pump, +0.15 and +4.21 % for products pumped with the 20 mL pump, and +0.24 % and –3.67 % for products pumped with the 10 mL pump (Table 3). Highest viscous products and lowest requested volumes gave the worst intermediate precision.
Three test batches were produced with the same formula. Relative biases were very low (<1 %) for weight and osmolality based on expected values calculated as mentioned above (Table 4). They were less than 3 % for sodium and potassium concentrations. Relative standard deviation associated with precision and intermediate precision were also very low (<1 %) for all measured parameters.
Changing pumping speed, discharge speed and pause do not impact the precision and accuracy of the pumped volume (<0.55 % error). Pumping speed had to be adjusted to avoid the suction of air bubbles from air intakes during pumping operations, which mainly affect the precision and accuracy of the pump. This explains why we haven’t tested pumping speeds above 100 mL/min.
The mean bias observed with the twelve pumps ACD remains very low for volume above 5 mL (<1 %), and a recalibration of the pump could improve systemic negative or positive bias. Overall precision of the twelve pumps ACD remains satisfactory with tested products except for vitamin E, which exhibit a relative standard deviation of mean delivered mass around +3.5 %. The vitamin E formulation contains a surfactant (PEG-40 hydrogenated castor oil) which can display a drag-reducing effect that may account for this positive bias. High relative biases were observed with 50 mL pumps and 0.5 mL requested volumes, but no specifications are given by the manufacturer for requested volume below 5 mL with the 50 mL pump (Table 5). The good accuracy and precision of the 12-pumps ACD were also confirmed with routine controls of three test batches. Nonetheless, a low (<3 %) negative systemic bias has been observed with sodium and potassium concentration. We thought that these biases could be partly attributable to the analytical instrument which lacks good linearity at low values and tends to undervalue concentrations (manufacturer data). A different instrument or method such as capillary electrophoretic methods may be more accurate for quantification of cations in parenteral nutrition formulation than our routine method which is designed for the analysis of urine or serum samples (9).
In comparison with our findings, the EM2400 ACD (Baxter) with a peristaltic pump claims a “repeatable” accuracy of 10 % for 0.2 mL pumped volumes, of 5 % for volumes of 0.4 mL, and of 3 % for volumes greater than 1 mL (data manufacturer). However, information on the “accuracy” of the ACD lack precision and clarity on the parameters studied (e. g. mean relative bias or relative standard deviation). This point should be explicit in manufacturers documentations.
The single-pump and multi-pump ACD can deliver PN nutrients or electrolytes with precision and accuracy. They can be sterilized with common products used within pharmaceutical isolators such as peracetic acid or hydrogen peroxide (manufacturer information), which also make them interesting for use in other central intravenous additive services (CIVAS) or cytotoxic compounding units.
1. Dehmel C, Braune SA, Kreymann G, Baehr M, Langebrake C, Hilgarth H, et al. Do centrally pre-prepared solutions achieve more reliable drug concentrations than solutions prepared on the ward? Intensive Care Med 2011;37:1311–16. Web of ScienceGoogle Scholar
2. Hecq J-D. Centralized intravenous additive services (CIVAS): the state of the art in 2010. Ann Pharm Fr 2011;69:30–7. Google Scholar
3. Zegbeh H, Pirot F, Quessada T, Durand T, Vételé F, Rose A, et al. Accuracy, precision and speed of parenteral nutrition admixture bags manufacturing: comparison between automated and manual methods. Ann Pharm Fr 2011;69:38–44. Google Scholar
4. Pharmaceutical compounding – sterile preparations. In: The United States Pharmacopeia and National Formulary (USP 37 - NF 32). Rockville: The United States Pharmacopeial Convention, 2014:567–610. Google Scholar
5. Mondet L, Boursier A, Dubaele JM, Marçon FP. 314: Mise au point et validation de formules de nutrition parentérale standard à destination des prématurés et des nouveau-nés. Nutr Clin Métabolisme 2014;28:S235. Google Scholar
6. Hubert P, Nguyen-Huu J-J, Boulanger B, Chapuzet E, Chiap P, Cohen N, et al. Harmonization of strategies for the validation of quantitative analytical procedures. A SFSTP proposal–part II. J Pharm Biomed Anal 2007;45:70–81. Google Scholar
7. Hubert P, Nguyen-Huu J-J, Boulanger B, Chapuzet E, Cohen N, Compagnon P-A, et al. Harmonization of strategies for the validation of quantitative analytical procedures. A SFSTP proposal–part III. J Pharm Biomed Anal 2007;45:82–96. Google Scholar
8. R Development Core Team. R: A language and environment for statistical computing. R Found Stat Comput 2014;1:409. Google Scholar
9. Nussbaumer S, Fleury-Souverain S, Bouchoud L, Rudaz S, Bonnabry P, Veuthey J-L. Determination of potassium, sodium, calcium and magnesium in total parenteral nutrition formulations by capillary electrophoresis with contactless conductivity detection. J Pharm Biomed Anal 2010;53:130–6. Google Scholar
About the article
Terry Hennache is a PharmD candidate. He has been resident in the sector of drug manufacturing and control at Amiens University hospital’s pharmacy. Currently, he is resident at Lille University Hospital since November 2015.
Gwenaelle Clabaut is currently pursuing a double degree in pharmaceutical and engineering sciences at University of Picardy Jules Verne (PharmD degree) at the International Mines Albi Graduate school of Engineering since September 2015 (Master degree). She was an intern in the sector of drug manufacturing and control at Amiens University hospital’s pharmacy.
Lina Mustapha is a PharmD candidate. She has been resident in the sector of drug manufacturing and control at Amiens University hospital’s pharmacy. Currently, she is resident at the Henri Becquerel Center since November 2015.
Jean-Marc Dubaele is a pharmacist in the sector of drug manufacturing and control at Amiens University hospital’s pharmacy. He completed is PharmD degree at University of Paris XI (Châtenay-Malabry) where he also completed his Master’s degree in structures and analyses of pharmaceutical plastics. He is involved in the development of paediatric medicine and has cofounded the GREPP research group.
Frédéric Marçon is a lecturer at the University of Picardy Jules Verne and pharmacist in the sector of drug manufacturing and control at Amiens University hospital’s pharmacy. Former resident at Amiens University Hospital, he obtained is PharmD degree in 2009 and completed his PhD in biopharmacy and pharmaceutical sciences in 2013. His work and researches are focused on the development of paediatric medicines (age-appropriate formulations, pharmacokinetic study, efficacy and safety study). He is a cofounder of the GREPP research group.
Published Online: 2016-02-12
Published in Print: 2016-03-01
Conflicts of interest statement: The authors state no conflict of interest. They have read the journal’s Publication ethics and publication malpractice statement available at the journal’s website and hereby confirm that they comply with all its parts applicable to the present scientific work.