We have chosen to study furosemide-midazolam incompatibility, because these drugs are widely used in anesthesia and intensive care units. This incompatibility is due to an acid–base reaction. In an equivolume (1:1) mixture, the formation of a visible milky-white precipitate is immediate [8
]. Because furosemide-midazolam incompatibility is pH-dependent, the impact of furosemide concentration is predictable. Furosemide in 10 mg/mL of saline is an alkaline solution (pH = 8.77). Mixing a furosemide solution with an acidic solution (i.e., 5 mg/mL midazolam, pH = 3.47) decreases the pH of the mixture sufficiently to cause furosemide precipitation [9
]. Furosemide and midazolam concentrations were simultaneously determined in the solution by UV spectrophotometry coupled with partial least square (PLS) regression [10
Furosemide (10 mg/mL Furosemide, Renaudin, France), midazolam (5 mg/mL Midazolam, Mylan, France), and saline (500 mL Freeflex®, Fresenius Kabi, France) were simultaneously infused using syringe pumps connected to a three-lumen infusion device (VSET + M, Doran International, France) consisting of a central tube with an antireflux valve for saline and two flexible low dead volume tubes reserved for the furosemide and midazolam infusions (Figure ). An extension line (diameter = 1 mm, length = 25 cm) simulating the central venous catheter was added at the distal end of the infusion set. A 1.2-μm porosity filter (Lipipor TNA, Pall, France) was added or not at the end of the infusion line. Three 50-mL syringes were prepared for each experiment: one filled with furosemide diluted in saline at 10 or 2.5 mg/mL, one filled with midazolam diluted in saline at 1 mg/mL, and one with saline only. The final drug concentration was checked using a spectrophotometric method before infusion.
Three-lumen infusion device (VSET + M) with the central lumen reserved for carrier fluid and the two flexible low dead volume tubes reserved for furosemide and midazolam infusions.
Using unpublished results, we defined two infusion conditions leading or not to visible particle formation. The choice of drug concentrations was in accordance with clinical practice and infusion sets were used with no filter:
1) 10 mg/mL of furosemide at 2 mL/h infusion rate, 1 mg/mL of midazolam at 2 mL/h, and saline at 100 mL/h (condition leading to visible particle formation).
2) 2.5 mg/mL of furosemide at 8 mL/h infusion rate, 1 mg/mL of midazolam at 2 mL/h, and saline at 50 mL/h (condition not leading to visible particle formation).
The infusion was first subjected to visual inspection, and then, in the absence of visible particles, a 25-mL sample was collected at the egress of the extension line. Particle counts were taken using a particle counter (APSS-2000, PMT, France). Tests were performed under the conditions described in Chapter 2.9.19 of the 7.5th
European Pharmacopeia [11
]. The infusion condition complied with the subvisible particle count test for a high volume as long as the average number of particles present in the sample tested did not exceed 25 per mL for particle sizes ≥10 μm and 3 per mL for particle sizes ≥25 μm. Each infusion condition was subjected to the visual inspection test three times. Three particle counts per sample were performed for the condition not leading to visible particle formation. All tests were performed at room temperature between 18°C and 22°C.
Our study was divided into two parts. In the first, we revalidated the two infusion conditions, following the same methods. In the second, for each infusion condition tested, we determined the mass flow rates of furosemide and midazolam on the infusion line with and without filter. Five trials were made per infusion condition tested. Drug concentrations in the mixture at the egress of the infusion line were determined using UV spectrophotometry (model UV-2450, Shimadzu, France) and partial least square (PLS) analysis. All information from the spectrophotometer was collected with UVProbe 2.21 software (Shimadzu, France). A partial least square (PLS) method on UV spectra was used to determine simultaneously the concentrations of the two drugs at the egress of the terminal extension line.
PLS regression is a simple and powerful multivariate method based on factor analysis and is used for building regression models based on latent variable decomposition relating a block of independent variables, x (spectra), to a block of dependent ones, y (concentrations). PLS regression was obtained using the PLS module of XLSTAT software version 2011.2.01 (Addinsoft, France). The 220–320 nm spectral zone was used to obtain the best model. The recovery percentage was in the 100.14–101.25% range. Detection limits (LOD = 3.3 x standard deviation/gradient) of drugs in mixtures were 0.19 μg/mL for furosemide and 0.36 μg/mL for midazolam. The quantification limits (LOQ = 10 x standard deviation/gradient) were 0.57 μg/mL for furosemide and 1.10 μg/mL for midazolam. Selectivity, calculated from the net analyte signal, was equal to 0.18 and 0.2 for furosemide and midazolam respectively. Because these exceeded the spectrophotometer’s linear range, they were diluted in saline. The drug mass flow rate (expressed as mg/h) was calculated as the product of drug concentration against total flow rate. Observed/theoretical mass flow rate ratios for each drug (%) also were determined per infusion condition. Measurements of pH were performed on drug solutions at the egress of infusion device using a pH meter (PHM201 MeterLab, Radiometer Analytical, Villeurbanne, France).
The Student’s t test was used to compare observed and theoretical drug mass flow rates and filtered and nonfiltered data after performing the Shapiro-Wilk test to check that the data observed was normally distributed. Results are expressed as mean values ± standard deviations (± SD) of mass flow rates. The level of significance was established at 0.05.