Antibiotics
Intravenous antibiotics for human use were bought from reputable local drugstores as needed. All products had been properly licensed by the drug regulatory agency of Colombia. We used the innovators of amikacin (Amikin®, Bristol-Myers Squibb, Guayaquil, Ecuador), gentamicin (Garamicina®, Schering-Plough SA, Bogota, Colombia) and vancomycin (Vancocin CP®, Eli Lilly & Compañia de Mexico SA de CV, Mexico), as well as the available generic products: four of amikacin, fifteen of gentamicin and three of vancomycin (N = 22 generic products). All products were reconstituted with calibrated micropipettes (Transferpette®, BRAND, Wertheim, Germany) following the manufacturer instructions for clinical use. Pure-grade reference powders of both aminoglycosides (Sigma Aldrich, St Louis, MO) and vancomycin (United States Pharmacopeia, Rockville, MD) were used as controls. The former were weighted in an analytical balance (Sartorius, Goettingen, Germany) and the latter was diluted to a final concentration of 50 mg/mL in distilled water.
Media, strains and inocula
Testing strains were Bacillus subtilis ATCC 6633 for amikacin (and some experiments with vancomycin), Staphylococcus aureus ATCC 6538p for vancomycin, and Staphylococcus epidermidis ATCC 12228 for gentamicin. Seeding agar were Difco™ Antibiotic Media (Becton Dickinson & Co, Franklin Lakes, NJ) numbers 5 (amikacin), 8 (vancomycin) and 11 (gentamicin) [24].
To grow S. aureus and S. epidermidis, we followed CLSI protocols [25]. Briefly, each bacterial stock previously stored and frozen at -70°C was resuscitated on solid medium (in two successive Trypticase Soy Agar plates, Becton Dickinson, Sparks, MD) and five colonies were selected and passed to 10 mL of Trypticase Soy Broth (five 16 × 125 glass tubes labeled 1 to 5 with successive 1:10 dilutions). After overnight incubation at 37°C, we made a second transfer from the last tube with visible growth into 10 mL of fresh liquid medium (three 16 × 125 glass tubes labeled 6 to 8 with successive 1:10 dilutions). This second set was grown to attain an OD580 nm = 0.300 for S. aureus and 0.450 for S. epidermidis (Spectro 22®, Labomed Inc., Culver City, CA, USA), equivalent to a log-phase culture with approximately ~108 CFU/mL. We ruled out significant impact of S. aureus' cell clusters [26] by multiple standardizations of growth curves correlating OD with CFU count (data not shown). For B. subtilis, a spore suspension was prepared by growing the microorganism for one week at 37°C in a large bottle on Difco™ Antibiotic Media number 1. Then, the spores were suspended in sterile distilled water and heated for 30 minutes at 65°C, washed three times, and reheated at 65°C for 30 minutes before re-suspension in sterile distilled water. The final suspension containing ~108 CFU/mL was temporarily maintained at 15°C prior to inoculation of the bioassay agar.
Device and pouring of glass-plates assay
A 36 × 36 cm glass plate previously described was modified to allow simultaneous runs in duplicates of all assays of the different generic products and innovator [27]. The device was routinely cleaned (iodine soap and water), disinfected (70% ethanol) and sterilized (steam autoclaving at 121°C). The seeding agar (3 mm of depth) was prepared following the manufacturer instructions, dispensed in a sterile 300 mL flask, and placed into a water bath (50°C) to maintain the agar in liquid state until poured. We inoculated the bioassay agar adding 2 mL of log-phase suspension of S. aureus or S. epidermidis in 100 mL of the corresponding antibiotic media or adding 3 mL of B. subtilis spore suspension for each 100 mL of antibiotic medium. These inoculum sizes (~2 × 106 CFU/mL) ensured sharply defined zone edges and a good slope steepness of the log dose-response line.
The solutions were applied in a simple sequential fashion down the columns, as described by Bennett et al. [27]. Incubation was carried out at 35°C for 18–22 h depending on the bacteria-agar combination, according with the instructions of the antibiotic media manufacturer [24]. The same researcher measured the zone sizes in all assays using an electronic caliper (Mitutoyo Corp., Kawasaki, Japan).
The use of mouse serum was approved, as well as the complete protocol, by the University of Antioquia Animal Experimentation Ethics Committee.
Statistical Analyses
We determined the linearity, limit of quantification, precision, accuracy, and specificity to validate the method for testing pharmaceutical equivalence. For this purpose, the log-transformed concentrations (x-axis, log10 mg/L) of each product were plotted against their respective inhibition zone sizes (y-axis, mean diameter in mm); intercept and slope of the best straight line were obtained fitting the data to a linear model (expressed by the equation y = b + mx, where b is the y-intercept and m is the slope) by least-squares regression using SigmaPlot 9.0 (SPSS Inc., Chicago, IL). The concentrations used ranged from 0.5 to 256, 0.125 to 64, and 0.25 to 128 mg/L for amikacin, gentamicin and vancomycin, respectively. The goodness of fit to the model (linearity) was expressed as coefficient of determination (r2) and standard error of estimate (Syx). We also calculated the x-intercept (log10 mg/L) and slope of the regression line with 95% confidence intervals (95% CI) [28]. The regression approach to analysis of variance was used to determine the statistical significance of intercept and slope. Normality and constant variance assumptions were checked using Kolmogorov-Smirnov and Levene's tests, respectively (SPSS 15.0, SPSS Inc., Chicago, IL). In case of non-equal variance, we applied Welch's test to confirm the significance of the parameters derived from the linear model [29].
Using a symmetrical parallel-line assay we tested the pharmaceutical equivalence comparing slope and intercept of each generic product with those of the innovator by Overall Test for Coincidence of the Regression Lines, a statistical technique for Curve Fitting Analysis (CFA) (Prism 5.0, GraphPad Software, Inc., San Diego, CA) [30]. We defined potency as the slope of the linear regression and concentration as the anti-log of the x-intercept when y = 0. We also estimated the relative concentration when y is at the mid-point of the linear regression, because this is the point of minimal variation using the 95% confidence interval of the predicted line. Assuming that generic and innovator are the same product, a pharmaceutically equivalent generic must display a parallel and overlaid curve with respect to the innovator (P > 0.05 by CFA). On the other hand, parallel curves with different intercepts meant identical APIs but at significantly higher or lower concentration, while lack of parallelism implied that different products were being tested, without further considerations. We also calculated the relative potency of each generic to innovator as the x-distance between the two lines, as recommended by the international pharmacopoeias [20].
The limit of quantification was defined visually as the smallest amount of drug that still produced a clearly distinguishable inhibition zone around the diameter of an empty well (3.5 mm). The limit of detection was calculated using the standard deviation and slope method recommended by ICH guidelines.
The repeatability of the assay was determined using a minimum of three concentrations of each antibiotic by triplicates during the same day or under similar experimental conditions but with different biological matrices (water instead of serum to dilute samples) and plates (intra-assay precision), and comparing the results of assays on different days (inter-day precision). These were expressed as means with standard deviations and coefficients of variation (CV) [31].
To test the ability of the assay to detect significant differences in concentration of the API (accuracy), we prepared a 5-point standard curve (two-fold dilutions from 128 to 8 mg/L) using vancomycin reference powder (USP) and compared it with samples containing 10, 15 and 20% more vancomycin. Briefly, a vial containing 100.5 mg of the antibiotic was dissolved in 2.01 mL of distilled water to a final concentration of 50 mg/mL and serially diluted to achieve 1 mL at 128 mg/L. Then, 0.1 mL aliquots of this solution were diluted to obtain a set of three extra concentration groups over three selected points in the standard curve (16, 32 and 64 mg/L): Group I with 10% more (17.6, 35.2 and 70.4 mg/L), Group II with 15% more (18.4, 36.8 and 73.6 mg/L) and Group III with 20% more (19.2, 38.4 and 76.8 mg/L) vancomycin. Intercepts and slopes of linear regressions produced by the reference standard and each extra group were compared by CFA [32].
To show the ability of the microbiological assay to unambiguously assess the API in presence of all other expected components in a pharmaceutical-grade product for clinical use (specificity), it was necessary to subject the innovator of vancomycin to specific stress conditions to increase degradation subproducts. In this case, we compared by CFA the intercept and slope of a freshly prepared standard curve (16 to 256 mg/L) against those obtained after 2, 4, 8 and 16 hours of heating at 80°C.