Dosage regimens of antimicrobial drugs should be optimized first to achieve clinical and bacteriological cure and second to provide minimal opportunities for the selection of resistant organisms. The successful outcome of therapy depends on both the pharmacokinetic (PK) profile, which determines the drug concentration-time course at the site of infection and the pharmacodynamic (PD) profile that is drug effect on the pathogen [1–4]. Hence, a rational approach to dose prediction, for subsequent confirmation or adaptation in disease models and clinical trials, is dependent on PK/PD integration and modelling methods. These establish PK/PD breakpoints for pre-determined levels of kill [2–6].

Minimum Inhibitory Concentration (MIC) is the primary PD parameter used to determine potency; it is the lowest drug concentration that inhibits visible bacterial growth after 16–24 h incubation. Standardised methodologies for MIC determination are described in European Committee on Antimicrobial Susceptibility Testing (EUCAST) and Clinical and Laboratory Standards Institute (CLSI) [7] guidelines [VET01-A4 (formerly M31-A3)]. These ensure reproducibility between individual analysts, between laboratories and across both geographical regions and time [5, 8]. CLSI and EUCAST methods require the use of broths formulated to standardises the growth of bacteria of each bacterial species. Thus, the almost universal use of internationally recognised guidelines, methods and standards for MIC determination is highly beneficial. However, when the objective of potency determination is prediction of dosage for clinical efficacy based on PK/PD breakpoints, conditions should be representative of in vivo pathological circumstances. Zeitlinger et al. [9] commented that “bacteria with appropriate and well-defined growth in the selected medium should be employed” and “in order to be able to extrapolate data from various models to in vivo situations, models should always attempt to mimic physiological conditions as closely as possible”.

Whilst serum is not identical to extravascular infection site fluids, it is likely to be closer to the biophase than broths in chemical composition and indeed in respect of immunological components also [9, 10]. A comparison of broth MICs with potency determined in biological fluids is therefore relevant to PK/PD breakpoint estimation, when the aim is optimal dose prediction for clinical use. For some drugs and pathogens, calculation of a scaling factor to bridge between broth and serum MICs may be warranted [4, 6, 10].

A second consideration, in relation to dose prediction, is that CLSI and EUCAST guidelines have limitations regarding accuracy for individual isolates, because they require use of two-fold dilutions, giving potential for approaching 100% error. For most purposes, this is wholly acceptable, and indeed is necessary, when large numbers of isolates are examined to establish MIC distributions. When plotted on a histogram, using a log-base 2 distribution, they are log-normal. These histograms facilitate the identification of wild-type distributions. Standardisation of methodology thereby enables determination of epidemiological cut-offs (ECOFF) by EUCAST and wild type cut-offs (CO_WT) by CLSI. However, modification of the two-fold dilution approach may be appropriate, when MICs are correlated with PK data for establishing PK/PD breakpoints. Based on methods previously described [11, 12] five overlapping sets of doubling dilutions were used in this study to decrease inaccuracy from up to 100% to no more than 20% for a small number of individual isolates. Hence, the quantitative determination potency with improved accuracy and in biological fluids may be appropriate for some drug classes and some microorganisms, when calculating PK/PD indices [6, 10, 12–16].

It is the hypothesis of this investigation that MICs determined in pig serum (potentially reflecting more closely concentrations required to achieve clinical efficacy than those determined in artificial broths) might differ from MICs determined in broths. The aims were to: (1) compare MICs measured in five broths with MICs determined in pig serum, using five sets of overlapping two-fold dilutions for three drugs, florfenicol, marbofloxacin and oxytetracycline, used in the treatment of pig pneumonia caused by the pathogens, Actinobacillus pleuropneumoniae and Pasteurella multocida; (2) determine whether any broth/serum MIC differences were attributable solely to drug binding to serum protein; (3) compare chemical compositions of CLSI recommended broths and pig serum; (4) determine rates of growth A. pleuropneumoniae and P. multocida in broths and pig serum in the absence of drugs. Aims 3 and 4 were included to investigate whether chemical composition differences and/or bacterial growth rate differences between broth and serum might account for matrix MIC differences over and above those attributable binding of the drugs to serum protein.

For some drug classes, MICs measured in broths using internationally standardised procedures (for example, those of CLSI and EUCAST) may not differ significantly from potency estimated in biological fluids. The absence of growth matrix differences implies no significant impact on PK/PD breakpoint estimation, as a basis for dose determination, provided that the protein bound serum drug concentration is corrected for [4, 16, 20]. However, broth/biological fluid differences in potency do occur for other drug classes, even after correction for drug binding to proteins in biological fluids. For example, Honeyman et al. [21] compared potencies of several tetracyclines in broth and a 50% broth: 50% serum mixture. They established marked differences in MIC for the two growth matrices. Moreover, MICs differed and had to be determined on a matrix-by-matrix, species-by-species, compound-by-compound basis. Likewise, Brentnall et al. [10, 14] reported for oxytetracycline, after correction for serum protein binding, for a calf isolate of Mannheimia haemolytica, a MIC in serum 6 times greater than the broth MIC. In marked contrast, Toutain et al. [6] reported, after correction for serum protein binding, MICs some 80-fold smaller in calf serum compared to broth for tulathromycin for M. haemolytica and P. multocida isolates from calves. Similar findings, with quantitatively even lower MICs for Mycoplasma mycoides mycoides in calf serum compared to broth, were reported for the macrolides, tulathromycin and gamithromycin by Mitchell et al. [22–24]. Thus, whilst correction for serum protein binding is always necessary, it is not always sufficient to account for potency differences between matrices.

These published data suggest: first, that serum MIC values should be considered on both a drug-by-drug and bacterial species-by-species basis to allow for the inactive protein bound fraction; and second, corrected serum values may not be the same as, and therefore might be used in preference to, the broth MIC for PK/PD breakpoint estimation.

Only free drug is microbiologically active and therefore protein binding is a major factor, and unfortunately sometimes the only factor, considered in seeking to explain growth medium differences in antimicrobial drug potency [2, 3, 9, 16, 20, 25–27]. The magnitude of drug binding to serum protein can vary with methodology [16] and, moreover, for single drugs, intra-species differences have been reported. In addition, it is necessary to consider possible differences in protein binding in serum obtained from different sources, differing animal breeds, ages and indeed between healthy and diseased animals. Such variations must be borne in mind in considering the present findings, for which protein binding was determined for pig serum from a single source in healthy animals. In fact, for all three drugs investigated, the degree of binding was independent of total concentration over therapeutic ranges [17–19].

For both organisms investigated in this study, correcting the serum MIC for protein binding for marbofloxacin, florfenicol and oxytetracycline yielded fu serum MICs differing from broth MICs in most instances. Fu serum:broth MIC ratios were: marbofloxacin 0.63:1 (P. multocida) and 0.35:1 (A. pleuropneumoniae); florfenicol 0.19:1 (P. multocida) and 0.38:1 (A. pleuropneumoniae); oxytetracycline 6.30:1 (P. multocida) and 0.35:1 (A. pleuropneumoniae). Thus, for marbofloxacin and florfenicol there were small to moderate trends for both pathogens of increased (1.6- to 5.2-fold) potency in serum compared to broth. Likewise, for oxytetracycline and A. pleuropneumoniae the greater potency in serum was 2.8-fold, whereas for P. multocida potency was 6.3-fold lower in serum. Consequently, correction of serum MIC for protein binding is, as stated, necessary but not sufficient for determination of potency differences between the two matrices for these two bacterial species and these three drugs. The consistent finding of differences, which were nevertheless unpredictable in direction (increased or reduced potency) and magnitude indicates the possibility of similar differences for human pathogens also more frequently than is commonly recognised.

Possibly the most important difference in chemical composition between all broths and serum was in albumin concentration, which was much higher in serum. Correction for drug binding to serum albumin revealed broth/serum potency differences for all drugs. It is therefore necessary to consider whether other differences, as well as albumin content, in chemical composition might explain the protein binding corrected potency differences. The chemical analyses indicated wide differences in composition of five broths for electrolytes, iron and organic compounds. Possibly surprisingly, these frequently large inter-broth differences resulted in, at most, very minor differences in MICs, with one exception, namely marbofloxacin, and for P. multocida only. However, even for this drug and species, MICs were similar for two broths and different but again similar for other three broths. In contrast, the broth and corrected serum MICs differed, despite the finding that chemical composition of pig serum for every analyte was within the lower and upper ranges for the five broths. Therefore, chemical composition indicated no readily apparent explanation for the broth/serum MIC differences.

In addition to the chemical differences between broths and serum, for circulating blood there are other differences e.g. the presence of white and red cells as well as a wide range of immunological components in blood. Future studies should therefore be directed towards evaluating which of these might interact with drugs, and how, to modulate drug potency. One factor, pH, is known to influence the rate and extent of growth of microorganisms [28, 29]. An effect of pH on weakly basic drugs such as macrolides and triamilides is well recognised [6].

Zeitlinger et al. [26] compared growth curves of Staphylococcus aureus and Pseudomonas aeruginosa in MHB and serum. Slower logarithmic growth was obtained for both species in serum compared to broth, and this might be expected to provide more rapid kill in serum, as a consequence of a smaller microbial challenge [5, 30]. Indeed, Illambas et al. [31, 29] reported significant effects of inoculum count (high, medium and low) on MIC for marbofloxacin and florfenicol and the calf pneumonia pathogens, M. haemolytica and P. multocida, while Dorey et al. [17–19] reported similar dependency of MIC on inoculum strength for the three drugs and the two pig pathogens investigated in this study.

In the present study, bacterial growth was initially greater in broths than serum for both P. multocida and A. pleuropneumoniae. This difference persisted for P. multocida at 24 h, whereas for A. pleuropneumoniae logarithmic growth was similar for the two media at this time. Slower growth, resulting potentially in lesser challenge to bacterial kill, therefore might explain, at least in part, the greater potency in serum compared to broth in five of six instances. The one notable exception, however, was the 6-fold reduction in potency of oxytetracycline in serum compared to broth for P. multocida. This finding indicates that some serum factor, presently unknown, operates to reduce potency of this drug for this species, whilst conversely potency was increased in serum for A. pleuropneumoniae.

The differences in MIC between serum and broth reported in this study do not provide a rationale for abandoning broths by diagnostic laboratories reporting MIC distributions of wild type organisms. This would be impractical and unnecessary. Rather, the present data suggest that, for marbofloxacin, florfenicol and oxytetracycline and the two bacterial species studied, it will be appropriate and possible to apply a scaling factor, to bridge between MICs in broths and pig serum when calculating PK/PD breakpoints.

For three commonly used drugs, florfenicol, oxytetracycline and marbofloxacin, and two pathogenic species of bacteria causing pig pneumonia, A. pleuropneumoniae and P. multocida, differences between serum and broth MICs were obtained, after correction was made for binding to serum protein. In contrast, major differences were not obtained when MICs were determined in five broths, marked differences in chemical composition between the broths notwithstanding. In the absence of drugs, the rate of logarithmic growth of A. pleuropneumoniae and P. multocida was slower in pig serum than in broth. It is suggested that, in using MIC data as a pharmacodynamic parameter to determine PK/PD breakpoints, the use of a scaling factor to bridge between broth and biological fluids may be required more commonly than currently recognised.