Clostridium perfringens is a common cause of intestinal diseases in humans, animals, fish, and their environment. It has caused serious damage to the global economy in the last few years (Allaart et al., 2014; Nan et al., 2017; Yadav et al., 2017; Yan et al., 2017). C. perfringens can be divided into five types, including A, B, C, D, and E types, based on the differences between its toxigenic genes and pathogenicity (Hatheway 1990). C. perfringens diseases in pigs are generally caused by type A and type C (Lee et al., 2014). Type A is often linked to diarrhea in suckling piglets with mild necrotizing enterocolitis (Springer et al., 2012); hemorrhagic necrotic enteritis is induced by type C in piglets aged 0–2 weeks (Waters et al., 2003; Nan, Hao et al., 2017).

Avilamycin is an orthosomycin family antibiotic derived from the fermentation products of Streptomyces viridochromogenes (Mertz et al., 1986; Saito-Shida et al., 2018), including the major active factor avilamycin A and 15 other small factors (JECFA, 2004). Avilamycin is one of the EU-approved antimicrobial agents in the feed industry and is currently only used in animals to control bacterial enteric infections and multidrug-resistant gram-positive bacteria (Ho et al., 2013; Lv et al., 2013; OIE, 2015). Avilamycin has a strong treatment efficacy against necrotic enteritis (Vissiennon et al., 2000; Paradis et al., 2016). It does not display cross-resistance with any other antimicrobial agents, suggesting that this type of antimicrobial agents may represent an avenue for the for development of new antimicrobial agents (Aarestrup et al., 2000; Weitnauer et al., 2001; Arenz et al., 2016; Krupkin et al., 2016). It can be placed at the frontline of drugs due to its low environmental toxicity, extensive metabolism in vivo, and reduced ecological hazards (Arenz, Juette et al., 2016; Krupkin, Wekselman et al., 2016).

In order to reduce the occurrence of drug resistance, a reasonable dosing regimen is necessary. Pharmacokinetics—pharmacodynamics (PK-PD) study are very important for developing a reasonable dosing regimen (Yoshii et al., 2016). Many PK-PD studies mainly focus on the analysis of PK data in the plasma, because the data in the plasma is relatively stable and easy to obtain, but the Clinical Laboratory of Standard Institute (CLSI) also mentions that when conditions permit, the PK of target tissues can also be used (CLSI 2007a). For infections of C. perfringens, the drug concentration in the ileal contents (an unformed stool sample taken from the ileum) should be used. Compared with disease animal models, healthy animal models have better stability and reproducibility, which is even more important (CLSI 2007a).

The clinical breakpoint (CBP) is of great significance to antimicrobial susceptibility testing and monitoring of resistance (OIE 2016; Toutain et al., 2017; Lei et al., 2018). The determination of interpretive CBPs needs a comprehensive analysis of relevant information, including wild-type cutoff value (CO_WT)/epidemiological cutoff value, PK-PD cutoff value (CO_PD), and clinical cutoff value (CO_CL). The CO_WT needs large numbers of in vitro minimum inhibitory concentrations (MICs) or zone diameter tests (Turnidge et al., 2007), and it allows the detection of resistance as a biological phenomenon (Toutain, Bousquet-Mélou et al., 2017). The CO_PD is the relationship between drug concentrations and microbial PK parameters ex vivo (Kahlmeter et al., 2003; CLSI 2007a). The CO_CL is the clinical outcomes with MIC from prospective clinical studies (Ambler et al., 2008). However, the clinical data is not easy to achieve from current studies; the CO_WT could be used to separate susceptible isolates from isolates with resistance when there is a lack of clinical data (Espinel-Ingroff et al., 2011; Peng et al., 2016). But the CO_WT cannot represent the CO_PD and clinical efficacy (Barbour et al., 2010). When the three cutoff values are established, the decision tree from CLSI is applied for establishing the CBPs (CLSI, 2007a).

The purpose of the current study is to formulate the optimal regimens of avilamycin treatment for C. perfringens infection based on the CBP, and the CBP has been based on the CO_WT, CO_PD, and CO_CL of avilamycin against C. perfringens.

C. perfringens is an environmental bacterium existing in the breeding industry and responsible for enteric disease with focal necrosis in the intestine. Also, it is a pathogen to public health through the food chain (Silva et al., 2009). Because of the antibacterial used for growth promotion or to prevent pathogenic effects of C. perfringens, extensive microorganism susceptibility researches have been conducted. Several studies showed that the C. perfringens antibiotic resistance was high to penicillin, bacitracin, tetracycline, and lincomycin (Akhi et al., 2015). But the oligosaccharide antimicrobial agent, avilamycin, has shown excellent activity against C. perfringens. The MICs of 30 C. perfringens isolates for avilamycin were tested in the local poultry industry, and the concentrations ranged from 4 to 512 μg/ml (Redondo et al., 2015). The MICs of avilamycin for 50 C. perfringens type A strains isolated from 1- to 7-day-old piglets, with or without diarrhea, ranged from 1 to 256 μg/ml, and the MIC with the highest number of strains was concentrated in 4 μg/ml (Salvarani et al., 2012).

In this study, the MIC distributions of avilamycin against the 120 clinical strains of C. perfringens ranged from 0.015 to 256 μg/ml, which had a wide scope than former research. Fitting the reported MIC data with our MIC distributions together, we got the same conclusion that the CO_WT was 0.25 μg/ml. The low MIC value (0.015 μg/ml) might be explained by regional disparities due to avilamycin not being popularly used in China. Meanwhile, the MIC distributions present a double peak, and there were a small number of resistant bacterium; the resistance mechanism of C. perfringens to avilamycin should be the focus and taken into account when establishing the CO_WT in future studies (Zheng et al., 2017). The mode of action of avilamycin is not well elaborated. It has been suggested that avilamycin acts by binding to the 30S part of the ribosome to inhibit bacterial protein synthesis (Wolf et al., 1973). However, a recent study indicated that resistance to avilamycin is associated with variations in the ribosomal protein L16 and thereby probably interacts with the peptidyl transferase activity. Some research showed that avilamycin binds to the 50S subunit, resulting in decreased susceptibility to avilamycin in bacteria (Mcnicholas et al., 2000).

We usually need to adopt certain statistical methods to define the CO_WT by the NRA (Hao et al., 2013; Smith et al., 2014). Conventionally, the MIC values were logarithmically transformed, and the cumulative distributions were subjected to regression simulation. The gradient was added to each fitting from the smallest distributed MIC to the most distributed MIC. The fit was kept between the optimal range between the actual values and estimated values, with a 95% confidence interval. Then the probability was predicted above the upper limit and below the lower limit, respectively, with the CO_WT as the upper limit of the wild-type distributions. In this experiment, the ECOFFinder program based on the NRA, was used to gradually replace and simplify the fitting process. We only needed to put the MIC distributions into the software to automatically fit and obtain CO_WT.

PK-PD explored the relationship between the PK parameters and bacterial inhibition or killing and even extended it to the clinical outcome (Drusano, 2004). Over the last 30 years, a clear understanding of PK-PD on the many classes of antimicrobial agents had been researched, including the β-lactams, aminoglycosides, quinolones, macrolides, lincosamides, tetracyclines, and glycopeptides (Ambrose et al., 2007; Turnidge and Paterson 2007). In this study, we used healthy animals to establish the PK-PD model because the PK data were often generated in healthy representatives of the target species and the healthy animal model was more stable than the disease model, as well as showed better repeatability (CLSI 2007a). Furthermore, we suggest that the animal model can also be established in diseased animals at the same time and compared with healthy animal models if permitted.

The index T > MIC was used for time-dependent antibacterial agents, but the indexes AUC_0–24h/MIC and C_max/MIC were used for concentration-dependent antibacterial agents. The selected PK-PD index determined the PD parameter that best predicts efficacy in the animal model (Jr et al., 2007). The human medicine, evernimicin, which belonged to the same class as avilamycin, had been used to evaluate the bactericidal and sterilizing efficacies. Evernimicin exhibited a concentration-dependent killing kinetics, as the ratio AUC0–24h/MIC was the best index correlated with a reduction of bacterial count (Drusano et al., 2001). The bacteriostatic characteristic of avilamycin against C. perfringens indicated concentration dependence too. Hence, the PK-PD index (AUC_0–24h/MIC) has been considered to calculate avilamycin ex vivo activity.

Most of the current studies on PK development have focused on the analysis of data in the plasma because these data are relatively stable and relatively easy to obtain. The concentrations of the drug at the target sites were generally estimated on the basis of the concentrations of the serum or plasma because the drug concentrations at the infection sites were largely correlated to the serum or plasma (Craig 1998; Ping et al., 2002). In our study, the drug concentrations in the blood might not accurately reflect the drug concentrations of the target tissue. For some digestive tract drugs, the T-shape ileal method had been used in swine for a more accurate PK study of the drugs in the digestive tract which could provide the target animals with maintaining a normal physiological state. Furthermore, researchers had shown that the measurement of free drug concentrations was more important than the total drug concentrations for evaluating the antimicrobial activity (Andes et al., 2002).

At present, the most common method for setting CO_CL was to collect a large number of clinical cases (Toutain et al., 2010; Lee et al., 2016), integrate the data, and finally use statistical analysis methods to obtain CO_CL. Because of the limited clinical treatment of C. perfringens in pigs caused by avilamycin, the clinically challenged test had proceeded under laboratory conditions. The therapeutic dose adopted PK-PD recommended dosages which were linked to PK-PD with clinical treatment effect. The CO_CL formula POC = 1/[1 + e-a+bf (MIC)] proposed by EUCAST had not been applied in practice at present (Toutain et al., 2017). CO_CL reflected the upper limit of the MIC values associated with a high likelihood of clinical success (when POC equals to 90%). Based on this principle, five different MICs nearing MIC₅₀, MIC₉₀, CO_WT, sensitive, and resistant strains were selected for conducting the clinical treatments. Three statistical analysis methods, including the “Window” approach, NRA, and CART analysis, were used to analyze the results of the clinical trials (Bhat et al., 2007; Cuesta et al., 2009; Cuesta et al., 2010; Turnidge and Martinez 2017), predicting the POC corresponding to the MIC distributions.

Comparing these three cutoff values, it was not difficult to find that the CO_PD was much higher than CO_WT and CO_CL. Because the dose used in the PK-PD study was the recommended dose, which was a uniform dose for piglet diarrhea. C. perfringens was sensitive to avilamycin, and the drug was widely distributed in the digestive tract, leading to high AUC values and CO_PD. Specially, when the CO_WT equaled CO_CL, the CO_PD did not influence the final CBP. When the CO_WT did not equal CO_CL, the CO_PD was used as a weighting factor, but the CBP would be established by the primary determinant CO_WT or CO_CL (CLSI 2007a). There was no clear methodology for establishing the CBP in veterinary medicine. In this study, we provided a train of thought and suggested that if there were obvious differences between CO_CL, CO_WT, and CO_PD, the number of clinical samples should be expanded, and there is a need for expert subcommittee member deliberation on the relative weighting of these three cutoffs.

The rational use of antibiotics is becoming more and more important in veterinary clinics, and the abuse of antibiotics is the main reason for the development of bacterial resistance. After establishing three cutoff values in this study, the CBP of avilamycin against C. perfringens in swine was set to be 0.25 μg/ml, which can easily and clearly distinguish drug-resistant bacteria. At the same time, the findings of this study established that the optimal dosage of avilamycin for the treatment of pigs with C. perfringens infection is 5.2 mg/kg mixed feeding, and if the daily doses are given at 12 h intervals, the expected effect can be achieved.