When evaluating outcomes of mastitis therapy, the spectrum of activity of the antimicrobial should be considered relative to the etiology of the cases enrolled in the study. In most countries, a variety of IMM antimicrobials are approved and veterinarians in some countries have access to systemically administered drugs that are able to penetrate mammary gland barriers (21). Most approved antimicrobials are relatively narrow spectrum (target Gram-positive organisms) but in some countries, broader spectrum products such as 3rd and 4th generation cephalosporins and some quinolones are approved. Dairy farmers in the U.S. have access to 7 approved IMM antimicrobial products but no systemic products are approved for treatment of clinical mastitis (limited extralabel usage of some products is allowed). One approved IMM product is classified as a lincosamide (pirlimycin) while 6 IMM products are classified as beta-lactams. The beta-lactams include 1st (cephapirin) and 3rd (ceftiofur) generation cephalosporins, aminopenicillins (amoxicillin and hetacillin), penicillin G, and a penicillinase-resistant penicillin (cloxacillin) (31). All approved IMM products are expected to have efficacy against most common Gram-positive mastitis pathogens and most are labeled as efficacious against Streptococci and Staphylococci. While there is limited research to differentiate among products, based on the antimicrobial classes, it is unlikely that there are significant differences among approved IMM products in efficacy for treatment of common Gram-positive pathogens. Use of antimicrobials for treatment of non-severe mastitis caused by E. coli is generally not needed (26) because these cases have a high rate of spontaneous cure (25, 26), thus very few cases of clinical mastitis benefit from use of a broader spectrum antimicrobials.

Before antimicrobials are approved for treatment of mastitis they are required to demonstrate efficacy for pathogens that are listed on the product label. Before antimicrobials are used to treat pathogens that are not listed on the label, the possibility of intrinsic resistance should be considered. Intrinsic resistance occurs when a bacterial genus or species lacks targets or possess defenses to render an antimicrobial ineffective. For example, Gram-negative bacteria are intrinsically resistant to pirlimycin because their cell walls lack a binding site. Most Gram-negative bacteria are intrinsically resistant to penicillin G, many Klebsiella spp. are intrinsically resistant to aminopencillins and intrinsic resistance to ampicillin and 1st and 2nd generation cephalosporins are common in Enterobacter spp. (32). Two IMM products (ceftiofur and hetacillin) have label claims that include efficacy for E. coli, but no products have efficacy claims for treatment of mastitis caused by Klebsiella spp. Some studies evaluating antimicrobial therapies of mastitis have included cases that are intrinsically resistant to the product (33, 34) and in these instances outcomes cannot be attributed to antimicrobial therapy. Knowledge of etiology is necessary to ensure that the spectrum of activity of an antimicrobial is appropriate for the case.

Acquired resistance refers to acquisition of resistance by a normally susceptible bacterial strain through some kind of genetic modification and is usually recognized by bi-modal distribution of minimum inhibitory concentration (MIC) values. Monitoring acquired resistance is useful to identify the potential of transmission of genetic determinants of resistance into environments and food systems. While there is little evidence that most mastitis pathogens found in N. American dairy herds have acquired widespread resistance to most IMM antimicrobials (35), a bimodal distribution of MIC values was observed for about one-third of Klebsiella pneumoniae included in a recent trial (25) and susceptibility of pathogens that are not included on product labels should be monitored on a regular basis.

Therapeutic success is driven by both pathogen factors and cow characteristics (30, 36–38). Effective bacterial clearance depends on a robust immune response and factors such as parity, stage of lactation, and history of previous clinical or subclinical mastitis cases should be considered when evaluating efficacy of treatments administered for clinical mastitis.

Studies included in this review were retrieved by searching databases and web platforms using PubMed, Web of Science, and Google Scholar. Addition studies were added by reviewing bibliographies of relevant papers. Boolean search terms were used and included mastitis, bovine, clinical, randomized, non-inferiority and treatment. Only natural exposure clinical trials that utilized randomized or systematic allocation to evaluate antimicrobial treatments and were published since 2000 in English language journals were retrieved. Only studies from the last 20 years were included because there have been considerable changes in the distribution of pathogen from contagious organisms (such as S. aureus and S. agalactiae), to a more diverse mix of environmental organisms (5), use of antimicrobials is increasingly discouraged and management of clinical mastitis has gradually shifted to selective therapy of clinical cases using on-farm culture.

Studies that did not define the antimicrobial therapy or included only non-antimicrobial therapies were excluded. Studies evaluating homeopathic or herbal treatments were also excluded. While thorough, the search was not systematic and although no studies that met inclusion criteria were excluded, it is likely that some qualifying studies were missed.

Despite the global importance of bovine mastitis, relatively few clinical trials that evaluated antimicrobial treatments of clinical mastitis were identified (Table 1). While a systematic review or meta-analysis is the ideal method to summarize and compare studies, the wide diversity of study designs, variation in outcomes, differences in pathogens and treatment protocols included in mastitis trials creates a challenging situation relative to use of this method. The limited number of trials evaluating antimicrobial therapy used for treatment of bovine mastitis has been previously noted by authors of a systematic review who were unable to identify sufficient papers to establish networks to evaluate bacteriological cure and were unable to reach a conclusion about clinical efficacy of antimicrobials (59). Studies included in this review were conducted in the U.S. or Canada (n = 9), European Union (n = 8), New Zealand (n = 6), Brazil (n = 2) and Mexico (n = 1). The 26 studies included 65 study arms that included IMM therapy containing a single antimicrobial (n = 28), IMM therapy containing combination products [n (60) = 13], combined IMM and systemic therapy (n = 5), systemic therapy only (n = 11), non-antibiotic treatments (n = 2) and no treatment (negative control; n = 6). Antibiotic classes included penicillins and extended spectrum penicillins, penicillin and aminoglycoside combinations, 1st−4th generation cephalosporins, lincosamides, a lincosamide combined with an aminoglycoside, tetracycline combined with several other classes, fluoroquinolones and macrolides. No studies were replicated, and a variety of outcomes were reported.

Mastitis efficacy trials are challenging to perform and use different study designs. Some trials are performed using experimental challenge where cows are purposefully infected using a well-described bacterial strain (61, 62). Experimental challenge studies are useful for answering narrow research questions, but natural exposure trials are preferred for evaluating treatments and no challenge studies were included. Natural exposure trials randomly allocate cows to treatments and compare outcomes to either a non-treated control group or a positive control group. Non-treated (“negatively-controlled”) control groups are considered the gold standard for determining efficacy and are able to determine if treatments improve (“are superior to”) outcomes as compared to non-treated cows (or cows treated with a placebo). The null hypothesis in a superiority trial states that treatments are equally effective while the alternative hypothesis states that they differ. The inclusion of non-treated animals allows the determination of spontaneous cure so the additional benefit of treatment can be determined. Welfare concerns about not treating cows affected with mastitis have limited the number of negatively controlled trials that have been performed and only 6 of the 26 trials included in this review were negatively controlled. Only one of the negatively controlled studies included Gram-positive pathogens (39).

Positively controlled clinical trials are frequently used to assess mastitis treatments (Table 1) and can be designed to demonstrate superiority or non-inferiority. Of positively-controlled studies included in this review (n = 20), 6 were specifically designed to test non-inferiority, 1 stated that they were testing a superiority hypothesis, while the remainder (n = 13) compared outcomes among treatments but did not define the type of hypothesis that they were testing. The lack of a non-treated control group makes it impossible to separate treatment effects from spontaneous cures and superiority is rarely demonstrated. The null hypothesis in a non-inferiority trial states that the treatments differ while the alternative hypothesis states that they are non-inferior. Non-inferiority studies, include a pre-defined margin of non-inferiority for each outcome (usually 15%) and conclude that the new treatment is superior, non-inferior, inconclusive or inferior to the comparison treatment (55, 63). It is important to recognize that the inclusion of culture-negative cases in trials will skew the results toward positive outcomes (regardless of treatment) as the majority of these cases have achieved spontaneous bacteriological cure at the time of detection (22, 23). Mathematical realities dictate that inclusion of a large proportion of culture-negative and non-severe Gram-negative cases in non-inferiority trials will almost always result in a finding of non-inferior unless the “true-efficacy” of one of the products is very low. Of 6 non-interiority trials included in this review, 4 enrolled cases regardless of etiology (including culture-negative) and all 6 concluded that the “new treatment” was non-inferior (or inconclusive) to the comparator product. All of the non-inferiority trails included in this review evaluated commercially available products which infers that the drug approval process resulted in an acceptable level of efficacy for at least some outcomes. Outcomes of non-inferiority trials should always be evaluated relative to the distribution of pathogens enrolled in the study with emphasis on the number of enrolled cases that would likely be within the spectrum of activity of the products being compared.

The purpose of antibiotic treatment is to enhance clearance of bacterial pathogens and treatment efficacy is initially evaluated based on estimates of BC. This outcome is very relevant for approving new products but is rarely evaluated in a clinical setting. Bacteriological cure is assessed by comparison of recovery of bacteria from milk samples collected when the case is detected to recovery of the same isolate from milk samples collected at various intervals after treatment is completed. Sampling strategies and intervals used to define BC vary among studies (Table 3). Some researchers defined BC based on results of a single post-treatment milk sample, while other studies require the absence of the causative pathogen in multiple samples (usually collected at 7-d intervals). In the included studies, apparent BC ranged from about 27–95%, but it is important to recognize that comparisons among studies are not accurate due to differences in the distribution of pathogens and sampling periods. The overall means and ranges of BC were 69% (27–95%; n = 35), 68% (33–91%; n = 13) and 60% (38–87%; n = 6) for all IMM antimicrobial therapies, systemic or systemic and IMM therapies combined or no antimicrobial treatment, respectively.

Bacteriological cures result from the combined effect of the host immune response (spontaneous cure) and the effect of treatment (treatment cure) and the value of antimicrobial therapy is greatest for pathogens that have a low rate of spontaneous cure and high rate of treatment cure (such as IMI caused by S. agalactiae). Among mastitis pathogens, expected rates of spontaneous bacteriological cure vary widely. The greatest contrast is between expectations of spontaneous bacteriological cure of IMI caused by S. aureus (close to zero) and Escherichia coli (about 90%) (25, 26). Limited efficacy of antibiotic therapy is well-documented for IMI caused by S. aureus (38, 64) and some pathogens lack targets for antimicrobial therapy (e.g., yeast, Prototheca bovis, Mycoplasma spp. and others) and are considered to be intrinsically resistant to all approved antimicrobials. It is important to reiterate, that even with highly efficacious drugs the benefit of antimicrobial therapy is only for the cases that do not achieve spontaneous bacteriological cure; thus, the marginal value of antibiotic therapy decreases for cases caused by E. coli or other pathogens with high rates of spontaneous cure.

Among the 23 trials that evaluated bacteriological cure, statistically significant differences among treatments were observed in only 4 trials while 16 reported no significant differences and 3 trials concluded the evaluated treatments were non-inferior (Table 3). Two of the trials that reported significant differences in BC enrolled only Gram-negative cases and compared IMM antimicrobial treatment to non-treated controls (25, 40). The distribution of pathogens in both studies included a considerable proportion of Klebsiella spp. which influenced overall BC of the combined groups. Fuenzalida and Ruegg (25) identified an interaction of pathogen by treatment group and reported BC of 18% (non-treated K. pneumoniae), 74% (treated K. pneumoniae), 97% (non-treated E. coli) and 99% (treated E. coli). Schukken et al. (40) reported significant differences in BC for cases caused by both E. coli and Klebsiella spp. but this study is unique in reporting exceptionally low spontaneous cure rates of E. coli cases, which the authors attributed to persistent infections. In spite of finding significant differences in BC, neither of these studies reported biologically important differences in important clinical outcomes. Taponen et al. (46) reported BC of clinical mastitis caused by S. aureus that were either β-lactamase negative or positive and were treated with either systemic penicillin & an IMM combination product containing penicillin & neomycin or received systemic penicillin alone. While a significant difference was identified among treatments, BC was dramatically decreased for β-lactamase positive organisms. Truchetti et al. (50) compared shorter (2d) vs. longer (5d) therapy using IMM ceftiofur and reported a significant difference in BC but no differences in any clinical outcomes. Over 30% of cases enrolled in this study were culture-negative and 26% were caused by S. aureus thus the impact of 3d difference in therapy was likely biologically irrelevant. In general, no clear relationship between BC and important clinical outcomes (such as new IMI, clinical cure, recurrence, SCC, milk yield or culling) were apparent (Table 4). Thus, while achieving BC is the goal of antimicrobial therapy, the finding of differences in BC in research trials does not appear to be predictive of differences in clinical outcomes.

New Intramammary infections are typically defined based on recovery of a different pathogen in follow-up milk samples but in the study that enrolled culture negative cases (22), NIMI was defined as recovery of any pathogen in the 14 and 21 d follow-up samples. While this outcome was not significant in any trials, recovery of both the etiological agent and new pathogens from follow-up samples after treatment is usually greater at earlier sampling periods (22, 25) as compared to samples collected after 14 d and in some instances may reflect differences among pathogens in duration of time to achieve both spontaneous and treatment clearance.

Almost all studies (n = 24; Table 5) evaluated “clinical cure” but the definition of this outcome varies enormously. Most researchers (67%) defined CC based on observations that the milk and/or the mammary gland returned to normal appearance, but the day of observation varied from 2 to 28 days after treatment and some relied on single observations, while other researchers required multiple observations. Of studies that defined CC based on observation of clinical signs, the proportion achieving CC was least (CC<15%) for studies that performed observations very early and evaluated systemic therapies (27, 41). For CC estimated based on visual observations after day 3, values of CC ranged from 25% to 98% with a median value of 81%. Other definitions of CC included retreatment within a defined time period (4 studies), use of scoring systems (2 studies) or a combination of methods (2 studies). Of 7 studies that reported significant differences in CC among treatments, 4 defined CC based on retreatment, 2 used comparison of defined scoring systems and 1 evaluated CC within 4 days of treatment. While achieving CC is the practical goal on farms, this outcome is not useful to determine effectiveness of antimicrobial therapy. In most cases of clinical mastitis, inflammation is self-limiting and regardless of BC, and for the majority of cases, milk will return to normal appearance by day 7 (30). There is almost no evidence that selection of an antimicrobial has a significant impact on this outcome and CC should not be used to make decisions about treatment efficacy.

Similar to CC, a variety of definitions and sampling periods were used to assess SCC responses in the 14 studies that included this outcome (Table 6). Seven studies each assessed SCC at the quarter or composite level and dilution of healthy milk from unaffected quarters likely influenced results of studies that used composite milk samples. Sampling periods ranged from 7 to 90 days after treatment, and all studies that assessed SCC at multiple periods reported a gradual decline in SCC as time passed. While some researchers compared linear scores, other researchers compared the proportion of samples that were defined as “healthy” based on a threshold or either 200,000 or 250,000 cells/mL. Of the 14 studies that included this outcome, only 3 reported significant differences in their measure of SCC. One researcher used the California Mastitis Test on quarter milk samples collected at day 36 post-treatment and reported significantly fewer quarters below “trace score” for quarters that did not receive IMM treatment but received frequent milking (39). Interestingly, no difference was seen in non-treated control quarters that were not frequently milked. In another study, fewer quarters that received IMM antibiotics (as compared to systemic) achieved SCC <250,000 cells/mL by 22 days post-treatment (45) and a 3rd study reported significantly lower SCC at days 7 and 14 but those effects were not significant by day 22 (52). When assessing SCC responses, it is important to recognize that quarter-level measurements will more accurately reflect ongoing inflammation that may indicate persistent IMI. When BC is achieved SCC will gradually decline and the speed of return to a “healthy” level is influenced by etiology. Assessment of SCC responses should be performed at the quarter level and should continue for at least 21 days. Somatic cell count responses are a practical outcome that can be used as an indicator of treatment success on farms, but a gradual (rather than immediate) decline should be expected. When using composite milk samples, a lower threshold (<150,000 cells/mL) may help prevent misclassification of on-going subclinical infections that can result after failure to cure clinical cases.

A few studies have evaluated other important clinical outcomes. Only 4 studies reported recurrence as a distinct outcome (22, 25, 33, 43) but several included recurrence in their definition of clinical cure (or “clinical failure”) (53, 54). Like other outcomes, recurrence can be defined at either the quarter or cow level, but when used to assess treatment this outcome should always be assessed relative to the affected quarter. Recurrence ranged from about 5–30% and was strongly influenced by additional risk factors such as parity (older cows are at greater risk of recurrence), etiology (culture positive are at greater risk as compared to culture negative), and increased milk yield. While two studies reported significant differences in recurrence based on treatment, this outcome is influenced by many other factors and should be interpreted cautiously. Post-treatment milk yield is an obviously important outcome that requires a prolonged follow-up period and has been evaluated infrequently in mastitis trials (22, 25, 39, 40, 43, 54). Of the 6 studies evaluating this outcome, 4 included a non-treated control group and the only significant finding was one study that reported non-treated control animals had the greatest post-treatment milk yield (39). Similarly, a significant difference in retention (or culling) after treatment was reported in only 1 of 8 studies that evaluated this outcome (22, 25, 27, 33, 40, 49, 55, 65). Culling is a very difficult outcome to assess as it is influenced by many factors including non-blinded studies that allow farmers to remove cows without a withholding period if they have not received antimicrobial therapy. While all of these outcomes are relevant and useful for dairy farmers, there is insufficient evidence to suggest that they are influenced by choices made about mastitis treatment.