Materials and Methods

Design—The study was designed as a randomized, double-blinded, placebo-controlled clinical trial and was approved by the University of Tennessee Institutional Animal Care and Use Committee.

Animals—Mares were enrolled by participating veterinarians from 6 private equine practices during the 2004 and 2005 breeding seasons: 4 practices were located in Lexington, Ky, and 2 practices were located in Ocala, Fla. Criteria for inclusion in the study included a signed consent form and a diagnosis of persistent endometritis that had been made on the basis of uterine cytologic findings from a specimen obtained prior to or at least 5 days after breeding. Mares were excluded if the cytologic specimen had been obtained during a foal heat (ie, first estrus after foaling) or if external genital or cervical conformation was abnormal to the extent that, in the opinion of the attending veterinarian, the likelihood of pregnancy would be adversely affected. Mares that were not bred within 30 days of administration of the first treatment with P acnes or placebo, those previously treated with an immune stimulant during the current breeding season, and those that were donors or recipients for embryo transfer were also excluded from the study.

Laboratory testing—Microbial culture and cytologic evaluation were performed on uterine specimens obtained transcervically per vagina with a sterile guarded culture swab. The swab was placed in transport medium and underwent standard microbiologic methods for microbial culture. The cytologic smear was air dried, stained,^c and evaluated by the participating veterinarian or by a laboratory technician from the participating hospital under oil immersion at 100X power. Uterine cytologic findings were graded as < 2, 2 to 4, or r 5 neutrophils/mean hpf. The criterion for cytologic diagnosis of endometritis (positive cytologic results) was detection of ≥ 5 neutrophils/mean hpf.

Procedures—Participating veterinarians were given a packet for each mare enrolled. The packet included an informed consent form, data collection form, adverse reaction reporting form, and 3 vials containing either placebo or a commercially available preparation of P acnes. The number of packets provided to participating veterinarians for each breeding season was determined on the basis of the estimated number of mares they expected to enroll during that season. Mares were stratified by veterinarian and randomized in blocks of 2 as they became eligible for the study by use of a sequential number on each packet that indicated the order of treatment. The placebo treatment consisted of a 1:100 dilution of a 20% fat emulsion^d labeled for IV administration and used in parenteral nutrition that was compounded at a local pharmacy to resemble the P acnes suspension in color and consistency. The placebo emulsion was placed in vials identical to those containing P acnes. Vials (3/mare) were labeled with a random number code to identify the treatment as P acnes or placebo, and the key to the treatment code was available only to the principal investigator (BWR). The P acnes suspension or placebo was administered IV in the left jugular vein at a dose of 1 mL/113 kg (1 mL/250 lb) on days 0, 2, and 6. The right jugular vein was to be used if the left jugular vein was not accessible. No additional treatment with P acnes or placebo was given during the breeding season, and no attempt was made to alter additional treatments prescribed by attending veterinarians other than to prohibit subsequent administration of an immunostimulant. Owners had the option to withdraw a mare from the study at any time without penalty and to have the treatment code broken.

Measurements—Data collection forms included questions relating to demographics (eg, farm, veterinarian and mare identification, breeding history, treatments given and dates of administration, breeding dates, and date and results of pregnancy examinations). Treatments administered, in addition to P acnes or placebo, were recorded and grouped into 1 of the following 3 categories: uterine lavage, oxytocin or cloprostenol administration, and antimicrobial administration. Pregnancy status was defined as a diagnosis made on the basis of ultrasound examination findings ≥ 14 days or manual palpation of the uterus ≥ 45 days after the most recent breeding date. A live foal was defined as a foal born alive that was able to stand and nurse within 12 hours of birth as verified by the attending veterinarian, owner, or breeding farm manager or, for registered Thoroughbred mares, a foal registered with the Jockey Club (the registering body for Thoroughbreds in the United States, Canada, and Puerto Rico) within 1 year after the expected foaling date. For mares without a live foal reported to the Jockey Club and mares that were not Thoroughbreds, the participating veterinarian or breeding farm manager, mare owner, or both were contacted to verify the mare's foaling status. Fetal loss was defined as a mare that had a diagnosis of pregnancy during the current breeding season but did not produce a live foal in the subsequent year. Adverse reactions to P acnes or placebo were recorded each time the mare was treated and at the first pregnancy examination. Adverse reactions were defined as tremors, inappetance, signs of pain or swelling in the neck area, rectal temperature > 39.2°C (102.5°F), or other signs (to be specified) that were observed by the attending veterinarian, owner, or farm personnel after treatment with P acnes or placebo. Reactions were recorded at any time after administration of the first injection of P acnes or placebo and up to 7 days after the third injection.

Statistical analysis—Univariate associations of historical, demographic, and breeding management variables of a categoric nature with treatment group and pregnancy status were examined by use of a χ² or Fisher exact test,^e depending on whether an expected table cell frequency was < 5. Continuous variables were evaluated with a t test or the Wilcoxon 2-sample test,^f depending on the distribution of data. The folded F test was used to determine equality of variance for t test comparisons. Normality was visually evaluated with a stem-and-leaf diagram and by use of the Shapiro-Wilk W test.^g Multivariate logistic regression models^h were constructed to evaluate the effect of treatment on pregnancy resulting from the first cycle bred, pregnancy at any time during the current breeding season, and live foal rate among mares that were bred within 2 days before or up to 30 days after administration of the first P acnes treatment or placebo. Variables evaluated for inclusion in the logistic regression models were year entering the study, treatment with P acnes or placebo, mare age, cervical and external conformation, current reproductive status (eg, barren, maiden, or foaling), presence of uterine infection, additional treatments administered (eg, antimicrobials, oxytocin or cloprostenol, uterine lavage, or specified combinations of treatments), first breeding within 2 days before to 8 days after the first treatment with P acnes or placebo, number of cycles bred after treatment with P acnes or placebo, and days between first day of treatment and first breeding. Categoric variables were coded as 0 if present and 1 if absent. Model building was performed by use of backward selection. Variables with the highest P value associated with the Wald χ² statistic were removed from the model in a sequential manner, and the effect on deviance in the −2 log likelihood ratio was observed. The criterion for retention of initial main effect variables in the preliminary model was an observed deviance of ≥ 2.072 and P ≤ 0.15. Variables retained in the preliminary model were subject to the same backward selection procedure. Once main effects were identified, all 2-way interactions among main effect variables were evaluated by use of the same backward selection process used to identify main effects. Each variable included in the final model was evaluated by use of backward selection with a cut point of 2.706 (P ≥ 0.10) for retention. The adjusted effect of treatment for the years 2004 and 2005 was evaluated with contrast statements that included the effects of treatment and the interaction term treatment × year. The estimate of the effect of treatment over both years was determined with a contrast that included the intercept, treatment, and age of the mare.

Results

Ninety-five mares were enrolled in the study. The number of mares enrolled in 2004 (n = 54 [57%]) was similar to the number enrolled in 2005 (41 [43%]). Mean ± SD age of all mares was 12.3 ± 5.6 years, and 85 of the 95 (89%) mares were Thoroughbreds. Eighty-one (85%) mares had been bred during the previous year, and 89 (94%) had been bred at least once previously. The numbers of mares that were barren (not in foal; n = 42 [44%]) and that had foaled (46 [48%]) were similar, and 7 (7%) mares were maidens (had not been bred). External genital and cervical conformation were considered to be normal in 74 (78%) and 89 (94%) mares, respectively. In the remainder, conformation defects were judged to not be severe enough to affect reproductive performance. Uterine microbial culture was not performed in 1 mare, but in 37 of the 94 (39%) remaining mares, results of microbial culture were negative. Most (57/94 [61%]) mares had 1 or more infections; 43 of 57 (75%) mares had a bacterial infection, 10 (18%) had a yeast infection, and 4 (7%) had bacterial and yeast infections. Among mares with bacterial infections, the most frequent cause was E-hemolytic Streptococcus spp, either alone or in combination with other agents (n = 31/57 [54%]). In addition to P acnes or placebo, additional treatment with oxytocin or cloprostenol was administered in 84 (88%) mares, antimicrobials were administered in 82 (86%) mares, and uterine lavage was performed in 72 (76%) mares. All but 1 mare received ≥ 1 treatment in addition to P acnes or placebo. The predominant breeding method (85/95 [89%]) was natural service.

Characteristics of mares in the experimental group were not significantly different from characteristics of control mares (Table 1). The timing and frequency of breeding was also similar between the 2 groups (Table 2). When demographic characteristics, physical examination findings, additional treatments administered by veterinarians, microbial culture results, and breeding procedures were compared between pregnant and nonpregnant mares, the only significant finding was that mean age of pregnant mares (mean ± SD, 11.3 ± 5.3 years) was significantly lower than that of nonpregnant mares (15.0 ± 5.5 years; P = 0.01; Table 3). As expected, the days to last breeding (P = 0.05) and total number of cycles bred were greater for nonpregnant mares; however, the total number of cycles bred was not significantly different between pregnant and nonpregnant mares (P = 0.08). No adverse reactions were reported for mares in either the P acnes or placebo-treated control group.

Pregnancy rates among mares that were first bred on a cycle that occurred during a 10-day interval from 2 days before to 8 days after administration of the first treatment with P acnes or placebo were evaluated. Mares treated with P acnes and bred during this interval were significantly (10/13; P = 0.03) more likely to become pregnant at first service, compared with placebo-treated mares (8/21; Table 4). Mares treated with P acnes in which the first breeding occurred during this same 10-day interval were also significantly (13/13; P = 0.01) more likely to become pregnant at any time during the given breeding season, compared with placebo-treated mares (13/21). Pregnancy rates at first service or later among mares whose first breeding was on a cycle within 30 days of the first treatment were not significantly different between groups.

Results of the multivariate analyses to assess the effect of treatment with P acnes on pregnancy and live foal rates were summarized (Table 5). Factors retained in the models to assess the effect of P acnes on pregnancy were treatment, breeding for the first time during the period from 2 days before to 8 days after the last treatment with P acnes or placebo, mare age, year, and the 2-way interaction between mare age and year and treatment and year. Pregnancy rates on first service for mares first bred within 30 days of the first treatment with P acnes were significantly higher (OR, 6.2; 95% CI, 1.6 to 24.9) than for placebo-treated controls in 2004; however, results for mares bred in 2005 and for mares bred during 2004 and 2005 combined were not significantly different between groups. Pregnancy at first service or any subsequent service during the 2004 breeding season was significantly greater among mares treated with P acnes (OR, 26.6; 95% CI, 2.9 to 241) than among control mares, and when data from 2004 and 2005 were combined, the rate was also higher for treated than for control mares (OR, 10.8; 95% CI, 3.2 to 36.2). Fewer cycles bred after treatment with P acnes or placebo and younger age on enrollment in the study were the only factors significantly associated with live foal rate and included in the logistic regression model. Two mares in the P acnes group were excluded from the analysis of live foals. The first mare was a non-Thoroughbred and was not pregnant when lost to follow up during 2004 and the second mare was pregnant during 2005 but died from an unrelated cause prior to the expected foaling date. Observed live foal rates were 32/44 (73%) among all P acnes treated mares and 21/49 (43%) among controls (P = 0.004). Pregnancy loss occurred in 11 of 32 (34%) placebo-treated mares and in 6 of 38 (16%) mares in the P acnes group, a difference that was not statistically significant (P = 0.07). In the multivariate analysis, mares treated with P acnes were significantly more likely to produce a live foal (OR, 3.9; 95% CI, 1.6 to 9.8), compared with placebo-treated controls.

Discussion

Results of the present study indicated that administration of P acnes was effective in increasing the pregnancy rate at first service and the overall pregnancy rate in 2004. In addition, the overall pregnancy and live foal rates were higher for the entire study period. These data also indicated that the effect on pregnancy rate was most noticeable when mares were bred during the interval from 2 days before to 8 days after the first P acnes treatment.

Data revealed a significant difference in the effect of P acnes treatment among mares enrolled during the 2004 and 2005 breeding seasons. Although the cause could not be determined with certainty, there was a significantly greater proportion of mares bred within 8 days of the first treatment with P acnes or placebo in 2004, compared with 2005. Subgroup analysis also revealed that mares treated with P acnes were significantly more likely to become pregnant when bred within 8 days of first treatment. The difference in pregnancy rates in 2004 and 2005 may be related to the time of breeding relative to initiation of treatment.

In a previously published field study¹⁸ on use of P acnes, mares with endometritis were not randomly allocated to treatment groups, no additional treatments for endometritis were described, and study investigators were not blinded. However, results of that study provide indirect evidence in support of our findings. Although the conception rate among nontreated mares was not reported in that study,¹⁸ the conception rate among mares treated with P acnes was high (86%). In addition, the author reported on a subset of mares in which uterine cytologic evaluation yielded positive results for endometritis on the first heat after treatment in 15 of 16 (94%) control mares and only 1 of 12 (8%) mares treated with P acnes.

When administered IV, P acnes is readily taken up by macrophages and persists longer inside the macrophage than most antigens before being degraded. The effect on the immune cascade includes release of cytokines that increase activity in the general immune defense system.¹² It was recently reported¹⁹ that P acnes induces a systemic immune response in healthy horses through CD4⁺ lymphocyte activation in response to increased interferon-G and interleukin-2 gene expression and a modest increase in interleukin-10 expression.

There is a growing body of knowledge to suggest a mechanism for the action of P acnes, both systemically and locally, on equine pulmonary mucosal surfaces. Prior reports suggest that P acnes,²⁰ interferon,²¹ and inactivated bacterial products²² downregulate the inflammatory response while increasing mucosal immune function. In healthy horses, P acnes modulated the inflammatory response in broncheoalveolar lavage fluid.²⁰ The numbers of CD4⁺, CD8⁺, CD5⁺, and major histocompatibility complex-II lymphocytes and opsonized phagocytic activity of leukocytes were all decreased in broncheoalveolar lavage fluid on day 14 after treatment. The authors concluded that administration of P acnes modulated the pulmonary mucosal immune response by activating immunoregulatory cell function without exacerbating inflammation and initiating tissue damage.²⁰ Results of a previous study²³ involving immunofluorescence and immunoperoxidase techniques confirmed that the uterus is part of a common mucosal system. Therefore, administration of P acnes may act to improve fertility among mares with persistent endometritis by modulating the inflammatory response and transforming the uterus into a more hospitable environment for fertilization, implantation, and maintenance of pregnancy.

Studies have been performed to evaluate the integrity of the uterine mucosal immune system in horses. Most investigators agree that there is reduced phagocytic capability in uterine neutrophils among mares susceptible to persistent endometritis.^24–27 However, there is less agreement on whether the chemotactic ability of these neutrophils is impaired. The response among mares in the present study that were treated with P acnes supports the hypothesis of lowered uterine mucosal immune function in mares with persistent endometritis; however, the exact defect in mucosal function and mechanism of action for P acnes remain unclear.

The strength of the present study lies in the design as a double-blinded randomized clinical trial. Randomization tends to balance treatment groups with regard to known and unknown factors that may influence pregnancy and live foal rates, so differences observed in these outcomes will more accurately represent the effect of treatment. When historical and clinical characteristics of mares were compared among treatment groups, no significant differences were observed, suggesting that the randomization procedure was effective. In addition, when adjustments were made for significant variables in the logistic regression models, there was a strong association between treatment with P acnes and pregnancy on first cycle bred or any time ≤ 30 days after the first treatment in 2004. There was also a significant increase in pregnancy and live foal rates for mares bred ≤ 30 days after first treatment with P acnes when data were combined for both years. Although lower, the overall pregnancy rate among mares treated with P acnes (74%) in this study was not substantially different from the 87% reported in the earlier study.¹⁸ The ability of mares to eliminate bacteria from the uterus has been correlated with age.²⁸ The significant decline in fertility associated with the age of the mares in the present study was identified in univariate analysis and in multivariate models to evaluate factors associated with pregnancy and was consistent with findings from prior studies.^29,30

There are several potential sources of bias that could have affected outcome in the present study. Treatments other than P acnes or placebo were given to mares at the discretion of the attending veterinarians and may have influenced the outcomes measured. Furthermore, only treatments that were administered concurrently with P acnes or placebo were recorded. We attempted to control for the effect of additional treatments by stratification and randomization of mares to treatment groups by veterinarians. Mares were blocked in groups of 2 and randomized within blocks so that an individual veterinarian would have an equal number of mares in each treatment group. The assumption was that treatment protocols would be more consistent within than among participating veterinarians. Moreover, the proportions of mares that were initially treated with uterine lavage and antimicrobial treatment, with or without oxytocin or cloprostenol, in addition to the P acnes or placebo treatments were similar among treatment groups. Data on treatment with oxytocin or cloprostenol were combined because there is no clear evidence at present that either treatment is superior for removal of uterine fluid or subsequent fertility.^31,32 The study design did not take into consideration the effect of stallion on the measured outcomes; however, it is customary to conduct fertility examinations on purebred stallions prior to the breeding season, and our assumption was that all were fertile or that randomization would control for the effect of stallion.

Causes of endometritis can be subdivided into 4 etiologic categories, including endometritis (ie, chronic degenerative endometritis), sexually transmitted diseases, persistent mating-induced endometritis (persistent endometritis), and chronic infectious endometritis. These categories are not absolute. Mares can change categories within or between breeding seasons or fit into more than 1 category.^33,34 Mares in certain categories may not have the same probability of becoming pregnant or carrying a live foal to term. Our criteria for enrollment did not enable identification of mares by subcategory, and we relied on randomization to balance the distribution of categories among the treatment groups. In a recent study,³⁵ it was found that a small proportion (4%)ⁱ of mares with uterine cytologic findings negative for endometritis had positive culture results for microbial infection. Such mares would have been excluded from the present study. Finally, 2 mares from each treatment group were bred on 2 days, and 2 mares were bred on 1 day prior to the first treatment being administered. Administration of P acnes upregulates the genes for interferon, interleukin-1, interleukin-6, and tumor necrosis factor within approximately 3 hours of administration, and multiple doses elicit pulses of immune stimulation.³⁶ Because the embryo does not reach the uterus until 5 to 6 days after ovulation in mares, administration of P acnes 1 to 2 days after breeding should allow sufficient time for activation of the uterine mucosal immune system and changes to occur in the uterine environment prior to implantation.

Our findings apply to broodmares that meet the cytologic criteria for a diagnosis of endometritis and are housed, bred, and managed under good-to-optimal conditions. The effect on pregnancy was most pronounced when mares were bred b 2 days before to 8 days after administration of the first P acnes treatment. Additional studies are needed to estimate the effect of P acnes administration on conception and live foal rates under different conditions of broodmare management and to determine whether repeated administration on subsequent breeding cycles affects overall pregnancy and live foal rates.