SEARCH

SEARCH BY CITATION

Keywords:

  • Trichomoniasis;
  • GRADE;
  • Systematic review

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

This review assesses the efficacy of whole cell Tritrichomonas foetus vaccine to prevent and treat trichomoniasis in beef cattle. Three databases were searched in June 2012. Eligible studies compared infection risk, open risk, and abortion risk in heifers or infection risk in bulls that received vaccine compared with no vaccine. Study results were extracted, summary effect measures were calculated, and the quality of the evidence was assessed. From 334 citations identified, 10 were relevant to the review. For heifers, there was limited evidence of moderate quality to assess the impact of vaccination on infection risk (RR, 0.89; P = .16; 95% CI, 0.76–1.05; 6 randomized and 4 nonrandomized studies; 251 animals) and open risk (RR, 0.80; P = .06; 95% CI, 0.63–1.01; 6 randomized and 5 nonrandomized studies; 570 animals). The quality of the body of work describing the impact of vaccination on abortion risk was low (summary RR, 0.57; P = .0003; 95% CI, 0.42–0.78; 3 randomized and 2 nonrandomized studies; 176 animals). The quality of evidence was very low for duration of infection (mean difference, −23.42; P = .003; 95% CI, −38.36 to −7.85; 2 randomized and 3 nonrandomized studies; 163 animals). Although the summary effect measures suggest a benefit to vaccination, due to publication bias the effect reported here is likely an over estimate of efficacy. For bull-associated outcomes, the evidence base was low or very low quality.

Abbreviations
RR

risk ratio

CI

confidence interval

A critical review and meta-analysis were performed to estimate the efficacy of killed, whole-cell Tritrichomonas foetus (T. foetus) vaccine with regard to incidence of T. foetus infection in heifers and bulls, duration of T. foetus infection in heifers and bulls, pregnancy percentage and abortion risk in heifers, and the ability of the vaccine to clear T. foetus-infected bulls of the infection. The motivation for conducting the review was the reemergence of T. foetus infection as a cause of reproductive failure in US Midwest cow-calf herds.[16-18]

Bovine trichomoniasis is a sexually transmitted disease caused by Tritrichomonas foetus.[1-3] In bulls, T. foetus lives in the smegma of the epithelial lining of the penis, prepuce, and distal urethra, and is transmitted to females through infected preputial secretions.[4-6] Infected bulls older than 3–4 years of age often are chronically infected. In cows and heifers, the most common sequela to infection is reproductive failure, but overt clinical signs of infection can include endometritis, salpingitis, placentitis, abortion, and potential subsequent pyometra.[11-13] It may take months for cattle to regain fertility.[12] Based on a simulation model, a herd with a 20–40% prevalence of T. foetus infection in the breeding bulls might expect a 14–50% reduction in annual calf crop size, a growing period decreased by 12–30 days, and weaning weights decreased by 22–53 pounds.[14] The result is wide variability in weaning weights, forcing producers to sell calves at lower weights or incur higher feeding costs.[7]

Common approaches to prevention of T. foetus introduction into a herd include biosecurity practices: limiting the potential for bulls from neighboring properties to mate with the herd, purchasing only virgin bulls, purchasing older bulls confirmed to be T. foetus negative, and artificial insemination.[7, 19] When biosecurity measures are not practiced, the efficacy of these measures cannot be ensured, or if further assurance of a T. foetus free herd is desired, vaccination may also be considered. Currently, 1 T. foetus vaccine is available on the US market,1 a killed, whole-cell protozoan vaccine indicated for vaccination of healthy cattle as an aid in the prevention of disease caused by T. foetus.[7, 20]

Given the emergence of T. foetus, the review question was, “What is the magnitude of reduction of infection risk, open risk, abortion risk and duration of infection in heifers, bulls or both that received a whole-cell killed T. foetus vaccine compared with no vaccine?” Furthermore, we used a slightly modified GRADE Summary of Findings and Evidence Profile Table[21, 22] to present the results in a format previously not used in veterinary science.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

Literature Selection

The approach to the review was designed by 1 of the authors (AOC). No protocol was registered or externally reviewed. One author (PB) had experience conducting 1 systematic review and another author (AOC) had conducted numerous systematic reviews and meta-analyses. One author had no prior experience with systematic reviews. The review question followed the PICO[23] format for systematic reviews: the relevant population (heifers or bulls), intervention (a whole-cell killed T. foetus vaccine), comparator (no vaccine), and outcomes (magnitude of reduction of infection risk, pregnancy loss, and duration of infection). The review was limited to whole-cell T. foetus vaccines, because the only commercially available product is a whole-cell vaccine.1 Any study that described the use of a whole-cell vaccine was used (ie, both commercially available and in-house laboratory-based whole-cell vaccines were considered relevant to the review).

To identify relevant primary research, the citation indexes PubMed, CAB Abstracts, and Agricola were searched in the first 2 weeks of June 2012. The PubMed search terms were “(Cattle OR Bovine) AND (Tritrichomonas foetus) AND (Vaccine* OR Vaccinate* OR Immunization OR Control OR Prevention).” Analogous search terms were used in CAB Abstracts and Agricola. No language or date restrictions were imposed during the search. Retrieved citations were imported into Endnote Web.2 Within Endnote Web, duplicates were removed based on title match only, manuscripts published in languages other than English were removed, as there were no funds for translation of articles, and articles published before the 1950s were excluded as prior experience suggests that few of these studies include control groups. We also hand-searched the reference lists of previously published reviews about T. foetus for relevant studies.[8-10] We contacted the manufacturer of the only commercially registered whole-cell killed T. foetus vaccine in the United States to request studies about vaccine efficacy.

To determine the outcomes considered in the review, a list of possible outcomes was extracted from the several articles considered likely to be included in the review. Three beef production experts at Iowa State University College of Veterinary Medicine were asked to rank the outcomes by relative importance, highest to lowest. The experts ranked titers to T. foetus and other immunological measures as nonimportant outcomes. Therefore, data from antibody tests were not extracted from any study, although many studies reported this outcome.

Screening of the remaining retrieved citations was conducted independently by 2 authors (PB and HN), DVM students working in a summer research program. Both authors screened all retrieved citations based on the following eligibility criteria:

  1. Did the study describe primary research?
  2. Was the research conducted in cattle associated with beef production?
  3. Did the study assess the efficacy of killed whole cell T. foetus vaccine for the prevention or control of Trichomoniasis?
  4. Did the study utilize a treatment and control group, the latter of which did not receive the vaccine?

If the response to each question was “yes”, the study was included in the review. When it was not possible to determine relevance based on the abstract and title, the full text articles were obtained and evaluated based on the same four questions.

Data Extraction

From the full text of relevant publications data on whether trials reported randomization, the intervention, and relevant outcomes were extracted by 2 authors (PB and HN). We did not extract information on study population demographics or other design features because the inclusion criteria were very narrow. The unit of concern for data extraction was the study. Some manuscripts had more than 1 study. When clarification about an outcome or data was needed, the 3rd author was consulted. Data were extracted for the following outcomes of interest: infection risk in heifers and bulls, duration of infection in heifers and bulls, open risk and abortion risk in heifers, and the ability of the vaccine to clear infected bulls of infection. The open risk refers to the number of heifers not pregnant divided by those bred.

When extracting data for days of infection, sometimes these data were reported as group means and standard error of the mean (SEM). Other times, the individual animal test results over time were reported and we manually calculated the mean days of infection and standard deviation. Furthermore, when studies reported the SEM for the days infected, we used Review Manager (RevMan)3 software to back-calculate the standard deviation for each treatment group based on the number of animals in each group and SEM.

Data Analysis

For outcomes with more than 1 study, the extracted data were entered into and analyzed using RevMan. For dichotomous outcomes, infection risk, open risk, and abortion risk, a risk ratio with a 95% CI for each study was determined. The summary effect measure comparing the risks was the Mantel-Haenszel summary risk ratio from a random effect model.[23] For the average number of days infected, a continuous outcome, the summary effect measure was the mean difference in days infected. The hypothesis was that the overall summary effect was equal to the null value (risk ratio = 1 or mean difference = 0). For all outcomes, it was expected that vaccination would decrease adverse events; therefore, the risk ratio should be <1.0 and the mean difference <0 if the vaccine is effective. A subgroup analysis based on the length of time for follow-up was conducted for 2 outcomes, infection risk and open risk in heifers. This aim of the subgroup analysis was to assess if length of follow-up was a source of heterogeneity.

Heterogeneity was assessed using the chi-square test for heterogeneity overall and within the subgroups. The null hypothesis was that heterogeneity was not present. If the P value for heterogeneity between the subgroup effects was significant (P < .1), we concluded that the subgroups were different and only then assessed and interpreted the P value for heterogeneity within the subgroup. If the P value for subgroup heterogeneity was >0.1, we concluded that the subgroup was not a source of heterogeneity. The subgroups were collapsed, and heterogeneity then was assessed across the entire population. We also reported the I2, which describes the percentage of variation across studies because of heterogeneity rather than chance.[24, 25] The data were also used to create a forest plot and funnel plot of each outcome with >1 study as well as a cumulative meta-analysis plots in meta and rmeta packages in R.4

After meta-analysis, the data were exported from RevMan into GRADEpro,5 a software package designed to guide reviewers through the process of assessing the quality of the scientific literature contributing to evidence base for each outcome, and to generate a Summary of Findings and Evidence Profile table. The GRADE system stands for “The Grading of Recommendations, Assessment, Development, and Evaluation”.[21]

The GRADE system is based on the presence of inconsistency,[26] indirectness,[27] imprecision,[28] and risk of bias[29, 30] existing in evidence base contributing to each outcome. The evidence base is considered to show evidence of inconsistency if there is a wide variation in point estimates, lack of overlap in CI, or evidence of heterogeneity among studies.[26] The evidence base is considered to have evidence of indirectness if the study populations, interventions, or outcomes are different from those of interest.[27] The evidence base is considered to have evidence of imprecision if the studies have wide CI. This could result from a sample size that is smaller than the number generated by a conventional sample size calculation for an adequately powered trial.[28] The evidence base is considered to have evidence of risk of bias if the studies in the review fail to report concealment of allocation, blinding, have incomplete accounting of subjects, large loss to follow-up, show selective outcome reporting, or other factors such as recruitment bias, stopping early, or using unvalidated outcome measures.[30]

The presence of inconsistency,[26] indirectness,[27] imprecision,[28] or risk of bias[29, 30] decreases the quality of evidence grade assigned to the evidence base. The quality of the reviewed body of work can be increased if the observed effects are large, if the only possible confounding bias is toward the null and if evidence of a dose response is available. The GRADE quality scales and interpretations that GRADE attaches to these terms are as follows:

  • Very low quality–indicates that we have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect.
  • Low quality–indicates that our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect.
  • Moderate quality–indicates that we are modestly confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.
  • High quality–indicates that we are very confident that the true effect lies close to that of the estimate of the effect.

For this review, the outcomes reported were not subjective (ie, infection detected by culture or pregnancy or abortion), and thus the impact of failure to blind on the risk of bias was considered minimal. Failure to randomize was considered important, but its impact was most likely to be seen as publication bias. The impact of multiplicity was also considered important, but it also was thought that it would manifest itself as selective reporting of outcomes with P values <.05 or publication bias. Suspicion of risk of publication bias was considered present if there was evidence that studies showing positive results were more likely to be published, whereas negative results were more likely to be excluded from the literature.[29] Publication bias may be inferred using a funnel plot and cumulative meta-analyses.

The rankings of the outcomes obtained from the three production specialists at the start of the review process were used to categorize the outcomes as critical or important in GRADE ranking system. Critical outcomes included the prevention of T. foetus infection in heifers and bulls. Important outcomes included clearance of infection in bulls after therapeutic use of the vaccine, and all other outcomes in heifers.

After ranking the outcomes and grading the body of evidence, the Summary of Findings table and Evidence Profile were created from the GRADEpro work. GRADE has 2 components, 1 that refers to the grading and presentation of evidence using the Summary of Findings table and Evidence Profile table, employed here. The 2nd component of GRADE is formal approach to making recommendations based on the evidence, which we did not employ. The GRADE Summary of Findings tables were augmented with the forest plot and meta-analysis from RevMan to create a comprehensive summary of the evidence base for each outcome. When the P-value for the hypothesis testing that the subgroup was a source of heterogeneity was >.1, we collapsed the data and only assigned a GRADE to the combined body of work, not the individual subgroups. However, when subgroups were available, we still included these in the forest plot as a means of enabling end-users to visually assess the evidence for heterogeneity by subgroup. The information in the tables should be used by end-users to determine if they would recommend vaccine use. Based on local value and preferences, resource availability, end-users may reach different conclusions about the value of the vaccine.

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

The results of the search are reported in Figure 1. For all outcomes, the GRADE Summary of Findings is presented in Table 1. The GRADE Evidence Profiles are provided in Table 2. Meta-analyses are provided Figures 2-6. Funnel plots are presented in Figure 7.

Table 1. Summary of findings for 2–3 subcutaneous doses of whole-cell killed Tritrichomonas foetus (T. foetus) vaccine
OutcomeIllustrative Comparative Risksh (95% CI)Relative Effect (95% CI)No of Participants (Studies)Quality of the Evidence (GRADE)
Assumed riskCorresponding risk
No vaccineWhole-cell killed T. foetus Vaccine
  1. CI, confidence interval; RR, risk ratio.

  2. a

    Some studies were very small.

  3. b

    A moderate effect in small studies.

  4. c

    Single study with n = 24.[31]

  5. d

    Studies of shorter duration showed less difference between groups, which would be expected given the study design.

  6. e

    The funnel plot suggests larger studies showed effects closer to the null. Furthermore, studies that reported more protective effects tended to be older. This provides evidence of regression toward the mean.

  7. f

    Studies that reported short-term pregnancy percentages would most likely also have access to calving percentages and abortion risks. However, fewer studies reported calving percentages and abortion risks and all that did report, reported an outcome favorable to the vaccine. Selective reporting of vaccine favorable outcomes is therefore a possibility. (Note: several studies that reported pregnancy percentages euthanized animals before calving.) Other studies slaughtered animals before 180 days.

  8. g

    Downgrade based on concerns about the validity of the data form; ie, SEM and standard deviations derived from time to event data.

  9. h

    The basis for the assumed risk is the median of control group risk across studies. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

Risk of T. foetus infection in heifersStudy populationRR 0.89 (0.76–1.05)251 (10 studies[32-39])

⊕⊕⊝⊝

Low because of inconsistency and publication biasd,e,f

Range of follow-up after challenge: 1-18 weeks 82 infections per 100 susceptible heifers73 infections per 100 susceptible heifers (63–87)
Moderate risk population
50 per 10044 per 100 (38–52)

Open risk

Range of follow-up after challenge: 35 days to calving

Study populationRR 0.80 (0.63–1.01)570 (11 studies[32-37, 39])

⊕⊕⊕⊝

Moderate because of potential for selective reportingf

39 open heifers per 100 bred heifers31 open heifers per 100 bred heifers (25–39)
Moderate risk population
50 per 10040 per 100 (31–50)
Duration of infection in heifers in daysThe mean duration of infection in days when exposed to T. foetus in the control heifers was 83 daysThe mean duration of infection in days when exposed to T. foetus in the vaccinated heifers was 60 days−23.10 (−38.36 to −7.85)163 (6 studies[34-37, 39])

⊕⊕⊕⊝

Moderate because of potential for risk for biasg

Abortions riskStudy populationRR 0.57 (0.42–0.78)176 (5 studies[32, 34, 35, 37, 39])

⊕⊕⊝⊝

Low because of imprecision and selective reportinga, f

68 abortions per 100 susceptible pregnant heifers38 abortions per 100 susceptible pregnant heifers (28–53)
Low risk population
10 per 100 6 per 100 (4–8)
Moderate risk population
50 per 10028 per 100 (21–39)

Risk of T. foetus infection in bulls

Max length of follow-up after challenge: 8 weeks

Study populationRR 0.41 (0.17–0.99)68 (3 studies[15, 31])

⊕⊕⊝⊝

Low because of imprecision and publication bias a,b

80 infections per 100 T. foetus susceptible bulls33 infections per 100 T. foetus susceptible bulls (14–79)
Low risk population
10 per 100 4 per 100 (2–10)
Moderate risk population
50 per 10020 per 100 (9–50)

Clearance of T. foetus infection in bulls

Max length of follow-up after vaccination: 14 weeks

Study populationRR 0.36 (0.16–0.77)24 (1 study[31])

⊕⊝⊝⊝

Very low because of serious potential for imprecision, and publication biasc

88 infections remain at 14 weeks per 100 infected bulls32 infections remain at 14 weeks per 100 infected bulls (14–67)
Table 2. Evidence profile summaries from the GRADE profiler for whole-cell killed Tritrichomonas foetus vaccine
No. of StudiesDesignRisk of BiasgInconsistencyIndirectnessImprecisionOther ConsiderationsQuality
  1. a

    Some studies were very small.

  2. b

    Single study with n = 24.[31]

  3. c

    Studies of shorter duration showed less difference between groups which would be expected given the study design.

  4. d

    The funnel plot suggests larger studies showed effects closer to the null. Furthermore, studies that reported more protective effects tended to be older. This provides evidence of regression toward the mean.

  5. e

    Studies that reported short-term pregnancy percentages would most likely also have access to calving percentages and abortion risks. However, fewer studies reported calving percentages and abortion risks and all that did report, reported an outcome favorable to the vaccine. Selective reporting of vaccine favorable outcomes is therefore a possibility. (Note: several studies that reported pregnancy percentages euthanized animals before calving.) Other studies slaughtered animals before 180 days.

  6. f

    Downgrade based on concerns about the validity of the data form; ie, SEM and standard deviations derived from time to event data.

  7. g

    As the outcomes reported were not subjective ie, infection detected by culture or pregnancy, the impact of failure to blind was considered minimal. The studies reported minimal loss to follow-up. Failure to randomize was considered important; however, its impact was most likely to be seen as publication bias. The impact of multiplicity was thought also to be manifest as selective reporting or publication bias.

Outcome: Risk of infection in susceptible heifers (follow-up 18 weeks), a critical outcome
106 randomized and 4 nonrandomized challenge studiesNo serious risk of biasSeriouscNo serious indirectnessNo serious imprecisionPublication biasd

⊕⊕⊝⊝

Low

Outcome: Open risk in susceptible heifers: an important outcome
116 randomized and 5 nonrandomized challenge studiesNo serious risk of biasNo serious inconsistencyNo serious indirectnessNo serious imprecisionSelective reporting biase

⊕⊕⊕⊝

Moderate

Outcome: Duration of infection in days in susceptible heifers: an important outcome
52 randomized and 3 nonrandomized challenge studiesSerious riskNo serious inconsistencyNo serious indirectnessNo serious imprecisionNo concerns

⊕⊕⊕⊝

Moderatef

Outcome: Abortion risk — pregnant heifers suffering reproductive losses: an important outcome
53 randomized and 2 nonrandomized challenge studiesNo serious risk of biasNo serious inconsistencyNo serious indirectnessSeriousaSelective reporting biase

⊕⊕⊝⊝

Low

Outcome: Risk of infection in susceptible bulls (follow-up 34 months), a critical outcome
32 nonrandomized and 1 randomized challenge studiesNo serious risk of biasNo serious inconsistencyNo serious indirectnessSeriousaPublication biasa

⊕⊕⊝⊝

Low

Outcome: Clearance of infection from infected bulls (follow-up 36 months); an important outcome
11 nonrandomized challenge studiesNo serious risk of biasNo serious inconsistencyNo serious indirectnessVery seriousbPublication biasb

⊕⊝⊝⊝

Very low

image

Figure 1. Flow chart of the search and identification of papers relevant to the review.

Download figure to PowerPoint

image

Figure 2. Meta-analysis and forest plot for effect of 2–3 subcutaneous doses of whole-cell killed Tritrichomonas foetus vaccine for risk of infection in T. foetus susceptible heifers.

Download figure to PowerPoint

image

Figure 3. Meta-analysis and forest plot for effect of 2–3 subcutaneous doses of whole-cell killed Tritrichomonas foetus vaccine on open risk in T. foetus susceptible heifers.

Download figure to PowerPoint

image

Figure 4. Meta-analysis and forest plot for effect of 2–3 subcutaneous doses of whole-cell killed Tritrichomonas foetus vaccine on duration of infection in heifers.

Download figure to PowerPoint

image

Figure 5. Meta-analysis and forest plot for effect of 2–3 subcutaneous doses of whole-cell killed Tritrichomonas foetus vaccine on abortion risk in pregnant heifers.

Download figure to PowerPoint

image

Figure 6. Meta-analysis and forest plot for effect of 2–3 subcutaneous doses of whole-cell killed Tritrichomonas foetus vaccine for prevention of infection with T. foetus in bulls.

Download figure to PowerPoint

image

Figure 7. Funnel plots for risk of infection in Tritrichomonas foetus susceptible heifers (A); open risk in T. foetus susceptible heifers (B); abortion risk in pregnant heifers (C); prevention of infection with T. foetus in bulls (D); and cumulative meta-analysis for risk of T. foetus infection in heifers (plot E).

Download figure to PowerPoint

Eight manuscripts reported risk (cumulative incidence) of T. foetus infection in heifers as an outcome,[32-39] 2 of which reported 2 separate studies for a total of 10 studies.[36, 38] The quality of evidence was considered low because of the presence of inconsistency based on the overall heterogeneity and possible publication bias (Tables 1 and 2). The funnel plot (Fig 7A) suggested that larger studies were more likely to report effects closer to the null value (no effect). The cumulative forest plot suggested that more recently conducted studies also were likely to demonstrate decreased efficacy compared to older smaller studies (Fig 7E). We assessed length of follow-up as a source of subgroup heterogeneity. The test for subgroup heterogeneity was not significant (chi-square P value for subgroup heterogeneity = .28); therefore, the data were combined into a single summary effect measure. The Mantel-Haenszel summary risk ratio was 0.89 (95% CI; 0.76–1.05), suggesting a 10% decrease in the risk of infection in vaccinated exposed heifers compared with nonvaccinates (P value for overall effect = .16). The uncertainty about this effect extends from protective to not protective (95% CI; 0.76–1.05).

Seven manuscripts evaluated the open risk as an outcome,[32-37, 39] 4 of which involved 2 separate studies[33, 35-37] (ie, 11 studies with relevant outcomes). The quality of the evidence in the 7 studies was considered moderate. The factor that downgraded the evidence base was the potential for selective reporting bias (Tables 1 and 2).The test for subgroup heterogeneity was not significant (chi-square P value for subgroup heterogeneity = .93). Therefore, the data were combined into a single meta-analysis. The Mantel-Haenszel summary risk ratio was 0.80 (95% CI; 0.63–1.01), suggesting a 20% decrease in the risk of infection in vaccinated exposed heifers compared with nonvaccinates (P value for overall effect = .06). The uncertainty about this effect extends from protective to barely protective (95% CI; 0.63–1.01).

Five manuscripts reported 4 studies evaluating the effect of whole-cell T. foetus vaccine on the duration of infection in heifers,[34-37, 39] 1 of which involved 2 different studies[36] (ie, 6 studies with relevant outcomes). The quality of evidence was considered moderate (Tables 1 and 2). The reason this body of evidence was not given a high rating was associated with our concerns about the meta-analysis approach rather than the analyzed studies. For this outcome, the data were poorly reported, and it was often unclear if the measures of variation reported were SEM or standard deviations. Furthermore, it was unclear that such data were normally distributed. Therefore, the validity of these as measures of variation was unclear. Some studies did not report summary data, but reported individual data. Therefore, we calculated average days infected from these data. When we calculated the data, we only included animals that were infected in calculations, but it was not clear if others had done so. Finally, although the average mean differences in days infected were normally distributed, a survival analysis could have been used to better capture the distribution of the outcome. Consequently, the body of evidence was downgraded from high to moderate. The forest plot is presented in Figure 4. The data suggested the mean difference of duration of infection was decreased by almost 23 days (95% CI; −38.36 to −7.85) in vaccinates compared with nonvaccinates (P value for overall effect <.00001). There was no evidence of heterogeneity among studies (P value for overall heterogeneity = .11).

Five manuscripts reported 5 studies evaluating the effect of whole-cell T. foetus vaccine on the incidence of abortion in pregnant heifers.[32, 34, 35, 37, 39] The quality of the evidence was rated as low because of imprecision and possible selective reporting bias (Tables 1 and 2). The forest plot is presented in Figure 5 and the Mantel-Haenszel summary risk ratio was 0.57 (95% CI; 0.42–0.78), suggesting a 40% reduction in the incidence of abortion among vaccinated pregnant heifers compared with nonvaccinated pregnant heifers (P value for overall effect = .0003). There was no evidence of heterogeneity among studies (chi-square P value for overall heterogeneity = .58).

Two manuscripts evaluating risk of infections in bulls were identified.[15, 31] The quality grade for this evidence was low because of imprecision (3 small studies with total n = 68) and possible publication bias because only 3 studies were available, and the smaller studies had larger effects (Fig 7D). The magnitude of the Mantel-Haenszel summary risk ratio was 0.41 (95% CI; 0.17–0.99), suggesting an approximately 64% decrease in risk of infection among vaccinates (P value for overall effect = .05). There was little evidence of statistical heterogeneity across these studies (chi-square P value for overall heterogeneity = .20, I2 = 39%).

Only 1 manuscript assessed therapeutic use of whole-cell killed T. foetus vaccine in infected bulls.[31] The quality grade for the evidence was very low because of very serious imprecision (a single study with 12 animals per group) and a strong possibility of publication bias (Tables 1 and 2) (ie, it seems unusual that this finding has not been replicated since its original publication in 1983). The reported magnitude of the Mantel-Haenszel risk ratio for this single study was 0.36 (95% CI; 0.16–0.77), implying an approximately 60% decrease in infection (ie, fewer remaining infections in vaccinates; chi-square P value for overall effect = .009).

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

This review summarized the body of literature describing the efficacy of killed, whole-cell Tritrichomonas foetus vaccine and its effects on incidence of infection, duration of infection, open risk, and abortion risk. With respect to the quality of the body of work for the outcomes related to females, none were considered high-quality bodies of evidence. Consistent issues identified included small studies, nonrandomized studies, and possible selective reporting and publication bias. Consequently, the conclusion is that there is limited or no evidence that vaccines decrease infections or open risk in heifers as the magnitude of the effect observed was small and the body of work of moderate quality. There was some evidence that vaccination decreases abortion as the magnitude of the effect was large; however, the body of work was of low quality. Our reservations about this finding stem mainly from the potential for selective reporting. Most of the studies assessed short-term pregnancy percentage, but only a few reported final cumulative incidence of pregnancy. The concern is that the studies that observed, but did not report, long-term pregnancy percentages or abortion risks did so because nothing of interest was observed (ie, the vaccine had no effect on the abortion risk). If available, such results would have made the vaccine look less effective.

The quality of the work describing the outcomes in bulls was even poorer than those for female-associated outcomes. For infection risk in bulls, the number of studies identified was only 3, 2 of which were not randomized. The only randomized study had 4 vaccinated animals and 8 unvaccinated animals. It is very concerning that veterinarians should be expected to conclude that vaccines are associated with an approximate 60% reduction in infection risk based only on 1 randomized study with 4 vaccinates. Information about clearance of T. foetus infection in bulls was only available from 1 study, and therefore although the study was reasonably large by veterinary standards, this is still only 1 observation, and it is not possible to judge if the findings from that study are representative of the true vaccine effect. Again, it is difficult to believe that veterinarians should be expected to rely upon data from a single nonrandomized study for reliable decision making.

With respect to the execution of the review, we did not call this a systematic review preferring instead to use the term critical review because several features of a systematic review are missing, in particular external review of a protocol and a panel with diverse expertise. The review, however, does have many aspects of a systematic review to enhance transparency. We conducted a comprehensive electronic search, but did not manually search journals for articles that may have been overlooked. We did, however, verify that no relevant citations from the prior reviews were missed by the electronic search. Therefore, the potential for bias was minimized. We did not hand search conference proceedings of the World Buiatrics Annual Meeting, and relevant studies may have been missed in that publication. However, because no relevant studies from that conference were identified by prior reviews, it seems unlikely that any were overlooked. We relied upon indexing in CAB abstracts to capture articles published in Bovine Practitioner and the conference proceedings of the Annual Meeting of the American Association of Bovine Practitioners.

Studies that were nonrandomized were included in summary of the evidence. This was done because there were few studies that randomized (Table 2) and also met our inclusion criteria. Also consistent with GRADE, our aim was to provide a comprehensive summary of the literature for others to evaluate. The likely direction of bias by failing to exclude these nonrandomized studies is away from the null, as the studies that identified a protective effect of the vaccine because of confounding would have been more likely to be published because of publication bias. An alternative strategy, exclusion of nonrandomized articles, would have limited the information available to make meaningful conclusions. Failure to blind also was not used as an exclusion criterion. The rationale was that the observed outcomes were objective, and therefore most likely unaffected by blinding. Veterinarians should consider their interpretation of the impact of randomization on vaccine efficacy when considering whether to recommend vaccination to producers.

The use of the GRADEpro system and the GRADE approach to grading the evidence base is novel. GRADE has 2 components, one that refers to the grading and presentation of evidence, employed here. The 2nd component of GRADE is a formal approach to making recommendations based on the evidence, and was not employed in this project. Our ability to extend the project to include recommendation-making was limited by the fact that the project was a summer project for a student, whereas recommendation-making needs engagement of stakeholders. The value of the GRADE tables is that they enable others to develop recommendations based on this evidence. GRADE refers to this feature as globalization of evidence, but localization of recommendations.

It is not possible here to review all of the publications about GRADE tables and recommendation-making process, and the reader is referred to other publications.[21, 22, 26-30, 40-44] With GRADE, assessments of the quality level of each evidence base were partitioned into 4 separate assessments about the presence of inconsistency, indirectness, imprecision, and risk of bias. In doing so, these judgments become transparent to the reader. Similar judgments are made in all reviews, but rarely communicated to the end-user. The recommendation-making process not included in this review includes considerations of evidence of efficacy and values and preferences and more judgments.

The GRADE working group encourages the use of the Summary of Findings and Evidence Profile tables (Tables 1 and 2). End-users need short and concise information, but reviewers also need to convey the nuances of interpretation that have gone into the quality judgment; these tables aim to balance these goals. Our decision to use GRADE was largely based on the use of absolute risk measures to convey the impact of the interventions.[22, 44] Absolute risk measures are considered easier to interpret than relative measures of effect such as the risk or odds ratio.[45, 46] By presenting populations of different risk, producers and veterinarians should be able to better evaluate the decision to use the vaccine. For example, in a population with 10% risk of infection, it is expected that 4 cases would be prevented by vaccination, but this number could be as low as 2 or as high as 10, given the uncertainty of the estimate (Table 1). Producers may decide not to use the intervention because the cost of 100 doses of vaccine may exceed the savings from 6 cases, given the uncertainty of the estimate. Alternatively, in a high-risk population, the reduction from 80 to 34 cases (95% CI; 14–79 cases) attributable to vaccination may be considered highly economical.

Another concern about this body of work is that all studies used the individual as the unit of treatment allocation, rather than the herd. This raises the possibility of interference (ie, herd immunity).[47-49] Such interference usually biases the body of work toward the null. We hypothesize that this effect would be minimal in this review. Herd immunity occurs because transmission of the organism is decreased. However, for this disease, infected bulls are the major source of transmission and are not vaccinated in the heifer studies. Therefore, if the vaccine was 100% effective in preventing trichomoniasis in heifers, the incidence of infection in nonvaccinated heifers would be unaffected by proximity to large numbers of vaccinated heifers. Of course, if the major mode of transmission was from infected heifer to noninfected heifer, with bulls acting only as vectors of the organism without becoming infected themselves, this may result in a bias toward the null (ie, herd immunity). Our understanding is that the organism lives in the prepuce of the bull, and it is the bull that is the main source of organism.

For all outcomes, the evidence bases were designated as moderate to very low quality. Based on this review, our conclusion is that there is a lack of conclusive evidence to support the use of this vaccine in areas where good biosecurity practices are in place. Others may use the GRADE tables to reach similar or different conclusions, depending upon local circumstances. For example, some veterinarians still may recommend the vaccine in settings where biosecurity is difficult may elect to use vaccine, as the circumstances differ. Such a judgment could only be reached if the veterinarian placed lower value on the small study sizes, quality of the evidence and sources of bias than we do, and placed higher value of the estimates of effect. Although there has been research on the effect of the vaccine in heifers, the effect in cows has not yet been assessed. Also, only 3 unconvincing studies have been published assessing the efficacy of this T. foetus vaccine in bulls. To improve the current body of research, herd level studies of multiple year duration would need to be conducted in production systems at equal risk of infection because of certain management practices. This would greatly add to the understanding of the efficacy of whole-cell killed T. foetus vaccine in a susceptible cattle population.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

The source of funding was investigator resources (Baltzell) and the ISU ILHAC Summer Scholars program (Newton).

Conflict of Interest: Authors disclose no conflict of interest.

Footnotes
  1. 1

    TrichGuard and TrichGuard V5L; Boehringer Ingelheim Vetmedica, Inc., St. Joseph, MO

  2. 2

    EndNote Web. [Computer program]. Version 3.3 edn; Thomson Reuters, New York, 2012.

  3. 3

    Review Manager (RevMan) [Computer program]. Version 5.1; Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, Denmark, 2011.

  4. 4

    Schwarzer, G., meta: Meta-Analysis with R, 2010, R package version 1.6-1. http://CRAN.R-project.org/package=meta},

  5. 5

    GRADEpro. [Computer program]. Version 3.2 for Windows; Jan Brozek, Andrew Oxman, Holger Schünemann, 2008.

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References