| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Institute for Animal Health, Compton, Newbury, RG20 7NN, United Kingdom
Ecolab Inc., Redmond, WA
1 Corresponding author: eric.hillerton{at}dexcel.co.nz
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Key Words: National Mastitis Council protocol • teat dip • noninferiority test
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
This clinical trial was conducted to demonstrate that UG Pt is efficacious in preventing bovine mastitis under natural exposure conditions in a positive-control test. Because UG Pt is the only acidified sodium chlorite postmilking teat dip product currently registered in the United Kingdom, no directly comparable products were available to act as positive controls. Consequently, an iodine-containing product, licensed as a teat dip in the United Kingdom and used widely within the European Union, was selected for use as a positive control.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Animal Supply and Management
All cows in milk at the Institute for Animal Health’s Cheseridge Farm (Newbury, UK) were included in the trial if eligible. Eligibility required that cows be free of clinical mastitis for 21 d before the trial, not in receipt of antimicrobial treatment during the 21 d before the trial, and free of teat lesions.
Cows were managed in groups according to their production levels (Table 1
). Cows of both treatment groups were mixed, bedded in sand freestalls or straw yards, and managed in an identical fashion. The lowest yielding management groups were moved to pasture in mid-April. Each production group contained approximately equal numbers of cows receiving control and test treatment products (Table 1). Cows were not sorted at milking time, but were identified by a colored leg band to indicate treatment group.
|
Allocation to Group
All lactating cows were examined for IMI using duplicate quarter samples taken 5 to 7 d before the trial. All newly calved cows were examined for IMI using duplicate quarter samples taken within 24 h of calving and then again within 7 d of calving. Infection was identified on the isolation of the same bacterial species from both samples, or from a third sample if the results differed between the first 2 (International Dairy Federation, 1981). Cows were categorized as either uninfected or infected in 1 or more quarters with Staphylococcus aureus, Streptococcus dysgalactiae, Streptococcus uberis, gram-negative bacteria, CNS, or Corynebacterium bovis. No other infections were confirmed at this time.
All uninfected cows were assigned to treatment group balancing lactation number (parity) and then lactation stage. The cows were managed in production (lactation stage) groups so that each group contained approximately the same number of cows on each treatment at approximately the same risk of infection (Table 1
).
Infected cows were classified according to category of mastitis pathogen. Within each pathogen group the cows were sorted by lactation stage then by parity. These cows were assigned sequentially and randomly to either the test or control product group. Cows unpaired each week were pooled as closely as possible under the same lactation stage and sequentially and randomly assigned. All newly calved cows were added to treatment or control groups based on infection status and then parity. They were included in the first-lactation group or the high-yielding group. Until allocation to treatment group, all newly calved cows were teat dipped with the control product only.
Eligibility of quarters for new infection meant that any one quarter could become infected only once by any 1 of Staph. aureus, Strep. dysgalactiae, Strep. uberis, gram-negative bacteria, CNS, or C. bovis. Duplicate quarter milk samples were taken from each cow 5 to 7 d before the start of the trial, every 28 d during the trial, and at the conclusion, and were examined to determine IMI. On confirmation of a new IMI, the quarter was excluded from eligibility for further IMI. When clinical mastitis occurred, the affected quarter was sampled for bacteriology, treated with antibiotic, and the cow excluded. Any teats showing accidental damage were excluded until integrity of the teat skin was restored.
All cows were milked in a double-8 milking parlor. The higher yielding management groups were milked thrice daily and the lower yielding groups milked twice daily. Before each milking all cows in each group had all teats dipped in water and dried with a single-service paper towel; each teat was then fore-stripped. The milk and the cow were observed for abnormal signs before the cluster was attached (NMC, 1999).
Products and Administration
The test product components (activator and base) and the control product (Iosan) were stored at ambient temperature before use. The UG Pt product was prepared fresh before each milking by combining the activator and base in equal volumes and mixing thoroughly to a uniform color. The mixed test disinfectant was placed in clean, dry, nonreturn teat-dip cups for application to the teats. The control disinfectant was ready to use and was the product normally used in the herd. Teats were dipped to at least 50% of the teat length as soon as possible following removal of the cluster.
The quantity of product applied to each teat was estimated by preparation of teat cups containing test or comparator product, weighed when full and then again after dipping all teats of subgroups of 4 cows of a total of 20 cows treated with each product.
Teat Condition Scoring
A minimum of 203 cows, selected from the higher yielding management groups (including heifers), was assessed for skin integrity (Mein et al., 2001) 3 d before the trial started and then every 28 d during the trial. These animals included the cows that had most recently calved; cows in this group were milked thrice daily and were most likely to show a response in teat skin to treatment. Inspection determined skin coloration, skin roughness, and orifice condition for hyperkeratosis. The particular traits of relevance that were scored were skin surface, as normal, dry, or rough; skin color, as normal, reddened, or otherwise discolored; and hyperkeratosis on a scale of none (0) to extremely rough (5).
Laboratory Procedures
All milk samples were examined for pathogenic bacteria (International Dairy Federation, 1981). Subclinical infections were defined based on the same bacteria identified from each duplicate sample or in 1 sample at each of 2 consecutive samplings. To be eligible the quarter had to be free of the identified pathogen in the pretrial sampling.
Statistical Methods and Assessment
The statistical methods recommended in the NMC protocol (Nickerson et al., 2004) were followed to compare the relative proportions of new IMI or clinical mastitis occurring in quarters treated with the test or comparator product.
The Z-value for the comparison was calculated together with its relative probability. This allowed comparisons between the 2 products to determine if one was superior to the other. A 95% one-sided lower limit confidence interval (LLCI) for the difference between proportions was calculated using the normal approximation and the critical one-tailed Z value of 1.645. Statistics were determined as follows:
Letting:
 |
then calculating:
 |
The proportion of infections in quarters treated with UG Pt was then compared with the proportion of infections in quarters treated with Iosan by estimating the magnitude of the difference between these proportions, based on the observed values. The following formula was used:
 |
Because this estimate was calculated using observed values, there is a possibility that it was arrived at by chance. To determine confidence in this result, the LLCI (at the 95% level) was used to calculate the potential difference between proportions in a "worst-case scenario" (i.e., at the lower level of the confidence interval). The following formula was used: Largest possible difference (at 95% LLCI) = LLCI/p1 x 100.
Teat condition was compared by ANOVA to determine any difference by treatment and time.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Allocation of Cows
A total of 348 cows met all eligibility criteria and were allocated to one of the treatment groups. The groups were balanced as closely as possible for age, stage of lactation, and infection status (Table 1). During the trial period, 24 cows were excluded because of incomplete or insufficient udder health data or because they were being treated with antibiotics for nonmastitis infections. Cows were removed from the trial at dry-off or when mastitis was detected (uncontrollable removal). Sufficient data for study inclusion were collected for 176 cows teat dipped with UG Pt and 172 cows teat dipped with Iosan.
Infections
Streptococcus uberis caused more than half of all subclinical infections (Table 2). Gram-negative bacteria were the most common cause of clinical mastitis (27%) with no pathogen isolated from 27% of cases. No significant differences were found between treatment groups in causes for either new infections or clinical mastitis.
|
). Hence, one product was not superior to the other. This outcome is not sufficient to demonstrate equivalence or noninferiority.
|
). When the LLCI was used to predict the largest possible difference between the proportion of infected quarters, the true proportion of infections in the Iosan-treated group might be lower (27.1%) than in the UG Pt-treated group. Nonetheless, if this value falls within the limits of acceptable difference, the treatments could still be considered equivalent.
|
Initial hyperkeratosis scores were not different between fore and hind teats (P > 0.07) for cows dipped with Iosan (Table 5). Scores did not change with time for both treatment groups and were never significantly different. Redness of the skin was not observed in either treatment group (Table 6). At 4 wk a small increase in number of cows with skin dryness was found in both groups, but not found later. A leathery appearance of the teat skin increased slightly in number of teats dipped with UG Pt only at 4 wk. Weather data from the local meteorological station showed that the first 4 wk of the trial were the wettest and coldest conditions of the year, providing the biggest weather challenge to teat-skin condition. Overall, the trial was conducted during a wide range of climatic conditions (a period below freezing to a period of warm weather, combined with wet and dry periods) typical of those in the south of the United Kingdom.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
The groups dipped with UG Pt or Iosan had few new IMI (Table 3). It is important that Iosan resulted in such a low rate because this suggests that the product was a highly effective positive control and hence, a true measure of equivalence for UG Pt. The eligible cows suffered clinical mastitis at a rate equivalent to 23 cases/100 cows per year. If ineligible cows were included, the herd rate over the adverse housing period would be equivalent to 33 cases/100 cows per year, about 75% of the national annual average (Kossaibati and Esslemont, 1997), suggesting that the test was conducted in a herd already maintaining good management and hygiene. The rate of clinical mastitis was determined with high accuracy, whereas the detection rate in commercial herds may have an error of underestimation approaching 50% (Dodd et al., 1969).
Statistical methods for assessing noninferiority in clinical efficacy trials are still very much in the development stage, and there is no recommended method entirely suited to teat-dip research. The NMC-recommended statistical analysis provides an appropriate tool to assess whether one treatment is superior to another (Z-test), but it is not appropriate for assessing noninferiority. This is because the guidelines do not provide recommendations for defining an acceptable margin of difference between the 2 treatments, delta (
). Lower level confidence intervals were calculated in accordance with NMC guidelines. All calculated values were small and negative (Table 4), but the various guidelines are unclear as to how such values should be handled and offer conflicting statements in relation to their interpretation.
The European Agency for the Evaluation of Medicinal Products (EMEA) has produced guidelines relating to the analysis of noninferiority trials (EMEA, 2005). They suggest that to demonstrate noninferiority, a value of should be chosen before the study onset as the maximum acceptable difference between the 2 treatments. If an appropriate margin were chosen, then a confidence interval that lies entirely between – and 0 (i.e., the test product is inferior to the reference product) or not more than (the reference product is inferior to the test product) would be adequate to demonstrate noninferiority.
The EMEA (2005) suggests that to demonstrate noninferiority, a may be constructed which summarizes the information known about the relative efficacy of the test and comparator products. Because this was the first study of its kind using the test product, such information was not available. According to Gomberg-Maitland et al. (2003), it is preferable that the value of is established by independent expert consensus. Subsequent to this study, the EMEA now advises that may be determined from historical data using the same or a similar reference product tested in trials of a similar size and over a similar period (EMEA, 2005).
Despite the lack of a predefined value, it is clear from the observed results that the incidence of clinical and subclinical mastitis was virtually identical in the UG Pt and Iosan groups. Hence, it is reasonable to assume that the treatments are at least equivalent.
In the absence of a predefined value of agreed by an independent expert consensus, an alternative approach to assessing noninferiority might be to compare the results of the current study with those of a study with a similar design, whose conclusion was that 2 treatments were similar. The work of Hogan et al. (1995) is cited as an example of noninferiority testing in the NMC-recommended protocols (Nickerson et al., 2004). That trial showed that for a number of bacterial classes, there was no difference between the control and test products in terms of the proportion of infected quarters. With this in mind, the data presented by Hogan et al. (1995) were used to calculate the largest possible difference in proportion of infected quarters, based on the LLCI at 95%, for each of the bacterial classes in which the authors had concluded that there was no difference between treatments. This calculation was performed in the same way as for values given in Table 4 in this report. The resulting values were –20.8% for Staph. aureus, –17.5% for streptococci, –52.3% for gram-negative bacteria, and –19.4% for yeasts.
In the absence of any direction as to the value that should be used for , it would seem reasonable to use a value that falls within the range of these values (i.e., between –17.5% and –52.3%) because the values were generated by a study of similar design. If the midpoint of this range were selected, this would equate to – being set at around –35%. This is consistent with the recent EMEA advice (EMEA, 2005) validating the approach used here to confirm noninferiority of the test product.
The largest possible difference in the proportion of infected quarters in the Iosan treated group relative to the UG PT group could have been as much as –27.1% for subclinical infections or –24.7% for the total number of infections (Table 4). If – is set at –35%, then UG Pt is clearly not inferior to Iosan in terms of the proportion of subclinical or total infections, because the largest possible difference in proportions falls between – and 0.
This argument highlights major difficulties in that no clear precedents exist to define for this kind of study, and that products that are believed at least equivalent require substantiation for an absurd value of . The observed difference in the proportion of teats that developed clinical mastitis between the 2 groups was very small. However, this difference could potentially be as large as –61.5% when the lower level confidence interval is taken into account (Table 4). This value lies beyond the value we have arbitrarily selected for –, despite the incidence of clinical cases (13) being the same in both groups. These tests appear more appropriate for testing equivalence when products are relatively ineffective than for testing noninferiority when the products are highly effective. For example, had each treatment resulted in 50 cases of mastitis rather than 13, then the largest potential difference between the number of infections would have been –29.5%, which would fall within our chosen limits of equivalence. A high power requires a trial with a large sample size mitigated in part by use of the noninferiority comparison. Large trials are not usually practical, being limited by herd size and the economic value of product testing compared with the value of the market for a licensed product.
It is important that consensus is reached on appropriate statistical methods for comparative trials of teat disinfectants to satisfy regulatory authorities. The Z-statistic may be usable in large-scale human trials; hence its adoption by EMEA and others, but it is not practical for experimental testing of teat disinfectants on grounds of feasibility and cost, because it relies on the proportion of infections and sample size, without taking account of time. If 2 products were truly equivalent, the proportion of infections should become more similar with time.
Teat Condition
The UG Pt product had no deleterious effects on teat condition. The degree of hyperkeratosis varied somewhat between determinations, but the cows examined were a dynamic sample from a dynamic population. Overall, the amount of hyperkeratosis decreased with each treatment with time. Significant care was necessary in the preparation of teats dipped with UG Pt in the determination of hyperkeratosis. If the teat ends were not cleaned properly, extra material was felt by touch around the orifice; this material proved to be dried UG Pt.
Redness of the skin was not observed on teats in either group. Some dryness of the teat barrel skin was found 4 wk after the start of the field trial, and was similar for both treatment groups; this quickly disappeared. The only skin thickening, leatheriness, was found in the first half of the trial and on very few cows (more so on those cows dipped in UG Pt). Some skin reaction is not unusual when the teat dip is changed to one of a different chemical formulation, and in this case, a very mild hyperplasia occurred. Previous experience suggests that this leatheriness was trivial and quite unlike the frequency or degree found with hypo-chlorite-based disinfectants. The slight deteriorations in skin condition coincided with a period when ambient temperatures were lowest. Such changes are common following a change to a new product.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
Received for publication May 25, 2006. Accepted for publication November 21, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
