The ELISA and fecal culture results are summarized in Table 2
for each season of sampling: fall, winter, spring, and summer, and by sampling cohort within season. For the subset (top 20% of ELISA S/P values at first test) of cattle with results from both fecal culture and ELISA in the same month, 234 of the samples exhibited both negative fecal culture and ELISA results, 30 samples had a negative fecal and a positive ELISA, 7 had positive fecal and a negative ELISA, and 21 were positive for both fecal culture and ELISA. Agreement between the 2 tests (ignoring within-cow dependence) was moderate (Kappa = 0.47; 95% CI = 0.32 to 0.61). Overall, for 4 observed seasons, only 3 possible transition periods occurred (i.e., fall to winter, winter to spring, spring to summer). Of the 86 positive ELISA test results occurring during fall, winter, or spring, 18 had a negative test result sometime later during the study. Of the 947 negative results, 31 changed to a positive result. For all cows, sampling cohort; sampling season; sampling month; and mean, mean daily minimum, and mean daily maximum monthly temperatures in the month of, month before, and 2 mo before sampling did not affect S/P ratio. Though non-significantly associated with the risk factors as listed previously, both negative-to-positive and positive-to-negative test results were observed in ELISA as well as fecal cultures for all 3 transitions. For fecal cultures, 155 cow-period transitions did not demonstrate a change in fecal culture test result; whereas 8 cow transitions from positive to negative and 8 cow transitions from negative to positive were observed. It is well documented that the sensitivity of the culture test is low, ranging from 33 to 50% (Sockett et al., 1992; Whitlock et al., 2000); thus, these types of transitions should be expected.
In the present study (conducted from October 2002 to September 2003), no consistent across-cohort seasonal effects were detected for S/P ratios or proportion sero-positive to MAP, which had been observed previously in the historical (and less valid) cross-sectional time-series data collected in 2001. In the mixed model analysis of S/P ratio, however, a change (P in S/P from the beginning to the end of the study was detected. Although the season-cohort interaction was not significant, the increasing S/P ratios were likely due to the first cohort (1) in which a trend toward increasing S/P ratio over time was shown throughout the course of the study (Table 2
). This is consistent with reports (Collins, 1996; Holmes et al., 2004), which suggested that serological response increases over time with advancement of the disease associated with advancing age; thus, the changes noted in our study are likely due to the stage of disease rather than season, because neither of the other 2 cohorts exhibited this increase in S/P ratio during the course of the study. It is also possible, however, that this finding was consistent with a seasonal effect. The single cohort involved with the significant change was the cohort first sampled in October 2002. It is possible that low S/P values in that month were due to an extended lagged effect of the previous season. The other 2 cohorts began in November and December, respectively. However, because the results were inconsistent for the other models, it cannot be concluded that season had a significant effect on S/P ratio results.
In the evaluation of the transitional model, no significant effects were found associating change in serological status with temperature. Variation in seropositivity within repeated samples per cow was noted on several occasions; however, these were consistent with other findings (Hirst et al., 2002). In addition, we found no evidence to support a hypothesis linking seasonality to the risk of MAP fecal culture positivity or change in status.
Two possible reasons exist for our inability to reproduce the results seen historically on this farm. First, the summer depression in both S/P ratio and the proportion of cows seropositive to JD observed previously might simply have been an artifact, and not truly associated with seasonality. The other possibility is that an actual historical association with S/P ratio and seasonality existed, but shortcomings in the present study proved sufficient to thwart proper assessment of the association. Discussion of each of these possibilities follows.
Artifact.
The historical data, for which declines in both S/P ratio and risk of seropositivity were observed during months of elevated temperature (Figure 1), were evaluated as a cross-sectional study with a single blood sample collected from specified cows only once per lactation. In other words, an entirely different group of cows was sampled during each subsequent month. It is possible, therefore, that the large decreases in S/P ratio observed in the summer months of 2001 were related only to the features (e.g., age, infection status) of the specific groups of cows sampled at those times, rather than to climatic conditions at that time. It is for this reason that the present investigation was designed as a cohort study with repeated sampling to evaluate this observed phenomenon more properly.
An additional problem with the historical data may relate to the manner in which the blood samples were analyzed. The historical samples were analyzed on a monthly basis with no measures in place to eliminate or reduce potential plate-to-plate or inter-operator variation. Extremely high values for S/P ratio, some >6.0, were recorded for certain cows. This result alone might have accounted for the decrease in the following months when the samples were analyzed on different plates and possibly by different operators (no automated devices were used by the laboratory at that time). Again, it is for these reasons that methods suited to reducing inter-plate and inter-operator variability in the evaluation of blood samples were specified in the present study.
True association.
The historical cross-sectional data were obtained from a cooperator herd approximately 1 yr before the beginning of the present study. Mean temperatures that cows were subjected to during 2001 to 2002 were somewhat greater than those during 2002 to 2003, but the absolute temperature difference was not extreme. Rather, the key difference was in relative humidity, which was greater in 2001 compared with 2002. Because of this humidity difference, little or no significant cooling-off period occurred during the night, and the daily temperature humidity index (THI) was usually much greater. According to some authors (Kelley, 1982; Johnson, 1987), if the nighttime temperature cools to an acceptable thermoneutral zone (e.g., THI = 72; Johnson, 1987), effects of any heat stress incurred during the day can be alleviated. As is noted in Figure 2, THI for the area and time period of the previous study were significantly greater than those during the current study, reaching almost 80 during certain periods, and especially >72 during the summer months. It could not be determined whether the cattle in the present study experienced periods of climatic heat sufficient to cause immunological suppression. It seems unlikely that a strictly linear dose-response relation between heat stress and the MAP test status would exist. Rather, a threshold effect at a critical THI would instead be expected. Therefore, statistical analyses that are predicated on treating heat indices as continuous (as opposed to categorical) variables could be problematic. Another aspect that might have exaggerated the results from the previous study was the fact that results were obtained during a particularly hot summer, with temperatures greater than the normal average, even for that area (Figure 2).
A second issue that might have obscured any real relationship of S/P ratio with heat stress is that different groupings of cows were sampled in each of the respective studies. In the present study, one group of cows was followed over time; whereas in the previous study, different cows were sampled 1 mo/yr, with eligibility for sampling established by their pregnancy status. Overall, mean S/P ratios for cows in the previous study were greater (P with mean values of 0.158 for the historical study and 0.089 in the present study. Mean S/P ratios for the previous study were inflated by some of the extremely elevated observations in S/P ratio mentioned earlier, and so mean and median results (less susceptible to outliers) are shown in Figure 3. In addition, cows in the previous study had a mean monthly seropositive proportion of 0.099, whereas the present study had 0.067 seropositive. Prior to, and concurrent with the conduct of the historical cross-sectional study, whose data were collected in response to a management-driven desire to manage colostrum from seropositive cows differently from seronegative animals, control measures for JD were implemented. Therefore, it is possible that, although the owner indicated he was not culling based on JD test results, some added culling pressure was applied to JD seropositive cows. In addition, the herd had expanded from approximately 2000 to 3400 cows from the time of the historical study to the current study; thus, unidentified stressors related to the expansion might have resulted in increased culling of JD positive cows.
A third issue that might have affected the results is the finding that Johne’s positive cows were at a much greater risk for culling. In other words, a so-called healthy-worker survivor effect, common to occupational health studies of human workers (McMichael et al., 1976), might aptly have been applied in this longitudinal study design of dairy cows. This phenomenon in cohort studies of production livestock is well described (Dohoo et al., 2003). Because seropositive cows were being culled at a greater rate than seronegative cows, they were more likely to be lost to follow-up than seronegative or healthy cows. In the present study, the odds of a culled cow being positive for MAP infection by ELISA were approximately 5 times greater than that for cows that were retained in the herd, adjusted for lactation number. This was similar to results from another study (Wilson et al., 1993) in which a 6-fold increase in culling rate for seropositive cows was identified. In addition, cows with S/P results in the top quintile (i.e., >0.0819) were culled from the herd at a faster rate (hazard) and were at a greater overall lactation risk (32.1%) of culling when compared with those having smaller S/P ratios. For example, cows in the lowest S/P quintile ( a lactation culling risk of 18.3% (see 305-d survival curves in Figure 4, adjusted for lactation age). These culling risks were somewhat less than those identified by Goodell et al. (2000) in which cows with S/P ratios >0.10 were removed from the herd at a rate of 35.8 to 50.0% compared with cows with an S/P ratio of in our study with greater S/P ratios were less likely to be present in the herd toward the end of the study, it might have been more difficult to describe any associations with season accurately. If elevated S/P ratio results were missing from the already reduced mean S/P ratios for the study population, it might have reduced the ability to discriminate significant fluctuations in S/P ratio or proportion-infected data. It was imperative, in a seasonal study such as ours, to be able to follow the same cows over time (as best as is possible), and when a significant proportion of the cows were lost to follow up, seasonal variation was much more difficult to assess.
Because this 1-yr study began in the fall and ended in the summer, any subsequent increase in S/P ratio or proportion of cows seropositive or fecal negative or positive to JD when colder months ensued could not be observed. If future studies were conducted, a springtime start date with a 2-yr study period would be ideal. The timeframe of funding for this study precluded a study period of more than a single year. In addition, the natural herd pressures of culling on seropositive animals (Figure 4) during a 2-yr period would make such studies difficult to undertake.
Answers to all your Biology Questions
