From the Editor |
This column continues
my discussion of how to critically read the scientific literature.
In this month's column, I will discuss the pertinent issues to
consider when evaluating field trials.
A field trial is a study in which exposure to disease organisms is left to nature but the treatment of the pig is randomly assigned to one group of pigs and the other pigs receive either the standard treatment or a placebo. This differs from laboratory experiments in which both the exposure of the pig to the disease-causing organisms and the treatment are controlled.
Study protocol
The purpose of the field trial must be clearly defined. The materials and methods must describe the study in sufficient detail for you to be able to completely understand and repeat the trial. This should include mention of the method the investigators used to assign pigs to a treatment. The pigs used in a trial should be "typical" of those used in industry. The success of a treatment in a disease-ridden barn may differ substantially from that in a minimal-disease barn.
Uniformity
The treatment and control pigs should differ only by their treatment. Pigs need to be randomly assigned to the treatment and control groups so that the groups do not differ by known factors, such as size and age, or by unknown factors. Left to the producer's discretion, the sickest pigs would receive treatment. However, all pigs need to have an equal opportunity to be placed in the treatment or the control groups. Even when pigs are randomly assigned to pen and to treatment, there are factors that could affect the outcome measurement. These factors should be controlled for in the statistical analysis. For example, if the outcome is average daily gain, the gender and the starting weight of the pigs should be included in the analysis.
Blinding
The producer, researcher, and statistician should not know which pigs are in the treatment group and which are in the control group. This is especially important if the outcome is a subjective measurement, such as the consistency of feces. An observer may call the same feces "normal" from a treatment pig, but "loose" from a control pig. Control pigs should receive a placebo, such as an injection of sterile water.
Follow-up
Treatment and control pigs need to have an equal opportunity to stay in the study. If the people involved in the study know whether a pig is in the treatment or control category, they will tend to treat it differently. For example, a sick pig in the control group may be culled whereas the same pig in the treatment group may be kept, "knowing" that the treatment will cure it. Pigs that are rejected either before or during the study need to be carefully described. The criteria for rejection should be equal in the treatment and control groups and should make biological sense from an industry standpoint. The outcome for the rejected pigs must be compared to the outcome of those that stayed in the trial.
Compliance
Typically, problems occur during field trials. For example, a feeder might malfunction, or a control pig might accidentally be given a treatment. These problems need to be carefully documented. Both the analysis and the overall conclusions need to reflect the affect of these problems.
Unit of concern
Statistical analysis should be conducted at the smallest level at which the treatment was applied. If the study is a feed trial, and pigs were fed on a pen basis, then the smallest group that can receive the treatment is the pen. If the barn has two feed lines, then the smallest group that can receive the treatment is the side of the barn. Researchers doing work in such barns need to repeat the study on many turns of the barn.
Statistical significance and power
If the statistics indicated a significantly different (P <.05) outcome for the treatment pigs compared to the control pigs, the next question to answer is whether the statistically significant difference observed is biologically important. With sufficient animals in a study, a researcher can find very small differences significantly different. If there is not a statistical difference, then perhaps the study did not have sufficient power to find a difference. A treatment may be very effective but if only four pens of pigs were used for the trial, the study was too small to detect a difference statistically. Also, if the normal variation in the outcome measurement is high--for example, in the farrowing rate--then a large number of animals will have to be included in the study.
Outcome measurement(s)
The best studies have one primary outcome, such as average daily gain, and one or more secondary outcome measurements, such as morbidity and mortality. The outcome measurements need to be biologically significant and directly related to the treatment.
Field trials are the source of unbiased, objective information regarding the benefit of a new procedure or product. Your interpretation of the "success" of the treatment will depend on how well authors have accommodated the concerns listed above, and the extent to which their results are consistent with results previously reported in the literature.
--Cate Dewey