StemCell Therapy

Veterinary medicine is an applied science. We take available research information and use it to make the best possible decisions in real world settings. This can be very challenging. But the entire process by which we try to expand the knowledge base in our field is fascinating in its own respect.

We rely heavily on statistics. On a basic level, I think this is very intuitive. What is a veterinarian, or a physician, or an educator, or an engineer, or a builder, or a farmer, but a keen observer in their field? Each respective profession can observe that treating a condition a certain way is more likely to achieve the desired result. Recording observations over a long period of time allows us to make accurate predictions, for example, that 90% of cows with gangrenous mastitis will die, and create benchmarks, for example, that incidence of displaced abomasum in fresh cows should be less than 5%.

Sometimes, though, there are more complicated questions for instance, we may observe the incidence of fresh cow DA, but now we want to know why cows get DAs so that we can do something about it. These more complicated questions also require more sophisticated methods of analysis.

A few years ago, we did a study in our practice. I would like to use it as an example to demonstrate the nature of science.

I was talking to my partner, Dr. Bill Croushore, one day and asking him what criteria he uses to determine whether or not a cow with a displaced abomasum was a good candidate for surgery. There are many identified risk factors for poor outcomes following surgery mastitis and lameness at time of surgery, for example, are poor prognostic indicators. That doesnt mean the surgery wont go well. It may mean that those cows are likely to be culled because of concurrent issues.

Science does not answer all the questions at once. As a matter of fact, in order to create a valid study, the investigators need to be very clear and specific about the question they are setting out to answer.

We wondered whether or not there might be a reliable way to quickly determine how a cow with a displacement would fare following surgery. Dr. Bill designed and authored a study about the use of a simple cow-side blood test as a prognostic indicator for outcome of LDA surgery.

There are some weak similarities between diabetic ketoacidosis in humans and ketosis because of negative energy balance experienced by lactating dairy cows. A human meter for a type of ketone known as beta-hydroxybutyrate is available commercially, and research has also validated the use of this meter for ketosis in cattle. This was the simple blood test we chose to use. It requires a few drops of blood, costs about $6, and gives a result in about 10 seconds. We felt that spending $6 for prognostic indication before investing several hundred dollars was worthwhile.

The first step in creating your study is to clearly define what you are trying to answer. In our case, we wanted to know whether or not we could use a BHB level to evaluate our surgical candidates. Simple question, right?

Now it starts getting more complicated. Next, we formulate what we call the null hypothesis. This is usually the opposite of what we think or are trying to prove, and it will probably never be stated in the final research paper. Our null hypothesis would have been something like, There is no relationship between the measured BHB value and the outcome, measured by 30-day survivability, following LDA surgery.

The goal of our study, then, is to allow us to reject this null hypothesis and prove that there is, in fact, a relationship between BHB and surgical outcome. Remember how I said we rely on statistics?

Statistical analysis of the data can be done in many ways, but in our case, it will give a value known as p. This is a probability. Science does not deal in absolutes. The p value is the probability that we can correctly reject the null hypothesis. By convention, we are willing to accept a 5% chance of error in rejecting the null hypothesis. This corresponds to a p value of 0.05 or less (5% expressed as a decimal).

In other words, if our study is designed correctly and if we use the correct statistical analysis, and p<0.05, we can reject the null hypothesis and state that BHB is, in fact, related to surgical outcome. This is referred to as statistical significance. If the p value is <0.01 or even 0.001, the relationship is even stronger.

A familiar example of this type of significance can be found on the milk you buy in the store. A few years ago, some co-ops and dairies decided they would require their producers to produce milk without the use of rBST. They advertised the milk as such.

The advertising was seen as harmful to the sale of milk that was produced using rBST and so following research on the topic, the non-rBST milk was required by law to put a statement on their labeling saying that research had shown no significant difference in milk from rBST and non-rBST treated cows. That does not mean that somebody eye-balled them and they seemed pretty similar. It means specifically that in all the parameters they tested, they were unable to reject the null hypothesis that the milks were no different.

Moving through our study, we had to establish a surgical protocol. We collected blood samples, recorded physical exam data on our patients, and decided what follow-up treatments they would receive. To be included in the study, the cows had to meet the criteria, and they had to be treated according to the protocol to eliminate confounding of data because of inconsistencies.

After 30 days, we called to follow up and see how the cows were doing and if they were still in the herd.

One of the challenges we face is sample size, which is denoted by the letter n. Statistical significance is affected by how many test subjects you use. For example, if you ran a study and looked at only two cows, your results would not mean much. By looking at larger groups of animals, you can minimize the variation because of an individual animal and get a truer reflection of the population. Our goal was to get about 150 animals (n=150). It took a while.

In the end, we went through all the work and found that there is, indeed, a useful relationship between BHB and surgical outcome, but the p value does not indicate what that relationship is. One of my favorite things about our study and its design was something we never expected.

The basic statistics allowed us to confirm the relationship we were looking for. We had a lot of help from several professional biostatisticians who co-authored our study. Remember all the data I said we collected? They were able to use that data to characterize the relationship.

In short, we were banking on a higher BHB value being an indication of fat storage in the liver and a poor prognostic indicator, and assuming that lower BHB values were better surgical candidates. The data showed otherwise.

Lower BHB values indicated a poorer prognosis and we speculate that this is because of more long-standing conditions, such as DAs that had not been found as quickly, and cows that were no longer able to mount the appropriate ketone response.

The statisticians were able to suggest a useful cut point for BHB (i.e., if the BHB is above a certain level, the data suggest this cow is a reasonable surgical candidate, and vice versa).

Such is the nature of science. Because the study was properly formulated and analyzed, we were able to confirm with statistical certainty what we suspected, but also tipped off to where our expectation was not correct.

In preparing the final version that was published in Journal of the American Veterinary Medical Association, Dr. Bill did a literature search and framed our study in light of the available scientific information to date.

Since its publication, our study has been cited a number of times in other research papers. By asking the right questions and performing thorough and meticulous investigation, our profession is able to build on itself, moving forward in small increments.

On the contrary, overreaching where information is scarce and leaping forward based on anecdotal information and consensus can be very problematic but thats a topic for next month.

See more here:
Veterinary medicine and the nature of science | Columns - Daily American Online

This entry was posted on March 28, 2020 at 7:46 am and is filed under Veterinary Medicine. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed. |