Just because a website or practitioner states that there is "research" behind something, how do you determine if the research that has been conducted is in fact any good? While no means exhaustive, here's a good checklist to help you get you started.
1. Is the research question clear and informs us of its aims?
2. Is the study a systematic review or meta-analysis?
These will provide the highest level of evidence as they evaluate the results of all available studies on the same subject. Ten or more studies with the same conclusion provides much better evidence than just one. While a randomised controlled trial that has been well conducted with a large sample size is good, the more studies showing a positive effect, the more we can trust in the research.
3. How many participants are included in the study?
Studies with a high number of subjects are always better than those with less subjects, as it’s more likely to be a better representative of the general population. For instance, a trial with 500 participants is much more likely to have a statistically valid conclusion than a trial with 20 participants. Unfortunately, in many horse studies this can be challenging, and there are many papers published that have less than 10 horses included. Be wary of basing too much emphasis on the results of small sample sizes.
4. How were participants recruited and were they randomised to groups?
Participants should be randomly assigned to the experimental and control conditions (e.g., sealed envelope assignment or computer generated random assignment). Look also at how participants were recruited. The sample can be biased when researchers use volunteers, especially those targeted through social media or special interest groups. Those who volunteer to participate in studies do not necessarily represent the general population, as they will often be people who already have an interest or bias towards the study subject.
5. Is the study blinded?
Trials should involve blinding, meaning that researchers, experimenters, subjects, and assessors should not know which group subjects are in during the experiment. Say for example one of the measurements in a study was lameness of the horse. If the person assessing the lameness was aware of which horses had been in the treatment group and which horses were in the control group, there is an increased risk of bias in their assessment of that lameness (i.e. they may be more inclined to “see” an improvement in the horses they know had the treatment).
6. Is there a good description of the methods used?
It should clearly state in the methods if subjects were randomised to groups and whether the subjects and assessors were blinded. If the information is not included this can suggest that the trial’s methodology is unknown (or the researchers are purposefully leaving it out) and therefore may be questionable.
7. Was there a control group?
Without one it’s very difficult to conclude that the results were due to the intervention applied.
8. Is there risk of bias?
This can include:
9. Does the study have validity?
A study should have 3 different types of validity. These are:
10. Where was the study published?
A study published in a well-known and respected journal will always be preferable to one published in an unknown publication. A “study” that has only been published on the website of the manufacturer who developed the product being researched should always been viewed with scepticism.
An impact factor is a measurement that shows how often articles within a journal have been cited by other articles. A higher impact factor means that studies published within that journal are more likely to be seen as important within their field. While a study published in a low impact factor journal isn’t necessarily going to be a poor, chances are it won’t have had the same level of rigorous review that a study published in a high impact journal will have been through.
11. Has the research been peer reviewed?
Peer reviewed research has been evaluated by external experts with experience in the subject matter. It is considered higher quality.
12. Are there appropriate statistical methods?
Statistics are complicated! Proper and accurate analysis of data requires appropriate statistical tests. The tests used to analyse the data must be appropriate for the type of study and the research question they are trying to answer. Any tables and figures should be clearly labelled. Ideally, effect sizes should be included throughout giving a clear indication of the variables’ impact.
13. Are the findings statistically significant and/or clinically significant?
This can be confusing, but it’s important to understand the difference. Statistical significance indicates the reliability of the study results and quantifies the probability of the study’s results being due to chance. Clinical significance reflects the impact on clinical practice and refers to the magnitude of the actual treatment effect.
The “P” value is frequently used to measure statistical significance. It is the probability that the study results are due to chance rather than to a real treatment effect. A statistically significant “P” value is usually 0.05 (or 5%). What a P < 0.05 implies is that the possibility of the results in a study being due to chance is <5%, and therefore more likely to be due to treatment effect (i.e. the treatment very likely is what made a difference).
While there are established, traditionally accepted values for statistical significance testing, such as the P value, this is lacking for evaluating clinical significance. More often than not, it is the judgement of the clinician (and the patient/client/rider) which decides whether a result is clinically significant or not. This can include whether the change from having the treatment makes a real difference to the subject lives, how long the effects last, consumer acceptability, cost-effectiveness, and ease of implementation.
For example, a study may show that an electrotherapy treatment demonstrated a statistically significant improvement in the appearance of a lesion in a tendon on ultrasound 3 weeks and 6 weeks post-treatment compared to a control group (who had the same type of tendon lesion but who did not receive the treatment). The clinical relevance of this study is the “treatment effect”, which looked at the time at which the horse returned to competition. The results showed that the treatment group returned to competition only 5 days earlier than the control group, which most researchers, clinicians and riders would agree is a clinically irrelevant “improvement” in outcomes.
While the study may have shown that initially the tendon appeared visually to have been healing faster (statistical significance), there was actually no real difference between groups in the time that the horses were able to get back to competition (clinical significance). In this instance the clinician and horse owner must decide if the time and financial investment is worth the horse having the treatment or not, which will differ from person to person depending on their goals and financial position.
14. Is the conclusion appropriate?
In the discussion and analysis of data, researchers should note whether findings are statistically significant and if they consider there is any clinical significance. They should be careful not to make the outcomes seem more relevant than they really are. It’s a common mistake to emphasise results that are in accordance with the researcher’s expectations while failing to focus on the ones that are not. Will it can be tempting to jump straight to the conclusion when reading a research paper, make sure you read the results carefully first to see if you draw the same conclusions yourself. Even in a well-designed trial, further research and affirmation of outcomes in equivalent studies are needed before trial outcomes can be accepted as factual. Limitations of the study should also be mentioned.
15. Last but not least, were the ethical standards met?