Research Rudiments Part II: Observational Studies
Covering the basics of observational research
Last time in this series, we covered animal research, the advantages of using it, and the areas where it comes up short. Here, we’ll continue the discussion with observational studies in humans.
What Is Observational Research?
Human research can largely be broken down into three categories: analyses of existing data (i.e. reviews and meta-analyses), observational, and interventional (i.e. experimental). (I)
Of these, observational studies sit in the middle of the spectrum in terms of passivity. Unlike analyses, they do involve collecting new data, but they don’t involve implementing an intervention to one portion of the subjects and tracking the outcomes that follow, like interventional studies.
Instead, as the name implies, human observational studies entail recording characteristics of a group of subjects then observing the differences in outcomes amongst this group.
Researchers can do this prospectively, which means following that group of subjects over a given period of time and recording what they do or what happens to them.
For example, researchers conducting prospective observational studies on artificial sweetener consumption and weight management ask subjects a bunch of questions at the beginning of the study about their demographics, medical history, and lifestyle, including questions related to artificial sweetener consumption. Then, they follow the subjects over a given period of time (often many years), in some cases reaching out to them to ask the same questions periodically, and identify any patterns or lack thereof. In this instance, they would look for relationships between artificial sweetener consumption and changes in weight over time–whether consuming more or less artificial sweeteners is associated with change of weight in one direction or another.
They can also do this retrospectively, which involves looking back in time and comparing outcomes that already occurred based on subjects’ previous characteristics.
Continuing with the example above, this would mean looking at that group of subjects and identifying who got cancer. Then, the researchers would look back in time at the subjects’ characteristics at the start of that prospective period, as well as throughout it, and look for any patterns. For example, the individuals who consumed more artificial sweeteners may have been more or less likely to develop cancer than the subjects who consumed less artificial sweeteners.
The key difference between prospective and retrospective studies is whether the outcomes you analyze occur before or after the completion of the study (II). At the beginning of a prospective study, the outcomes–think weight gain, cancer diagnoses, heart attacks, etc.–have not happened yet; whereas, all of the outcomes have already occurred at the beginning of a retrospective study.
One useful analogy here is a basketball game, where a prospective design would be like writing down all of the players names, jersey numbers, heights, and grades before the game starts then observing which players score more points, get more assists, make more steals, and commit more fouls.
On the other hand, in this case, a retrospective study design would be like looking back at the stat sheet of a game that occurred a year ago, collecting the players’ information, and drawing conclusions on what types of players were more or less likely to play well.
So, what are the pro’s and con’s of observational studies?
One key aspect of observational research is that it can only show correlative relationships, not causative relationships. In other words, those researchers above could identify that people who consume artificial sweeteners as part of a weight-loss diet tend to lose more weight, but they couldn’t confirm that it’s the artificial sweetener consumption that is causing the weight-loss.
Observational studies are limited in this ability to show causation because they don’t involve a controlled intervention where all variables other than that main variable of concern are isolated and accounted for. Those researchers cannot be certain that individuals who are more likely to consume artificial sweeteners aren’t simply more likely to lose weight because they are ALSO more likely to do something else that promotes weight loss, like exercise more or eat less high-calorie foods. From another angle, they cannot be certain that individuals who are less likely to consume artificial sweeteners aren’t ALSO more likely to do something that prevents weight-loss, like exercise less or eat more high-calorie foods.
On the upside, though, since they don’t involve a bunch of control, observational studies can include a lot more subjects than interventional studies. In other words, when the only things subjects have to do to contribute to the study is to complete a survey once every few months, it’s much easier to include thousands to hundreds of thousands of subjects. Similarly, this lack of commitment from the subject’s point of view makes it easier to carry out observational studies across years and even decades.
This is not the case when taking part in the study means living in a laboratory setting for weeks, following a restrictive diet, getting blood drawn, or giving muscle biopsies, as is the case with interventional research.
Additionally, compared to an experiment in a laboratory setting, observational studies generally occur in what are called free-living conditions. In this way, findings from observational studies can be more transferable to real-life applications, in that the subjects were likely behaving similarly to how people behave in the real-world. In comparison, interventional studies tend to lose some applicability points because the way subjects behave and perform in laboratory settings may be different than how they behave and perform in real-life.
Now, on the downside, as I alluded to above, observational studies cannot perfectly control for confounding variables–additional variables that are influencing the results–partly because they lack a control group.
In theory, an interventional study involves changing a single variable between two groups while keeping all other variables equal–in the artificial sweetener example, this would mean both groups follow the same exact diet, except one group swaps artificial sweeteners for sugar. This makes it possible for researchers to identify whether the one variable they changed in the experimental group (this is called the independent variable) caused any differences in the results. In comparison, the fact that an endless amount of variables could be fluctuating between groups in an observational study makes it impossible to point to one as the definitive cause of an outcome.
For this reason, in some cases, observational studies serve as a step prior to interventional trials, providing insights into relationships that seem to exist based on observations. Then, researchers can pursue those insights with interventional studies that weed out causative factors from confounders.
Returning to our artificial sweetener example once more, this process might look something like this: observational studies suggesting that people who consume more artificial sweeteners tend to be overweight, so researchers carry out interventional trials where they test whether giving artificial sweeteners to one group leads to differences in weight as compared to a similar group that they don’t give artificial sweeteners too.
Conclusion
Well, that’s observational studies at a glance. They’re useful for their power in numbers, long-duration, and realistic settings; whereas, their main drawbacks include lack of control and ability to show causation.
Next in this series, we’ll focus on the details of the interventional studies I frequently used for comparison here.
P.S. Is it a ridiculous idea to subscribe for more useful health and fitness content for FREE?
If not, hit that subscribe button below.