Why is it that health studies often seem to contradict one another? One day we're told something is bad for us; the next we're told it might actually make us live longer.
The answer often comes down to how medical studies are performed. Some types of study are simply considered more reliable, which is why the advice from our doctors can change with the emergence of new studies.
So what is the best kind of study: a prospective, longitudinal, observational study, or a randomized, double-blind, controlled trial – and what the heck do any of those words mean anyway?
Here are a few of the terms that appear every day in medical journals, and what they mean to you.
Observational study
This is a type of study in which people or patients are watched over time to see how a particular treatment or behaviour affects them – for example: how drinking wine over many years can affect the risk for a heart attack.
An observational study is considered somewhat less reliable than a controlled experiment, because it cannot prove that a behaviour or treatment causes a result since it can't exclude other factors that might have played a role. An observational study can only notice a pattern of links between behaviours and effects. That's why one often hears the phrase "correlation does not imply causation" in relation to observational studies.
But observational trials have their strengths. For example, they're indispensable when testing the effect of something over the course of decades, and/or over large groups of people. They're also a key study method to look for a side effect that crops up only rarely. In both cases, a controlled experiment would neither be practical nor feasible.
Prospective observational study
A prospective observational study, also called a "cohort study," is one type of an observational study.
This type of study follows a group of patients, called a "cohort," over time, gathering data from the time the study begins and then observing the effects over time. A cohort study is often used to measure a number of behaviours and their effects over many years.
Example:
The Framingham Study is a well-known, long-term cohort study looking at what causes heart attacks in people living in the town of Framingham, Mass. It began in 1948 and continues to this day. Much of what has been learned in the last half-century about how diet, exercise, smoking and cholesterol affect the heart came out of this study.
Retrospective observational study
A retrospective observational study, also called a "case-control study," looks at outcomes that have already occurred before the study has begun, and then looks back to gather evidence. So if a group of people is developing a similar type of cancer, researchers might match them with people who do not have the cancer, and then collect data to figure out what characteristics distinguish two groups.
Example:
The conclusion that smoking causes lung cancer first began with retrospective studies, in which doctors noted a link between a rise in lung cancer rates and rises in smoking rates.
That led to prospective studies of smokers and how they developed lung cancer over time, and then to controlled experiments using mice exposed to smoke.
Since retrospective studies tackle problems backwards, they are prone to some kinds of errors and are not always ideal. Indeed, the conclusions of many case-control studies have later been contradicted. But retrospective studies are often the first step in noticing patterns.
Longitudinal study
A longitudinal study is a type of observational study – either prospective or retrospective -- that involves repeatedly "checking in" on the same items or same subjects over time — often over many decades. This helps track trends across the life span.
While longitudinal studies can offer lots of insight, they are often quite expensive and are prone to errors as participants sometimes drop out of the study.
Example:
The Framingham Heart Study is an example of a longitudinal study, with researchers following the same volunteers – then their offspring and even their grandchildren – over decades. In fact, more than 1,200 articles in leading medical journals have come out of the Framingham Study.
Randomized controlled trial
Randomized controlled trials are considered the most reliable kind of study. They are commonly used to test if a medicine or a medical technique works. They involve assigning one group to try the new drug or treatment, and the other group to receive a "control": either a placebo drug (dummy pill) or placebo procedure; the usual drug or treatment; or no treatment at all.
In a randomized trial, participants are screened so that they are as similar as they can be, and then they are randomly chosen to be part of one arm of the study or another.
Example:
A recent randomized controlled trial looked at how drinking water can affect weight. A group of adults were divided into two groups: one group drank two cups of water prior to every meal, while the other did not. Over 12 weeks, the water drinkers lost more weight than the non-water drinkers.
While randomized controlled trials are considered the "gold standard" of medical experiments, they can be expensive and complicated to conduct. As well, not every study is suitable to be an RCT. If one wanted to know if babies exposed to alcohol in the womb later developed developmental problems, it wouldn't be ethical to conduct a controlled experiment to test the theory.
Blinded study
A controlled experiment is "blinded" when the study volunteers are not told they'll be getting the real drug or treatment, or a fake one -- thereby, hopefully, eliminating a possible "placebo effect."
A trial is "double-blinded" if both the participants and the study staff don't know who's receiving the experimental drug or treatment. This not only reduces the likelihood for the placebo effect, it also eliminates the possibility that the researchers will favour one group over another in some way.
While blinded trials are considered ideal, blinding is not always feasible. For example, a study meant to compare a clot-busting drug to clot-busting surgery cannot be reasonably blinded.
The opposite of a blinded study is an "open label study."
Confounding factors
A confounding factor in a study is something that can "creep in" to affect results. A confounder may falsely suggest a link between a behaviour and an effect where no real link exists, or it can mask an actual link.
Example:
A study that notices a link between coffee drinkers and high lung cancer rates might conclude that the coffee is causing the lung cancer. But further investigation might also reveal that people who drink a lot of coffee also smoke a lot of cigarettes. So in this case, smoking would be a confounding factor.
Ideally, researchers do their best to filter out confounding variables. For example, a well-planned randomized trial would screen participants to ensure that they are as similar as possible.
Even a prospective observational study can control for confounding factors, by matching participants as much as possible. So if smoking is thought to be a factor in a coffee=lung cancer experiment, then the researchers would re-analyze the results, comparing high and low coffee intake among smokers, to high and low coffee intake in non-smokers.
Yet controlling for confounding factors is not always possible, since it requires knowing what to control for.
