Epidemiology 101
Insights from a former MPH student
Over the past 5 years, I have studied nutritional science and public health.
Want to know a secret?
I learned these 5 important insights during that time 🧵👇
Insight #1: Cause & Effect Are Complex
Randomized control trials (RCTs) are not the only way for determining cause & effect. When it comes to the topics such as chronic diseases, RCTs may be too short and not practical for answering questions related to that.
The mistake is thinking RCTs are superior to observational research when in reality similar limitations apply to RCTs depending on the study design. This is extremely apparent in free-living study designs where participants are allowed to live out their lives. These types of studies are subjected to confounding just as much as a cohort study.
But we can infer causality from epidemiology. We can use theoretical frameworks like The Bradford Hill Criteria. Hill’s criteria include 9 considerations for inferring causality in epidemiology:
Strength (effect size): A small association does not mean that there is not a causal effect, though the larger the association, the more likely that it is causal.
Consistency (reproducibility): Consistent findings observed by different persons in different places with different samples strengthen the likelihood of an effect.
Specificity: Causation is likely if there is a very specific population at a specific site and disease with no other likely explanation. The more specific an association between a factor and an effect is, the bigger the probability of a causal relationship.[1]
Temporality: The effect has to occur after the cause (and if there is an expected delay between the cause and expected effect, then the effect must occur after that delay).
Biological gradient (dose-response relationship): Greater exposure should generally lead to a greater incidence of the effect. However, in some cases, the mere presence of the factor can trigger the effect. In other cases, an inverse proportion is observed: greater exposure leads to lower incidence.[1]
Plausibility: A plausible mechanism between cause and effect is helpful (but Hill noted that knowledge of the mechanism is limited by current knowledge).
Coherence: Coherence between epidemiological and laboratory findings increases the likelihood of an effect. However, Hill noted that "... lack of such [laboratory] evidence cannot nullify the epidemiological effect on associations".
Experiment: "Occasionally it is possible to appeal to experimental evidence".
Analogy: The use of analogies or similarities between the observed association and any other associations.
As a student of philosophy, I can tell you that causality itself is difficult to understand. We are just scratching the surface of the subject.
Nevertheless, these criteria helped researchers understand the relationship between lung cancer and cigarette smoking.
Insight #2: Risk versus Odds Ratio
Odds ratios are used for cross-sectional studies. They tell us the ratio of the odds of an event occurring in an exposure group to the odds of an event occurring in an unexposed group.
This is different than risk ratios or relative risk used for cohort studies. Relative risk tells us the ratio of the probability (risk) of an event occurring in an exposed group to the probability (risk) of an event occurring in an unexposed group.
The mistake lies in confusing these two measurements.
The odds ratio refers to the number of events divided by the number of non-events. This also refers to the probability of an event divided by the probability of a non-event.
Compare this to the relative risk which is the probability of an event occurring divided by the probability of the event not occurring.
Both risk and odds are important for epi. It’s just the time course for each is different as odds would be more appropriate for an event that is currently happening as we see in cross-sectional studies. This is different from relative risk which deals with probabilities and accounts for changes across time, as you would see in a cohort study.
Insight #3: Measurements Have To Be Validated
Epidemiology often relies on survey data collected from a large section of a population.
This is important as survey instruments have to be valid. Validity merely refers to does the measurement actually measure what its suppose to. Another aspect of a good survey is reliability.
A measuring instrument should be able to give consistent results regardless of when you take it. For example, your scores on a test should not be wildly different if you take it multiple times. While you obviously want to improve your scores on a test, those scores shouldn’t jump from 83 to 100 to 0 on three different attempts.
In nutritional science, we often rely on food frequency questionnaires (FFQ). The FFQ has to be valid and reliable in order for researchers to draw any meaningful inference from them.
FFQs are typically validated by more accurate measurements of dietary intake such as 24-hour food recalls or food weigh in. An FFQ should correlate with these more accurate measures in order for it to be seen as valid.
How an instrument is validated is important for determining its validity and reliability. This is important regardless of study type.
Insight #4: Nutrition Relies On Epi
Numerous disease and nutrition outcomes take time to study. For example, the relationship between diet and cardiovascular disease, the relationship between coffee intake and diabetes, the relationship between fat intake and cancer, etc.
Just pick and choose which diet or nutrient and chronic disease, then you have the premise of a cohort study.
The mistake is in thinking RCTs are superior again for testing out these relationships. But conducting an RCT for 20 years just isn’t feasible from a financial and methodological standpoint.
This plays into what we established previously, an RCT has its limitations as well. While cohort studies aren’t perfect they give us an easier route for testing these relationships between nutrition and disease.
Epidemiological studies can follow a lot of people across a longer period of time compared to an RCT. Not to mention epidemiological studies often agree with RCTs the majority of the time.
Insight #5: People Cherry Pick Epi
Numerous critics of nutritional epidemiology or epidemiology, in general, have no qualms about using it when it states what they want.
A great example is carnivore diet supporter Shawn Baker, who criticizes nutritional epidemiology but readily uses it when it supports his points.
Shawn isn’t the only one to do this though, many people (typically low carb-associated individuals) try to dismiss nutritional epidemiology. Yet these criticisms are so blatantly obvious to researchers and are often adjusted for in the research.
One example is the healthy user bias, this bias entails those who volunteer for a study pertaining to health often have health habits the general population does not. But this is a bias not inherent to epidemiological studies but to all study types.
Critics also ignore the fact randomization, statistical adjustment, using comparator groups, and other methods can reduce the influence of this bias.
Even more troublesome, are those who cry “healthy user bias” while at the same time demonizing certain food groups or food ingredients are undermining their own message. If health behaviors are so powerful they can overcome the latest food trend to avoid.
TL;DR: Use these 5 insights to learn about epidemiology (epi).
• Cause & Effect Are Complex
• Risk versus Odds Ratio
• Measurements Have To Be Validated
• Nutrition Relies On Epi
• People Cherry Pick Epi
See you again next week.
Whenever you're ready, there are 3 ways I can help you:
1. Learn the basics of critical thinking here. (60+ happy learners)
2. Learn in-depth nutrition information here. (12,300+ health enthusiasts)
3. Support this newsletter by becoming a premium subscriber.