Tag Archive for: food frequency questionnaire

The FFQ: Still Too Vague

/in /by

I spent a long time examining validation and reliability studies on the Food Frequency Questionnaires (FFQ). It was interesting to compare the original validation studies with a new FFQ that was published in early 2024; researchers asked subjects in those studies that began decades ago to participate in this recent validation study.

The Stats

I learned more about a variety of statistics that I don’t typically encounter: coefficient of correlation, and then attenuated and deattenuated coefficient of correlations, and more. The researchers concluded that the “study showed that the FFQ used in prior studies has reasonably high reproducibility and validity in measuring food and food groups intakes among both women and men.” I disagree.

The coefficient of correlation is important (COC) because it gives an indication of the association of the variable with a standard, in this case a 7 Day Dietary Recall. The best COC is 1.0 or -1.0, which means it’s perfectly correlated or not correlated with the standard. A COC greater than 0.8 is considered a strong relationship, but a relationship of 0.6 – 0.79 is considered moderate.

The COC for most categories of food was well below 0.6. How can that in any way be valid? It may be reproduceable, but you’re reproducing the same mistake over and over again.

How Dangerous Is Meat?

High level analytics like this aren’t my area of expertise, but logic dictates that you can’t get precision even with large numbers of subjects. This is especially true when using FFQ data to correlate nutrition with disease. Remember the study on red meat intake and type 2 diabetes? The Hazard Ratio was only 10% per 100-gram serving of red meat. If the meat intake is moderately correlated, how much does any error of intake impact the HR?

Whether researchers are trying to estimate how much of each type of meat a person eats or trying to calculate the heme-iron content of that meat, the FFQ doesn’t have enough precision to be used in determining those values. Remember, the increase in HR was 10% per 100 grams—that’s 3.3 ounces—of unprocessed red meat per day. If a patty were 100 grams, a reasonable size, and you ate six patties every day, that would be 600 grams or over 1.5 pounds of hamburger patties per day. Would that raise the HR to 60% based on that single answer? What about a vegan who gets no heme iron? Would they never get type 2 diabetes? We know that’s not true either.

One more thing: People under-report what they eat. It can be 100 to 200 calories per day, or even up to 500 calories per day. No after-the-fact adjustment of the food intake can make up for that kind of imprecision.

The Bottom Line

What we’re left with is this: There may be a relationship between red meat, and subsequently, heme iron intake, and the risk of type 2 diabetes, but we don’t know how much. That’s about it. We’re going to need much better studies to nail that down before we make a pronouncement. For now, you’re probably safe eating red meat, especially if you keep this in mind: eat better, eat less, and move more.

What are you prepared to do today?

        Dr. Chet

References:
1. Am J Epidemiol. 1985;122(1):51–65.
2. Am J Epidemiol. 2024;193(1):170–179

Why Errors in Food Intake Matter

/in /by

What is the big deal about errors in food intake in studies most people never hear about? It’s a problem because decisions on gaps in the diet, impact of nutrient intake, and the potential benefit or hazards of food and supplement intake are based on studies that use these techniques. I’ll give you a couple of examples, but let me start with something that has become common knowledge.

Obese subjects underreport food intake at a greater rate than subjects who are normal weight. Female obese subjects are more likely to underreport food intake than male obese subjects. Don’t assume they intend to deceive; I think many people are simply unaware of how much they eat, especially when they graze or sample food as they eat, pick at the kids’ leftovers, or eat little snacks at work.

Diet Change and Heart Disease

The Women’s Health Initiative is one of the largest studies done on examining the role of diet and heart disease in women. Results published in 2006 demonstrated that after a number of years on a low-fat diet, there were no differences in the rates of different forms of heart disease. What struck me at the time was that the goal was to reduce fat intake to 20% of total caloric intake, but using a form of dietary recall, the experimental subjects were able to lower their percentage fat intake from 35% down to 28%. That’s still much higher than the goal of 20%. If we were to estimate an average error in food intake based on dietary recall, it could very well be that these subjects actually had well over 30% fat intake due to under-reporting.

Why is that a big deal? Two reasons. First, they were not on a diet that was designed to reduce fat intake enough to impact cardiovascular disease. Second, in a review just published in 2021, a scientist is calling for the repudiation of the results of that study claiming that a low-fat diet does not work to reduce heart disease (or type 2 diabetes either). Based on the results of that WHI trial, we don’t know that for sure because of the potential for under-reporting food intake including fat, as well as the inability of the subjects to meet the goal of 20% fat in the diet. One error begets more errors. In this case, it’s being used to suggest that low-fat diets are not the way to reduce heart disease, and I’m just not ready to make that leap.

Nutrient Studies

There are a number of studies that have used FFQ as the method of assessing food intake in individual nutrient trials. Aside from the “How many portions of beef did you eat per week over the past year” type of questions, the total number of questions ranges from 138 to 164 on most FFQs. The degree with which people will report that accurately is suspect to begin with. Add to that the potential under-reporting of food intake when you’re trying to assess iron, calcium, folic acid, and other nutrients in the diet can provide significant errors in determining how much nutrients people are getting. As the saying goes: garbage in, garbage out.

One more thing. The FFQ were validated by three-day diet histories, which are also prone to significant error.

The Bottom Line

Research that examines dietary intake may be prone to errors. It doesn’t make it worthless; it just means we have to interpret the results carefully. This is especially true when determining whether any specific diet can help reduce disease or prove whether a nutrient is beneficial or not.

What we can do is speak in global terms. Eat better. Eat less. Move more. Do that first and worry about the details later. Even with the potential errors in assessing food intake, there’s no question about that.

So here’s what I challenge you to do: for the next month, make a strong effort to eat better than you do right now. I think if you take this first step, you’ll feel the difference.

What are you prepared to do today?

        Dr. Chet

References:
1. Front. Endocrinol. 2019. 10:850. doi: 10.3389/fendo.2019.00850
2. JAMA. 2006;295:655-666.
3. Open Heart 2021;8:e001680. doi:10.1136/openhrt-2021-001680

How to Assess Food Intake

/in /by

If you’ve been reading the Memo for any length of time, you know that I’m not fond of the methods used for determining food intake in free-living individuals, especially the Food Frequency Questionnaire. When looking at the validity of the doubly-labelled water technique for last week’s Memos on metabolism, I happened upon a review article that examined several methods of collecting food intake in nutritional studies; they also assessed metabolism to see if the calories used equaled the calories taken in.

Researchers from Australia reviewed the published research and selected 59 articles that examined which method of assessing food intake was the most valid as verified by metabolic data. Besides the FFQ, with and without food models, they examined food diaries, food histories, and 24-hour diet recall with and without the use of technology.

They found that with a couple of exceptions, every method of collecting food intake underreported energy intake by 1.5% to 47%. The researchers concluded that while every method had high variability, 24-hour diet recalls were the most accurate with a variability of 8% to 30%. The highest degree of underreporting? The FFQ which had one study top 47%—that’s right, almost half of the food eaten wasn’t reported!

It makes sense to me; Paula and I rarely have pizza or bacon, but do we remember how many times we’ve had those in the last year? Of course not. Here’s a harder recall issue: Paula has a bite or two of chocolate almost every day, but rarely eats a whole chocolate bar. How would she report that accurately? And I often eat Riley’s leftovers—how do I report those two chicken nuggets or one-eighth ear of corn? In the real world, it’s hard enough to accurately record what you’re eating right now, let alone a month or a year ago.

Besides my personal satisfaction of being correct, the real question is: “What does this mean in the real world?” I’ll cover that on Saturday.

What are you prepared to do today?

        Dr. Chet

Reference: Front. Endocrinol. 2019. 10:850. doi: 10.3389/fendo.2019.00850

My Clementine

/in /by

In doing the background research on whether citrus increases the risk of melanoma enough to be a real concern, there was one more important study that was published this year. The same basic group of researchers who did the original examination of citrus intake in the two large healthcare professional studies added one more component. They used a database of FURs (furocoumarins) levels from 10 different foods that were included in the food frequency questionnaire and re-examined the data. This time, there was no relationship between FUR and melanoma; there was a small relationship with basal cell carcinoma but nothing of great significance.

I’m still going to eat my clementine every morning and drink my energy drink that has grapefruit juice in it when I want to. But these studies did raise at least two questions.

The Furocoumarins Database

The database was constructed by selecting foods that were suspected of having high FUR levels, which makes sense. The chemical analysis of each food is above my pay grade, but let’s assume it’s accurate since no one questioned it in the years since that study was completed. They purchased the foods for analysis from 17 different grocery stores in Connecticut and then analyzed three different samples of the same food purchased from different stores. All good.

The issue is that phytonutrient content can vary based on the season of the year, the ripeness of the fruit or vegetables, even the time of day it was harvested. FURs are stress-induced molecules. After harvesting, FUR levels can be affected by many factors ranging from ultraviolet light exposure to insect infestation. The levels can also vary by the variety of the fruit or vegetables. Then, when it comes to juice, there are processing and storage factors to consider. None of that was considered in the database study.

To be fair, buying foods that we actually eat is the right thing to do. But as a grocery shopper, you know that you can purchase two identical containers of food with two different expiration dates. That has to be accounted for, even if only to find out it doesn’t make a difference. In the case of FURs, it appears the date matters, based on prior research.

Food Frequency Questionnaires

If you’ve been reading the Memo for any length of time, you know that I’m not a big fan of food frequency questionnaires for the way they’re typically used today. The FFQ was designed to give an overall estimate of what people eat, but it’s not a measurement of what people actually ate. When you want to know the overall servings of citrus a person eats, fine. When you begin to break it down into extremely discrete serving portions, that’s where things fall apart in my opinion. Think of how many of these large studies that have used the FFQ that have gotten results that indicate nutrients are associated with some disease or another. In other studies, the reverse is true.

The reason that these are imprecise is because they were validated with a low number of subjects: 173 for one FFQ and 150 for another. The correlation coefficients are just not high enough to put a lot of faith in the FFQ for anything other than a general idea of diet. I’ve got a lot more to say about the topic but this isn’t the forum.

In addition, I think we’d be foolish to overlook the fact that the places where citrus grows in the backyard, and is therefore more convenient and available, are the same places that get more sunny days every year and more intense sun. The researchers controlled for self-reported sun exposure, but I’m almost as skeptical about that as about FFQs.

The Bottom Line

Based on the sum total of all the research, FURs are in some way related to the development of skin abnormalities, especially if you spend a lot of time in the sun. But given all the issues with the FFQ, at least in my opinion, it isn’t enough to avoid citrus if you typically eat or drink it. Pass my clementine, please.

What are you prepared to do today?

        Dr. Chet

References:
1. J Clin Oncol 2015:33:2500-2508.
2. Nutr Cancer. 2020;72(4):568-575.
3. J Food Science. 2011. doi.org/10.1111/j.1750-3841.2011.02147.x
4. J. Agric. Food Chem. 2017, 65, 24, 5049-5055.