← Sigma Nutrition Radio
Sigma Nutrition Radio
#615: Why Processing Doesn't Determine the Healthfulness of Plant-Based Diets – Alysha Thompson, PhD
00:56:18
Loading summary
Transcript33 lines
- [00:00]
A
Hello and welcome to Sigma Nutrition Radio. This is episode 615 of the podcast. My name is Danny Lennon and you are very welcome to the show. Today we're going to be discussing a set of really interesting and often debated topics within nutrition science at the moment through the lens of a recent paper that was published that sheds light on some of these questions. And I think, as we'll see, there are many interesting questions embedded within the original research question. And so we're going to be taking a look specifically at a prospective cohort study that was titled Plant Based Diets, Ultra Processed Foods and Risk of Mortality and Major Chronic Diseases. And so to walk through not only the paper which is really interesting in and of itself, but some of the larger implications and some of the bigger research questions that it relates to, I'm going to be talking with the lead author on that paper, Dr. Alicia Thompson. Dr. Thompson is a research fellow at the Coe center for Sustainable Food Systems and Institute for Global Food Health at Queen's University Belfast. She has a PhD in nutritional epidemiology and most of her work relates to working with large prospective data sets. For example, today we'll be looking at her work that used a UK biobank and she's involved in a range of work that relates to plant based diets, flavonoids, and how some of these things connect to chronic disease risk like type 2 diabetes, cardiovascular disease, cancer and kidney related outcomes. For those of you who are premium subscribers of the podcast, you will be able to get a set of detailed study notes that breaks down the paper, goes through everything that we mention in today's episode as well, in addition to useful graphs, explanations of certain key terms, definitions and so on. So that will be available linked in the description box where you're currently listening. And after this episode you will hear a key ideas roundup for those of you listening on the public feed of the podcast. If you're interested in getting more out of your podcast listening and be able to get access to our detailed study notes for each of these episodes, as well as a premium exclusive episode each month, then take a look at Sigma Nutrition Premium. It's our membership that not only supports the podcast directly and allows us to keep doing this, but gives you these extra educational resources that will not only supplement your learning, but hopefully allows you to recap over that much easier, to revise this content much more effectively and to be able to retain more of what you actually listen to. So that will be linked up in the description box where you're listening right now or just check out SigManutrition.com and we'll have some details there. And for anyone that considers becoming a member, thank you and hopefully you enjoy the content. So with that out of the way, let's jump into this conversation between myself and Dr. Alicia Thompson. A very big welcome to the podcast to Dr. Alicia Thompson. Thank you so much for taking the time to come and talk to me.
- [03:08]
B
Thank you. Thank you for having me. It's great, great.
- [03:10]
A
I think, as I mentioned to you, as I'm sure we'll get to a bit later on after reading one of your publications, I thought not only was an excellent paper that we'll talk about, but also raises some really important points that are big topics in nutrition science at the moment. And so it's a good gateway into those types of conversations. But before we get into any of my questions, can you let people know a bit about your work, your research interests, anything else you think might be relevant to them?
- [03:36]
B
Thank you for having me here, Danny. My name is Alicia Thompson and I'm a research fellow at the COE Centre for Sustainable Food Systems at Queen's University Belfast. And broadly my work is within nutritional epidemiology. So I'm interested in how habitual dietary patterns relate to long term health outcomes, particularly chronic diseases like type 2 diabetes, cardiovascular disease, cancer and also mortality. So a lot of my work thus far has been on plant based diets, but really trying to move, move beyond the simple question of whether someone eats a plant based diet or not. I'm interested in the quality within the plant based dietary pattern because two people can have very different plant based diets, they eat quite different foods nutritionally within that as well. Alongside that I've also worked on flavonoids, flavonoid rich foods which fit quite naturally into this area. So foods like tea, berries, citrus fruits and other plant foods that contribute to a range of bioactive compounds and we're interested in these patterns are also associated with better long term health. And then more recently this paper on plant based diets and ultra processed foods has come about and again we're just again delving into the area of plant based diet quality and what contributes to this. So yeah, that's the sort of quick overview of my work so far.
- [05:00]
A
And as we'll get to this prospective cohort study that you've published, there's a lot of really interesting layers beneath it, as I'm sure I'll get to a bit later on. But from an overview level, when we're looking at not only this healthfulness of plant based diets. But this interaction then with in this case a classification of ultra processing or not, can you maybe lead off with for you, what was the research question that you were trying to answer that went behind the design of the study in the first place?
- [05:28]
B
Yeah, great. So essentially a lot of our work has looked at plant based diet quality previously. And then obviously with this increasing trend of looking at ultra processed foods, we thought how does this fit in the context of a healthy plant based diet? So our main question essentially was whether ultra processed foods or the content changes or influences or modifies the association between plant based diet quality and long term health outcomes. So I would frame the study as setting the intersection of two active areas of nutrition research. So the first being plant based diet quality, where the literature is increasingly showing that healthier plant based diets are associated with better health outcomes and that quality does matter. A diet can be plant based, but still quite different in nutritional content. A diet rich in whole grains, fruits and legumes is quite different to one that's higher in refined grains and sugar sweetened beverages, for example. The second area of ultra processed foods is growing in evidence linking UPFs to adverse health outcomes. And we wanted to better understand how this fits in the context of plant based diet quality. So is it the processing itself? Is it nutrient profile, energy density, palatability? There's so many different components and drivers of this group. So yeah, our question deliberately was more nuanced than are plant based diets healthy or are UPFs harmful? We wanted to know whether the UPF content of plant based dietary patterns influenced the associations with mortality and major chronic disease risk.
- [07:05]
A
And I'm sure as we'll get to much later on in this conversation, even that raises a whole host of other questions that are really interesting to consider in, in the fact that not only are we considering now healthfulness as this broad concept, and we can think of more healthful or unhealthful versions of, let's say a plant based diet. Now it ties in the question of if people are interested in this ultra processed as a label based nova or not, is that the right way to make that demarcation between healthful or not? And so I'm interested to ask you a bit about that later on, but to give people the overview of what the study actually looked like, can you maybe walk us through the main elements of the study design, how you went about that and then as we do that maybe some explanations for why it was designed in the way it was?
- [07:52]
B
Sure, yeah. So this was a prospective cohort study that we use, the UK Biobank. I think a lot of people are familiar with this study now, it's widely used across the world. So we assessed diet first and then followed participants for all cause mortality, type 2 diabetes, cardiovascular disease and cancer. So I call that the core four because we've done this a lot in, in my research anyway, looking at these four core endpoints. So again, yet the study used the UK Biobank. It's a great resource, a great prospective study and it was based in the uk. We it recruited over half a million study participants between 2006 and 2010 and for this analysis we included participants aged 40 to 70 years old. So it's a middle aged adult cohort and participants in analysis were included if they completed at least two 24 hour dietary assessments. So the thing with the UK Biobank is that great resource recruited half a million people and at baseline, all study participants completed a baseline assessment which included completing various touchscreen questionnaires, providing biological samples and imaging, for example. However, in the touchscreen questionnaire they asked several questions on diet and lifestyle. However, the dietary questions were very basic and minimal and this was a food frequency questionnaire and I think it covered 21 questions on diet. So this was very minimal, not very detailed for us to carry out detailed analysis. So a few years later, clearly researchers decided, oh, actually the dietary data isn't so great with what we have for all half a million participants, let's implement a more detailed and comprehensive, particularly chronic diseases like type 2 diabetes, cardiovascular disease, cancer and also mortality. So a lot of my work thus far has been on plant based diets, but really trying to move beyond the simple question of whether someone eats a plant based diet or not. I'm interested in the quality within the plant based dietary pattern because two people can have very different plant based diets, they eat quite different foods nutritionally within that as well. Alongside that I've also worked on flavonoids, flavonoid rich foods which fit quite naturally into this area. So foods like tea, berries, citrus fruits and other plant foods that contribute to a range of bioactive compounds. And we're interested in how these patterns are also associated with better long term health. And then more recently this paper on plant based diets and ultra processed foods has come about and again we're just again delving into the area of plant based diet quality and what contributes to this. So yeah, that's the sort of quick overview of my work so far as
- [10:43]
A
we'll get to this prospective cohort study that you've published. There's a lot of really interesting layers beneath it, as I'm sure I'll get to a bit later on. But from an overview level, when we're looking at not only this healthfulness of plant based diets, but this interaction then with, in this case a classification of ultra processing or not, can you maybe lead off with for you, what was the research question that you were trying to answer that went behind the design of the study in the first place?
- [11:11]
B
Yeah. Great. So essentially a lot of our work has looked at plant based diet quality previously. And then obviously with this increasing trend of looking at ultra processed foods, we thought how does this fit in the context of a healthy plant based diet? So our main question essentially was whether ultra processed foods or the content changes or influences or modifies the association between plant based diet quality and long term health outcomes. So I would frame the study as setting the intersection of two active areas of nutrition research, so the first being plant based diet quality, where the literature is increasingly showing that healthier plant based diets are associated with better health outcomes and that quality does matter. A diet can be plant based, but still quite different in nutritional content. You know, a diet rich in whole grains, fruits and legumes is quite different to one that's higher in refined grains and sugar sweetened beverages, for example. The secondary of ultra processed foods is growing in evidence, evidence linking UPFs to adverse health outcomes. And we wanted to better understand how this fits in the context of plant based diet quality. So is it the processing itself? Is it nutrient profile, energy density, palatability? There's so many different components and drivers of this group. So yeah, our question deliberately was more nuanced than are plant based diets healthy or are UPFs harmful? We wanted to know whether the UPF content of plant based dietary patterns influenced the associations with mortality and major chronic disease risk.
- [12:48]
A
And I'm sure as we'll get to much later on in this conversation, even that raises a whole host of other questions that are really interesting to consider in the fact that not only are we considering now healthfulness as this broad concept, and we can think of more healthful or unhealthful versions of let's say a plant based diet. Now it ties in the question of if people are interested in this ultra processed as a label based nova or not, is that the right way to make that demarcation between health flow or not? And so I'm interested to ask you a bit about that later on, but to give people the overview of what the study actually looked like, can you Maybe walk us through the main elements of the study design, how you went about that, and then as we do that, maybe some explanations for why it was designed in the way it was.
- [13:34]
B
Sure, yeah. So this was a prospective cohort study that we used the UK Biobank. I think a lot of people are familiar with with this study now it's widely used across the world. So we assess diet first and then follow participants for all cause mortality, type 2 diabetes, cardiovascular disease and cancer. So I call that the core four, because we've done this a lot in, in my research anyway, looking at these four core endpoints. So again, yet the study used the UK Biobank. It's a great resource, a great prospective study and it was based in the UK. It recruited over half a million study participants between 2006 and 2010. And for this analysis we included participants aged 40 to 70 years old. So it's a middle aged adult cohort. And participants in our analysis were included if they completed at least two 24 hour dietary assessments. So the thing with the UK Biobank is that great resource recruited half a million people. And at baseline, all study participants completed a baseline assessment, which included completing various touchscreen questionnaires, providing biological samples and imaging, for example. However, in the touchscreen questionnaire they asked several questions on diet and lifestyle. However, the dietary questions were very basic and minimal and this was a food frequency questionnaire and I think it covered 21 questions on diet. So this was very minimal, not very detailed for us to carry out detailed analysis. So a few years later, clearly researchers decided, oh, actually the dietary data isn't so great with what we have for all half a million participants. Let's implement a more detailed and comprehensive 24 hour dietary assessment. So this was called the Oxford Web Q Dietary Assessment and it was issued three years after the initial recruitment and it went on between 2009 and 2012. So for the course for about four years, up to four, five dietary assessments were issued to the participants. However, this was in a subset of participants. So this leaves us with around 210,000 participants who've completed at least one dietary assessment. So I find that a little bit disappointing because compared to the half a million people we could have been working with, we're now sort of brought down to that smaller figure, but a great resource all the same. So yet in this analytical sample for this study, we were left with just under 125,000. And this was because we restricted our analysis to those who completed a minimum of two 24 hour dietary assessments. And this was just to Remove a little bit of the noise associated with one dietary assessment and how this may not capture a great idea of participants habitual dietary patterns. So we included those who completed a minimum of two for that reason. So yeah, the participants reported what they ate and drank over the previous 24 hours. And we averaged the intakes for the food items that we required for our scoring system across all five of the dietary assessments that were eligible. And after characterizing participants diets and following them using the linked health records, we then were left with the outcomes that we decided on which were all cause mortality, cardiovascular disease, type 2 diabetes and incident cancer. And our follow up ranged from about 8.3 to 10.5 years depending on the outcome. And our statistical analysis approach was using Cox proportional has regression models and, and these models were used to estimate whether people in the higher adherence group of the dietary pattern had a higher or lower rate of developing an outcome over this follow up period compared to those in the lower adherence group.
- [17:25]
A
Can you maybe speak a bit more for people about what you ended up having then for those dietary patterns at the end and that kind of interaction of both the nutritional quality side of it and then the kind of processing emphasis as well and where that left you with these different patterns?
- [17:40]
B
Sure, yeah. This is where it gets a bit complicated and actually if you take a look at the study it's quite mind blowing. There was a lot of different dimensions to this. So to break it down simply, hopefully we combined the two frameworks. So the plant based diet indices which were previously created and published by Harvard colleagues back in 2016 and we used this scoring approach, the plant based diet index or PDI as I'll probably refer it to us and we replicated this in the UK Biobank. So again using these this indices first and then we brought in the NOVA classification system to capture the processing level of the food items that were included within each food group included in the plant based diet index. So that was crossing the two dimensions which gave us four different dietary patterns. So yes, so this left us with a high UPF hpdi, a low UPF hpdi, a high UPF UPDI and a low UPF updi. And each score was characterized characteristic of different healthier plant foods or unhealthy plant foods. And just to note that within this dietary pattern and within this plant based diet index, although this is a plant forward or a plant rich dietary pattern, this is not omitting animal based foods at all. So I think that's a bit of a misconception that's had whenever we talk about plant based. There are so many different types of plant based diets and not all of them omit animal based foods. And this was quite nice within this scoring approach is that we didn't have to exclude people who consumed or reported consuming animal based foods. This included everybody. But the scoring system was catered towards positive scores for those consuming plant foods and negative scores for those consuming animal based foods. And then this was broken down again based on whether it was an HPDI or updi. So, so this was how the score is broken down. Just to detail, within the second dimension of the processing level, we classified the foods according to nova. We separated the foods into UPFS and non UPFS or low UPFS as we'll describe it, because within this group we also include groups that are included for Nova 1 to 3. So yeah, for the purpose of the modified scores, UPFS were Nova group 4 and non UPFs were Nova groups 1 to 3. So when we crossed the two dimensions, we were left with the four dietary pattern scores. So this meant that instead of only asking whether someone ate a more healthful plant based diet, we could then ask whether the healthful plant based diet dietary pattern was higher or lower in UPFs. And I think the important thing to mention here is that these are research indices. So a phrase like high UPF healthy plant based diet sounds quite strange in ordinary language, but technically it means a healthier plant based pattern whether the foods where the foods contribute to the score are classified as ultra processed or not. So it doesn't necessarily mean that every UPF is healthy in this context.
- [20:54]
A
But I do think it raises an important point that I'm sure we'll get to later on where now, particularly in a lot of mainstream messaging we're seeing oversimplified messages of basically be equating the amount of UPFs with whether it's healthy or not. And as we can see, there's a lot of nuance here that I'm sure we'll get to one of the specific questions, just as you mentioned the NOVA classification and you mentioned the use of the Oxford WEBQ as well, given when a lot of this dietary information would have been collected and then the ability to go and then assess that and then knowing some of the nuances. And there's still debate within the utility or not the utility, but how we assess certain foods on nova, if it's not specifically designed to do that, it can be quite a challenge from a research perspective to accurately put it in one of those categories. Can you maybe talk about how you went about that. And was that indeed something that was difficult to classify, what people had reported eating back when that data was collected?
- [21:50]
B
Yeah, sure. That was definitely one of the challenges and limitations of the study. So again, again, a huge challenge with the work. NOVA is useful in some parts, so applying it to self reported dietary data isn't always straightforward because these recalls often lack full product and ingredient level detail that we require in order to categorize these foods into Nova Group 4. So Foods reporting in large cohorts involves some uncertainty, always, especially within the plant based diet index as well, across different cohorts. It's just, although we're trying to replicate, there's no such thing as perfect. So again, participants report with the eight, but they may not always provide details on the brand, the full ingredient list, preparation method, or whether food was homemade or packaged, fortified, sweetened. And this is quite important for the UPF or NOVA categorization. So NOVA classification only requires information about the extent and purpose of the processing. And for some foods this isn't straightforward. So for example, fruits and vegetables or legumes are clearly not UPFs. And sugar sweetened beverages and or those drinks and confectionery type products are usually easier to classify as UPFs because usually they are all in that category. But some foods sit a bit closer to this boundary. So we found that a little bit tricky when it came to some of the food groups within the plant based diet index. For example, fruits, we couldn't classify any fruit item as a UPF in the UPF category. So this meant that later on in our adjustments and models, we had to adjust for fruit whenever we were looking at the sole dietary pattern of upf. So this is another limitation of the study and something we just had to consider and we're very transparent about. So again, the classification involved a lot of judgment, a lot of subjective judgment. And I think this is what's seen across many studies that are applying the NOVA system to dietary data. And yeah, it's not unique in our study, it is a broader challenge of UPF research and it just means that transparency and sensitivity are important in carrying out these sensitivity analyses to better understand how UPFS fall within this broader idea of diet and habitual diet. But it also means that NOVA shouldn't be interpreted as a perfect measure of diet quality. And I think this is where we have got a little bit mixed up in recent research maybe is looking at UPFS as a measure of diet quality. And essentially it is, but alongside or in parallel with other measures, I think that's important. And I think balanced position would be that Nova adds useful information about food processing that traditional nutrient profiling may miss. But applying it to real world dietary data sets requires careful judgment and cautious interpretation. And I think sometimes the media grabs hold of that and runs with it.
- [24:55]
A
And I do want to return to that point later on because there is much to discuss about not only that misinterpretation, but the implications and for not only research but public policy going forward. Forward. But if we return to the study, so we've talked about this study design these different patterns and we can look at from a top line level what were some of the main results around your kind of primary outcomes as well as for you, any of the results that are important to raise to people or that stood out in some way.
- [25:23]
B
Some nice and interesting findings from the paper. So essentially the broad pattern was that healthier plant based dietary patterns were associated with more favorable outcomes, especially all cause mortality and type 2 diabetes, whether or not they were higher or lower in UPFs. And that's the interesting thing to note is that we find these beneficial associations regardless of whether the healthy plant based dietary pattern was higher or lower in upf. So more specifically for all cause mortality we found that higher adherence to healthier plant based patterns were associated with lower risk. And that was true for both high UPF HPDI and low low UPF hpdi. And these the hazard ratios were very similar between the two dietary patterns when it came to risk of all cause mortality comparing highest to lowest adherence groups. When we looked at type 2 diabetes again we found very similar associations but a little bit stronger in the strength of association. The findings for healthier plant based dietary patterns were that we found this association with lower risk, but the strongest associations were for the low UPF hpdi. So this also again emphasizes that whenever the underlying dietary pattern and nutrient density is high, there is an additional benefit of low UPF diet in the context of type 2 diabetes risk. So although we find that a high UPF healthful plant based diet showed lower risks of type 2 diabetes, we find a little stronger association for the low UPF hpdi. So there's there might this association shows us there might be an additional benefit of consuming low UPF products in this context, moving on to cardiovascular disease. So this was a little bit of a complicated finding and maybe one that people would question why we found that the high UPF HPDI was associated with lower risk whereas the low UPF HPDI was not clearly associated. In fully adjusted models this result sounds quite counterintuitive. So I would explain it, explain it by, I suppose, returning to what the high UPF HPDI captured. So the this meant that participants were consuming higher whole grain intake, lower refined grains, sugar, sweetened beverages, sweets, desserts and animal foods. So we can propose or suggest that these more beneficial associations we see with cardiovascular disease risk were potentially driven by the higher whole grain intake that we see in the UPF category. So this covers breads, whole grain breads, the these cereals like muesli and these oat based cereals as well. So that's all we can propose right now because we haven't carried out additional or more mechanistic analyses to look into these specific associations. But that's what we can propose and that's what we find when it comes to cancer. Again, like many previous analyses, looking at this endpoint, the findings were less clear. The healthier plant based dietary patterns were not clearly associated with total cancer risk. Again, this might be because total cancer is heterogeneous as an endpoint, different cancer sites have different causes and dietary relationships. So total cancer can mask site specific associations. So maybe in future studies it would be more interesting to look at different subtypes of cancer rather than total cancer overall. However, in this study we didn't have capacity to do this. Following on from that, the flip side of these associations is also important. So less healthy plant based patterns were generally associated with higher risk. So this comes back to the initial point of not all plant based diets are equal. Just because you consume a plant based diet doesn't necessarily mean that it's a healthy one. So when we split it into healthy and unhealthy plant based diets, we see differing risk outcomes in these large prospective cohort analyses. So, so again, the main message here is not simply that plant based diets are protective, it's the quality of the plant based dietary pattern that matters strongly. Another analysis, it's worth noting, or another sort of section of the paper, was alongside plant based diet indices and NOVA classification, we also used a modified nutrient quality index, or mnqi, which we refer to it as in the paper. And this allowed us to look at whether the high and low UPF versions of the plant based dietary pattern differed nutritionally. So this index included positive components like fiber and protein and less favorable components such as saturated fat, total sugars and salt. So it gave us a way of asking, are these high and low UPF plant based patterns actually very different in nutrient quality or are some of them more similar than the processing label alone might suggest? So what was interesting was that Both high UPF and low UPF healthy plant based diet indices had similarly high nutrient quality scores. So you can see in the paper in Figure 2, this is a really nice bar chart where we compare the nutrient quality of the high and low UPF PDIs and you can see that there's really no difference between the two. This helps us explain why high UPF healthy plant based dietary patterns could still show favorable associations because of this underlying nutrient density of a healthy plant based dietary pattern despite it containing upfront food items. It's not simply that a diet high and low quality, it's more than simply a diet that's high or low in upf. So looking at this nutrient density gives us an additional nuance into this idea of plant based diet quality. And then on the again the flip side, in contrast, when we looked at the unhealthful plant based diet indices, these had poor nutrient profiles including lower fiber, protein, calcium, vitamin B12, particularly in some of the lower upfront variants. But again, comparing the high and low UPF unhealthful plant based diet indices, these nutrient profiles look quite similar between the two. So this was a really interesting finding. Although nothing complex or hard to carry out. It was just quite nice to see what does a high and low UPF plant based diet look like when it comes to nutrient density.
- [32:00]
A
So there's a lot to work through and I think the first thing people will see here is that if we are again, there's nuance within each of these different outcomes and I don't want to reduce any of that, but from an overview level we can see that perhaps it might be fair for someone to think if we are looking at those distinctions between the healthful and unhealthful patterns, those are relatively straightforward and sit with what we know from wider literature as well, that those healthful plant based patterns are consistently shown these associations versions with positive outcomes relative to the unhealthful versions, where it starts to get interesting, depending on the outcome, is this question of whether thinking about the amount of UPFs or Category 4 ANOVA foods adds any refinement to that or gives us any extra information. Then rather just looking at the nutrient profile or a categorization of a healthy and unhealthy pattern, and at least for some of those outcomes like diabetes, there could be some signal there. But again, there's still a lot of work to get to it and you started doing some of that work to try and distinguish why we might be seeing that. So I guess from your perspective, given the various hypotheses that are out there. What do you think may be good potential hypotheses that explain these potential differences and if they really are inherent to NOVA classification of or not. And I realize that's such a broad, wide question, but feel free to go in whatever direction you think might be useful.
- [33:28]
B
Yeah, a really good question and complex to answer. So again, it's hard to distinguish what among the high and low UPF diets may be driving the associations. But what we can say is that whenever we looked at the food group analysis, there were specific food groups. So we broke it down to the food group level and we carried out several sensitivity analyses within this paper, including a subgroup analysis breaking down each of the 17 food groups that make up the plant based diet index. So what we found here were that there were specific food groups that drive these associations that we see. So coming back to a lot of the literature again, and a lot of the papers I've published previously have shown the same thing, especially when it comes to type 2 diabetes. It's the sugar sweetened beverages that typically drive the adverse associations and it's the whole grains and fruits and vegetables that typically drive the positive associations that we see with type 2 diabetes specifically. But again, it's quite a broad, it's quite a broad message that we've seen in previous literature too when we broke down each of the food groups into their high and low UPF counterparts. And actually again, it emphasizes what we concluded in our paper was that high and low UPF whole grains were beneficial for these hard endpoints. The likes of meat was adversely associated whether it was high or low UPF, especially for type 2 diabetes. We didn't find any associations for the likes of dairy eggs. We found a beneficial association for fish in the high UPF category. So it was very mixed with the associations that we found in the end when we broke it down to food group. But whenever we carried out additional sensitivity analyses which were vast, we carried out a leave one out analysis again to emphasize which of these food groups may be driving these associations that we see, whether they're adverse or a positive association. And again with the leave one out analysis. Although these results were pretty consistent with what we saw in the main analysis. Again we find that when we remove sugar sweetened beverages, this is what sugar sweetened beverages seem to be driving a large part of the adverse associations or the positive associations. Because lack of sugar sweetened beverages could have a beneficial impact on our associations. We also carried out additional sensitivity analysis looking at independence of associations from nutrient quality. So we adjusted for this modified nutrition nutrient quality index in our primary analysis and again our results were broadly consistent with what we saw in the primary analysis. We also looked at the ratio of low to high UPFs against the PDI scores to see whether higher UPFs against low UPFs may be, there may be some differing association. And again, there were no consistent associations with the outcomes. And we carried out this additional association we or this additional analysis. We also carried out potential modification between PDI scores as a raw score and UPF intake to see whether the HPDI and UPDI was modified by total UPF intake. And again we found no significant effect modification by UPF intake. We then went on another level. We looked at effect modification by sex and assessed through sex stratified analyses and we had consistent associations with our primary results. So again we went through these different levels of sensitivity and subgroup analyses to really dive into and discover the underlying, what might be driving these underlying associations. And it really points to, it really just points to this basic message of there are specific food groups within a plant based diet and within the NOVA classification system that do drive these associations that we see. And that's why it's important to not just look at the Nova 4 as one group in isolation, but to consider them alongside diet quality, alongside nutrient quality. Context is really important. Looking at the food environment, why are people consuming these foods? I think we need to ask more questions rather than just taking it for face value. And I think that's what we have concluded from this paper is that there's so many more questions to ask and so many more analyses to be done to really pull this apart and see what more is there to do with NOVA diet quality that we can figure out and analyze.
- [38:30]
A
And before I return to your specific study to maybe one more question pondering on this utility of nova, because this is a huge debate within nutrition science right now, all the way from, in some cases people who are very much on board with the use of NOVA for all sorts of things, things to the extreme of very, I would say in good faith putting forward very good points around maybe not only a lack of utility, but putting us even in the wrong direction and then a whole host of opinions in between. And so we've seen some other work from various groups that has indicated some of the things that you have touched on where the level of processing or whether something is in category four or not doesn't explain differences or seeing. And so maybe isn't refining that message. But then again, there's other people in that point to just these associations. We're seeing with ultra processed foods and how can we completely negate this? And so foreseeably there could be people coming across your study now and maybe listening here that have, depending on their belief or position, could try and interpret it in two different ways. Right. One could be that, well, maybe whether something is high or low, you UPF still does give us this additive piece of information and we layer it in on top of some of this other stuff and it's useful where someone else might interpret it to say, well really is it adding much more than once we know the food groups that determine healthfulness or not and we account for those and we know that's going to be driving much of these associations. What foreseeably could be in a plant based diet that would fall into this healthful pattern, but still is high UPF that is going to cause a problem or vice versa. We could point to examples where that's clearly not the case. Right. And on an individual food level. And so given that there's these different types of interpretations, at least from the data that you've put forward, what do you think is the, the fairest interpretation people should come to or which one of those ends of the spectrum that say, do you find yourself, yourself falling more towards?
- [40:33]
B
Really good question. And yeah, there seems to be two sides at the moment. But with what we were trying to understand with this paper was a very neutral story. And I think that's what is shown with the paper with our interpretation, I suppose. And really what we want to conclude is that whenever the underlying diet quality is high, UPFs can fall into a healthy dietary pattern. Whenever this underlying diet and nutrient quality is high. I think that NOVA can provide useful information, but I don't think it should be used alone. It's a useful lens, but not a complete measure of diet quality. And it's maybe again, as I mentioned previously, it's maybe being used that way. But again, there's limitations to every study, but its strength is that it does capture something traditional nutrient profiling often misses. So that's what was interesting in our study, that we could then specifically split these UPF high UPF and low UPF foods according to the food groups that are present in the plant based diet index. So that was a nice nuance to the paper. However, coming back to the limitations of the study, trying to split these food items into how they are actually categorized as UPF is quite tricky when we don't have the formulation information, the ingredient information. We have a basic 24 hour dietary recall and that's what we're working with and our output is only as good as, as the input. So I think that's important to emphasize that we are just using the dietary data that we're being provided with and that's, that's the best that we can do. So that's why it's important to keep running additional analyses, asking more questions in this area. And I think for nutrition research, NOVA can help us ask questions that nutrient based approaches may miss alone. And for public health, I think it can help describe patterns in the food supply and identify how industrial or industrially formal formulated foods contribute to population diets. But its limitation is that novo groups together foods that may be very different nutritionally. So again, a sugar sweetened beverage or confectionary product is very different in fiber, protein or saturated fat. So my position would be, is that NOVA does not explain everything. And NOVA is not useless. But the useful question would be is what does NOVA add beyond nutrient quality, food groups and dietary pattern analysis? And when does it improve our interpretation of diet disease relationships? And I think these are the key questions we should ask when running these analyses and breaking these analyses down.
- [43:15]
A
Great point. And I think that refers to what you'd previously said about many of these questions we still need answers to because NOVA may be picking up something, but right now we actually don't know what that necessarily is. And as you said, we have these really heterogeneous groups. And so, so really we're bucketing things together that maybe isn't the case. If we're saying it's down to the degree of processing. Well, there's other ways to assess that don't take into account the intention behind it. And so maybe we need a different way to assess that, but lots of really interesting open questions. I did want to return to one thing about your particular study because as in any of these cases, when we see a prospective cohort study, the way that adjustment is done for a number of these factors is critical, crucial. And you guys did a really nice job on some of this covariate adjustment. Can you maybe talk a bit about that, how you went about that, given the complexity of the exposures that we're talking about here?
- [44:07]
B
This is a complicated point and something that our group does have a lot of experience in. And we have looked at these endpoints previously when it comes to plant based diet quality. But again, when we're pulling in this additional dimension of ultra processed foods, that's when things change a little bit. So we adjusted for a wide range of sociodemographic Lifestyle, health and dietary factors to reduce this confounding that we talk about, while recognizing that some variables can also sit partly on the causal pathway. So in again, observational research, as many of your listeners will already know, people with different dietary patterns often differ in many other ways. People with healthier diets may be more physically active, less likely to smoke, have different educational or socioeconomic profiles, or drink alcohol differently, use medications differently, have a different baseline health status. And those factors can really confound this diet, disease relationship, or associations that we see. So we adjusted for this broad set of covariates based off of a literature search, and we update this literature search frequently because we're in the area of research and the papers are coming out vastly all the time, and we have to keep up to date with what covariates are important when it comes to our endpoints and our exposure. So we adjusted for, for the key confounders, in addition to the likes of medication use or polypharmacy, long term health conditions or multimorbidity, the number of dietary assessments that were completed by the participants, because again, someone who has completed all five dietary assessments is probably more interested in their health. So these were other confounders that we had to consider. But also it's an important point to make and one that we have made in previous studies is the mediator question. And this is more nuanced again. And in previous studies we have ran mediation analyses. It would have been quite nice to actually run mediation analyses in this paper too. However, for capacity reasons, it just would have blurred the lines and possibly made the paper a little bit more hard to interpret than it already is because of the different layers and dimensions of the scores. So this mediator idea, it's something that may lie on pathway between diet and disease. And BMI is a great example for this, as it's typically, typically this mediator plays this mediator role and it can confound the association because body weight is related to both diet and disease risk. But it could also mediate the association of diet influences body weight and body weight then influences type 2 diabetes or CBD risk, for example. So this means that adjusting for BMI in our case probably attenuated the associations and may have produced more conservative estimates if part of the effective diet operates through body weight. And this is the same conceptual issue applies to lipids, inflammation, glucose metabolism, or other biomarkers, depending on the research question and timing. However, mediation in this case for our paper wasn't, wasn't part of the question or Hypothesis. So BMI was considered as a confounder alongside the other health lifestyle medication confounders as well.
- [47:28]
A
That's really useful and I think it's something that people can keep an eye on out for when reading any paper that this concept of confounders and knowing the difference between that and these mediating factors, but not only that, how that connects to the actual research question. And as you perfectly put it, that in some cases, depending on that research question, it would be categorized as this confounder. And we have to adjust for the maybe in another situation we may not do that. And so understanding that particularly in such an complex exposure that we have, and especially when we're looking at chronic disease outcomes that are again complex as well, to look at trying to get enough control from appropriate adjustment without over adjusting is this fine balance. So you guys did a really nice job of that. So Alicia, when we think about where this work then leaves us from, some of the takeaways that you not only had from that, but then those other research questions that it maybe opens up that either you have planned for your group to do or maybe you'd like to see see other groups take on. What do you think are some of those pertinent next research questions that over the next couple years hopefully we can start to see some work, try and answer.
- [48:34]
B
Yeah. So I think it's a really exciting space now in nutrition research. Despite the UPF story and narrative being taking over a little bit, there's so much more to look at and there's so much depth in that too. So I think future research needs to move beyond the broad labels, especially for plant based diets. We need more details, detail on quality, nutrient adequacy, processing level, affordability and substitutions. I think that's a really interesting area now, although it's quite a, it's not a new thing. Substitution analysis. I think it's coming back. I think we see these trends in nutrition, these different areas come back again and again. So for upfs, I think we need to look at subgroups rather than total UPF intake alone. And this, these analyses are coming through now in the literature. I think, yeah, that's one clear direction. I think total UPF intake is used, but it can hide this heterogeneity, you know, that we see across the different food groups a second direction, substitution analysis. I think in nutrition the comparator really matters. And if someone reduces one food, what are they replacing it with? And that's the key question. And that's just, that's real life context that we're maybe missing a little bit in nutrition science at the moment. The third direction is nutrient adequacy. I think in our nutrient data or dietary data, we need to consider this nutrient element. I know nutrients are now old fashioned and we're looking at whole foods as one and dietary patterns, but I think nutrients play a key role in addition or in parallel to these dietary patterns. We need to look at how these dietary patterns actually how they look when it comes to nutrient adequacy. Fourthly, I think we need to point in the direction of applying these analyses to more diverse populations. And I think again, us researchers are saying this all of the time. Much of the evidence is coming from the high income, predominantly white cohorts. And we need research across different cultures, food environments, socioeconomic contexts and types of plant based products as well. The food environment has changed massively, especially from a lot of this cohort data, the dietary data that we're using, which for Example in the UK Biobank was reported between between 2006 and or 2009, 2012. The Food Environment has changed drastically since then. And that's another limitation of applying this nova framework to this old dietary data is that our food environment is very different now and this data maybe doesn't capture this UPF framework that we are trying to apply it to. So that's a challenge in itself. I think we'd also like to look on a deeper level, especially within, in our group we also look at which of the components of a plant based diet or a healthy dietary pattern may be driving these associations, not just on a food group level, but on a compound level. So looking at polyphenols, more specifically flavonoids, our group does a lot of work into flavonoids and different health outcomes and has done so within the Harvard cohorts and UK Biobank. And this is just another level. And I think what would be really interesting now is to pull in the omics. What does your protein signature look like? This proteomics signature biomarkers? I think there's so much more than just diet. I think we can look at this underlying biological, these underlying biological mechanisms too. And I think there's so much more in this area to discover and it's quite an exciting space at the moment, especially with the high quality data that we have access to and can work with.
- [52:13]
A
Absolutely. As you said, there's no shortage of really interesting research questions and I think there's lots of groups doing some really nice elegant work on that. So hopefully the next couple years. We'll continue to see more excellent work. Answer some of these questions before we get to the very final question. Alicia, for people that are listening, is there anywhere you could send their attention towards if they want to learn more about you or your group or any of the work that you guys are doing?
- [52:37]
B
Yeah, sure. For my work, I mostly keep LinkedIn up to date. I try my best posting papers that we've published recently. If you want to look at some of the Co center for Sustainable Food Systems work which covers not only nutrition but the whole food systems, which is also important and something that obviously we don't have enough time to talk about today. But another area of research that I'm also looking into with regards to healthy and sustainable diets. If you want to point listeners to the Co center for Sustainable Food Systems website that will keep you updated with what we're doing in Belfast and what's going on here and all of the exciting things to come.
- [53:16]
A
So that'll be linked up for everyone listening in the description. And of course I will link to the study that we have discussed throughout this episode today. I highly encourage you go and read the full text of that and go through that in more detail. So with that, Dr. Alicia Thompson, we get to the very final question that I always end the podcast on, which can be completely outside of what we've discussed today. It's quite a broad open question, so apologies for putting you on the spot, but it's simply if you could advise people to do one thing each day that might have a positive impact on any area of their life, what might that one thing be?
- [53:49]
B
Oh wow, throw me. You've thrown me now, let me see. So I think not related to nutrition or work. Let's keep it away from that. I think to be honest, the secret to a good life is realizing how good it already is. So that's something I like to live by. Look up, up, be thankful and don't worry too much.
- [54:09]
A
Fantastic. A perfect way to round this out. Dr. Alicia Thompson, thank you so much for first of all giving up your time to come and talk to me today for the excellent explanations you've given and more so for your wider work and the excellent work that you're doing. So thank you for that. It's really been a pleasure.
- [54:24]
B
Thanks so much. Foreign.
- [54:38]
A
Thanks so much for listening into today's episode. Before you go, I just wanted to remind you about Sigma Nutrition Premium, our subscription. For those of you podcast listeners who want to significantly deepen your understanding of nutrition science and become truly confident in your knowledge. So what's the idea of this subscription? Essentially, it was created with the goal of allowing you to more deeply understand the material you're hearing on the podcast episodes themselves, to be able to retain more of that after you've finished listening or reading through the notes, and then be able to easily and efficiently revise over that so that in the future you can be able to remember that information, to be able to reuse it, to be able to create your own content or ideas using things that you have learned. And at the core of the subscription is our detailed study notes that you get to each episode that is full of useful descriptions, background content, context diagrams, charts, etc. To allow you to more deeply understand some of the concepts that were mentioned throughout that particular episode, as well as them linking them back to previous episodes. You also get premium only episodes, so for full details on this then check out the link in the description box, wherever you're currently listening right now, or just go to sigma nutrition.com and you can see all the details there. And of course, your support is what keeps Sigma Nutrition going. We don't run ads, we don't sell supplements, anything, anything like that. So your support is what allows me to continue to do this. So thank you for that. I hope you do come back for the next episode regardless. And until then, have a great week. Stay safe and take care of.