A new study published in the Stroke journal has linked the intake of artificial sweeteners with an increase in the incidence of stroke.
Historically, sweeteners have been recommended as a risk free replacement for sugar. Patients with diabetes are perhaps more likely to make the switch from sugar to sweeteners because the effect on blood glucose levels is negligible.
This evidence, if true, could change the landscape of recommendations for patients. I’ve therefore been very keen to get my hands on this article to take a look. This week I finally managed to do just that and I’m ready to report my findings.
The study comes from the women health initiative (WHI). I’ll quote the website here so not to get this wrong. “The Women’s Health Initiative (WHI) is a long-term national health study focused on strategies for preventing heart disease, breast and colorectal cancer, and osteoporotic fractures in postmenopausal women. Launched in 1993, the WHI enrolled 161,808 women aged 50-79 into one or more randomized Clinical Trials (CT), testing the health effects of hormone therapy (HT), dietary modification (DM), and/or calcium and Vitamin D supplementation (CaD) or to an Observational Study (OS). At the end of the initial study period in 2005, WHI Extension Studies (2005-2010, 2010-2020) continued follow-up of all women who consented.
First of all this is an observational study. I’ve blogged on these before but I’ll explain the concept again. An observational study is when you look at a group of people over time and observe their outcomes.
This is different from an intervention study. An intervention study chooses 2 groups of people and adds an intervention into only one of the groups. This then allows the researchers to compare cause and effect of the intervention they have added. Intervention studies are more favourable than any other type of research because we can test the effect of the variable we have added in. Intervention studies also allow us to control more extrinsic variables which may impact the results.
Observational studies on the other hand do not intervene in anyway. You might look at 2 nations and observe their dietary habits. You may then look at their health outcomes and attribute this to diet. A good example might be smoking data and lung cancer.
The obvious downside is there may be other variables contributing to those outcomes. This doesn’t apply so much to smoking data because one group has a very obvious habit the other group doesn’t.
Observational studies usually allows us to recruit many more participants and can still be a useful tool in research. However, they aren’t the best type of research. So keep that in mind.
Not looking at sweeteners as a primary outcome
It may not have passed you by that the primary outcomes the WHI is looking at doesn’t include the effect of sweeteners on health.
This was added later at year 3 of the study and then tracked over time. Ladies were asked to self report their sweetener intake and were followed up over a mean period of 11.9 years.
This already rings alarm bells because the trial has not been set up to look at sweetener intake and stroke events. Exclusion criteria may not be robust enough to account for those individuals who may skew the data. Prime examples include people at high risk of stroke such as patients with diabetes or existing cardiovascular disease who just so happen to have a lot of sweeteners.
However, I can see how it may be appealing to add in some other measures whilst you have a captive audience of 81,714 ladies.
The study group also is only looking at post menopausal women. In research regardless of your results you can only draw conclusions in the group of people you have studied. It might seem counterintuitive but to say these results then apply to 25 year old women is bad practice.
Intriguing study results
The study was the largest of its kind. The researchers found a positive association between higher intakes of sweeteners (twice or more daily) and stroke events.
People were also more likely to suffer a stroke if they were of black origin.
The study quotes similar findings elsewhere including a large study which looked at data from the Framingham study (a large study). This studied 4085 patients and found an association between increased sweetener intake and stroke.
Long term study
Observational studies are tricky ones to interpret. A lot of our nutritional guidance and advice is actually based on observational studies. It is very hard to perform an intervention study with diet. Well, over the long term anyway.
You can get people to eat a certain way for a matter of months. However, to get a large group of people to follow a specific diet for years on end is difficult. This is why most dietary intervention studies usually last a year or less.
The problem is diet exerts its effect on the body over many years. You don’t eat one cake and suddenly have a heart attack. It is the accumulation of repeated dietary behaviours that positively or negatively effect the body. Thus, this is how we must study diet.
So the sweetener study ticks this box.
Remember I mentioned its hard to control extrinsic variables when using an obversataonal study. This becomes quite significantly when interpreting data like this study because they haven’t controlled for many external variables by not using a robust exclusion criteria i.e. the diabetes and CVD patients for example.
So how do such studies account for this?
They use regression models. This sounds complicated but bear with me here and I’ll try to explain. A regression model tries to account for all those extrinsic variables that may also have an influence on your results.
The way you do this is to start grouping participants based on certain characteristics. Lets use a simple example to start. We look at stroke incidence and sweetener intake. However, half of our participants have diabetes which also increases the risk of stroke.
So what do we do.
We separate the non diabetes and diabetes patients into 2 groups and analyse them separately. If the group without diabetes with a high sweetener intake also has an increased risk of stroke the outcomes start to look more convincing. Make sense?
So why is this a problem in this study?
Let me start with the factors they accounted for first to highlight the issue. The study tried to account for, age, ethnicity, smoking status, body mass index and weight, diabetes, cardiovascular disease and hypertension, dietary quality, energy intake, waist to hip ratio, alcohol intake, income and education.
To assess purely the effect of sweeteners on stroke risk the researchers need to separate out the people with none of the above risk factors. This doesn’t include the other factors they haven’t thought about because there are always others.
What are the odds you will find a significant number of people left over after excluding all these people? I’m no mathematician but assuming a lot less than the original study number.
In fact, I’d probably argue it would be very difficult to find even one person in this study population without at least 2 risk factors.
Therefore, the researchers had to try to account for factors as best they could.
Unsurprisingly, the researchers reported the women who had higher levels of sweetener intake were more likely to be overweight or obese, had lower levels of exercise, had a higher energy intake from their diet, had lower diet quality, and were more likely to be past smokers, drinkers and have a history of diabetes, heart attack or stroke.
These are all huge risk factors for stroke.
The researchers tried to account for some of these variables like diabetes and history of CVD. However, they couldn’t exclude all the risk factors.
For them to conclude increased sweetener consumption is correlated with increased risk of stroke I’m not sure is appropriate. The other study quoted from the Framingham study data also experiences similar problems.
With so many risk factors active in this study I’m surprised they even tried to study this in the first place.
How can we ever trust an obversataonal study then?
This returns us to the exclusion criteria. If this study and the Framingham data study weren’t piggy backing other larger studies for their data they would have more control at the start.
You could launch an observational study but exclude most people with the above risk factors. You could even follow healthy individuals from middle age and track them over time. As time progresses you can exclude any individual who develops particular risk factors such as diabetes so you are left with a more purified data set.
Then if people with higher levels of sweetener experience more strokes, I’ll be more convinced.
I’m open to change with dietary advice. I really am. However, the evidence needs to be high quality and convincing.
I’m not convinced this study or the studies it references are robust enough for me to be changing my opinion on sweetener intake just yet.
My stance on sweeteners has always been they can be a good substitute to sugar. Particularly if you have diabetes. However, if you don’t eat sugar or sweeteners in the first place, don’t put them into your diet as their is no benefit to this.
Perhaps in the future more robust evidence will emerge. For now, I’d say if you have them in moderation and don’t over do it, you’ll likely be fine.