Note: A shorter version of this post appears on Weighty Matters.
An opinion article in the journal Trends in Endocrinology and Metabolism by Dr. Susan Swithers about artificial sweeteners made for scary articles last week. Here is a small sampling:
- “Do Diet Drinks Mess Up Metabolisms?” at NPR’s The Salt
- “Diet soda may do more harm than good” at CNN (in which the journalist says Stevia isn’t a chemical and Swithers is quoted saying saccharin is a derivative of tar- in fact it was discovered by a researcher working on coal tar derivatives, but I guess it sounds scarier the other way)
- “Artificial sweeteners could lead to obesity, diabetes” at CBS News
- “Study: Diet soda doesn’t help you lose weight” at USA Today (not a study!)
- Dr. Swithers even appeared on NPR’s Science Friday last week.
And many, many more. Her message reached millions.
These news pieces were largely based on a press release from the journal publisher (Cell Press) titled “The dark side of artificial sweeteners“. The press release and article argue that artificial sweeteners increase risk of obesity, diabetes, and metabolic syndrome (and therefore heart disease). It is true that some research associates artificial sweeteners with these outcomes, but it is also true that some research does not. With a look at the article, there are a number of misleading things about it, and it does not reflect all the evidence. Below I review the article’s sections on artificial sweeteners and obesity, diabetes, metabolic syndrome, and heart disease as I am relatively familiar with the literature. Dr. Swithers and others have done interesting neuroscientific rodent work with artificial sweeteners which is the second half of her article which I do not review here.
Here is what is missing from media stories about this publication:
- This was not a systematic review of all research. In fact, the oldest paper on artificial sweeteners cited was published in 2007. There are a lot of studies published prior to 2007 on this topic.
- All of the cited studies were described in the text in a way that supports a view that artificial sweeteners are harmful, and all study results suggesting artificial sweeteners have no relationship with disease were ignored or relegated to footnotes under the table as you can see below. Some of this is apparently because of the word limit.
- Strangely, these are not the fault of Dr. Swithers, but the requirements of the journal: “Opinions should present a personal viewpoint on a research-related topic, rather than a balanced review of this topic. The aim should be to stimulate debate or new research, cover controversial topics, or provide a new framework for, or interpretation of, an old problem or current issue, or speculate on the implications of some recent research.” They also note that you should not cite meta-analyses, and there is a reference limit.
- So the journal specifically requests non-balanced articles of only recent research, then puts out hyped press releases without saying this, frightening millions of people. To me, scientific “opinions” should be based on a dispassionate analysis of the whole relevant evidence-base, and this is simply irresponsible, and it should change.
- Regarding observational studies that purportedly show a relationship between artificial sweeteners and weight gain, Swithers cites 2 studies (Fowler et al., 2008, and Laska et al., 2012).
- It is notable that the former made no statistical adjustments for any dietary variables (artificial sweeteners are often associated with other dietary factors in other studies), and in fact the authors write that “artificial sweeteners use might therefore simply be a marker for individuals already on weight-gain trajectories… This is the most obvious possible explanation of our findings.”
- In the second study, total energy intake was statistically adjusted for, and it found that diet soda was cross-sectionally associated with increased BMI in males and females but not longitudinally (Swithers notes this), which is a stronger design. So this suggests that reverse causality is especially a concern in artificial sweetener research.
- On the other hand, there are stronger study designs not cited such as Mozaffarian et al. (2011) that analyzed data from 3 large prospective cohorts of a total of 120,877 people and included statistical adjustments for many dietary variables and found a small but statistically significant reduction in weight associated with increased diet soda consumption. As another example, de Koning et al. (2011) also included statistical adjustments for many dietary variables (including diet quality) and analyzed separately by if the participants had lost or gained weight. There was an association between greater weight loss and higher artificially sweetened beverage intake, but also one with greater weight gain and higher intake. Are these artefacts or are there subgroup differences? Further research is needed, but overall the data don’t suggest that artificial sweeteners will lead to weight gain, at least across the board, especially when interventional trials are considered too.
- For a good review of the artificial sweetener/weight research, see Anderson et al. (2012). They conclude “there is no evidence that low-calorie sweeteners can be claimed to be a cause of higher body weights in adults.“
- Next Swithers discusses observational studies associating artificial sweeteners with metabolic syndrome. She cites 4 studies in support (Duffey et al. (2012), Dhingra et al. (2007), Lutsey et al. (2008), and Nettleton et al. (2009).
- The first I blogged about and I would not say it shows an association. They found that people with a better diet quality who didn’t drink diet beverages had lower metabolic syndrome incidence compared to those of a worse diet quality who drank diet beverages. But people of poor diet quality who drank diet beverages had the same metabolic syndrome incidence as those with a poor diet quality and did not drink diet beverages. And when analyzed by individual metabolic syndrome components, there was no consistency. This was a well-designed study because it looked at differences by diet quality, where many do not.
- The other 3, which I mentioned in this post, did find positive associations between diet soda and metabolic syndrome, but all of the papers caution about interpretation and suggest that diet soda could be a marker of poor diet/lifestyle habits and/or confounding by unknown variables could explain the results. In addition, the 3rd (Lutsey et al. (2008)) warns about reverse causation as diabetics consume 3 times more diet soda per day than non-diabetics (Mackenzie et al. (2006)), so it may be that many of these associations are the result of people with new disease diagnoses switching to diet soda rather than diet soda promoting disease. Some but not all studies attempt to test for this.
- In the next section of observational studies association artificial sweeteners with diabetes, Swithers cites the aforementioned de Koning et al. (2011) study as evidence of a positive association, but it actually found no association with after full multivariate adjustment (though it is listed as such in Table 1), and the authors support artificially-sweetenened in place of sugar-sweetened beverages in the discussion.
- In another of the papers Swithers cites (Bhupathiraju et al., (2012)), the study involved 2 cohorts, and she only highlighted the one with a positive association (Nurses’ Health Study) in her article and ignored the negative one (Health Professionals Follow-Up Study), and the positive one was only significant after full multivariate adjustment for artificially sweetened caffeine-free/carbonated beverages and not artificially sweetened caffeinated/carbonated beverages (not noted in the article text, but noted in Table 1).
- Swithers also cites Romaguera et al. (2013) as supporting a positive association between at least 1 artificially sweetened beverage per day and diabetes, but there was no overall relationship with full statistical adjustments. Swithers then notes that in participants of normal weight at baseline there was a positive association, however, she doesn’t note that there was no relationship in overweight and obese groups in the text (but this is noted in Table 1).
- The last study she cites (Fagherazzi et al., 2013) does show a positive association.
- Finally, 4 studies are cited that purportedly associate artificial sweeteners with hypertension and cardiovascular disease (Fung et al., (2009), de Koning et al., (2012), Cohen et al., (2012), and Gardener et al., (2012)).
- For the first, Swithers writes that 2 artificially sweetened beverages per day are associated with coronary heart disease “in age-adjusted models“, but doesn’t note that there is no longer an association when fully adjusted for.
- The second study similarly does not show any association after all statistical adjustments are included (it is when only age is adjusted for) but it is not described like this in the text. I blogged about that one here. These are both noted in Table 1 as showing no association with cardiovascular disease, however.
- The third does support a positive association, and indeed in 3 large cohorts. But we don’t have good biological plausibility for why this is- the papers I have read struggle to reference a good mechanistic explanation.
- The last one I also blogged about here, which paradoxically found that sugar-sweetened beverages do not increase heart attack and stroke risk but artificially-sweetened beverages do. This study has a small number of participants and therefore a small number of cardiovascular events in each group and therefore I am not sure it is appropriate to draw conclusions without more studies.
Human Intervention Studies
- Swithers specifically notes in the section looking at interventional studies that only studies within the past 5 years were looked at (2 of them).
- The first discussed is by de Ruyter et al. (2012) and was a double-blind, randomized, 18-month trial in children who were assigned 1 diet beverage or 1 sugar-sweetened beverage per day. The diet beverage group did not gain as much weight as the sugar beverage group at 18 months, but as Swithers notes, there was no control/water group, and she seems to dismiss it. However, it still suggests that children should at least be encouraged to switch to diet drinks from sugary drinks if they won’t go with water, and this was a relatively long trial.
- The second study (CHOICE; Tate et al. (2012)) is another randomized trial in people who substituted some of their sugar-sweetened beverage calories for either (4 servings/day were provided) 1) water, 2) artificially sweetened drinks, or 3) a non-intervention group (“attentional control”) which received similar attention time as the other groups but received general weight-loss counseling instead of being given beverages to substitute in their diet. The study is described by Swithers like so: “overweight and obese adults who substituted water or [artificially sweetened beverages] for [sugar sweetened beverages] lost no more weight at 6 months than an attentional control group“. This is true; all groups reduced their calorie intake and weight loss was similar in all groups. However, subjects in the diet drink group but not the water group had an increased chance (twice as likely) of achieving a 5% weight loss than the attentional control group did, and the authors say this was the result of better adherence, but this was not noted with Swithers. This is a very important result and suggests that switching people from sugary drinks to artificially sweetened ones rather than to plain water may improve the chances they reach a 5% weight loss, which a is clinically important target. Additionally, another paper from just published from the same cohort (CHOICE, (Piernas et al. (2013))) analyzed diet patterns in the diet beverage group to see if it increased preferences for sweet-tasting foods and drinks. There were no differences between the diet and water groups for macronutrients or specifically total or added sugar. Interestingly, the dessert intake in the diet drink group was lower than the water group.
- Swithers does not cite a meta-analysis (because the journal requirements wouldn’t let her anyway) by de la Hunty and colleagues (2006) of short-term human trials examining the substitution of sugar with aspartame/other sweeteners on energy intake which suggests that total calorie intake is reduced when artificial sweeteners are used in place of sugar. More recent humans studies are discussed here, and overall tend to suggest that when used in place of sugar, artificial sweeteners slightly reduce total calorie intake.
When all evidence is considered, it is a much more nuanced story. In my opinion, the evidence is weighted toward artificial sweeteners being fine. I would recommend the scientific statement by the American Heart Association and the American Diabetes Association published last year for a good review of human research on the subject. We do need more long-term human studies on metabolic outcomes and weight, but it seems like substituting sugar with artificial sweeteners does generally result in a small reduction in calorie intake.