For discussion

Sugar tax: plausibility and value of the studies

The authors of a study [1] that extrapolated the potential effects of various tax models, including a tiered manufacturer's levy on sugary drinks, concluded that such a tax would have some effects in Germany between 2023 and 2043 (see box: "Expected Effects"). Professional societies such as the German Diabetes Society (DGS) are calling for the introduction of such a tax, claiming that its benefits have been proven. This study evaluates the plausibility and value of the observational studies used for the optimistic calculations in [1].

Non-controversial issues
>> There is no doubt:
• The availability of energy-dense foods is essential for increasing body weight in an affluent society.
• Dietary habits and risk of disease are influenced by many factors, such as depression, medication, psychological stress, shift work, poverty, education and other factors.
• When calculating the relationship between sugar consumption, weight, type 2 diabetes and associated diseases, epidemiological studies cannot distinguish between those who are ill “with” or “because of” the disease.
• It is undisputed that in the case of illness many measurable parameters (e.g. LDL cholesterol, blood sugar, weight) are altered and can be influenced by diet and exercise.

Controversial issues
It is disputed whether
• the effects extrapolated by simulation, as in study [1], actually occur and are clinically relevant;
• the “cardiometabolic effect” assumed by the authors actually exists or is merely a laboratory finding that correlates with disease;
• there is a pathophysiological basis due to which a change in sugar intake can have effects on cells and vessels;
• government measures are justified if studies justifying the sugar tax are selectively chosen and higher-quality studies, a Cochrane review and the experiences of others are not taken into account.
Why one should make the statements
must take a position
If the authors’ statements [1] were correct, the following consequences would arise:
1. The detection of disease risks and early stages of disease would fall within the purview of observational epidemiology and policy-making at the population level. This would only be justifiable if the evidence suggests a relevant health and economic effect (apart from government tax revenue).
2. It is a matter of balancing the rights of the individual against the rights of the state to intervene.
General background
How is type 2 diabetes mellitus defined?
The definition uses a laboratory-measurable threshold for glucose, the glucose increase in the exercise test, and the HbA1c value. These values were set so low that, when the threshold is reached, diabetic complications occur only extremely rarely [2–7]. These parameters were chosen based on the assumption that normalizing them would prevent diabetes-associated diseases. If this assumption were correct, diabetes-associated diseases could be prevented by normalizing blood glucose levels. However, this only applies to diabetic coma, but not to all other diabetic complications. The reasons for this are the many different subgroups that can be differentiated using classical markers [7, 8] and pathophysiological criteria, e.g., DNA damage [9, Diabetologie und Stoffwechsel 2022; 17(05): 349–352], and the knowledge of the need for multifactorial therapy in type 2 diabetes [11–14].
Type 2 diabetes prevention and
-Early detection – status of
Intervention studies
1. Can diabetes prevention prevent subsequent complications?
Many studies have shown that comprehensive lifestyle interventions can reduce the incidence of diabetes by approximately 50% in individuals at risk [15-39].

1. a) Diabetes Prevention Program and its Outcome Study (DPPOS) [16, 28, 29]:
3,234 subjects with impaired glucose tolerance received either 850 mg of metformin twice daily, an intensive lifestyle intervention, or a placebo for 3 years, followed by an 18-month, slightly less intensive lifestyle intervention. The effect on cardiovascular disease was examined. The results after 21 years:
Neither metformin nor the lifestyle intervention reduced the incidence of cardiovascular events (non-fatal myocardial infarction, stroke, and death) during follow-up (up to 21 years), despite demonstrable effects on the incidence of diabetes, blood glucose levels, smoking cessation, and weight loss.

1. b) Finish National Diabetes Prevention Program (FIN-D2D) [30-37]:
This study of 10,149 subjects at high risk for type 2 diabetes resulted in a reduction in diabetes risk after a median follow-up of 7.4 years. Despite diabetes prevention, which was still evident after 13 years [33], weight loss, and improvement in cardiometabolic parameters, this intervention study did not show a reduction in the incidence of cardiovascular events or death from any cause (36–37).

1. c) Da Quing study [35, 36, 40, 41]:
This study included only 438 adults with impaired glucose tolerance in the intervention group and 138 in the control group (n = 138) [35]. The intervention led to a reduction in the incidence of type 2 diabetes by 51% and by 43% in the follow-up period (20 years) [35]. However, after 20 years [35], no significant effect of the intervention on the occurrence of a cardiovascular event or death was detectable. After 23 years, the result was only significant for women. After 30 years, it is reported [36, 40] that the intervention resulted in fewer cardiovascular events (HR 0.74, CI 0.59–0.92), fewer microvascular complications (HR 0.65, CI 0.45–0.95), fewer deaths (HR 0.74, CI 0.61–0.89), and a mean life extension of 1.44 years (CI 0.2–2.68) [36].
There are several reasons to doubt the reliability of the data from the Da Quing study:
• The small number of cases: In the 30-year evaluation [36], only 59 of 138 subjects in the control group and 220 of 438 subjects in the intervention group were still alive.
• The 95% confidence interval ranges from no effect to a good effect of prevention.
• The groups studied were not comparable. Starting in the second year (Figure 3C in 40), the curves for cardiovascular death diverged.
• The incidence rate of the Da Quing study is not comparable with the other two studies and European experience.

2. Can earlier diagnosis and treatment of type 2 diabetes prevent diabetes-associated diseases?
In a 10-year endpoint study involving 16,047 patients, no effect of early diagnosis and treatment of diabetes mellitus could be observed after 10 years [42].
The lack of clinical significance of cardiometabolic markers in
Example of the long-term blood sugar value HbA1c
1. Can a reduction in long-term blood glucose (HbA1c) be used to determine clinical efficacy?
It was claimed that the UKPDS study would show that each further reduction in HbA1c of 1% reduces the risk of diabetes-related disease by 2% [43]. Since the UKPDS study was never designed to do this, calculations were necessary to reach this conclusion. A closer look at Figure 2 of this study [43] reveals something striking: The difference in the incidence of myocardial infarction or microvascular complications between an HbA1c of 7% and 8.5%, i.e., the clinically relevant range, is less than 10% in each case, and not—as claimed—2%.
For the sugar tax thesis, the difference in disease incidence per 1,000 patient-years, i.e., between an HbA1c of 5.5% (normal) and 6.5% (slightly elevated), is relevant. It is less than 5%. Therefore, the effect assumed in (1) is not to be expected. This example and others [44-47] show that translating observational data into statements supposedly proving causality, such as "this leads to...", leads to false conclusions.

2. Does lowering HbA1c prevent disease in non-diabetic patients?
There are no good intervention studies to support this assumption. The Women's Health Study, a 10-year prospective cohort study, investigated whether a slightly higher HbA1c level in non-diabetic women predicts cardiovascular disease and the onset of diabetes [48]. After multivariate analysis (smoking, ethnicity, hypertension, BMI, etc.), the increased risk for future diabetes persisted, but there was no evidence of a prediction of cardiovascular events. One reason for this is the considerable differences between individuals: at a given glucose level, some produce significantly more HbA1c than others [49]. Thus, 21% of the differences in HbA1c levels are not explained by glucose levels (“glucose gap”).

3. Can an intensified lifestyle program reduce the incidence of cardiovascular disease in type 2 diabetics?
The Look-AHEAD study investigated whether a more profound lifestyle intervention than a tax that only slightly reduces sugar consumption has effects on disease outcomes (and not just laboratory values). [50-54] Despite achieving effects on parameters predicting cardiometabolic risk, there were no effects on cardiovascular endpoints.
Effects of reducing the
Sugar consumption in
Intervention studies
1. Intervention studies with a reduction in sugar consumption
In the study by C. B. Ebbeling [55, 56], 224 overweight to obese adolescents received a one-year intervention to reduce their intake of sugar-sweetened beverages. A reduction in sugar-sweetened beverage intake was observed in the intervention group, which remained approximately 50% even in the second year.
In the first year of the intervention, the intervention group gained 1.6 kg, while the control group gained 3.5 kg. The difference was 1.9 kg and decreased to 0.8 kg after the second year (although the intervention group continued to consume significantly fewer sweetened beverages). Because the adolescents were still growing, the BMI difference is important: After one year, it was 0.57, and after two years, only 0.3 in favor of the intervention group, and was not statistically significant. The subgroup analysis showed that only those of Hispanic background showed a significant effect of -8.8 kg after two years and a BMI decrease of -2.35. Since the effect of sugar reduction diminishes over time, a sugar tax must be assumed to be a failure.
In addition, the study participants seem to look for alternative calorie sources after some time: When normal-weight children are examined over 18 months [57], the children in the sugar-free
The group experienced a slight weight loss during the first six months. However, after six months, weight gain resumed, almost parallel to the control group. At the end, the groups differed by only slightly more than 1 kg.
One explanation is the genetically determined relationship between obesity and the intake of sweetened beverages [58]. If sweetened beverages are eliminated, the calorie requirement is met in other ways. Overall, the study situation is inconclusive [57-68], with effects either undetectable or so small that they cannot be clinically relevant. This does not change when meta-analyses such as that by Malik in Am J Clin Nutr 2013, 98:1084-1102 calculate that the withdrawal of sweetened beverages has a stronger effect than school-based health education. The small effect and a 25% dropout rate [69] in studies make the success of sugar-
containing beverages is unlikely. Nevertheless, in some reviews, the conclusion (“Reducing sugar intake in sweetened beverages is effective”) does not agree with the previously presented results (“We found low-certainty evidence”) [69].
A similar approach is taken in an expert statement (doi: 10.2903/j.efsa.2022.7074). After reviewing the literature, the working group comes to the conclusion: "Based on a systematic review of the literature, prospective cohort studies do not support a positive relationship between the intake of dietary sugars, in isocaloric exchange with other macronutrients, and any of the chronic metabolic diseases or pregnancy-related endpoints assessed. ... The level of certainty is moderate for obesity and dyslipidaemia (> 50-75% probability), low for non-alcoholic fatty liver disease and type 2 diabetes (> 15–50% probability) and very low for hypertension (0–15% probability)...A level of sugar intake at which the risk of dental caries/chronic metabolic diseases is not increased could not be identified….”
The correct conclusion would be: It
Given the low level of evidence, nothing has been proven or disproven.

2. The success of a dietary change to a Mediterranean diet
Sugar consumption is also reduced in the Mediterranean diet. The best-known intervention study investigating this is the Predimed study [70, 71] with 7,447 participants. Cardiovascular events occurred in 4.4% of the control group, 3.8% in the olive oil group, and 3.4% in the nut group. Comparing the best with the worst group, the difference is 1.1% with a confidence interval of 0.54–0.95 in the nut group and 0.53–0.91 in the olive oil group. The absolute numbers of 109 events in the control group, 96 in the olive oil group, and 83 in the nut group cast doubt on whether the group differences (see also size of the confidence interval) after the diet could be significant or even clinically relevant without extrapolation to many patient years.
Pathophysiological considerations
1. The increase in glucose and the importance of food composition
• Interindividual differences in the rise in blood values after eating are very pronounced in healthy individuals [72]: The difference for triglycerides is 10.3%, glucose 6.8%, and insulin 5.9%. For example, the composition of the food components explains only 3.6% of the increase in fat after eating. Given the large intraindividual differences in the rise in sugar levels after a meal and the minor importance of food composition for fluctuations in levels [72], a sugar tax with only a small reduction in daily sugar consumption (in England, only 15 grams per household per week) will not have clinically relevant effects.
• Increased glucose metabolism (glucose flux). A brief and slight increase in glucose after eating is no explanation for cell damage. Biochemical studies show, even if they do not differentiate between the isoforms of hexokinase [73, 74], that glucose flux is already saturated at physiological glucose concentrations. Consequently, a brief and slight increase in glucose through the consumption of sugary drinks cannot lead to an increased glucose flux and cellular activation in either a healthy person or a diabetic. This also explains why, in cell experiments, glucose effects only occur at glucose concentrations that are far above the "normal" increase in glucose after eating. Thus, the theory of positive "cardiometabolic" effects of a tax lacks a biochemical basis.
• How can the different results of the modeling and the intervention studies be explained? The crucial difference is the study methodology. The studies that calculate a health-relevant success of reducing sugar consumption [1, 75–80] are observational studies and registry studies. The studies that argue against the health success of a sugar tax are intervention studies (for an understanding of the advantages and disadvantages of observational studies compared to randomized intervention studies, see also https://ww.sokrates-rationalisten-forum.de/erkenntnistheoretische-grundlagen-und-ihre-beachtung-in-der-heutigen-wissenschaft). If the quality of the conduct is the same, intervention studies are of higher quality because they form the study groups randomly. Especially for diseases with many different causes, the study groups are thus more comparable. Confounding variables can be eliminated. No study can control for the variety of influencing factors, which are composed of various medical, psychosocial, and economic factors. No study team has the expertise and resources to simultaneously assess all risk factors from the fields of medicine, sociology, psychology, and economics. Even if these many factors could be captured, the group size would be too small to make a statistically reliable conclusion. Therefore, the methodology used (observational studies in Study 1 versus intervention studies in this review) is crucial for the different conclusions.

If the authors of [1] had incorporated the results of higher-quality intervention studies into their modeling instead of the “success reports” from observational studies, the result would have been completely different. If randomized intervention studies [81] do not
Although we cannot prove an effect, there is much to suggest that it does not actually exist, because intervention studies are more likely to overestimate an effect than to underestimate it.

Below are some examples of results from intervention studies that contradict observational studies:
• In men with low testosterone, the TRAVERSE study [82, 83] failed to prevent diabetes or bone fractures. Although observational studies have found that low vitamin D3 levels correlate with cardiovascular disease and cancer, intervention studies in children and adolescents found no effect of vitamin D3 supplementation on the risk of bone fracture [84] or protection against cancer and cardiovascular disease. A study that examined the difference between low-molecular-weight and unfractionated heparin is instructive for the importance of the value of the study design. The more rigorous the study criteria and study quality, the smaller or even nonexistent the benefit of low-molecular-weight heparin [85].
• If a confounding variable is not taken into account, the application of sophisticated mathematical models such as microsimulation [1] creates the risk of misestimating an effect. The more basic data are processed, the more input errors, for example due to a lack of comparability between groups, become noticeable in the quality of the output due to multiplication in the calculation process [81, 86, 87]. In the case of multifactorial diseases, i.e. diseases caused not only by medically accessible factors (such as cardiovascular diseases), the risk of distorting the results due to group differences in observational studies cannot be denied and leads to incorrect conclusions. The microsimulation model applied for Germany in (1) is based on the result of English data (UK IMPACT NCD), which are just as little based on intervention studies as the assumptions regarding BMI-independent effects of sugar consumption on diabetes and cardiovascular diseases [67, 75-78, 80, 88-98].
If the value of the study designs is ignored
The reader of this study [1] and similar studies [75–80, 88–99] is led to believe that essential health data and risk factors were included in the calculation. This is not true, because other, probably even more important factors that also influence dietary behavior were not taken into account. These include many psychosocial factors, such as education, optimism, loneliness, finding meaning in life, and much more [100–110]. Therefore, it is conceivable that the "cardiometabolic" data included by the authors [1] in the calculation have no clinical significance, but are merely epiphenomena.
This is demonstrated by a randomized intervention study on psychosocial intervention with the endpoints of weight and HbA1c [109]. Unlike the above-cited studies on reducing sugar intake, diabetes prevention, and the Mediterranean diet, the social intervention showed relevant effects on weight and HbA1c.
Experiences with the
Introduction of a sugar tax
The goal of taxing sugar or sweetened beverages is to prevent clinically relevant diseases. Therefore, the effect of this tax must be measured by the following outcomes: reducing the number of people with a BMI above 25 and above 30, and reducing the number of people with high blood pressure, diabetes, heart attacks, and strokes.
But in England [110], the effect of a tax on soft drink manufacturers on children's weight loss was disappointing. Among girls, the reduction in obesity prevalence was only 1.6% with a confidence interval of 1.1 to 2.1 (!), and among boys, it was undetectable. The control group is important in the study: Since there was no control group after the introduction of the tax, the increase in obesity before the introduction of the tax was extrapolated to the period afterward.
This results in:
1. No one lost weight, but the weight gain was just less than expected.
2. The calculated expectation is falsely high, because in Germany the increase in obesity during this period without tax was much lower.

A study that assumes that a reduction in sugar intake of 15 grams per week per household would have relevant effects has been widely misinterpreted. This study does not provide any measurements for any diseases, but only extrapolations that assume 15 grams per week per household (not per person) and use this to calculate an expected weight loss.
In a systematic review and meta-analysis published in JAMA in 2022, Andreyeva and colleagues only came to the following conclusion: “In this systematic review and meta-analysis of implemented SSB taxes worldwide, SSB taxes were associated with higher prices and lower sales of taxed beverages.” Teng and colleagues reported in Obesity Reviews 2019: “Based on real-world evaluations, SSB taxes introduced in jurisdictions around the world appear to have been effective in reducing SSB purchases and dietary intake.” So it's just about buying and consuming sugar, with no evidence of clinically relevant (!) effects! This also corresponds to the experiences from Mexico, South Africa, and other countries.
Therefore, Thow and colleagues come to the correct conclusion in their review in Nutrition Reviews 2014: “Future research should use prospective study methods to determine the effect of taxes on diets and focus on the effect of taxation in conjunction with other interventions as part of a multisectoral strategy to improve diets and health.”
How unclear and open the issue of a health-promoting effect of any kind of sugar tax still remains is evident in a Cochrane review from 2020 (doi.org/10-1002/14651858.CD012333.pub2). After analyzing all studies available at the time, the review concluded: "The available evidence on the review question can be considered very limited, and the certainty of the evidence is very low."
The responsibility of the authors
The authors [1] failed to live up to their responsibility on several levels:
Level 1: The one-sided selection of “suitable” observational studies and lack of plausibility checks based on the higher-quality intervention studies.
Level 2: Readers are not informed that cardiometabolic markers are only markers and that their treatment through lifestyle interventions cannot yet be proven with clinical endpoints.
Level 3: There is no long-term study that demonstrates a long-term effect of sugar reduction on body weight and other diseases.
Level 4: Given the weakness of the observational studies and the lack of conclusive studies, Emmert-Fees and co-authors wrote quite correctly in the abstract of their publication [1] under “Conclusions”: “In this study we found, that SSB taxation in Germany could help to reduce the national burden of noncommunicable diseases …” The subjunctive “could” makes the difference between obvious overinterpretation and honest, albeit overly optimistic, interpretation.
Level 5: The authors of the study [1] explain
in PLoS: “No tax on SSBs is currently enacted in Germany and the national government is preparing a new national strategy on food seeking evidence-based recommendations to establish policy priorities until 2050.” It remains unclear whether this is a paid report (to obtain new research funding) or a truly “free” research effort.
Level 6: If a measure extends life, then increased care costs, medical costs and pension payments must be taken into account, as well as possible side effects (increase in anorexia and
of orthorexia among young people who want to live a particularly health-conscious life). <
from:
Prof. Dr. med. Dr. hc Peter Paul Nawroth
(ORCID 0000-0002-6134-7804)
Varun Kumar PhD
(ORCID 0000-0003-2827-1024)

Literature available online at www.m-vf.de:
https://t1p.de/n543u