Was the causal effect between the sugar consumption and diabetes prevalence confirmed?

Please make yourself comfortable, this also will be long.

This particular article aims to comment on the effort of Dr Lustig to convince the audience that he has proven the causative link between the diabetes incidence and sugar/fructose consumption. It is not aimed to contradict the found link.

Dr Lustig proudly presented the study in which he participated and its design at 31:34 minute: 

He explained why it is so great and what they have examined. It looks good so far. They even mentioned the oils in the food supply, which is a surprise after the previous attempts based on the EPIC InterActive analysis that failed to mention this at all. 

He also announced they estimated that sugar explained 25% diabetes cases worldwide:

Similarly, he said that 29% of diabetic prevalence in the U.S. is explained by sugar and sugar alone, irrespectively of its calories and irrespective of its effect on weight. My critical mind is asking: what about the remaining 75 or 71%? This was not explained. 

After reading the original study I can tell you that they did a hell lot of analysis and statistical tests. I am not going into depth of the details. What I would like to say is that there were so many limitations, mainly about the quality of the data from all named sources, that even the team ended their discussion with this: 

...any of the findings we observe here are meant to be exploratory in nature, helping us to detect broad population patterns that deserve further testing through prospective longitudinal cohort studies in international settings, which are only now coming underway.

Therefore, although they found statistical significance in the link between the sugar AVAILABILITY (not consumption) and the prevalence of diabetes, independently from the BMI or total calories availability, even when adjusted for other potential confounders (including fat), this was just an extensive statistical analysis of the data of a limited quality, despite allegedly being the best they had at that time. 

Moreover, the basic rule, which is taught in the first epidemiology lecture, is that the causation effect cannot be obtained from the observational data and based on the associations as this particular study was. So what causation effect was Dr Lustig talking about? 

Oh yes, the Medical Causation Inference, or causation inference in short. What do they say about it? 

Causal inference is the process of drawing a conclusion about a causal connection based on the conditions of the occurrence of an effect. The main difference between causal inference and inference of association is that the former analyzes the response of the effect variable when the cause is changed. The science of why things occur is called etiology: the study of causation, or origination.

From this it emerges that all he did (and he also said it) was that he expanded on the single snapshot of the EPIC study correlations with the 10 years long temporal relationship (still based on associations) that the changes of diabetes prevalence followed the previous changes in the availability of sugar and not the other way round. Nothing else. This may have captured the causative effect or it may have not. 

Here is more information about what the inference actually is. So even if they statistically found a seemingly causative effect, it came with the discussed limitations and the nature of the data (observational) and therefore it is not a solid proof about the true causative effect. And, as they said: this is only the beginning and the hypothesis outlined here needs further testing by longitudinal prospective cohort studies. 

Although I said I was not going into the details of the study, I would like to highlight one point of the announced Bradford Hill Criteria, which was not included in the four points Dr Lustig mentioned. The strength. Look at the diagram:  

Despite the link between the sugar availability and the prevalence of diabetes among all these countries was found to be statistically significant, the strength of the correlation coefficient looks quite low. The direction of the scatter plot seems to have very small angle against the x-axis, but this was calculated by the computer, so let’s accept it. The low strength here does not mean the causation is non-existent, but the higher coefficient means that the causation is more likely, according to the Bradford Hill criteria on the Wikipedia. This was just to highlight another weak point in the conclusion for the causal effect of sugar consumption on the prevalence of diabetes. 

Many scientists often stick to the statistical significance and push this finding through, despite the correlation coefficient is often found small. And, just for the comparison: try to recall the critique of Dr Lustig towards the scatter plot of Ancel Keys. He had an apparently stronger association between the fat intake and coronary heart disease mortality, did he not? Yet that much stronger coefficient was named as VERY MINOR by Dr Lustig. How would you then evaluate the one above? Moreover, Dr Lustig criticized few outliers in the Ancel Keys study, but here you can also find a few.  

One thing came to my mind: the links between other factors (dietary, lifestyle, economic) with the diabetes were found either statistically insignificant or non-existent. For those that showed some link, albeit insignificant (there is a difference between the statistical significance and clinical relevance), I would be curious to find out whether some of these also would come out positive in the causation inference test. This was not done.

And finally: at the 36th minute you could hear Dr Lustig saying that "this is proximate cause". And here you can learn the difference between the proximate and ultimate cause. The proximate cause that the ship is sinking is the hole in its body below the waterline, but the ultimate cause was that it hit a rock that made the hole. And, as Wikipedia says:

"Separating proximate from ultimate causation frequently leads to better understandings of the events and systems concerned."

So we now have to find the ultimate causation, since we already have the proximate one (or we think we have it). How? 

We now know that their statistical analysis found other individual factors (dietary, lifestyle, economic) as insignificant or non determining in the relationship with the diabetes prevalence. Maybe those 71 and 75 unexplained percents are needed to be explained? What factors do they refer to? Maybe a cluster of various factors that are naturally present with the actual sugar consumption (not average availability per capita). Examining each of these other factors separately might not give the objective result, because in reality we consume composite foods within a certain lifestyle and they interact in the etiology of the disease. Where some may not come up significant on their own, they may help other factor becoming significant, or these coupled together in real living conditions, especially when they commonly occur together in the society with increasing prevalence of diabetes, or any other metabolic disease. And now even Dr Lustig knows that a 'donut diet' (sugar and fat) coupled with the lack of physical activity and positive energy balance is detrimental to our health. Yet their analysis points only at the sugar and sugar alone, but only within about the quarter of the diabetes cases prevalence.

Finally, later during the lecture, at 56:20 you could hear: 

"90% of what we know in medicine today is causal medical inference. Only 10% is scientific proof."

Well, I did not know that, but OK. As I said at the beginning, this article was written to comment on the claim that we have a proof about the causal influence of sugar only on the diabetes prevalence, independently of the factors that were already discussed. I have no problem with accepting it if it was correct but there is still a space for doubt and that was my point. Unless we will do a proper scientific study (which we will not do in this case for obvious reasons also outlined by Dr Lustig), we can only say: this is the best we have and it looks solid. Not that it IS a definitive evidence. I still have a sort of feeling that dietary fats consumed along sugars in excess do have a significant impact on the development of insulin resistance and also other metabolic disruptions. 

Remember that liver is the major organ for energy metabolism and it not only metabolizes fructose but the dietary fats as well. 

Now I would like to cite my own work based on similar analysis of dietary patterns worldwide, performed by Popkin et al  2012: 

"...the worldwide trends, the increase in obesity, MetS, CVD and other metabolic illnesses between the years 1962 and 2000 came along increased estimated per capita availability of calories from sweeteners by 74 kcal/d but the TEI rose by 403 kcal/d, indicating that the consumption of other energy rich nutrient had also increased."

And the study presented by Dr Lustig also said: 

"The impact of sugar on diabetes was independent of sedentary behavior and alcohol use, and the effect was modified but not confounded by obesity or overweight."

Leaving behind the questionable quality of the data mentioned earlier, these two quotations need an explanation in a specific context. The Popkin et al study was related to the developing countries. However, as you could see in one of my other articles, in the US the trend of sugar consumption also did not particularly correlate with the ever increasing rise of metabolic diseases for the past 14 years. There certainly was other dietary factor or factors that made this trends worse and it not only could have modified these trends but actually contribute to them. Just because one study did not find a statistical significance it does not mean that a different study performed by different scientists and maybe based on better quality data would not find it. It is quite common that studies do not always come to the same conclusion or they even report the opposite results. 

And that was my point. I did not disprove the findings of this study because I do not see behind all of it. They might have found a true cause-effect relationship, or may not. We cannot be sure for sure because of the limitations associated with it. But my 'feeling' does not let me to leave out the substantially increased fats and oils consumption along the sugar in the world, and the actual drop of sugar consumption in the U.S. while the metabolic diseases, including diabetes, have been increasing. I agree though, that there are groups of people who consume way too much sugar and these are far more likely to develop diabetes. However, their diet is not only rich in sugar but also rich in processed fats and poor in essential nutrients and as Dr Lustig said, this also was not examined in their study. 

After watching one of the older videos from 2012, Dr Lustig announced that there were some sensitive slides not to be published before the main work was published. These were not presented to the audience and were cut off the video. However, from the Q&A at the end of the video at 40:00 minute I have realized that this secret information was related to this econometric study. What I have learned from this video, Dr Lustig announced that they did not have the data available to evaluate the role of glycaemic index of the diet and its possible role in the diabetes development. He also said that the category of roots, pulses, nuts, etc, which have a low glycaemic load did not show a protective effect in the MODEL and that he was surprised, or rather shocked, as he said. Well, perhaps something was wrong with the data? Can we trust the final outcomes then? I leave it up to you. Only future will show the truth. 

At 41:44 minute you can hear the answer on my suggested question earlier in this article: the team had no answer for the remaining 71 or 75% of the causes for diabetes. They simply did not know. He mentioned the genetics... but nothing about dietary fat or the combination of fat with sugar as is for the DONUT or fast food outlets... The remaining three quarters of the 'causes' of type 2 diabetes remained a complete mystery and he tried to make the audience happy with those 25% as pretty good. Something smells here...