いわゆる発展途上国18ヶ国に住む35～70才の男女135,335名を対象に、食生活に関するアンケート調査を行い7.4年間の追跡調査の結果、「果物、野菜、豆は、ほどほどの量つまり3～4人前（約375～500g）食べると死亡リスクが最も低く、もっと多く食べても更なる恩恵は殆どない」、同様に、「高脂肪食（総エネルギーの約35％）は低脂肪食より死亡リスクが低かった。しかし、高炭水化物食（総エネルギーの60％以上）は、心血管疾患リスクはないものの死亡率が高い」という内容の２つの研究論文（PURE1，PURE2）が、カナダMcMaster大学／Hamilton Health Sciencesから発表されランセット誌に掲載されました。
「この結果を踏まえ、世界的な食事ガイドラインを再検討すべきである」と研究者は提言していますが、これぞまさしく前回の記事 “ごまかし科学” で述べたspin／バイアス（偏り）以外の何物でもありません。
この研究は“prospective cohort study（前向きコホート研究）と呼ばれる観察研究であり、観察研究では要因と結果との相関の強さは定量的に測定できますが、因果関係を証明することはできません。
統計学／科学では “Correlation does not imply causation” という言葉が使われます。つまり、“相関関係は因果関係を含意しない”と云う意味で、相関関係があるだけでは因果関係があるとは断定できず、因果関係の前提に過ぎません。
米国Brigham and Women's HospitalのMichelle L. O'Donoghueも、2017年10月11日付け Medscapeで、この研究論文の問題を次のように厳しく指摘しています。
Hello. This is Dr Michelle O'Donoghue, reporting for Medscape. Today I want to issue an urgent plea to science journalists to please stop publishing stories about all these observational studies that have created a significant amount of confusion in the nutritional literature.
It seems that every day on the news, we hear a conflicting report about whether or not a particular diet strategy or food group is beneficial. The researchers may have the best of intentions, but the findings are rarely reliable for a variety of reasons. As we know, diet records are quite subject to recall bias—namely, people either cannot remember or may slightly misrepresent what they have been eating. More important, it is almost impossible to disentangle the large number of confounding factors, such as income and education, that heavily influence what we eat and are known to be linked to adverse outcomes.
Some argue that with careful multivariable adjustment, it is possible to account for these potential confounders in their analyses. Unfortunately, we have learned time and time again that the results of observational studies cannot be heavily relied upon.
For example, for a long time hormone replacement therapy (HRT) was believed to be beneficial for women. This belief was based on a large number of very well-conducted observational studies that suggested that HRT was associated with improved cardiovascular risk.
I cite this as an example simply to highlight that time and again, we have been fooled by observational results, and these should not be laying the groundwork for the evidence base that we use to make clinical treatment decisions.
PURE: Fats Good, Carbs Bad?
Most recently, at the European Society of Cardiology annual meeting, there was a great deal of interest in and conversation about the results of the PURE study. The investigators concluded that high intake of carbohydrates was associated with a higher risk for death, whereas increased intake of fats, in particular saturated fats, was associated with a lower risk for death. This came as quite a surprise, so of course it had people talking.
“Why does it seem that, when the results of an observational study support what we want to hear, people are also more willing to embrace the findings?”
First, the study's strength: PURE was a huge study. More than 100,000 patients were enrolled in this study, and they were all followed for more than 7 years.
Although I greatly respect these researchers, they themselves acknowledge several limitations to the study design. Most important, they were unable to distinguish the type of carbohydrates a person was eating, and thus a carbohydrate from whole grains was considered the same as one from heavily processed foods or soft drinks. I think we would all agree that not all carbohydrates are created equal, so it becomes nearly impossible to group all carbohydrates together into a single type of food group.
Because it was a multinational study, they observed that carbohydrate intake was actually highest in low- and middle-income countries. They tried to adjust for these types of differences in their analyses, but nonetheless, we know that access to healthcare and risk for death are profoundly different throughout the world and in different countries, and many of these factors that affect outcome are quite unrelated to diet.
Despite these limitations, the study was published in a very high-profile journal and garnered a lot of media attention. Some have even suggested that this study should form the basis for entirely new guideline recommendations. But why should the results of this single observational study carry more weight than other observational studies that have been published? Why does it seem that, when the results of an observational study support what we want to hear, people are also more willing to embrace the findings? This may be a larger observational study, which makes it seem more compelling, but we always have to accept that there are limitations. The confidence intervals may be tighter around a confounded result, but nonetheless, can we really rely on the findings themselves?
Guidelines and RCTs
In my opinion, our guideline recommendations should be based purely on randomized data, which is the most reliable and rigorous. But therein lies the rub: We lack randomized controlled trials in the nutritional study arena. I hope more will be forthcoming in the near future.
Unfortunately, we also lack funding for such types of randomized trials. Many of the studies that are being conducted are supported by food groups and boards that have different interests in mind in terms of promoting their own products. In fact, many of those studies create enough conflicting reports in the literature that they lead to more inaction about nutrition. Every time we turn on the news, we hear a different conflicting report, and when we hear different conflicting reports, it leads people to believe that nothing we can do will change our cardiovascular risk in a meaningful way and that diet does not really matter.
I would love to have a call for more rigorous randomized trials in the nutritional literature. Until we have such studies, however, I would be very hesitant to say that any guideline recommendations should be based on observational studies alone. Again, this is not a criticism of any of the researchers. I believe these dietary questions have been approached in as rigorous a way as possible, but nonetheless, we need to recognize the limitations of observational data and realize that they could lead us astray, as they have in the past.
PURE study challenges the definition of a healthy diet: but key questions remain