普段ならば質問と返答の一往復で終わる筈なのだが、この時は、なぜか、論文著書のreplyの後に、わざわざもうひとつのletterが追加された。勿論、論文に対して否定的な内容なのだが、問題点が網羅的に指摘されていて、当該論文をどう評価すべきかについての総まとめといった性格を持っている。
言わんとするところを「忖度」しつつ内容を紹介すると、
＊ 質問票自体、かなり精度低いものが用いられている。このタイプの質問票が広く評価されている根拠として筆者が挙げている論文は、他の調査方法と比較した上で、むしろ、このタイプの質問票の精度が低いことを明らかにしたものである。
＊ 精度の低い質問票であっても、定期的に繰り返し調査されているのならば、参考にならないこともない。しかし、このケースでは、調査開始時に一度だけ、しかも、去年１年間の食生活を思い出しながら記入するという方法がとられているため、その後の変化が全く反映されない上に、記入された内容自体の信用度も低い。
＊ 被験者の年齢を考慮すると、追跡期間中に定年退職を迎えている筈である。定年退職の前と後で食生活が大きく変化することは調査を通じて既に明らかにされているのだが、この点についても考慮した形跡がない。
＊ 加糖飲料のデータから得られた相対危険度は、年齢や体重などの条件で補正していっても大きな違いが生じないが、人口甘味料を用いた飲み物のデータから得られた相対危険度は、補正を加える毎に大きく変化している。このことは、人口甘味料を用いた飲み物以外の要因が相対危険度に寄与していることを強く示唆する。
＊ 数多くの調査によって、果汁100%のジュースであっても、摂取量が多ければ糖尿病リスクが上昇することが知られている。しかるに、この論文では、全くリスクが上昇しないことになっている。これは、あまりにも不自然だ。
＊ 常識的には、やはり、太っている人や血糖値が高い人が人口甘味料を用いた飲み物を多く飲むようになると考えるのが妥当だろうが、なぜか著書は、あれこれ理由を挙げて、この考え方を退けようとしている。
＊ 生理学的側面の考察のくだりで、文献を挙げつつ、人口甘味料にインスリン分泌を増加させる効果があるのではないか、などと、途方もないことを言い始めるのだが、そこで挙げられている論文は、インスリンを増加させる効果は無さそうだという内容のもので、ろくに内容を読まずに参考文献に挙げていることが疑われる。
＊ また、空腹感を強める可能性についても言及しているが、この点については既に調査が行われていて、否定的な結果が出ている。
もう充分だろう。調査の方法や結果の解釈にこれだけ問題点を抱えていたのでは、そもそも「論文」として成り立っていない。学術的な価値はゼロである。
しかし、このような応酬が、学術誌の誌上で、誰の目にも見えるような形で行われたことには、大きな意味があると言えるのではないだろうか？
専門家がどのように調査を行い、その結果をどのように分析し解釈するのかという点は、門外漢には、少々掴み難いところがある。だが、このような形でのやり取りを見ることで、少しずつ理解を深めていくことは可能だ。これは、おそらく、専門家を志す若い人にも当てはまるだろう。
･･･というようなことを考えながらこのletterを読み終えようとしていたのだが、締めくくりの文の含意がすぐには理解出来ず、一瞬とまどってしまった。
しばらくの沈思黙考ののち、ようやく、これが「トドメの一撃」だということに気がついた(←鈍い！)
言わんとするところを「忖度」すると、以下のような意味になるだろうか？
「もし、人口甘味料は体に悪いという結論ありきで、膨大のデータから都合のいいところを探して、切り取って、論文の形に仕上げたのだとしたら、調査終了から論文の発表までの間が不自然に長かったり、方法論や解釈に歪みが生じたりするのは当然だけど、君のこの論文はどうなの？」
Consumption of artificially and sugar-sweetened beverages and incident type 2 diabetes: methodologic concern about a recent epidemiological study
The American Journal of Clinical Nutrition, Volume 98, Issue 1, July 2013, Pages 251–252, https://doi.org/10.3945/ajcn.113.064386
Dear Sir:
In a recent issue of the Journal, Fagherazzi et al (1) revived the controversy about the safety of intense sweeteners. It has been repeatedly proven that increased sugar intake leads to weight gain (2) and that consumption of sugar-sweetened beverages (SSBs) is significantly associated with an increased risk of type 2 diabetes (T2D; reviews in reference 3). From the data on the E3N French cohort of 66,118 menopausal women, monitored between 1994 and 2007, Fagherazzi et al extend this risk to the consumption of artificially sweetened beverages (ASBs). However, several methodologic weaknesses cast doubt on the validity of their conclusions.
First, it is unlikely that a single diet-history questionnaire from 1993 on eating and drinking habits during the previous year was sufficiently precise to quantify the amount of different beverages consumed during the following 14 y of medical consultations. The reference the authors provided to justify their survey methodology (4) shows a limited reproducibility of the data provided by such a questionnaire, as indicated by low correlation of coefficients between data of 2 diet-history questionnaires recorded at 1-y intervals. This reference also indicates a very poor relative validity because of the systematic differences between data provided by the questionnaires and the data of 12 repeated 24-h recalls. In addition, in the E3N questionnaire, the beverage recording was only partial because consumption at the traditional 2 main meals in France, lunch and dinner, was not taken into consideration.
Another weakness identified was that the questionnaire only allowed respondents to answer either “only SSB consumption,” or “only ASB consumption,” without the possibility of indicating the respective quantity of both beverages that a large proportion of subjects probably consumed. This is confusing and may have introduced a significant distortion in the respective quantification of the SSB and ASB intakes. Furthermore, one diet-history questionnaire carried out at the beginning of the 14 y cannot be said to be sufficient to identify changes in the respondents’ dietary habits. Changes over a significant period of time may have been important to note, considering the commonly broadcast advice of reducing energy intake, particularly sugar intake, by public health stakeholders in France during the follow-up period. In addition, the age and the professional activity of the participants may have favored significant changes. All were female national education employees, aged 52.6 ± 6.6 y at baseline, who stopped their professional activity during the follow-up period. Different French national surveys indicate strong changes in life and eating habits at this age (5), and body weight frequently increases during this period in women (6).
In this way, it is detrimental that Fagherazzi et al’s (1) article did not provide any follow-up of the participants’ body weight, especially because they reported twice as much obesity among ASB consumers (6.8%) as in SSB consumers (3.1%). Taking into consideration the results, it must be noted that the relation between sweetener consumption and diabetes was largely reduced after adjustment for confounding variables (model 1), especially after adjustment for energy intake (model 2) and even further reduced after adjustment for BMI (model 3), with a disappearance of the risk for the middle levels of ASB consumption. It is surprising that the risk of T2D associated with ASB intake followed a U-shaped curve, which was significant only in women in the first quartile (<1 soda every 2 wk) and women in the fourth quartile (weekly consumption of >2 portions) but not for consumers between these levels. Unfortunately, Fagherazzi et al did not give any explanation for this surprising observation, which made hazardous the quadratic spline regression model of risk of T2D presented in their Figure 1. Another tentative conclusion is the argued absence of association of T2D with the consumption of fruit juices. As indicated in the article’s Table 1, after adjustment for the confounding variables including body weight, the OR of diabetes appearance increased significantly for consumers who drank 3–4 portions of fruit juice (448–967 mL) on a weekly basis, in accordance with other works (7).
Finally, an association between the consumption of intense sweeteners and T2D does not indicate a causal relation. As underlined by Elfhag et al (8), a reverse relation seems more probable because overweight or obese persons with prediabetic conditions tend to favor ASBs to SSBs to reduce their sugar consumption. However, by discussing the causal mechanism of T2D induced by ASBs, the authors suggest that intense sweeteners may promote insulin secretion. This claim is supported by a reference from Anton et al (9). Unfortunately, this work showed the contrary, ie, that sparing caloric intake by substituting sugar with intense sweeteners significantly reduced postprandial glucose and insulin concentrations, with no compensation in eating more during the following meals. Numerous other studies confirmed that intense sweeteners do not increase hunger, consumption, or insulin secretion (reviewed in reference 10).
In conclusion, it is possible that methodologic weakness may explain the surprising observation reported by Fagherazzi et al (1) as well as the significant delay between the end of the survey (June 2007) and the publication of the findings (January 2013).
The author, as an expert in physiology and nutrition, is a member of a scientific-reflection group appointed by the French section of the International Sweetener Association.
