肥胖不是你吃太多？」兩大學派激辯病因真相，打破熱量迷思

本翻譯僅作學術交流用，無商業意圖，請勿轉載，如有疑議問請來信

雖然「少吃多動」仍是主流觀點，但《歐洲臨床營養期刊》最新評論強調，肥胖可能源於高碳水飲食引發的胰島素失調。研究呼籲科學界應跳脫「熱量進出」思維，認真檢視飲食組成對代謝的深層影響。

Competing paradigms of obesity pathogenesis: energy balance versus carbohydrate-insulin models

肥胖病發病機制的競爭範式：能量平衡與碳水化合物-胰島素模型

Ludwig DS, Apovian CM, Aronne LJ, et al. Competing paradigms of obesity pathogenesis: energy balance versus carbohydrate-insulin models. Eur J Clin Nutr. 2022;76(9):1209-1221. doi:10.1038/s41430-022-01179-2

https://pubmed.ncbi.nlm.nih.gov/35896818/

摘要 Abstract

儘管持續的公共衛生運動旨在減少能量攝入（“少吃”）和增加能量消耗（“多運動”），但肥胖疫情仍然持續不斷。這一失敗的解釋之一是，基於能量平衡概念的當前方法並未得到公眾的充分接受。另一種可能性是這一方法基於一個錯誤的範式。能量平衡模型（EBM）的新公式，與之前的版本類似，將過度飲食（能量攝入 > 消耗）視為肥胖的主要原因，並強調“複雜的內分泌、代謝和神經系統信號”在無意識層面上控制食物攝入。該模型將肥胖率上升歸因於便宜、方便、能量密集的“超加工”食品，這些食品富含脂肪和糖。另一種觀點，即碳水化合物-胰島素模型（CIM），提出對高度加工碳水化合物的激素反應會將能量分配轉向脂肪組織的沉積，從而使身體的代謝需求可用的卡路里減少。因此，增加的脂肪量導致過度飲食以補償被隔離的卡路里。 在這裡，我們強調能量平衡模型（EBM）和碳水化合物-胰島素模型（CIM）在肥胖病理生理學方面的強烈對比，並考慮 EBM 中妨礙範式測試和改進的缺陷。糾正這些缺陷應該優先考慮，因為需要進行建設性的範式衝突來解決長期存在的科學爭議，並為設計新的模型以指導預防和治療提供信息。然而，公共衛生行動不必等待這場辯論的解決，因為這兩種模型都將加工碳水化合物視為肥胖的主要驅動因素。

The obesity pandemic continues unabated despite a persistent public health campaign to decrease energy intake (“eat less”) and increase energy expenditure (“move more”). One explanation for this failure is that the current approach, based on the notion of energy balance, has not been adequately embraced by the public. Another possibility is that this approach rests on an erroneous paradigm. A new formulation of the energy balance model (EBM), like prior versions, considers overeating (energy intake > expenditure) the primary cause of obesity, incorporating an emphasis on “complex endocrine, metabolic, and nervous system signals” that control food intake below conscious level. This model attributes rising obesity prevalence to inexpensive, convenient, energy-dense, “ultra-processed” foods high in fat and sugar. An alternative view, the carbohydrate-insulin model (CIM), proposes that hormonal responses to highly processed carbohydrates shift energy partitioning toward deposition in adipose tissue, leaving fewer calories available for the body’s metabolic needs. Thus, increasing adiposity causes overeating to compensate for the sequestered calories. Here, we highlight robust contrasts in how the EBM and CIM view obesity pathophysiology and consider deficiencies in the EBM that impede paradigm testing and refinement. Rectifying these deficiencies should assume priority, as a constructive paradigm clash is needed to resolve long-standing scientific controversies and inform the design of new models to guide prevention and treatment. Nevertheless, public health action need not await resolution of this debate, as both models target processed carbohydrates as major drivers of obesity.

“競爭範式的支持者在不同的世界中從事他們的工作……兩者都在觀察這個世界，而他們所觀察的並沒有改變。但在某些領域，他們看到的事物不同，並且彼此之間的關係也不同。這就是為什麼一條甚至無法向一組科學家證明的定律，有時對另一組科學家來說似乎直觀上是顯而易見的。” – T. Kuhn, 1970 [ 1].

“[P]roponents of competing paradigms practice their trades in different worlds … Both are looking at the world, and what they look at has not changed. But in some areas they see different things, and they see them in different relations one to the other. That is why a law that cannot even be demonstrated to one group of scientists may occasionally seem intuitively obvious to another.” – T. Kuhn, 1970 [1].

新的能量平衡模型——專注於食物攝入
The new energy balance model—a focus on food intake

這兩種肥胖模型有一個共同特徵：假定對一個關鍵生理參數的穩態調節，以促進最佳功能 [10, 11]。在能量平衡模型中，體重（或體脂肪）是被調節的變量，這一可能性在某種程度上得到了進化的支持：在食物匱乏的時期，適當的體脂肪對生存是必要的，而過多的脂肪可能會增加被捕食的風險。

Both models of obesity share a common feature: presumed homeostatic regulation of a critical physiological parameter to promote optimal functioning [10, 11]. In the EBM, body weight (or body fat) is the regulated variable, a possibility with some evolutionary support: whereas adequate body fat is needed for survival during times of food scarcity, excessive fat might increase risk of predation.

Hall 等人提出的新能量平衡模型 [9] 認為，大腦通過“複雜的內分泌、代謝和神經系統信號來控制食物攝入，以調節體重，這些信號根據身體的動態能量需求以及環境影響而起作用。”這一控制系統集中於“獎勵、食慾和感官處理”，涉及“顯著性、渴望和動機，這些主要在我們的意識下運作。”肥胖的結果是“各種便宜、方便、能量密集的超加工食品的可獲得性和市場營銷增加，這些食品的份量大、脂肪和糖含量高，而蛋白質和纖維含量低。”這些暴露導致過度進食，能量過剩被儲存為體脂肪。

The new EBM of Hall et al. [9] proposes that the brain controls food intake to regulate body weight through “complex endocrine, metabolic, and nervous system signals acting in response to the body’s dynamic energy needs as well as environmental influences.” This control system centers on “reward, appetite, [and] sensory processing” involving “salience, wanting, and motivation that primarily operate below our conscious awareness.” Obesity results from “increased availability and marketing of a wide variety of inexpensive, convenient, energy-dense, ultra-processed foods that are high in portion size, fat, and sugar, and low in protein and fiber.” These exposures cause overeating, with the energy excess deposited into body fat.

早期的公式特徵性地將能量平衡的兩個組成部分視為一體 [ 12, 13, 14, 15, 16, 17]。霍爾等人的能量平衡模型 [ 9] 與這些公式不同，主要集中於食物攝入的控制，而對能量消耗的關注較少。這個新的能量平衡模型暗示，並且相關的評論明確指出 [ 2, 18, 19, 20]，在這個模型中，所有卡路里在代謝上是相似的。例如，霍爾和郭 [ 19] 斷言，“就實際而言，當涉及到體脂肪和在碳水化合物與脂肪比例不同的控制等熱量飲食之間的能量消耗差異時，‘一卡路里就是一卡路里’。”雖然承認飲食成分會影響各自宏量營養素的氧化速率，但能量平衡模型認為，飲食最終通過增加總能量攝入來驅動脂肪沉積，而不是通過對底物分配的獨立於卡路里的影響。

Earlier formulations characteristically considered both components of energy balance in concert [12,13,14,15,16,17]. The EBM of Hall et al. [9] differs from these formulations, with a primary focus on the control of food intake and less attention to energy expenditure. This new EBM implies, and related reviews explicitly state [2, 18,19,20], that all calories are metabolically alike in the model. For instance, Hall and Guo [19] assert that, “for all practical purposes, ‘a calorie is a calorie’ when it comes to body fat and energy expenditure differences between controlled isocaloric diets varying in the ratio of carbohydrate to fat.” While acknowledging that dietary composition influences oxidation rates of respective macronutrients, the EBM holds that diet ultimately drives fat deposition by increasing total energy intake, not through calorie-independent effects on substrate partitioning.

碳水化合物-胰島素模型——代謝範式的一個特例
The carbohydrate-insulin model—a special case of the metabolic paradigm

CIM 代表一種對立的範式，其起源於二十世紀初 [ 7, 21, 22, 23, 24, 25, 26, 27, 28]，認為血液中的代謝燃料供應（作為燃料氧化的代理）是被調節的參數。雖然足夠的體脂可能有助於在饑荒期間生存，但考慮到所有組織，尤其是大腦，對持續燃料供應的依賴，獲得代謝燃料是立即生存所必需的 [ 28, 29, 30]。

The CIM represents an opposing paradigm, with origins in the early twentieth century [7, 21,22,23,24,25,26,27,28], that considers the supply of metabolic fuels in the blood (as proxy for fuel oxidation) the regulated parameter. Whereas adequate body fat may aid survival during famine, access to metabolic fuels is required for immediate survival, in view of the dependency of all tissues, and especially the brain, on a continuous fuel supply [28,29,30].

CIM [ 8, 31, 32] 提出高升糖負荷 (GL) 飲食——即含有大量快速可消化碳水化合物（即自由糖、加工穀物、大多數澱粉類蔬菜）——會引發激素反應，抑制脂肪動員（脂解作用）並促進脂肪在脂肪組織中的沉積。正如最近詳細說明的 [ 8]，攝取高 GL 餐會產生高比例的胰島素與胰高血糖素分泌，以及 GIP 與 GLP-1 分泌。這種高度合成代謝的激素特徵使底物分配向沉積轉移，導致可用於代謝活躍組織（包括大腦）的能量減少，特別是在餐後晚期 [ 33, 34]。大腦對這種代謝狀態的反應是啟動控制飢餓和其他食慾反應的途徑 [ 35, 36] 以促進能量攝取。如果個體通過限制食物來抵抗進食的驅動，則通過減少能量消耗來保存代謝燃料，表現為疲勞（導致久坐行為）、減少非運動性活動熱產生、增加肌肉效率及其他機制。 在沒有超過大多數人能夠維持的卡路里限制的情況下，脂肪積累是因為能量持續分配到脂肪組織。因此，碳水化合物-胰島素模型解釋了卡路里限制飲食效果不佳的原因，超出了因享樂和獎勵影響而缺乏遵守的因素。

The CIM [8, 31, 32] proposes that a high-glycemic load (GL) diet—one with large amounts of rapidly digestible carbohydrates (i.e., free sugar, processed grains, most starchy vegetables)—elicits hormonal responses that inhibit fat mobilization (lipolysis) and promote fat deposition in adipose tissue. As recently detailed [8], consumption of a high-GL meal produces a high ratio of insulin to glucagon secretion, and of GIP to GLP-1 secretion. This highly anabolic hormonal profile shifts substrate partitioning toward deposition, leaving less energy available for metabolically active tissue including the brain, especially in the late postprandial period [33, 34]. The brain responds to this metabolic state by activating pathways controlling hunger and other appetitive responses [35, 36] to promote energy intake. If an individual resists the drive to eat by restricting food, metabolic fuels are conserved through reduced energy expenditure manifesting as fatigue (leading to sedentary behavior), decreased non-exercise activity thermogenesis, increased muscular efficiency, and other mechanisms. Without a degree of calorie restriction beyond most people’s ability to sustain, fat accumulation results because of continued partitioning of energy into adipose tissue. Thus, the CIM offers an explanation for the poor efficacy of calorie-restricted diets beyond lack of adherence due to hedonic and reward influences.

除了 GL 之外，CIM 提供了一個概念框架，以理解其他飲食因素、行為和環境暴露如何通過代謝機制影響體重，而不是對能量攝入或消耗的主要影響；這些因素包括果糖[37, 38, 39, 40]、蛋白質量[41]、脂肪酸類型、纖維、餐內食物順序[42]、用餐時間[43]、身體活動以及內分泌干擾的食品添加劑和污染物[44, 45]。CIM 還假設了一種飲食-表型互動，即內源性胰島素分泌高、葡萄糖穩態失調以及對胰島素介導的脂肪細胞脂解抑制高度敏感的個體，特別容易受到高 GL 飲食的不利代謝影響，這可能解釋了對以宏量營養素為重點的減重飲食反應的顯著異質性[46, 47, 48]。

In addition to GL, the CIM provides a conceptual framework for understanding how other dietary factors, behaviors and environmental exposures may affect body weight through metabolic mechanisms rather than primary effects on energy intake or expenditure; these include fructose [37,38,39,40], protein amount [41], fatty acid type, fiber, food order within a meal [42], meal timing [43], physical activity and endocrine-disrupting food additives and pollutants [44, 45]. The CIM also postulates a diet-phenotype interaction, such that individuals with high endogenous insulin secretion, disorders in glucose homeostasis, and high sensitivity to insulin-mediated suppression of adipocyte lipolysis would be especially susceptible to the adverse metabolic effects of a high-GL diet, potentially explaining some of the marked heterogeneity in response to macronutrient-focused weight loss diets [46,47,48].

這種病理生理學的觀點與常見肥胖形式的發展相符。微小的基質分配偏向脂肪儲存將解釋緩慢但逐漸的體重增加，直到脂肪組織的胰島素抵抗發展到足夠的程度。脂肪組織的胰島素抵抗將抵消高升糖負荷飲食的過量胰島素分泌，導致體重停滯，但代價是異位脂質沉積和全身代謝功能障礙，這與脂肪組織擴展性假說一致[49]。這一擴展的公式提供了詳細的機制和眾多可檢驗的假設，以指導研究[8]。

This view of pathophysiology accords with the development of common forms of obesity. A small shift of substrate partitioning favoring fat storage would account for slow but progressive weight gain, until adipose tissue insulin resistance develops to a sufficient degree. Adipose tissue insulin resistance would counterbalance the excessive insulin secretion of a high-GL diet, resulting in a weight plateau, but at the cost of ectopic lipid deposition and systemic metabolic dysfunction, consistent with the adipose tissue expandability hypothesis [49]. This expanded formulation provides detailed mechanisms and numerous testable hypotheses to inform research [8].

與這兩種模型相關的證據
Evidence pertaining to the two models

肥胖的自然過程通常需要數年到數十年發展，涉及平均每天過量儲存約 1 到 2 克脂肪——這對於短期的代謝餵養研究（即≤2 週）來說，實在太小而無法測量。雖然這一效應在較長期的門診試驗和觀察性研究中可能可觀察到，但由於對測試飲食的遵守不佳和混雜因素，從這些數據中進行因果推斷的能力可能受到限制。此外，鮮有研究專注於兒童，這是一個肥胖發展的動態階段[50]。雖然動物研究可以闡明機制，但其向人類的轉化仍然存在問題。基於這些原因，關於肥胖病因的廣泛文獻可以被選擇性引用以支持相反的觀點，因為這場辯論的每一方都聲稱對方的觀點。

The natural course of obesity, which usually develops over years to decades, involves excessive storage of ∼1 to 2 g fat/d on average—far too small to measure in short-term metabolic feeding studies (i.e., ≤2 weeks). Whereas this effect could be observable in longer-term outpatient trials and observational studies, causal inference from these data may be limited by poor adherence to test diets and confounding. Furthermore, few studies have focused on childhood, a dynamic stage of obesity development [50]. Although animal studies can elucidate mechanisms, their translation to humans remains problematic. For these reasons, the vast literature on obesity pathogenesis can be selectively cited to make opposing points, as each side of this debate has claimed of the other.

在本節中，我們並不旨在提供文獻的全面回顧，而是強調與 Hall 等人[9]的主要分歧，考慮到研究設計的限制。表 1 總結了區分這些模型的關鍵特徵，以便於進行評估。先前的評論提供了對[12, 13, 14, 15, 16, 51, 52, 53, 54, 55, 56, 57, 58]和反對[6, 31, 32, 59, 60, 61, 62, 63, 64, 65, 66, 67]早期版本的 EBM 的多種觀點。

In this section, we do not aim to provide a comprehensive review of the literature, but rather highlight main disagreements with Hall et al. [9], considering study design limitations. Table 1 summarizes key features distinguishing the models to facilitate this assessment. Prior reviews offer a range of perspectives for [12,13,14,15,16, 51,52,53,54,55,56,57,58] and against [6, 31, 32, 59,60,61,62,63,64,65,66,67] earlier versions of the EBM.

儘管啮齒類動物和人類並未進化出相同的飲食，但實驗動物研究在這場辯論中被考慮在內。Hall 等人[9]提出的證據反對碳水化合物-胰島素模型（CIM）是觀察到 70%碳水化合物和 10%脂肪的飲食能保護啮齒類動物免於肥胖，而 20%碳水化合物和 60%脂肪的飲食在某些實驗條件下會產生最多的體重增加。同樣，最近一項針對 5 種小鼠品系的研究得出結論，增加飲食中的脂肪，而非碳水化合物或蛋白質，與食物攝入量和體重的變化更大相關[68]。然而，Tordoff 和 Ellis[69]發現，啮齒類動物的飲食中碳水化合物和脂肪的能量比例相等時，最具肥胖性，無論朝任何方向的偏差都會減少體重增加。進一步增加這種異質性，Kennedy 等人[70]得出結論，對小鼠而言，極低碳水化合物飲食（蛋白質含量較低）“誘導出一種與減重相符的獨特代謝狀態”。 顯然，考慮到近親繁殖品系的特異性、未控制的飲食暴露所造成的混淆以及啮齒動物和人類不同的營養需求，這項研究必須謹慎地推廣到人類身上[ 71, 72, 73, 74]。例如，飽和脂肪和糖通常佔據高脂啮齒動物飲食中的大部分熱量，這種組合會導致下丘腦發炎和全身性胰島素抵抗[ 75, 76, 77, 78, 79, 80, 81, 82]。

Although rodents and humans have not evolved to eat the same diets, experimental animal research has been considered in this debate. Hall et al. [9] present as evidence against the CIM the observation that a diet of 70% carbohydrate and 10% fat protects rodents from obesity and one with 20% carbohydrate and 60% fat produces the most weight gain in some experimental conditions. Similarly, a recent study with 5 mouse strains concluded that increasing dietary fat, but not carbohydrate or protein, was associated with greater variations in food intake and body weight [68]. However, Tordoff and Ellis [69] found that rodent diets with equal amounts (by energy) of carbohydrate and fat were most obesogenic and deviations in either direction reduced weight gain. Adding to this heterogeneity, Kennedy et al. [70] concluded that a very-low-carbohydrate diet (with lower protein content) in mice “induces a unique metabolic state congruous with weight loss”. Clearly, this research must be extrapolated to humans with caution, in view of well described limitations involving idiosyncrasies of inbred strains, confounding from uncontrolled dietary exposures and dissimilar nutrition requirements of rodents and humans [71,72,73,74]. For instance, saturated fat and sugar often comprise most calories on high-fat rodent diets, a combination that causes hypothalamic inflammation and systemic insulin resistance [75,76,77,78,79,80,81,82].

這些方法論問題可以通過直接檢查因果方向來避免。雖然對於大宗營養素的激素反應可能因物種之間的進化差異而異，但影響脂肪儲存的生物機制是高度保守的，增強了將啮齒動物研究轉化為人類的潛力。在能量平衡模型中，飲食通過增加食物攝入來驅動脂肪沉積。因此，當在肥胖飲食下的動物與在等熱量對照飲食下的同窩兄弟進行配對餵養，確保相同的能量攝入時，對身體組成的影響應該是相同的。
These methodological issues can be avoided by direct examination of causal direction. Whereas hormonal responses to macronutrients may differ among species due to evolutionarily divergence, biological mechanisms affecting fat storage are highly conserved, enhancing potential translation of rodent studies to humans [83,84,85]. In the EBM, diet drives fat deposition by increasing food consumption. Therefore, when animals on an obesogenic diet are pair-fed to littermates on an isocaloric control diet, ensuring the same energy intake, effects on body composition should be identical.

這一預測經常失敗。Petro 等人 [ 86] 將小鼠分為 58%與 11%脂肪飲食，持續 11 週，並觀察到高脂肪組的脂肪量更大（24.1%對 18.5%，P < 0.001），這與其他研究結果一致 [ 87, 88, 89]。在高糖飲食中也觀察到了類似的與卡路里無關的效果 [ 90, 91, 92, 93]。儘管有人可能會質疑這些數據的含義，認為啮齒類動物對這種代謝效應更敏感，但這一論點將削弱啮齒類宏量營養素研究在理解人類肥胖方面的有效性。

This prediction often fails. Petro et al. [86] pair-fed mice 58% vs 11% fat diets for 11 weeks and observed greater adiposity in the high-fat group (24.1 vs 18.5%, P < 0.001), consistent with other findings [87,88,89]. Similar calorie-independent effects have been observed with high-sugar diets [90,91,92,93]. Although one could challenge the implications of these data by arguing rodents are more susceptible to such metabolic effects, that argument would undermine the validity of rodent macronutrient studies for understanding human obesity in the first place.

研究血糖指數（GI）提供了另一種繞過物種特異性差異的宏量營養素代謝的方法。在一系列涉及幾種齧齒動物品系和物種的研究中，通過替換澱粉類型來檢查 GI 的影響，同時控制宏量營養素、飽和脂肪、糖和微量營養素[82, 94, 95, 96, 97]。這些研究顯示，在消耗高 GI 與低 GI 飲食的動物之間，出現以下變化，順序為：高胰島素血症、基質分配的轉變偏向脂肪沉積、能量消耗減少、脂肪增加和體重增加——所有這些都發生在能量攝入增加之前。當能量攝入受到限制以防止體重增加時，高 GI 組仍然出現異常的身體組成。儘管攝入的卡路里較少，但這些動物的體脂肪卻更多，犧牲了瘦體組織[96]。儘管多種機制（例如，腸道微生物組）可能介導這些影響，但它們與能量平衡模型（EBM）的基本前提相矛盾，即飲食成分對脂肪沉積沒有卡路里獨立的影響。

Studies of glycemic index (GI) offer another way to circumvent species-specific differences in macronutrient metabolism. In a line of investigation involving several rodent strains and species, the effects of GI were examined by substitution of starch type, controlling for macronutrients, saturated fat, sugar, and micronutrients [82, 94,95,96,97]. These studies demonstrate the following changes among animals consuming high- vs. low-GI diets, in this sequence: hyperinsulinemia, a shift in substrate partitioning favoring fat deposition, decreased energy expenditure, increased adiposity and weight gain – all prior to an increase in energy intake. When energy intake was restricted to prevent weight gain, the high-GI group still developed abnormal body composition. Despite consuming fewer calories, these animals had more body fat at the expense of lean body tissues [96]. Although multiple mechanisms (e.g., gut microbiome), may mediate these effects, they contradict a fundamental premise of the EBM, that diet composition has no calorie-independent effects on fat deposition.

最後，Hall 等人[9]將胰島素作用的研究視為不具區別性，但這些研究提供了另一個機會來直接測試模型預測。在碳水化合物-胰島素模型中，較高的胰島素分泌通過直接的外周機制促進脂肪儲存[8]。能量平衡模型則專注於激素的中央作用，似乎預測相反，考慮到胰島素在大腦中的食慾抑制作用[98, 99, 100, 101, 102]。這些涉及慢性胰島素給藥和減少胰島素分泌的基因模型的肥胖研究支持碳水化合物-胰島素模型[103, 104, 105, 106, 107, 108, 109]。淡化這些發現的重要性——即外周的與卡路里無關的作用主導中央的與卡路里有關的作用——有可能創造出一個過於一般化的能量平衡模型，以至於無法進行測試，特別是因為 Hall 等人[9]將人類鼻用注射後胰島素的中央食慾抑制效應解釋為反對碳水化合物-胰島素模型的證據。

Finally, Hall et al. [9] dismiss studies of insulin action as non-discriminating, but these provide another opportunity to test model predictions head-to-head. In the CIM, greater insulin secretion promotes fat storage through direct peripheral mechanisms [8]. The EBM, with its focus on the central actions of hormones, seems to predict the opposite, in view of the anorectic actions of insulin in the brain [98,99,100,101,102]. These studies of adiposity, involving chronic insulin administration and genetic models of reduced insulin secretion, support the CIM [103,104,105,106,107,108,109]. Downplaying the significance of these findings—that the peripheral calorie-independent actions dominate central calorie-dependent ones—risks creating an EBM so general as to be untestable, especially as Hall et al. [9] interpret the central anorectic effects of insulin following nasal injection in humans as evidence against the CIM.

雖然正如 Hall 等人所述「神經系統已進化以控制能量攝入」，但大腦也幾乎控制了所有代謝的各個方面，包括葡萄糖代謝，這在 1850 年代由克勞德·伯納德著名地描述過。事實上，飲食成分對身體組成的影響與碳水化合物-胰島素模型一致，這在肥胖的動物模型中普遍表現出來，如表 2 所示。當能量攝入限制在與對照組相同或更低的水平時，實驗模型中觀察到脂肪增加，這影響了許多被認為介導食物攝入的大腦通路，顯示出所謂的「飢餓」或「飽足」激素的外周代謝作用的存在。在這些模型中，過度的脂肪自發性地發展，而不需要增加食物攝入或體重。這些發現似乎與對人類基因研究的普遍解釋相矛盾，該解釋將大腦中肥胖相關多態性的更高流行率歸因於能量平衡模型。

While “nervous systems have evolved to control energy intake,” as Hall et al. [9] state, the brain also controls virtually all aspects of metabolism [110,111,112,113], including glucose metabolism, as famously described by Claude Bernard in the 1850s [114]. Indeed, effects of dietary composition on body composition consistent with the CIM manifest commonly among animal models of obesity, as exemplified in Table 2. With restriction of energy intake to levels at or below that of controls, increased adiposity has been observed in experimental models affecting numerous brain pathways thought to mediate food intake, demonstrating the existence of peripheral metabolic actions of putative “hunger” or “satiety” hormones. In some of these models, excessive adiposity spontaneously develops without increased food intake or body weight. These findings seem at odds with a common interpretation of human genetic studies that attributes the greater prevalence of obesity-related polymorphisms in the brain vs. adipocyte as evidence for the EBM.

顯然，遺傳因素影響人類肥胖風險，根據全基因組測序，BMI 的遺傳率估計為 30% [ 115]。這種遺傳率中只有一小部分可以通過約 290 個單核苷酸多態性的已知常見變異來解釋，而這些多態性的大多數生理後果仍然未知。在某些情況下（例如，MC4R），靠近已知引起單基因肥胖的基因的常見變異顯示了中樞神經系統的關鍵重要性 [ 116]——儘管這些並不排除與 CIM 一致的途徑（表 2）。一些相關基因在大腦中廣泛表達，其他基因則普遍表達（例如，FTO）。還有一些基因在大腦外的表達更為突出（例如，MSX1，TMEM18，SEC16B，ADCY3）。事實上，途徑分析顯示，對肥胖的遺傳易感性可能涉及“胰島素分泌/作用、能量代謝、脂質生物學和脂肪生成” [ 117]。

Clearly, genetic factors influence human obesity risk, with BMI heritability estimated at 30% based on whole genome sequencing [115]. Only a small component of this heritability can be explained by known common variation at ~290 single-nucleotide polymorphisms and the physiological consequences of most of these polymorphisms remain unknown. In some cases (e.g., MC4R), common variation near genes known to cause monogenic obesity illustrates the critical importance of the central nervous system [116]—although these do not exclude pathways consistent with the CIM (Table 2). Some implicated genes are expressed widely in the brain and others are ubiquitously expressed (e.g., FTO). Still others are more prominently expressed outside the brain (e.g. MSX1, TMEM18, SEC16B, ADCY3). Indeed, pathway analysis showed that genetic susceptibility to obesity can involve “insulin secretion/action, energy metabolism, lipid biology and adipogenesis” [117].

Hall 等人 [9] 引用的多態性作為反對碳水化合物-胰島素模型 (CIM) 的證據，仍然存在替代解釋。例如，ATGL 的純合突變導致脂解功能缺陷，但似乎並不增加肥胖的風險。然而，這種突變也會損害脂肪生成，導致不僅脂肪動員減少，脂肪儲存也減少。正如 Schreiber 等人 [118] 所總結的，“脂質分解與合成的相互依賴為 ATGL 缺乏的小鼠和人類缺乏肥胖提供了合理的解釋。”而 FTO 基因的等位基因與食慾或食物攝入相關，這一觀察並未提供有關代謝途徑或因果方向的信息。

For polymorphisms cited by Hall et al. [9] as evidence against the CIM, alternative interpretations remain viable. Homozygous mutations in ATGL, for instance, resulting in defective lipolysis do not appear to increase risk for obesity. However, this mutation also impairs lipogenesis, resulting in not only less fat mobilization, but also less fat storage. As Schreiber et al. [118] conclude, “Interdependence of lipid catabolism and synthesis provides a rational explanation for the lack of obesity in ATGL-deficient mice and humans.” Whereas alleles of the FTO gene are associated with appetite or food intake, this observation provides no information regarding metabolic pathways or causal direction.

因此，基因學研究表明涉及肥胖的途徑在大腦內部和外部運作；在許多情況下，這些似乎與 CIM 一致。總的來說，基因表達數據並未明確區分這兩種模型，因為大腦在控制食物攝入和能量代謝方面的作用，以及身體與大腦之間通過神經、代謝和激素信號的交流。

Thus, the genetics studies indicate pathways involving obesity that operate within and outside the brain; in many cases, these appear consistent with the CIM. Altogether, genetic expression data do not definitively differentiate between the two models, in view of the brain’s role in controlling both food intake and energy metabolism and the communication between the body and brain through neural, metabolic, and hormonal signals.

儘管設計限制使得無法通過觀察性研究直接測試能量平衡模型（EBM）與碳水化合物-胰島素模型（CIM）之間的因果機制，但如果謹慎解讀，這些研究仍然可以提供有用的信息。Hall 等人[9]指出，“沒有證據表明碳水化合物攝入解釋了國家之間體重差異”，但這些生態比較對於像體重這樣的變量價值不大。例如，碳水化合物攝入量高的國家往往貧窮，且相當比例的人口營養不良、缺乏營養，並從事自給農業。此外，Hall 等人忽視了長期且豐富的觀察歷史，這些觀察將常見慢性疾病的出現（包括肥胖）與通常包括增加消耗高度精製穀物、糖和含糖飲料的人口營養轉變聯繫起來[119, 120]。在美國，1970 年至 2000 年間，BMI 增長最快，這段期間也恰逢精製穀物、糖和總碳水化合物消耗的顯著增加[121, 122]。然而，這些世俗趨勢可能受到身體活動和其他相關行為變化的干擾。

Although design limitations preclude a direct test of causal mechanisms in the EBM vs CIM with observational research, these studies can still be informative if interpreted with the necessary caution. Hall et al. [9] state that “evidence to suggest that carbohydrate intake explains between-country differences in body weight is nonexistent”, but these ecological comparisons are of little value for a variable like body weight. Countries with high carbohydrate intake, for instance, tend to be poor, with a substantial proportion of the population undernourished, malnourished, and engaged in subsistence agriculture. Moreover, Hall et al. disregard a long and rich history of observations linking the emergence of common chronic disorders, obesity among them, to population-wide nutrition transitions that typically include increased consumption of highly refined grains, sugar, and sugary beverages [119, 120]. In the USA, BMI increased most rapidly from 1970 to 2000, also concurrent with marked increases in consumption of refined grains, sugar, and total carbohydrate [121, 122]. These secular trends, though, may be confounded by changes in physical activity and other relevant behaviors.

前瞻性隊列研究提供了更大的能力來控制混雜因素，特別是包括社會經濟地位，儘管可能仍然存在殘餘混雜。此外，體重和其他脂肪量的測量特別容易受到反向因果關係的影響（人們因體重增加或肥胖而改變飲食的傾向，而不是因為飲食改變導致體重增加或肥胖）。此外，典型的前瞻性設計比較基線飲食與未來體重變化，將無法檢測在飲食評估時已達到穩態的先前變化。在這種情況下，可能會出現偏向於零關聯的偏差；因此，涉及 GI 和 GL 的隊列研究中缺乏一致的關聯難以解釋[123]。為了更好地模擬干預研究，可以檢查飲食變化與體重隨時間變化之間的關係。在這些分析中，精製穀物、馬鈴薯產品和含糖飲料的攝入量較高——這些是 GL 的主要貢獻者——在經過廣泛調整潛在混雜的飲食和生活方式因素後，與三個大型隊列中的體重增加有關[124]。 （這些研究中，紅肉和加工肉類也與體重增加有關。）

Prospective cohort studies provide greater ability to control for confounding factors, notably including socioeconomic status, although residual confounding may remain. In addition, body weight and other measures of adiposity are especially susceptible to reverse causation (the tendency for people to change their diets as a result, rather than a cause, of weight gain or obesity). Furthermore, the typical prospective design comparing baseline diet with future weight change will not detect prior changes that have reached steady state by the time of the dietary assessment. In this situation, bias toward null associations may ensue; thus, the lack of consistent association involving GI and GL in cohort studies is difficult to interpret [123]. To better simulate an interventional study, the relationship of change in diet to change in weight over time can be examined. In such analyses, higher intakes of refined grains, potato products, and sugar-sweetened beverages—the main contributors to GL—were associated with greater weight gain in three large cohorts after extensive adjustment for potentially confounding dietary and lifestyle factors [124]. (Red and processed meats were also associated with greater weight gain in these studies.).

Hall 等人 [9] 結論認為，流行病學數據“與能量平衡模型一致，表明存在多種潛在的飲食驅動因素導致過量熱量攝入……”然而，Mozaffarian [125] 提出了對這一概念化的新問題，至少在目前美國肥胖流行病的階段是如此。根據全國代表性調查，Mozaffarian 指出，自 2000 年以來，能量攝入已經趨於平穩或下降，儘管肥胖率仍在上升，身體活動卻適度增加。（由於女性腰圍的增長不成比例，根據 BMI 評估的肥胖趨勢可能低估了自 2000 年以來流行病的進展程度 [126]）。他認為，這些趨勢呼籲考慮替代的因果解釋，包括涉及代謝功能障礙的解釋。

Hall et al. [9] conclude that the epidemiological data, “consistent with the EBM, suggest a variety of potential dietary drivers of excess calorie intake…” However, Mozaffarian [125] raises new questions about this conceptualization, at least as pertains to the current stage of the obesity epidemic in the USA. Based on nationally representative surveys, Mozaffarian notes that energy intake has plateaued or declined since 2000, and physical activity has increased moderately, even as rates of obesity continue to rise. (Because of disproportionate increases in waist circumference in women, obesity trends as assessed by BMI may underestimate the extent to which the epidemic has advanced since 2000 [126]). These trends, he argues, call for consideration of alternative causal explanations, including those involving metabolic dysfunction.

最近的一項行為試驗的綜合分析報告顯示，宏量營養素為重點的飲食在長期減重方面沒有差異 [ 127]，如 Hall 等人所引用 [ 9]，而其他比較低碳水化合物與高碳水化合物飲食的綜合分析則表明前者有顯著但適度的優勢 [ 128, 129, 130, 131]。然而，對這些證據的解釋往往將療效與行為實施混為一談 [ 132]。大多數行為減重試驗缺乏足夠的干預強度，以在組間獲得宏量營養素攝入的強烈對比，且組間的初始減重差異迅速減弱。在現代食品環境中，維持飲食改變可能很困難，但這一挑戰並非不可克服。隨著對療效的更好了解，可以設計出更強有力的行為和環境干預，以促進長期遵循。在少數採用強化干預（例如部分食物供應）的試驗中，低 GL 飲食在整個方案期間的減重效果優於高 GL 飲食 [ 133, 134]。

A recent meta-analysis of behavioral trials reported no difference in long-term weight loss among macronutrient-focused diets [127], as cited by Hall et al. [9], whereas other meta-analyses comparing low- vs. high-carbohydrate diets suggest a significant, if modest, advantage to the former [128,129,130,131]. However, interpretation of this evidence tends to conflate efficacy with behavioral implementation [132]. Most behavioral weight loss trials lack sufficient intervention intensity to obtain strong contrasts in macronutrient intakes between groups, and initial differences in weight loss between groups wane rapidly. Maintenance of dietary change can be difficult in the modern food environment, but this challenge is not insurmountable. With better knowledge of efficacy, more powerful behavioral and environmental interventions can be designed to facilitate long-term adherence. Among the few trials that employed intensive interventions (e.g., partial food provision), weight loss was greater on low- vs. high-GL diets for the duration of the protocols [133, 134].

自由生活試驗的局限性原則上可以通過代謝病房試驗來克服，後者對遵守規範和混雜因素保持嚴格控制。然而，由於成本和後勤挑戰，這些試驗通常持續時間較短，這引發了對涉及慢性影響的無根據推斷的擔憂。霍爾[20]認識到需要至少持續幾個月的試驗，他觀察到：

The limitations of free-living trials can be, in principle, circumvented by metabolic ward trials that maintain strict control over adherence and confounding factors. However, due to cost and logistical challenges, these trials are usually short in duration, raising concern for unfounded inference involving chronic effects. The need for trials of at least several months duration was recognized by Hall [20], who observed that:

“即使是能量消耗和宏量營養素平衡的微小差異，理論上也可以導致體重和組成的顯著差異，只要飲食在長期內保持不變。僅僅 100 kcal/d 的能量消耗差異，可能會導致約 10 g/d 的初始體脂不平衡。根據當前的體組成方法，檢測這樣的體脂差異需要持續約 100 天的時間。然而，這一可能性仍需進一步研究。”
“even small differences in energy expenditure and macronutrient balance can theoretically lead to significant differences of body weight and composition if the diets are maintained over long periods. A 100 kcal/d difference in energy expenditure alone could lead to an initial body fat imbalance of about 10 g/d. Using current body composition methods, it would require a sustained period of about 100 days to detect such a difference in body fat. Nevertheless, this possibility requires further investigation.”

此外，對於大營養素變化的代謝適應可能需要幾週到幾個月的時間 [ 135, 136, 137, 138, 139, 140, 141]。最近的一項綜合分析報告顯示，在比較低碳水化合物飲食與高碳水化合物飲食的研究中，持續時間≥2.5 週的研究中總能量消耗較高，且異質性低 [ 142]。在持續時間<2.5 週的研究中未顯示出有意義的飲食效果，且異質性顯著，這加強了對短期試驗價值的擔憂。代謝病房的人工環境也可能獨立於潛在的生理機制影響飲食行為。

Furthermore, metabolic adaptations to macronutrient changes may require several weeks to months [135,136,137,138,139,140,141]. A recent meta-analysis reported higher total energy expenditure, with low heterogeneity, among studies ≥2.5 weeks duration comparing low- vs. high-carbohydrate diets [142]. No meaningful dietary effect was evident in studies <2.5 weeks, with substantial heterogeneity, reinforcing concerns about the value of short trials. The artificial setting of a metabolic ward may also affect eating behavior independently of underlying physiological mechanisms.

Hall 等人 [9] 將兩項為期兩週的住院試驗解釋為與碳水化合物-胰島素模型不一致。在其中一項試驗 [143] 中，隨意能量攝入在“超加工”飲食與“未加工”飲食之間約為 500 kcal/d 的差異 [9]。然而，這一差異迅速減弱，在“超加工”飲食上以 -25 kcal/d 的斜率下降，這表明該效應可能在額外的兩週後消失。此外，最初的能量攝入差異完全可歸因於能量密度的巨大差異，這是一個影響短期但不影響慢性攝入的因素（見下文）。與此相關的擔憂是無法區分關鍵的宏量營養素機制。雖然食物加工的程度對消化速率、荷爾蒙反應和高碳水化合物食物的健康影響有很大影響，但對高脂肪和高蛋白食物的生理意義較小（表 3）——這意味著超加工食品的不良影響可以更好地用碳水化合物-胰島素模型來解釋，而不是能量平衡模型。

Hall et al. [9] interpret two 2-week inpatient trials as inconsistent with the CIM. In one of these trials [143], ad libitum energy intake was ~500 kcal/d greater on an “ultra-processed” vs. “unprocessed” diet [9]. However, this difference waned rapidly, with a slope of −25 kcal/d on the “ultra-processed” diet, suggesting the effect could extinguish after an additional 2 weeks. Furthermore, the initial difference in energy intake was fully attributable to the large difference in energy density, a factor that affects short-term, but not chronic, intake (see below). Related to this concern is the inability to distinguish crucial macronutrient mechanisms. Whereas the extent of food processing greatly affects digestion rate, hormonal response, and health impacts of high-carbohydrate foods, processing has lesser physiological significance for high-fat and high-protein foods (Table 3)—implying that the adverse effects of ultra-processed foods can be better explained by the CIM than by the EBM.

在第二個為期兩週的病房試驗中，對比低脂肪與低碳水化合物飲食時，觀察到類似的效果減弱模式，這可能與代謝適應和能量密度有關[ 144]。在確定的研究結果出爐之前，似乎明智的做法是不要假設這些減弱的效果會穩定並在長期內影響體重。

A similar pattern of effect attenuation, potentially related to metabolic adaptation and energy density, was observed in a second 2-week ward trial comparing low-fat vs. low-carbohydrate diets [144]. Pending definitive research, it seems prudent not to assume that these waning effects would stabilize and influence body weight over the long term.

胰島素在脂肪細胞生理學中的主導作用，包括脂肪生成和脂肪分解，已被認識了幾十年[145]。在糖尿病患者中，胰島素及增加胰島素分泌或對脂肪組織代謝作用的藥物會導致體重增加[146]。這些影響中的一些可能涉及與能量平衡模型相容的其他機制，例如減少糖尿病尿糖。然而，降低分泌的藥物所引起的體重減輕[147]表明，胰島素對脂肪儲存的作用在啮齒類動物中[103, 104, 105, 106, 107, 108, 109]的觀察也發生在人類中。例如，α-葡萄糖苷酶抑制劑[148]，通過降低碳水化合物的血糖反應，產生約 1 公斤的體重減輕，同時降低 HbA1c，這與一些其他導致體重增加的糖尿病藥物（包括胰島素）形成對比。在沒有糖尿病的人中，降低胰島素分泌的藥物也會導致體重減輕[147]。此外，兩項新的研究表明，胰島素在人體中抑制脂肪組織的線粒體呼吸[149, 150]。
A dominant role of insulin on adipocyte physiology, including lipogenesis and lipolysis, has been recognized for decades [145]. In patients with diabetes, insulin and drugs that increase insulin secretion or action on adipose tissue metabolism cause weight gain [146]. Some of these effects may involve other mechanisms compatible with EBMs, such as reduced glycosuria. However, the weight loss induced by drugs that lower secretion [147] suggests that the action of insulin on fat storage seen in rodents [103,104,105,106,107,108,109] occurs in humans. For instance, alpha-glucosidase inhibitors [148], which lower the glycemic response to carbohydrate, produce weight loss of ~1 kg, while also lowering HbA1c, in contrast to some other diabetes drugs (including insulin) that cause weight gain. Drugs that lower insulin secretion in people without diabetes also cause weight loss [147]. Furthermore, two new studies suggest that insulin suppresses adipose mitochondrial respiration in humans [149, 150].

Hall 等人 [9] 認為 GLP-1 受體激動劑對肥胖的有效性是反對碳水化合物-胰島素模型的證據，因為這種腸促胰島素能迅速增強葡萄糖刺激的胰島素分泌。然而，GLP-1 還具有其他相關的生物作用，包括降低胃排空速率（這會降低血糖反應）[151]。事實上，GLP-1 受體激動劑會長期減少總胰島素分泌的指標 [152, 153]，儘管這種效果是直接還是間接仍不清楚。無論如何，飲食中的 GL 強烈影響腸促胰島素的分泌特徵，而腸促胰島素對脂肪細胞的胰島素敏感性具有直接作用。基於這些原因，GLP-1 位於碳水化合物-胰島素模型的中心因果路徑上 [8]。
Hall et al. [9] consider the effectiveness of GLP-1 receptor agonists for obesity as evidence against the CIM, because this incretin acutely potentiates glucose-stimulated insulin secretion. However, GLP-1 has other relevant biological actions, including reduced gastric emptying rate (which lowers glycemic response) [151]. In fact, GLP-1 receptor agonists chronically reduce measures of total insulin secretion [152, 153], although whether this effect is direct or indirect remains unclear. In any event, dietary GL strongly affects the incretin secretion profile and incretins have direct actions on adipocyte insulin sensitivity. For these reasons, GLP-1 lies on the central causal pathway in the CIM [8].

關於脂解的抑制，Hall 等人 [9] 引用了一項研究，顯示阿西匹莫克對人類體重沒有影響 [154]。然而，這種尼古丁酸受體激動劑具有複雜的生物作用，使試驗的解釋變得困難。阿西匹莫克增加了反調節激素的分泌，促進蛋白質分解，並誘導葡萄糖氧化的補償性增加 [155]。值得注意的是，使用各種藥物抑制脂肪酸氧化會刺激實驗動物和人類的食物攝取 [156, 157, 158, 159, 160, 161]。

Regarding inhibition of lipolysis, Hall et al. [9] cite a study showing no effect of acipimox on weight in humans [154]. However, this nicotinic acid receptor agonist has biological actions that complicate interpretation of the trials. Acipimox increases counter-regulatory hormone secretion, promotes protein breakdown, and induces a compensatory increase in glucose oxidation [155]. Of note, inhibition of fatty acid oxidation with various agents stimulates food intake in experimental animals and humans [156,157,158,159,160,161].

總結有關這兩種模型的證據，動物數據顯示過量的脂肪沉積顯然可以與能量攝入脫鉤，這與能量平衡模型的基本前提相對立。在涉及不僅是飲食，還有被認為介導食物攝入的腦路徑的動物模型中，肥胖可以在不增加食物攝入的情況下發生。然而，人類數據存在重大方法學限制，迄今為止，這些限制阻礙了對這兩種模型的明確測試。為了推進科學，需要進行具有足夠持續時間和互補設計的研究，包括：（1）能夠區分瞬時與慢性宏量營養素效應的機制導向餵養研究（≥1 個月）；（2）具有足夠干預強度以產生有意義的長期行為改變的有效性試驗（≥1 年）；以及（3）理想上從兒童時期開始的肥胖自然歷史的縱向觀察研究（≥10 年）。

To summarize evidence pertaining to the two models, the animal data demonstrate that excessive fat deposition can evidently be disassociated from energy intake, opposing a fundamental premise of the EBM. In animal models involving not only diet, but also brain pathways considered to mediate food intake, obesity can occur without increased food intake. However, the human data have major methodological limitations that have, so far, precluded a definitive test of the two models. To advance science, studies with adequate duration and complementary designs will be needed, including: (1) mechanistically oriented feeding studies capable of distinguishing transient from chronic macronutrient effects (≥1 month); (2) efficacy trials with adequate intervention intensity to produce meaningful long-term behavior change (≥1 year); and (3) longitudinal observational studies, ideally beginning in childhood, of the natural history of obesity (≥10 years).

臨床轉化與公共採納
Clinical translation and public adoption

雙方在這場辯論中都同意，食品環境的根本變化推動了肥胖疫情。新的能量平衡模型（EBM）對如此廣泛的飲食因素的關注提供了很少的新可行見解（無處不在、便宜、方便、能量密集、超加工食品，份量大、脂肪和糖含量高，蛋白質和纖維含量低）。隱含的建議是避免垃圾食品，這已經被倡導多年[ 56, 162, 163, 164, 165, 166]。特別令人擔憂的是，Hall 等人[ 9]所針對的飲食因素與慢性體重增加之間的因果關係尚未得到證明，除了那些也涉及碳水化合物-胰島素模型（CIM）相關途徑的因素（即，含有高升糖指數（GL）和果糖的糖；降低共同攝入碳水化合物升糖指數（GI）的纖維；以及降低共同攝入碳水化合物升糖指數（GI）並刺激胰高血糖素分泌的蛋白質）。剩餘的 EBM 特定飲食目標包括：

Both sides of this debate agree that fundamental changes in the food environment have driven the obesity pandemic. The new EBM’s focus on such a broad range of dietary factors offers few new actionable insights (ubiquitous, cheap, convenient, energy-dense, ultra-processed foods high in portion size, fat, and sugar, and low in protein and fiber). The implicit advice, to avoid junk foods, has been advocated for years [56, 162,163,164,165,166]. Of particular concern, causal relationships with chronic weight gain have not been demonstrated for the dietary factors targeted by Hall et al. [9] other than those that also involve CIM-related pathways (i.e., sugar, which is high in GL and fructose; fiber, which lowers the GI of co-ingested carbohydrates; and protein, which lowers the GI of co-ingested carbohydrates and stimulates glucagon secretion). The remaining EBM-specific dietary targets include:

能量密度。能量密度的急性變化會影響短期攝入。例如，Bell 等人[167]在一項交叉設計中給予 18 名女性不同能量密度但控制了宏量營養素的飲食。這些女性在每個條件下消耗的食物體積相同，導致能量密度增加 31%，而在高能量密度與低能量密度條件下的能量攝入也相應增加 31%。Hall 等人[9]引用了幾個干預措施和一項觀察性分析，以建議一個重要的慢性影響。在一項干預研究[168]中，97 名肥胖女性被建議單獨減少脂肪攝入，或減少脂肪攝入並增加低能量密度的水果和蔬菜。經過 1 年，低能量密度組的完成者比對照組多減少了 1.5 公斤，但這一效果與瘦體重的更大損失有關。這些組別在總脂肪量或腰圍上並無差異。在另一項干預研究[169]中，200 名成年人被建議遵循能量限制飲食，其中一些被指示消耗不同量的低能量密度湯與高能量密度固體零食。 在這裡，1 年後的體重差異仍然 modest。然而，零食組的參與者消耗了極高的 GL 項目（“餅乾、烤馬鈴薯片、烤玉米餅片、貝果片和椒鹽脆餅”）；不出所料，這組的碳水化合物消耗量更高，排除了任何相關的因果推斷。此外，在這個問題上最大的和最長的試驗（n = 2718）中，干預組之間的能量密度顯著差異持續了 4 年，對能量攝入或體重沒有影響[ 170]。關於能量密度的觀察數據[ 171]，Bes-Rastrollo 等[ 172]強調了混淆和普遍性方面的主要擔憂。

Energy density. Acute changes in energy density affect short-term intake. For example, Bell et al. [167] gave 18 women, in a cross-over design, diets differing in energy density but controlled for macronutrients. The women consumed the same volume of food during each condition, resulting in a 31% increase in energy density and a corresponding 31% increase in energy intake on the high- vs. low-energy-density conditions over 2 days. Hall et al. [9] cite several interventions and one observational analysis to suggest an important chronic effect. In one interventional study [168], 97 women with obesity were counseled to decrease fat intake alone or to decrease fat intake and increase low-energy-density fruits and vegetables. After 1 year, completers in the low-energy density group lost 1.5 kg more than those in the comparison group, but the effect related to greater loss of lean mass. The groups did not differ in total fat mass or waist circumference. In another interventional study [169], 200 adults were counseled to follow energy-restricted diets, with some instructed to consume varying amounts of low-energy-density soups vs. high-energy-density solid snacks. Here again, there was a modest difference in body weight at 1 year. However, participants in the snack group consumed exceedingly high-GL items (“crackers, baked potato chips, baked tortilla chips, bagel chips, and pretzels”); not surprisingly, carbohydrate consumption was greater in this group, precluding any relevant causal inference. Furthermore, in the largest and longest trial of this question (n = 2718), a significant difference in energy density between intervention groups was maintained for 4 years, with no effect on energy intake or body weight [170]. Regarding observational data on energy density [171], Bes-Rastrollo et al. [172] highlight major concerns about confounding and generalizability.

飲食脂肪。對於能量密度在肥胖中的角色的假設在很大程度上促使了從二十世紀末開始的公共衛生建議中專注於減少飲食脂肪。然而，低脂飲食在與肥胖相關的結果上並未顯示出優越性，一些綜合分析得出與高脂飲食相比在減重方面的劣勢。美國農業部幾乎已經放棄了減少總飲食脂肪的公共衛生運動。

Dietary fat. Assumptions about the role of energy density in obesity motivated, in large measure, the focus on reducing dietary fat in public health recommendations from the late twentieth century [173,174,175,176,177]. However, low-fat diets have not shown superiority for obesity-related outcomes [178,179,180], and some meta-analyses conclude inferiority vs. higher-fat diets for weight loss [128,129,130]. The USDA has virtually abandoned the public health campaign to reduce total dietary fat [181].

食品加工。在一項為期兩週的試驗中，食物攝取量在“超加工”飲食與“未加工”飲食之間的比較中更高[ 143]。然而，這一效果，約 20%的增長，僅歸因於非飲料能量密度的約 85%增長，根據 Bell 等人的研究[ 167]。Poti 等人[ 182]對觀察數據的系統評估得出結論：“目前尚不清楚與肥胖的關聯是否可以歸因於加工本身或超加工食品的營養成分……而且潛在的殘餘混淆可能性很高。”如表 3 所示，宏量營養素組成影響食品的原生基質和結構的破壞如何改變健康效果，這表明 CIM 機制比 EBM 更能解釋超加工食品與肥胖之間的關聯。

Food processing. Food intake was greater in a 2-week trial with consumption of an “ultra-processed” vs. “unprocessed” diet [143]. However, this effect, a ∼20% increase, is attributable to the ∼85% increase in non-beverage energy density alone, based on the findings of Bell et al. [167]. A systematic review of observational data by Poti et al. [182] concludes, “It remains unclear whether associations [with obesity] can be attributed to processing itself or the nutrient content of ultra-processed foods.… and the potential for residual confounding was high.” As demonstrated in Table 3, macronutrient composition affects how disruption of native matrix and structure of a food alters health effects, suggesting that CIM mechanisms offer a better explanation for the associations of ultra-processed foods with obesity than those of the EBM.

儘管肥胖盛行率的持續上升可能歸因於公眾採納的缺乏，而非能量平衡模型本身的任何內在缺陷，但過去一個世紀以能量平衡模型為指導的治療結果卻顯示出相反的情況。1959 年，精神科醫生和肥胖研究者阿爾伯特（“米奇”）斯坦卡德與梅維斯·麥克拉倫-休姆進行了一項為期 30 年的文獻回顧，追溯到 1920 年代最初使用卡路里計算進行體重控制的時期。他們得出結論，報告之間的結果“驚人地相似且驚人地糟糕”，並且這些結果“雖然看起來糟糕，但仍然[可能]比一般醫生所獲得的結果要好。”明確針對能量平衡的概念，作者寫道：

Although the continuing increases in obesity prevalence might be attributable to lack of public adoption rather than any inherent deficiency of the EBM itself, the results of EBM-guided treatment throughout the last century suggest otherwise. In 1959, psychiatrist and obesity researcher Albert (“Mickey”) Stunkard with Mavis McLaren-Hume [183] conducted a 30-year literature review dating back to the original use of calorie counting for weight control in the 1920s. They concluded that the outcomes among reports were “remarkably similar and remarkably poor” and that these results “poor as they seem, are nevertheless [probably] better than those obtained by the average physician.” Explicitly addressing the notion of energy balance, the authors wrote:

“許多年前，詳細的代謝研究證明，人類並不違反熱力學定律，過多的體脂肪是由於熱量攝入超過熱量消耗所致。這一不無道理的發現隨即被確立為‘所有肥胖都源於過度飲食’的格言……醫生的工作似乎只是解釋半飢餓會減少脂肪儲備，為此開處方飲食，然後靜觀其變。如果患者如預期般減輕體重，這僅僅確認了治療肥胖實際上是一件相當簡單的事情。然而，如果患者如常常發生的那樣未能減重，他便被視為不合作或被責備為貪吃。能夠考慮到不遵循療程本身可能是一個醫療問題的醫生是非常罕見的。”

“Many years ago detailed metabolic studies demonstrated that human beings do not defy the … law of thermodynamics and that excessive body fat results from an excess of caloric intake over caloric expenditure. This not unreasonable finding was thereupon enshrined as the dictum that ‘all obesity comes from overeating’… The physician’s job, it seemed, was simply to explain that semistarvation reduces fat stores, to prescribe a diet for this purpose, and to sit by. If the patient lost weight as predicted, this merely confirmed the comfortable feeling that treatment of obesity was really a pretty simple matter. However, if, as so often happened, the patient failed to lose weight, he was dismissed as uncooperative or chastized as gluttonous. It was the rare physician who entertained the possibility that failure to follow a regimen might in itself be a medical problem.”

在 1992 年，國家衛生研究院舉辦了一次關於自願減重和控制方法的共識發展會議，參加者包括許多肥胖領域的領先專家。當時，飲食脂肪限制被認為是“達到健康體重的最佳方法……因為它更容易在不必吃小份量的情況下攝取更少的卡路里”[184]，這一觀點在當代學術評論中經常被提及[173, 174]。然而，共識會議發現，肥胖治療的效果並沒有比 Stunkard 和 McLaren-Hume 所回顧的結果好得多[183]。會議紀要總結道：“參加減重計劃的參與者通常會減少約 10%的體重……[大部分]的體重在 1 年內會恢復，幾乎所有的體重在 5 年內會恢復”[185]。此外，Mozaffarian 的分析[125]提供了定量證據，顯示在最近幾十年中，美國人至少在整體上遵循了 EBM 的基本“少吃”建議——即使肥胖率仍在上升。

In 1992, the National Institutes of Health sponsored a Consensus Development Conference on Methods for Voluntary Weight Loss and Control, including many of the leading experts in obesity. At that time, dietary fat restriction was considered “The best means of achieving a healthy weight … preferred because it is easier to eat fewer calories without having to eat small portions” [184], a view frequently espoused in contemporary academic reviews [173, 174]. However, the Consensus Conference found little evidence that obesity treatment achieved much better outcomes that those reviewed by Stunkard and McLaren-Hume [183]. Conference proceedings concluded that “participants who remain in weight loss programs usually lose approximately 10% of their weight…. [much] of the weight is regained within 1 year, and almost all is regained within 5 years” [185]. Moreover, the analysis of Mozaffarian [125] provides quantitative evidence that, in recent decades, Americans have adhered to the fundamental “eat less” recommendation of the EBM, at least on a population basis – even as obesity rates continue to increase.

公理地說，專注於因果驅動因素（沿著機制途徑的上游）的疾病治療應該比針對下游後果和表現的治療更有效，對患者也更可持續。如果發燒被類比為“熱平衡”的失調，那麼可以合理地開處方冷水淋浴來降低體溫。這種治療會暫時有效（如果能說服發燒的患者嘗試），但身體會通過劇烈顫抖和血管收縮來補償熱量損失。一旦患者離開冷水淋浴，發燒就會回來。退燒藥通過解決熱量積累的生物驅動因素來更有效且更愉快地為患者服務。同樣，如果肥胖是由於燃料分配失調所致，那麼治療該問題的措施（例如，通過降低胰島素與胰高血糖素的比率）將比限制卡路里更能獲得更好的遵從性，因為患者在減重過程中會感受到更少的饑餓感和較小的能量消耗減少。

Axiomatically, disease treatment focused on causal drivers (upstream along the mechanistic pathway) should be more effective, and more sustainable for the patient, than those targeting downstream consequences and manifestations. If fever were, by analogy, considered a disorder of “heat balance,” one might rationally prescribe a cold shower to reduce body temperature. This treatment would work temporarily (if one could convince a febrile patient to try it), but the body would compensate for the heat loss by severe shivering and blood vessel constriction. Once the patient got out of the cold shower, the fever would return. Antipyretics work more effectively, and more pleasantly for the patient, by addressing the biological driver of heat accumulation. Similarly, if obesity results from a disorder of fuel partitioning, then measures to treat that problem (e.g., by reducing the insulin-to-glucagon ratio) would achieve better adherence than calorie restriction, because the patient would experience less hunger and a lesser reduction in energy expenditure with weight loss.

混淆的範式衝突 Muddling paradigm clash

維持這些競爭模型之間的對比對於澄清思路、指導研究議程以及確定有效的預防和治療手段至關重要。Hall 等人 [9] 通過將碳水化合物-胰島素模型（CIM）降級為能量平衡模型（EBM）的“特例”來混淆這一對比。這一主張掩蓋了模型之間最根本的可能差異：因果方向和因果機制（圖 1）。以這種方式將 CIM 納入 EBM 需要將 EBM 解釋得過於寬泛，以至於使其無法被證偽，因此作為科學假設毫無用處。正如卡爾·波普爾所說：“一個解釋一切的理論，什麼也不解釋。”

Maintaining the contrast between these competing models is critical to clarify thinking, inform a research agenda, and identify effective means of prevention and treatment. Hall et al. [9] muddle this contrast by relegating the CIM to “a special case” of the EBM. This claim belies the most fundamental possible differences among models: causal direction and mechanisms of causality (Fig. 1). To subsume the CIM in this way requires construing the EBM so broadly as to make it unfalsifiable, and consequently useless as a scientific hypothesis. As Karl Popper reportedly said, “a theory that explains everything, explains nothing.”

Hall 等人 [9] 也聲稱，CIM 已經放棄了基本的原則，提到之前的「脂肪中心」公式僅考慮胰島素在脂肪組織中的作用。然而，這種描述並不是 CIM 支持者所提出的，並且提供了一個錯誤的區分。多種荷爾蒙、自主神經系統和其他影響對脂肪組織生物學的控制已被認識了幾十年 [27]。事實上，高 GL 和高糖飲食的生理作用早已被概念化為涉及多個器官之間的綜合關係，超越脂肪組織和多種荷爾蒙，超越胰島素 [6, 29]。

Hall et al. [9] also claim that the CIM has abandoned fundamental precepts, referring to prior “adipocentric” formulations said to consider only the actions of insulin in adipose tissue. However, this characterization was not made by CIM proponents and offers a false distinction. The control of adipose tissue biology by multiple hormonal, autonomic and other influences has been recognized for decades [27]. Indeed, the physiological actions of high-GL and high-sugar diets have long been conceptualized as involving integrated relationships among multiple organs beyond adipose tissue and numerous hormones beyond insulin [6, 29].

對於 CIM 修訂的擔憂與他們承認的事實形成對比，即“EBM 的發展仍然需要闡明在動態食品環境中最負責引發肥胖的因素，以及這些因素如何改變控制食物攝入的腦回路的機制”[9]。事實上，基於 EBM 的建議的飲食目標已經從 20 世紀初的卡路里計算[186]，轉變為 20 世紀末對飲食脂肪限制的全面關注[173, 174, 175, 176, 177, 187]，再到認為所有卡路里都是相同的[2, 18, 19]，再到新的公式[9]，副標題為“超越卡路里進入，卡路里消耗”，現在將一系列現代飲食因素歸咎於肥胖。為了使科學模型保持相關性，它們必須隨著知識的積累而發展。

This concern about CIM revision contrasts with their acknowledgment that “development of the EBM [still] requires elucidation of the factors in the dynamic food environment that are most responsible for instigating obesity [and] the mechanisms by which these factors alter the brain circuits controlling food intake” [9]. Indeed, dietary targets of EBM-based recommendations have changed from calorie counting in the early twentieth century [186] to an overarching focus on dietary fat restriction in the late twentieth century [173,174,175,176,177, 187], to the notion that all calories are alike [2, 18, 19], to the new formulation [9], subtitled “beyond calories in, calories out,” that now blames a host of modern dietary factors. For scientific models to remain relevant, they must grow as knowledge accrues.

即使 Hall 等人[9]批評 CIM 的來源，他們的 EBM 也存在重大缺陷，包括：
Even as Hall et al. [9] criticize the provenance of the CIM, their EBM has major deficiencies, including:

• 缺乏明確的可測試假設。如何對因果途徑中的關鍵步驟進行調查？哪些研究將區分所提出的因果途徑（過度飲食驅動慢性體重增加）與 CIM 中的對立假設？當人類或動物在實驗中被過度餵食時，他們最初會增加體重。但是，飢餓感和能量消耗的變化會對持續的體重變化產生對抗作用；在強迫餵食結束後，個體通常會減少進食，直到體重回到基線[188, 189, 190, 191, 192, 193]。換句話說，推入脂肪組織的多餘能量並不會“保持不變”[4]，然而在習慣性飲食中累積的多餘脂肪質量卻保持著驚人的穩定。

• Lack of explicit testable hypotheses. How will key steps along the causal pathway be interrogated? What studies will differentiate the proposed causal pathway (overeating drives chronic weight gain) from the contrasting hypothesis in the CIM? When humans or animals are experimentally overfed, they gain weight initially. But changes in hunger and energy expenditure oppose ongoing weight change; after the force-feeding ends, individuals characteristically undereat until body weight returns to baseline [188,189,190,191,192,193]. In other words, the excess energy “pushed” into adipose tissue doesn’t stay “put” [4], yet excess adipose mass accumulated over time on habitual diets remains remarkably stable.

• 同義反覆。當辯稱 EBM 的反對者將物理學與病理生理學混淆時，Hall 等人聲稱，“EBM 納入了能量分配的生理機制……因此，整體能量不平衡主要反映為脂肪不平衡，無論飲食成分如何。”他們還聲稱“全身脂肪不平衡最終主要反映為脂肪儲存的變化。”這樣做，他們傳播了這種混淆。如上所述，能量守恆定律認為，能量平衡的變化必須與脂肪和脂肪組織質量的相應變化共存（身體的主要能量儲存生物分子和儲存庫，分別）。這些同義反覆並未提供任何機制上的洞見。
  Tautologies. While arguing that opponents of the EBM confuse physics with pathophysiology, Hall et al. assert that, “the EBM incorporates physiological mechanisms underlying energy partitioning … such that overall energy imbalances are primarily reflected as fat imbalances regardless of the composition of the diet.” They also assert that “whole-body fat imbalances end up primarily reflected as changes in adipose tissue fat storage.” In so doing, they propagate this confusion. As considered above, the law of energy conservation holds that a change in energy balance must coexist with a commensurate change in fat and adipose tissue mass (the body’s main energy storage biomolecule and depot, respectively). These tautologies provide no mechanistic insight.

• 關鍵模型組件涉及的機制稀缺。新的能量平衡模型如何解釋人口水平體重的快速增加，以及個體之間隨時間的大變異？生理調節變量（例如，體溫、血清鈉）在極端條件下以外的情況下特徵為穩定。哪些研究能區分假定的介導因素（例如，獎勵、享樂影響）與碳水化合物-胰島素模型中的那些（對宏量營養素組成的激素反應）？此外，如果對美味食物的愉悅反應導致慢性過度攝食，為什麼獨立證明可口性對肥胖的影響如此困難[194, 195, 196, 197, 198, 199, 200, 201]？

• Paucity of mechanisms involving key model components. How does the new EBM explain the rapid population-level increase in weight, and large variations within individuals over time? Physiologically regulated variables (e.g., body temperature, serum sodium) are characterized by stability except under extreme conditions. What studies would distinguish the putative mediators (e.g., reward, hedonic influences) from those in the CIM (hormonal response to macronutrient composition)? Moreover, if pleasure-related responses to tasty foods cause chronic overconsumption, why has it been so difficult to demonstrate an independent effect of palatability on obesity [194,195,196,197,198,199,200,201]?

• 忽視已建立的代謝機制。對於肥胖個體，能量限制在體脂肪儲存達到正常水平之前就會引發飢餓反應的特徵（包括降低的能量消耗）。可口食物的享樂和獎勵方面如何觸發代謝反應？

• Disregard of well-established metabolic mechanisms. For individuals with obesity, energy restriction elicits hallmarks of the starvation response (including reduced energy expenditure) long before body fat stores reach a normal level. How do the hedonic and reward aspects of palatable food trigger metabolic responses?

• 難以解釋肥胖的自然歷史。大多數肥胖形式在多年內發展，與每日約 10 到 20 卡路里的正能量平衡有關（相當於 1 茶匙糖的能量含量）。自 1970 年至今，美國的能量攝入量增加約 200 卡路里/天（12 盎司葡萄汁）[ 122, 125, 202]。考慮到過度體重的心理社會及其他負擔，為什麼如此少的人能夠成功通過有意識的控制來補償這些小的每日影響？畢竟，成年人經常抵抗愉悅的誘惑（例如，性、毒品），這些誘惑也會招募潛意識的驅動力？

• Difficulty accounting for the natural history of obesity. Most forms of obesity develop over many years, associated with a positive energy balance of ∼10 to 20 kcal/d (the energy content in 1 teaspoon of sugar). The secular increase in energy intake from 1970 to the present in the U.S. is ∼200 kcal/d (12 oz grape juice) [122, 125, 202]. Considering the psychosocial and other burdens of excessive weight, why do so few people successfully compensate by conscious control for these small daily effects? After all, adults routinely resist pleasurable temptations (e.g., sex, drugs) that also recruit subconscious drives?

• 依賴於不區分模型的假設。新的能量平衡模型（EBM）解釋證據表明大腦控制體重，支持過度飲食在肥胖中的因果作用。如上所述，大腦還影響幾乎所有能量代謝和脂肪細胞生物學的方面。

• Reliance on assumptions that do not differentiate among models. The new EBM interprets evidence that the brain controls body weight as supporting a causal role of overeating in obesity. As considered above, the brain also influences virtually all aspects of energy metabolism and adipocyte biology.

結論 Conclusions

對於難以解決的公共衛生問題，科學模型的目的是指導信息豐富的研究設計，並通過幫助闡明因果機制，建議有效的預防或治療方法。新的能量平衡模型並未達到這一目的。至少，未來的模型應該 (1) 指定可測試的、以機制為導向的預測，以檢查因果途徑；(2) 解釋為什麼增加的人口水平 BMI 受到代謝反應的支持；以及 (3) 演示臨床研究所建議的、動物模型所證明的飲食的卡路里獨立效應如何能夠整合進這一模型中。
For intractable public health problems, the purpose of scientific models is to guide the design of informative research and, by helping to elucidate causal mechanisms, suggest effective approaches to prevention or treatment. The new EBM does neither. At a minimum, future formulations should (1) specify testable, mechanistically oriented predictions that examine the causal pathway; (2) explain why the increased population-level BMI is defended by metabolic responses; and (3) demonstrate how calorie-independent effects of diet suggested by clinical research and demonstrated by animal models can be integrated in this model.

能量平衡模型及其前身在近一個世紀以來主導了思考[7]——影響了科學設計、實驗結果的解釋、公共衛生指導方針和臨床治療——在很大程度上排除了其他觀點。例如，國立衛生研究院（NIH）贊助了多項針對肥胖相關結果的低脂飲食的多中心試驗[178, 179, 180]（所有試驗的主要結果均為負面），但對於低升糖飲食卻沒有類似的研究。隨著傳統策略無法遏制肥胖相關疾病的上升趨勢，應該研究新的因果模型，而不是因為誇張的聲稱已經駁斥了它們而被壓制[2, 9, 18, 19, 57, 58, 203, 204, 205]。
The EBM and its precursors have dominated thinking for nearly a century [7]—influencing scientific design, interpretation of experimental findings, public health guidelines, and clinical treatment—largely to the exclusion of other views. For instance, the NIH has sponsored numerous multi-center trials of low-fat diets for obesity-related outcomes [178,179,180] (all with negative primary outcomes), but nothing comparable for low-GL diets. With the inability of conventional strategies to stem the rising toll of obesity-related disease, new causal models should be studied, not suppressed by hyperbolic claims of having disproven them [2, 9, 18, 19, 57, 58, 203,204,205].

無可否認，對於複雜的科學問題的辯論可能會產生兩極化，雙方都有選擇性地引用不確定證據的傾向。這一問題因研究肥胖自然歷史中小的日常影響而變得更加複雜。為了科學進步和公共健康，這場辯論的所有方面應該共同努力，制定互相可接受的競爭模型版本，並設計無偏見的研究來對其進行嚴格測試。通過認識到在某些實驗環境中對一個模型的證據並不會使另一個模型在所有環境中失效，並且人類的肥胖發病機制可能包含兩者的元素，可以促進建設性的範式衝突。
Admittedly, debate on complicated scientific questions may polarize, with a tendency for both sides to cite selectively from inconclusive evidence. This problem is exacerbated by difficulties in studying the small daily effects that characterize the natural history of obesity. In the interests of scientific advancement and public health, all sides of this debate should work together to formulate mutually acceptable versions of competing models and design unbiased studies that would put them to a rigorous test. A constructive paradigm clash may be facilitated with the recognition that evidence for one model in certain experimental settings does not invalidate the other model in all settings, and that obesity pathogenesis in humans may entail elements of both.

最後，我們要強調的是，這一範式衝突不應延遲公共衛生行動。精製穀物和添加糖類約占美國和歐洲能量攝入的三分之一。這兩種模型都將這些高度加工的碳水化合物視為體重增加的主要驅動因素——儘管原因不同。無論這場辯論如何發展，現在已經存在共識，即在預防和治療肥胖方面，需要用最少加工的碳水化合物或健康脂肪來取代這些產品。
Finally, we would emphasize that this paradigm clash should not delay public health action. Refined grains and added sugars comprise about one-third of energy intake in the US and Europe. Both models target these highly processed carbohydrates—albeit for different reasons—as major drivers of weight gain. Regardless of how this debate may evolve, common ground now exists on the need to replace these products with minimally processed carbohydrates or healthful fats in the prevention and treatment of obesity.

.