For the primary focus of this review, PubMed was searched for meta-analyses of human controlled intervention trials investigating the effects of sugar sources on cardiometabolic outcomes. The sources to be considered were sugar-sweetened beverages, fruit juice, honey, and whole fruit. Where meta-analysed data from controlled trials were available, these were only used when at least 3 studies comprising the meta-analysed data to provide more conservative inferences.

What are the key nutritional differences between sugars sources?

With the exception of honey, when expressed per 100 g of food or per 100 mL of fluid, there is little difference between sugar sources in the energy or macronutrient content (Table 1). However, there are some notable differences in fibre, potassium, and polyphenol content (Table 1) [3, 11, 12]. In particular, 100% fruit juices and whole fruit display higher fibre, potassium and polyphenol content than SSBs and whole fruit contain more fibre than fruit juices. Whilst honey appears to have a high potassium and polyphenol content when expressed per 100 g, it is arguably more relevant to interpret these per g of sugar. When the nutritional composition is expressed per g of sugar, fruit juices - and especially whole fruit - display a markedly higher fibre, potassium, and polyphenol content than both SSBs and honey (Table 1). The potential relevance of these differences will be discussed after an overview of the evidence regarding the physiological effects of these sugar sources. It should also be noted that the portion size of these sugar sources varies and can therefore alter the likely intakes of nutrients and the glycaemic load (Table 2).

Table 1 Nutrient composition of sugar sources expressed per g or per mL of food item and as per g sugarTable 2 Nutrient composition of sugar sources expressed per UK portion sizeSugar sources and the relationships to cardiometabolic healthBlood glucose and insulin sensitivity

The glycaemic index of SSBs and honey are both in the moderate range, whereas the glycaemic index of orange and apple juice, and whole oranges and apples are all in the low range (Table 1) [13], suggesting that the addition and/or combination of factors in fruit-sources of sugars can lower the immediate glucose response when normalised for the amount of carbohydrate ingested. This acute response, however, does not seem to directly translate into chronic responses, whereby meta-analyses suggest that honey consumption can lower fasting glucose concentrations [15] (Fig. 2A). Whilst less clear than with acute responses, data do still show a broad pattern which is consistent with a role of food matrix effects on glycaemic control, such as an increase in fasting glucose concentration with addition of excess energy from liquid sources of sugars such as SSBs, which is not seen with addition of whole fruit (Fig. 2A) [16]. Indeed, substitution of whole fruit for other sources of energy in the diet can reduce HbA1c by ∼ 0.19% (95%CI: -0.03 to -0.35%; Fig. 2B) [16].

Fig. 2

Effects of experimental addition and/or substitution of various sugar sources into the diet on fasting glucose concentrations, glycated haemoglobin (HbA1c) and fasting insulin concentrations. Data are mean differences ± 95%CI redrawn from Ahmed et al. for honey [15] and Choo et al. for all other food sources [16]. SSB, sugar-sweetened beverages. Fructose-containing sugar doses in Choo et al. was a median of 15% energy intake for 4.5 weeks in substitution trials, and 12.2% energy intake for 6 weeks in addition trials. Honey doses in Ahmed et al. were at a median of 40 g of honey for 8 weeks

Glycaemic control is largely influenced by insulin secretion and insulin sensitivity. In people with normal β-cell function, for a given glucose concentration, higher insulin concentrations can be a marker of lower insulin sensitivity. Therefore, fasting insulin concentrations are often used marker of insulin sensitivity. Meta-analyses demonstrate that addition of excess energy as SSBs, but not substitution, can increase fasting insulin concentrations by 5 pmol/L (95%CI: 1 to 9 pmol/L; Fig. 2C) [16]. There is some indication that the source of sugars may be important in this regard, since evidence does not indicate that addition of fruit increases fasting insulin concentrations (mean difference − 0.3 pmol/L; 95%CI: -5 to 4 pmol/L) [16].

Direct comparison between sugar sources on glycaemic control and insulin sensitivity has been performed in several studies, albeit as secondary or tertiary outcomes [3, 4]. When intake of two whole apples per day for 8 weeks was compared to a sugar-matched control beverage made from apple juice, the treatment effect on fasting glucose and insulin concentrations was − 0.06 mmol/L (95%CI: -0.15 to 0.03 mmol/L) and − 0.05 pmol/L (-0.10 to 0.00 pmol/L), respectively [3]. Furthermore, when intake of 550 g/d whole apples was compared to 500 mL of either clear or cloudy apple juice, the changes in insulin concentrations were − 3 ± 17 pmol/L with whole apples, 8 ± 10 pmol/L with cloudy apple juice, and 3 ± 11 pmol/L with clear apple juice (treatment effect p > 0.05) [4]. Taken together these data indicate that the source of sugars may play a role in glycaemic control, whereby fruit sources, and in particular whole fruit may improve glycaemic control and insulin sensitivity, when compared with mixed comparators. These data are consistent with observational evidence demonstrating negative associations of whole fruit intake with development of diabetes, and a neutral association with 100% fruit juice [17,18,19]. Evidence from direct comparisons between sugar sources, however, is limited and therefore the certainty of causality between the food matrix effects of sugars on glycaemic control is constrained.

Blood lipids, lipoproteins, and inflammation

Low-density lipoprotein cholesterol (LDL-c) and chronic systemic inflammation are central drivers of atherosclerotic cardiovascular disease (CVD) [20]. Meta-analyses indicate that high fructose intakes can increase plasma apolipoprotein B and triglyceride concentrations during hypercaloric feeding trials, but evidence does not indicate such increases during isocaloric feeding trials [21]. When sources of sugars have been directly compared, consuming two whole apples per day for 8 weeks has been shown to reduce LDL-c concentrations by 0.14 mmol/L (0.02 to 0.26 mmol/L) and fasting triglyceride concentrations by 0.05 mmol/L (0.01 to 0.08 mmol/L) when compared with a sugar-matched control beverage comprised of fruit juice [3]. Somewhat consistent with this, consumption of 550 g apples per day for 4 weeks lowered LDL-c concentrations by > 0.3 mmol/L compared with 500 mL of clear apple juice per day, with similar LDL-c reductions when whole apples were compared with cloudy apple juice. The evidence did not indicate any significant differences in triglyceride response between intake of whole apples (-0.06 ± 0.38 mmol/L) compared with clear (0.03 ± 0.34 mmol/L) or cloudy (0.01 ± 0.36 mmol/L) apple juice [4].

The role of sugar source may also play a role in inflammatory marker responses to sugar intake. Meta-analysis of sugar sources demonstrates that C-reactive protein (CRP) concentrations are not lowered by either substitution or addition of SSBs [22]. However, whole fruit can lower CRP with either substitution or addition to the diet [22]. Similarly, whereas the data did not support a decrease in TNF-α concentrations with addition of SSBs or fruit juice to the diet, addition of whole fruit can lower TNF-α concentrations (Fig. 3B) [22]. Finally, for interleukin-6 (IL-6), the evidence did not suggest that SSBs, fruit juice, or whole fruit increased or decreased IL-6 concentrations (Fig. 3C) [22]. Interestingly, honey intake was demonstrated to increase IL-6 concentrations [15]. Direct comparisons of sugar sources do not provide evidence that either CRP or TNF-α concentrations differ with addition of whole fruit compared to fruit juice [3, 4]. Since energy balance status can influence inflammatory markers and possible mask potential effects of a dietary intervention, it is notable, that substitution studies with fruit juice were performed in either neutral or negative energy balance [22], whereas both substitution and addition studies of whole fruit were performed in either neutral or positive energy balance [22]. This suggests that these effects of fruit juice and of whole fruit can be seen within the context of changes in energy balance.

Accordingly, there is good evidence that the source of sugars can influence circulating LDL-c responses, whereby a more complex, whole/intact, food source can lower LDL-c concentrations compared with simpler, processed sources of sugars. Effects on triglyceride concentrations are less consistent, as are effects on inflammatory markers, with some suggestions of potential for fruit juice or whole fruit to lower some circulating inflammatory markers, albeit with less direct evidence.

Fig. 3

Effects of experimental addition and/or substitution of various sugar sources into the diet on circulating inflammatory marker concentrations. Data are mean differences ± 95%CI redrawn from Ahmed et al. for honey [15] and Qi et al. for other food sources [22]. SSB, sugar-sweetened beverages. Fructose-containing sugar doses in Choo et al. was a median of 9% energy intake for 6 weeks in substitution trials, and 8% energy intake for 5 weeks in addition trials. Honey doses in Ahmed et al. were at a median of 40 g of honey for 8 weeks

Blood pressure and vascular function

Blood pressure and vascular function play a major role in cardiometabolic health [23, 24]. Meta-analyses of sugar sources demonstrates that substitution or addition of SSBs or honey to the diet do not lower either systolic or diastolic blood pressure, whereas the addition of either fruit juice or whole fruit can lower both systolic and diastolic blood pressure (Fig. 4A and B) [15, 25]. Direct comparison of fruit sources does not provide evidence of differences between whole fruit compared with fruit juice consumption on either systolic or diastolic blood pressure [4]. Accordingly, it may be possible to achieve the blood pressure lowering effects of fruit from either fruit juice or from whole fruit.

Fig. 4

Effects of experimental addition and/or substitution of various sugar sources into the diet on systolic blood pressure, diastolic blood pressure and body mass. Data redrawn from Ahmed et al. for all outcomes with honey [15], Qi et al. for blood pressure outcomes with other food sources [25], and Chiavaroli et al. for body mass with other food sources [26]

Vascular structure and function play a key role in cardiometabolic health and blood pressure regulation. Changes in each of three layers of the artery can regulate vascular structure and/or function. These include the central and peripheral arterial stiffness of the tunica adventitia, captured by pulse wave velocity, distensibility and β-stiffness [27], smooth muscle function of the tunica media, captured by nitroglycerine-mediated dilation [28], and endothelial function of the tunica intima, captured by flow-mediated dilatation [29]. Changes in vascular function can influence cardiometabolic health in several ways. These include glycaemic control via delivery of insulin and glucose to skeletal muscle, and regulation of blood pressure via the relationship between blood flow, vascular resistance, and blood pressure.

Direct comparison of sugar sources has demonstrated some effects on markers of vascular function in healthy people. For example, consumption of 200 mL per day of orange juice for 2 weeks increased flow-mediated dilation compared with a SSB, without a detectable change in blood pressure [30]. Furthermore, consumption of 500 mL orange juice per day for 4 weeks lowered diastolic blood pressure by ∼ 5 mmHg compared with equivalent ingestion of a SSB, and that the addition of the flavonoid hesperidin (∼ 300 mg, equivalent to 500 mL orange juice) to the SSB can also lower diastolic blood pressure by ∼ 5 mmHg relative to placebo [31]. Whilst no evidence of chronic changes in vascular function were observed when measured in the overnight fasted state, acute increases in postprandial microvascular endothelial reactivity were observed with both orange juice and a hesperidin-fortified beverage versus a SSB [31]. The endothelium-dependent microvascular vasodilatory response to acetylcholine has also been shown to increase with supplementation of two whole apples per day for 8 weeks, compared with an apple juice-based control beverage, alongside a reduction in intracellular cell adhesion molecule-1 (ICAM-1), although no evidence for differences in other adhesion molecules was observed [3].

There is consistent evidence that the source of sugars can influence blood pressure and vascular function with some effects apparent within hours of consumption. Pure fruit juices, particularly orange, grapefruit and grape juice, and whole fruit show generally favourable responses, such as lower blood pressure and increases in flow mediated dilatation and microvascular endothelial reactivity which are not observed with SSBs and honey. Further improvements in endothelial function have been observed with whole fruit, yet this did not yield further reductions in blood pressure.

Appetite and energy intake

Long-term changes in body weight and fat mass primarily reflect energy balance, that is, energy intake minus energy expenditure. Therefore, effects of sugar sources on appetite and energy intake have implications for the regulation of body mass. The control of appetite and energy intake is complex and comprises many factors. However, a primary driver of energy intake is energy density [32]. Diets high in free sugars have been shown to increase self-reported energy intake [33], which is likely to be largely explained by the energy density of the diet. Recently, additional factors have been suggested to play a role such as the degree of food processing. Interestingly, according to a commonly used version of the NOVA (not an acronym) classification system, “carbonated drinks” are classified as group 4 (“ultra-processed”), whereas fruit juices are classified as group 1 (“unprocessed or minimally processed”) despite having very similar energy and sugar contents (Tables 1 and 2) [34]. Consequently, US portions of fruit juice, equivalent to 240 mL, end up providing more energy and sugars than does a portion of fruit, yet the difference is smaller with UK portion sizes (Table 2). When this classification is considered in light of some evidence that largely ultra-processed diets can increase energy intake and body mass compared with largely unprocessed diets matched for presented energy, energy density, macronutrients, sugar, sodium and fibre [35], this may have relevance for sugar sources, appetite and energy intake. Nevertheless, is should be noted that in the only current RCT of ultra-processed diets on body mass, energy density of foods was higher in the ultra-processed condition and therefore, there is a need to understand whether ultra-processed foods increase energy intake independent from energy density. Furthermore other evidence suggests that factors such as food texture and physical structure may be at least as important as processing [36, 37], and these factors - alongside energy density - are a more objective, and thus operationally useful way of characterising foods than the NOVA system [37].

When evidence from RCTs of addition of sugar calories or substitution of sugar calories in the diet are examined, the source of sugar may modulate effects on body mass. Indeed, meta-analyses indicate that addition of SSBs can increase body mass, whereas addition of either fruit juice or whole fruit can lower body mass (at least when comprising < 10% of total energy intake; Fig. 4C) [26]. It should, however, be noted that the wide confidence interval for the effect of fruit juice on body mass means this effect size should be interpreted somewhat cautiously until more evidence is available. Notwithstanding this, in direct comparisons of whole fruit with fruit juice, there is no evidence for a difference in the body mass responses with whole apples compared with apple juice a juice-based control beverage [3, 4].

One of the first studies to directly compare sugar sources on appetite found that ingestion of whole apples resulted in a higher satiety rating compared with apple puree and apple juice, although the effects were short-term and no longer apparent after 2 h [38]. Direct comparisons of fruit sources of sugars with SSBs are rare. However, when examined in isolation, SSBs preloads do not normally produce any compensation in subsequent energy intake, thereby tending to produce passive overconsumption [39]. Recent, direct comparisons of apples in different forms, such as whole fruit versus puree versus juice, consistently find that whole apples result in greater satiety ratings than apple juice [40, 41]. Furthermore, when these were tested in a preload-test meal design to assess energy intake, apple juice (with or without fibre) preloads resulted in compensation such that total energy intake (preload plus test meal) did not differ from control. In contrast, apple sauce lowered total energy intake, which was reduced further still by whole apples [40]. Consistent with this consumption of a mixture of fruits consumed in liquid (apple and grape juice) versus solid (apple, grapes, and raisins) form, results in weaker acute satiation and satiety responses, particularly in people with overweight/obesity, although differences in appetite ratings were not detected after 8 weeks of supplementation. Somewhat consistent with this, the addition of ∼ 500 kcal of fruit and vegetables to the diet for 8 weeks resulted in increases in body mass in the region of 1.5-2 kg, with the difference between consumption as liquids versus solids being 0.6 ± 14.9 kg (mean ± SD, p = 0.19) [42]. These data suggest that extrapolation from acute appetite responses to longer-term changes in body mass requires caution [43]. Taken together, these data suggest that sugar source can influence appetite and energy intake responses in the short term, while meta-analyses suggest that sugar source can make a difference to body mass. However, direct translation from acute appetite responses to longer-term changes in body mass is not warranted.

Potential mechanisms by which sugar sources influence cardiometabolic health

The potential mechanisms by which sugar source may influence the physiological responses described above could include oral processing, gastric emptying, digestion and intestinal absorption rates, sodium/potassium balance, modulation of the gut microbiome, and/or alterations in appetite-related gut hormones. The properties of sugar sources which could modulate these mechanisms include the physical structure, and the content and type of carbohydrates, fibre, polyphenols, fats, proteins, water and micronutrients within the food or beverage. These properties will be discussed in relation to the potential mechanisms which may mediate the physiological responses to sugar sources.

Oral processing, gastric emptying and intestinal absorption

The physical structure of a food (liquid versus solid, and textures of solid and semi-solid foods) affect bite size, number of chews per bite, and the duration of oro-sensory exposure [44]. In turn these responses can affect rates of eating and energy intake. Faster eating rates are associated with increases in total energy intake within a meal, which can contribute to the ways in which the physical structure of a food can influence overall energy intake. Solids are typically consumed more slowly than semi-solids and liquids (10–120 g/min versus up to 600 g/min) [45, 46], and within solid foods, harder foods are typically consumed more slowly than softer foods [47].

Gastric emptying rates could play a key role many of the effects of sugar source on health outcomes including reductions in blood glucose and increases in satiety. Slower gastric emptying can slow down the rate of nutrient delivery to the intestine and thereby contribute to slower intestinal absorption rates. In turn, this can be one mechanism by which postprandial glucose and/or insulin concentrations are lowered. However, one consideration is that, despite differences in rates of digestion and absorption, some foods can still elicit similar postprandial glucose concentrations, as increases in glucose appearance rates can be offset by increases in insulin-stimulated glucose clearance rates [48]. This may, in part, explain why the glycaemic index of fruit juices and whole fruit are reported as broadly similar despite the former being classed as a source of free sugars (Table 1). Indeed, when whole apples and apple juice were directly compared, apple juice produced a higher postprandial insulin response, in the presence of a similar peak glucose concentration [38]. Using magnetic resonance imaging (MRI) it has been shown that gastric emptying rates are slower with whole apples (half-life: ∼65 min) when compared with either apple puree (∼ 41 min) or apple juice (∼ 38 min) [41]. Unfortunately, plasma glucose kineti