Sugar restriction: the evidence for a drug-free intervention to reduce cardiovascular disease risk



Simon Thornley, Section of Epidemiology and Biostatistics, Level 4, School of Population Health, Tamaki Innovation Campus, The University of Auckland, Private Bag 92019, Auckland 1142, New Zealand.




Uncertainty exists about what dietary component is most likely to cause coronary heart disease. Over the last thirty years, attention has focused on saturated fat and salt as guilty parties. More recently, evidence suggests that excess sugar intake is more likely than either traditional factor to lead to atherosclerotic disease. Some researchers have also speculated that sugar is addictive, in a similar manner to caffeine and established drugs of abuse.


Here we review the epidemiological, biochemical and psychological evidence that implicates excess sugar intake as an important cause of ill-health.


We found relatively consistent evidence of association between markers of sugar intake and risk factors for cardiovascular disease, or the disease itself. This evidence contrasted with rather weaker evidence which linked either saturated fat or salt with cardiovascular disease endpoints. We also found some evidence of a sugar addiction syndrome.


We suggest that advice to restrict sugar intake should be a routine part of clinical care, particularly when patients are being counselled about cardiovascular risk.


Authorities in the field of nutrition have made contradictory statements about the role of sugar in the aetiology of diabetes, a condition linked to cardiovascular disease:

It is not yet clear whether any single attribute of the Western way of life is particularly important in increasing the risk of diabetes. Excess sucrose has largely been exonerated as an important dietary factor in the aetiology of type 2 diabetes … (2003).

J. I. Mann and A. S. Truswell[1]

Originally proposed as the ideal sweetener for people with diabetes … Fructose … has been indirectly implicated in the epidemics of obesity and type 2 diabetes (2009).

The American Heart Association[2]

So what can we conclude? Is sugar a problem or isn't it?

What do local authorities have to say? Institutions in Australasia charged with preventing premature cardiovascular death offer advice to reduce a person's chance of becoming ‘just another statistic’. The Australian Heart Foundation suggests the following: eat a variety of foods; include vegetables, whole grains, fruit, nuts and seeds; choose healthier fats and oils; try to limit sugary, fatty and salty takeaway meals; and drink mainly water. The New Zealand counterpart solicits several steps to a healthy heart: eat three meals a day with more plant material, and less dairy fat, meat fat or deep-fried food; eat low-fat products (lean meat); and avoid excesses of sugar or salt.

This advice seems sensible, but does it withstand closer scrutiny? Should a clinician blindly adhere to this advice? What is the evidence for all or part of the guidance? Underlying this counsel is the belief that saturated fat, sugar and salt cause coronary artery disease, or the broader term ‘cardiovascular disease’ (which encompasses stroke and peripheral vascular disease events as well). A cursory glance at such guidelines suggests that the evidence for restricting all components of food is equal and should be considered on par. If anything, saturated fat avoidance is prioritised in this advice. Here, we look closer at the biological and epidemiological evidence that excess sugar intake causes cardiovascular disease, briefly contrast such evidence with that for modifying fat and salt intake, and consider what advice should be given to patients to reduce their risk.

The varied counsel offered by Australasian heart health agencies glosses over much debate and uncertainty over nutritional exposures and their effects on heart health.The consequences of sugar intake have divided the scientific community over the past 30 years. Cleave first proposed that scientific evidence implicated excess sugar and refined carbohydrate intake in a range of non-communicable diseases, in his influential work published in 1976.[3] If one looks back over a longer period of time, some experts have drawn attention to the adverse effects of sugar consumption since at least the early 1960s.[4, 5] From this time, nutrition attention shifted to fat, and interest in sugar was largely abandoned until recently. As an example, Professor Jim Mann, an international nutrition expert, describes sugar (sucrose) as ‘exonerated’ in the aetiology of type 2 diabetes when writing in the prestigious Oxford Textbook of Medicine. The American Heart Association, which in 2002 had taken a similar view to Mann,[2] later changed its mind, reporting that fructose (half of the sucrose molecule) was now facing a guilty verdict, whereas it had earlier received a pardon.[6]


During the last 20 years, physiological properties of carbohydrates have attracted little interest due to saturated fat dominating the nutritional horizon. Present in highest concentrations in animals, saturated fats are generally solid at room temperature and have all carbon chains (fatty acids) filled (or saturated) with hydrogen. The warnings of the National Heart Foundations in New Zealand and Australia to avoid meat and dairy fat are made on the basis that such sources contain high concentrations of saturated fat. Fat is also energy-dense, carrying twice the calories of either protein or carbohydrate per unit weight. Both the presence of ‘atherogenic’ saturated fat and the high density of energy from this source have made it a favourite target of public health nutritionists seeking to improve population health.

Some properties of carbohydrates have been studied, however. In the early 1980s, the properties of glucose-containing foods were classified using the term ‘glycaemic index’ (GI). This measure is the average area-under-the-curve of plasma glucose measurements drawn immediately before, and up to 1 h after, the intake of a food standardised by the weight of the carbohydrate it contains. The index discriminates between starchy foods, made up of high volumes of linked glucose chains. Refined carbohydrates, such as white bread, white rice and pasta, are scored high on the GI (>65), while unrefined carbohydrates, such as whole grain breads and lentils, score lower (<55). Foods that are low in carbohydrate content but high in fat or protein tend to also have low GI due to the effects of the other macronutrients on gastric emptying. Also, the carbohydrate fructose, found in highest concentration in sugar, is rated low on a GI scale (∼20), simply because it counts as carbohydrate but has a small effect on serum glucose during the first hour of the postprandial phase.

What is sugar? Until recently, this term has encompassed all mono and disaccharides that are composed of either six- or five-ring sugar molecules. This is a subclass of carbohydrates that differentiates ‘sugar’ from small chains of carbohydrates (oligosaccharides) and complex carbohydrate made up of long chains of sugar molecules (starch and fibre). In the last 10 years, however, researchers have focused on the properties of one molecule: fructose. This monosaccharide is most commonly found in disaccharide form in table sugar, also known as sucrose. Table sugar consists of one glucose joined to one fructose molecule and is commonly added to a variety of manufactured foods (yoghurt, breakfast cereals, sauces, cakes, pastries and soft drinks), and to tea and coffee. Fructose is present in a restricted range of foods: fruit, honey, refined sugar and high-fructose corn syrup – a sugar-like substance derived from corn – often used as a sugar replacement in North America. Nutritionists frequently distinguish between ‘intrinsic’ sugar, monosaccharides present in fruit and vegetables, and ‘added’ sugar, which is usually sucrose, high-fructose corn syrup or glucose, used as an ingredient in the manufacture of many foods.

A historical perspective

Before we review the epidemiological and clinical evidence linking food intake with cardiovascular disease, we briefly explore the history of the modern-day ‘diabesity’ epidemic. The recent rising tide of type 2 diabetes, and the obesity epidemic, starting in the early 1980s have not occurred in a nutrition vacuum, nor has the epidemic been equally spread over all nations. If we examine the evolution of obesity prevalence in a range of developed countries, using the Organisation for Economic Co-operation and Development (OECD) data from the early 1980s to 2007, we see a stark trend (Fig. 1).[7] First, the rate of rise in obesity prevalence has occurred mostly in English-speaking countries lead by the United States, but closely followed by the UK, New Zealand, Australia and Canada. In each of these countries (except Canada), between a quarter and one-third of all adults are classified as obese. Continental European countries are the next lowest prevalence region, with the lowest levels of obesity observed in the only two Asian countries in the OECD: Japan and South Korea (<5%). We also note that the epidemic probably started in the late 1970s and early 1980s, for it was in this period that significant efforts were made to measure the prevalence of obesity in OECD countries.

Figure 1.

Adult obesity prevalence (body mass index ≥ 30 kg/m2), by year, for selected Organisation for Economic Co-operation and Development countries (1980–2007).[7]

What was the dominant change in nutrition policy that occurred in the 1960s and 1970s that predated the obesity epidemic? In the 1960s, the diet-heart hypothesis was becoming rapidly accepted and enshrined in public health nutrition messages, first in the United States. The message spread subsequently to other English-speaking countries, to increase the percentage of daily energy eaten from carbohydrate and reduce intake of animal-derived fat.[8, 9] In the late 1970s, the U.S. government's aspiration was to increase carbohydrate consumption to account for 55–60% of energy and reduce fat intake to between 30% and 40% of energy.[9] While the instigator of the hypothesis, Professor Ancel Keys and the American Heart Association warned the public to forego animal fat, a small but vocal minority argued that sugar was much more likely to cause coronary heart disease, and that this substance should be the target of public health preventative campaigns.[3-5]

We know that since this time (the late 1970s), the prevalence of diabetes and obesity has risen considerably, particularly in English-speaking countries, and that sugar intake has increased substantially over the same period. Professor of economics, Barry Popkin, used food disappearance data to document a global rise in sugar intake of about 30% over the period between 1970 and 2000.[10]

During the early 1980s, scientific exploration of the health effects of starch began to be summarised in the form of the widely known ‘glycemic index’, coined by David Jenkins at the University of Toronto.[11] The index discriminated between foods, based on their effects on after-meal serum glucose levels. In patients with diabetes, control of blood glucose has been a cornerstone of reducing the incidence of complications of the disease. In some nutrition circles, low-glycaemic diets (restricting diet to low and medium GI foods) was recommended to improve glycaemic control and also for reducing the incidence of complications of diabetes. As a consequence of such a focus on postprandial glycaemia, fructose-containing foods, such as those that contain high sugar levels, were often classified as medium GI and ‘healthy’. Because the GI of sugar (sucrose) is lower than flour and other refined sources of starch, sugar became, paradoxically, exonerated from the list of suspects as a cause of the epidemics of diabetes, obesity or coronary heart disease.[1] In fact, leading researchers of GI advocated that people with diabetes continue to eat and drink high levels of sugar to improve their postprandial serum glucose control.[12]

More recently, the tide of evidence and scientific opinion is turning in favour of sugar as a significant vascular disease risk factor, and the substance continues to attract attention as a potential cause of the epidemic of non-communicable disease of the 21st century. The changing attitude of large scientific bodies is perhaps best highlighted in the change of opinion expressed by the American Heart Association between 2002 and 2009. The organisation, charged with a mission to rid the United States of cardiovascular disease and stroke, compiled a summary statement in 2002 for health professionals about the effects of sugar on cardiovascular and metabolic health. The article concluded that there was ‘no definitive evidence’ to limit sugar intake, and that individual physicians should ‘rely on professional judgement’. The stance of the authors of the report was best summed up by the following statement:

Consuming fructose either free or in the form of sucrose has neither beneficial or adverse effects.[2]

By 2009, however, opinions had changed. The organisation concluded that the weight of scientific evidence had turned against sugar, concluding that men should consume no more than nine teaspoons per day, with six recommended for women.[6] Risk of weight gain and developing diabetes were the main concerns associated with excess sugar intake, summarised in the following statement:

Originally proposed as the ideal sweetener for people with diabetes … Fructose … has been indirectly implicated in the epidemics of obesity and type 2 diabetes.[6]

In the remainder of this paper, we consider the accumulating evidence that the fructose component of sugar is responsible for risk factors for cardiovascular disease; what factors may make sugar intake difficult to limit; and contrast this evidence with more established nutrition exposures, such as to saturated fat and salt.

Physiology of fructose

Although fructose is a carbohydrate-like glucose, it has a completely different metabolic profile. From the work done on GI, researchers have observed that fructose has almost no effect, in the short term, on postprandial glycaemia in contrast to glucose-rich foods, such as breads, rice and pasta. If fructose intake does not result in conversion to glucose, what is its metabolic fate? From small clinical and animal studies, after ingestion, fructose is absorbed from the small intestine, almost completely taken up by the liver from the portal circulation, in an insulin-independent manner. Fructose is then slowly converted to both glucose (∼50%)[13] and fatty acids, which are then released into the peripheral circulation as a combination of triglycerides and very low-density lipoproteins. During the metabolism of fructose to fructose-1-phosphate, intracellular energy is depleted (adenosine triphosphate converted to adenosine diphosphate) and uric acid is produced.[14] This contrasts with the metabolic fate of glucose and starch-containing foods in which only about 20% of the absorbed carbohydrate is taken into liver cells, with the remaining 80% taken up by skeletal and smooth muscle, and other organs in the presence of insulin. Only about 5% of starch is converted to fatty acids by liver cells, compared with almost 50% of fructose. After continued ingestion of high doses of fructose, the presence of high levels of triglycerides within liver cells results in insulin resistance, non-alcoholic fatty liver disease and long-term deterioration in glycaemic control.

Epidemiological associations between fructose intake and risk factors for cardiovascular disease

With our biochemical understanding of fructose, we might expect that fructose intake is associated with a range of conditions, such as weight gain, high levels of serum urate (and gout), diabetes, dyslipidaemia and possibly cardiovascular disease. Do the epidemiological studies support such assertions? Several exposure–disease (or risk factor for disease) associations have consistently emerged from such studies. Before we consider the evidence, however, we reflect on possible barriers that distort nutrition-related exposure–disease relationships. The first is the measurement error associated with recording what individuals actually eat. A 24-h dietary recall, considered by some to be the most accurate of all nutrition exposure assessments (although resource-intensive), in one examination underestimated sugar intake by about 20%.[15] Put simply, people have bad memories for what they eat and drink. Part of this may be explained, as we discuss in later sections, by the subconscious drives to eat so that much intake is not consciously recorded.

What effect does this inaccuracy have on exposure–disease effect estimates? Measurement error, if random, tends to lower the magnitude of these relationships and bias towards a null (or ‘no effect’) value. If the error is systematic, the effect estimate may be biased in either direction.

Another major impediment to accurate estimation of risk of nutrition exposures is the cost, complexity and ethical restrictions placed on randomising exposure to varying diets. For this reason, most nutritional studies have been observational, and lack the necessary component (randomisation) to balance unmeasured confounders between the experimental and control groups. The presence of unmeasured residual confounding is an ever-present threat to the conclusion of observational epidemiological studies (cross-sectional, cohort or case control).

With such limitations in mind, we approach the epidemiological literature to evaluate the strength of evidence of adverse health outcomes, linked to high intakes of fructose. Because fructose does not enter the diet in pure form, and has only recently been identified as a nutritional exposure likely to cause disease, many publications point the finger at exposures, which are likely indicators of high levels of fructose exposure, such as soft drink or sugar intake.

We will not attempt to fully describe the epidemiological evidence for disease from excess fructose intake; instead, we focus on summary studies, such as systematic reviews and meta-analyses. Such studies enable us to assess where the weight of the scientific evidence stands, and help us avoid possible biases created by distorted results that may emerge from individual investigators. Table 1 outlines the patterns that emerge consistently from the summary studies (where these are absent, single study results have been reported), and we refer the reader to the original articles to explore further the nature and strength of the relationship reported between markers of fructose intake and outcome measures.[16-29]

Table 1. Positive associations between indicators of exposure to fructose and important risk factors for cardiovascular disease or the disease itself
ExposureDisease or risk factor for diseaseNature of evidenceStrength of effect and comparison
  1. CHD, coronary heart disease; CI, confidence interval; HDL, high-density lipoprotein; LDL, low-density lipoprotein; RR, relative risk; SSB, sugar-sweetened beverages.
Soft drinks (SSB)Weight gain

Systematic review and meta-analysis of observational studies.[16, 17]

Randomised controlled trial[18]

SSB: correlation coefficient: r = 0.08; P < 0.001
Soft drinksGout and hyperuricaemiaCohort studies[19]

Comparing highest and lowest quintile

Multivariate RR for gout: 2.02 (1.49–2.75; P for trend <0.001)

FructoseRise in blood pressureSingle randomised crossover study[20]6.2 mmHg increase 30 min after administration of 60 g of fructose, compared with no change after 60 g of glucose
FructoseHigh levels of triglyceridesMeta-analysis of small randomised crossover studies in patients with diabetes.[21]Compared with starch; standardised mean difference 0.24 mmol/L (95% CI 0.05–0.44)
Sugar-sweetened beveragesType 2 diabetes (and weight gain)Meta-analysis of observational studies[22]1–2 SSB/day versus <1 serving/month; relative risk 1.26 (95% CI 1.12–1.41)
Caloric sweetenerDyslipidaemia (high LDL, low HDL, high triglycerides)Cross-sectional study[23]Comparing daily intakes of <5% energy with >25% energy; mean difference in HDL was 0.3 mmol/L
Sugar intakeDental cariesSystematic review of literature[24-26]Pooled effect not carried out.
Dental cariesCardiovascular diseaseMeta-analysis of observational studies[27]Compared adults with periodontal disease to those without: relative risk 1.14, 95% CI 1.07–1.21)
Serum urateCHD and all cause mortalityMeta-analysis of observational studies[28, 29]Compared adults with hyperuricaemia to those without: pooled RR for CHD 1.34, 95% CI 1.19–1.49
Sweetened beveragesCoronary heart diseaseCohort studyCompared upper quartile of consumption to lower quartile (adjusted RR 1.20; 95% CI: 1.09 to 1.33)[29]

Although individual observational studies support no association between fructose intake and weight gain or other indices associated with cardiovascular disease,[30, 31] the majority of systematic reviews support a positive association. If consistent associations are found among many studies, which use different methods, then this strengthens the evidence for causal claims.[32] We identified several systematic reviews and meta-analyses of observational studies that found evidence of positive associations between sugar intake and outcomes, such as weight gain, hyperuricaemia and gout, high levels of triglycerides, and dental caries. Dyslipidaemia, a recognised association with coronary disease and a variable used to help predict disease from popular Framingham-derived equations, was positively associated with intake of sugar-sweetened beverages. One systematic review is revealing about the nature of the study and reported effect: it showed that sugar and beverage industry-funded and cross-sectional studies were less likely to show positive associations with weight gain than longitudinal-designed studies which were independently resourced.[16] Cross-sectional studies are more open to error, such as from survivor bias and the lack of a temporal separation between exposure and disease. Case control studies suffer from the uncertainty that comes from the need to select and recruit a suitable control group. Therefore, cohort studies, which incorporate a time component, are generally considered more reliable than either cross-sectional or case control designs, and were more likely to show positive associations in our review.

Sugar intake is probably best known for causing dental caries, and this effect is relatively uncontroversial.[26] We note with interest the consistent positive association observed between dental caries and cardiovascular disease incidence.[27, 32] While most authors argue that an immunological mechanism is likely to be responsible for the association involving pathogenic bacteria present in the mouth, we argue that the association observed between rotten teeth and incident cardiovascular disease is more likely to be due to a third (or confounding) variable in the manner depicted in Figure 2.[34] At the time of writing, a cohort study appeared which reported a direct association between sugar sweetened beverage intake and incident cardiovascular disease (adjusted RR 1.20; 95% CI: 1.09 to 1.33), comparing upper with lower quartiles of intake. The study included 42 883 men with 3 683 outcomes and over 22 years of follow up.[35] Such a finding supports our idea presented in Figure 2.

Figure 2.

A plausible causal diagram that may explain the nature of the association between dental caries and coronary artery disease.[33] Solid arrows indicate proposed direction of causal relationship between variables, while dashed arrow shows apparent (epidemiological) association.

In a similar way, as proposed for dental caries, we speculate from the evidence in Table 1 that dyslipidaemia, hyperuricaemia, a diagnosis of type 2 diabetes, hypertension and obesity – all of which are themselves associated with incident cardiovascular disease – all have evidence of a statistical association with excess intake of fructose. Evidence of association is usually considered a prerequisite for making a case that an exposure causes disease, although other considerations should be taken into account.[32] Indeed, this information suggests that routine clinical measures of cardiovascular disease risk (except cigarette smoking) are assessing, indirectly, indicators of dietary fructose intake. If the major risk factors for cardiovascular disease are associated with fructose intake, this supports the hypothesis that the disease itself is caused by this upstream exposure.

Sources of fructose in the diet

How much sugar do we consume on average, and where does most of it come from? Data from the United Nations Food and Agriculture Organisation which estimate average food disappearance (annual production + import – export/the number of people in the population) suggest that we eat and drink, on average, between 30 and 40 teaspoons of ‘added’ sugar per day. In 2007, the amount of sugar and sweeteners, on average, consumed in Australia was estimated at 33 teaspoons per day, with the corresponding figures, 38 for New Zealand and 46 for the United States. Survey estimates of daily sugar intake are generally lower, which may reflect underreporting, common in nutritional survey instruments (Table 2).[36] Men generally consume more than women, with a pattern of declining intake in older age groups. Soft drinks are a major source of added sugar in the US. diet, accounting for about one-third of reported intake. Fruit juice, confectionary and breakfast cereals are other major sources (Table 3).

Table 2. Mean U.S. reported sugar intake (teaspoons/day)[34]
Age group (years)MalesFemales
Table 3. Major sources of added sugars in the U.S. diet[34]
CategoryProportion (%) of added sugar intake
Regular soft drinks33.0
Sugars and candy16.1
Cakes, cookies, pies12.9
Fruit drinks9.7
Other grains (toast and waffles)5.8

Evidence for competing dietary causes of cardiovascular disease

Although we do not focus on this issue, we draw attention to the relative paucity of evidence which indicates statistical associations between other potential causes of cardiovascular disease, such as excess saturated fat and salt intake. This is important because these exposures (saturated fat and salt) are usually accorded higher status than limiting sugar intake in dietary guidelines and public health campaigns, such as those prepared by national heart health organisations.

Briefly, systematic reviews and quantitative meta-analyses have been disappointing when assessing the link between modifying saturated fat intake and incidence of cardiovascular disease.[37] A ‘Cochrane review’ of the effect of modified or reduced fat on total and cardiovascular mortality reported no effect in the pooled analysis of randomised studies (pooled relative risk (RR): 0.98, 95% confidence interval (CI) 0.93–1.04), no reduction in cardiovascular mortality (pooled RR: 0.94, 95% CI 0.85–1.04), but an overall reduction in total cardiovascular events (pooled RR: 0.86, 95% CI 0.77–0.96).[38] Of interest is that the most objectively recorded outcomes (cardiovascular mortality and total mortality) showed no benefit, and the benefit from the overall reduction in total cardiovascular disease (CVD) events was small (14% reduction). The inconsistency of effect for the three similar outcomes was also noted.[39]

Some people in the scientific community discount these results and focus on positive associations observed between fat-modifying diets and indices of cardiovascular risk, such as adverse lipid profiles. Others focus on the outcome of reports which report subgroup analyses. One study, for example, which pooled a limited selection of cohort data for which individual-level data were available, showed a positive association between replacing saturated fat with polyunsaturated fat in reducing total cardiovascular events (pooled hazard ratio 0.69; 95% CI 0.59, 0.81) and cardiovascular mortality (pooled hazard ratio 0.57; 95% CI 0.42–0.77).[39, 40] The findings were, however, inconsistent in that replacing saturated with monounsaturated fat resulted in no association with coronary death, with a similar null result reported for replacing saturated fat with carbohydrate. If we limit our discussion to this single study, the findings raise the question of whether saturated fat is truly the causal exposure because the nature of the replacing nutrient (carbohydrate, monounsaturated or polyunsaturated fat) should have little effect on the risk of cardiovascular disease. The results, to us, are more consistent with polyunsaturated fat protecting individuals from developing disease. The fact that researchers are investigating such exposure subgroups rather than reporting overall pooled results consistent with the original hypothesis (that saturated fat causes coronary disease) suggests post hoc analysis and enthusiasm on the part of some researchers to ‘prove a hypothesis’ in the face of generally unsupportive statistical evidence. We are unclear as to why this one positive association is so widely reported when the overall picture from a range of systematic reviewers shows little support for such a statistical association, let alone a causal effect.[37, 41]

Similarly, a recent Cochrane review of randomised trials that studied the effects of salt restriction showed that, in normotensive and hypertensive cohorts, no significant decline occurred in incidence of cardiovascular disease (RR among normotensive groups: 0.71, 95% CI: 0.42–1.20; RR among hypertensive groups 0.84, 95% CI: 0.57–1.23), with an increase in mortality risk in one trial of salt restriction that enrolled subjects with congestive heart failure (RR 2.59, 95% 1.04–6.44).[42] Despite such negative statistical evidence, salt restriction remains an established part of nutritional wisdom, frequently disseminated from public health authorities.

If replacing saturated fat with other nutrients or discarding salt is a tenuous strategy to reduce cardiovascular disease risk, is there any evidence for other dietary exposures? Intake of starch and glucose may confer disease risk, and a physiological index of absorption of these nutrients is embodied in the GI. Diets that restrict intake of high GI foods show relatively consistent positive associations with reduced risk of cardiovascular disease. A comparative meta-analysis, which evaluated the evidence for effects on cardiovascular disease risk from a range of nutritional exposures, reported best evidence for a Mediterranean diet, a ‘high quality diet’, increased intake of vegetables, nuts and diets that restrict high GI foods or reduce a participant's glycaemic load.[41] The first two exposures are complex, so they make it difficult to establish a causal exposure. Nuts and vegetables tend to have low GI so their effect is consistent with the hypothesis that reducing refined carbohydrates improves CVD risk profile. One other meta-analyses (of observational studies) has similarly reported a reduction in risk of a range of chronic diseases when different levels of glycaemic food intake are compared, with the pooled ratio of risk (RR) between highest and lowest quintile groups for incidence of CVD calculated at 1.25 (95% CI: 1.00, 1.56).[43] High-level evidence, from randomised controlled trials, is unavailable for assessing the effect of such diets on CVD disease incidence. A Cochrane review of randomised studies indicated, however, that diets based on glycaemic load or index were more effective than standard (often low-energy) diets at achieving weight loss at follow-up in short-term studies (from 5 weeks to 6 months).[44]

The advice, therefore, from the Australian Heart Foundation, to eat vegetables, nuts and whole grains does have some empiric support. However, a little-known fact is that intake of fat (and protein) reduces GI (due to slowed gastric emptying), and so the advice to eat unrefined food but also reduce fat intake is somewhat contradictory when viewed from a GI perspective. If one adds advice to lower fat intake on a GI-restricted diet, the overall effect will increase GI. We have also mentioned the assumption that the ‘glycemic-index philosophy’ implicitly makes: that the health properties of a food may be determined from its short-term effect on serum glucose. As discussed, high-fructose intake, which has little acute effect on serum glucose, in epidemiological studies showed statistical associations with deterioration of long-term glycaemic control and increased risk of weight gain and being diagnosed with type 2 diabetes.

Evidence for a sugar addiction syndrome

Evidence from a completely different line of enquiry may shed light on what effect the campaign against fat intake has had on population and individual metabolic health. With evidence that sugar harms health, why do people in Western societies consume the substance in such large quantities, and why has it assumed such a dominant position in the modern diet? Most lay people are familiar with the term ‘sugar hit’ and find sugar especially palatable. Addiction theory, which has been developed to explain the obsessive use of various substances (such as tobacco and opiates), may also partially explain the global rise in sugar intakes observed over the last 30 years.[10, 45] Curiously, the best evidence for an addictive component of food includes the simple sugars, both fructose (as sucrose) and glucose.

What is addiction?

Addiction, as an entity, is often poorly understood. Nick Heather defines the syndrome as ‘repeated failures to refrain from drug use despite prior resolutions to do so’. He describes three features of addiction: (i) neuroadaptation to the substance, (ii) craving for the drug and (iii) ‘akrasia’ – failures of resolve to stop using the substance.[46] Such a definition suggests that physical dependence is a related phenomenon – associated with adaptation to a drug that is taken to prevent the onset or relieve established withdrawal symptoms. The Diagnostic and Statistical Manual-IV[47] criteria for substance use are commonly used to adjudicate addiction in the individual. These criteria, summarised, are a maladaptive pattern of substance use manifest by three or more of the following, present over the same 12-month period: (i) taking larger amounts, (ii) unsuccessful efforts to cut down, (iii) overinvestment of time, (iv) giving up important social activities, (v) continued use despite negative consequences, (vi) tolerance (greater need) and (vii) use to avoid unpleasant withdrawal symptoms. The essential features of addiction are, therefore, a combination of clinical impairment, loss of control, tolerance and a withdrawal syndrome when the substance is discontinued.

Substance use, in people with addictions, is often associated with a reward perceived as ‘a hit’ or ‘feeling high’. As well as the positive symptoms, drug use alleviates negative withdrawal symptoms, which are thought to lead to subconscious learning and deeply entrenched behaviour. Neurophysiological studies shed light on how this behaviour occurs. The loss of control that accompanies addiction is mediated, in part, by operant conditioning or instrumental learning in the mesolimbic dopaminergic pathway, which connects the nucleus accumbens with the ventral tegmental area in the midbrain. Learning by positive reinforcement involves linking an association between a behaviour and a positive reward. That then leads to the behaviour becoming subconscious (such as smoking and experiencing a ‘hit’ from nicotine). Negative reinforcement also results in subconscious learning but with the stimulus and reward reversed: behaviour that avoids negative stimuli is reinforced (such as the unpleasant withdrawal symptoms that accompany tobacco abstinence in dependent smokers). This reward is likened to the pleasure that comes from taking off tight shoes. The onset of withdrawal (such as irritability or craving) prompts the smoker to light a cigarette to temporarily relieve these symptoms. The latter mechanism is thought to be the dominant cause of automatic, learned, smoking behaviour.

Drugs that produce dependence influence dopamine concentrations in the nucleus accumbens. Balfour describes nicotine effects on two parts of the nucleus: the core and medial shell which have distinct effects on behaviour.[48] Increases in dopamine concentration in the core result in physiological reward, making the behaviour more likely (such as puffing on a cigarette). The shell, in contrast, is thought to mediate stimulus response (‘Pavlovian’) action, so that both the behaviour and associated sensory stimuli are linked by reward. Such pathways help explain why environmental cues (such as the smell of tobacco) can lead to subconscious urges to take the drug.

We have introduced the biological basis and clinical observations associated with established addictions. Does eating behaviour show similarities that suggest it too can be considered in the same light?

What are the parallels between food consumption and addiction?

First, the same neural circuits linked to reward from addictive drugs are also implicated in the physiology of appetite and hunger. In slow positron emission tomography studies, eating stimulates neural activity in the mesolimbic-dopaminergic pathway known to mediate cocaine and nicotine addiction.[49] Reduced dopamine (D2) receptor availability is strongly correlated with increased body mass index (Pearson correlation coefficient 0.71), indicating that increased dopamine levels are likely to be found in these regions of the brain. This paucity of receptors is likely to reflect the drug tolerance observed in addicted individuals. That is, increasing quantities of the substance are required to achieve a similar reward response due to the presence of fewer receptors. Low levels of mesolimbic free D2 receptors have similarly been reported in individuals addicted to cocaine, opiates and alcohol.

Observation of people experiencing drug withdrawal has suggested a link between hunger and established addictions, such as smoking. Restricting food intake increases cigarette smoking,[50] and restricting intake while trying to stop smoking is linked to an increased risk of relapse.[51, 52] Smoking acutely reduces hunger,[53-55] and in some studies, decreased desire for and consumption of sweet-tasting foods.[56] Hunger is also a symptom of nicotine withdrawal, and people may gain an average of 7 kg after stopping smoking.[57, 58]

When frequent users of drugs, such as alcohol, tobacco and cocaine, abstain from their use, they experience a cluster of symptoms known as withdrawal. When obese people abstain from sugary or high GI food, do they experience such symptoms? Opiate withdrawal is characterised by nausea, stomach cramps, muscular spasm and twitching, feelings of coldness, and heart pounding. What is common to many forms of withdrawal is the time-course. In the case of unassisted or cold-turkey withdrawal, the symptoms peak in 48–72 h and largely evaporate after 30 days if abstinence is maintained.[59] For nicotine, the withdrawal syndrome that accompanies abstinence is well described.[58] Common symptoms include urges, craving, reduced concentration, irritability, increased appetite and depressed mood. Symptoms usually peak in the first few days after quitting and largely subside after a month of continuous abstinence. We know relatively little about whether an equivalent withdrawal syndrome exists for food. Although carbohydrate craving has been defined,[60] a withdrawal syndrome has not been similarly described. However, some brief case reports exist, often present in the lay literature. For example, Atkins[61] described a real estate executive who was unable to lose weight despite use of emetics and laxatives, or even undergoing obesity surgery:

The executive recalls ‘often I would shake until I could put some sugar in my mouth’. Cues are also described – ‘I had an hour's drive from my office to my home, and I knew every restaurant, every candy machine and every soft drink dispenser along the whole route’.

This may be an extreme case, but the patient described tremor linked to short-term restriction from sugar. Such symptoms mirror some of those experienced after acute opiate abstinence.[59] The executive recounted relief from eating or drinking sugar, along with cues to eat encountered on his way home from work. The increased attention paid to such cues is typical of addiction and abstinence. Although formal definition of sugar or carbohydrate withdrawal, to our knowledge, does not exist, a case report has been published. The features of the syndrome were tremors, headaches and abdominal pains that lasted about 1 month after abstinence from ‘processed sugar and flour’, which were voluntarily removed from the correspondent's diet.[62] Symptoms were reported to be most intense during the third day after abstinence.

Given the ubiquity of sugar and carbohydrates in modern western diets, obese people may not experience withdrawal symptoms unless prolonged abstinence (>48 h) is tried. Instead, the experience of obese individuals may be subtle symptoms similar to those of caffeine withdrawal, such as irritability, poor concentration and urges that accompany short-term abstinence from sugary food. These may be underrecognised, as is the case in addictions such as tobacco, where often the threat of such symptoms prompts subconscious drives to perform the addictive behaviour.[48]

Some researchers advocated that obesity be considered a type of addiction. Volkow,[63] for example, analysed the similarities between the neural mechanisms underlying obesity and drug addiction, and argued that obesity be included as a specific subtype of the psychiatric disease: substance abuse.[64]

The main determinants of an addictive substance are its concentration and speed of absorption (time-to-hit). Given that foods with added sugar are likely to have higher concentrations of fructose than natural sources, we speculate that manufactured sugary foods and drinks (which take less time to consume) are most addictive.


What do we conclude from this survey of the evidence? First, sugar intakes have increased substantially against a nutritional backdrop that has focused on reducing fat intake and salt to reduce the incidence of cardiovascular disease. Second, excess intake of fructose, due to the accumulating, consistent epidemiological evidence of links with risk factors for cardiovascular disease, suggests that substantial health gains will result from limiting intakes. Third, from the parallels among drugs of abuse, overeating and carbohydrate addiction, we speculate that many patients will find it difficult to limit their intake of sugar due to stimulation of reward pathways in the brain, and the experience of unpleasant withdrawal symptoms that accompany attempts to restrict intake.

The American Heart Association has published guidelines that suggested limiting intake of sugar to no more than six teaspoons per day for women and nine for men.[6] From food disappearance data, average daily sugar consumption is between 30 and 40 teaspoons per day in English-speaking countries, such as the UK, United States, Canada, Australia and New Zealand. The implications of the advice are enormous: most adults should reduce their intake by between 1/6 and 1/3 of their current consumption. As we have shown, the largest source of added sugar in the United States comes in liquid form, either from soft drinks and fruit juice which may be overlooked by patients.

One of the authors (RT) has considerable experience of advising patients how to cut down their intake of sugar. He suggests making patients aware of their intake by translating weight (grams), which is often reported on the nutrition panels on manufactured foods, into teaspoons. Four grams of sugar is about 1 teaspoon. When patients understand how many teaspoons are in commonly consumed food portions, they are often surprised. Many people are taken aback when the sugar content of soft drink, fruit juice, breakfast cereals and seemingly healthy sweetened yogurts is revealed. For the clinician advising people to cut down their intake of sugar, we recommend first advice about how to reduce intake of added sugar. This includes fruit juice, soft drink, cordials, sweetened yoghurts and breakfast cereals, as well as the better understood sources in chocolate, sweets, desserts, cakes and biscuits.

From the published evidence of a likely sugar withdrawal syndrome, we also suggest warning patients that they are likely to suffer withdrawal symptoms when they attempt to restrict their sugar intake. Such symptoms are likely to include irritability, loss of concentration, hunger, craving for sugar and restlessness. Cues left around the house, such as the presence of available sugary foods, arelikely to prompt consumption especially in the early phases (<1 month) of restriction. We, therefore, suggest removing sugary foods from the house and work environment, reducing the chance that the patient's resolve to forego sugar will be broken.

This paper suggests a deviation from widely accepted practice for many cardiologists, general physicians and family doctors concerned with reducing the CVD risk of the patient that they have before them. Rather than reaching for the prescription pad, we suggest a brief conversation about the perils of a high-sugar diet and practical advice about how to cut down.

Conflicts of interest

The authors have declared no potential conflicts of interest. No funding was received for the preparation of this manuscript.