Introduction

Of the numerous psychoactive compounds that humans ingest, none is more popular than caffeine. Indeed, caffeine is unusual amongst psychoactive compounds in being part of the daily diet of most people. With more than 80% of people worldwide consuming caffeine daily [95], current usage transcends almost every social barrier, including age, gender, geography, and culture. The popularity of caffeine exceeds that of any other psychoactive substance, whether it is nicotine, alcohol, or illicit drugs. Caffeine occurs naturally in a number of plant species, where it serves as a toxin to defend against herbivores. A common but erroneous belief, sometimes implied in advertisements for caffeine products, is that caffeine has always been widely present in the human diet. In fact, it was not until after European colonization in the seventeenth and eighteenth centuries that caffeine products, previously unavailable to most people, became widely accessible. That is, the ubiquitous presence of caffeine in the human diet is a phenomenon of fairly recent origin.

The aim of this chapter is to provide an overview of caffeine and its use, with particular attention being given to consequences for health and well-being. In that context, the emphasis throughout is on dietary use, taking account of both acute and chronic effects. Following relevant background, including mention of the main sources of caffeine and prevailing patterns of usage, attention is given to the pharmacology of caffeine, including the main mechanism of action and the key processes of physical dependence and tolerance. This is followed by a discussion of the psychopharmacology of caffeine, with particular attention being given to effects on psychomotor performance and mood, and the processes of withdrawal and withdrawal reversal. The remainder of the chapter deals mostly with the health consequences of dietary caffeine, beginning with mental health and well-being. That section is followed by two separate sections dealing with physical health, the first of which is concerned with cardiovascular disease, and the second with cancer, maternal use, and potential adverse interactions between caffeine and other drugs. Questions as to whether caffeine is addictive and whether there is a level of consumption that may be considered safe are examined, and processes for reducing and quitting caffeine consumption are reviewed. In the section thereafter, attention is given to emerging interests in potential health benefits of caffeine products, especially in relation to Type 2 diabetes mellitus and Parkinson’s disease, and the growing interest in compounds other than caffeine in caffeine beverages. The section preceding the conclusions considers processes that threaten the integrity of caffeine science, a topic that to date has received far too little attention.

Main Sources of Caffeine and Patterns of Consumption

The main dietary sources of caffeine are tea and coffee beverages, and increasingly, soft drinks (e.g., colas) and energy drinks. The tea plant is indigenous to regions of China, South Asia, and India. Written accounts in China of tea leaves being used to brew a beverage date to as early as 350 A.D., and by about 600 A.D. tea had been introduced to Japan from China. It is unclear, however, to what extent tea was consumed by the general population of either country during these early periods. In the seventeenth century, the Dutch introduced tea to Europe and America, and today tea is cultivated commercially in about 30 countries. Coffee is indigenous to Ethiopia from where it was transported for cultivation to Arabia in the fifteenth century. By the early sixteenth century, the practice had been established in the Islamic world of extracting caffeine by infusing ground roasted beans. The Dutch brought coffee plants to Europe in the early seventeenth century, and established plantations in the Dutch East Indies. Subsequent colonization by other European powers led to new and extensive plantations being established in the West Indies, Latin America, Africa, and India.

By the late-eighteenth century, coffee replaced tea in popularity in the United States, and today coffee is the main source of caffeine globally. Tea continues to be consumed more widely, but qualifies as the second main source because its caffeine content is generally lower than that of coffee. Other common sources of caffeine include cocoa and chocolate (in both solid and beverage form), but the caffeine content of these is generally low and represents a negligible fraction of the total amount of caffeine consumed. In addition, although the daily intake of caffeine from sources specific to particular regions (e.g., maté in parts of South America) may be substantial for individual consumers, the overall intake from such sources is small relative to total global consumption of the drug. Similarly, some medications, both prescribed and over-the-counter, contain as much as 200 mg (approximately 2–4 cups of coffee or tea) per tablet or capsule, and could be an important (even the main) source of caffeine for some individuals. For the general population, however, caffeine-containing medications are typically taken intermittently, or not at all, thereby contributing little to total population caffeine intake. Notwithstanding variations in per capita consumption between geographic regions, intake for the majority of consumers ranges from about 200 to 400 mg of caffeine per day (the approximate equivalent of 2–6 cups of coffee or tea per day).

Caffeine soft drinks are an increasingly important source of the drug, and often the main source for children. The more recently developed so-called “energy” drinks are also increasing in importance as a source of caffeine for young people. Whereas the caffeine in sodas and energy drinks sometimes partly derives from plant products involved in manufacture (e.g., cacao, cola nut, guarana), most of the caffeine content of such drinks is added in refined form. That is, these products, which are targeted primarily at children and adolescents, are explicitly designed to be psychoactive. The seeming inexorable growth in the consumption of caffeine by children has become a cause for concern in its own right (e.g., [72]) as well as giving rise to concerns that caffeine in the form of sodas and energy drinks may serve as a gateway to increased use of other drugs.

Although consumption patterns relating to the various main sources of caffeine may change during the lifespan (e.g., an individual may switch from drinking sodas during childhood to coffee in adulthood), exposure to caffeine is essentially lifelong for the majority of people. Indeed, the first exposure for most people precedes birth. Caffeine crosses the placenta [17, 236], and because most women consume caffeine while pregnant, the majority of newborns show pharmacologically active levels of plasma caffeine [36]. Exposure typically continues during childhood, with patterns of use tending to consolidate during adolescence and early adulthood. Thereafter, usage tends to stabilize, generally undergoing little change for the remainder of life [95]. The unparalleled prevalence of caffeine use introduces multipliers in relation to the possible impact of the drug. At the individual level, lifelong use could lead to effects accumulating over the lifespan. Furthermore, considering the near-universal use of caffeine, individual effects, even if small, could have a substantial cumulative impact when assessed across entire populations.

Pharmacology of Caffeine

Caffeine belongs to a family of purine derivative methylated xanthines often referred to as methylxanthines or merely xanthines. At room temperature, caffeine is a white odorless powder with a bitter taste [238]. Caffeine was first isolated from green coffee beans in 1820 by Ferdinand Runge in Germany, and later was found to be present in a variety of other species (e.g., tea, mate, cacao). Figure 1 shows the structure of caffeine (1, 3, 7-trimethylxanthine) and the three dimethylxanthine primary metabolic products of caffeine in humans. Following oral ingestion, caffeine is rapidly absorbed into the bloodstream from the gastrointestinal tract [7]. Approximately 90% of the caffeine contained in a cup of coffee is cleared from the stomach within 20 min [25], and peak plasma concentration is typically reached within about 40–60 min [177].

Fig. 1
figure 1

Caffeine and its dimethylated metabolites in humans (arrow widths indicate approximate relative proportions of the metabolites in plasma)

Full size image

Once ingested, caffeine is readily distributed throughout the body, and the concentrations attained in blood are highly correlated with those found in the brain, saliva, breast milk, semen, amniotic fluid, and fetal tissue [95]. The drug has an elimination half-life of about 5 h in adults [171], and typical consumption patterns of 3–4 doses (e.g., cups) per day, result in plasma concentrations that remain at pharmacologically active levels for most of the waking hours. In adults, caffeine is virtually completely transformed by the liver, with less than 2% of the ingested compound being recoverable in urine [214]. Although the beverages and foods that contain caffeine may have other constituents (e.g., sugar, milk) that possess nutritional value, it should be noted that caffeine itself has no nutritional value.

Main Mechanism of Action

Caffeine exerts a variety of pharmacological actions at diverse sites, both centrally and peripherally, which are generally believed to be due mostly to competitive blockade of adenosine receptors [37]. Adenosine is a neuromodulator that acts on specific cell-surface receptors distributed throughout the body [19, 151, 197, 244]. Due to similarities in the molecular structure of caffeine and adenosine, caffeine occupies adenosine receptor sites, with A1 and A2A receptors appearing to be the primary targets. Table 1 summarizes some of adenosine’s main actions, which are generally to inhibit physiological activity. At typical dietary levels of intake, caffeine blocks adenosine receptors, producing effects broadly opposite to those summarized in Table 1 [14, 20, 53, 135]. It appears, also, that A1 and A2A receptors may interact in functionally important ways with dopamine receptors [44, 59]. In particular, A2A receptors may be involved in the control of the dopaminergic signaling system essential to motor control [31]. In addition, caffeine has been reported to stimulate neuroendocrine activity, especially the catecholamine stress hormones of epinephrine and norepinephrine (e.g., [129]). Increases in serum cortisol and/or urinary cortisol metabolites have also been reported [123, 142, 143, 173, 174]. However, findings have not been entirely consistent in that some investigators have found cortisol levels to be unresponsive to caffeine [121, 165]. It may be that the inconsistencies indicate that the typical challenge of about 250 mg (2–3 cups of coffee) represents a “borderline dose” to which some people may be unresponsive. For example, in one study, 250 mg of caffeine had no effect, whereas 500 mg increased plasma cortisol levels [216].

Table 1 Some acute biological effects of adenosinea
Full size table

Physical Dependence

Repeated use of caffeine, such as occurs in the context of dietary use, generally leads to the development of physical dependence, evidenced by the appearance of behavioral, physiological, and subjective “withdrawal” effects provoked by abrupt cessation of use [111]. Although incompletely understood, the mechanism responsible for caffeine physical dependence is believed to involve adenosine. Repeated exposure to caffeine, including dietary use, is thought to result in an increased number of adenosine receptors and/or enhanced affinity, resulting in hypersensitivity during abstinence [14, 168, 242]. Sleepiness, lethargy, and headache are common symptoms of caffeine withdrawal in humans [40, 58, 80, 97, 122, 127, 172, 218, 220, 234], and cessation of as little as 100 mg (1 cup of coffee) per day, and possibly considerably less can produce symptoms (e.g., [140, 207]). These may be felt within about 12–16 h, with a peak at around 24–48 h, generally abating within 3–5 days, and only infrequently extending for up to 1 week [67, 81, 82]. Notably, studies show that decreases in psychomotor performance (not necessarily discernible to the individual) are detectable after as little as 6–8 h since caffeine was last ingested [73].

Tolerance

Drug tolerance refers to the progressive reduction in responsiveness which sometimes accompanies repeated exposure to a drug. It is evidenced by a decline in efficacy, whereby the same drug dose has less effect following repeated use or an increased dose is required to produce effects previously experienced. Although caffeine tolerance has been shown in relation to the locomotor stimulant effects of the drug in rats [46, 79], there have been relatively few empirical demonstrations of caffeine tolerance in humans. One focus of attention in relation to caffeine tolerance in humans has been the drug’s cardiovascular effects [34, 90, 91], which it is widely believed undergo tolerance. The most often (and frequently, only) cited source for the claim of hemodynamic tolerance is a study by Robertson and colleagues [181], which is widely misquoted as having demonstrated complete hemodynamic tolerance to dietary caffeine. James [89] (pp. 111–113) has shown that the Robertson et al. [181] study did not demonstrate complete tolerance to caffeine, and that due to its many methodological shortcomings the study could not have demonstrated complete tolerance. On the contrary, as discussed in more detail below, empirical evidence from diverse sources converges to show that blood pressure remains reactive to the pressor effects of caffeine despite repeated exposure such as that which occurs when caffeine is part of the daily diet (e.g., James [90, 91, 99]).

Overall, it appears unlikely that complete tolerance occurs in relation to most effects arising from typical patterns of caffeine consumption. Importantly, the response magnitude to successive doses of the drug is generally inversely proportional to plasma caffeine level [211, 212]. Also, it is notable that overnight abstinence, which characterizes usual patterns of consumption, results in almost complete depletion of systemic caffeine by early morning [138, 171, 201]. Several lines of inquiry suggest that the pattern of diurnal depletion of systemic caffeine experienced by most consumers contributes to tolerance, if it occurs at all, typically being partial rather than complete. Indeed, the very fact that many hundreds of published experiments have reported significant caffeine-induced behavioral, physiological, and subjective effects provides strong evidence that usual patterns of consumption do not produce complete tolerance. Most participants in such experiments have been typical caffeine consumers who arrive at the experimental laboratory following a brief period of abstinence. Notwithstanding the brevity of the typical period of abstinence (e.g., overnight) employed in experimental studies of the acute effects of caffeine, participants are generally observed to be caffeine responsive. As is discussed in the following section, some caffeine-induced responses (especially enhanced performance and mood) are attributable to withdrawal reversal. However, other responses, particularly increased blood pressure, are not attributable to withdrawal reversal. By definition, any observed caffeine-induced effects not attributable to withdrawal reversal provide proof positive that tolerance, if it has developed at all, cannot have been complete.

Psychopharmacology of Caffeine: The Critical Processes of Caffeine Withdrawal and Withdrawal Reversal

The earliest systematic examinations of the psychopharmacology of caffeine were conducted about a century ago [77, 78]. The strong consensus for most of the intervening period has been that caffeine is a stimulant capable of enhancing aspects of human psychomotor performance and mood. In recent years, however, that traditional view has been essentially disproved. Recent advances in knowledge about the dynamics of caffeine withdrawal and withdrawal reversal have radically transformed our understanding of caffeine psychopharmacology. In a typical study, behavioral and psychological outcomes are measured in healthy volunteers before and after double-blind administration of caffeine and placebo, and (compared with baseline and placebo) changes have often been reported in post-caffeine outcomes. This has been particularly evident in studies of performance and mood, wherein it has often been concluded that caffeine has enhancing properties. However, a critical appraisal of the typical study design shows that the findings yielded by such studies are, at best, ambiguous [92, 94, 105].

Paralleling the time-honored practice of placebo-controlled studies of medications, caffeine is typically withheld for a period prior to testing for effects, with the aim of ensuring all participants are equivalent in systemic drug levels at time of testing. Such efforts to achieve experimental control appear especially relevant to the assessment of caffeine effects, because the drug is used daily by most people. Typically, caffeine is consumed in separate portions throughout the day, with fewer portions consumed later in the day, followed by overnight abstinence [95]. With the half-life of caffeine in healthy adults being approximately 5 h [171], overnight abstinence usually leads to complete or near-complete elimination of systemic caffeine by early morning [138, 139]. Consequently, when employing the placebo-controlled paradigm, caffeine researchers have frequently made use of naturally occurring overnight abstinence by asking participants to forgo their usual morning caffeine beverage prior to laboratory testing.

What has not been fully appreciated until recently is that, having avoided caffeine since the evening before, study participants are generally entering the early stages of caffeine withdrawal by the time they are tested in the laboratory (typically, at least 12–14 h since caffeine was last ingested) (see [92] and [105]). As mentioned above, habitual use of caffeine produces physical dependence, evidenced by the appearance of readily measurable withdrawal symptoms following periods of abstinence (e.g., Juliano and Griffiths [111]). Thus, the crucial question is: To what extent do effects (e.g., enhanced performance and mood) generally attributed to caffeine represent genuine net effects of the drug or reversal of withdrawal effects induced by short periods of abstinence? [92]

Performance and Mood

The fact that caffeine is consumed daily by most people as part of a “normal” diet presents formidable methodological obstacles when trying to accurately isolate the net effects of the drug. Although the problem was largely ignored for decades, systematic attempts have begun to tackle key methodological challenges posed by caffeine withdrawal and withdrawal reversal. Approaches have varied, but generally fall into three broad categories, consisting of studies that compare consumers and low/non-consumers, pre-treatment and ad lib consumption studies, and long-term withdrawal studies [95, 105]. The first two approaches (studies comparing consumers with low/non-consumers and pre-treatment/ad lib consumption studies) have been shown to involve substantial limitations (for a discussion see James [95] and James and Rogers [105]). In contrast, the third approach (long-term withdrawal) has proven successful. This has entailed taking the core features of the traditional drug-challenge paradigm, with its attendant strengths of double blinding and placebo control, and extending them to include alternating periods of daily caffeine use and non-use (abstinence).

Table 2 summarizes the core design features of an experimental paradigm employed successfully by James and colleagues (e.g., James [88, 90, 91, 97] and Keane et al. [117]) to elucidate caffeine’s net effects using “long-term” withdrawal. During caffeine phases of that paradigm, participants ingest the approximate equivalent of 1 cup of coffee three times daily, thereby simulating the typical population pattern of caffeine consumption. The protocol employs six consecutive days of placebo/caffeine intake to achieve stability of responding before “challenging” participants on the 7th day of each alternating 1-week period. The 1-week time frame was chosen on the grounds that studies of caffeine tolerance in humans have generally found that effects plateau within 3–5 days of continuous use [34, 90, 91, 181]. In addition, there is a strong body of evidence showing that withdrawal effects generally abate within a similar time frame of 3–5 days (e.g., [66, 81]). The full research design, as shown in Table 2, offers the substantial benefit of being able to examine and compare the separate acute and chronic effects of caffeine in the one experiment. An abridged version of the design has also been used, consisting of the “PP” and “CC” conditions outlined in Table 2 without the “PC” and “CP” conditions (e.g., James and Gregg [100, 101] and James et al. [102]). While not elucidating the more detailed processes of withdrawal and tolerance, the abridged design allows key questions concerning caffeine’s net effects to be addressed.

Table 2 Summary of a double-blind placebo-controlled crossover protocol incorporating alternating periods of “long-term” caffeine exposure and abstinencea
Full size table

Long-term caffeine withdrawal studies have provided strong support for the withdrawal reversal hypothesis in relation to performance and mood [105]. That is, overnight caffeine abstinence has been found to be detrimental to performance and mood, with these adverse effects being removed when caffeine is re-ingested (restoration due to reversal of withdrawal effects). Importantly, recent studies have yielded consistent evidence of caffeine having little or no net beneficial effect on performance and mood under conditions of sustained caffeine use versus sustained abstinence [97, 101, 102]. Several other studies, which may not all strictly qualify as “long-term” studies, have reported similar results in relation to performance and mood in adults [98, 110, 180, 184, 185] and children [72].

Sleep and Wakefulness

Former strong beliefs about caffeine being capable of enhancing psychomotor performance and mood are matched by equally strong beliefs that caffeine is effective in reversing negative effects of sleep loss [103]. Until very recently, however, studies of caffeine and sleep failed to take account of the processes of withdrawal and withdrawal reversal. Employing the abbreviated version of the experimental paradigm summarized in Table 2 (i.e., “PP” versus “CC” as defined in the table), James et al. [102] examined the effects of dietary caffeine in healthy volunteers who alternated weekly between placebo and caffeine and who were either rested or deprived of more than 50% of their usual nighttime sleep on the evening before testing. Confirming previous studies, caffeine was found to have no significant net enhancing effects for either performance or mood when participants were rested, while also having no net restorative effects when performance and mood were negatively affected by sleep restriction. Indeed, James and Gregg [101] found that caffeine exacerbated the marked adverse effects of sleep restriction on mood.

Similarly, after controlling for caffeine withdrawal effects, Rogers et al. [185] found that cognitive performance was unimproved by caffeine in participants who were sleep restricted. Acute (overnight) caffeine withdrawal was found to impair performance on tasks requiring sustained attention, and subsequent caffeine intake merely prevented further deterioration in performance (withdrawal reversal). In contrast, the significantly better levels of performance on the same tasks shown by long-term (3 weeks) withdrawn participants were not improved by caffeine. Additionally, acute caffeine withdrawal had a variety of negative effects on mood. More recently, Keane et al. [117] examined the effects of caffeine on patterns of electroencephalographic activity in a rare example of a study of electroencephalography in which caffeine withdrawal and withdrawal reversal were controlled. While again finding little evidence of positive stimulant effects, Keane et al. [117] found some similarities in effects on brain activity following caffeine ingestion (challenge) and acute caffeine withdrawal. As such, these findings are consistent with results from studies of performance and mood in which caffeine withdrawal and withdrawal reversal had been controlled. That is, rather than having positive stimulant effects, a change in drug state, whether in the form of acute caffeine challenge or acute caffeine withdrawal, may disrupt normal electrophysiological activity in the brain, which may in turn be the substrate for the observed negative effects on performance and mood.

The terms “sleep” and “wakeful” lack precise definition, and are sometimes used as if they were exact antonyms of one another. Possibly everyone, however, has had the experience of being both sleepy and wakeful (i.e., tired but unable to sleep, for example, during periods of acute worry). This should not be surprising, since it is unlikely that a single mechanism controls the processes of sleepiness and wakefulness. As such, caffeine may directly interfere with an aspect of sleep (e.g., block receptors in the adenosine mechanism) and thereby forestall sleep without necessarily or appreciably benefiting wakefulness. At the same time, sleepiness is a reliable effect of even brief periods of caffeine abstinence.

One source of confusion concerning caffeine’s putative anti-soporific effects is the fact that withdrawal-induced sleepiness is reversible by ingesting caffeine, thereby creating the illusion that caffeine is effective in “stimulating” wakefulness and overcoming sleepiness. In reality, the overall effect of caffeine on the sleep cycle is likely to be disruptive, involving an increased risk of caffeine-induced sleep delay and withdrawal-induced periods of sleepiness. The former, caffeine-induced sleep delay, is possibly largely avoided by the majority of consumers who typically do not ingest caffeine after early evening. In contrast, although sleepiness induced by caffeine withdrawal is possibly widely experienced, most people are probably unaware of it (i.e., unaware of caffeine withdrawal as a cause of daytime sleepiness). Indeed, there is a strong possibility, yet to be verified, that sleepiness induced by caffeine withdrawal is a common, though largely unrecognized, cause of fatigue-related traffic and industrial accidents.

Mental Health and Well-Being

Major systems of medical and psychiatric diagnosis give formal recognition to “disorders” of psychological function arising from caffeine misuse, noting that “misuse” in this context includes levels of use falling within the range seen in the general caffeine-consuming population. Since formal diagnoses can only be made after affected persons come to the attention of relevant professionals, it follows that a sizable proportion of the general public may be engaging in caffeine “misuse” even if a formal diagnosis has not been made. The 10th revision of the International Statistical Classification of Diseases and Related Health Problems [252] has a specific diagnostic classification of mental and behavioral disorders due to use of “other stimulants”, including caffeine, which includes subcategories of acute intoxication, dependence syndrome, and withdrawal state. Similarly, under the label of caffeine-related disorders, under the broader rubric of substance-related disorders, the Diagnostic and Statistical Manual of Mental Disorders, 4th edition, text revision [5] has classifications for caffeine intoxication, caffeine-induced anxiety disorder, and caffeine-induced sleep disorder.

Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, Text Revision

Considering the Diagnostic and Statistical Manual of Mental Disorders, 4th edition, text revision more specifically, the essential features of caffeine intoxication are shown in Table 3, which cites recent consumption of caffeine and five or more symptoms that develop during, or shortly after, caffeine use. As the name implies, the classification of caffeine-induced anxiety disorder refers to the occurrence of symptoms of anxiety (e.g., nervousness, worry, apprehension) associated with, and believed to be precipitated by, the consumption of caffeine. Caffeine-induced sleep disorder typically refers to insomnia (e.g., increased sleep latency, decreased sleep time, fragmented sleep) provoked by caffeine consumption. However, as explained above, periods of reduced caffeine intake or abstinence can also lead to bouts of sleepiness (hypersomnia). Thus, on the one hand, caffeine-induced sleep disorder refers to nighttime wakefulness, which many people may recognize as having experienced. On the other hand, caffeine-induced sleep disorder also refers to the occurrence of withdrawal-induced daytime sleepiness due to caffeine abstinence or reduced caffeine intake. Again, as suggested above, given that caffeine-induced nighttime insomnia is easily avoided by not consuming caffeine latter in the day or evening, withdrawal-induced daytime sleepiness is possibly a more common occurrence, even if (or possibly because) it is less often recognized by consumers as a symptom of their caffeine use.

Table 3 Diagnostic criteria for caffeine intoxication
Full size table

The Diagnostic and Statistical Manual of Mental Disorders, 4th edition, text revision [5] also makes reference to a more general category of caffeine withdrawal, but refers to it as a syndrome under consideration and not yet a formal diagnosis. As pointed out by James (1997), this is an ironic position, because there is substantially more empirical evidence for the existence of caffeine withdrawal as a specific syndrome, than for any of the accepted caffeine diagnoses. Furthermore, the proposed diagnosis of caffeine withdrawal includes headache as a defining symptom. In reality, headache is a common, though not universal, symptom of caffeine withdrawal (e.g., [101]). By excluding cases of caffeine withdrawal headache not accompanied by other symptoms, and cases in which other withdrawal symptoms (e.g., lethargy, tiredness, irritability) are experienced without headache, the proposed diagnosis of caffeine withdrawal seems bound to lead to under-diagnosis.

The symptoms listed in Table 3 fall into two broad categories considered to be of lesser or greater seriousness. Less serious symptoms include restlessness, nervousness, excitement, insomnia, flushed face, diuresis, and gastrointestinal complaints, and may occur following daily use of as little as 100 mg of caffeine (about 1 cup of coffee). More serious symptoms include muscle twitching, rambling flow of thought and speech, tachycardia or cardiac arrhythmia, periods of inexhaustibility, and psychomotor agitation, said to occur at levels of intake of 1 gram or more per day. Although substantially above average dietary levels, consumption at this higher level of intake is not rare, possibly involving about 10% of the population. Indeed, Hughes et al. [83] interviewed 162 randomly selected caffeine users and concluded that 7% met the criteria for caffeine intoxication. Among those who had tried to stop caffeine permanently, 24% satisfied the criteria for caffeine withdrawal. The Diagnostic and Statistical Manual of Mental Disorders, 4th edition, text revision [5] stipulates that reported symptoms must cause clinically significant distress or impairment in social, occupational, or other important areas of functioning. These stipulations, however, remain open to interpretation. Thus, of the many caffeine-induced dysphoric effects that are experienced in the population, the relative proportions that fall above and below the threshold of clinical significance remain unknown.

The Epidemiology of Caffeine Disorders

Given what appears to be a relatively low level of awareness of caffeine-induced dysfunction in the general as well as professional communities, there is a strong suspicion that caffeine “disorders” remain substantially undiagnosed despite the existence of formal diagnostic protocols. Moreover, in addition to clinical diagnosis, caffeine ingestion and withdrawal appear to have a variety of other commonplace psychological and behavioral outcomes that can be serious. For example, as well as the possibility that withdrawal-induced sleepiness contributes to fatigue-related accidents, caffeine-induced hand tremor has been found to undermine surgical precision. In a double-blind placebo-controlled crossover study, Urso-Baiarda et al. [229] found that moderate amounts of caffeine had a detrimental effect on microsurgical ability due to the adverse effect of the drug on hand steadiness. Also within a general surgical context, evidence indicates that patients commonly experience perioperative caffeine-withdrawal headache due to the requirement that they fast (and therefore do not receive their usual caffeine intake) prior to anesthesia [41, 57, 162]. Prophylactic administration of caffeine appears to provide a simple and effective remedy [69, 245]. Indeed, the reversal of headache under such circumstances is further evidence of the role of caffeine withdrawal in the development of headache. Findings such as these contribute to the impression that the population prevalence of caffeine-induced disorders far exceeds that which would be implied by the frequency with which such problems are diagnosed in clinical settings.

Dietary Caffeine and Physical Health: Cardiovascular Disease

When considered in totality, the large and diverse body of relevant scientific literature is conclusive in pointing to adverse acute effects of caffeine on cardiovascular function, especially blood pressure. Though evidence of chronic effects is less conclusive (for reasons outlined below), the available evidence nevertheless provides strong grounds for concluding that dietary caffeine is a significant factor in the development of cardiovascular disease. The extent of the evidence is such as to suggest the need for primary prevention at a population-wide level, including appeals to consumers to avoid caffeine in the interests of cardiovascular health. Such action, however, has been largely absent and it is important to examine possible reasons. Accordingly, this section provides an overview of relevant experimental and epidemiologic findings, and considers reasons why the evidence may not have caught the attention of health authorities to the extent that it should. Two main reasons for this neglect, considered below, appear to be: confusion regarding the epidemiology of caffeine and cardiovascular disease due to exposure misclassification, confounders, and possible misunderstanding of putative threshold effects; and the belief that habitual caffeine use leads to the development of tolerance to the cardiovascular effects of the drug.

Concerns for cardiovascular health in the context of dietary caffeine have a firm foundation in demographics. Cardiovascular disease is the major cause of mortality and morbidity in the developed countries of the world accounting for approximately half of all deaths [233, 250, 251]. It is expected also that within the near future cardiovascular disease will be the major cause of death and disease throughout much of the developing world [169, 182, 183]. As well as being of high prevalence, cardiovascular diseases are generally of long latency, and have complex multifactorial causation involving lifestyle variables including diet. As mentioned above, the prevalence of dietary caffeine is extremely high and essentially lifelong. In that context, it is especially noteworthy that adenosine has an important role in the regulation of cardiovascular function as well as being the main mechanism of action for caffeine, thereby providing strong biological plausibility for a possible link between the two. Blood pressure level is of particular concern, because it is the single most important predictor of cardiovascular disease [156, 182], and compared with several other key indices of cardiovascular function blood pressure is particularly responsive to dietary caffeine.

Acute Effects of Caffeine on Blood Pressure

It has been shown conclusively that caffeine increases blood pressure acutely, with reports generally indicating increases in the range of 5–15 mg Hg systolic and 5–10 mg Hg diastolic. This acute pressor effect occurs across a wide age range, with effects lasting for up to several hours in healthy men and women (see James [95, 99] for a discussion). In addition, the pressor effect of caffeine is additive to that of cigarette smoking [54, 104, 210], is at least additive (e.g., France and Ditto [52], Greenberg and Shapiro [64], James [88, 90], and Lane and Williams [129]) and may be synergistic (e.g., [3, 128]) to the pressor effect of psychosocial stress, and is evident in persons with hypertension as well as normotensives (e.g., [71, 200]). Furthermore, studies show that caffeine produces acute increases in aortic stiffness and enhances wave reflection, both of which contribute to increased blood pressure as well as being independent risk factors for cardiovascular disease [116, 149, 239, 240]. Again, these effects have been observed in persons who are normotensive as well as those being treated for hypertension, and appear to be synergistic to similar effects of smoking [241].

Epidemiology of Caffeine and Cardiovascular Disease

More than 100 large epidemiologic studies in more than a dozen countries have reported data on the relationship between dietary caffeine and cardiovascular function, morbidity and/or mortality. Taken as a whole, epidemiologic findings suggest that dietary caffeine is detrimental to cardiovascular health [99]. However, one feature of this large literature is the many inconsistencies in the reported findings. The response of some commentators and reviewers to this inconsistency has been to dismiss concerns about caffeine, a response that is neither logical nor consistent with overall findings. Dismissing concerns is particularly unjustified in light of the large and consistent body of evidence from experimental studies. By their nature, experimental studies afford a greater level of control than epidemiologic approaches. Indeed, by integrating experimental and epidemiologic findings, the former help to clarify inconsistencies in the latter.

Considered comprehensively, the experimental and epidemiologic findings raise concerns over the implications of dietary caffeine for population cardiovascular health. The assumption that clear consistency should have emerged in the epidemiologic findings, if caffeine were having substantive effects on blood pressure and other indices of cardiovascular function, fails to take account of the many methodological shortcomings in the epidemiologic studies published to date. In particular, there are grounds for concluding that the many “null” reports (i.e., non-significant associations) in the epidemiologic literature on caffeine and cardiovascular disease reflect a high rate of Type II error (i.e., failure to observe a real effect when one exists).

Misclassification

A major shortcoming of many studies is poor measurement of the key “exposure” variable, namely, caffeine consumption. Although this shortcoming has long been the subject of criticism (e.g., [61, 89, 198]), relatively little improvement or innovation has been undertaken by epidemiologists over the past three decades to try to overcome the problem. Although at least half of the relevant epidemiologic studies conducted to date collected blood samples (mostly for the purpose of measuring serum lipid levels), none took the obvious next step of measuring systemic levels of caffeine or its metabolites [99]. As such, use has not been made of the fact that good estimates of dietary caffeine levels can be obtained by analyzing plasma and saliva caffeine (or paraxanthine, the major metabolite in humans) using high-performance liquid chromatography (e.g., [1]) or enzymeimmunoassay techniques [100].

Furthermore, although dietary caffeine levels can be measured reliably using detailed self-report inventories [107], many studies have employed poor self-report protocols and have shown little regard for the reliability of the measurements employed. Since the inadequate methods frequently used are likely to have produced largely undifferentiated (i.e., random) measurement error, the effect in many epidemiologic studies will have been to underestimate the true association between caffeine and cardiovascular disease or to report “no association”. Thus, while overall epidemiologic findings suggest that dietary caffeine has a modest detrimental effect on cardiovascular health, actual effects may be larger considering the often imprecise methods that have been employed [95, 96, 99].

Confounding in Epidemiologic Research

A frequent erroneous observation about the epidemiology of caffeine and cardiovascular health is that much of the research has ignored the influence of confounders. This “confounder myth” [95] asserts that reports of significant positive correlations between caffeine consumption and cardiovascular disease are the result of failure to control confounders, especially cigarette smoking. As well as being a cardiovascular risk factor, smoking has been found to be positively correlated with caffeine use (e.g., [119, 167, 226]). The myth, however, arises from the fact that, for the past three decades, epidemiologic studies of caffeine have routinely controlled for cigarette smoking. Excepting one or two early studies, virtually all of the literature reporting a positive correlation between caffeine consumption and cardiovascular disease controlled for the influence of cigarette smoking.

In the context of population studies there is always a risk of unanticipated influence of an as yet unidentified confounder. The level of such risk is probably lower in epidemiologic studies of dietary caffeine and health than in many other areas, because the list of potential confounders controlled for in caffeine studies (including those that reported positive findings) is very long, including: age, gender, cigarette smoking, alcohol consumption, body mass index, dietary factors, serum cholesterol, blood pressure, medical history, use of oral contraceptives, family history of heart disease, physical activity, personality, region of residence, education level, and religion [89]. Indeed, rather than being inadequately controlled for confounder effects, there has probably been a tendency toward overadjustment for confounders in epidemiologic studies of caffeine (e.g., [120, 190]). In particular, findings have frequently been adjusted for blood pressure and cholesterol, which may be caffeine-related and coffee-related causal pathways in their own right. Thus, as with measurement error, the likely effect of overadjustment for confounder effects would be to increase the risk of Type II error; that is, to underestimate the actual strength of the association between caffeine consumption and cardiovascular disease.

Threshold Effects

It is common in epidemiologic studies of caffeine to stratify according to level of reported caffeine use. A proportion of studies adopting that approach have reported the existence of a “threshold”, whereby a positive association is observed in consumers reporting higher levels of intake (e.g., “6 or more cups of coffee” per day) but not in consumers reporting lower levels of intake. Although such reports may be reassuring for “average” consumers, the notion of an actual threshold in this context is not persuasive. Experimental studies of caffeine have consistently found the acute hemodynamic effects of caffeine to be proportional to systemic caffeine level (e.g., [211, 212]). In the absence of other intervening variables, this dose-response effect would be expected to result in a relatively continuous relationship between caffeine and cardiovascular health outcomes rather than one marked by a threshold. Unreliability in the data, especially due to imprecise measurement of dietary exposure (as outlined above), is a more likely explanation of the threshold effects sometimes reported in epidemiologic studies of caffeine and population cardiovascular disease.

Epidemiology of Caffeine and Blood Pressure

As part of the much larger body of epidemiologic research on caffeine and cardiovascular disease, James [99] identified 18 population studies that were specifically concerned with caffeine and blood pressure. Of these, 5 reported no association between dietary caffeine and blood pressure, 6 reported a significant positive association for systolic and/or diastolic pressure, and 7 reported an inverse association for either systolic or diastolic pressure. The diverse findings are not explained by differences in the study populations, as these were similar in demographics and socioeconomics. Indeed, the level of inconsistency highlights the extent of the shortcomings that exist in the epidemiologic findings, which contrast the largely consistent pattern of pressor effects reported in experimental studies (discussed below). Of particular concern is the fact that epidemiologic studies have generally ignored issues related to the plasma caffeine concentration time course and associated pressor effects. The general pattern is shown in Fig. 2, which is a schematic representation of the estimated 24-h plasma caffeine concentration time course, assuming an elimination half-life of 5 h and ingestion of the approximate equivalent of 1 cup of coffee in the morning, mid-morning, and mid-afternoon.

Fig. 2
figure 2

Schematic representation of estimated 24-h plasma dietary caffeine concentration time course and associated change in blood pressure. Estimated plasma caffeine concentration assumes an elimination half-life of 5 h and ingestion of 1 cup of coffee after awakening and at two further time points (arrows) during the earlier part of the day (with evening and overnight abstinence). Associated blood pressure changes are relative to caffeine-free levels

Full size image

Figure 2 helps to show that the strength, and even the sign, of the correlation between dietary caffeine and blood pressure level depends on the timing of blood pressure measurement relative to when caffeine was last ingested [99]. Using 24-h ambulatory monitoring, James [91] found that overnight abstinence produced transient modest decreases in blood pressure. Thus, taking a cross-section of the population, recent caffeine consumption is likely to have a pressor effect (positive association), whereas brief caffeine abstinence (10–12 h) may have no effect, and longer periods of abstinence (12–24 h) may decrease blood pressure modestly (inverse association due to withdrawal). In view of this analysis, a noteworthy feature of several of the studies in which dietary caffeine was said to have been protective (i.e., inverse association between intake and blood pressure) is that participants were asked to fast before being examined [99]. Specifically, participants in 5 of the 7 relevant studies were reported to have fasted, while one reported non-fasting and one omitted to report whether participants fasted or not. Thus, in the majority of the studies involved, caffeine consumers’ blood pressure readings were likely to have been transiently lower (due to withdrawal) than “normal” for themselves and potentially lower also than their non-consuming counterparts.

Although interpretation of the findings of epidemiologic studies of caffeine and blood pressure depends crucially on knowing when blood pressure was measured relative to when participants ingested caffeine, with one exception [203], none of the relevant studies provides that level of detail. In the one exception, an overall analysis revealed no association between caffeine consumption and blood pressure level after adjustment for age, body mass, cigarette smoking, alcohol consumption, serum cholesterol, and family history of hypertension [203]. On closer examination, however, the authors reported that participants who had consumed caffeine during the 3 h prior to measurement had significantly elevated blood pressure compared with participants consuming no caffeine for the same period. Importantly, because the increases in blood pressure associated with recent ingestion of caffeine were independent of average daily intake (a measure of habitual use), the results also confirm experimental findings that habitual caffeine consumption does not lead to complete tolerance to the pressor action of the drug.

Chronic Effects of Dietary Caffeine on Blood Pressure

Before the last decade, there had been little direct (experimental) examination of the chronic hemodynamic effects of dietary caffeine. Among the first studies to undertake such an examination, modest sustained decreases in blood pressure were reported when caffeine beverages were either removed from the diet [10] or replaced by decaffeinated alternatives [235]. Similar results were reported in a number of subsequent studies in which ambulatory monitoring was used to measure blood pressure level for extended time periods [63, 91, 109, 176, 221].

Moreover, it is known that blood pressure responses of similar magnitude may be accompanied by different patterns of change in cardiac output and total peripheral resistance, and these differences in hemodynamic profile may be implicated in cardiovascular pathology [65]. Speculation has existed as to whether caffeine-induced pressor effects are due to cardiac stimulation of contractility leading to increased cardiac output, or vasoconstriction leading to increased total peripheral resistance. Findings generally suggest that the blood pressure-elevating effect of caffeine is due primarily to increased vascular resistance [30, 56, 70, 100, 209]. Because greater risk has been attached to hemodynamic reactivity in which vascular, rather than myocardial, responses predominate [112], findings of caffeine-induced vascular resistance add to concerns regarding the possible implications of dietary caffeine for cardiovascular health.

Dietary Caffeine and Population Blood Pressure Levels

If, as this review indicates, dietary caffeine contributes to statistically significant elevations in blood pressure, it should be noted that such increases are modest in absolute terms, amounting to possibly 2–4 mm Hg for most waking hours of the day. The question, therefore, that needs to be considered is whether such increases are likely to have an appreciable effect on population cardiovascular mortality and morbidity. It is sometimes presumed that increases of such magnitude are not meaningful, on the grounds that blood pressure level is inherently variable. However, it should be remembered that the effects of caffeine are at least additive, and possibly synergistic, to blood pressure increases due to a variety of other factors (e.g., smoking, hypertension, stress). In this sense, caffeine represents a preventable additional burden on the cardiovascular system.

The clearest insight into the contribution of blood pressure increases to cardiovascular disease is provided by population statistics describing the relationship between blood pressure level and cardiovascular mortality and morbidity. Since the association between the population distribution of blood pressure and cardiovascular disease is primarily linear, any contribution by caffeine to population blood pressure level may be expected to contribute to the overall incidence of cardiovascular mortality and morbidity [147, 148, 175, 183, 233]. It is important to remember that exposure to caffeine is generally long (essentially lifelong for most consumers), the prevalence of exposure is high (more than 80% in most countries), and the incidence of cardiovascular disease is high throughout the world. While reduced blood pressure associated with reductions in dietary caffeine may be expected to be modest in absolute terms, even modest absolute changes in population levels of blood pressure translate to significant changes in the population burden of cardiovascular death and disease.

For example, it has been estimated that a downward shift of 2–3 mm Hg in the population distribution of blood pressure would produce life-saving benefits equal to the cumulative benefits achieved by antihypertensive treatment [183, 188]. It has also been estimated that population-wide reductions of 2 mm Hg could avert 5% of deaths from coronary heart disease and 15% of stroke deaths [182, 183]. More specifically, James [87] estimated that if caffeine consumption had the effect of elevating average population blood pressure by 2–4 mm Hg (a reasonable inference considering the relevant experimental data (e.g., [96, 99, 109, 221]), extrapolation based on epidemiologic blood pressure data [147, 148] suggests that population-wide cessation of caffeine use could lead to a reduction of 9–14% of premature deaths from coronary heart disease and 17–24% of premature deaths from stroke. If caffeine were removed from the diet in populations where coffee specifically is widely consumed, additional benefits would be achieved due to the adverse impact of that beverage on serum cholesterol and homocysteine [99].

Dietary Caffeine and Physical Health: Non-Cardiovascular Disease

Cancer

Although numerous studies of cultured cells in vitro have demonstrated the mutagenic potential of caffeine, in vivo studies of intact nonhuman animals have suggested variously that caffeine: is not a carcinogen, is carcinogenic under some conditions, and is antitumoric under other conditions [95]. Moreover, the relevance of the in vitro and in vivo findings to lifelong dietary use of caffeine in humans remains unclear. Overall, there is a strong consensus that the experimental evidence as a whole suggests that the drug is not a significant carcinogen in humans. In addition, there has been extensive epidemiologic study of caffeine beverage consumption and cancer. Most of this research has been primarily concerned with coffee consumption, although over the past decade tea has also been a focus of attention. Because comparatively few studies have examined caffeine specifically, it is necessary to treat the findings for coffee and tea consumption as being only indirectly suggestive of the carcinogenic potential of caffeine.

All Cancers

Cancer is not a single disease, and therefore it is not surprising that most studies have been concerned with cancers located at one or a small number of specific sites rather than overall cancer rates. However, regarding overall rates, studies have tended to suggest no adverse impact of caffeine on cancer mortality (e.g., [137, 152]). On the other hand, studies of specific sites indicate more complex associations than that suggested by examination of the relationship between caffeine consumption and overall cancer incidence.

Lower Urinary Tract

Following an early report by Cole [29] of a significant association between coffee consumption and cancer of the lower urinary tract (renal pelvis, bladder, and urethra), there has been considerable epidemiologic interest in coffee as a possible cause of bladder cancer. The substantial body of literature that has accumulated tends to suggest a positive but weak association [84, 89]. However, although the association has been reported intermittently in different populations during the past two decades [26, 33, 158], widespread doubt exists as to whether the association is causal.

Pancreas

Early studies reported a relationship between caffeine beverages and pancreatic cancer [146, 219], one of the most rapidly fatal of human malignancies. Subsequent epidemiologic studies, however, yielded mixed results. In a review of relevant research conducted prior to 1990, the International Agency for Research on Cancer [84] concluded that the evidence was suggestive of a weak relationship between high levels of coffee consumption and the occurrence of pancreatic cancer, but cautioned that even this association could be due to bias or confounding. More recent studies have tended not to support the existence of even a modest positive correlation [55, 114, 202, 253, 255], although in one meta-analysis it was concluded that small amounts of coffee may be protective while high intake increases disease risk [163].

Breast

The epidemiology of caffeine and breast disease is somewhat mixed, especially among older studies, with some reporting a modest increased risk associated with caffeine consumption [131, 134, 150, 187] and others reporting no association [144, 189, 196]. More recent studies, however, have tended increasingly to report no association [47, 155, 208] and, more recently still, reports have appeared of an inverse association (i.e., “protective” effect) between caffeine consumption and breast cancer. Unfortunately, however, the pattern of findings has been inconsistent, with one study reporting an inverse association in premenopausal women and no association post-menopause [11], and another study reporting a weak inverse association in postmenopausal women but no association for the cohort overall [155].

Colon

Results of studies of caffeine consumption and cancer of the colon and/or rectum have also been highly varied. Some reported no association between coffee consumption and increased risk of disease [136, 164, 166], whereas others reported an increased risk [205, 213]. Still others, however, have reported a reduced risk [1, 12, 23, 86, 132, 133, 206, 223]. The frequency of reports of reduced risk (i.e., potential protective effect) has led to speculation about possible mechanisms of action, including rates of bile acid secretion and colonic motility [86, 223].

Other Sites

Results for other sites also tend to be mixed, with a pattern seeming to emerge of more recent studies reporting no association, or even a protective effect in some instances, thereby negating earlier findings of adverse effects. For example, Armstrong and Doll [6] reported a positive correlation between coffee consumption and cancer of the kidney, whereas later studies, with the exception of Asal et al. [8], have mostly failed to observe any relationship between coffee and/or tea consumption and kidney cancer. Similarly, whereas several earlier studies reported significantly increased risk of ovarian cancer in coffee consumers [130, 225, 249], more recent studies have tended to report no association [76, 215, 222].

Maternal Use of Caffeine

As mentioned above, caffeine readily crosses the placenta during pregnancy. Thus, throughout pregnancy, the developing fetus is exposed to concentrations of the drug equal to systemic levels in the mother. Naturally, questions arise regarding the implications of this exposure, especially considering the known pharmacological actions of caffeine. In 1980, responding to reasonable suspicions and early empirical findings, the United States Food and Drug Administration issued a warning advising pregnant women to restrict, or eliminate, coffee consumption. The focus of this warning was in relation to gross morphological (i.e., physical) abnormalities that had been observed in animal studies. However, animal studies usually involved dosing levels higher than those typical of human dietary use, and the consensus today is that dietary levels are unlikely to result in morphological abnormalities [95].

Notwithstanding reassurance regarding gross defects, the question arises as to what represents an appropriate margin of safety for intrauterine exposure to caffeine in the human fetus. The usual safety standard employed by the Food and Drug Administration in relation to the human consumption of food additives is one-hundredth the maximum safe level of exposure in animals [227]. By that standard, virtually any pattern of regular caffeine consumption by a woman who is pregnant would put her unborn child at risk. That is, applying the Food and Drug Administration’s usual standards, pregnant women should abstain from caffeine completely. Moreover, teratology (the scientific study of conditions caused by the interruption or alteration of normal development) includes not only the study of physical defects, but also the study of more subtle behavioral and emotional anomalies. Although a wide range of caffeine-induced developmental effects on behavior and neurochemistry have been demonstrated in animals, there have been very few reported studies in humans. The results that have been reported point to the need for further studies to examine caffeine as a potential behavioral teratogen [95].

Pregnancy Outcome

In addition to concerns about possible teratogenicity, there are concerns that maternal caffeine use could have adverse effects on pregnancy outcomes. Several studies have reported a positive association between maternal caffeine use and spontaneous abortion [28, 35, 60, 178, 247], whereas some others have found no association [43, 157]. It has been suggested that positive findings could be due to confounding from pregnancy-induced nausea, which is less frequent in pregnancies that miscarry than those that go to term. It is plausible that women who experience nausea might respond by reducing their caffeine intake. Consequently, it could be this “loss of taste” for caffeine rather than reduced caffeine per se that might be the basis for the observed positive correlation between higher caffeine use and spontaneous abortion. However, the nausea hypothesis has not been supported by studies that took account of nausea experienced during pregnancy [42, 60].

The mixed, sometimes contradictory nature of the findings for fetal loss is also characteristic of the findings for other major pregnancy outcomes. In particular, several studies have reported an inverse association between maternal caffeine use and fetal growth [159, 217, 237], while others have found no association [27, 68, 204]. Although the findings are far from consistent, it appears that the current weight of evidence is suggestive of an adverse effect of caffeine in that at least two meta-analyses have concluded that caffeine consumption is associated with a significant decrease in birth weight [45, 195]. Even then, it remains a possibility that the seemingly adverse effects of caffeine on particular pregnancy outcomes could have been due to the influence of confounders (e.g., recall bias). Overall, however, the available evidence points to maternal caffeine use being associated with increased risk of adverse pregnancy outcome, especially increased spontaneous abortion and lower birth weight.

Notwithstanding evidence of an association between caffeine consumption and adverse pregnancy outcomes, advice as to the need for caution regarding caffeine intake during pregnancy has tended to be heavily qualified. This appears to be partly due to the fact that several of the relevant studies have observed significant associations only for higher levels of intake, which has contributed to the belief that any causal involvement of caffeine is subject to a threshold. For example, in its most recent position statement on “nutrition and lifestyle for a healthy pregnancy outcome”, the American Dietetic Association has specified a threshold of 300 mg/day, advising that pregnant women should avoid only higher levels of intake [113]. However, although there appears to be good agreement that caffeine can be harmful at higher levels of intake, there is no clear evidence-based reason that explains why immunity from harm is conferred at lower dietary levels. Moreover, as discussed above in relation to cardiovascular disease, unreliability in the data due to imprecise measurement of dietary exposure to caffeine would appear to be a more likely explanation of any threshold of harm pertaining to maternal caffeine use. Indeed, with regard to the particular threshold advised in this context, 300 mg/day cannot reasonably be regarded as “high”, since that level of intake can be readily reached and exceed by consuming as little as 2 cups of brewed coffee. All things considered, abstinence, as frequently recommended in relation to tobacco and alcohol, would appear also to be the most appropriate recommendation regarding caffeine use during pregnancy.

Adverse Interactions Between Caffeine and Other Drugs

Considering the near-universal use of caffeine, it is inevitable that the taking of other drugs will often coincide with that of caffeine. Regarding recreational drugs, it is commonplace to see smokers light up when drinking a caffeine beverage, and indeed cigarette smokers consume more caffeine on average than non-smokers. Similarly, alcohol is sometimes consumed in conjunction with caffeine, either as separate beverages or, as appears to be increasingly popular among younger-age groups, in a single beverage containing both alcohol and caffeine. Caffeine is also sometimes used to “cut” illicit drugs such as heroin, cocaine, and amphetamine, with the users of those drugs sometimes consuming substantial amounts of caffeine even when not intending to do so. Particular concerns, however, arise in relation to pharmaceuticals with which caffeine may interact adversely or whose therapeutic efficacy may be undermined by caffeine (e.g., benzodiazepines and some antibiotics) [89, 95].

Is Caffeine Addictive, and Is There a Safe Level of Consumption?

The evidence reviewed above indicates that dietary caffeine is a probable risk to cardiovascular health, poses a threat to fetal growth, interacts adversely with common therapeutic drugs, and produces dysphoric effects after brief abstinence. Therefore, taking account of its widespread and persistent use, should caffeine be considered a drug of addiction? Physical dependence is a common feature of drugs widely regarded as addictive. On this point, the evidence is conclusive; the occurrence of a characteristic syndrome of abstinence effects shows that repeated caffeine use leads to the development of physical dependence. Accordingly, it may reasonably be said that caffeine is a drug of addiction. On the other hand, the term “addiction” has wide currency, and carries a variety of emotive connotations (e.g., illegal importation, criminal syndicates, and violent crime) that have little relevance to dietary caffeine. Accordingly, it might be prudent not to be too strident in labeling caffeine an “addictive” substance. This stance, however, should not distract us from the evidence that dietary caffeine is harmful.

Considering the evidence of harm, it is appropriate to ask: Is there a safe level of consumption? As previously stated by this author, a balanced (if unpopular) answer to this question is that there is no daily level of intake that can be regarded safe [95]. The equivalent of as little as 1 cup of coffee produces modest increases in blood pressure lasting 2–3 h, which over the course of a lifetime is likely to contribute to increased cardiovascular disease; any exposure to caffeine during pregnancy exposes the fetus to a dose equivalent to that received by the mother; caffeine interacts negatively with therapeutic medications, and dietary use produces physical dependence.

Reducing and Quitting Caffeine Consumption

Despite strengthening evidence that dietary caffeine is harmful, few reports exist of systematic efforts for assisting habitual consumers to reduce or cease their use of the drug. Indeed, following a brief rise in interest about 2 decades ago, reports of systematic attempts to manage caffeine intake appear to have all but disappeared from the literature. One early commentary on the subject more than a century ago advised that negative withdrawal effects could be avoided by a gradual reduction of caffeine [18]. That advice has stood well the test of time. Using a single-subject experimental design, Foxx and Rubinoff [51] reported favorable results for three participants who received a program of behavioral intervention based on nicotine and cigarette “fading” methods that the same research group had developed for smokers (e.g., [49, 50]. Treatment consisted of a combination of self-monitoring and a series of predetermined step-wise reductions in daily caffeine consumption in the direction of a specified terminal goal of reduced daily intake. Subsequently, Foxx [48] obtained follow-up data from the three original participants, reporting that the reduced intake of all three was substantially maintained 40 months following the termination of treatment. Bernard et al. [13] employed similar procedures with a single subject, and again reported favorable results.

These generally promising initial findings were confirmed in a larger study by James et al. [107] in which 27 chronic heavy caffeine consumers were monitored before and during a 4-week treatment program and at 6- and 18-week follow-up. However, because the results of this and previous caffeine-reduction studies were expressed solely in terms of participant self-reports, the reliability of the findings could be open to question. Accordingly, James et al. [107] reported plasma concentrations of caffeine and its primary demethylated metabolites (paraxanthine, theophylline, and theobromine) as well as self-reported caffeine intake during the course of a caffeine-fading regimen similar to that employed in the previous study by the same authors [106]. Overall, the 12 subjects, each with a history of heavy caffeine use, provided highly reliable self-reports of caffeine intake during the course of the 18-week program. However, unlike the earlier studies in which follow-up data had been obtained, participants in the James et al. [107] study showed signs of relapse at 12 weeks follow-up.

It has long been known that the accuracy of self-reports is enhanced when subjects are aware that their behavior may be independently checked (e.g., [141, 161]. Hence, the independent measurement of plasma caffeine levels in the James et al. [107] study may have encouraged subjects to be more accurate than participants in previous studies in reporting follow-up caffeine intake. If accurate and generalizable, the relapse reported by James et al. [107] is broadly consistent with reports of treatment outcomes for other dependence-producing substances. Although the reasons for the relapse observed by James et al. [107] remain unclear, relapse would not appear to have been due to the direct influence of withdrawal effects, since the resumption of higher levels of consumption did not occur until many weeks after the original treatment goal had been achieved. As such, firm statements cannot be made at this time regarding the long-term level of success of attempts to reduce caffeine intake. Nevertheless, it is clear from the available evidence that motivated individuals wishing to reduce or quit their use of caffeine can do so without experiencing pronounced (if any) negative withdrawal effects, provided that intake is reduced in a graduated (step-wise) fashion rather than abruptly (as when “going cold turkey”).

Does Caffeine Have Health Benefits?

There has long been interest in caffeine beverages as possible sources of benefit, and much of that interest has centered on the putative benefits of caffeine for psychomotor performance and mood. However, as discussed above, there is now a firm body of evidence showing that caffeine has little or no net benefits for performance or mood. At the same time, at least partly fostered by industry-sponsored research, there has been substantial growth in interest in caffeine beverages as possible sources of benefit for physical health, especially in relation to diabetes and Parkinson’s disease.

Type 2 Diabetes Mellitus

Several epidemiologic studies in the United States and Europe have reported significant dose-dependent reductions in the risk of developing Type 2 diabetes mellitus in association with caffeine and coffee consumption (e.g., [2, 191, 193, 231, 232]). While findings have prompted some authors to claim that caffeine and coffee protect against Type 2 diabetes, it is important to note that the studies in question were non-experimental and shared many of the same potential confounder effects that have generally undermined interpretation of epidemiologic studies of caffeine and health. More importantly, experimental studies have found the opposite pattern of results than would be expected from the population studies.

Double-blind placebo-controlled trials have consistently found that caffeine impairs glucose tolerance and decreases insulin sensitivity, and the findings have been reported for a wide range of participant groups including persons with diabetes and those without (e.g., [62, 118, 124–126, 140, 170, 232]). As such, it is difficult to reconcile how caffeine could offer protection against Type 2 diabetes when experimental studies have shown that it compromises glucose metabolism both before and after development of the disease. Thus, although caffeine appears distinctly unlikely to confer any protection against the development of diabetes, one issue is whether there may be a compound other than caffeine in coffee that offers such benefit. If such a compound exists, to be of benefit, it would need to be sufficiently potent not only to negate, but to exceed, the negative effects of caffeine.

Parkinson’s Disease

There is a substantial body of recent epidemiologic evidence of an inverse association between caffeine consumption and the development of Parkinson’s disease (e.g., [9, 74, 192]). This finding has been widely assumed to be causal, and has contributed to speculation about the “neuroprotective” action of caffeine. In particular, attention has focused on interactions between the dopaminergic and adenosinergic systems and caffeine’s putative ability to forestall dopaminergic neuron degeneration through its action on the A2A adenosine receptor (e.g., [21, 24, 115, 154, 199, 224]). An earlier population study by Jarvis [108] is sometimes cited as supportive of the idea that caffeine has neuroprotective properties. In a cross-section of the population, Jarvis reported that higher caffeine intake was positively related to better performance on certain psychomotor and cognitive tasks, and the effect was reported to have been larger in older participants. However, a more recent prospective study involving a larger population sample found little evidence of improved performance associated with caffeine consumption or of reduced age-related cognitive decline [230].

Moreover, Evans et al. [39] recently suggested that the inverse association between caffeine consumption and Parkinson’s disease, as well as the similar relationship that exists with cigarette smoking (which has fostered the belief that nicotine is neuroprotective), may be “epiphenomena” rather than causal. Broadly, Evans et al. [39] argued that confounding due to individual differences in the personality disposition of impulsive sensation seeking may have led to misunderstanding of the findings. The authors cited evidence that sensation seeking is inversely associated with Parkinson’s disease, with higher sensation seeking also being associated with higher caffeine consumption and smoking. Evans et al. [39] hypothesized that there are biological features characteristic of low-sensation-seeking individuals that also predispose to Parkinson’s disease. Thus, rather than indicating any neuroprotective capability, higher caffeine and nicotine intake may simply be two behavioral manifestations of a generalized personality disposition, namely, impulsive sensation seeking, which itself is the expression of a biological substrate that confers a level of protection against the development of Parkinson’s disease.

“Other” Active Compounds in Caffeine Beverages

Notwithstanding the strength of the evidence that dietary caffeine poses a number of significant risks to health, an important caveat arises when other compounds in caffeine beverages are considered. Whereas caffeine is generally accepted as being the main biologically active ingredient of those beverages, the presence of other compounds also having biological effects has become a focus of interest. Of course, the “other” active compounds could have either positive or negative implications for health. An example of the latter is the presence of a cholesterol-raising factor in unfiltered brewed coffee [194, 228, 243, 248]. However, influenced by industry-sponsored research over the past decade, interest has been strengthening in the search for beneficial effects from non-caffeine active compounds in coffee and tea (e.g., the relation between coffee and Type 2 diabetes mentioned above).

Accordingly, any assessment of the overall health implications of caffeine beverages must take account of the benefits, if any, of these other compounds. For example, it is claimed that polyphenols, especially chlorogenic acid, in coffee have potential cardiovascular benefits due to antioxidant properties (e.g., Bonita et al. [16]). Similarly, theanine, a non-proteinic amino acid, has been posited as having a blood pressure-lowering effect (e.g., [186]). At the same time, it must be emphasized that a notable feature of research into the benefits of caffeine beverages is the involvement of the caffeine industry at all levels of research, including basic and applied animal and human studies, and the production of published scientific articles including empirical studies and literature reviews. By any measure, industry involvement is extensive, even pervasive, and the conflicts of interest inherent in industry-sponsored research (including the dissemination of research findings) raise serious questions regarding the increasing frequency of “scientific” claims for the benefits of caffeine-containing beverages. In short, for some time, the integrity of caffeine science has been under threat.

Threats to the Integrity of Caffeine Science

Industry Influences on Research

The available experimental evidence, and to a lesser extent the epidemiologic evidence, supports the conclusion that caffeine use is a likely risk factor for health. Notwithstanding the importance of the implications of this conclusion for population health, there is little organized effort to inform and to advise the public on ways consistent with the magnitude of that threat. Indeed, it is evident from a close examination of the caffeine literature that this is a field of enquiry marred by a considerable amount of misinformation and misrepresentation. In particular, it is necessary to confront the reality that the academic pursuit of research on caffeine is extensively linked to the trade in caffeine products. Each of the main sources of caffeine, namely, coffee, tea, soft drinks, and energy drinks, is a multinational, multibillion dollar enterprise. By their own account, these industries have sought to lessen the impact of scientific findings that could threaten their commercial interests (see James [93, 98]).

Over the past quarter-century, various methods have been employed by industry to influence public opinion about caffeine and caffeine products. Such attempts include dissemination of selective information and funding for selected caffeine research [93]. During the 20 years from 1962 to 1982, the average number of cups of coffee consumed per day in the United States declined 39% [153], and it is evident from caffeine-industry publications that manufacturers attributed much of that decline to increased public awareness of scientific concern about possible caffeine-induced harmful effects [93]. Around 1990, there was an arrest in the downward trend, and thereafter a reversal evidenced by substantially increased sales of all categories of caffeine beverages. Manufacturers of caffeine products appear to have been in no doubt about the reason for the improved commercial outlook for caffeine products. Industry representatives congratulated themselves on the success of their campaign to counter scientific findings that threatened their interests [75, 179]. In this regard, there appear to be parallels between actions by the caffeine industry to protect its commercial interests and similar activities by the tobacco and alcohol industries.

One influential industry body is the International Life Sciences Institute, which lists its Committee Members as including Coca-Cola, Kraft Foods, Mars, Nestlé, Procter & Gamble, Unilever, and others having commercial interests in caffeine products [85]. The International Life Sciences Institute actively pursues affiliations with the United Nations Food and Agriculture Organization and the World Health Organization, and is directly and extensively involved in publicly funded European Union research in areas of interest to its members [98]. The International Life Sciences Institute and the companies it represents commission scientific research into caffeine, and take an active role in sponsoring the production of scientific literature on caffeine [85]. Although affiliation with industry and material assistance from industry do not themselves constitute evidence of wrongdoing, such collaboration is worryingly commonplace in caffeine research (e.g., [15, 22, 85, 186]).

Despite being reported to have been extensively involved in assisting the tobacco industry to counter the World Health Organization’s efforts to promote tobacco controls, especially in developing countries, the International Life Sciences Institute describes itself as having ongoing close involvement with the World Health Organization’s activities [76]. A World Health Organization Committee of Experts on Tobacco Industry Documents reported that for many years tobacco companies operated with the “purpose of subverting the efforts of the World Health Organization to address tobacco issues [and that the] attempted subversion has been elaborate, well financed, sophisticated and usually invisible” [254] (p. 18). Subsequently, the Tobacco Free Initiative, a World Health Organization project, identified the International Life Sciences Institute as one such group [145, 160]. Moreover, the International Life Sciences Institute has been the subject of editorial criticism for its reticence in declaring a possible conflict of interest regarding its involvement in a publication concerned with health issues related to alcohol consumption [38]. The picture that has emerged is of an “institute” presenting itself as dispassionate and independent, while actually serving as a “third party” representative of commercial interests [98].

The research community needs to heed the dangers of industry influence on research. It is important that ways are found for ensuring exposure of possible conflicts of interest where they are not freely declared. Where possible conflicts do exist, ways must be found to safeguard against resulting threats to scientific integrity [98]. The importance and urgency of steps by the scientific community to counter such threats is highlighted by empirical evidence of bias attributable to pharmaceutical industry involvement in biomedical research. For example, in a study of the association between funding source and conclusions in randomized medication trials, Als-Nielsen et al. [4] found that after adjustment for study characteristics, industry-sponsored trials were 5 times more likely to yield conclusions favorable to industry’s commercial interests than trials funded by nonprofit organizations.

Conflict of Interest and the Self-Serving Bias

A conflict of interest exists when an ethical or professional interest clashes with a pecuniary self-interest. Although a necessary prerequisite for openness, the mere declaration of a conflict of interest is unlikely to foil outcome bias in industry-sponsored research. For one thing, a simple declaration provides no basis for consumers of scientific research, including scientists, policy makers, and the public, to judge the nature and extent of any consequential bias. Indeed, drawing on relevant experimental findings from social psychology, Dana and Loewenstein [32] have argued that declaring a conflict of interest can actually be counterproductive by exacerbating the declarer’s bias. Dana and Loewenstein [32] explained that part of the difficulty in dealing with the problem is that it is usually assumed that bias founded on a conflict of interest is a matter of deliberate choice. This perspective contributes to the indignation that is sometimes expressed when the subject is raised. Unfortunately, however, the “deliberate choice” view of bias arising from conflicts of interest is inconsistent with empirical findings, which show that even when individuals try to be objective their judgments are subject to an unintentional self-serving bias [32]. In other words, self-serving bias is part of human nature. It is the role of the scientist to safeguard the integrity of research in the face of human limitations.

Indeed, unintentional self-serving bias might help to explain some apparent contradictions alluded to above. Weinstein [246] has shown that “behavioral performance tends to produce perceptions supportive of the behavior” (p. 2). If so, it is likely that caffeine consumers will be more readily accepting of conclusions consistent with their own extant caffeine-consuming behavior than findings that conflict with such behavior. Thus, it is possible that a subtle inherent self-serving bias inclines consumers of caffeine products to be more influenced by neutral or positive findings concerning caffeine than is engendered by more objective assessments. Furthermore, since most people consume caffeine daily, it is likely that the large majority of researchers, reviewers, and editors of scientific literature are caffeine consumers. As such, the resulting impact of an inherent self-serving bias on the way scientific findings are promulgated could be pervasive. For example, experimental findings are ordinarily accepted as providing stronger evidence of causal relationships than epidemiologic findings. Yet, the opposite view could be said to have been in operation in a number of important areas of caffeine research. In relation to cardiovascular disease and Type 2 diabetes mellitus, in particular, there appears to have been a tendency to ignore experimental findings of likely harm in favor of accepting epidemiologic findings of no harm. Unfortunately, however, there is relatively little published literature addressing the topic of self-serving bias in science, and as such little systematic knowledge has accumulated as to the extent of industry-based threats to scientific integrity.

Conclusions

Claims that dietary caffeine is of little importance to health are ill founded. Short-term withdrawal of caffeine has negative effects on psychomotor performance and mood, and these effects may reoccur chronically in habitual consumers. Caffeine produces modest increases in blood pressure that have long-term implications for cardiovascular health, caffeine interacts adversely with some medicines, and there is suggestive evidence of increased risk of spontaneous abortion and lower birth weight associated with caffeine use in pregnancy. Conversely, there is little or no satisfactory evidence of net benefits of dietary caffeine. Although further evidence is needed, it is unlikely that adverse effects are necessarily limited to groups characterized as “heavy” consumers. Notwithstanding the need for further research, the extensive involvement of industry bodies in that research effort raises questions concerning the integrity of caffeine science.