Additional related information may be found at:
Neuropsychopharmacology: The Fifth Generation of Progress

Back to Psychopharmacology - The Fourth Generation of Progress

Caffeine : A Drug of Abuse?

Roland R. Griffiths and Geoffrey K. Mumford


The widespread use of culturally sanctioned caffeine-containing foods presents an intriguing paradox. On one hand, it is the experience of most regular caffeine users that caffeine produces only rather subtle effects that are generally so well-woven into the fabric of daily experience that they are not clearly differentiated from the changes in mood and behavior associated with normal experience.

On the other hand, however, caffeine is arguably the most robust form of drug self-administration known to man. Consider the facts: (a) Historically, use of caffeine-containing foods has been long-term (possibly dating back 4700 years), with use of these foods spreading worldwide despite recurring efforts, motivated on moral, economic, medical, or political grounds, to restrict or eliminate their use (4, 45). (b) Currently, regular daily consumption of behaviorally active doses [e.g., >80% of adults in the United States (44, 69)] is widespread throughout the world (38); the extent of caffeine usage far exceeds that estimated for alcohol and nicotine, which rank second and third, respectively, as the most widely consumed psychotropic drugs worldwide (39). (c) Habitual consumption of behaviorally active doses of caffeine occurs in widely different vehicles (e.g., drinking of coffee, tea, soft drinks, matJ; chewing kola nuts) and under widely different social and cultural conditions; therefore, caffeine ingestion is a remarkably generalized form of drug self-administration occurring across a broad range of dietary and contextual conditions.

It is this paradox of seemingly subtle pharmacological effects juxtaposed against its robustness as a self-administered drug that makes caffeine among the most intriguing psychotropic substances for research into the behavioral-pharmacological mechanisms underlying the habitual use of drugs generally, and possibly drugs of abuse in particular. The purpose of this chapter is to provide an updated review of the rapidly emerging research literature concerning the behavioral pharmacology of caffeine as it relates to our understanding of self-administered and abused substances. Specifically, the chapter will review infrahuman and human research on the reinforcing effects of caffeine and human research on caffeine subjective effects, tolerance, and physical dependence. The review will conclude with a discussion about whether it is meaningful to consider caffeine to be a drug of abuse. Neurochemical mechanisms underlying some of caffeine's pharmacological effects are discussed in Purinergic Mechanisms in Nervous System Function and Disease States on purinoceptors in this volume.


Reinforcing efficacy of a drug refers to the relative effectiveness in establishing or maintaining behavior on which the delivery of the drug is dependent (52). Over the last 25 years, reliable experimental models of drug-taking behavior in laboratory animals have been developed that provide valid information about the relative reinforcing effects of psychoactive drugs (10, 48) (see Animal Models in Drug Addiction ).

Intravenous self-injection has a high degree of face validity and is often regarded by specialists as providing the most direct and unequivocal assessment of a drug's reinforcing effect (11). Intravenous self-injection involves implanting a venous catheter in animals and allowing them to self-administer a drug. A behavioral response (often a lever press) is followed by intravenous drug injection. The ability of the injection to reinforce behavior is assessed by examining the establishment or maintenance of responding, usually relative to vehicle control.

As shown in Table 1, intravenous drug self-injection studies indicate that caffeine can function as a reinforcer under some conditions. Of the nine self-injection studies (all of which were conducted in nonhuman primates and rats), all but two demonstrated caffeine self-injection in at least a portion of the animals studied. The two negative studies either tested only one dose level of caffeine (61) or tested caffeine self-injection for a relatively short period (1 or 3 days) after substitution for a psychomotor stimulant drug (SPA) that maintained high rates of self-injection (110).

Of the seven positive studies, four demonstrated self-injection in all animals (25, 28, 52; Griffiths, Sannerud, and Kaminski, unpublished data; Table 1), whereas three (3, 20, 96) showed that only a subset of animals (25-33%) self-injected caffeine. A characteristic sporadic pattern of caffeine self-injection has emerged in three studies with nonhuman primates that examined self-injection over an extended period of consecutive days (25, 52; Griffiths, Sannerud, and Kaminski, unpublished data; Table 1). This pattern is characterized by periods of relatively high rates of intake alternating irregularly with periods of low intake. Fig. 1 illustrates this pattern in three baboons.

These intravenous self-injection studies clearly show that caffeine can function as a reinforcer under some conditions. However, the inconsistent results across animals and studies contrast with the results reported with the classic abused stimulants (e.g., amphetamine and cocaine) that have more consistently been shown to maintain intravenous self-injection across a wide range of species and conditions (52). The variation in results with caffeine is analogous to that reported in self-injection studies with nicotine: Nicotine has not reliably maintained self-injection across animals and studies. It has been concluded that although nicotine can serve as a reinforcer (40, 59), it does so under a more limited range of conditions than do drugs such as cocaine (28, 40). A similar conclusion seems warranted for caffeine (28).


As in the animal drug self-administration laboratory, over the last 25 years, reliable procedures have been developed in the human laboratory for examining the reinforcing effects of drugs (7, 23, 48, 60). These procedures have permitted the demonstration of drug reinforcement and the examination of correlates and determinants of reinforcement.

Demonstration of Caffeine Reinforcement

Although the circumstantial evidence suggesting the reinforcing effects of caffeine is compelling (see introductory paragraphs at beginning of this chapter), clear experimental evidence has only recently been reported. Table 2 summarizes 10 studies that examined whether caffeine could function as a reinforcer in humans. These studies provide unequivocal evidence of the reinforcing effects of caffeine. These studies demonstrated caffeine reinforcement under double-blind conditions using various subject populations (moderate and heavy caffeine users with and without histories of alcohol or drug abuse), using a variety of different methodological approaches (variations on both choice and ad libitum self-administration procedures), when caffeine was available in coffee, soda, or capsules, when subjects did and did not have immediate past histories of chronic caffeine exposure, and in the context of different behavioral requirements after drug ingestion (vigilance versus relaxation activities). Fig. 2 illustrates caffeine reinforcement in three subjects who chose between capsules containing caffeine and placebo.

Incidence of Caffeine Reinforcement

Table 2 shows that caffeine reinforcement has been demonstrated in 100% of subjects with histories of heavy caffeine use and abuse of alcohol or drugs (49, 50, 51) and in a somewhat lower proportion (about 45%) of subjects with histories of moderate and heavy caffeine use alone (57, 65, 66, 68, 88, 89, 100). However, one recent study (100) demonstrated caffeine reinforcement in 100% of moderate users when they were given caffeine discrimination training and required to perform a vigilance task after capsule administration.

Caffeine Reinforcement Is an Inverted U-Shaped Function of Dose

Caffeine reinforcement appears to be an inverted U-shaped function of dose. A dose as low as 25 mg per cup of coffee has been shown to function as a reinforcer under conditions in which subjects could repeatedly self-administer that dose within the day (66, 89). A robust finding has been that increasing doses beyond 50 or 100 mg tends to decrease choice of caffeine or rates of caffeine self-administration (51, 57, 66, 74, 102). Relatively high doses of caffeine (e.g., 400 or 600 mg as a single dose) have been shown to produce significant caffeine avoidance (57).

Subjective Effects of Caffeine and Placebo Co-vary with Reinforcement or Choice

A robust finding from studies of caffeine reinforcement has been that qualitative ratings of subjective effects have generally co-varied with measures of reinforcement or choice. A good example is provided from a choice study by Evans and Griffiths (32) that examined the subjective effects of placebo and caffeine on forced-exposure days. When these data were retrospectively categorized into caffeine choosers and nonchoosers, a face-valid profile of changes in subjective effects emerged. It was found that (a) choosers showed "positive" subjective effects of caffeine relative to placebo (e.g., increased alert, content, energetic, liking), (b) nonchoosers showed "negative" effects of caffeine relative to placebo (e.g., increased anxiety, mood disturbance, jittery), and (c) choosers showed "negative" effects of placebo (e.g., increased headache, fatigue).

The results of most caffeine reinforcement and choice studies reported to date are consistent with these generalities drawn from Evans and Griffiths (32). For example, in studies involving subjects with histories of heavy coffee drinking and drug abuse, experimental conditions that have maintained higher levels of self-administration or greater choice have generally been associated with higher ratings of liking or "positive" subjective effects (49), 50, 51). Analysis of subjective effect data from a choice study by Stern et al. (102) also showed that caffeine produced positive subjective effects in choosers and negative effects in nonchoosers. That study provided little evidence that placebo was associated with negative effects in choosers, probably because subjects were not abstinent from their normal dietary caffeine. Also consistent, Griffiths and Woodson (57) reported that caffeine choice was positively related to ratings of content/satisfied and negatively related to ratings of mood disturbance and dislike effects following caffeine. Finally, Hughes et al. (67) analyzed data from a series of four previous choice studies and showed that caffeine choice was positively related to ratings of headache and drowsiness following placebo.

Despite these reasonably consistent findings across a range of different studies, it is important to recognize that the relationship between reinforcement or choice and subjective effects has not been invariant. For example, although Hughes et al. (67) demonstrated an inverse relationship between caffeine choice and the occurrence of drowsiness, they also provided examples in which this relationship was absent.

Future research should determine whether subjective effect changes in response to caffeine and placebo challenges can prospectively predict subsequent, experimentally independent caffeine reinforcement. The measures of subjective effects in all of the studies cited above have been obtained at the same time that reinforcement or choice has been assessed. In the choice studies, for example, subjective effects have been obtained during the same forced-exposure days which provided the basis for the subsequent choice.

Finally, as a historical note, the profile of subjective effects that have emerged from reinforcement studies to date are entirely consistent with the results of a survey study (41) and a companion caffeine administration study (43) that did not investigate reinforcement per se. These studies showed that after overnight caffeine abstinence, heavy coffee users reported positive subjective effects of coffee drinking or caffeine administration in contrast to coffee abstainers or light users, who reported unpleasant and undesirable subjective effects. In addition, compared to abstainers or light users, heavy users reported undesirable dysphoric effects of coffee abstinence or placebo.

Behavioral Context Modulates Caffeine Reinforcement

One recent study demonstrated that caffeine reinforcement can be enhanced by explicitly manipulating the behavioral requirements following drug ingestion (100). In that study, all subjects demonstrated caffeine reinforcement in one condition or more when a computer vigilance activity was required following drug ingestion. In contrast, only two of seven subjects showed caffeine reinforcement when a relaxation activity was required following drug ingestion.

Physical Dependence May Potentiate Caffeine Reinforcement

An important implication of the findings described above showing a relationship between abstinence-associated headache, fatigue and drowsiness, and either choice/reinforcement or heavy caffeine use is that physical dependence may potentiate the reinforcing effects of caffeine. In this regard, a retrospective analysis by Hughes et al. (67) of data from four previous studies involving choice between caffeinated coffee (100 mg/cup) and decaffeinated coffee showed that subjects who reliably reported more headache with decaffeinated than with caffeinated coffee were 2.6 times more likely to show reliable choice of caffeinated coffee.

A direct experimental test of the hypothesis linking physical dependence to reinforcement was provided in a choice study by Griffiths et al. (49) in heavy caffeine users who first received forced exposure to caffeinated and decaffeinated coffee on different days and then, on a subsequent choice day, made a mutually exclusive choice between the two coffees. When subjects had received only caffeinated coffee for a week or more before the choice tests (and presumably were physically dependent), subjects reliably chose, and reported more liking for, caffeinated over decaffeinated coffee. In contrast, when subjects had received only decaffeinated coffee for a week or more before the choice tests (and presumably were not physically dependent), subjects did not reliably choose caffeinated coffee, nor were there pronounced differences in ratings of liking.

Evans and Griffiths (32) failed to replicate the effect of caffeine physical dependence on caffeine choice in a study in which two groups of normal subjects had a choice test between capsules containing 600 mg caffeine or placebo after being exposed to either 900 mg/day caffeine or placebo for 18 days. Although the chronic caffeine group was shown to be both tolerant and physically dependent (i.e., showed withdrawal), the groups did not differ in the percentage choice of caffeine (44% versus 31% in the chronic caffeine and placebo groups, respectively). The authors speculated that the duration of placebo exposure was too short to result in maximal withdrawal after the high chronic dose of caffeine (900 mg/day). Also, the dose of caffeine used in the choice tests (600 mg) was higher than that shown to be a reinforcer in previous dose-effect studies (57, 66).

Thus, although individuals who report caffeine withdrawal are more likely to show caffeine reinforcement, the two studies (32, 49) that attempted to demonstrate a potentiation of caffeine reinforcement by providing a history of chronic caffeine administration have provided equivocal results. Finally, it is also clear that a history of chronic caffeine administration is not a necessary condition for caffeine to function as a reinforcer. One study demonstrated caffeine reinforcement in a subject who normally abstained from all caffeine (57), and two other studies demonstrated caffeine reinforcement in subjects after they were caffeine-free for several weeks (50, 100).

Effects of Prestudy Caffeine Intake on Caffeine Reinforcement

Although a reasonable a priori prediction would be that amount of prestudy caffeine use should predict the incidence of reinforcement in experimental studies, such a relationship has not been demonstrated (57, 67, 102), and there has been one instance in which caffeine reinforcement was demonstrated in an individual who normally abstained from caffeine use (57). Given that subjective effects have been shown to co-vary with caffeine reinforcement, these results would appear to be at variance with studies by Goldstein and colleagues (41, 43), who showed that after overnight caffeine abstinence, heavy coffee users (5 or more cups/day) reported pleasant and desirable effects of coffee drinking and caffeine in contrast to coffee abstainers or light users, who reported more unpleasant and undesirable subjective effects. It seems likely that the subject samples investigated to date in caffeine reinforcement studies have been too small and homogeneous to demonstrate what appears to be a relatively weak relationship between prestudy caffeine use and caffeine reinforcement.

Effects of Prestudy Anxiety Levels on Caffeine Reinforcement

In two studies, caffeine choice has been significantly negatively correlated with prestudy anxiety levels (32, 57). In one of these studies, the two subjects with the highest prestudy anxiety had the lowest incidence of caffeine choice and demonstrated significant caffeine avoidance at a lower dose than any other subject (57). These results are consistent with questionnaire studies that have shown that anxiety and panic disorder patients report less consumption of caffeine than controls (9, 76, 77, 104), presumably because such patients are particularly sensitive to anxiogenic and related dysphoric subjective effects of caffeine (13, 17, 104).


Subjective effects of a drug usually refer to drug-induced changes in an individual's experiences or feelings that are not accessible to independent verification by an observer. Reasonably sophisticated methods for assessing human subjective effects of drugs were first developed about 40 years ago (5, 72) and have been extensively used to evaluate caffeine. Typically, these methods involve having subjects self-rate their moods, feelings, or behaviors on questionnaires after double-blind administration of drug.

Qualitative Subjective Effects of Caffeine

Laboratory examination of subjective effects (i.e., self-reported ratings of mood) of caffeine has often led to equivocal and sometimes confusing results. Many studies have failed to find significant subjective effects of caffeine either at usual dietary doses (40-100 mg) (e.g., see refs. 14, 18, and 79) or at doses as high as 200-500 mg (e.g., see refs. 18, 79, 81, 82, and 86). Other studies have shown that high doses of caffeine (200-800 mg) produce a predominantly "dysphoric" profile of subjective effects characterized by increases in anxiety, nervousness, or jittery (e.g., see refs. 15, 16, 30, 80, 83, 87, and 91).

The total lack of subjective effects and the dysphoric subjective effects of caffeine are at variance with the common experience of regular caffeine users who generally report positive or desirable subjective effects after consuming caffeine (41). In contrast to the studies cited above, an increasing number of studies have reported a profile of predominately positive subjective effects after caffeine administration (53, 75, 78, 85, 99, 100; also see studies cited in section on subjective effects and caffeine reinforcement). The types of positive subjective effects that have been significantly affected by caffeine include (a) increases in ratings of well-being, energy/active, alert, concentration, self-confidence, motivation for work, and desire to talk to people and (b) decreases in ratings of sleepiness and muzziness/not clearheaded (53, 85, 99).

There seem to be at least three factors that may increase the likelihood of demonstrating such positive subjective effects of caffeine: (i) testing in caffeine deprivation or total abstinence; (ii) testing low caffeine doses; and (iii) testing under conditions (or in populations) in which caffeine functions as a reinforcer.

First, most studies that demonstrated positive subjective effects of caffeine have tested caffeine after a significant period of caffeine deprivation or abstinence [e.g., overnight or 24 hr (53, 75, 78, 85, 99, 100)]. That studies have also shown such effects in subjects who were maintained on an otherwise caffeine-free diet (85, 99) indicates that physical dependence is not a necessary condition for demonstrating positive subjective effects of caffeine. Second, the positive subjective effects of caffeine have been most often demonstrated at relatively low caffeine doses [e.g., 20-200 mg (53, 75, 78, 85, 99, 100)]. Although there appear to be wide individual differences in relative sensitivity, it is clear that dysphoric/anxiogenic subjective effects of caffeine emerge in a dose-related fashion (15, 30, 87). Third, the positive subjective effects of caffeine are more likely to be demonstrated in individuals in whom caffeine is presumably functioning as a reinforcer (32, 43, 57); see section on subjective effects and caffeine reinforcement).

When caffeine produces positive changes in subjective effects, the profile of these changes (e.g., increases in well-being, energy/active, alert, concentration, self-confidence, motivation for work, desire to talk to people) is remarkably similar to that produced by d-amphetamine and cocaine (34). A major difference in the subjective effect profile appears to be that caffeine is more likely to produce dysphoria/anxiety with increases in dose than d-amphetamine and cocaine.

There is some evidence suggesting a dissociation between the positive/energy effect and the dysphoric/anxiogenic effect of caffeine. Two studies that administered 300 mg of caffeine to a group of subjects showed virtually no correlation between caffeine-induced stimulation and anxiety (14, 42). Examination of individual subject data from these and other studies (30, 53, 99) indicates that these two types of subjective effects can occur alone or together in the same individual.

Finally, results of questionnaire and experimental studies suggest that panic disorder patients may be particularly vulnerable to the anxiety- and panic-producing effects of caffeine. Questionnaire studies indicate that such patients self-report consuming less caffeine and experiencing more anxiety than control subjects (9, 76 77, 104). An experimental study by Charney et al. (17) showed that, compared to normal controls, panic disorder patients had more anxiety and a greater rate of panic attacks after blind oral administration of caffeine [10 mg/kg caffeine citrate: equivalent to 5 mg/kg (about 350 mg) caffeine base] after being instructed to follow a caffeine-free diet for 2 weeks. Similar results have been described by Uhde (104), who compared the effects of 480 mg of caffeine in panic disorder patients and normal controls.


Tolerance refers to an acquired change in responsiveness of an individual as a result of exposure to drug such that an increased dose of drug is necessary to produce the same degree of response, or that less effect is produced by the same dose of drug. The development of tolerance is most unambiguously concluded from studies that compare full dose-response curves in the presence and absence of the tolerance-inducing manipulation (73). As reviewed elsewhere (62), preclinical research has clearly demonstrated tolerance to response-rate-decreasing, behavioral stimulant, and discriminative stimulus effects of caffeine, with the extent and rate of tolerance development differing across the different measures.

Information about the development of caffeine tolerance in humans is relatively easily obtained by measuring any pharmacological response before and after caffeine exposure. As described below, "complete" tolerance development (i.e., caffeine effects no longer different from placebo) has been demonstrated after repeated administration of relatively high caffeine doses spread across the day for a number of consecutive days. Beyond these demonstrations, however, detailed quantitative knowledge of caffeine tolerance in humans is quite fragmentary.

Tolerance to Caffeine Subjective Effects

Tolerance to the subjective effects of caffeine has been demonstrated only in one recent study (32), in which two groups of subjects received either caffeine (300 mg t.i.d.) or placebo (t.i.d.) for 18 consecutive days. During the last 14 days of chronic dosing, the caffeine and placebo groups did not differ meaningfully on ratings of mood and subjective effects. Furthermore, after chronic dosing, compared to placebo, caffeine (300 mg b.i.d.) produced significant subjective effects (including increases in tension-anxiety, jittery/nervous shaky, active/stimulated/energetic and strength of drug effect) in the chronic placebo group but not in the chronic caffeine group, suggesting the development of complete tolerance at these doses. Fig. 3 shows these effects for ratings of tension-anxiety and strength of drug effect.

Tolerance to Caffeine Sleep Disruption

Unambiguous evidence for the development of tolerance to caffeine sleep-disrupting effects is also quite recent. Although Colton et al. (21) showed that heavy coffee drinkers were less sensitive to the sleep-disturbing effects of caffeine than light coffee drinkers, the study design did not allow differentiation of tolerance from other preexisting differences between these self-selected subject populations. Two recent studies provided direct experimental evidence for caffeine tolerance to sleep disruption by demonstrating decreases in caffeine-induced disruption of objective measures of sleep after caffeine dosing of 250 mg b.i.d. for 2 days (111) or 400 mg t.i.d. for 7 days (8). By day 7 in the latter study, a number of sleep measures (e.g., total sleep time, sleep efficiency, number of awakenings) were no longer different from baseline, suggesting the development of complete tolerance at these doses.

Tolerance to Caffeine Physiological Effects

There is good evidence for decreased responsiveness to physiological effects of caffeine with repeated daily caffeine administration. Such tolerance development has been demonstrated to various physiological effects of caffeine, including diuresis (29), parotid gland salivation (108), increased metabolic rate (oxygen consumption) (8), increased blood pressure (2, 93), increased plasma norepinephrine and epinephrine, and increased plasma renin activity (93). Compared to a group that received placebo, complete tolerance to caffeine (250 mg t.i.d.) was demonstrated in 1-4 days on blood pressure, plasma norepinephrine and epinephrine, and plasma renin activity (93). Results consistent with these have been reported on other cardiovascular and physiological responses (2, 24), although whether complete tolerance development occurs has been questioned (24). The only human study to quantitatively assess the extent of caffeine tolerance by constructing dose-effect curves (as has been done frequently in infrahuman research) was a study that showed a two- to threefold decrease in the minimally effective dose required to produce diuresis after a period of abstinence from caffeine (estimated 170-340 mg/day) (29).

Parametric Determinants of Tolerance

Beyond the several demonstrations of complete tolerance development at high caffeine doses described above, there is little quantitative knowledge about the parameters that determine caffeine tolerance. As with the development of tolerance to most drugs, the degree of tolerance development to caffeine can be expected to depend on the caffeine dose, the dose frequency, the number of doses, and the individual's elimination rate (97). There is some indication that the rate and/or extent of tolerance development differ across different measures (8, 24). The rate of tolerance development has been estimated to be quite rapid for blood pressure [t1/2 = 1 hr (97)], with complete tolerance to various cardiovascular effects occurring in 2-5 days (2, 24, 93). Little is known about the rate of loss of tolerance except that it was shown to be rapid for blood pressure (t1/2 = 1 hr) in one study (97), but in another study (29) was suggested to be quite slow for caffeine-induced diuresis.

Relationship Between Caffeine Tolerance and Withdrawal

Tolerance and physical dependence are sometimes thought to be functionally interrelated, reflecting common neuroadaptive changes in response to repeated drug administration. The only relevant study to address this issue is the one described above that demonstrated tolerance to caffeine subjective effects in a group of subjects that received caffeine (300 mg t.i.d.) compared to a group that received only placebo (32). In that study, the group of tolerant subjects showed withdrawal (a time-limited elevation in headache) when switched to placebo compared to the nontolerant placebo group. Overall, this study provides only limited information suggesting the covariation of tolerance and withdrawal.


Physical dependence is manifested by time-limited biochemical, physiological, and behavioral disruptions (i.e., a withdrawal syndrome) upon termination of chronic or repeated drug administration. As has been reviewed elsewhere (56), preclinical research has clearly demonstrated time-limited behavioral disruptions following cessation of chronic caffeine administration. As for human research, 37 case reports and human experimental studies of caffeine withdrawal spanning the period 1833 to 1987 have been reviewed by Griffiths and Woodson (56). The following section updates that review based on 16 additional recent studies that have significantly advanced our understanding of caffeine withdrawal (22, 31, 32, 33, 35, 54, 63, 65, 67, 68, 69, 84, 92, 98, 103, 105).

Signs and Symptoms of Caffeine Withdrawal

The most frequently reported withdrawal symptom is headache (also cerebral fullness), which is characterized as being gradual in development, diffuse, throbbing, and sometimes severe (56). Other symptoms, in roughly decreasing order of prominence, are: (a) drowsiness (increased sleepiness and yawning; decreased energy and alertness), (b) increased work difficulty (decreased motivation for tasks/work, impaired concentration), (c) decreased feelings of well-being/contentment (decreased self-confidence; increased irritability), (d) decreased sociability/friendliness/talkativeness, (e) flu-like feelings (muscle aches/stiffness; hot or cold spells; heavy feelings in arms or legs; nausea), and (f) blurred vision (54, 56, 98). In addition to these symptoms, composite scales of depression and anxiety may be elevated (98) and psychomotor performance may be impaired (56, 92, 98). The occurrence of headache as a withdrawal symptom does not necessarily correlate with the occurrence of other symptoms (e.g., tiredness), suggesting that other signs and symptoms are not merely epiphenomena of headache (54, 56).

Severity of Caffeine Withdrawal

The severity of caffeine withdrawal is an increasing function of caffeine maintenance dose (31, 56). When symptoms of caffeine withdrawal occur, the severity can vary from mild to extreme. At its worst, caffeine withdrawal is incompatible with normal functioning and is sometimes totally incapacitating (56).

Incidence of Caffeine Withdrawal

The incidence of caffeine withdrawal is an increasing function of caffeine maintenance dose (33, 35, 56). The best estimates of the incidence of caffeine withdrawal in the general population come from a recent survey study and an experimental study. A recent random digit-dial telephone survey in Vermont showed that among current users of caffeine who reported that they had abstained from caffeine for 24 hr, 27% reported withdrawal headaches when they abstained (63, 69). The experimental study (98) involved 62 individuals from the general community with a distribution of caffeine intake similar to that of the general population in the United States (mean caffeine intake of 235 mg/day). The study involved an approximately 48-hr, double-blind caffeine abstinence trial under conditions which obscured that the purpose of the study was to investigate caffeine. During caffeine withdrawal, 52% reported moderate or severe headache, and 8-11% showed abnormally high scores on standardized depression, anxiety, and fatigue scales. The incidence of headache observed from the survey and experimental study in the general population (27-52%) is in the range of that observed in several other recent studies conducted in special subject populations [19-57%, (54, 67, 105)].

Minimum Dosing Parameters for Caffeine Withdrawal

Although the incidence and severity of caffeine withdrawal are an increasing function of caffeine dose, two recent studies have shown that caffeine withdrawal can occur after relatively long-term administration of caffeine doses as low as 100 mg/day (31, 54). Studies also indicate that caffeine withdrawal headache may occur after termination of high doses of caffeine (terminal doses of {ewc MVIMG, MVIMAGE,!greateq.bmp}600 mg/day) administered for as few as 6-15 days (27, 49).

Time Course of Caffeine Withdrawal

The caffeine withdrawal syndrome follows an orderly time course (56). Onset has been usually reported to occur 12-24 hr after terminating caffeine intake, although onset as late as 36 hr has been documented (54). Peak withdrawal intensity has generally been described as occurring 20-48 hr after abstinence. The duration of caffeine withdrawal has most often been described as ranging between 2 days and 1 week, although longer durations have been occasionally noted (32, 54, 56, 105). Fig. 4 illustrates the time course of increased headache and decreased energy/active ratings after substitution of placebo capsules for caffeine capsules (100 mg/day).

Clinical Significance of Caffeine Withdrawal

The importance of caffeine withdrawal is derived in part from the large population at risk: 82% of adults in the United States consume caffeine regularly, with mean intake estimated at 227 mg/day (44). Even if a conservative 10% incidence of clinically significant withdrawal were assumed, the size of the vulnerable population is enormous. Caffeine use and abstinence should be considered in the differential diagnosis of commonly reported symptoms of headache, fatigue, and mood disturbances. Research is needed to determine the necessity of the common medical practice of requiring short-term caffeine abstinence before various laboratory tests (e.g., serum lipid profiles, fasting blood sugar) as well as in preparation for operations and a variety of procedures such as endoscopies or colonoscopies. Finally, more research is needed to explore possible situations involving caffeine withdrawal morbidity. For example, a neonatal withdrawal syndrome following chronic maternal caffeine ingestion has been proposed (84) and caffeine withdrawal headache has been implicated in postoperative morbidity (33, 35), in a weekend migraine syndrome (22), and as a possible mechanism underlying inappropriate chronic daily use of caffeine-containing analgesics (12, 19, 46).


Throughout history, caffeine has intermittently been labeled as a drug of abuse (see refs. 4, 37, and 58), and analogies have been made to classic abused substances. This suggestion is usually controversial and is rebutted with the following observations: The majority of caffeine use is consistent with socially accepted limits and patterns; gross overconsumption of any article of diet can be harmful; and the adverse effects of excessive intake seem to be largely transient (26). Certainly there is an important distinction to be made between legal, socially domesticated drugs (such as caffeine, nicotine, and alcohol) and illegal drugs (such as cocaine, heroin, and phencyclidine). Instances of appropriate use usually can be identified and agreed upon with the former but not the latter. Beyond the legal distinction, however, definitions of what constitutes a drug of abuse or dependence are often complex and open to alternative interpretations, as has been clearly illustrated in a recent debate about whether or not nicotine is meaningfully considered to be a drug of abuse (94, 95, 107). Ultimately, whether or not caffeine is labeled as a drug of abuse is a somewhat arbitrary social-political decision subject to revision over time. This section will consider the question of caffeine as a drug of abuse from two current perspectives: a clinical psychiatric perspective and a perspective based on a broader reinforcement/adverse effects analysis.

Clinical Psychiatric Perspective

Generic criteria for making a diagnosis of Psychoactive Substance Dependence Disorder have been developed by the American Psychiatric Association (DSM-III-R and DSM-IV) (1) to facilitate treatment and research on abuse of behaviorally active drugs. In considering caffeine, the most relevant criteria are: (a) persistent desire or unsuccessful efforts to cut down or control use (b) use despite knowledge of physical or psychological problem caused or exacerbated by use, (c) marked tolerance, and (d) characteristic withdrawal syndrome or taking substance to relieve or avoid withdrawal symptoms. These criteria largely overlap with those promulgated by the World Health Organization (ICD-10) (109) for also making a diagnosis of psychoactive substance dependence. Recently, Hughes et al. (70) reviewed studies relevant to making a diagnosis of caffeine dependence and concluded that although there was a substantial research database suggesting the validity of caffeine dependence, an appropriate clinical patient database was almost completely absent. Even more recently, these same investigators (69) reported the results of a random-digit telephone survey in Burlington, Vermont, in which they applied the generic DSM-III-R criteria for drug dependence to current caffeine users. Surprisingly, 44% of current users (about 36% of the general population) were diagnosed as drug-dependent on caffeine by fulfilling three or more of the generic criteria. The most common criterion endorsed was persistent desire or unsuccessful efforts to cut down or control use. Although the sample size from this study was small and the validity of applying to caffeine the generic criteria for psychoactive substance dependence needs to be addressed, the estimate that about one-third of the general population may be dependent on caffeine is startling and worthy of further investigation.

Reinforcement/Adverse Effects Analysis

A broader approach to considering whether it is meaningful to label caffeine as a drug of abuse is to examine the extent to which caffeine has the defining characteristics of a drug of abuse. As discussed elsewhere (55), drugs of abuse have two major characteristics: (i) they have reinforcing effects and (ii) they produce adverse effects (i.e., they have the capacity to harm the individual and/or society). According to this model, the relative abuse liability of a drug can be considered to be a multiplicative function of the degree of reinforcing effect and the degree of adverse effect.

With regard to the first characteristic of abused drugs (i.e., reinforcing effects), as has been reviewed in previous sections, it is now clear from research with laboratory animals and with humans that caffeine can function as a reinforcer under some conditions. The results of intravenous caffeine self-injection studies in nonhuman primates and rats indicate that caffeine functions as a reinforcer under a more limited range of conditions than do classic psychomotor stimulant drugs of abuse such as d-amphetamine or cocaine. This conclusion about caffeine is similar to that previously reached for nicotine (28, 40). The results of self-administration and choice studies in humans clearly demonstrate the reinforcing effects of low and moderate doses of caffeine. Studies of subjective effects have demonstrated that amphetamine generally produces greater elevations than caffeine in ratings indicating "euphoria" and "well-being" (14, 15, 106). The only human study to directly compare the self-administration of caffeine with that of another drug was a study that examined the effects of response requirement on consumption of coffee and cigarettes, and it concluded that the reinforcing effects of coffee and cigarettes were comparable (6). Overall, the available animal and human research suggests that caffeine is a reinforcer, but a less robust reinforcer than cocaine or d-amphetamine, and perhaps most similar to nicotine.

With regard to the second characteristic of abused drugs (i.e., adverse effects), a balanced discussion of adverse effects of caffeine is beyond the scope of this chapter and has been the focus of a number of recent books (36, 71, 101). Briefly, the adverse effects of caffeine use include caffeine intoxication, caffeine withdrawal signs and symptoms, and exacerbation of medical conditions. Acute or chronic caffeine intoxication, also known as caffeinism, is a DSM-IV diagnostic category which may be indistinguishable from manic episodes, panic disorder, and generalized anxiety disorder (1), and which may occur in about 10% of the population (47, 69). In chronic or recurrent forms of caffeinism, patients often will not recognize the ingestion of caffeine-containing foods as problematic and may seek medical treatment for anxiety, insomnia, or cardiovascular, gastrointestinal, and other somatic complaints (47). Caffeine withdrawal signs and symptoms, which have been discussed in a previous section, may occur at clinically significant levels in about 10% of caffeine consumers who abruptly abstain. In terms of exacerbating medical conditions, the majority of medical specialists recommend reducing or eliminating caffeine for a variety of conditions including, anxiety, insomnia, palpitations, tachycardia, arrhythmia, fibrocystic disease, esophogitis/hiatal hernia, ulcers, and pregnancy (64), although the scientific basis for at least some of these recommendations appears ambiguous (36, 71, 101). Importantly, significant health risk from nonreversible pathological consequences of chronic caffeine use (e.g., heart disease, cancer, human reproduction) has not been conclusively demonstrated, although here too the status of the present data remains somewhat ambiguous (36, 71, 101).

This analysis indicates that caffeine has the two defining characteristics of prototypic drugs of abuse. Inasmuch as the relative abuse liability of a drug can be considered to be a multiplicative function of the degree of reinforcing effect and the degree of adverse effects, the modest reinforcing effect and modest adverse effects documented to date would suggest a low abuse potential relative to more widely recognized drugs of abuse. This analysis also predicts that if future research reveals significant additional caffeine-associated health risk, the perception of the relative abuse liability of caffeine would increase appreciably, as has occurred with nicotine over the last 20 years.


Preparation of this review was supported, in part, by National Institute on Drug Abuse grants RO1 DA01147 and RO1 DA03889.

published 2000