Effect of Caffeine on the Recognition of and Responses to Hypoglycemia in Humans

Methods

Participants

Eight healthy, nonobese volunteers (5 men; age range, 20 to 33 years) gave written, informed consent for the study, which was approved by the Human Investigations Committee of Yale University. None had any history or clinical evidence of medical illness or was taking any prescription or over-the-counter medications. Each volunteer participated in two studies performed at least 1 week apart. Seven participants drank no tea or coffee, and their caffeine consumption from other sources (usually cola) was less than 300 mg per month. One participant consumed coffee on a regular basis (two to three cups per day). All participants abstained completely from caffeine-containing food and drinks for 3 days before each study.

All studies were done at the Yale Clinical Research Center on the morning of the study after a 10-hour overnight fast. Two intravenous catheters were inserted under local anesthesia: one in an antecubital vein for infusion of insulin and 20% glucose solution and a second, inserted retrograde into a dorsal hand vein, for sampling of arterialized venous blood. The hand was placed in a heated box (65 C) and the cannula was kept patent by a 154-mmol/L NaCl infusion. Potential distractions, such as conversation and other background noise, were minimized throughout each study. After the cannulae were inserted, the following baseline measurements were made.

Measurements

The middle cerebral artery velocity (V_MCA) was measured using a transcranial Doppler technique (Transpect Medasonics; Freemont, California). The left middle cerebral artery velocities were continuously monitored throughout the study, and the mean velocity was recorded every minute as previously described [9]. Mean V_MCA measurements during the final 20 minutes of each stage of the study were averaged. Assessment of V_MCA appears to be a reliable indicator of cerebral blood flow during hypoglycemia, and the changes in blood flow and velocity have been reported to be symmetric between cerebral hemispheres during insulin-induced hypoglycemia [10].

Cognitive function was examined using P300 auditory evoked responses (Medelec Ltd; Surrey, UK) as previously described [11]. The P300 component of event-related potentials appears to be a reliable indicator of cognitive function and appears to relate incoming sensory information to memory updating processes. P300 latencythe interval from the stimulus to the peak of the evoked potentialremains stable within individuals during repeated testing and is unaffected by gender [12, 13]. Briefly, scalp electrodes were placed on the vertex and right and left mastoids. P300 evoked potentials were obtained with an auditory (oddball) categorization task in which participants silently counted a number of high-pitched clicks delivered in a train of frequent low-pitched and infrequent high-pitched clicks. Following the P300 measurements, the participants completed a visual analog assessment of symptoms characteristically associated with hypoglycemia (facial flushing, palpitations, tingling, trembling, sweating, hunger, light-headedness, weakness, anxiety, difficulty concentrating, and sleepiness) [13]. They were asked to grade the symptoms on a scale of 1 to 100 with 100 being the most severe. Awareness of hypoglycemia was determined using a similar visual analog scale. Symptoms and cognitive function were assessed at the end of the euglycemic and both hypoglycemic periods. Participants had been told at recruitment about some of the sensations that they might experience during hypoglycemia and that they may notice a nonspecific difference in the way they felt. They were also told at that time that their blood glucose would be lowered at some stage during each of the visits but they were not told the level at any time.

To measure hormone and caffeine levels, blood was taken from the heated hand vein for subsequent measurement of glucagon, epinephrine, norepinephrine, cortisol, growth hormone, insulin, and caffeine levels every 15 to 30 minutes. Plasma insulin, C-peptide, glucagon, cortisol, and growth hormone levels were measured by double-antibody radioimmunoassays and catecholamines by a radioenzymatic technique (Amersham; Arlington Heights, Illinois). Plasma caffeine levels were quantitated by EMIT (Enzyme Multiplication Immunoassay Technique; Syva, Palo Alto, California) using a BMD Hitachi 717 autoanalyzer (Indianapolis, Indiana).

Arterialized venous PCO ₂ and hematocrit were measured at baseline, at the end of the euglycemic period, and at the end of both hypoglycemic periods because alterations in these variables can affect cerebral blood flow [14].

After the baseline variables were measured, a stepped euglycemic, hypoglycemic, hyperinsulinemic (2 mU/kg body weight per minute) glucose clamp procedure [15] was used to examine the effect of caffeine on the responses to euglycemia and hypoglycemia. With this technique, exogenous glucose is infused at a variable rate to maintain plasma glucose at predetermined levels, based on duplicate glucose measurements made every 5 minutes at the bedside by a glucose oxidase method (Beckman Instruments; Fullerton, California). For each protocol, plasma glucose levels were clamped at 5 mmol/L for 90 minutes (euglycemic phase). After the first 30 minutes of euglycemia, participants consumed, in a randomized, double-blind design, either caffeine-free diet cola alone or the same cola to which 400 mg of caffeine had been added, and plasma glucose was maintained at 5 mmol/L for a further 60 minutes. After completion of the euglycemic phase, plasma glucose was lowered to 3.8 mmol/L over 20 minutes and kept there for the remainder of the first hypoglycemic step (80 minutes in total). Thereafter, plasma glucose was lowered further to 2.8 mmol/L and maintained at that level for the second 80-minute hypoglycemic step.

As a further control, four of the participants also participated in a third study, which examined the effects of acute ingestion of caffeine on V_MCA and counter-regulatory hormone levels in the absence of exogenous infusions of insulin or glucose. After abstaining from caffeine-containing foods and drink for 3 days, a single retrograde cannula was inserted as described above. Baseline measurements were made of plasma glucose, insulin, and counter-regulatory hormone levels. Participants then consumed 400 mg of caffeine diluted in caffeine-free diet cola, and V_MCA, glucose, and hormone measurements were repeated every 15 to 30 minutes over the next 4 hours. No insulin or exogenous glucose was given.

Statistical Analyses

Statistical analyses were done using repeated measures analysis of variance. Where analyses of variance (ANOVA) showed a significant group time interaction, contrasts in means were compared using the paired Student t-test. When data were not normally distributed, comparisons were made after logarithmic transformation or with Kruskal-Wallis or Wilcoxon signed-rank tests. Results are expressed as individual means with point estimate of differences between means and 95% confidence intervals presented as percentage differences. Otherwise, data are shown as either mean SE, mean (95% confidence interval), or median (25% to 75% range).

Results

In both clamp studies, plasma glucose levels in patients receiving caffeine and placebo were identical at baseline (4.9 0.1 mmol/L in the caffeine study and 4.9 0.1 mmol/L in the placebo study); during the final 20 minutes of the euglycemic period (5.0 0.1 mmol/L versus 4.9 0.1 mmol/L); and during the final 20 minutes of the first (3.7 0.1 mmol/L versus 3.8 0.1 mmol/L) and second (2.8 0.1 versus 2.8 0.1 mmol/L) hypoglycemic stages (Figure 1). At the same times, plasma insulin levels were similar between studies (930 54 versus 942 30; 864 54 versus 918 30 pmol/L and 882 84 versus 948 78 pmol/L, respectively). Coefficients of variation for plasma glucose levels achieved during each stage were less than 5%. Arterialized PCO ₂ and hematocrit were unchanged from baseline throughout each study (data not shown). Caffeine levels were undetectable at the start of both hypoglycemia studies. After ingestion of caffeine, plasma levels increased to 7.8 1.3 g/mL by 60 minutes and remained elevated for the duration of the study (Figure 1).

View larger version:

• In this window
• In a new window

Figure 1. After 30 minutes, a drink containing caffeine-free diet cola with or without caffeine (400 mg) was given. At baseline in the caffeine study and throughout the placebo study, plasma caffeine levels were below the limit of detection of the assay. Mean ( SE) plasma glucose and caffeine levels during both studies.

Comparisons of counter-regulatory hormone levels between studies were made during the final 30 minutes of each period (baseline, 5, 3.8, and 2.8 mmol/L). Catecholamine, growth hormone, and cortisol levels were similar at baseline and did not change statistically during the first 90 minutes (euglycemia) of the placebo study (Figure 2). After caffeine was ingested, only epinephrine levels were significantly higher at the end of the euglycemic period compared with placebo (311 versus 186 pmol/L; point estimate of difference, +256 pmol/L [CI for mean percentage paired difference, 35% to 290%], P < 0.02 difference). When plasma glucose was lowered to 3.8 mmol/L, however, epinephrine levels were two to three times higher (1725 versus 791 pmol/L) in the caffeine study (+934 pmol/L [76% to 158%], P < 0.001 by repeated measures ANOVA) compared with placebo. Similarly, norepinephrine (1.58 versus 1.12 nmol/L; +0.46 nmol/L [CI, 26% to 60%], P < 0.002), cortisol (676 versus 411;+265 nmol/L [CI, 26% to 78%], P < 0.008) and growth hormone (20.9 versus 13.1; +7.8 g/L [CI, 16% to 143%], P < 0.05) levels were markedly elevated in the group that had ingested caffeine compared with the placebo group. At a plasma glucose level of 2.8 mmol/L, levels of epinephrine (4072 versus 2915 pmol/L; +1157 pmol/L [CI, 15% to 85%], P < 0.05), norepinephrine (2.27 versus 1.79 nmol/L; +0.48 nmol/L [4% to 74%], P < 0.05), and cortisol (963 versus 775 nmol/L; +188 nmol/L [CI, 8% to 42%], P < 0.05) were higher after caffeine compared with placebo, whereas growth hormone did not differ (32.1 versus 37.9 g/L; +7 g/L [CI, 38% to 25%], P > 0.05). Glucagon levels were almost identical at baseline (122 15 versus 115 13 ng/L [mean SE]) and throughout both studies; 83 12 versus 82 12 ng/L at the end of the euglycemic period, 96 9 versus 92 10 ng/L when plasma glucose was lowered to 3.8 mmol/L, and 125 12 versus 131 14 ng/L when plasma glucose was 2.8 mmol/L.

View larger version:

• In this window
• In a new window

Figure 2. When plasma glucose was lowered to 3.8 mmol/L, epinephrine ( < 0.001), norepinephrine ( < 0.002), cortisol ( < 0.008), and growth hormone ( < 0.05) responses were greater after ingestion of caffeine compared with placebo. At a plasma glucose level of 2.8 mmol/L, epinephrine, norepinephrine, and cortisol (all < 0.05) levels were higher in the caffeine study. Glucagon levels were almost identical throughout each study (see text). Levels (mean SE) of epinephrine, norepinephrine, cortisol, and growth hormone.PPPPP

Baseline mean middle cerebral artery velocities (V_MCA) were identical in both studies (Figure 3). Following caffeine intake, the V_MCA immediately decreased from 64 to 49 cm/s ( 15[CI, 10 to 21 cm/s]; P < 0.001), which was sustained for the duration of the study. In contrast, no significant change in V_MCA was noted during any of the steps after placebo.

View larger version:

• In this window
• In a new window

Figure 3. The mean values ( SE) are shown at baseline (5 mmol/L), at the end of euglycemia, and when plasma glucose was lowered to 3.8 and to 2.8 mmol/L. Data shown are the average values for the final 20 minutes of each period. *** < 0.001 and * < 0.05 caffeine compared with placebo. Black bars depict caffeine; white bars, placebo. Middle cerebral artery velocity and P300 latency.PP

Symptoms were unchanged from baseline at the end of the euglycemic period in both studies (Table 1). In the placebo study, awareness of hypoglycemia (movement along the visual analog scale from blood sugar normal to blood sugar very low) was unchanged from baseline and no symptoms were reported when plasma glucose was lowered to 3.8 mmol/L. In contrast, at the same glucose level, participants who had ingested caffeine felt hypoglycemic (P < 0.01 versus placebo and baseline); the sensations of trembling (P < 0.001), sweating (P < 0.005), weakness (P < 0.05), facial flushing (P < 0.05), and palpitations (P < 0.002) had increased from baseline. At a plasma glucose of 2.8 mmol/L, awareness of hypoglycemia (P < 0.01), hunger, trembling, sweating, weakness, and poor concentration were more intense in the caffeine study (all P < 0.05 compared with placebo).

The latency of the P300 evoked potential, an index of cognitive function, was similar at the start of both studies (282 8 and 286 8 ms) and did not change after placebo even when glucose was lowered to 2.8 mmol/L (see Figure 3). However, P300 latency decreased slightly immediately after caffeine intake and later became significantly prolonged when plasma glucose was lowered to 2.8 mmol/L [308 versus 289 ms; 17 ms (CI, 2% to 13%), P < 0.05].

Control Visit

The four participants who consumed caffeine without exogenous glucose or insulin had a rapid and sustained increase in plasma caffeine levels to values that were similar to those achieved in the same participants during the clamp studies (9.2 1.3 compared with 7.7 1.5 g/mL). Caffeine ingestion was associated with a slight increase in plasma glucose (from 4.8 to 5.2 mmol/L; +0.4 mmol/L [CI, 0.1 to 0.7 nmol/L], P < 0.05), but no change in plasma insulin levels was found. Caffeine produced a rapid decrease in V_MCA (from 66 to 48 cm/s; +18 cm/s [10 to 24 cm/s]), which was sustained for the duration of the study. Plasma epinephrine levels increased slightly (from 171 to 389 pmol/L; +218 pmol/L [CI, 167 to 291 pmol/L], P < 0.05) within 30 to 45 minutes but no significant change occurred in the levels of other counter-regulatory hormones.

Discussion

In this study, acute ingestion of caffeine (in a dose equivalent to 2 to 3 cups of dripped-brewed coffee) caused a rapid and sustained reduction in middle cerebral artery velocity associated with a markedly stimulated hormonal counter-regulatory response when plasma glucose was lowered to 3.8 mmol/L. At this plasma glucose level, participants developed warning symptoms (trembling, sweating, and palpitations) and felt hypoglycemic after intake of caffeine but not after placebo. The disparate hormonal and symptomatic responses persisted when glucose levels were lowered further into the frank hypoglycemic range (2.8 mmol/L). After 1 hour at 2.8 mmol/L, cognitive function became impaired in the caffeine study but not after placebo. The absence of a clinically significant change in hormone levels, symptoms, and cognitive function during the euglycemic phase of the clamp studies and the control visit suggests that the effect of caffeine ingestion to alter awareness of and the physiologic responses to hypoglycemia requires that peripheral glucose levels decrease below a critical value.

As results of the cortical evoked potentials suggest, the enhanced hormonal and symptomatic responses to hypoglycemia induced by caffeine may be related to increased vulnerability of the central nervous system to neuroglycopenia. Because the brain cannot store glucose, normal central nervous system function requires that a steady supply of glucose from the blood be maintained. In the placebo study, the supply of glucose to the brain was directly related to the plasma glucose concentration because the V_MCA did not differ throughout the study. In contrast, as described by others [17, 18], acute ingestion of caffeine caused a substantial reduction in cerebral blood flow, an alteration that persisted during hypoglycemia. Thus, during the caffeine study, the supply of glucose to the brain was jeopardized by both the decrease in plasma glucose concentration and the reduction in cerebral blood flow. Although the sustained reduction in cerebral blood flow induced by caffeine is not severe enough to cause ischemia in healthy volunteers, it is unclear whether caffeine increases the risk for ischemia in high-risk individuals who have cerebrovascular disease [19]. It should be noted that the effects of caffeine on the brain are not restricted to alterations in blood flow. Caffeine probably acts as a central nervous system stimulant primarily through its action as a central adenosine receptor antagonist [7]. Competitive antagonism of adenosine receptors increases synaptic activity and stimulates release of neurotransmitters, effects that cause a net increase rather than a decrease in brain glucose requirements [20-23]. Preliminary data from humans who have depth electrodes and microdialysis probes implanted for the management of seizure disorders [24] indicate that acute ingestion of less than 200 mg of caffeine is associated with a twofold increase in lactate levels in hippocampal extracellular fluid (During MJ. Unpublished observation), an index of brain glucose use [25].

In our study glucagon release during hypoglycemia was unaffected by caffeine. Whereas release of growth hormone, cortisol, and catecholamines during insulin-induced hypoglycemia is thought to be mainly dependent on central rather than peripheral glucopenia [26], in vitro studies using the isolated, buffer-perfused pancreas suggest that hypoglycemia stimulates glucagon release at least in part via direct effects at the level of the cell [27]. In this study, peripheral (and therefore cell) glucose levels were similar in both studies and the glucagon responses to hypoglycemia were almost identical, suggesting that the glucagon release in response to a moderate reduction in blood glucose levels may be controlled at the level of the cell rather than centrally. The failure of glucagon to increase above baseline levels despite lowering plasma glucose to 2.8 mmol/L, may, in part, have been a consequence of sustained hyperinsulinemia [28].

In clinical practice, the Whipple triad is used to diagnose hypoglycemia. The results of this study suggest that previous ingestion of caffeine may cause persons to feel hypoglycemic at a glucose level not usually considered to be low. Thus, in defining the Whipple triad, the threshold for hypoglycemia considered as being clinically relevant, that is, associated with the development of warning symptoms, may have to be redefined. The ability of caffeine to induce symptoms of hypoglycemia at a plasma glucose level within the normal range is likely to be mediated, in part, by the early and exaggerated catecholamine response to a slight reduction in plasma glucose levels.

Although all but one of our participants did not usually ingest caffeine in their diet, the average daily consumption of caffeine in the United States is 200 to 300 mg, with more than 10% of the population ingesting more than 500 mg/d, an amount similar to the test dose of caffeine used in this study [29]. Furthermore, drug dosage is related to body size or weight. When a child drinks a can of cola, caffeine intake is comparable to an adult drinking four cups of instant coffee [30]. Our findings suggest that healthy persons who ingest moderate caffeine loads may be more likely to develop autonomic symptoms if plasma glucose levels fall into the low-normal range in the late postprandial period. Glucose levels within this range frequently occur in otherwise healthy persons who have eaten a large carbohydrate load [31]. Indeed, the failure to observe a close relationship between plasma glucose levels and symptoms in some participants with reactive hypoglycemia [3, 4] may be related in part to a failure to control for differences in caffeine intake as well as other factors that influence the timing and magnitude of catecholamine responses to decreasing glucose levels. For example, a high caffeine intake may pose particular problems for healthy children and patients with poorly controlled insulin-dependent diabetes mellitus who, because of their young age or their adaptation to chronic hyperglycemic release of epinephrine, have hypoglycemic warning symptoms at higher glucose levels than do healthy adults [32, 33].