Systematic review of randomised controlled trials of the effects of caffeine or caffeinated drinks on blood glucose concentrations and insulin sensitivity in people with diabetes mellitus



N. Whitehead, Department of Nutrition and Dietetics, Leeds Metropolitan University, Leeds LS1 3HE, UK.

Tel.: +44 (0)113 8129233




Compounds other than macronutrients have been shown to influence blood glucose concentrations and insulin sensitivity in people with diabetes, with caffeine being one such substance. The present study systematically reviewed the evidence of the effects of caffeine on blood glucose concentrations and/or insulin sensitivity in people with diabetes.


Four databases, including MEDLINE and EMBASE, were searched up to 1 February 2012. Randomised controlled trials (RCTs) investigating the effects of caffeine on blood glucose and/or insulin sensitivity in humans, diagnosed with type I, type II or gestational diabetes mellitus (GDM), were included. Quality assessment and data extraction were conducted and agreed by both authors.


Of 253 articles retrieved, nine trials (134 participants) were identified. Trials in people with type II diabetes demonstrated that the ingestion of caffeine (approximately 200–500 mg) significantly increased blood glucose concentrations by 16–28% of the area under the curve (AUC) and insulin concentrations by 19–48% of the AUC when taken prior to a glucose load, at the same time as decreasing insulin sensitivity by 14–37%. In type I diabetes, trials indicated enhanced recognition and a reduced duration of hypoglycaemic episodes following ingestion of 400–500 mg caffeine, without altering glycated haemoglobin. In GDM, a single trial demonstrated that approximately 200 mg of caffeine induced a decrease in insulin sensitivity by 18% and a subsequent increase in blood glucose concentrations by 19% of the AUC.


Evidence indicates a negative effect of caffeine intake on blood glucose control in individuals with type II diabetes, as replicated in a single trial in GDM. Larger-scale RCTs of longer duration are needed to determine the effects of timing and dose. Early indications of a reduced duration and an improved awareness of hypoglycaemia in type I diabetes require further confirmation.


Worldwide, the prevalence of diabetes mellitus continues to rise. A systematic analysis of global trends has estimated that the number of people diagnosed with diabetes increased from 153 million in 1980 to 347 million in 2008, equating to age-standardised prevalence values of 9.8% for men and 9.2% for women, and resulting in an increasing health, social and economic impact (Danaei et al., 2011).

The escalating health and economic costs related to diabetes are directly linked to poorer glycaemic control and the microvascular and macrovascular complications that ensue (Saydah et al., 2004). Nutritional research to date has concentrated heavily on the glycaemic impact of the type and amount of dietary carbohydrate but, more recently, attention has focused on the impact that other dietary components may have on blood glucose concentrations and insulin sensitivity in diabetes. Systematic reviews have examined the influence of various herbs and dietary supplements (Yeh et al., 2003), including bitter gourd (Ooi et al., 2010) and cinnamon (Akilen et al., 2010), with inconclusive results.

Caffeine (1,3,7-trimethylxanthine) has also attracted increasing attention. It is widely known for its stimulating properties (Smit & Rogers, 2002), appears frequently in the diet of many developed countries (Barone & Roberts, 1996; NDNS, 2008/2009) and is a compound recognised as one of the most commonly consumed, biologically active, dietary ingredients throughout the world (Heckman et al., 2010). Importantly, it has also been shown to disrupt normal glucose metabolism in both healthy adults and those with diabetes through mechanisms that enhance insulin resistance.

Although short-term studies in healthy adults continue to show that caffeine ingestion results in acute transient insulin resistance and impaired glucose tolerance (Louie et al., 2008; Moisey et al., 2008), epidemiological studies would appear to contradict these conclusions. Heavy coffee consumption (four to six cups of coffee daily) in healthy adults is associated with a reduced risk of type II diabetes (Huxley et al., 2009). A number of explanations have been proposed, including a beneficial counter effect of phytochemical compounds present within coffee that is protective against diabetes (Tunnicliffe & Shearer, 2008). However, with no causal link established, these studies remain purely correlational in nature.

In individuals with existing diabetes, it is proposed that caffeine may impact on blood glucose concentrations through several mechanisms. Caffeine may act by hindering the transportation of glucose from the blood into the muscles through its role as an adenosine receptor antagonist (Fisone et al., 2004), subsequently inhibiting glucose uptake into muscle cells, even in the presence of insulin (Vergauwen et al., 1994; Crist et al., 1998). In addition, it is suggested that raised blood glucose concentrations post-consumption of caffeine may be a result of elevated epinephrine (adrenaline), which may induce insulin resistance through impairment of glucose uptake in peripheral tissue and the stimulation of hepatic glucose production (Keijzers et al., 2002; Thong & Graham, 2002).

Despite uncertainty regarding the theoretical basis of the mechanisms of action, further research suggests that individuals should not exceed caffeine intakes of more than 200–400 mg day−1 (equivalent to two to four cups of instant or brewed coffee daily, depending on the amount of coffee used); the latter for nonpregnant individuals [Nawrot et al., 2003; Food Standards Agency (FSA), 2011]. Reasons include the potential for hypertension (Nurminen et al., 1999; Riksen et al., 2009) and an increased risk of spontaneous abortion or impaired foetal growth in pregnant women (Higdon & Frei, 2006). There is, however, no current guidance or conclusive evidence to suggest if this recommendation should be altered for people with diabetes and no systematic reviews within this area (Cochrane Library,; accessed on 1 February 2012; DARE Database,; accessed on 1 February 2012).

The present systematic review therefore aims to evaluate randomised controlled trials reporting the impact of caffeine ingestion on glycaemic outcome and/or insulin sensitivity in adults with type I, type II and gestational diabetes mellitus (GDM).

Materials and methods

Eligibility criteria

Trials were included in the systematic review if they met certain criteria. (i) The trial involves human participants who had been previously diagnosed with type I, type II or GDM. (ii) The trial involves the ingestion of caffeine (anhydrous powder or anhydrous powder in capsule form) or a caffeinated drink versus a placebo control with blood glucose and/or insulin concentrations or sensitivity measured. (iii) There is evidence of random allocation and evidence of having a control or placebo group for comparison. (iv) The trial is published in the English language.

Information sources

Four online electronic databases; MEDLINE 1950, EMBASE 1980, CINAHL 1981 and Cochrane Central Register of Controlled Trials 2012, were searched for randomised controlled trials investigating the effects of caffeine ingestion on blood glucose concentrations and/or insulin sensitivity in people with diabetes up to 1 February 2012. The following search terms were used to search for relevant publications: ‘DIABETES OR DIABETIC’ and ‘GLUCOSE OR HYPERGLYC* OR HYPOGLYC* OR INSULIN OR HYPERINSULIN* OR HYPOINSULIN*’ and ‘CAFFEINE OR COFFEE’. The search strategy is provided in the Supporting information (Data S1).

To explore the ‘grey literature’ and minimise publication bias, the following online resources were searched on 1 February 2012; Bandolier (, CAB Abstracts (, (, DARE Database (, Diabetes UK (, Google Scholar (, ICTRP: International Clinical Trials Registry Platform Search Portal WHO (, NHS Evidence ( and Open SIGLE: System for Information on Grey Literature in Europe ( The search terms ‘CAFFEINE’ or ‘COFFEE’ and ‘DIABETES’ were used. Publication titles and abstracts obtained were screened for relevant articles with full texts checked for inclusion criteria. Secondary references were also checked.

Study selection

Studies were initially selected from the electronic databases according to the search criteria. Potentially relevant publication titles and abstracts were then identified and screened for retrieval from the Cochrane Library (; accessed on 1 February 2012). Publications were discarded if they were deemed irrelevant to the review's initial objectives, duplicate publications, reported an inappropriate population type, did not report defined outcomes, used an alternative study design or were not published in the English language (Fig. 1). The identified literature was assessed independently by each of the two investigators using the Critical Appraisal Skills Programme (CASP, 2006) and joint agreement on inclusion was reached.

Figure 1.

Flow diagram of the screening process. n, number.

Data collection and extraction

Data were extracted independently by the two investigators using defined measures, including sample size, caffeine dose, method of caffeine ingestion, glucose load administered, blood glucose concentrations and/or insulin concentrations for the duration of monitoring for participants versus control/placebo groups (Table 1). For standardisation and more direct comparison between trials, individual series data were converted to the incremental area under the curve (iAUC) where available raw data made this possible (Jankelson et al., 1967) and insulin measured as pm was converted to μU per mL (Robinson et al., 2004, 2009). The CASP (2006) tool was used to assess risk of bias at trial and outcome level.

Table 1. Study characteristics and outcomes for the nine trials meeting full inclusion criteria
StudyType of trialSubjects (n) Type and Duration of diabetesWashout period between treatmentsHabitual caffeine intake (daily) pretrialSubject characteristics [sex (male/female), BMI (kg m–2)]Total caffeine dose and method of administrationGlucose load givenEnd point of monitoringMethod of measurementGlucose values,mean (SD)Insulin values, mean (SD)Outcome of caffeine on glucose and insulin parameters
  1. AUC, area under curve; BMI, body mass index; HbA1c, glycated haemoglobin; iAUC, incremental area under curve; IV, intravenous; n, number; v, versus.

Jankelson et al. (1967)



controlled cross-over trial

n = 9

Type II

Duration not reported

Usually 1 week or moreNot reported


BMI not reported

~224 mg

drink of coffee/water placebo

0.5 g kg−1 body weight IV1-h post-glucose loadMean incremental AUC (derived from raw data)

6.7 (0.33) versus 5.69 (0.31) mm (caffeine versus placebo)

P = 0.001

39.4 (1.5) versus 43.4 (9.5) μU mL−1 (caffeine versus placebo)

P = 0.009

Glucose levels ↑ (iAUC 20% higher in caffeine versus placebo)

Insulin levels ↓ (iAUC 9% lower in caffeine versus placebo)

Lane et al. (2004)Double-blind placebo controlled cross-over trial

n = 14

Type II

Duration ≥6 months

Within 2 weeks


526 mg (144 mg)

(7-day diary)

Male (11) Female (3)

BMI not reported

375 mg

of caffeine/placebo capsules

75 g of carbohydrate

liquid meal

2-h post-glucose loadMean incremental AUC

3.87 (0.3) versus 3.2 (0.36) mm

(caffeine versus placebo)

P = 0. 04

66.73 (10.49) versus 45.17 (5.98) μU mL−1

(caffeine versus placebo)

P = 0.01

Glucose levels ↑ (iAUC 21% higher in caffeine versus placebo)

Insulin levels ↑ (iAUC 48% higher in caffeine versus placebo)

Robinson et al. (2004)


randomised placebo controlled cross-over trial

n = 12

Type II

Duration 3.1 (0.9) years

1 week

Low to moderate

0–500 mg



BMI 32 (1)

5 mg kg−1 body weight = ~510 mg

of caffeine/placebo capsules

75 g of dextrose beverage3-h post-glucose loadTotal AUC

864 (57) versus 729 (67) mm

(caffeine versus placebo)

P = 0.003

6007 (1340) versus 4495 (757) μU mL−1

(caffeine versus placebo)

P = 0.04

Glucose levels ↑

(AUC 16% higher in caffeine versus placebo)

Insulin levels ↑ (AUC 25% higher in caffeine versus placebo)

Insulin sensitivity ↓ (by 14%, P = 0.02)

Lee et al. (2005)


randomised placebo controlled cross-over trial

n = 8

Type II

Duration <5 years with one subject diagnosed 10 years earlier

~1 week

Low to moderate

(0–5 cups of coffee/tea)



BMI 29.9 (3.2)

5 mg kg−1 body weight

(~465 mg)

caffeine/placebo capsules in 250 mL of water

IV glucose to maintain levels of

~5 mm

3-h euglycaemic clampContinuous blood glucose monitoring

37% reduction in glucose uptake at baseline

(caffeine versus placebo)

P < 0.05

Before exercise

50 (11) versus 51 (17) μU mL−1

(caffeine versus placebo)

After exercise:

56 (12) v

58 (13) μU mL−1

(caffeine versus placebo)

Glucose uptake ↓

Insulin sensitivity ↓ before and after chronic exercise

Lane et al. (2007)Double-blind placebo controlled cross-over trial

n = 20

Type II

Duration 7.5 (6.3) years

Within 2 weeks

486 mg

(250 mg)

(7-day diary)

Male (9)

Female (11)

BMI 32.0 (7.2)

250 mg

of caffeine/placebo in decaffeinated coffee

75 g of carbohydrate

liquid meal

2-h post-glucose loadMean incremental AUC

2.8 (0.3) versus 2.2 (0.3) mm

(caffeine versus placebo)

P = 0.02

50.9 (6.0) versus 42.9 (6.0) μU mL−1

(caffeine versus placebo)

P = 0.02

Glucose levels ↑ (iAUC 28% higher in caffeine versus placebo)

Insulin levels ↑ (iAUC 19% higher in caffeine versus placebo)

Lane et al. (2008)Double-blind placebo controlled cross-over trial

n = 10

Type II

Duration ≥6 months

Not reported

520 mg

(419 mg)

Male (5) Female (5)

BMI 31.9 (5.6)

500 mg

of caffeine/placebo capsules

90 g of carbohydrate

liquid meal then variable

(as meals)

Continuous blood glucose monitoring (Interstitial 72 h)Mean daytime blood glucose values

8.0 versus 7.4 mm

(caffeine versus placebo)

P < 0.0001

Glucose levels ↑ (8.1% higher mean glucose values in caffeine versus placebo)
Watson et al. (2000)

Double-blind randomised placebo

cross-over controlled trial

n = 34

Type I

Duration 15 (1.7) years

No washout period

395 mg

(37 mg)

(7-day diary)

Male (22) Female (12)

BMI 25.4 (0.5)

400 mg

(200 mg × 2)

of caffeine/placebo capsules


(as meals)

3 months durationFrequency of hypoglycaemia and HbA1c

1.3 versus 0.9 episodes/week

(caffeine versus placebo)

P < 0.03

44% ↑ in hypoglycaemic episodes associated with more intense warning symptoms with caffeine inclusion

HbA1c levels were unaffected

Richardson et al. (2005)

Double-blind randomised

placebo controlled cross-over trial

n = 19

Type I

Duration 19.2 (10.4) years

12 daysNot reported

Male (9) Female (10)

BMI not reported

500 mg

(250 mg × 2)

of caffeine/placebo capsules

Variable (as meals)48 h

Interstitial glucose levels

Total time hypoglycaemic/24 h

Frequency of mild, moderate and prolonged nocturnal hypoglycaemia

90 versus 195 min

(caffeine versus placebo)

P = 0.044

0.13 versus 0.61

moderate overnight hypoglycaemic episodes

(caffeine versus placebo)

P = 0.011

Duration of nocturnal hypoglycaemia overall 49 versus 132 min

(caffeine versus placebo)

P = 0.035

Total time hypoglycaemic ↓ with caffeine

Duration of moderate nocturnal hypoglycemic episodes ↓ with caffeine

Overall duration of nocturnal hypoglycaemia decreased

Robinson et al. (2009)


randomised placebo controlled cross-over trial

n = 8

Gestational diabetes

1 weekNot assessed


BMI 27.1 (1.2)

3 mg kg−1 body weight

(~200 mg)

caffeine/placebo capsules

75 g of dextrose2 hTotal AUC

616 (42) versus 518 (35) mm

(caffeine versus placebo)

P = 0.001

9367 (1747) versus 7479 (1274) μU mL−1

(caffeine versus placebo)

P = 0.07

Glucose levels ↑

(AUC 19% higher in caffeine versus placebo)

Insulin levels ↑ (AUC 29% higher in caffeine versus placebo)

Insulin sensitivity ↓ (by 18%, P = 0.01)

Summary outcome measures

Glycaemic outcome measures are presented as mean (SD) blood glucose and/or insulin concentrations, with and without caffeine or coffee ingestion, to allow direct comparison of the data (Table 1). Differences in the total area under the curve (AUC) and iAUC were then expressed as the percentage differences in glucose and insulin values for caffeine or caffeinated drink compared to placebo within studies (Table 1).

In trials examining outcomes as hypoglycaemic events, total minutes, intensity and number of events of hypoglycaemia were recorded to examine the effects of caffeine effects on blood glucose concentrations in relation to this parameter. Outcome measures regarding the effects on hypoglycaemic warning systems have been commented on; however, these were not directly relevant to the initial objectives of the review. For all values, P ≤ 0.05 was considered statistically significant.

Synthesis of results

The results were synthesised by constructing a descriptive summary of the included trials in table form (Table 1), with narrative synthesis (CRD, 2009). A meta-analysis was not performed given the differences in process and outcome measures in the included trials. Risk of bias was included within the narrative synthesis and was based on: randomisation, blinding, incomplete outcome data, selective reporting and carry-over effect (Higgins & Green, 2011). The PRISMA (2009) checklist was used to structure the review.


A total of 253 articles were found on an initial search of the electronic databases and the Cochrane Library (; accessed on 1 February 2012), with a total of 12 full papers retrieved for more detailed evaluation (Fig. 1).

Nine trials were included in the final review: six trials investigated people with type II diabetes; two with type I diabetes; and one trial reported women with GDM. A common feature of the trial interventions was their cross-over design, with the administration of both caffeine or a caffeinated drink and a placebo before a glucose or carbohydrate load to enable monitoring of post-prandial blood glucose concentrations (Table 1).

Type II diabetes

Six trials in people with type II diabetes met the full inclusion criteria (Jankelson et al., 1967; Lane et al., 2004, 2007, 2008; Robinson et al., 2004; Lee et al., 2005). Each trial primarily investigated the effects of caffeine on blood glucose concentrations, with all but one trial (Lane et al., 2008) also monitoring insulin concentrations and/or sensitivity, after a glucose load. All were of relatively small sample size with a mean of twelve participants per trial.

All trials consistently demonstrated a negative impact on blood glucose control in type II diabetes, in subjects who were predominantly classified as overweight or obese (Table 1). Although the methods by which caffeine and glucose loads were administered differed between trials, it was uniformly shown that caffeine or a caffeinated drink had no impact on glucose and/or insulin concentrations before the ingestion of a glucose/carbohydrate load, although, after glucose ingestion, caffeine caused an elevation of blood glucose concentrations compared to placebo and did so at all time points measured, from 1 h (Jankelson et al., 1967) to 3 h (Robinson et al., 2004; Lee et al., 2005). Lane et al. (2008) reported that caffeine raised mean glucose during ‘daytime’ hours on the single day in which it was administered (compared to a single day of placebo administration). In addition, 3-h post-prandial glucose responses to all three meals on that day were elevated.

Jankelson et al. (1967) were the first to analyse the effects of caffeine on glycaemic control in individuals with ‘adult onset diabetes’ after the injection of intravenous glucose, following two cups of instant caffeinated coffee, compared to the oral ingestion of water. Despite a significant increase in blood glucose concentrations following caffeine ingestion at 1 h [6.7 (0.33) versus 5.69 (0.31) mm; P = 0.001], no significant differences were found for concentrations of insulin in this trial. Lane et al. (2004) supported these findings by concluding that the acute administration of caffeine impaired post-prandial glucose metabolism at 2 h [3.87 (0.3) versus 3.2 (0.36) mm; P = 0.04]. There was a significant increase as measured by mean iAUC in blood glucose (21% increase in the iAUC; P = 0.04) and in insulin (48% increase in the iAUC; P = 0.01) for caffeine compared to placebo (Table 1). Robinson et al. (2004) subsequently confirmed increased glucose concentrations for up to 3 h after caffeine ingestion and a glucose load, compared to placebo, indicating a prolonged effect of caffeine on blood glucose values [864 (57) versus 729 (67) mm; P = 0.003] and insulin concentrations [6007 (1340) versus 4495 (757) μU mL−1; P = 0.04], which equated to a 16% and 25% increase in the AUC, respectively. The insulin sensitivity index was subsequently 14% lower (P = 0.02) in the caffeine trial compared to placebo [4.3 (0.7) versus 4.9 (0.7)].

Lee et al. (2005) confirmed these findings by using the euglycaemic clamp technique to measure insulin sensitivity in obese type II diabetics before and after exercise training and comparing them with obese and lean individuals without diabetes. Caffeine ingestion resulted in a significant reduction in glucose uptake over the final 30 min of the 3-h clamp procedure compared to placebo both before (37%; P < 0.05) and after (36%; P < 0.05) 3 months of aerobic exercise training. Insulin sensitivity was therefore significantly lower after caffeine ingestion compared to placebo, with exercise training failing to alleviate the negative effects of caffeine on insulin-mediated glucose uptake.

The two most recent published trials to meet the review criteria were by Lane et al. (2007, 2008). Lane et al. (2007) again showed caffeine (administered in decaffeinated coffee) to exaggerate post-prandial hyperglycaemia (by 28% of the iAUC; P = 0.02) and hyperinsulinaemia (by 19% of the iAUC; P = 0.02) compared to placebo at 2 h. Lane et al. (2008) demonstrated that caffeine impaired glucose metabolism, producing higher mean daytime glucose concentrations (8.0 versus 7.4 mm; P < 0.0001) and exaggerated post-prandial glucose responses at 3 h (by 0.7 mm after breakfast; 1.0 mm after lunch; 1.8 mm after dinner) by measuring 3-day interstitial glucose concentrations with a continuous glucose monitor, although no difference in the effects of caffeine was shown across the three meals.

The characteristics of the trial participants varied; half were conducted in men alone (Jankelson et al., 1967; Robinson et al., 2004; Lee et al., 2005), and half in both men and women (Lane et al., 2004, 2007, 2008). Participants did not require insulin but used a variety of oral hypoglycaemic agents and/or diet and exercise both within and between trials. Similarly, the participants' duration of diabetes varied between trials, with diagnosis occurring ≥6 months (Lane et al., 2004, 2008), 3.1 (0.9) years (Robinson et al., 2004), 0–10 years (Lee et al., 2005) and 7.5 (6.3) years (Lane et al., 2007) prior to each trial. Reported glycated haemoglobin (HbA1c) levels indicated a mixed level of diabetes control: 8.4% (0.9%) (Robinson et al., 2004), 7.2% (1.4%) (Lane et al., 2007) and 6.4% (0.5%) (Lane et al., 2008). Where HbA1c levels were not reported, mean fasting glucose levels were given: 7.5 (1.6) mm (Lane et al., 2004) and 7.9 (2.2) mm (Lee et al., 2005), with the exception of Jankelson et al. (1967) who failed to report either. Reported habitual caffeine intakes ranged from low to moderate, both within and between trials, with the exception of Jankelson et al. (1967) who failed to report on this. The remaining five trials reported estimated habitual daily caffeine (reported in mg) or cups of coffee/tea intakes to be: 526 (144) mg (Lane et al., 2004); 0 to 300–500 mg (Robinson et al., 2004); 0–5 cups (Lee et al., 2005); 486 (250) mg (Lane et al., 2007); and 520 (419) mg (Lane et al., 2008).

When examining the impact of participant characteristics, Lane et al. (2007) found that a standard dose of 250 mg of caffeine (administered in decaffeinated coffee) had similar effects in all participants, irrespective of the level of glycaemic control. The duration of diabetes was the only characteristic that related to the effect of caffeine. Lane et al. (2007) comment that it is logical to hypothesise that participants with a longer history of diabetes may have less available insulin reserves, and therefore have a reduced capability of overcoming the insulin resistance caused by caffeine, which subsequently leads to exaggerated post-prandial hyperglycaemia. Trials showed that habitual caffeine consumption did not lead to the development of a tolerance of its effects on blood glucose concentrations (Lane et al., 2004, 2007, 2008), with Lane et al. (2007) additionally finding heavy caffeinated coffee drinkers to respond similarly to those who drank less caffeinated coffee on a daily basis.

The methods of caffeine administration varied between the trials. Jankelson et al. (1967) used drinks of caffeinated coffee as the source of caffeine (approximately 224 mg), whereas the majority of trials used caffeine capsules to administer set doses: 375 mg; 250 mg plus an additional 125 mg to maintain drug levels (Lane et al., 2004), 5 mg kg−1 body weight; approximately 510 mg (Robinson et al., 2004), 5 mg kg−1 body weight; approximately 465 mg (Lee et al., 2005); and 500 mg (Lane et al., 2008). Lane et al. (2007) did not use caffeine capsules and, instead, administered 250 mg of caffeine in 475 mL of decaffeinated coffee and gained a similar hyperglycaemic and hyperinsulinaemic response (Table 1).

The methods and amounts of glucose given to participants also varied between trials, preventing direct comparison of precise blood glucose values. The intravenous injection of glucose solution (50% in water) at a dose of 0.5 g kg−1 of body weight (Jankelson et al., 1967) had an instant effect on blood glucose concentrations, and a maximal response at the comparatively earlier time of 10 min as opposed to 1 h (Lane et al., 2004, 2007), 1.5 h (Robinson et al., 2004) and 2–3 h (Lane et al., 2008) following the oral ingestion of glucose. Oral administration of glucose also varied; 75 g of glucose/carbohydrate via a beverage/liquid meal (Lane et al., 2004, 2007; Robinson et al., 2004) and 90 g of carbohydrate via a liquid meal for breakfast, with lunch and dinner ad libitum (Lane et al., 2008), as a result of the increased length of the trial in comparison to others.

Type I diabetes

Two trials in people with type I diabetes met the full inclusion criteria (Watson et al., 2000; Richardson et al., 2005), with both investigating the effect of caffeine on the frequency and/or awareness of hypoglycaemia. Hypoglycaemia warning signs include feeling hungry, trembling, sweating, feeling anxious, palpitations, tingling of the lips and blurred vision (Diabetes UK, 2011). Sample sizes were small (Table 1), although both studies employed a cross-over design, with fewer participants required to obtain the same statistical power (CRD, 2009). Watson et al. (2000) additionally reported a robust sample size calculation, ensuring the trial was of adequate sample size to detect a true difference.

Hypoglycaemic episodes were reported according to differing criteria. Watson et al. (2000) recorded the prevalence and intensity of hypoglycaemic (capillary blood glucose concentration of <3.5 mm) episodes daily over each of the 3-month periods that participants took caffeine and placebo capsules. They monitored through the use of four times daily premeal blood glucose readings, supplemental readings where hypoglycaemia was suspected and the use of a nine item validated questionnaire to assess intensity and number of symptoms. Episodes were then divided into into the number of symptomatic and biochemical hypoglycaemic events. Richardson et al. (2005) similarly defined hypoglycaemia as a blood concentration of <63 mg dL−1 (directly equivalent to <3.5 mm) but measured interstitial blood glucose concentration through continuous blood glucose monitoring and did so only during the last 2 days of each 2-week intake of caffeine or placebo capsules. The intensity of hypoglycaemia was defined as mild (>53 and <63 mg dl−1), moderate (<53 mg dl−1 for four consecutive readings) and severe (when external help was required). The duration of hypoglycaemia was assessed as complete when four consecutive interstitial glucose readings were within normal range following a recorded concentration of <63 mg dl−1.

Watson et al. (2000) demonstrated that the ingestion of modest amounts of caffeine enhanced the intensity of hypoglycaemia warning systems in people with type I diabetes (P < 0.05) without altering the prevailing standard of glycaemic control, measured by HbA1c. The total numbers of hypoglycaemic episodes reported were more frequent with caffeine (1.3 versus 0.9 episodes per week; P < 0.03); however, fewer severe hypoglycaemic episodes were experienced. The 44% increase in the number of mild hypoglycaemic episodes reported when caffeine was included in the diet, was notably associated with the more intense warnings.

By contrast, although over a shorter time period, Richardson et al. (2005) demonstrated that overall interstitial hypoglycaemia was less frequent in those taking caffeine (P = 0.044). Caffeine also significantly reduced the duration of nocturnal hypoglycaemia, compared to placebo, with a mean duration of 49 versus 132 min, respectively (P = 0.035) and reduced the number of moderate glycaemic events at night (0.13 versus 0.61; P = 0.011).

Both studies recruited participants of similar age (between 30 and 60 years of age) and incorporated both sexes. The mean (SD) duration of diabetes was reported to be 15 (1.7) years (Watson et al., 2000) and 19.2 (10.4) years (Richardson et al., 2005), with the earlier study reporting more optimum HbA1c levels of 7.9% (0.2%) and an absence of complications other than background retinopathy compared to the poorer HbA1c levels of 8.4% (1.1%) (Richardson et al., 2005); indicating two study populations who were at slightly differing stages of disease and complication risk.

Caffeine was similarly administered as capsules in doses of 400 and 500 mg (taken as 200 or 250 mg twice a day) (Watson et al., 2000; Richardson et al., 2005 respectively), although only the former study reported habitual intake of caffeine, defining their population as moderate caffeine drinkers [395 (37) mg day−1]. Each study ensured that participants were established on low caffeine dietary intakes prior to the start of caffeine or placebo capsules; equating to <15 mg caffeine daily for 2 months (Watson et al., 2000) and <50 mg caffeine daily for 2 weeks (Richardson et al., 2005). Neither trial controlled for the glucose loads (i.e. meals) consumed by participants, which were neither monitored nor reported.

Gestational diabetes

One trial, involving eight women diagnosed with GDM, was selected as meeting the full review criteria, with results showing that approximately 200 mg of caffeine impaired insulin sensitivity following a standard 2-h 75 g oral glucose tolerance test (Robinson et al., 2009). The total AUC analysis showed 19% higher [616 (42) versus 518 (35) mm; P = 0.001] glucose responses after 3 mg kg−1 of caffeine than after placebo, with a significant decrease in insulin sensitivity (by 18%; P = 0.01).

Fasting blood glucose concentrations were 4.9 (0.2) mm with women not taking any diabetes medication, denoting a group who had relatively good levels of glycaemic control. Levels of HbA1c were not reported. Participants were instructed to refrain from caffeine containing products for 48 h before each experiment, with habitual daily caffeine intake not assessed. Caffeine was ingested by participants in doses of 3 mg kg−1 of prepregnancy body weight (mean dose of approximately 200 mg) via a capsule in 250 mL of water. Robinson et al. (2009) importantly justify using a lower dose of 3 mg kg−1 (reported to be equivalent to one to two cups coffee) than that previously used in trials conducted with men (i.e. 5 mg kg−1) (Robinson et al., 2004), aiming to account for the alterations in caffeine pharmacokinetics in pregnancy (i.e. caffeine's half-life is increased) (Knutti et al., 1982).

Despite a small sample size for those recruited with GDM (n = 8), the study benefited from a pregnant control population (n = 19), a robust cross-over design for both groups of participants, and demonstrated highly significant glucose responses associated with caffeine ingestion in women with GDM.

Risk of bias

In all studies, outcome measures were predetermined prior to trials commencing, with the reporting of both significant and nonsignificant findings (reducing reporting bias). The blinding of both participants and assessors (double-blinding) took place in all studies except one (Jankelson et al., 1967), which ensured that both intervention and control (placebo) groups received similar treatment throughout (reducing performance and detection bias).

In type I diabetes and GDM, studies specifically reported randomisation of participants to either the treatment or control groups first. Despite only two out of the six studies in type II diabetes explicitly reporting that participants were randomised to the intervention group first (Robinson et al., 2004; Lee et al., 2005), selection bias was minimised as a result of the cross-over nature of the trial designs.


Caffeine is one of the most commonly consumed, biologically active dietary ingredients throughout the world. It is a stimulant found naturally in coffee, tea, chocolate, some soft drinks, energy drinks and certain medicines, and is estimated as being present in the diets of more than 85% of British and North American adults (Henderson et al., 2002; Frary et al., 2005). Under experimental conditions, moderate intakes of caffeine have been shown to disrupt glycaemic control in healthy individuals, therefore supporting the hypothesis that caffeine impairs glucose metabolism in patients with existing diabetes. The aim of this review was to systematically evaluate the evidence for the effects of caffeine ingestion on glycaemic outcome and/or insulin sensitivity in adults with type I, type II and GDM.

Glycaemic impact of caffeine

Consistently elevated post-prandial blood glucose concentrations after the acute ingestion of moderate to high doses of a single daily caffeine supplement (Lane et al., 2004, 2007, 2008; Robinson et al., 2004; Lee et al., 2005) and caffeinated coffee (Jankelson et al., 1967) were observed in type II diabetes and with caffeine supplementation in the single study in GDM (Robinson et al., 2009), indicating a similar impaired glucose response to caffeine and caffeinated drinks that has been reported in healthy nondiabetic controls. Repeated exaggerations of post-prandial glucose, resulting from a single daily dose of caffeine, showed higher mean glucose concentrations (Lane et al., 2008), with the potential to increase HbA1c and, ultimately, the risk of diabetes-related complications.

The effects of caffeine remained present for up to 3 h after caffeine consumption and following a glucose load (Robinson et al., 2004), with regular aerobic exercise failing to alleviate the negative effects (Lee et al., 2005). Caffeine itself had no impact on blood glucose concentrations independently before the administration of a glucose load, further supporting the evidence that caffeine consumption alongside a glucose/carbohydrate load promotes heightened and prolonged hyperglycaemia, independent of factors such as exercise, which might be expected to alleviate the effect.

Although this systematic review clearly demonstrates an enhanced glycaemic effect of caffeine in type II diabetes under experimental short-term conditions, it is important to note that, in daily living, caffeine is ingested not as alkaloid caffeine capsules but predominantly as coffee in adults and as carbonated drinks in children and adolescents (Frary et al., 2005). With the exception of Jankelson et al. (1967), who administered caffeine as a coffee beverage, alkaloid caffeine capsules were used as the supplement of choice and generally in a high single dose. In four out of six studies in type II diabetes, caffeine doses of >350 mg were administered and, for the two studies in type I diabetes, the dose was >400 mg (Table 1), corresponding to the maximum advised daily dose (Health Canada, 2010). Battram et al. (2006) have shown that, in healthy individuals, caffeine exerts a glycaemic effect of up to 40% compared to the same dose of caffeine naturally present within a coffee beverage, suggesting that, in its more common dietary presentation, caffeine induces a more subdued effect. Interestingly, Jankelson et al. (1967) were not only the only study to use coffee as the means of caffeine supplementation, but also showed an enhanced post-prandial glucose effect similar in magnitude to studies using alkaloid caffeine, suggesting that a similar effect can be replicated with moderate doses of coffee consumption.

The dichotomy between the results of these experimental studies and the consistent findings of epidemiological studies showing an inverse relationship between heavy coffee consumption and risk of type II diabetes in healthy nondiabetic adults (Huxley et al., 2009) is conflicting. A number of explanations have been proposed. Coffee has been shown to be a poor marker of caffeine intake, resulting in misclassification (Brown et al., 2001), contains other compounds such as chlorogenic acid, trigonelline and antioxidants, which may counteract the glycaemic effect of caffeine, through their favourable impact on oxidative stress, gluconeogenesis, gut hormones or intestinal microflora (Tunnicliffe & Shearer, 2008), and more recent evidence suggests a potentially beneficial metabolic impact of coffee on adipocytes and liver function (Wedick et al., 2011). Despite various theories, the association between high coffee consumption and lower diabetes risk remains purely correlational. High coffee consumption may simply be a marker for other risk factors (Malik et al., 2010) or, for a variety of reasons, certain sectors of the population may be more prone to the effect of caffeine/caffeinated beverages.

Habitual coffee consumption has been proposed as inducing tolerance to caffeine (Robertson et al., 1981; Denaro et al., 1991) and the disappearance of acute daily side effects. In three studies (Lane et al., 2004, 2007, 2008), the estimated mean habitual coffee consumption ranged from 486 to 526 mg caffeine daily (Table 1), which is in excess of current daily recommendations of no more than 400 mg (Health Canada, 2010). Despite this high intake in older individuals aged 54–63 years, the ingestion of caffeine increased both the glucose response and insulin resistance in response to a carbohydrate challenge. Although estimates of habitual intake applied only to the immediate prestudy periods and cannot be relied on as an estimate of their long-term coffee consumption, it introduces the possibility that habitual coffee drinkers might be predisposed towards the more detrimental effects of caffeine over time.

This review highlights two specific groups who could be considered to be more prone to the negative effects of caffeine on blood glucose. First, Lane et al. (2007) demonstrated that the magnitude of the caffeine effect was correlated with the number of years since the diagnosis of diabetes. Although this trial requires replication to establish and quantify the effect, it suggests that those with type II diabetes may require greater dietary scrutiny of caffeine intake over time. Second, in women with GDM, it was shown that, even with lower ingested intakes of caffeine of approximately 200 mg, which corresponds to the safe recommended dose in pregnant women (FSA, 2011), an increased post-prandial glucose effect after caffeine ingestion and a carbohydrate challenge was apparent (Robinson et al., 2009). The direct correlation between glycaemic control and adverse foetal effects shown in gestational diabetes (Metzger et al., 2008) signifies a group in whom the preliminary results from this relatively high quality, yet small sample sized trial warrant further investigation.

Overall, the evidence for caffeine-induced impairment in the management of type II diabetes could warrant consideration of a change in current dietary practice for caffeine ingestion in this population group. Such guidelines may encourage those with poorly controlled type II diabetes to reduce their daily consumption of caffeine by choosing noncaffeinated beverages as their drink of choice. It can be postulated that, if habitual caffeinated coffee drinkers who have type II diabetes abstain from caffeinated beverages (i.e. lower their overall caffeine intake), then clinically beneficial reductions in post-prandial and overall blood glucose concentrations may be possible. Recently, a single-arm, pre–post design pilot study was conducted to test the effects of caffeine abstinence on blood glucose control over 3 months in participants with type II diabetes who drank caffeinated coffee daily (Lane et al., 2012). The primary outcome of this small trial (n = 7) was a reduction in HbA1c by 0.56% (P = 0.04), with the suggestion that habitual caffeine consumption increases chronic glucose concentrations, and caffeine abstinence may lead to beneficial improvements in chronic glycaemic control.

The significantly elevated post-prandial insulin concentrations following caffeine supplementation, compared to placebo, was observed in three studies in type II diabetes (Lane et al., 2004, 2007; Robinson et al., 2004) and in the single study in women with GDM (Robinson et al., 2009). Hyperglycaemia in conjunction with hyperinsulinaemia indicates a caffeine induced impairment in insulin sensitivity, which has previously been shown in healthy controls following both caffeine ingestion (Greer et al., 2001; Keijzers et al., 2002) and the consumption of caffeinated coffee (Moisey et al., 2008). This would support reduced insulin sensitivity as a mechanism of action for the negative impact of caffeine on glycaemic control.

Further research, in the form of randomised controlled trials, is warranted in individuals with type II diabetes to determine whether reducing or abstaining from caffeine altogether improves long-term glycaemic control (i.e. HbA1c) and, ultimately, the risk of diabetes-related complications. This is also true of GDM where the preservation of both foetal and maternal health is paramount. Determination of a safe ‘upper limit’ for caffeine ingestion is warranted (i.e. a specified amount of caffeine that, when ingested with or after a glucose/carbohydrate load, does not effect blood glucose concentrations). Current research suggests that this amount would be less than 200 mg (equivalent to one to two cups of instant coffee), with the potential to be reduced further for women with GDM where existing daily recommendations are already 200 mg (FSA, 2011) and 300 mg (Health Canada, 2010) for pregnant women. Replication of this study with trials of larger sample size is now required to determine the clinical implications of a regular intake or abstinence of caffeine during pregnancy on long-term glycaemic control and foetal outcomes.

In type I diabetes, there is early evidence of the potentially beneficial uses of caffeine in relation to both increased awareness, and the decreased duration of hypoglycaemic episodes indicating that HbA1c values remain unaffected by caffeine supplementation at 3 months. Patients with type I diabetes face an exacting task when trying to achieve normoglycaemia without inducing recurrent hypoglycaemic episodes that may impact on psychological health and cognitive function, with recurrent severe hypoglycaemia further being associated with depressed mood and poor quality of life (Frier, 2004). Debrah et al. (1996) were the first to indicate that caffeine might have a role in recognising the onset of hypoglycaemia in type I diabetes patients through augmentation of symptomatic and hormonal responses to a modest reduction in blood glucose. The highly significant increase in hypoglycaemic episodes with more intense warning symptoms indicates that caffeine may have the potential to have a beneficial impact.

Nocturnal hypoglycaemia accounts for approximately half of all severe episodes, and is particularly dangerous because the warning symptoms are blunted or absent during sleep (DCCT, 1991). It is of note that caffeine was observed to reduce overall daily hypoglycaemic episodes and also the duration of moderate overnight hypoglycaemic episodes (Richardson et al., 2005) through a validated measure of continuous blood glucose monitoring that can precisely quantify the duration and magnitude of hypoglycaemia. Further trials are now needed to confirm and quantify these results over prolonged time periods using caffeine supplementation and also to determine whether these can be replicated using dietary sources such as coffee or other caffeinated drinks.

The limited research within this area restricts any argument for change in current practice, policy or guidance for caffeine intake. Healthcare professionals should, however, be aware of the current research findings, and should await confirmation regarding the clinical implications of caffeine intake from future research trials. The long-term effects of caffeine intake on glycaemic control in type I diabetes are relatively unknown; however, published trials have shown that caffeine has the potential to play a beneficial role, especially if individuals with the disease either struggle to recognise episodes of hypoglycaemia or if nocturnal hypoglycaemia is problematic.

Further trials are warranted in this population with a greater emphasis on the effects of caffeine on blood glucose concentrations and/or insulin sensitivity. The mechanisms and physiology behind type I and type II diabetes vary greatly, and it is currently unknown whether caffeine exhibits the same effect in both diseases.


A number of limitations were associated with the diversity of methods of caffeine and glucose administration and differences in how glucose outcome measures were reported. In general, caffeine was administered in a single pharmacological dose, corresponding to the amounts reported for habitual consumption. It should be noted that this also met or exceeded the upper recommended limits for daily caffeine intake and that a single dose may not reflect the normal daily pattern of dietary consumption. Translation of the effects of pharmacological doses into a dietary context is therefore hindered.

Although the cross-over design of trials successfully addressed between-subject variability, and in so doing the sample size required in each study, there are limitations associated with the potential for a carry-over effect. To address this, seven trials were performed using a ‘washout period’ of at least 1 week or within a time period of 2 weeks (Table 1). Lane et al. (2008) failed to comment on the time between trial days, and Watson et al. (2000) assumed no carry-over effect. The ‘carry-over’ effect is arguable, given that caffeine is rapidly and completely absorbed and eliminated, with an mean half-life of 5 h (Charles et al., 2008), although the potential exists to bias results where a washout period is not defined.

Raw data were computed where possible to standardise individual series data into the iAUC (Jankelson et al., 1967). Further attempts were made to contact authors to gain original data that could be converted into iAUC measures for direct comparison, although these were unsuccessful. Agreement on standardised measures of the total AUC or mean iAUC would enable a more direct comparison between trials and the presentation of data as a meta-analysis, to quantify and compare effects and strengthen conclusions.

Further limitations at the review-level include the potential for language bias introduced through exclusion of trials that were not written in English.


In conclusion, current evidence suggests a negative effect of caffeine intake on blood glucose control in those with type II diabetes, with an increase in blood glucose concentrations and a decrease in insulin sensitivity being evident following the ingestion of a glucose load compared to placebo. Larger-scale, randomised controlled trials of a longer duration are needed to determine effects of timing and dose of daily caffeine ingestion on HbA1c, and also when taken in its more natural form of coffee or carbonated drinks. In individuals with type I and gestational diabetes, further high-quality research is needed to expand the current relatively small literature base.


The authors would like to thank the British Dietetic Association for accepting an initial abstract to the Dietitians New to Research Symposium 2012.

Conflict of interests, source of funding and authorship

The authors declare that there are no conflicts of interest.

No source of funding was sought for this work.

NW and HW contributed equally to this study. Both authors critically reviewed the manuscript and approved the final version submitted for publication.