Search tips
Search criteria 


Logo of brjsmedBritish Journal of Sports MedicineVisit this articleSubmit a manuscriptReceive email alertsContact usBMJ
Br J Sports Med. 2007 August; 41(8): 523–530.
Published online 2007 May 4. doi:  10.1136/bjsm.2007.035147
PMCID: PMC2465427

Effect of caffeine supplementation on haematological and biochemical variables in elite soccer players under physical stress conditions



To evaluate the effect of caffeine on white cell distribution and muscle injury markers in professional soccer players during exercise.


22 male athletes completed a placebo controlled double blind test protocol to simulate a soccer match, followed by a Yo‐Yo intermittent recovery test.


Exercise caused an increase in packed cell volume that was enhanced by caffeine. Caffeine and exercise had a synergistic effect on the blood lymphocyte count, which increased by about 38% after exercise, and by an additional 35% when combined with caffeine. Caffeine promoted an exercise independent rise in circulating monocytes, and a synergistic action of exercise and caffeine was observed on segmented neutrophils. Caffeine promoted thrombocytosis. Plasma adenosine deaminase, aspartate aminotransferase, and lactate dehydrogenase concentrations were enhanced by exercise, and alanine transaminase concentration was enhanced in both groups, with a synergistic effect of caffeine.


The pronounced increase in the white cell count in the group receiving caffeine appeared to be caused by greater muscle stress and consequently more intense endothelial and muscle cell injury. The use of caffeine may augment the risk of muscle damage in athletes.

Keywords: leucocytosis, thrombocytosis, muscle damage markers, creatine kinase, transaminase

Several studies have found encouraging ergogenic effects of caffeine (1,3,7‐trimethylxanthine) during endurance exercises by acting as a fatigue delayer and as an enhancer of the contractile strength of cardiac and skeletal muscle.1,2 Nevertheless, adverse effects may occur in athletes who are susceptible to xanthine because of hormonal and metabolic features.3,4

Caffeine is an antagonist of adenosine A1 and A2 receptors and is also a psychomotor stimulant that quickly penetrates through the blood–brain barrier, increasing performance and delaying fatigue.5 It is known that xanthine acts on the central nervous system (CNS) and on various different metabolic pathways, with the overall result that glycogen is conserved because of an increase in fat oxidation and a decrease in carbohydrate oxidation.3 Distinct effects of caffeine are caused directly by adenosine receptor antagonism or indirectly by an increase in plasma adrenaline.6 Caffeine changes the immune response, inducing leucocytosis, lymphocytosis, and neutrophilia along with an increase in metabolic rate.7 Its use together with exercise activates both the hypothalamic‐pituitary‐adrenal axis and the autonomic nervous system, stimulating fast β‐endorphin and cortisol release.8 In addition, caffeine decreases muscular pain perception, effort perception, and the reaction time to a stimulus.9

Exercise causes an increase in some white blood cells such lymphocytes and neutrophils, leading to a mild leucocytosis.10,11,12 Synergy between vascular endothelial growth factor (VEGF) released after vascular muscle damage and the products of cellular metabolism, especially ADP, stimulates mononuclear cells such as white cells to migrate to the circulatory system.13 The strong muscle contractions during exercise may also cause micro‐tears both in muscle and in the vascular endothelium, which also affect the migration of white cells.14 In addition, many enzymes are released during skeletal muscle cell damage and exert immunomodulatory actions, serving as messengers to the immune system, in addition to being involved in circulating and tissue‐bound leucocytes, cytokines, hormones, and growth factors causing muscle regeneration.15,16

Caffeine and caffeine based substances have been increasingly used as ergogenic supplements by athletes and soccer players, but its effects on the human immune system during physical exercise are still obscure. In the present study, we evaluated the effect of caffeine supplementation on the white cell count and muscle injury markers in soccer players during exercise.



Professional soccer players (n = 22) from a first division team affiliated to the Confederação Brasileira de Futebol (CBF, Brazilian Soccer Confederation) participated in this study voluntarily. These athletes had no medical history of health problems and were not using ergogenic substances or any other drugs. The team was in contention for the Brazilian Soccer Championship, which guaranteed that all the players had similar diets, training regimens, and resting and sleep conditions. This was essential to ensure the control of many experimental variables that could otherwise affect the results.

Clinical examinations, anthropometric measurements, and laboratory tests were carried out on the subjects to assure integrity and homogeneity among the three experimental groups (table 11).). The initial laboratory evaluation tests included haematological and biochemical analyses, which allowed us to identify alterations in metabolism that could affect the results or impair their interpretation.

Table thumbnail
Table 1 Age and anthropometric measurements in soccer players assigned to groups CEx (caffeine and exercise), LEx (lactose and exercise) and C (caffeine)

Experimental protocol

The study was double blind and randomised. It was approved by the ethics committee for human research from the University Castelo Branco and met the requirements for carrying out research on human subjects (Health National Council, Brazil, 1996). Written informed consent was obtained from the subjects, who were instructed as to the nature and procedures of the study.

Prior to training (D0), blood and urine were collected from the fasting subjects. A clinical examination was done and anthropometric measurements made. No caffeine, xanthines, or other substance that could mask the results were ingested by the athletes for 72 hours before blood collection.

On the 13th day of training (D13), blood was collected from fasting soccer players (before breakfast (PRE)). The players were randomly divided into two groups and received a specific breakfast diet and a physical exercise programme: CEx = caffeine and exercise; LEx = lactose and exercise. Three athletes who were avoiding exercise for non‐physical technical reasons were used as control (C = caffeine without exercise). We present these data but do not discuss them because of the small number of subjects.

After receiving breakfast and the supplements, the subjects were driven to the test place, which took 15 minutes. After 20 minutes of warming up (articular mobilisation and elongation exercises), the subjects undertook the test protocol under cardiac monitoring, simulating a soccer match (fig 11).).

figure sm35147.f1
Figure 1 Experimental design and temporal curve. Before training, the athletes were submitted to a clinical and biochemical evaluation (D0) as described in Methods. After 12 normal training days, blood was collected (PRE), and the athletes received ...

Specific procedures

Diet supplementation

The different supplements were in indistinguishable capsules so that the subjects were not aware of the substance they were ingesting. Caffeine (Purifarma, China) was given to groups CEx and C at a dose of 5−1 in two 500 mg capsules, which were then completed with lactose (Via Farma, Brazil). The control group, LEx, received two capsules with 500 mg lactose each.

Test protocol

The variable distance run protocol (VDR) was used to simulate a soccer match and was carried out for 45 minutes in a 50×50 m court with 5×5 m marks. After hearing a beep, the athletes started running at top speed and in any direction. They randomly ran any of the 66 distances determined by a voice command following the beep and were separated by resting periods (frequency × distance + resting time: 7×10 m + 15 s; 12×20 m + 25 s; 15×30 m + 35 s; 17×40 m + 45 s, and 15×50 m + 60 s). At the end, all the athletes had run the same distances at variable speeds, which were correlated with each subject and associated with his soccer position. The athletes were allowed to drink a solution of electrolytes and glucose (Gatorade®) ad libitum throughout their training. After the VDR, a Yo‐Yo intermittent recovery test (Yo‐Yo IRT)17 was carried out to drive the athletes to exhaustion. This test ended at different times for each athlete, and immediately afterwards blood was collected for laboratory analyses using a double blind procedure (POST).

Data collection

Preliminary analyses (D0)

The preliminary tests were carried out on D0 to ensure that the individuals taking part in the study were in good metabolic condition. Blood and urine were collected from fasting athletes on the morning of the first day of training. Clinical examinations and anthropometric evaluations were done at the same time ((tablestables 1–6).

Table thumbnail
Table 2 Haematological variables in the soccer players from groups CEx (caffeine and exercise), LEx (lactose and exercise) and C (caffeine), determined on day 0
Table thumbnail
Table 3 Biochemical variables in the soccer players from groups CEx (caffeine and exercise), LEx (lactose and exercise) and C (caffeine) determined on day 0
Table thumbnail
Table 4 Hormonal levels in the soccer players from groups CEx (caffeine and exercise), LEx (lactose and exercise) and C (caffeine) determined on day 0
Table thumbnail
Table 5 Enzyme levels in the soccer players from groups CEx (caffeine and exercise), LEx (lactose and exercise) and C (caffeine) determined on day 0
Table thumbnail
Table 6 Haematological variables in the soccer players from groups CEx (caffeine and exercise), LEx (lactose and exercise) and C (caffeine) determined on day 13

Blood sampling (D0 and D13)

Venous blood samples were collected from the forearm into heparinised tubes. Immediately after the collection, the blood samples were centrifuged for plasma separation. Plasma was quickly frozen and stored at −70°C. The samples were analysed by Laboratório Bittar Ltda (Niterói, Brazil).

A range of analyses was carried out to detect any variable that could affect the results; urinary myoglobin and troponin I were also determined for evaluation of muscle integrity (data not shown).

Statistical analyses

Student's t test or analysis of variance was used to compare the treatments, and significant differences were set at p<0.05. Data are expressed as mean (SEM).


Preliminary tests (D0)

Unless stated, all the baseline haematological and biochemical variables were within the normal range for the group.

Athletes from all the groups had normal red and white cell counts at D0 (table 22).). The results from all the other analyses were also normal, and no difference was found within or among the groups. Ketone bodies and lipid profiles were similar among the groups, and the high density lipoprotein (HDL) concentration was high in all groups (mean (SEM), mmol/l: CEx = 55.2 (2.8); LEx = 52.9 (2.8), C = 53.7 (3.8)). Although cholesterol concentrations were higher in CEx (199.5 (6.2) mmol/l) than in C (150.0 (16.9)), this difference was not considered further because the main comparison was between CEx and LEx.

Serum biochemical analyses and serum hormones were within the normal population range. The plasma sodium concentration was slightly lower in CEx than in the other groups ((tablestables 3 and 44).). Serum concentrations of creatine kinase (CK) and creatinine kinase MB isoform (CKMB) were above normal levels in 14 subjects (table 55),), and troponin I and urinary myoglobin were at normal levels for all athletes. C reactive protein concentration was also within the normal values for all the groups (data not shown).

Exercise test (D13)

The haematological profile was affected by exercise, with an additional effect of caffeine. Erythrocyte count, haemoglobin, and packed cell volume values increased by about 4% in CEx (table 66),), and the red cell count increased by nearly 2% in LEx.

The white blood cell population was modified by both exercise and caffeine supplementation. The leucocyte count increased by 34% in LEx and by 61% in CEx, with an overall increase of nearly 80% in the supplemented group compared with LEx. Segmented neutrophils increased by 31% in LEx and by 58% in CEx. Lymphocytes increased 45% in LEx and 77% in CEx, and circulating monocytes increased by 30% in C and 50% in CEx. The thrombocyte count increased by 11% in LEx and by 24% in CEx (table 66).

After exercise, the total protein increased by 11% in CEx and by 5% in LEx, and this effect was significantly more pronounced in the CEx group. Both β and γ globulins were increased by exercise, independent of caffeine, and the effect was greater for β globulins (table 77).

Table thumbnail
Table 7 Plasma protein concentrations in the soccer players from groups CEx (caffeine and exercise), LEx (lactose and exercise) and C (caffeine) determined on day 13

For plasma enzyme determinations, no differences were found for γGT (table 88),), troponin I, or plasma myoglobin (data not shown) between PRE and POST periods or among the groups. Exercise caused an increase of 34% in plasma lactate dehydrogenase (LDH) concentration in the CEx group, similar to the unsupplemented LEx group. The same behaviour was observed for aspartate aminotransferase (AST) and adenosine deaminase (ADA) concentration, but the increase was less pronounced. The increase in alkaline phosphatase (AP) was 24% in the CEx group. Alanine aminotransferase (ALT) also rose significantly in both groups, the increase being approximately 50% greater in CEX than in LEx. There was no change in γ‐glutamyl transferase (γGT) in any of the groups (table 88).

Table thumbnail
Table 8 Changes in enzymes in the soccer players from groups CEx (caffeine and exercise), LEx (lactose and exercise) and C (caffeine) determined on day 13

A rise in CK was observed between the initial and final levels in both exercise groups (CEx = 30%, LEx = 13%). We plotted the CK values as a function of ALT, AST, AP, and γGT to investigate the origin of the raised CK, but no concentration differences were found among these enzymes ((figsfigs 2–5). Figure 22 shows the enhancement in CK and ALT for the CEx and LEx groups, in both cases significantly different from C. An increase in AST and AP v CK was found only in CEx ((figsfigs 3 and 44),), and no differences were found by linear regression analysis among CEx, LEx, and C. To analyse the origin of AST and ALT we plotted CK against γGT concentrations, but no significant differences were detected (fig 55).

figure sm35147.f2
Figure 2 Correlation between creatinine kinase (CK) and alanine aminotransferase (ALT) measured before and after physical exercise. Filled square, CEx PRE; empty square, CEx POST; filled circle, LEx PRE; empty circle, LEx POST; filled triangle, ...
figure sm35147.f3
Figure 3 Correlation between creatinine kinase (CK) and aspartate aminotransferase (AST) measured before and after physical exercise. Filled square, CEx PRE; empty square, CEx POST; filled circle, LEx PRE; empty circle, LEx POST; filled triangle, ...
figure sm35147.f4
Figure 4 Correlation between creatinine kinase (CK) and alkaline phosphate (AP) measured before and after physical exercise. Filled square, CEx PRE; empty square, CEx POST; filled circle, LEx PRE; empty circle, LEx POST; filled triangle, C PRE; ...
figure sm35147.f5
Figure 5 Correlation between creatinine kinase (CK) and γ‐glutamyl transferase (γGT) measured before and after physical exercise. Filled square, CEx PRE; empty square, CEx POST; filled circle, LEx PRE; empty circle, LEx ...


Caffeine supplementation has been shown to enhance both performance and reaction time in athletes practising different physical activities.2,18,19 In this study we examined the effects of caffeine supplementation on haematological variables, plasma proteins, and liver enzymes in soccer players under physical stress condition (a simulation of a soccer match). The adequacy of the experimental model was assured by the similarity among the subjects, all of whom were professional soccer players from a first division team, under an intense pre‐game regimen and thus having the same training programmes, diet, and resting and sleeping hours.

What is already known on this topic

  • Caffeine is a well described ergogenic aid. Xanthines act as fatigue delayers and muscle strength enhancers, acting on the central nervous system.
  • Caffeine has several metabolic effects including a rise in fat oxidation and the maintenance of glycogen stores.
  • These features have led to an increase in caffeine use in athletes.

What this study adds

  • This study showed important effects of caffeine on the white blood cell count in exercising footballers, suggesting that it aggravates the effects of muscle injury.

Many studies on exercise metabolism have used apparently healthy subjects, but there have been no preliminary physiological analyses guaranteeing that these conditions held true. In our view, a preliminary evaluation is necessary for subject selection because metabolic disorders may lead to erroneous and controversial conclusions. In the present study, this assessment of the subjects assured their homogeneity as a group, not only in relation to clinical and anthropometric data, age, and sex, but also in metabolic terms. The 85 biochemical analyses undertaken allowed us to select individuals with similar carbohydrate, lipid, and protein metabolism, similar oxygen transport efficiency, similar macronutrient anabolism and catabolism indicators, normal water and electrolyte metabolism, no infection or parasitic infestation, a well balanced body water content; and undisturbed hepatic and renal function. Reinforcing the usefulness of pre‐evaluation, one athlete was excluded because asymptomatic hepatitis was detected. In addition the extensive analyses gave us reference data for future evaluations.

Based on previous data on caffeine pharmacokinetics and dose dependency during both rest and physical exercise,5,18 we used a 5−1 caffeine dose in this study. This is within the positive supplementation range of 3–9−1 body weight and thus allowed us to predict an improvement in the athletes' performance during the experiment, even though this was not the main focus of this study.

Endurance training may affect the red cell count and impair physical performance.13,20 We found a significant exercise induced increase in packed cell volume enhanced by caffeine in our investigation. In addition we did not find any change in blood volume, which indicates that the enhancement was a result of red blood cell mobilisation by exercise.

Exercise increases the circulating red and white cell counts, largely because of their mobilisation from blood storage sites.13 With respect to circulating leucocytes, we found a synergistic effect of caffeine in addition to exercise. The total white cell count in the blood increased almost twofold more in response to exercise than in the non‐supplemented group. We showed a release of lymphocytes with exercise, and caffeine also promoted a rise in circulating monocytes. A synergistic action of exercise and caffeine (group CEx) was observed on segmented neutrophils. Exercise enhanced the lymphocyte count by about 38%, and when the effect of caffeine was added, the count increased by an additional 35%. These results are reinforced by the findings of Ramanaviciene et al,7 who showed that caffeine improved immunological activity by increasing the mobilisation of lymphocytes.

Exercise causes thrombocytosis, and as platelets respond to stimuli that recruit white cells from blood storage sites, they are possibly modulated by factors that affect leucocytosis.11,21 Physical stress promotes cell damage, accompanied by an acute inflammatory process22 which results in the mobilisation of lymphocyte subpopulations into the blood stream after exercise; thus the lymphocyte count increases during activity and declines once exercise ceases.12 Platelets may have increased because of the muscle and vascular trauma that occurs during physical exercise. The additional effect of caffeine on platelets may reflect its action on purinergic receptors—a proinflammatory action that appears to be mediated by adenosine monophosphate and protein kinase, or to be caused by release from the spleen.14

We found an enhancement of plasma ADA, AST, and LDH in the exercise groups. ALT was also increased in both groups, with a synergistic effect of caffeine. The impact of exercise and caffeine on these enzymes could reflect a loss of hepatocyte or muscle cell membrane integrity with consequent leakage of the proteins into the blood. Concentrations of AP and γGT in the blood were not changed in any of the groups, in agreement with Haralambie's findings,23 ruling out the possibility of hepatic injury. Thus the increase in plasma LDH, ALT, and AST suggests that the muscle lesions caused by exercise are enhanced by caffeine.24,25

The biggest increase in blood CK and LDH concentrations is detected in a 24–72 h post‐exercise window.26 In our study, blood was sampled immediately after physical exercise because of our interest in metabolic changes. This early collection explains why the concentrations of these enzymes were much lower here than in other studies.25 The CK concentration increased more in the CEx group than in the LEx group, probably because CKMB increased significantly in the CEx group. These data indicate that the increases in the enzymes analysed in this study resulted from muscle injury, and that caffeine increased the exercise induced damage, in agreement with Hoffman et al.24

Our study suggests that the pronounced increase in the white cell count in CEx was caused by greater muscle stress and consequently more intense endothelial and muscle injury. The immune system may play a role in modulating skeletal muscle repair after exercise injury, so the degree of exercise induced muscle damage is reflected by the subsequent leucocytosis. The subacute inflammatory process in group CEx, coupled with catecholamine release, resulted in enhancement of ADA concentration. This could reflect a subacute inflammatory response associated with adrenaline release rather than an effect of caffeine, as caffeine acts directly on white cells.26,27 Thus the increase in the white blood cell count in athletes from groups CEx and LEx was probably caused by mechanical effects rather than by hormonal action, in agreement with Boyum et al.26 Overall, the use of caffeine appears likely to aggravate muscle injury in athletes and may be one of the causes for the observed changes in white cell response.


We are grateful to Marta Louzada for stimulating discussions in the immune system cells.


C - caffeine only group

CEx - caffeine plus exercise group

LEx - lactose plus exercise group

VDR - variable distance run protocol

VEGF - vascular endothelial growth factor


1. Conway K J, Orr R, Stannard S R. Effect of a divided caffeine dose on endurance cycling performance, postexercise urinary caffeine concentration, and plasma paraxanthine. J Appl Physiol 2003. 941557–1562.1562 [PubMed]
2. McLean C, Graham T E. Effects of exercise and thermal stress on caffeine pharmacokinetics in men and eumenorrheic women. J Appl Physiol 2002. 931471–1478.1478 [PubMed]
3. Graham T E, Helge J W, MacLean D A. et al Caffeine ingestion does not alter carbohydrate or fat metabolism in human skeletal muscle during exercise. J Physiol (Lond) 2000. 529837–847.847 [PubMed]
4. Cavalcante J W S, Santos P R M, Menezes M G F. et al Influence of caffeine on blood pressure and platelet aggregation. Arq Bras Cardiol 2000. 75102–105.105 [PubMed]
5. Evans S M, Griffiths R R. Caffeine withdrawal: A parametric analysis of caffeine dosing conditions. J Pharmacol Exp Ther 1999. 289285–294.294 [PubMed]
6. Battram D S, Shearer J, Robinson D. et al Caffeine ingestion does not imped the resyntesis of proglicogen and macroglicogen after prolonged exercise and carboydrate supplementation in humans. J Appl Physiol 2004. 96943–950.950 [PubMed]
7. Ramanaviciene A, Acaite J, Ramanavicius A. Chronic caffeine intake affects lysozyme activity and immune cells in mice. J Pharm Pharmacol 2004. 56671–676.676 [PubMed]
8. Motl R W, Dishman R K. Effects of acute exercise on the soleus H‐reflex and self‐reported anxiety after caffeine ingestion. Physiol Behav . 2004;80577–585.585 [PubMed]
9. Kruk B, Chmura J, Krzeminski K. et al Influence of caffeine, cold and exercise on multiple choice reaction times. Psychopharmacology 2001. 157197–201.201 [PubMed]
10. Ikarugi H, Shibata M, Shibata S. et al High intensity exercise enhances platelet reactivity to shear stress and coagulation during and after exercise. Pathophysiol Haemos Thromb 2003. 33127–133.133 [PubMed]
11. Li N, Walllen N H, Hjemdahl P. Evidence for prothrombotic effects of exercise and limited protection by aspirin. Circulation 1999. 1001374–1379.1379 [PubMed]
12. Pedersen B K, Toft A D. Effects of lymphocytes and cytokines. Br J Sports Med 2000. 34246–251.251 [PMC free article] [PubMed]
13. Wang J S, Chow S E, Chen J K. Strenuous, acute exercise affects reciprocal modulation of platelet and polymorphonuclear leukocyte activities under shear flow in men. J Thromb Haemost 2003. 12031–2037.2037 [PubMed]
14. Horrigan L A, Kelly J P, Connor T J. Immunomodulatory effects of caffeine: friend or foe? Pharmacol Ther 2006. 111877–892.892 [PubMed]
15. Howatson D G, Van Someren K A. The efficacy of ice massage in the treatment of exercise‐induced muscle damage. Scand J Med Sci Sports 2005. 15416–422.422 [PubMed]
16. Paulsen G, Benestad H B, Strom‐Gundersen I. et al Delayed leukocytosis and cytokine response to high‐force eccentric exercise. Med Sci Sports Exerc 2005. 371877–1883.1883 [PubMed]
17. Krustrup P, Mohr M, Amstrup T. et al The yo‐yo intermittent recovery test: physiological response, reliability, and validity. Med Sci Sports Exerc 2003. 35697–705.705 [PubMed]
18. Graham T E, Spriet L L. Metabolic, catecholamine and exercise performance responses to various doses of caffeine. J Appl Physiol 1995. 78867–874.874 [PubMed]
19. Graham T E, Hibbert E, Sathasivam P. Metabolic and exercise endurance effects of coffee and caffeine ingestion. J Appl Physiol 1998. 85883–889.889 [PubMed]
20. Stewart I B, Darren E R, Alastair W. et al Cardiovascular and splenic responses to exercise in humans. J Appl Physiol 2002. 941619–1626.1626 [PubMed]
21. Drela N, Kozdron E, Szczypiorski P. Moderate exercise may attenuate some aspects of immunosenescence. BMC Geriatrics 2004. 41–7.7 [PMC free article] [PubMed]
22. Pizza F X, Peterson J M, Bass J H. et al Neuthrophils contribute to injury and impair its resolution after lengthenig contractions in mice. J Physiol (Lond) 2005. 562899–913.913 [PubMed]
23. Haralambie G. Serum gamma‐glutamyl transpeptidase and physical exercise. Clin Chim Acta 1976. 72363–369.369 [PubMed]
24. Hoffman J R, Maresh C M, Newton R U. et al Performance, biochemical, and endocrine changes during a competitive football game. Med Sci Sports Exerc 2002. 341845–1853.1853 [PubMed]
25. Chevion S, Moran D S, Heled Y. et al Plasma antioxidant status and cell injury after severe physical exercise. Proc Natl Acad Sci USA 2003. 1005119–5123.5123 [PubMed]
26. Boyum A, Ronsen O, Tennfjord V A. et al Chemiluminescence response of granulocytes from elite athletes during recovery from one or two intense bouts of exercise. Eur J Appl Physiol 2002. 8820–28.28 [PubMed]
27. Hitoglou S, Frydas S, Hatzistilianou M. et al Response of ADA and its isoenzymes in mice infected by Trichinella spiralis and treated with mimosine. Int J Immunopathol Pharmacol 2004. 17191–200.200 [PubMed]

Articles from British Journal of Sports Medicine are provided here courtesy of BMJ Publishing Group