Paper | Published:

Chromium picolinate for reducing body weight: Meta-analysis of randomized trials

International Journal of Obesity volume 27, pages 522529 (2003) | Download Citation

Subjects

Abstract

The aim of this meta-analysis was to assess the evidence of chromium picolinate for reducing body weight. Literature searches were conducted on Medline, Embase, The Cochrane Library, Amed and Ciscom. Nine experts and four manufacturers of commercial preparations containing chromium picolinate were asked to contribute published and unpublished studies. There were no restrictions regarding the language of publication. The screening of studies, selection, data extraction, validation and the assessment of methodological quality were performed independently by two reviewers. To be included, studies were required to state that they were randomized, double-blind and placebo-controlled, and report on body weight. Ten trials met all inclusion criteria and provided data, which were suitable for statistical pooling. For body weight a significant differential effect was found in favour of chromium picolinate (weighted mean difference: −1.1 kg; 95% confidence interval (CI): −1.8 to −0.4 kg, n=489). Sensitivity analysis suggests that this effect is largely dependent on the results of a single trial (weighted mean difference: −0.9 kg; 95% CI: −2.0 to 0.2 kg, n=335). Three of the reviewed trials reported on adverse events, indicating their absence in the treatment groups. In conclusion, our meta-analysis suggests a relatively small effect of chromium picolinate compared with placebo for reducing body weight. The clinical relevance of the effect is debatable and the lack of robustness means that the result has to be interpreted with caution.

Introduction

Obesity has become one of the most important avoidable risk factors for morbidity and mortality.1 The risk of developing, for instance, cancer, diabetes or heart disease increases with the degree of overweight in both men and women.2,3,4,5 While recent data from the US National Health and Nutrition Examination Surveys (NHANES) show an increase in the prevalence of obesity from 22.9% in NHANES III (1988–1994) to 30.5% in NHANES 1999–2000,6 figures from the UK suggest an increase from 6–8% in 1980 to 15–16.5% in 1994.7,8 One of the major factors responsible for this phenomenon is a decrease in energy expenditure from physical activity.9,10 These considerations and the notoriously poor compliance with conventional weight management programs emphasize the importance of effective and safe therapeutic options to reduce body weight. It is therefore hardly surprising to see a plethora of slimming aids being marketed.

Chromium picolinate is used as an oral treatment to help reduce body weight and alter body composition. It is an organic compound of trivalent chromium, an essential trace mineral and cofactor to insulin, and picolinic acid, a naturally occurring derivative of tryptophan. Chromium enhances insulin activity and has been the subject of a number of studies assessing its effects in carbohydrate, protein and lipid metabolism.11,12,13 Effects include an increase in lean body mass, a decrease in percentage body fat and an increase in basal metabolic rate.11,14 Chromium picolinate is advocated in the medical literature for reducing body weight15,16,17and preparations are sold as slimming aids in the US and Europe, and through the internet. Whether chromium picolinate is effective for this condition, however, is a matter of debate. The aim of this meta-analysis is to assess the evidence from randomized controlled trials of chromium picolinate for reducing body weight in human subjects.

Methods

Systematic literature searches were conducted using the data sources Medline, Embase, The Cochrane Library, Amed and Ciscom. The search terms used were chromium, picolinate, dietary supplements, food supplements and derivatives of these. Each database was searched from its inception until January 2002. Additionally, nine experts and four manufacturers of commercial preparations containing chromium picolinate were contacted and asked to provide published or unpublished trials.18 Hand-searches of relevant medical journals, conference proceedings and of our own files were conducted. The bibliographies of all papers located were searched for further studies. No restrictions regarding the language of publication were imposed.

To be included, studies were required to state that they were randomized, double-blind and placebo-controlled, use monopreparations of chromium picolinate and report on body weight of human subjects. Trials were excluded if patients received chromium picolinate as a multiple intervention concomitantly with other medications, diet or exercise. Data were extracted systematically according to patient characteristics, interventions and results. Methodological quality was evaluated using the scoring system developed by Jadad,19 which quantifies the likelihood of bias inherent in the trials, based on the description of randomization, blinding and withdrawals. The screening and selection of studies, data extraction, validation and the assessment of methodological quality were performed independently by two reviewers (MHP, CS). Disagreements in the evaluation of individual trials were largely due to reading errors and were resolved through discussion.

The mean change of body weight compared with baseline was defined as the primary endpoint and was used to assess the difference between the chromium picolinate and the placebo groups. Means and 95% confidence intervals (CI) were calculated using standard meta-analysis software (RevMan 4.1, Update Software Ltd, Oxford, UK), which uses the inverse of the variance to assign a weight to the mean of the within-study treatment effect. For most studies, however, the information was insufficient to allow us to directly calculate the variance of the preintervention to postintervention change. Studies generally reported data on preintervention mean and standard deviation and postintervention mean and standard deviation, but not the standard deviation of the change. The Cochrane Collaboration suggests imputing the variance of the change by assuming a correlation of 0.4 between preintervention and postintervention values.20 The variance of the change was imputed using this correlation factor and then used to assign a weight to the mean of the within-study treatment effect. Summary estimates of the treatment effect were calculated using a random effects model. The χ2 test for heterogeneity was performed to determine whether the distribution of the results was compatible with the assumption that intertrial differences were attributable to chance variation alone. Publication bias was assessed using a funnel plot whereby effect estimates of the common outcome measure were plotted against trial sample size. Sensitivity analyses were performed to test the robustness of the overall effect.

Results

The literature searches revealed 31 potentially relevant trials, which were reported in 30 articles.21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50 No unpublished studies were found. Fourteen studies were excluded because they were either not randomized, double-blind and placebo-controlled,31,32,33,34,35 did not test a monopreparation of chromium picolinate,36,37,38,39,40 did not report on body weight41,42,43or were published in duplicate.44 Ten trials met all inclusion criteria and provided data for body weight that were suitable for statistical pooling (Table 1). Seven other studies, which were reported in six articles,45,46,47,48,49,50were also included but reported data that were not suitable for statistical pooling (Table 2). None of these trials reported a significant reduction of body weight in participants treated with chromium picolinate compared with placebo. The flowchart (Figure 1) provides an overview of all included and excluded trials.

Table 1: Randomized, placebo-controlled, double-blind trials of chromium picolinate with parallel group design
Table 2: Randomized, placebo-controlled, double-blind trials with parallel group design lacking data for meta-analysis
Figure 1
Figure 1

Flowchart.

The result of the meta-analysis for body weight is shown in Figure 2. It suggests a significant reduction in body weight in patients receiving chromium picolinate compared with patients receiving placebo (weighted mean difference: −1.1 kg; 95% confidence interval: −1.8 to −0.4 kg, n=489). The χ2 test for heterogeneity suggests that the distribution of the results was compatible with the assumption that intertrial differences were attributable to chance variation (P=1).

Figure 2
Figure 2

Effects of chromium picolinate for body weight reduction (random effects model). The mean differences in the change from baseline are given with 95% confidence intervals. The vertical line represents no difference between chromium picolinate and placebo.

A funnel plot of the mean difference in body weight reduction plotted against trial sample size is shown in Figure 3. Visual inspection implies that studies with a smaller sample size are asymmetrically distributed to the right of the weighted mean difference of all trials. The two largest trials22,23suggesting a mean difference of about 1.2 kg in favour of chromium picolinate more closely reflect the weighted overall estimate. Although too few studies are available for firm statements, the funnel plot is consistent with some degree of publication bias.

Figure 3
Figure 3

Funnel plot of the mean difference in body weight reduction in trials of chromium picolinate, plotted against sample size. The vertical line indicates the weighted mean difference of all trials.

Sensitivity analyses were performed to test the robustness of the main analysis. We first tested whether removing the data of one trial,23 which accounted for 58% of the overall effect, would alter the direction of the result. The meta-analysis of these data21,22,24,25,26,27,28,29,30suggests the absence of a differential effect (weighted mean difference: −0.9 kg; 95% CI: −2.0 to 0.2 kg, n=335). Another sensitivity analysis assessed six trials,21,22,23,26,27,30 which all included overweight or obese patients for treatment periods ranging between 6 to 13 weeks (weighted mean difference: −1.1 kg; 95% CI: −1.8 to −0.4 kg, n=385).

This meta-analysis focused on the effect of chromium picolinate on body weight. However, additional data were available in eight other trials,21,22,23,25,26,27,29,30 which were used in a secondary analysis to provide an indication of the effect on percentage body fat and lean body mass. The meta-analysis of these data suggests a nonsignificant effect for lean body mass (weighted mean difference: 0.4 kg; 95% CI: −0.1 to 0.8 kg, n=416) and a significant effect for percentage body fat (weighted mean difference: −1.2%, 95% CI: −1.7 to −0.6%, n=416). The lack of robustness of the effect for percentage body fat is apparent when data from one trial,23 which accounted for 61% of the effect, were removed (weighted mean difference: −0.8%; 95% CI: −1.7 to 0.2%, n=262).

Three of the reviewed trials reported on adverse events.45,46,47 All of these demonstrate the absence of adverse events in participants receiving chromium picolinate.

Discussion

Our meta-analysis suggests a significant differential effect for body weight in favour of chromium picolinate. The size of the effect, however, is of debatable clinical relevance and the result of the main analysis lacks robustness. Therefore, we feel that this finding has to be interpreted with caution.

Two trials22,23 accounted for much of the weight in the main analysis. Both were published by the same author and were among only two other studies, which did not advise on maintaining a habitual dietary pattern as a measure of controlling lifestyle factors. In addition, they were the only studies that did not include a specific weight-training program, while the participants in all other trials were required to perform such a program. It is conceivable that the weight training may have obscured a further possible weight-lowering effect through an increase in lean body mass. This notion, however, is not supported by our data. The differences in study sample size may be of more interest. Both trials22,23 were the largest in the meta-analysis and each included more than 120 participants, while all other trials were of considerably smaller sample size. A possible type II error may have prevented the smaller trials from detecting significant intergroup differences. Therefore, the possibility exists that the two larger trials22,23 found a true effect. None of the other studies — four included less than 30 participants and four others included less than 50 participants — suggest a trend towards a differential effect.

Even if one accepts that the overall result reflects a true specific effect, its size is not impressive. A body weight reduction of 1.1–1.2 kg during an intervention period of 10–13 weeks (i.e. 0.08–0.1 kg/week) seems too small to be clinically meaningful. By comparison, a diet with a provision of 3300 kJ/day (800 kcal/day) achieves a mean weight loss of about 1.5–2.5 kg/week, and a more moderate energy restriction of 5000 kJ/day (1200 kcal/day) would induce a weight loss of 0.5–0.6 kg/week.51

Lifestyle changes including dieting and regular physical exercise are generally regarded as the basis for successful long-term weight loss, and limited evidence exists to support the effectiveness of pharmacotherapeutic options other than orlistat.5,52 The popularity of complementary and alternative medicine,53 however, has created a ready market for nonprescription weight loss products, which are also increasingly sold through the internet. For instance, data from a US survey of a random population sample of 14 679 adults demonstrate the common use of nonprescription weight loss products particularly among young obese women. Interestingly, 8% of women with no excess body weight were also reported to use such products.54 Even though these preparations are popular, they frequently lack evidence of effectiveness55 and some may be associated with severe adverse events. Three of the reviewed trials assessed adverse events45,46,47 and all agree on the total absence of adverse events in participants receiving chromium picolinate. Several other authors also found chromium picolinate to be relatively safe.11,56 However, literature searches on Medline, Embase and Amed identified data suggesting potential risks through chromosome damage.57 These were subsequently not confirmed in animal experiments58 and in studies involving humans.59 Two cases of acute rhabdomyolysis were associated with chromium picolinate in young men who took it as a part of their exercise regimen.60,61 Another case of severe renal impairment in a 33-year-old woman was reported.62 She had self-administered a dose several times higher than the recommended daily allowance. Finally, a 32-year-old man was reported to have ingested 1 mg of chromium picolinate daily for 4 days. He subsequently experienced acute generalized exanthematous pustulosis.63

Despite the fact that all included trials were randomized, double-blind and placebo-controlled, the extent of methodological rigor of these studies was variable. None of the studies scored the maximum possible score on the Jadad scale,19 and one scored as low as two of possible five points.30 Study participants were described as healthy volunteers, athletes and obese patients. In five studies21,22,23,26,27 sufficient information was provided to calculate the body mass index identifying the participants as overweight or obese. Another trial30 classified women as obese who exceeded 20–25% body fat. Four other studies24,25,28,29 included apparently healthy subjects but provided insufficient information to calculate the body mass index. None of these studies suggest a trend towards a differential effect.

Limitations of this meta-analysis pertain to the citation tracking and its potential incompleteness. Although strong efforts were made to locate and retrieve all trials on the subject, it is conceivable that some were not uncovered. The distorting effects on systematic reviews arising from publication bias and location bias are well-documented.64,65,66 This includes suggestions that positive findings may be overrepresented in complementary medicine journals67,68 and that these journals favour positive conclusions at the expense of methodological quality.69 In addition, there is evidence for the tendency of positive findings to be published in English language journals,70 and for some European journals not to be indexed in major medical databases.71 Therefore, the possibility of exaggerated treatment effects exists. Restrictions of literature searches relating to the language of publications and databases are therefore problematic. Databases searched for the purposes of this study included those with a focus on the American and European literature and those that specialize in complementary medicine. There were no restrictions in terms of publication language. We are therefore confident that our strategy has minimized bias. The funnel plot, however, suggests some degree of bias. The two largest trials, which are also the studies that reflect the weighted overall estimate most closely,22,23 report the largest reductions in body weight. It is interesting to note that these studies and five others21,25,26,27,29 were at least partly funded by the same manufacturer of commercially available chromium picolinate preparations.

In conclusion, our meta-analysis suggests a relatively small effect of chromium picolinate compared with placebo for reducing body weight. The clinical relevance of the effect, however, is debatable. The lack of sufficient robustness means that the overall result has to be interpreted with caution.

References

  1. 1.

    World Health Organization. Obesity: preventing and managing the global epidemic. World Health Organization: Geneva; 1998.

  2. 2.

    National Task Force on the Prevention and Treatment of Obesity. Overweight, obesity and health risk. Arch Int Med 2000; 160: 898–904.

  3. 3.

    , , , , , , , . Impact of overweight on the risk of developing common chronic diseases during a 10-year period. Arch Int Med 2001; 161: 1581–1586.

  4. 4.

    , , , . The effect of diet on risk of cancer. Lancet 2002; 360: 861–868.

  5. 5.

    , , , , , , , . Obesity and the risk of heart failure. N Engl J Med 2002; 347: 305–313.

  6. 6.

    , , , . Prevalence and trends in obesity among US adults 1999–2000. JAMA 2002; 288: 1723–1727.

  7. 7.

    Department of Health: Obesity. Reversing the increasing problem of obesity in England. DoH: UK: 1995.

  8. 8.

    . As obesity rates rise, experts struggle to explain why. Science 1998; 280: 1367–1368.

  9. 9.

    , . Obesity in Britain: gluttony or sloth? BMJ 1995; 311: 437–439.

  10. 10.

    , . Divergent trends in obesity and fat intake patterns: the American paradox. Am J Med 1997; 102: 259–264.

  11. 11.

    . Effects of chromium on body composition and weight loss. Nutr Rev 1998; 56: 266–270.

  12. 12.

    , . Chromium in human nutrition. Ann Rev Nutr 1988; 8: 543–563.

  13. 13.

    . Chromium in human nutrition: a review. J Nutr 1993; 123: 626–633.

  14. 14.

    . Essentiality of chromium in humans. Sci Total Environ 1989; 86: 75–81.

  15. 15.

    , (eds). Textbook of natural medicine. Churchill Livingstone: London; 1999. pp 1434–1436.

  16. 16.

    , (eds). Encyclopedia of natural medicine (2nd rev. edn). Prima Publishing: Rocklin; 1998. pp 690–691.

  17. 17.

    (Ed). Encyclopedia of nutritional supplements. Prima Publishing: Rocklin, 1996. pp 317–319.

  18. 18.

    , , , , , , , . Review of the usefulness of contacting other experts when conducting a literature search for systematic reviews. BMJ 1998; 317: 1562–1563.

  19. 19.

    , , , , , et al. Assessing the quality of reports of randomised clinical trials: is blinding necessary? Control Clin Trials 1996; 17: 1–12.

  20. 20.

    , , , . Variance imputation for overviews of clinical trials with continuous response. J Clin Epidemiol 1992; 45: 769–773. Cited in: Clarke M, Oxman AD (eds). Cochrane Reviewers Handbook 4.1.5. In: The Cochrane Library, Issue 2. Update Software. Oxford; 2002.

  21. 21.

    , , , . Effect of chromium supplementation and exercise on body composition, resting metabolic rate and selected biochemical parameters in moderately obese women following an exercise program. J Am Coll Nutr 2001; 20: 293–306.

  22. 22.

    , , , , . A randomized, double-masked, placebo-controlled study of the effects of chromium picolinate supplementation on body composition: a replication and extension of a previous study. Curr Ther Res 1998; 59: 379–388.

  23. 23.

    , , , . Effects of chromium picolinate supplementation on body composition: a randomized, double-masked, placebo-controlled study. Curr Ther Res 1996; 57: 747–756.

  24. 24.

    , , , . Chromium picolinate effects on body composition and muscular performance in wrestlers. Med Sci Sports Exer 1998; 30: 1730–1737.

  25. 25.

    , , . The effect of chromium picolinate on muscular strength and body composition in women athletes. J Strength Cond Res 2001; 15: 161–166.

  26. 26.

    , , , , , . Effects of resistance training and chromium picolinate on body composition and skeletal muscle in older men. J Appl Physiol 1999; 86: 29–39.

  27. 27.

    , , , , . Effect of resistance training with or without chromium picolinate supplementation on glucose metabolism in older men and women. Metabolism 1999; 48: 546–553.

  28. 28.

    , , , . Effects of chromium picolinate on beginning weight training students. Int J Sports Nutr 1992; 2: 343–350.

  29. 29.

    , , , , , . Effects of chromium and resistive training on muscle strength and body composition. Med Sci Sports Exer 1996; 28: 139–144.

  30. 30.

    , , , . Chromium and exercise training: effect on obese women. Med Sci Sports Exer 1997; 29: 992–998.

  31. 31.

    , . Effect of chromium picolinate on body composition. J Sports Med Phys Fitness 1995; 35: 273–280.

  32. 32.

    , , , . Chromium supplementation and resistance training: effects on body compositon, strength and trace element status of men. Am J Clin Nutr 1996; 63: 954–965.

  33. 33.

    , , , , , , . The short-term therapeutic efficacy of treating obesity with a plan of improved nutrition and moderate caloric restriction. Curr Ther Res 1992; 51: 261–275.

  34. 34.

    , , . Effects of carbohydrate and chromium ingestion during intermittent high-intensity exercise to fatigue. Int J Sport Nutr Exer Metab 2000; 10: 476–485.

  35. 35.

    , , , , , , , , . Clinical evidence for effectiveness of phencal in maintaining weight loss in an open label, controlled 2-year study. Curr Ther Res 1997; 58: 745–763.

  36. 36.

    , . Effects of chromium supplementation on fasting insulin levels and lipid parameters in healthy, non-obese young subjects. Diabetes Res Clin Res 1995; 28: 179–184.

  37. 37.

    , , , , , . Chromium supplementation of human subjects: effects on glucose, insulin and lipid variables. Metabolism 1983; 32: 894–899.

  38. 38.

    , , . The effectiveness of long-term supplementation of carbohydrate, chromium, fibre, and caffeine on weight maintenance. Int J Obesity 1997; 21: 1143–1151.

  39. 39.

    . Effect of chromium supplementation on glucose tolerance and lipid profile. Saudi Med J 2000; 21: 45–50.

  40. 40.

    , , , , , . Potential antioxidant effects of zinc and chromium supplementation in people with type 2 diabetes mellitus. J Am Coll Nutr 2001; 20: 212–218.

  41. 41.

    , , . Effects of chromium picolinate supplementation on insulin sensitivity serum lipids and body composition in healthy nonobese older men and women. J Gastroenterol 2000; 55A: M260–M263.

  42. 42.

    , , , . Clinical use of the trace element chromium (III) in the treatment of diabetes mellitus. J Trace Elements Exp Med 1995; 8: 183–190.

  43. 43.

    , , . The effects of chromium picolinate supplementation on body composition in different age groups. Age 1991; 14: 138.

  44. 44.

    , , , , . Chromium picolinate supplementation and resistive training by older men: effects on iron-status and hematologic indexes. Am J Clin Nutr 1997; 66: 944–949.

  45. 45.

    , . Beneficial effects of chromium picolinate supplementation on serum triglyceride levels in NIDDM. Diabetes Care 1994; 17: 1449–1452.

  46. 46.

    , , , , , , . Elevated intakes of supplemental chromium improve glucose and insulin variables in individuals with type 2 diabetes. Diabetes 1997; 46: 1786–1791.

  47. 47.

    , , , , , , . Effects of chromium picolinate supplementation on body composition, strength and urinary chromium loss in football players. Int J Sports Nutr 1994; 4: 142–153.

  48. 48.

    , , . The effect of chromium picolinate on serum cholesterol and apolipoprotein fractions in human subjects. West J Med 1990; 152: 41–45.

  49. 49.

    , , , , . Effekt von Chromhefe und Chrompicolinat auf die Körperzusammensetzung bei übergewichtigen, nichtdiabetischen Patienten während und nach einer Formula-Diät. Acta Med Aust 1997; 24: 185–187.

  50. 50.

    . The effect of chromium picolinate on insulin controlled parameters in humans. Int J Biosocial Med Res 1989; 11: 163–180.

  51. 51.

    , , . Obesity management by life-style strategies. Br Med Bull 1997; 53: 389–408.

  52. 52.

    , . Obesity. BMJ 2001; 322: 1406–1409.

  53. 53.

    , , et al. Trends in alternative medicine use in the United States 1990–1997: results of a national follow-up survey. JAMA 1998; 280: 1569–1575.

  54. 54.

    , , . Use of nonprescription weight loss products. Results from a multistate survey. JAMA 2001; 286: 930–935.

  55. 55.

    , , . The effectiveness of popular, non-prescription weight loss supplements. Med J Aust 1999; 171: 604–608.

  56. 56.

    , , , , . Alternative treatments for weight loss: a critical review. Crit Rev Food Sci Nutr 2001; 41: 1–28.

  57. 57.

    , , , . Chromium(III) picolinate produces chromosome damage in Chinese hamster ovary cells. FASEB 1995; 9: 1643–1648.

  58. 58.

    , , . Lack of toxicity of chromium chloride and chromium picolinate in rats. J Am Coll Nutr 1997; 16: 273–279.

  59. 59.

    , , , , , , . Effect of supplementation with chromium picolinate on antibody titers to 5-hydroxymethyl uracil. Eur J Epidemiol 1998; 14: 621–626.

  60. 60.

    , . Suspected chromium picolinate-induced rhabdomyolysis. Pharmacotherapy 1998; 18: 860–862.

  61. 61.

    , , . Rhabdomyolysis associated with nutritional supplement use. J Clin Rheumatol 2000; 6: 328–332.

  62. 62.

    , , , , . Chromium picolinate toxicity. Ann Pharmacother 1998; 32: 428–431.

  63. 63.

    , , , . Acute generalized exanthematous pustulosis induced by chromium picolinate. J Am Acad Dermatol 1999; 41: 820–823.

  64. 64.

    . The existence of publication bias and risk factors for its occurrence. JAMA 1990; 263: 1385–1389.

  65. 65.

    , , , . Publication bias in clinical research. Lancet 1991; 337: 867–872.

  66. 66.

    , . Bias in location and selection of studies. BMJ 1998; 316: 61–66.

  67. 67.

    , . Alternative therapy bias. Nature 1997; 385: 480.

  68. 68.

    , , . Bias in alternative medicine is still rife but is diminishing. BMJ 2001; 323: 1071.

  69. 69.

    , , , . Location bias in controlled clinical trials of complementary/alternative therapies. J Clin Epidemiol 2000; 53: 485–489.

  70. 70.

    , , , , , . Language bias in randomised controlled trials published in English and German. Lancet 1997; 350: 326–329.

  71. 71.

    , . Bias against European journals in medical publication databases. Lancet 1999; 353: 1592.

Download references

Author information

Affiliations

  1. Complementary Medicine, Peninsula Medical School, Universities of Exeter and Plymouth, Exeter, UK

    • M H Pittler
    •  & E Ernst
  2. Department of Exercise and Health Sciences, University of Bristol, Bristol, UK

    • C Stevinson

Authors

  1. Search for M H Pittler in:

  2. Search for C Stevinson in:

  3. Search for E Ernst in:

Corresponding author

Correspondence to M H Pittler.

About this article

Publication history

Received

Revised

Accepted

Published

DOI

https://doi.org/10.1038/sj.ijo.0802262

Further reading