Meta-Analysis of Randomized and Nonrandomized Clinical Trials
Two remaining pediatric studies, although nonrandomized, were reviewed to increase the power of this analysis. Although the study participants were not randomized, there was no evidence of selection bias. The analysis was stratified to acknowledge this difference. In the study by Breese et al. (41), 10 consecutive patients had been treated with corticosteroids, and the next 18 patients were randomized to receive cyclosporin A or enteral nutrition. The study of Chafai et al. (42) was a retrospective analysis of patients treated with corticosteroids and enteral nutrition.
Papadopoulou et al. (31) completed a retrospective analysis of very similar treatment groups, yet their study included more than one treatment episode per patient. Repeated entry of the same patient could compound any selection bias, and thus we omitted this study from our analysis (despite its having the highest RR of 1.30 [nonsignificant CI] and favoring enteral nutrition). Nicholls et al. (30) randomized patients to receive either conventional treatment (steroids or enteral nutrition) or cyclosporin A. However, they then selected patients in the conventional treatment group with small bowel disease for treatment with enteral nutrition (n = 10) and those with colonic involvement for steroid therapy (n = 4). Although a 100% remission rate was achieved in both groups, we did not include this study in our final analysis.
Although treatment was clearly allocated in a nonrandom manner by Chafai et al., (42) the allocation was made in the best interests of the patient—that is, selection of the most effective and appropriate treatment for each patient. Although the comparison of outcome is less valid than if treatment allocation is randomized, the similarity of treatment groups in some way minimizes any selection bias.
The pooled relative risk for all seven studies was 0.982 (95% CI 0.73–1.33; values identical in fixed and random effects models). Heterogeneity Q statistic was 0.82 with 6 df (P = 0.991;Fig. 2, Table 3).
The inclusion of semirandomized studies assessed the sensitivity of our inclusion criteria. There was no significant change in RR or CI ranges and homogeneity was maintained (Table 3), confirming the two additional studies to be very similar to the truly randomized series. This calculation was extended to assess how many further studies would be required to reach a statistically significant treatment effect (Table 3). Only after the inclusion of 10 further studies (data not shown) that were identical in size and outcome to the largest reported pediatric study (n = 68) was a significant RR achieved in favor of steroid therapy (RR = 0.86 [95% CI 0.74–0.99]; Q = 0.83, 14 df;P = 1.0).
In view of the relatively limited number of children randomized, quantitative analysis of other factors had limited statistical value. We reviewed the available data with particular reference to speed of remission induction, duration of remission, nutritional endpoints, and height velocities. The only significant difference between the two treatment groups was in height velocities achieved at short-term follow-up. Thomas et al. (37) showed the children randomized to enteral nutrition to have a significantly higher mean height velocity standard deviation score (SDS) 6 months after treatment (+0.32 vs. −3.1). This was despite a significantly higher calorie intake in the steroid-treated group, indicating the growth suppressant effects of steroids are not overcome by simply increasing calorie intake.
An identical finding was reported by Sanderson et al. (13), who showed a significantly higher mean height velocity SDS at 3 months in the group randomized to enteral nutrition (+0.2 vs. −2.85). Chafai et al. (42) also confirmed a significantly higher growth velocity in the group receiving elemental diet.
In this analysis of 194 children with acute Crohn's disease, enteral nutrition was as effective as steroids at inducing a clinical remission (RR = 0.98, 95% CI 0.73–1.33).
Meta-analyses published in the adult literature (1–3) suggest a clear benefit of steroid therapy over enteral nutrition. These studies in adults with acute Crohn's disease all report an OR of 0.35 (95% CI 0.23–-0.53) for achieving a remission with enteral nutrition. The RR we calculated from data presented by Griffiths et al. (1) is 0.71. The 95% CI (0.56–0.90) remains below equivalence (RR = 1.0), yet closer to the range we present.
It is also of note that in almost all pediatric and adult studies a clinical disease activity index was used (44–47) to define a clinical remission. These indices, by including measures of a patient's clinical well-being, are likely to favor remission in the steroid-treated groups, given the effect steroids have on improvement of overall well-being.
A significant difference between the adult and pediatric studies is the degree of compliance. Up to 40% of adults were withdrawn from nutritional therapy because of intolerance of the diet (1). Not surprisingly this was highest in those taking an elemental or semielemental diet by mouth (18–41%). However, of 98 children included in this meta-analysis, only 8% did not complete the course of enteral nutrition, despite requiring a semielemental diet administered by nasogastric tube. Griffiths et al. report results excluding patients who withdrew from their study because of intolerance to the diet, yet the OR remains significantly in favor of steroid therapy, 0.57 (95% CI 0.35–0.94). Calculation of the pooled RR for the same data, however, fails to reach significance (RR = 0.86, 95% CI 0.68–1.09). This implies that adults who are able to tolerate nutritional therapy are as likely to go into remission as those treated with steroids, a fact also reported in large adult studies (48).
Compliance in excess of 90% is often reported when using enteral nutrition in children (13,30,31,38). Compliance with a prolonged course of exclusive enteral diet is clearly more difficult for an adult. In addition to the inconvenience, the possible benefits for growth and development do not apply.
Recent observational studies with more palatable polymeric diets are confirming excellent compliance in children who take the diet by mouth (49,50). Successful use of enteral nutrition also requires motivated nursing, dietetic, and medical staff to support children and families during therapy. This may well explain some of the better results obtained in pediatric units in which dietary therapy has become more commonplace (31,37,40,49).
A further difference from published adult data is that in most trials of adults, prevalent rather than incident cases are recruited. In our meta-analysis of pediatric trials a large proportion of randomized patients had newly diagnosed disease. Inclusion of patients with more long-standing or intractable disease may introduce a systematic bias into adult studies. Although not proven, patients with more recently diagnosed disease may respond better to nutritional therapy than those with more long-standing disease. This would make enteral nutrition more effective and better tolerated in the pediatric population.
A qualitative review of treatment effects on growth was also performed. Despite the well-described adverse effects that corticosteroids can have on development, growth, and pubertal maturation (51), only two of the randomized studies document the short-term effects of treatment on growth velocities. The randomized studies by Sanderson et al. (13) and Thomas et al. (37) documented significantly improved height velocity in the enterally fed groups at 3 and 6 months, respectively, although pubertal staging was not controlled for specifically.
It was also impossible to quantify the impact of enteral nutrition on different disease sites. In contrast to the widely held belief that enteral feeding is less effective in large bowel disease (13,52), available data from pediatric studies (38,40) suggests large and small bowel disease respond equally well to nutritional therapy. This is confirmed by the larger adult studies (33,53). As in the meta-analyses of about 400 adults with active Crohn's disease, it was also impossible to detect valid differences in speed of remission induction and time to first relapse.
The efficacy of polymeric diets is also still debated. There are insufficient randomized pediatric or adult data comparing elemental with whole protein, polymeric diets. Although the latter are the ideal clinical choice in palatability and cost, there are no comparative studies to confirm this.
Only 32 patients in the world literature have received polymeric diets in randomized trials comparing those diets with elemental diets; however, meta-analyses including studies of semielemental diets have been unable to document significant differences in efficacy between any one of the three types of diet (1,2). Even these analyses are only able to detect large differences in treatment effects. However, along with findings in multiple observational studies in children documenting the efficacy and palatability of polymeric diets (38,49,50), these findings provide some indication of their potential. Further randomized studies comparing polymeric and elemental diets may be justified to contribute to a statistically conclusive answer.
The inclusion of seven pediatric trials in the current analysis ensured sufficient power to detect a clinically significant difference between the two therapies. Further sensitivity analyses demonstrate that for steroids to achieve a statistically significant benefit over enteral nutrition, approximately 700 more children with acute Crohn's disease would require randomization. Results obtained from these children would also have to equal the most favorable outcome for steroid therapy reported so far (RR = 0.84) (40), if our current results were to be overruled. The volume of data necessary to overturn the evidence we report here greatly strengthens these findings (RR = 0.98, CI 0.23–0.53). The entire adult literature includes only approximately 400 randomized patients. With an incidence of less than 10 per 100,000 children (54), larger studies are unlikely to be feasible in children.
Besides reaching a statistically significant conclusion, a meta-analysis must assess and quantify interstudy variability for a conclusion to be valid. A best estimate that encompasses this heterogeneity may be much more robust than a result from a large, tightly controlled, one-center study. The seven studies included in this meta-analysis were found to be substantially homogeneous in their outcomes, making the pooled estimate less uncertain and more generalizeable. Although studies tend to become more diverse as researchers try to avoid duplication of work, this field of study has remained limited in its variability in view of the well-defined population and treatment strategies.
An inherent limitation of meta-analysis is the availability of good quality studies. A key factor in this is often the quality of randomization. Lack of allocation concealment and double-blinding can lead to substantially inflated results of evaluations of treatment effects (55).
Double-blinding in studies with enteral nutrition has never been possible, and although five studies stated that patients were randomized, none reported details of their randomization methods. Because most studies have reported no difference between the treatment effect of steroids and enteral nutrition, a publication bias is most unlikely. Studies demonstrating a clear benefit for either therapy are unlikely to have been overlooked. In addition, we selected studies in abstract form to ensure inclusion of all available data. Although this prevented a quality assessment of included trials as performed by some investigators (1,2), a quality score has not always been useful in a final analysis (1,3).
This systematic review provides both a quantitative and qualitative analysis of the data available on the use of enteral nutrition in children with active Crohn's disease.
If new research in this field is to contribute significantly to our practice and not merely support the current evidence, this analysis also provides an idea of the quality and volume of data required to achieve this. Allocation of funds and enrollment of patients, even if randomized, to studies merely supporting current good-quality evidence is inappropriate. In the absence of large, multicenter studies, however, this analysis demonstrates the benefit of small, well-designed randomized trials if the accumulation of quality evidence is to continue.
Corticosteroids clearly remain essential in the management of patients with Crohn's disease. However, despite their proven efficacy and ease of use, this should not preclude the use of an equally efficacious, but less-convenient therapy. The effect of steroids on growth and puberty make their repeated use particularly unsuitable for children approaching their major period of growth. Steroids also remain poor at healing bowel mucosa (56), whereas there are initial observational studies documenting improved mucosal cytokine profiles after enteral nutrition (28,50). Concerns about the long-term risks of malignancy in the face of active bowel inflammation continue (57–59). With more years of Crohn's disease ahead of them than adults, children may benefit more from achieving a better mucosal remission.
We therefore conclude that enteral nutrition is as effective as steroids in achieving a remission in children with active Crohn's disease. It is likely that a palatable, whole-protein polymeric diet is no less efficacious than an elemental, amino acid–based formula. This should make nutritional therapy a much more practical choice for both patient and clinician.
In view of the benefits of dietary therapy on growth, development, and probably the gut mucosa, an enteral diet should be recommended as first-line therapy in all children with active Crohn's disease.
The authors thanks Michael A. Stoto, Ph.D., Professor and Chair, Department of Epidemiology and Biostatistics, School of Public Health and Health Services, The George Washington University, Washington, DC, for his help with the statistical analysis.
Supported by grant T32-DK07477-16 from the the National Institutes of Health Training Grant in Pediatric Gastroenterology and Nutrition for training in clinical effectiveness (R.H.).
1. Griffiths AM, Ohlsson A, Sherman PM, Sutherland LR. Meta-analysis of enteral nutrition as a primary treatment of active Crohn's disease. Gastroenterology 1995; 108:1056–67.
2. Fernandez-Banares F, Cabre E, Esteve-Comas M, Gassull MA. How effective is enteral nutrition in inducing clinical remission in active Crohn's disease? A meta-analysis of the randomized clinical trials. JPEN J Parenter Enteral Nutr 1995; 19:356–64.
3. Messori A, Trallori G, D'Albasio G, Milla M, Vannozzi G, Pacini F. Defined-formula diets versus steroids in the treatment of active Crohn's disease: A meta-analysis. Scand J Gastroenterol 1996; 31:267–72.
4. Bernstein CN, Seeger LL, Sayre JW, Anton PA, Artinian L, Shanahan F. Decreased bone density in inflammatory bowel disease is related to corticosteroid use and not disease diagnosis. J Bone Miner Res 1995; 10:250–6.
5. Gokhale R, Favus MJ, Karrison T, Sutton MM, Rich B, Kirschner BS. Bone mineral density assessment in children with inflammatory bowel disease. Gastroenterology 1998; 114:902–11.
6. Wolthers OD, Riis BJ, Pedersen S. Bone turnover in asthmatic children treated with oral prednisolone or inhaled budesonide. Pediatr Pulmonol 1993; 16:341–6.
7. Walker-Smith JA. Dietary treatment of active Crohn's disease: Dietary treatment is best for children (letter; comment). BMJ 1997; 314:1827.
8. Savage MO, Beattie RM, Camacho-Hubner C, Walker-Smith JA, Sanderson IR. Growth in Crohn's disease. Acta Paediatr Suppl 1999; 88:89–92.
9. Ghosh S, Cowen S, Hannan WJ, Ferguson A. Low bone mineral density in Crohn's disease, but not in ulcerative colitis, at diagnosis (see comments). Gastroenterology 1994; 107:1031–9.
10. Walker-Smith JA. Management of growth failure in Crohn's disease. Arch Dis Child 1996; 75:351–4.
11. Navarro J, Vargas J, Cezard JP, Charritat JL, Polonovski C. Prolonged constant rate elemental enteral nutrition in Crohn's disease. J Pediatr Gastroenterol Nutr 1982; 1:541–6.
12. Morin CL, Roulet M, Roy CC, Weber A. Continuous elemental enteral alimentation in children with Crohn's disease and growth failure. Gastroenterology 1980; 79:1205–10.
13. Sanderson IR, Udeen S, Davies PS, Savage MO, Walker-Smith JA. Remission induced by an elemental diet in small bowel Crohn's disease. Arch Dis Child 1987; 62:123–7.
14. Beattie RM, Bentsen BS, MacDonald TT. Childhood Crohn's disease and the efficacy of enteral diets. Nutrition 1998; 14:345–50.
15. Griffiths AM. Inflammatory bowel disease. Nutrition 1998; 14:788–91.
16. Duerksen DR, Nehra V, Bistrian BR, Blackburn GL. Appropriate nutritional support in acute and complicated Crohn's disease. Nutrition 1998; 14:462–5.
17. O'Sullivan MA, O'Morain CA. Nutritional therapy in Crohn's disease. Inflamm Bowel Dis 1998; 4:45–53.
18. Murch SH, Walker-Smith JA. Nutrition in inflammatory bowel disease. Baillieres Clin Gastroenterol 1998; 12:719–38.
19. Griffiths AM. Crohn's disease in adolescents. Baillieres Clin Gastroenterol 1998; 12:115–32.
20. Walker-Smith JA, Murch SH. Diseases of the small intestine in childhood.
4th ed. Oxford, England: ISIS Medical Media, 1999.
21. Walker-Smith JA. Therapy of Crohn's disease in childhood. Baillieres Clin Gastroenterol 1997; 11:593–610.
22. Beattie RM, Camacho-Hubner C, Wacharasindhu S, Cotterill AM, Walker-Smith JA, Savage MO. Responsiveness of IGF-I and IGFBP-3 to therapeutic intervention in children and adolescents with Crohn's disease. Clin Endocrinol (Oxf) 1998; 49:483–9.
23. Khoshoo V, Reifen R, Neuman MG, Griffiths A, Pencharz PB. Effect of low-and high-fat, peptide-based diets on body composition and disease activity in adolescents with active Crohn's disease (see comments). JPEN J Parenter Enteral Nutr 1996; 20:401–5.
24. Belli DC, Seidman E, Bouthillier L, et al. Chronic intermittent elemental diet improves growth failure in children with Crohn's disease. Gastroenterology 1988; 94:603–10.
25. Morin CL, Roulet M, Roy CC, Weber A, Lapointe N. Continuous elemental enteral alimentation in the treatment of children and adolescents with Crohn's disease. JPEN J Parenter Enteral Nutr 1982; 6:194–9.
26. Ricour C, Duhamel JF, Nihoul-Fekete C. Continuous enteral feeding in children: Technique and indications in 170 cases from 1969 to 1975 (in French). Arch Fr Pediatr 1977; 34:154–70.
27. Azcue M, Rashid M, Griffiths A, Pencharz PB. Energy expenditure and body composition in children with Crohn's disease: Effect of enteral nutrition and treatment with prednisolone. Gut 1997; 41:203–8.
28. Breese EJ, Michie CA, Nicholls SW, et al. The effect of treatment on lymphokine-secreting cells in the intestinal mucosa of children with Crohn's disease. Aliment Pharmacol Ther 1995; 9:547–52.
29. Stober B, Nutzenadel W, Ullrich F. Basic diet in Crohn's disease (in German). Monatsschr Kinderheilkd 1983; 131:721–4.
30. Nicholls S, Domizio P, Williams CB, et al. Cyclosporin as initial treatment for Crohn's disease. Arch Dis Child 1994; 71:243–7.
31. Papadopoulou A, Rawashdeh MO, Brown GA, McNeish AS, Booth IW. Remission following an elemental diet or prednisolone in Crohn's disease. Acta Paediatr 1995; 84:79–83.
32. Summers RW, Switz DM, Sessions JTJ, et al. National Cooperative Crohn's Disease Study: Results of drug treatment. Gastroenterology 1979; 77:847–69.
33. Malchow H, Ewe K, Brandes JW, et al. European Cooperative Crohn's Disease Study (ECCDS): Results of drug treatment. Gastroenterology 1984; 86:249–66.
34. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Ins 1959; 22:719–48.
35. Breslow NE, Day NE. The design and analysis of cohort studies. In:Statistical methods in cancer research.
Vol. 2: Lyon, France: IARC Science Publications 1987;1–406.
36. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177–88.
37. Thomas AG, Taylor F, Miller V. Dietary intake and nutritional treatment in childhood Crohn's disease. J Pediatr Gastroenterol Nutr 1993; 17:75–81.
38. Ruuska T, Savilahti E, Maki M, Ormala T, Visakorpi JK. Exclusive whole protein enteral diet versus prednisolone in the treatment of acute Crohn's disease in children. J Pediatr Gastroenterol Nutr 1994; 19:175–80.
39. Seidman E, Lohoues MJ, Turgeon L, Bouthillier L, Morin CL. Elemental diet versus prednisone as initial therapy in Crohn's disease: Early and long term results (abstract). Gastroenterology 1991; 100:250.
40. Seidman E, Griffiths A, Jones A, Issenman R. Semi-elemental diet vs prednisolone in pediatric Crohn's disease (abstract). Gastroenterology 1993; 104:778.
41. Breese EJ, Michie CA, Nicholls SW, et al. Tumor necrosis factor alpha-producing cells in the intestinal mucosa of children with inflammatory bowel disease. Gastroenterology 1994; 106:1455–66.
42. Chafai S, Martin D, Goulet O, Mougenot JF, Ricour C, Schmitz J. Semi-elemental diet and corticosteroids in the treatment of Crohn's disease in children (abstract). J Pediatr Gastroenterol Nutr 1995; 220:465.
43. Breese E, MacDonald TT. TNF alpha secreting cells in normal and diseased human intestine. Adv Exp Med Biol 1995; 371B:821–4.
44. Hyams JS, Ferry GD, Mandel FS, et al. Development and validation of a pediatric Crohn's disease activity index. J Pediatr Gastroenterol Nutr 1991; 12:439–47.
45. Lloyd-Still JD, Green OC. A clinical scoring system for chronic inflammatory bowel disease in children. Dig Dis Sci 1979; 24:620–4.
46. Harvey RF, Bradshaw JM. A simple index of Crohn's-disease activity. Lancet 1980; 1:514.
47. Best WR, Becktel JM, Singleton JW, Kern FJ. Development of a Crohn's disease activity index. National Cooperative Crohn's Disease Study. Gastroenterology 1976; 70:439–44.
48. Malchow H, Steinhardt HJ, Lorenz-Meyer H, et al. Feasibility and effectiveness of a defined-formula diet regimen in treating active Crohn's disease. European Cooperative Crohn's Disease Study III. Scand J Gastroenterol 1990; 25:235–44.
49. Beattie RM, Schiffrin EJ, Donnet-Hughes A, et al. Polymeric nutrition as the primary therapy in children with small bowel Crohn's disease. Aliment Pharmacol Ther 1994; 8:609–15.
50. Fell JM, Hollis A, Paintin M, Arnaud-Battandier F, MacDonald TT, Walker-Smith JA. Normalisation of mucosal cytokine mRNA in association with clinical improvement in children with Crohn's disease treated with polymeric diet (abstract). J Pediatr Gastroenterol Nutr 1998; 26:544.
51. Brain CE, Savage MO. Growth and puberty in chronic inflammatory bowel disease. Baillieres Clin Gastroenterol 1994; 8:83–100.
52. Teahon K, Bjarnason I, Pearson M, Levi AJ. Ten years' experience with an elemental diet in the management of Crohn's disease. Gut 1990; 31:1133–7.
53. Lochs H, Steinhardt HJ, Klaus-Wentz B, et al. Comparison of enteral nutrition and drug treatment in active Crohn's disease: Results of the European Cooperative Crohn's Disease Study. Part IV (see comments). Gastroenterology 1991; 101:881–8.
54. Barton JR, Gillon S, Ferguson A. Incidence of inflammatory bowel disease in Scottish children between 1968 and 1983: Marginal fall in ulcerative colitis, three-fold rise in Crohn's disease. Gut 1989; 30:618–22.
55. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials. JAMA 1995; 273:408–12.
56. Modigliani R, Mary JY, Simon JF, et al. Clinical, biological, and endoscopic picture of attacks of Crohn's disease: Evolution on prednisolone. Groupe d'Etude Therapeutique des Affections Inflammatoires Digestives. Gastroenterology 1990; 98:811–8.
57. Leichtner AM, Jackson DW, Grand RJ. Crohn's disease and ulcerative colitis. In: Walker WA, Durie PR, Hamilton JR, Walker-Smith JA, Watkins JB, eds. Pediatric gastrointestinal disease: Pathophysiology, diagnosis, management. St Louis, MO: Mosby, 1996: 692–726.
58. Ekbom A, Helmick C, Zack M, Adami HO. Increased risk of large-bowel cancer in Crohn's disease with colonic involvement. Lancet 1990; 336:357–9.
59. Richards ME, Rickert RR, Nance FC. Crohn's disease-associated carcinoma: A poorly recognized complication of inflammatory bowel disease. Ann Surg 1989; 209:764–73.
Keywords:© 2000 Lippincott Williams & Wilkins, Inc.
Child; Crohn's disease; Enteral nutrition; Meta-analysis of publications; Steroids