Breastfeeding and risk of inflammatory bowel disease.

Abstract

The recently published study by Baron et al (1) certainly sheds new light on the association between breastfeeding and inflammatory bowel disease (IBD). This study, which was conducted with the use of excellent methods, fulfills the criteria for the best-quality category in the meta-analysis we recently published (2). Incorporating the results of this study into the pooled estimate calculation would diminish the significant results of protective breastfeeding on Crohn disease (CD) [Mantel-Haenszel odds ratio (ORMH): 0.62; 95% CI: 0.27, 1.43] and would not affect significantly the summary estimate of the protective association between breastfeeding and ulcerative colitis (ORMH: 0.62; 95% CI: 0.43, 0.91). However, more important than its effect on the pooled estimate was the high heterogeneity that is implied from its inclusion in the CD studies (P 0.001, chi-square heterogeneity test). In our study, the effects found by all of the studies had high heterogeneity, but this may have been partly attributed to the differences in studies quality, with heterogeneity in the highest-quality studies that showed no statistical significance. Inclusion of the study by Baron et al as one of the highest-quality studies implies high heterogeneity in this group as well. Why some studies show a significant protective effect of breastfeeding while others show no effect or even suggest that breastfeeding is a risk factor for CD is an enigma that may have several possible explanations. One explanation relates to the different genetic characteristics of the studies’ populations. The highest-quality studies reviewed by us were all conducted in Sweden or North America; the study conducted by Baron et al was performed in northern France. It was previously shown that the genetic background of the population has a significant influence on the effect of some risk factors. A good example is the lack of effect of smoking on the development of CD in Jewish populations as opposed to other populations (3). The second explanation relates to the fact that CD may be regarded as a cluster of diseases that have the same manifestations but that are caused by different etiologies (4). Thus, the heterogeneic effect of breastfeeding on CD may relate to its different interactions with the yet unknown various etiologies of this disease. The third explanation suggested by Jantchou et al may also account for the discrepancy between this study’s findings and those of previous studies; the components of breast milk in northern France may differ significantly from the components of breast milk in less industrialized areas. Baron et al are the first investigators to implicate breastfeeding as a risk factor for CD. This, however, is not the only new finding of this study. The observed association between some vaccinations and CD in this study is also novel. This observation and the high rate of CD in this area suggest that the population of this study is unique either in its environmental exposure or in its genetic background. Thus, we agree with Jantchou et al that breastfeeding should not be discouraged, especially on the basis of one study. On the contrary, on the basis of our meta-analysis (which showed a protective effect of breastfeeding on IBD), the biologic plausibility of this association, and the experimental evidence gathered in animal experiments (5), we still believe that breastfeeding should be encouraged. Baron et al’s study does, however, emphasize the need for further highquality studies of other population types to fully understand the association between breastfeeding and IBD.