American Journal of Epidemiology Advance Access originally published online on March 22, 2006
American Journal of Epidemiology 2006 163(9):872-873; doi:10.1093/aje/kwj135
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Letter to the Editor |
TWO OF THE AUTHORS REPLY
Clinic of Obstetrics, Division of "Experimental Obstetrics," Campus Virchow-Klinikum, Charité–University Medicine Berlin, 13353 Berlin, Germany
(e-mail: thomas.harder{at}charite.de)
We thank Ms. Quigley (1
) for her observations. Ms. Quigley raises concerns about our recent meta-analysis on duration of breastfeeding and risk of overweight (2). In essence, her comments focus on the possible impact and causes of heterogeneity and confounding and suggest that a meta-analysis on this research topic might have been inappropriate. While the majority of her concerns on sources of heterogeneity were discussed in our article (2), we feel that the arguments provided here further support our hypothesis of an independent influence of breastfeeding on risk of later overweight.
Ms. Quigley's first three points of criticism, regarding age, the definition of overweight, and the definition of breastfeeding as possible sources of heterogeneity, have already been addressed in the Results section of our article (2). In order to additionally exclude an effect of current breastfeeding, however, we recalculated the pooled odds ratio for all studies investigating probands at least 2 years of age by excluding studies with younger probands (3, 4). This led to an estimate completely identical to that obtained in the main analysis (odds ratio (OR) = 0.96, 95 percent confidence interval (CI): 0.94, 0.98). Similarly, the pooled odds ratio for all studies that used body mass index to define overweight was 0.96 (95 percent CI: 0.94, 0.98), which was not largely different from that for studies that used any other measure to define overweight (OR = 0.93, 95 percent CI: 0.87, 0.99). Furthermore, the pooled odds ratio from all studies that analyzed exclusive breastfeeding (OR = 0.94, 95 percent CI: 0.89, 0.99) did not differ largely from that obtained in the main analysis (OR = 0.96, 95 percent CI: 0.94, 0.98).
We agree with Ms. Quigley that variation in the composition of infant formula over time could be a reasonable source of heterogeneity. Given that adaption of formula composition to that of breast milk has been improved over the years, one would expect to find a smaller effect of breastfeeding on overweight risk in more recent birth cohorts than in precedent birth cohorts. Per example, however, after the studies are divided into two equal-sized subsets (containing five studies each), the pooled odds ratio for studies including probands born earlier (before 1982; OR = 0.98, 95 percent CI: 0.94, 1.02) does not differ largely from the pooled odds ratio for studies that used probands born later (i.e., during or after 1982; OR = 0.95, 95 percent CI: 0.93, 0.97). While these data speak against a possible effect of variations in formula, with regard to the outcome investigated here, namely change over time as a source of heterogeneity, they further confirm the robustness of the pooled estimate.
Finally, Ms. Quigley argues that a large proportion of the effect of breastfeeding duration on later overweight risk is attributable to confounders not considered in our analyses. In our opinion, however, Ms. Quigley's reanalysis (1) of our data does not support this view but rather corroborates our hypothesis of a causal relation between breastfeeding and later overweight risk: Irrespective of how the subgroups are defined, particularly regarding the definition of overweight, the pooled adjusted odds ratio does not differ in practical terms from the pooled unadjusted odds ratio. Therefore, in our opinion, a high stability of effect size, despite some heterogeneity in exposure and/or outcome definition, strengthens rather than weakens the validity of the relation evaluated.
Meanwhile, at least four different meta-analyses from three different groups of investigators, using different inclusion criteria and methodological frameworks, have independently shown an association between breastfeeding and later risk of overweight in the offspring (2, 5–7). The consistency of the effect observed in different settings and populations using unadjusted and/or adjusted data, as well as the observation of a biologically plausible dose-response relation (2), strongly supports the causality of this association, which could and should have considerable implications for public health, preventive medicine, and future research strategies.
ACKNOWLEDGMENTS
Conflict of interest: none declared.
References
- Quigley MA. Re: "Duration of breastfeeding and risk of overweight: a meta-analysis." (Letter). Am J Epidemiol 2006;163:870–2.
- Harder T, Bergmann R, Kallischnigg G, et al. Duration of breastfeeding and risk of overweight: a meta-analysis. Am J Epidemiol 2005;162:397–403.
[Abstract/Free Full Text] - Dubois S, Hill DE, Beaton GH. An examination of factors believed to be associated with infantile obesity. Am J Clin Nutr 1979;32:1997–2004.
[Abstract/Free Full Text] - Thorogood M, Clark R, Harker P, et al. Infant feeding and overweight in two Oxfordshire towns. J R Coll Gen Pract 1979;29:427–30.[Medline]
- Plagemann A, Harder T. Breast feeding and the risk of obesity and related metabolic diseases in the child. Metab Syndr 2005;3:192–202.
- Arenz S, Rückerl R, Koletzko B, et al. Breast-feeding and childhood obesity—a systematic review. Int J Obes 2004;28:1247–56.[CrossRef][ISI][Medline]
- Owen CG, Martin RM, Whincup PH, et al. Effect of infant feeding on the risk of obesity across the life course: a quantitative review of the published evidence. Pediatrics 2005;115:1367–77.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||