| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
LETTER TO THE EDITOR |
Departments of Pediatrics and of Epidemiology and Biostatistics
McGill University Faculty of Medicine
Montreal Children's Hospital
2300 Tupper Street (Les Tourelles)
Montreal, PQ H3H 1P3
Canada
E-mail: michael.kramer{at}mcgill.ca
My colleagues and I thank Dr Buyken and her colleagues for their comments on the Promotion of Breastfeeding Intervention Trial (PROBIT). They claim that our study "addresses only the effect of prolonging the duration of exclusive breastfeeding." That is inaccurate, however, because the breastfeeding promotion intervention we randomly assigned (based on the WHO/UNICEF Baby-Friendly Hospital Initiative) was designed to increase both the duration of any breastfeeding and the exclusivity of breastfeeding in the first 6 mo of life. Both of those goals were achieved in the experimental group in our trial.
Buyken et al also comment that we "did not provide information on the mean number of weeks for which children in the intervention and control groups had been breastfed." Given the left truncation and skewed distribution of breastfeeding duration, "mean number of weeks" is not a very useful metric. In fact, our earlier article, published in JAMA (1), describing breastfeeding and health outcomes in the first year of follow-up, provided detailed information on the proportion of infants breastfeeding at all weeks in the first year, as well as the proportion exclusively breastfeeding at both 3 and 6 mo. These data were summarized in our recent AJCN article summarizing the anthropometric and blood pressure outcomes at the 6.5-y follow-up (2).
Buyken et al express their "particular concern" that "imprecise measurements may well have masked the likely modest effects of breastfeeding prolongation on the health outcomes assessed" at 6.5 y. Although random measurement errors and inter-polyclinic differences certainly did reduce the precision of the trial group differences for the triceps skinfold-thickness and blood pressure measurements, null effects with extremely narrow CIs were observed for body mass index (BMI; in kg/m2) and subscapular skinfold thickness. In an individual-subject data meta-analysis of observational studies, Owen et al (3) were able to control for confounding by socioeconomic status, maternal BMI, and maternal smoking during pregnancy in 11 studies. After such control, they found no significant reduction in mean BMI (–0.01; 95% CI: –0.05, 0.03) between breastfed and formula-fed infants, a result that is entirely consistent with our results comparing 2 randomly assigned groups differing in breastfeeding duration and exclusivity. Although my colleagues and I agree that our results cannot address the potential benefits of any breastfeeding, compared with no breastfeeding, in the first weeks of life, the results of the systematic review of Owen et al, comparing any breastfeeding with formula feeding, cast serious doubt on any claim of a "programming" effect of early breastfeeding that confers long-term protection against obesity.
Finally, my colleagues and I urge researchers and other readers to be skeptical about all reports of conditional effects. Many biostatisticians and clinical trialists have cautioned against subgroup analysis, even in properly randomized controlled trials (4, 5). In the face of an overall null result, it is inevitable that post hoc stratification by some baseline factor (even a factor such as an astrological sign) will yield a statistically significant result. Unless such conditional (subgroup) effects are hypothesized a priori and are replicated in other studies, however, they are highly likely to reflect type 1 errors—ie, chance findings. Thus, the observation by Buyken et al from an observational (nonrandomized) study that breastfeeding was protective only in boys of overweight mothers should be regarded with skepticism unless and until such results are reported in other studies.
ACKNOWLEDGMENTS
The author had no personal or financial conflict of interest.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
