Abstract

Observational studies suggest a longer duration of breastfeeding to be associated dose dependently with a decrease in risk of overweight in later life. The authors performed a comprehensive meta-analysis of the existing studies on duration of breastfeeding and risk of overweight. Studies were included that reported the odds ratio and 95% confidence interval (or the data to calculate them) of overweight associated with breastfeeding and that reported the duration of breastfeeding and used exclusively formula-fed subjects as the referent. Seventeen studies met the inclusion criteria. By meta-regression, the duration of breastfeeding was inversely associated with the risk of overweight (regression coefficient = 0.94, 95% confidence interval (CI): 0.89, 0.98). Categorical analysis confirmed this dose-response association (<1 month of breastfeeding: odds ratio (OR) = 1.0, 95% CI: 0.65, 1.55; 1–3 months: OR = 0.81, 95% CI: 0.74, 0.88; 4–6 months: OR = 0.76, 95% CI: 0.67, 0.86; 7–9 months: OR = 0.67, 95% CI: 0.55, 0.82; >9 months: OR = 0.68, 95% CI: 0.50, 0.91). One month of breastfeeding was associated with a 4% decrease in risk (OR = 0.96/month of breastfeeding, 95% CI: 0.94, 0.98). The definitions of overweight and age had no influence. These findings strongly support a dose-dependent association between longer duration of breastfeeding and decrease in risk of overweight.

The prevalences of overweight and obesity are increasing nearly worldwide. Therefore, a high priority has been given to research strategies to prevent the development of obesity. We have published a meta-analysis (1, 2) that showed breastfeeding, compared with formula feeding, to be associated with a decreased risk of overweight. Recently, this finding was confirmed by others (3). However, neither study investigated whether a relation exists between the duration of breastfeeding and the risk of overweight. This issue is of particularly high importance since it might support the causality of this association. Furthermore, it is highly relevant to clinical practice to know whether a longer duration of breastfeeding could lead to a stronger decrease in risk of overweight in later life. Therefore, we performed a meta-analysis of the relation between the duration of breastfeeding and the risk of overweight.

MATERIALS AND METHODS

Study base

The meta-analysis was conducted according to the checklist of the Meta-analysis of Observational Studies in Epidemiology (MOOSE) group (4). We performed a literature search including the databases MEDLINE (US National Library of Medicine's database accessed through PubMed, 1966 through December 2003), CINAHL (Cumulative Index to Nursing and Allied Health Literature, 1982 through December 2003), SERFILE (bibliographic information on biomedical and health science serials, 2002–2003), and EMBASE (the Excerpta Medica database, 1989 through December 2003), using the terms breastfeeding, infant nutrition, weaning, overweight, obes* (truncated), and adipos* (truncated) in the full-text option, without language restrictions. Furthermore, a manual search was carried out on all the references cited in published original studies and in all reviews identified by the literature search (517). To be eligible, studies had to fulfill the following three inclusion criteria: 1) be an original report comparing breastfed subjects with exclusively formula-fed subjects (referent group) of any given age, 2) report the odds ratio and 95 percent confidence interval (or data to calculate them) of overweight or obesity associated with breastfeeding, and 3) report the duration of breastfeeding for at least one exposure group. Any definition of overweight or obesity was allowed. From review of the abstracts identified in the search, 49 articles were subjected to full review; 33 of these studies were excluded since they did not fulfill the inclusion criteria (18 studies did not provide data to calculate the odds ratio, nine studies did not use exclusively formula-fed probands as the referent, and six studies did not report the duration of breastfeeding). Details are available on request.

Of the 16 original reports that met the inclusion criteria (1833), one consisted of two independent studies (24), so that 17 studies (16 cohort studies, one case-control study) were included in this meta-analysis. From these reports, data were abstracted in duplicate, using a standardized form.

Statistical analysis

Unadjusted odds ratios and 95 percent confidence intervals were calculated directly from the data given in the articles, where possible. Otherwise, the published odds ratio and 95 percent confidence interval were used. We used three different approaches to investigate by means of meta-analytical techniques whether a relation exists between the duration of breastfeeding and the risk of overweight. First, a weighted meta-regression was performed (34), using the duration of breastfeeding as the independent variable and the weighted odds ratio for overweight in breastfed probands, compared with formula-fed subjects, as the dependent variable. Second, the pooled odds ratio for overweight in breastfed subjects was calculated separately for five predefined categories of duration of breastfeeding. Third, the pool-first method (35) was used to combine the regression coefficients obtained from the studies.

Meta-regression analysis.

For meta-regression analysis, all duration-specific odds ratios had to be related to the respective duration of breastfeeding. Since the duration of breastfeeding was reported as categorical data with a certain range in the studies (e.g., 1–3 months, 4–6 months, and so on), the median of the upper and lower limits of each category was assigned to the particular estimate in each study (35). Estimates were plotted against the respective duration of breastfeeding as the independent variable. Since the scatterplot revealed a linear relation, a weighted meta-regression (34) with duration of breastfeeding as the covariate was performed (random-effects model). The regression coefficient with its 95 percent confidence interval was delogarithmized for data presentation.

Categorical analysis.

A pooled odds ratio for overweight in breastfed subjects was calculated for the five separate predefined categories of duration of breastfeeding: less than 1 month, 1–3 months, 4–6 months, 7–9 months, and more than 9 months. Since the Cochrane Q-based test revealed significant heterogeneity in each case, a random-effects model was used throughout.

Trend estimation.

To studies that provided data for more than two categories of duration of breastfeeding, we applied the “pool-first method” (35) to quantify the dose-response relation. This was possible for 11 studies (1923, 25, 2832). After visual inspection of the plots to ascertain model adequacy, we calculated a study-specific regression coefficient and corresponding 95 percent confidence interval for each study by use of a log-linear model. After exponentiation, the resulting odds ratio and 95 percent confidence interval for change in risk for each month of breastfeeding were pooled with a random-effects model.

Subgroup analysis.

Two subgroup analyses were performed. First, we calculated separate estimates for all studies that used body mass index to measure overweight and for all that did not. Second, age-specific estimates were pooled in the predefined subgroups 0–5 years and 6 or more years by the random-effects model.

Publication bias and statistical software.

Publication bias was assessed by inspection of the funnel plot and by formal testing for funnel plot asymmetry using the Begg test and the Egger test. Calculations were performed using STATA, version 8, software (Stata Corporation, College Station, Texas).

RESULTS

Study characteristics of included reports are displayed in tables 1 and 2. From the 17 studies that reported duration of breastfeeding, 14 gave data for more than one category of duration of breastfeeding, leading to 52 estimates included in the meta-regression analysis. Visual inspection of the scatterplot revealed that the relation between duration of breastfeeding and risk of overweight was linear. In the weighted meta-regression, duration of breastfeeding was significantly negatively related to risk of overweight (regression coefficient: 0.94, 95 percent confidence interval (CI): 0.89, 0.98) (figure 1).

FIGURE 1.

Scatterplot and meta-regression line of log odds ratio of risk of overweight/obesity associated with breastfeeding, according to duration of breastfeeding. A total of 17 studies provided 52 estimates of duration of breastfeeding and overweight. Weighted meta-regression revealed a significant inverse linear relation between the duration of breastfeeding and the risk of overweight (regression coefficient = 0.94, 95% confidence interval: 0.89, 0.98).

FIGURE 1.

Scatterplot and meta-regression line of log odds ratio of risk of overweight/obesity associated with breastfeeding, according to duration of breastfeeding. A total of 17 studies provided 52 estimates of duration of breastfeeding and overweight. Weighted meta-regression revealed a significant inverse linear relation between the duration of breastfeeding and the risk of overweight (regression coefficient = 0.94, 95% confidence interval: 0.89, 0.98).

TABLE 1.

Characteristics of the 17 studies, ordered alphabetically by first author, that are included in this meta-analysis, according to origin, design, data source, age, study size, loss to follow-up, exposure assessment method, and type of breastfeeding


Reference
 

Origin
 

Study design
 

Data source
 

Age
 

Study size (final no.)
 

Lost to follow-up (%)
 

Exposure assessment
 

Type of breastfeeding
 
Armstrong et al. (18Great Britain Cohort Survey 3–4 years 32,200 38 Records Exclusive 
Czajka-Narins and Jung (19United States Cohort Hospital 2 years 409 Not reported Records Partial 
Dubois et al. (20Canada Case-control Hospital 4–9 months 89 Not reported Questionnaire Partial 
Gillman et al. (21United States Cohort Survey 9–14 years 15,341 55.1 Questionnaire Partial 
Hediger et al. (22United States Cohort Survey 3–5 years 2,685 18 Questionnaire Partial 
Langnäse et al. (23Germany Cohort Hospital 1, 2, and 5–7 years 1,326 48 Not reported Not reported 
Liese et al. (I) (24Germany Cohort Survey 9–10 years 1,046 48 Questionnaire Partial 
Liese et al. (II) (24Germany Cohort Survey 9–10 years 1,062 47 Questionnaire Partial 
O'Callaghan et al. (25Australia Cohort Hospital 4–6 years 4,062 45 Questionnaire Not reported 
Parsons et al. (26Great Britain Cohort Survey 33 years 11,407 27 Not reported Partial 
Poulton and Williams (27New Zealand Cohort Hospital 3, 5, 7, 9, 11, 13, 15, 18, 21, and 26 years 695–939 9.5–33 Not reported Partial 
Richter (28German Democratic Republic Cohort Survey 6–7 years 2,385 Not reported Not reported Not reported 
Thorogood et al. (29Great Britain Cohort Hospital 1 year 66 59 Records Not reported 
Toschke et al. (30Czechoslovakia Cohort Survey 6–14 years 33,768 Questionnaire Partial 
Von Kries et al. (31Germany Cohort Survey 5–6 years 10,240 23 Questionnaire Exclusive 
Wadsworth et al. (32Great Britain Cohort Survey 6 years 3,731 Not reported Not reported Not reported 
Yeung et al. (33)
 
Canada
 
Cohort
 
Survey
 
1, 3, 5, and 6 months
 
316
 
23
 
Not reported
 
Partial
 

Reference
 

Origin
 

Study design
 

Data source
 

Age
 

Study size (final no.)
 

Lost to follow-up (%)
 

Exposure assessment
 

Type of breastfeeding
 
Armstrong et al. (18Great Britain Cohort Survey 3–4 years 32,200 38 Records Exclusive 
Czajka-Narins and Jung (19United States Cohort Hospital 2 years 409 Not reported Records Partial 
Dubois et al. (20Canada Case-control Hospital 4–9 months 89 Not reported Questionnaire Partial 
Gillman et al. (21United States Cohort Survey 9–14 years 15,341 55.1 Questionnaire Partial 
Hediger et al. (22United States Cohort Survey 3–5 years 2,685 18 Questionnaire Partial 
Langnäse et al. (23Germany Cohort Hospital 1, 2, and 5–7 years 1,326 48 Not reported Not reported 
Liese et al. (I) (24Germany Cohort Survey 9–10 years 1,046 48 Questionnaire Partial 
Liese et al. (II) (24Germany Cohort Survey 9–10 years 1,062 47 Questionnaire Partial 
O'Callaghan et al. (25Australia Cohort Hospital 4–6 years 4,062 45 Questionnaire Not reported 
Parsons et al. (26Great Britain Cohort Survey 33 years 11,407 27 Not reported Partial 
Poulton and Williams (27New Zealand Cohort Hospital 3, 5, 7, 9, 11, 13, 15, 18, 21, and 26 years 695–939 9.5–33 Not reported Partial 
Richter (28German Democratic Republic Cohort Survey 6–7 years 2,385 Not reported Not reported Not reported 
Thorogood et al. (29Great Britain Cohort Hospital 1 year 66 59 Records Not reported 
Toschke et al. (30Czechoslovakia Cohort Survey 6–14 years 33,768 Questionnaire Partial 
Von Kries et al. (31Germany Cohort Survey 5–6 years 10,240 23 Questionnaire Exclusive 
Wadsworth et al. (32Great Britain Cohort Survey 6 years 3,731 Not reported Not reported Not reported 
Yeung et al. (33)
 
Canada
 
Cohort
 
Survey
 
1, 3, 5, and 6 months
 
316
 
23
 
Not reported
 
Partial
 
TABLE 2.

Characteristics of the 17 studies, ordered alphabetically by first author, that are included in this meta-analysis, according to duration of breastfeeding, outcome assessment, definition of overweight/obesity, and confounders


Reference
 

Duration of breastfeeding
 

Outcome assessment
 

Definition of overweight and obesity
 

Confounders
 
Armstrong et al. (186–8 weeks MS* Obesity: BMI* > 95th percentile; severe obesity: BMI > 98th percentile Sex, birth weight, and socioeconomic status 
Czajka-Narins and Jung (192–4, 5–7, 8–10, and 11–18 months MS Overweight: BMI > 18.5 kg/m2 None 
Dubois et al. (20<1, 1–3, and >3 months MS Obesity: >90th percentile of weight/age None 
Gillman et al. (21<1, 1–3, 4–6, 7–9, and >9 months SR* Risk of overweight: BMI = 85th–95th percentile; overweight: BMI > 95th percentile Age, sex, Tanner stage, television, physical activity, eating habits, weight cycling, concerns to gain weight, birth order, household income, daily energy intake, maternal body mass index, birth weight, and maternal smoking 
Hediger et al. (22≤2, 3–5, 6–8, and ≥9 months MS Risk of overweight: BMI = 85th–94th percentile; overweight: BMI > 95th percentile Birth weight, ethnicity, age, sex, maternal body mass index, and age at introduction of solid food 
Langnäse et al. (23≤6 and >6 months MS Overweight: BMI > 90th percentile None 
Liese et al. (I) (24<6 and 6–12 months (exclusive breastfeeding: <2, 2–4, and 5–6 months) MS Overweight: BMI > 90th percentile Age, sex, city, nationality, socioeconomic status, and smoking 
Liese et al. (II) (24<6 and 6–12 months (exclusive breastfeeding: <2, 2–4, and 5–6 months) MS Overweight: BMI > 90th percentile Age, sex, city, nationality, socioeconomic status, and smoking 
O'Callaghan et al. (25≤2 weeks, 3–6 weeks, 7 weeks–3 months, 4–5 months, and ≥6 months MS Moderate obesity: BMI = 85th–94th percentile; marked obesity: BMI > 94th percentile Birth weight, sex, small for gestational age, eating problems, and sleeplessness 
Parsons et al. (26>1 month MS Obesity: BMI > 30 kg/m2 Maternal body mass index, social class, and maternal smoking 
Poulton and Williams (27≤6 and >6 months NR* Overweight: 3–15 years: percentiles (not further specified); >15 years: BMI > 25 kg/m2 Sex, birth weight, maternal education, and maternal and paternal overweight 
Richter (28<3, 3–6, and ≥7 months MS Overweight: weight > 120% None 
Thorogood et al. (29<1, 1–2, 3–4, 5–6, and >6 months MS Overweight: weight/50th percentile of weight divided by length/50th percentile of length >110% None 
Toschke et al. (30<1, 2–3, 4–6, and >6 months MS Overweight: BMI > 90th percentile; obesity: BMI > 97th percentile Parental education, parental obesity, maternal smoking, birth weight of >4,000 g, daily television watching of >1 hour, sport outside school, and siblings 
Von Kries et al. (31≤2, 3–5, 6–12, and >12 months MS Overweight: BMI > 90th percentile; obesity: BMI > 97th percentile Parental education, maternal smoking during pregnancy, birth weight of <2,500 g, own bedroom, and consumes butter more than 3 times per week 
Wadsworth et al. (32≤2, 3–4, 5–10, and >10 months NR Overweight: BMI > 90th percentile; obesity: BMI > 97th percentile Socioeconomic status during childhood, birth weight of >2,500 g, no. of persons per room at 2 years, and fat consumption at 4 years 
Yeung et al. (33)
 
≥2 months
 
MS
 
Obesity: weight/length > 95th percentile
 
None
 

Reference
 

Duration of breastfeeding
 

Outcome assessment
 

Definition of overweight and obesity
 

Confounders
 
Armstrong et al. (186–8 weeks MS* Obesity: BMI* > 95th percentile; severe obesity: BMI > 98th percentile Sex, birth weight, and socioeconomic status 
Czajka-Narins and Jung (192–4, 5–7, 8–10, and 11–18 months MS Overweight: BMI > 18.5 kg/m2 None 
Dubois et al. (20<1, 1–3, and >3 months MS Obesity: >90th percentile of weight/age None 
Gillman et al. (21<1, 1–3, 4–6, 7–9, and >9 months SR* Risk of overweight: BMI = 85th–95th percentile; overweight: BMI > 95th percentile Age, sex, Tanner stage, television, physical activity, eating habits, weight cycling, concerns to gain weight, birth order, household income, daily energy intake, maternal body mass index, birth weight, and maternal smoking 
Hediger et al. (22≤2, 3–5, 6–8, and ≥9 months MS Risk of overweight: BMI = 85th–94th percentile; overweight: BMI > 95th percentile Birth weight, ethnicity, age, sex, maternal body mass index, and age at introduction of solid food 
Langnäse et al. (23≤6 and >6 months MS Overweight: BMI > 90th percentile None 
Liese et al. (I) (24<6 and 6–12 months (exclusive breastfeeding: <2, 2–4, and 5–6 months) MS Overweight: BMI > 90th percentile Age, sex, city, nationality, socioeconomic status, and smoking 
Liese et al. (II) (24<6 and 6–12 months (exclusive breastfeeding: <2, 2–4, and 5–6 months) MS Overweight: BMI > 90th percentile Age, sex, city, nationality, socioeconomic status, and smoking 
O'Callaghan et al. (25≤2 weeks, 3–6 weeks, 7 weeks–3 months, 4–5 months, and ≥6 months MS Moderate obesity: BMI = 85th–94th percentile; marked obesity: BMI > 94th percentile Birth weight, sex, small for gestational age, eating problems, and sleeplessness 
Parsons et al. (26>1 month MS Obesity: BMI > 30 kg/m2 Maternal body mass index, social class, and maternal smoking 
Poulton and Williams (27≤6 and >6 months NR* Overweight: 3–15 years: percentiles (not further specified); >15 years: BMI > 25 kg/m2 Sex, birth weight, maternal education, and maternal and paternal overweight 
Richter (28<3, 3–6, and ≥7 months MS Overweight: weight > 120% None 
Thorogood et al. (29<1, 1–2, 3–4, 5–6, and >6 months MS Overweight: weight/50th percentile of weight divided by length/50th percentile of length >110% None 
Toschke et al. (30<1, 2–3, 4–6, and >6 months MS Overweight: BMI > 90th percentile; obesity: BMI > 97th percentile Parental education, parental obesity, maternal smoking, birth weight of >4,000 g, daily television watching of >1 hour, sport outside school, and siblings 
Von Kries et al. (31≤2, 3–5, 6–12, and >12 months MS Overweight: BMI > 90th percentile; obesity: BMI > 97th percentile Parental education, maternal smoking during pregnancy, birth weight of <2,500 g, own bedroom, and consumes butter more than 3 times per week 
Wadsworth et al. (32≤2, 3–4, 5–10, and >10 months NR Overweight: BMI > 90th percentile; obesity: BMI > 97th percentile Socioeconomic status during childhood, birth weight of >2,500 g, no. of persons per room at 2 years, and fat consumption at 4 years 
Yeung et al. (33)
 
≥2 months
 
MS
 
Obesity: weight/length > 95th percentile
 
None
 
*

MS, weight and height were measured by investigators; BMI, body mass index; SR, weight and height were self-reported by probands; NR, not reported.

Unadjusted data are reported only for “overweight,” not for “at risk for overweight.”

Table 3 shows the results of categorical analysis. From 1 month of breastfeeding onward, the risk of subsequent overweight continuously decreased up to a reduction of more than 30 percent, reaching a plateau at 9 months of breastfeeding.

TABLE 3.

Duration of breastfeeding and risk of overweight: categorial analysis (random-effects model)


 

Duration of breastfeeding
 
    

 
<1 month
 
1–3 months
 
4–6 months
 
7–9 months
 
>9 months
 
No. of duration-specific study estimates 14 15 11 
Odds ratio for overweight 1.0 0.81 0.76 0.67 0.68 
95% confidence interval
 
0.65, 1.55
 
0.74, 0.88
 
0.67, 0.86
 
0.55, 0.82
 
0.50, 0.91
 

 

Duration of breastfeeding
 
    

 
<1 month
 
1–3 months
 
4–6 months
 
7–9 months
 
>9 months
 
No. of duration-specific study estimates 14 15 11 
Odds ratio for overweight 1.0 0.81 0.76 0.67 0.68 
95% confidence interval
 
0.65, 1.55
 
0.74, 0.88
 
0.67, 0.86
 
0.55, 0.82
 
0.50, 0.91
 

Figure 2 shows the forest plot with odds ratio and 95 percent confidence interval and the pooled estimate for the reduction in risk of overweight for each month of breastfeeding, calculated from trend analysis by a random-effects model. Each month of breastfeeding was found to be associated with a 4 percent decrease in risk (odds ratio (OR) = 0.96/month of breastfeeding, 95 percent CI: 0.94, 0.98). A fixed-effects model revealed a similar pooled odds ratio and a nearly identical 95 percent confidence interval (OR = 0.96, 95 percent CI: 0.95, 0.98).

FIGURE 2.

Odds ratios (with corresponding 95% confidence intervals in parentheses) for overweight, per month of breastfeeding. Studies are ordered alphabetically by first author. The pooled or “combined” odds ratio (OR) was calculated by a random-effects model.

FIGURE 2.

Odds ratios (with corresponding 95% confidence intervals in parentheses) for overweight, per month of breastfeeding. Studies are ordered alphabetically by first author. The pooled or “combined” odds ratio (OR) was calculated by a random-effects model.

In only two of these studies (22, 31) was the influence of the duration of exclusive breastfeeding analyzed. The pooled odds ratio for risk of overweight per month of exclusive breastfeeding was 0.94 (95 percent CI: 0.89, 0.99; random-effects model).

Subgroup analyses revealed that the definition of overweight influenced the estimate only slightly. In studies that used body mass index to define overweight, the pooled odds ratio was 0.96 (95 percent CI: 0.94, 0.98) for eight studies, while the odds ratio was 0.93 (95 percent CI: 0.87, 0.99) for the three studies that used another measure to define overweight or obesity. Similarly, the age at examination had only a marginal influence on the magnitude of the effect of duration of breastfeeding on risk of overweight. The pooled odds ratio from all five studies investigating probands up to or including 5 years of age was 0.97 (95 percent CI: 0.94, 0.99), while in older subjects aged 6 or more years, it was 0.96 (95 percent CI: 0.93, 0.99) for six studies. No evidence of publication bias was observed, as indicated by a symmetric funnel plot (not shown) and a nonsignificant Begg test (p = 0.64) and Egger test (p = 0.77).

DISCUSSION

Using three different techniques, we show that a longer duration of breastfeeding is associated with a larger decrease in risk of overweight. Each of the three methods used in our study has its own advantages and limitations. Meta-regression analysis is known to be highly flexible with regard to the shape of the dose-response relation (34). However, the duration-specific estimates from one study are not completely independent from each other as they have the same referent category in each case, which could bias the result. This disadvantage can be at least partly overcome by the use of separate categories of duration of breastfeeding. However, because of the smaller sample sizes in some of the categories, the power of the statistical analysis might be limited. The “pool-first method” (35) is highly flexible with regard to the shape of the dose-response-relation, and it ensures complete independence of all included estimates. However, as a minimum of three categories is needed to calculate the study-specific regression coefficient in the log-linear model, in the case of our meta-analysis, the use of this technique had to be restricted to studies from which study-specific regression coefficients could be calculated. Nevertheless, in essence, all three methods came to the same result of an inverse linear association between duration of breastfeeding and risk of overweight in later life. However, it has to be considered that all studies performed until now on breastfeeding and risk of overweight are secondary analyses of health surveys or of studies designed primarily to answer different questions.

One major methodological problem to overcome in a meta-analysis of breastfeeding and risk of overweight is the change of the definition of overweight over time. Following the proposal of the Meta-analysis of Observational Studies in Epidemiology group (4), we included any definition of overweight and investigated the possible consequences of this strategy by subgroup analysis. In fact, the definition of overweight had only a minor impact on the pooled estimate.

Eight of the studies (18, 22, 2427, 31, 32) gave duration-specific, confounder-adjusted estimates, some only for subsets of the data. Because of this small sample size and considering the fact that the type and number of confounders differed largely among these studies, we did not calculate a pooled estimate of the confounder-adjusted odds ratio. Remarkably, only three (19, 21, 22) of the 17 studies gave some basic information on ethnicity, mostly by declaring a “mixed” ethnic background of the population. Therefore, it is unclear to date whether the effect of breastfeeding on risk of overweight is restricted to certain ethnic groups and might be confounded by social class. Taken together, a statistically based decision on the role of confounding could not be derived from the data here. However, Savitz (36) has postulated that, in general, the existence of a dose-response relation reduces the likelihood of an association to be completely due to confounding, since increasingly implausible scenarios are required for the exposure-confounder association to exaggerate the dose-response gradient.

The mechanisms by which breastfeeding affects the risk of overweight are still unclear. Breastfeeding results in a lower body weight gain during the critical neonatal period, obviously caused by a lower mean caloric intake in breastfed infants, compared with formula-fed neonates (37). A lower body weight gain during neonatal life has been shown to lead to decreased risk of obesity in adolescence and adulthood (38). In animal experiments, the kind of neonatal nutrition was shown to influence the development of neuroendocrine circuits in the mediobasal hypothalamus that regulates appetite control and body weight, with long-term consequences for risk of obesity (for review, refer to reference 39). These mechanisms might also explain why a longer duration of breastfeeding is associated with a stronger decrease in risk of overweight in later life.

In summary, we found that the duration of breastfeeding is inversely and linearly associated with the risk of overweight. The risk of overweight was reduced by 4 percent for each month of breastfeeding. This effect lasted up to a duration of breastfeeding of 9 months and was independent of the definition of overweight and age at follow-up. Even if interpreted as being of relatively small size, this association, if causal, might be of importance for the general population. Since the majority of studies analyzed here used partially breastfed subjects, it might be concluded that, beyond exclusive breastfeeding, also longer partial breastfeeding up to 9 months leads to a greater decrease in risk of overweight in later life, which might be considered in future clinical recommendations.

Conflict of interest: none declared.

References

1.
Plagemann A, Harder T. Breast feeding and the risk of obesity and related metabolic diseases in the child. Metab Syndr (in press).
2.
Harder T, Plagemann A. Ernährung und perinatale Programmierung von Adipositas: Zur epidemiologischen Evidenzlage. (In German). In: Zabransky S, ed. SGA-Syndrom. Ernährung und Wachstum. Marburg, Germany: Jonas Verlag,
2004
:43–51.
3.
Arenz S, Rückerl R, Koletzko B, et al. Breast-feeding and childhood obesity—a systematic review.
Int J Obes Relat Metab Disord
 
2004
;
28
:
1247
–56.
4.
Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group.
JAMA
 
2000
;
283
:
2008
–12.
5.
Parsons TJ, Power C, Logan S, et al. Childhood predictors of adult obesity: a systematic review.
Int J Obes Relat Metab Disord
 
1999
;
23
(suppl 8):
S1
–107.
6.
Butte NF. The role of breastfeeding in obesity.
Pediatr Clin North Am
 
2001
;
48
:
189
–98.
7.
Dewey KG. Is breastfeeding protective against child obesity?
J Hum Lact
 
2003
;
19
:
9
–18.
8.
Dwyer JT, Mayer J. Overfeeding and obesity in infants and children.
Bibl Nutr Dieta
 
1973
;
18
:
123
–52.
9.
Nutt HH. Infant nutrition and obesity.
Nurs Forum
 
1979
;
18
:
131
–57.
10.
Schetelig H. Die Bedeutung des Stillens in der Ernährung des Säuglings [the value of breast feeding in infant nutrition]. (In German).
Fortschr Med
 
1979
;
97
:
349
–52.
11.
Edelman B, Maller O. Facts and fictions about infantile obesity.
Int J Obes
 
1982
;
6
:
69
–81.
12.
Kovar MG, Serdula MK, Marks JS, et al. Review of the epidemiologic evidence for an association between infant feeding and infant health.
Pediatrics
 
1984
;
74
(suppl):
615
–38.
13.
Hamosh M. Does infant nutrition affect adiposity and cholesterol levels in the adult?
J Pediatr Gastroenterol Nutr
 
1988
;
7
:
10
–16.
14.
Lucas A. Does early diet program future outcome?
Acta Paediatr Scand Suppl
 
1990
;
365
:
58
–67.
15.
Dewey KG. Growth patterns of breastfed infants and the current status of growth charts for infants.
J Hum Lact
 
1998
;
14
:
89
–92.
16.
Dewey KG. Growth characteristics of breast-fed compared to formula-fed infants.
Biol Neonate
 
1998
;
74
:
94
–105.
17.
Martorell R, Stein AD, Schroeder DG. Early nutrition and later adiposity.
J Nutr
 
2001
;
131
(suppl):
874S
–80S.
18.
Armstrong J, Reilly JJ; Child Health Information Team. Breastfeeding and lowering the risk of childhood obesity.
Lancet
 
2002
;
359
:
2003
–4.
19.
Czajka-Narins DM, Jung E. Physical growth of breast-fed and formula-fed infants from birth to age two years.
Nutr Res
 
1986
;
6
:
753
–62.
20.
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.
21.
Gillman MW, Rifas-Shiman SL, Camargo CA, et al. Risk of overweight among adolescents who were breastfed as infants.
JAMA
 
2001
;
285
:
2461
–7.
22.
Hediger ML, Overpeck MD, Kuczmarski RJ, et al. Association between infant breastfeeding and overweight in young children.
JAMA
 
2001
;
285
:
2453
–60.
23.
Langnäse K, Mast M, Danielzik S, et al. Socioeconomic gradients in body weight of German children reverse direction between the ages of 2 and 6 years.
J Nutr
 
2003
;
133
:
789
–96.
24.
Liese AD, Hirsch T, von Mutius E, et al. Inverse association of overweight and breast feeding in 9 to 10-y-old children in Germany.
Int J Obes Relat Metab Disord
 
2001
;
25
:
1644
–50.
25.
O'Callaghan MJ, Gilliams GM, Andersen MJ, et al. Prediction of obesity in children at 5 years: a cohort study.
J Pediatr Child Health
 
1997
;
33
:
311
–16.
26.
Parsons TJ, Power C, Manor O. Infant feeding and obesity through the lifecourse.
Arch Dis Child
 
2003
;
88
:
793
–4.
27.
Poulton R, Williams S. Breastfeeding and risk of overweight.
JAMA
 
2001
;
286
:
1449
–50.
28.
Richter J. Zum Zusammenhang zwischen Stillzeit und Körpergewichtsentwicklung [influence of duration of breast-feeding on body-weight-development]. (In German).
Ärztl Jugendkd
 
1981
;
72
:
166
–9.
29.
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.
30.
Toschke AM, Vignerova J, Lhotska L, et al. Overweight and obesity in 6- to 14-year-old Czech children in 1991: protective effect of breast feeding.
J Pediatr
 
2002
;
141
:
764
–9.
31.
Von Kries R, Koletzko B, Sauerwald T, et al. Breast feeding and obesity: cross sectional study.
BMJ
 
1999
;
319
:
147
–50.
32.
Wadsworth M, Marshall S, Hardy R, et al. Breast feeding and obesity.
BMJ
 
1999
;
319
:
1576
–7.
33.
Yeung DL, Pennell MD, Leung M, et al. Infant fatness and feeding practices: a longitudinal assessment.
J Am Diet Assoc
 
1981
;
79
:
531
–5.
34.
Thompson SG, Sharp SJ. Explaining heterogeneity in meta-analysis: a comparison of methods.
Stat Med
 
1999
;
18
:
2693
–708.
35.
Greenland S, Longnecker MP. Methods for trend estimation from summarized dose-response data, with applications to meta-analysis.
Am J Epidemiol
 
1992
;
135
:
1301
–9.
36.
Savitz DA. Interpreting epidemiologic evidence. New York, NY: Oxford University Press,
2003
.
37.
Heinig MJ, Nommsen LA, Peerson JM, et al. Energy and protein intakes of breast-fed and formula-fed infants during the first year of life and their association with growth velocity: the DARLING study.
Am J Clin Nutr
 
1993
;
58
:
152
–61.
38.
Stettler N, Zemel BS, Kumanyika S, et al. Infant weight gain and childhood overweight in a multicenter, cohort study.
Pediatrics
 
2002
;
109
:
194
–9.
39.
Plagemann A. ‘Fetal programming’ and ‘functional teratogenesis’: on epigenetic mechanisms and prevention of perinatally acquired lasting health risks.
J Perinat Med
 
2004
;
32
:
297
–305.