Article Text

Download PDFPDF

Infant feeding practice and childhood cognitive performance in South India
  1. Sargoor R Veena1,
  2. Ghattu V Krishnaveni1,
  3. Krishnamachari Srinivasan2,
  4. Andrew K Wills3,
  5. Jacqueline C Hill3,
  6. Anura V Kurpad2,
  7. Sumithra Muthayya2,
  8. Samuel C Karat1,
  9. Mahadevu Nalinakshi1,
  10. Caroline H D Fall3
  1. 1Epidemiology Research Unit, Holdsworth Memorial Hospital, Mysore, Karnataka, India
  2. 2St John's Research Institute, St John's National Academy of Health Sciences, Bangalore, India
  3. 3MRC Epidemiology Resource Centre, Southampton General Hospital, Southampton, UK
  1. Correspondence to Dr S R Veena, Epidemiology Research Unit, Holdsworth Memorial Hospital, PO Box 38, Mandi Mohalla, Mysore 570021, India; veenasr{at}


Aim Several studies have suggested a beneficial effect of infant breastfeeding on childhood cognitive function. The main objective was to examine whether duration of breastfeeding and age at introduction of complementary foods are related to cognitive performance in 9- to 10-year-old school-aged children in South India.

Methods The authors examined 514 children from the Mysore Parthenon birth cohort for whom breastfeeding duration (six categories from <3 to ≥18 months) and age at introduction of complementary foods (four categories from <4 to ≥6 months) were collected at the first-, second- and third-year annual follow-up visits. Their cognitive function was assessed at a mean age of 9.7 years using three core tests from the Kaufman Assessment Battery for children and additional tests measuring long-term retrieval/storage, attention and concentration, visuo-spatial and verbal abilities.

Results All the children were initially breastfed. The mode for duration of breastfeeding was 12–17 months (45.7%) and for age at introduction of complementary foods 4 months (37.1%). There were no associations between longer duration of breastfeeding, or age of introduction of complementary foods, and cognitive function at 9–10 years, either unadjusted or after adjustment for age, sex, gestation, birth size, maternal age, parity, socio-economic status, parents' attained schooling and rural/urban residence.

Conclusions Within this cohort, in which prolonged breastfeeding was the norm (90% breastfed ≥6 months and 65% breastfed for ≥12 months), there was no evidence suggesting a beneficial effect of longer duration of breastfeeding on later cognitive ability.

Statistics from


Several studies, summarised in three systematic reviews and a meta-analysis, have concluded that children who were breastfed rather than formula-fed in infancy have a small but significant advantage in cognitive ability, ranging from 2 to 8 developmental quotient points.1,,4 Some studies also reported a ‘dose–response’ effect of duration of breastfeeding on cognitive ability.5,,8 These effects have been attributed to breast-milk constituents (long-chain polyunsaturated fatty acids required for brain development) and/or environmental factors (better mother–baby bonding and sensory stimulation in breastfed infants).1 910

Evidence for a beneficial effect of breastfeeding on cognitive development comes mainly from observational studies in children,1,,8 10,,12 including those born small-for-gestational-age (SGA) and/or very-low-birth weight,5 10 adults13 and older individuals.14 Recently, a randomised trial of a population-based breastfeeding promotion programme reported a significant benefit of the intervention on childhood cognitive ability.15 However, some observational studies, and another randomised controlled trial among preterm infants, found no associations between breastfeeding and later cognitive ability.16,,23

What is already known on this topic

  • In high-income countries, children who were breastfed rather than formula-fed and breastfed for a longer duration, tend to score higher on tests of cognitive function.

  • It is controversial whether this is causal or due to confounding, as higher parental education or socio-economic status influences breast feeding and also predicts childhood cognitive ability.

  • Data examining associations between the age at introduction of complementary foods and later cognitive ability are scarce.

What this study adds

  • This study adds to a very small literature on this topic from low-income countries.

  • Unlike high-income countries, the duration of breastfeeding in this south Indian population was unrelated to parental school attainment and only weakly related to socio-economic status.

  • In this cohort, in which prolonged breastfeeding was the norm, we found no evidence that longer duration of breastfeeding promotes childhood cognitive development.

Parents' socio-economic status (SES) and intelligence/education are strongly related to childhood cognitive performance. In high-income countries, these factors are also related to initiation and duration of breastfeeding24 and may confound the association of breastfeeding with cognitive function through postnatal nutrition, stimulation, growth and development. Therefore, adjustment for these potential confounders is important to assess the role of breastfeeding on cognitive performance. Data from developing countries are very few,6 8 but may be helpful in addressing the confounding effects, since, in these populations, breastfeeding may be unrelated to, or inversely related to, maternal SES and education. In India, a recent survey reported that early termination of breastfeeding was associated with higher maternal SES and education.25 Apart from breastfeeding, data examining associations between the age at introduction of complementary foods in infancy and later cognitive ability are scarce.

We have used data from the Mysore Parthenon Study26 27 to examine whether duration of breastfeeding and age at introduction of complementary foods are associated with cognitive ability, independent of sociodemographic factors, among 9–10-year-old South Indian children.


The Mysore Parthenon study, a prospective birth cohort study, was initiated in 1997–1998 mainly to examine the incidence and determinants of gestational diabetes in India and its short- and long-term effects in the offspring.26 27 A total of 830 women attending the antenatal clinic at the Holdsworth Memorial Hospital (HMH), Mysore, South India participated in the study. Of these, 674 (81%) delivered their babies at HMH. Excluding seven stillborn babies, and four with major congenital anomalies, newborn anthropometry was performed on 663 babies as reported previously.26 27 Excluding 25 deaths and eight with major medical problems, 630 children were followed up, with repeat anthropometry, annually until the age of five, and every 6 months thereafter.

At 1, 2 and 3 years of follow-up, information on infant feeding was obtained by asking mothers the same set of questions: How was the baby fed from birth (breast, bottle, breast+bottle or other)? If breastfed, was the baby still being breastfed? If no longer breastfed, what was the age (months) at which breastfeeding stopped? In addition, at 1 year's follow-up, mothers were asked the age (months) at which their baby started taking solid foods regularly. We did not collect data on duration of exclusive breastfeeding or on complementary foods other than solid foods. Breastfeeding and age at starting complementary foods data were available for 568 and 482 children, respectively.

At 9–10 years of age (September 2007 to May 2008) children were invited for assessment of their cognitive function. Of the 630 children, excluding 88 children (61—unwilling, 17—moved away from Mysore and 10—untraceable; of the 88 only 54 had breastfeeding data), 542 (86%) underwent cognitive testing. The current analysis is restricted to 514 children (249 boys and 265 girls) with complete breastfeeding data and cognitive outcomes (figure 1).

Figure 1

Flow chart of the study participants.

Cognitive tests

The cognitive measures consisted of a series of neuropsychological tests applicable for use in school-aged children related to specific cognitive domains (memory, attention, fluid reasoning) consistent with the Carroll model.28 They included three core tests from the Kaufman Assessment Battery for Children—second edition, 2004 (KABC-II)29 and additional tests30,,32 that underwent an extensive adaptation process to ensure their applicability in the local cultural context.33 These tests are described in table 1 and covered the domains of short-term memory, long-term memory and retrieval ability, visuo-spatial ability and language production. These tests were administered at HMH research unit in separate rooms free from distraction by two trained masters' level child psychologists (unaware of the children's breastfeeding status) in the local Kannada language, to each child in a single session of 60–90 min. Intraclass correlation coefficients for agreement between the two psychologists were 0.89 (intrarater agreement) and 0.90 (inter-rater agreement).

Table 1

Description of the cognitive tests used in the study

Covariates and confounders

We considered the following as potential covariates and confounders: ‘maternal factors’ (maternal age, parity, BMI and height in pregnancy); ‘infant factors’ (gender, gestational age at birth and birth weight); ‘child factors’ (current age, BMI and height) and ‘parental factors’ (parents' school attainment, rural/urban residence and current SES, assessed using the Standard of Living Index (SLI) designed by National Family Health Survey-2 which derives a score based on type and size of the house, household sanitary facilities, source of water and power supply, cooking fuel used, ownership of house/property, land, livestock and household assets). In Indian terms, SLI scores of 0–14 indicate a ‘low’ standard of living; scores of 15–24 indicate a ‘medium’ standard; and scores of 25–67 indicate a ‘high’ standard.34 Most of the families in our study would be described in India as ‘middle class’ or ‘lower middle class.’ Ninety-five per cent of our families had a private water supply direct to the house, and 5% obtain water from a public tap/pump/well; 99% of families used electricity as the main source of household lighting; 59% owned a bicycle; 12% owned a car; 83% owned a colour television and 12% a black and white television; and 40% owned a refrigerator. None of the mothers had ever smoked or consumed alcohol.

The HMH Research Ethics Committee approved the study, and informed verbal consent was obtained from parents and children.

Statistical methods

Skewed distributions were examined based on histograms and skewness statistics. For maternal BMI and the child's Kohs Block Design score, log transformation was the most appropriate, while for the child's pattern reasoning score, square root transformation provided a better approximation of a normal distribution. All cognitive tests scores were then converted to z-scores, which represent the difference from the mean score for each individual and are expressed in units of SDs. All the children were initially breastfed, and the main breastfeeding exposure was total duration of breastfeeding in months. Categories for duration of breastfeeding were chosen based on their meaningfulness from a public health perspective (eg, having a bin for <3 months) and also from a statistical perspective (ie, having enough subjects in each bin), and to reduce the effect of outliers and maintain ordering, breastfeeding duration was split into six categories (<3, 3–5, 6–8, 9–11, 12–17 and 18+ months) and considered as a quantitative variable. Age at starting complementary foods was split into four categories (<4, 4, 5, ≥6 months). Initially, associations of breastfeeding duration (six categories) and age at starting complementary foods (four categories) with potential confounders and outcomes were analysed using univariable linear regression model. Tests of a departure from linearity were performed using likelihood ratio (LR) tests, and where a departure from linearity was found (indicated by a LR test (p<0.05)), we reported F tests of general association. The associations between breastfeeding duration or age at starting complementary foods with cognitive outcomes were then analysed using a series of multivariable regression models, to examine whether any association between exposures and cognitive outcomes could be acting through socio-economic and parental factors, infant factors and/or the child's current size. Stata v10 was used for all analyses.


All the children were initially breastfed, and very few stopped breastfeeding before the age of 3 months. The mode for duration of breastfeeding was 12–17 months and for age at starting complementary foods 4 months, similar in boys and girls (table 2). Duration of breastfeeding was similar among children who underwent cognitive testing and children who did not take part in cognitive testing (data not shown). Girls scored better than boys in tests of word order (p=0.03), pattern reasoning (p=0.003), verbal fluency names (p<0.001) and coding (p<0.001). One per cent of mothers were illiterate, approximately 35% had received primary school education; 51% had received secondary school education; and 13% were graduates or postgraduates/professionals. Corresponding figures for fathers were 3%, 35%, 40% and 22%, respectively. Approximately 74% of the families were from urban areas and 26% from rural areas.

Table 2

Characteristics of the study cohort

Associations of breastfeeding duration with covariates and cognitive outcomes

There were significant non-linear associations between duration of breastfeeding and maternal age at pregnancy, birth weight, the child's current BMI and urban/rural dwelling (table 3). Short and long duration of breastfeeding were associated with higher maternal age, birth weight and the child's current BMI. Rural mothers tended to breastfeed for longer duration than urban mothers. There were no significant associations between duration of breastfeeding and parental attained schooling or SLI score (table 3). However, we noted that SLI score rose across the first four categories of duration of breastfeeding, and this was statistically significant (p=0.044).

Table 3

Associations of duration of breastfeeding with covariates and cognitive function scores

There were no significant linear or non-linear associations, for any of the cognitive outcomes, with duration of breastfeeding (table 3). This remained true after adjusting, in a series of models, for all the covariates (table 4). We also rechecked our results using breastfeeding duration as a continuous rather than categorised variable; the findings were unchanged. We noted that, as for SLI score, cognitive test scores tended to rise across the first four categories of duration of breastfeeding (table 3); however, none of these trends was statistically significant. Since some studies reported stronger associations of breastfeeding duration with later cognitive performance among SGA children (gestation and sex-specific birth weight <10th percentile)5 or low-birth-weight children (full-term birth weight <2500 g),6 we repeated the above analyses limiting the sample to SGA children (gestation and sex-specific birth weight <10th percentile),27 and (separately) to low-birth- weight children (full-term (>37 weeks) birth weight <2500 g). Once again, there were no significant associations between duration of breastfeeding and cognitive measures in this group.

Table 4

Associations between duration of breastfeeding, age at introduction of complementary foods and tests of cognitive function: multiple linear regression analysis

Associations of age at introduction of complementary foods with covariates and cognitive outcomes

Breastfeeding duration correlated weakly with age at starting complementary foods (Spearman r=0.11; p=0.02). Earlier introduction of complementary foods was associated with higher children's current BMI, family's SES and maternal education. Primiparous and urban mothers introduced complementary foods to their infants earlier than with multiparous and rural mothers (table 5).

Table 5

Associations of age at introduction of complementary foods with covariates and cognitive function scores

There were no significant associations between the age at starting complementary foods and any of the cognitive measures either unadjusted (table 5) or adjusted, in a series of models, for all covariates (table 4).


The results of this study from India support the null hypothesis that there was no association between either duration of breastfeeding, or age at introduction of complementary foods, and cognitive abilities in 9- to 10-year-old South Indian children.

The strengths of the study were that in a large sample of children, we had a battery of cognitive function tests specifically adapted for, and validated in, a South Indian population and collected data on a variety of potential confounders including birth weight, gestational age, maternal age, parity, height and BMI at pregnancy, SES, parents' education and rural/urban residence. Breastfeeding duration was obtained by maternal recall at 1 year for 94% of children (of whom more than 60% of the children were still breastfed, making these data reasonably accurate) and for the remainder at 2 and 3 years of follow-up. The limitations are that the dataset was relatively homogeneous in terms of breastfeeding (the majority of the children were breastfed for a year or more), we were unable to differentiate between exclusive and partial breastfeeding, and we did not have any information on frequency of breastfeeding, nutritional quality of breast-milk or type and nutritional quality of complementary foods.

To our knowledge, only two studies from developing countries have examined cognitive performance in relation to breastfeeding. One study in the Philippines reported that a longer duration (12 months+) of breastfeeding was associated with higher cognitive performance among low-birth-weight children aged 8 and 11 years.6 Another study (in Chile) found a non-linear association, with higher cognitive abilities in 5½-year-old children exclusively breastfed for 2–8 months compared with those breastfed for <2 months or >8 months.9 Many observational studies in high-income countries have shown higher cognitive performance among children4 7 10,,12 16,,22 and adults13 who were breastfed compared with those not breastfed and/or breastfed for a longer duration. Adjusting for SES, parental education, intelligence and other confounders, these associations lost their significance in some studies16,,22 but in some other studies tended to diminish, although they remained significant.7 10,,12 It is therefore unclear whether there is a genuine biological effect of breastfeeding upon cognitive development, or whether the association results entirely from confounding and remains in the latter studies only because of an inability to measure and adjust for all the relevant confounding factors. One possible explanation for the negative findings in our study is that SES was only weakly related to breastfeeding duration in Mysore, and confounding was not an issue, thus revealing a genuine lack of effect of breastfeeding on cognitive ability. An alternative explanation is that we failed to detect an effect because of a lack of heterogeneity in breastfeeding duration in our population. A striking difference between the studies in high-income countries and ours was that most infants in the former stopped breastfeeding <6 months,7 11 12 and ‘longer duration’ could mean anything from 2+ to 8+ months; few of our children were breastfed for as short a duration as this, and 65% were breastfed for 12 months or more. If the first 6 months of life is a critical period in which breastfeeding can influence cognition, we may have had inadequate power to detect this because almost all our children were breastfed during that time. It is possible that the nutritional quality of breast milk is important in this context. We have no data on the docosahexaenoic acid (DHA) and arachidonic acid (AA) (fatty acids important for infant brain development1) concentrations in our study. Studies in India reporting fatty acid composition in breast milk are scarce, and only one recent study reported that milk DHA levels in Indian women (consuming predominantly a vegetarian diet) were similar to milk DHA levels reported from Western and European women. Levels of plasma DHA and AA in Indian women were lower than the levels in American and European women. Maternal plasma omega-3 and omega-6 fatty acids levels were positively associated with their respective levels in milk, though there was no direct association between maternal plasma and milk DHA or AA levels.35 The study suggests that levels of LCPUFA vary between different populations and may be dependent on their dietary intakes. Low levels of these or other nutrients in breast milk could be another explanation for our negative findings.

Randomised intervention studies are a way of overcoming confounding, but it is impractical to randomise healthy babies to be breastfed or formula-fed. In a large randomised trial of a breastfeeding promotion programme in Belarus, which led to significant differences in breastfeeding initiation and continuation between intervention and control groups, children in the experimental group had higher test scores of intelligence and teachers' academic ratings compared with the control group at 6.5 years.15 However, this study was criticised36 because the paediatricians who administered the cognitive tests were not blind to the intervention status of the children, although the teacher ratings and results based on audit data were blinded. In another trial among preterm infants, 8-year-old children who were breastfed during infancy had higher test scores for IQ than formula-fed children.37 However, this trial involved non-randomised comparisons between breast-milk-fed and formula-fed infants, which might be biased by the socio-biological differences between these groups. The same authors, in a subsequent randomised multicentre study of preterm infants, reported no differences in psychomotor and developmental indices at 18 months between those receiving donor's breast milk and those fed on nutrient-enriched preterm formula.23 They concluded that considering the lower nutrient value of donor's breast milk, their data add significant support to the opinion that breast milk promotes neurodevelopment.

We found no association between the age at introduction of complementary foods and cognitive performance. Two studies have investigated associations between cognitive function and diet quality in infancy based on dietary history collected from/after 4 months.38 39 One study reported that meat consumption from 4–12 and 4–16 months was positively associated with children's psychomotor but not mental developmental indices up to 24 months of age.38 The other study reported that 4-year-old children who consumed more fruits, vegetables and home-prepared foods at infancy (6 and 12 months) had higher full-scale and verbal intelligence scores, independent of confounding variables.39 Infancy is a period of rapid brain growth, and it seems likely that the nutritive quality of complementary foods, including their timely introduction, is important for cognitive development. More research is required in this area.

To conclude, within this cohort, most of whom were breastfed for >12 months, we found no evidence that a longer duration of breastfeeding promotes cognitive development. We highlight the fact that prolonged breastfeeding was the norm among these children, and since brain development is most rapid in the first 6 postnatal months, it may be that breastfeeding at this time is beneficial but that we were unable to detect any effect, since almost all children were breastfed throughout that period. In support of this, cognitive scores tended to rise across the first three categories of duration of breastfeeding (although none of the trends was significant) suggesting that longer breastfeeding up to the age of 8 months may benefit cognitive development. There was no evidence of any association between the age at introduction of complementary foods and cognitive performance. Our study adds to a very small literature on this topic from developing countries. Despite the negative findings in relation to cognitive function, we strongly support WHO guidelines on infant feeding practices (exclusive breastfeeding for 6 months, introduction of nutritious complementary foods from 6 months and continued breastfeeding up to 2 years), which have been clearly shown to reduce infant infections and mortality, and prevent stunting, in developing country populations.40


We are grateful to the families who participated in the study and to BDR Paul (former Medical Director). We acknowledge the substantial contribution made to the study by I Annamma, BB Baby, L Lalitha, P Savitha, TP Prathibha, VR Asha, MN Jayakumar, A Saroja, S Geetha, KJ Chachyamma, M Stephen, KN Kiran, K Rumana, J Pearce and P Coakley.



  • Funding This study was funded by the Parthenon Trust, Switzerland, Wellcome Trust, UK and Medical Research Council, UK.

  • Competing interests None.

  • Ethics approval Ethics approval was provided by the HMH Research Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Patient consent Parental consent was obtained.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.