Letters

Duration of breast feeding and adult arterial distensibility

Humans are primates, designed to breast feed for years not months

Explanation of findings and context before publication might have been helpful

Dose-response, cause and effect relation between breast feeding and heart disease seems unlikely

Authors did not discuss data from prospective studies

Does this study herald the return of national dried milk?

Breast feeding: distension or distortion?

Statistical analysis was unclear

Authors' reply

Summary of rapid responses

Humans are primates, designed to breast feed for years not months

EDITORWith respect to the article by Leeson et al on duration of breast feeding and arterial distensibility in early adult life, of course the duration of breast feeding mattersthe longer the better.¹

Humans are animals, mammals, and primates. Research on correlates of weaning age in non-human primates, such as adult body size, length of gestation, timing of permanent tooth eruption, timing of sexual maturity, and growth rates during childhood, predict that modern humans should be breast fed for between two and a half and seven years.²3 Humans have slightly longer durations of all stages of the life span than our nearest relatives, chimpanzees. We have slightly longer gestation, later dental eruption, later sexual maturity, and therefore would expect slightly later ages of weaning. Chimpanzees breast feed for four to five years. Around the world, many children are breast fed for two and a half to seven years, including some in the United States, Canada, and Great Britain.

Maybe a healthy start in life of several to many years of breast feeding should be followed by a lifelong diet low in animal protein and fat and high in physical exercise, to maximise heart health in adulthood. But we will not know this until researchers study the effects on blood vessel flexibility of normal durations of breast feeding (2.5-7.0 years), and of the combination of normal durations of breast feeding with different post-weaning diets and amounts of exercise. I find it appalling that researchers would suggest that more than four months of breast feeding could be harmful to children, when research shows that 2.5-7.0 years is clearly the normal and natural duration for our species. On a final note, it is always good advice to question the credibility of research and researchers funded by infant formula companies.

Katherine A Dettwyler, associate professor of anthropology and nutrition. 
Texas A&M University, College Station, TX 77843-4352, USA kadettwyler{at}hotmail.com

Competing interests: None declared.

Explanation of findings and context before publication might have been helpful

EDITORPerhaps it is not surprising that the report by Leeson et althat breast feeding followed by a high fat diet may later be associated with stiffer arteriesdrew so many responses.¹ Breast feeding is a sensitive topic. The normalising of artificial feeding by formula companies and the media requires efforts to protect breast feeding. But we should not allow our protective stance to become blindly defensive. It is easy to fall into the teleological trap of believing that breast feeding was designed, by God or nature, to be perfect. It is possible that evolution could have had this result. During our evolution we have had much shorter life spans than we have achieved recently and eaten much less fat. This is a sound study that adds a piece to the complex puzzle of how early nutrition may influence adult disease risks. It does not prove that breast feeding increases the risk of heart disease.

In their attempts to counter the study's conclusions, many of the respondents tilt at windmills. Some point out that rates of heart disease are low in developing countries where breast feeding for two years or more is common. But with changing diets and a higher proportion of elderly people, rates of heart disease are increasing rapidly in developing countries.² Many have dismissed the findings because they are based on maternal recall. This is an important epidemiological issue, which can be studied by comparing mother's recall with clinic records. Such studies show that mothers can accurately recall breastfeeding duration for as long as 29 years, but they are less reliable at recalling age at introduction of formula. ^{3 4} A Queensland study found that the differences in breastfeeding duration as recalled by 75 mothers (over one to 10 years) and recorded by the clinic were less than one month for 79% of children, and less than two months for 95% of children.⁵ They found no difference in accuracy of recall between mothers with different levels of education, or with numbers or ages of children.

Leeson et al described the limitations of their study and emphasised that it should not lead to any change in infant feeding recommendations. It is unfortunate but predictable that the media will sensationalise such research reports. If advocates of breast feeding fostered links with reputable infant nutrition researchers such as Lucas's team, perhaps an appropriate explanation of the findings and context could be prepared before publication. Midwives and breast feeding counsellors could then use this to reassure parents.

Wendy Holmes, women's and children's health specialist. 
International Health Unit, Macfarlane Burnet Centre for Medical Research, Fairfield, Victoria 3078, Australia holmes{at}burnet.edu.au

Competing interests: None declared.

Dose-response, cause and effect relation between breast feeding and heart disease seems unlikely

EDITOROnce again epidemiological data from the United Kingdom are leading to a claim that extended breast feeding may lead to later adverse cardiovascular outcomes. ^{1 2} Leeson et al say that their findings are consistent with those of Fall et al, which were widely publicised in the media. ^{1 3} The causal mechanism postulated by Fall et al was not found to hold in this study, and the results do not support a hypothesis of deranged blood lipid profiles in adulthood. Will this failure to confirm the previous hypothesis receive attention, or will the media say that this study "confirms" the findings of the previous one?

We will never know the impact of breast feeding on human health because it is unethical to randomise. Thus we have to be very careful to look for confounders when we do associative studies such as this, and Leeson et al made an effort to do so. Presumably, however, families with children who breast fed for longer periods in the United Kingdom 20-30 years ago differed from those who fed their babies closer to the norm of the time. Slightly over a quarter of British babies were breast fed for longer than four months in 1980⁴; similar to the proportion of those contacted who agreed to participate in this study.

The demographic and health survey data for South Asian countries show that about half the children are breast fed for longer than two years in India, two and a half years in Nepal, and three years in Bangladesh.⁵ Hundreds of millions of adults currently alive in that region were probably breast fed for even longer periods than this. If there were any dose-response, cause and effect relation between sustained breast feeding and heart disease, why is heart disease not at much higher levels there among those who reach old age than it is in rich countries? This study was conducted by a group that included the Medical Research Council childhood nutrition research centre, which has collaborated with the infant food industry for its outcome studies on nutrition. Their honesty in admitting this (or is it the exemplary BMJ insistence on such declarations?) may not allay our fears regarding the potential effects on the research of this kind of conflict of interest

Although Leeson et al point out that their findings cannot be interpreted as cause and effect, normally anything negatively associated with breast feeding quickly gets translated into just that by the media and receives wide dissemination. Let's see what happens with this one.

Ted Greiner, head, international nutrition research group. 
International Nutrition Research Group, Department of Women's and Children's Health, Uppsala University, S-75185 Uppsala, Sweden ted.greiner{at}kbh.uu.se

Competing interests: None declared.

Authors did not discuss data from prospective studies

EDITORLeeson et al propose a complex mechanism to explain their observations linking a putative marker of vascular risk with the duration of any breast feeding.¹ They suggest a dose-response relation between duration of any breast feeding and brachial artery distensibility and that extending breast feeding by two months has an effect on arterial distensibility broadly equivalent to that produced by a 4 mm Hg increase in blood pressure. The discussion of their findings is, however, not systematic. It neglects (as does the editorial by Booth²) to review important evidence. The observational findings by Leeson et al should be placed in the context of other epidemiological data relating directly to factors (in this case, blood pressure) whose link to adverse health outcomes are more clearly established than that of arterial distensibility.

One of the most important pieces of evidence comes from the seven year follow up by Wilson et al of the Dundee infant feeding study.³ In this study, systolic blood pressure at the age of 7 was found to be significantly raised in those children who had been exclusively formula fed for the first 15 weeks of life compared with those who had received any breast milk (mean 94.2 (95% confidence interval 93.5 to 94.9) mm Hg v 90.7 (89.9 to 917) mm Hg). These findings run counter to the observations by Leeson et al on distensibility, from which the opposite findings would be expectednamely, that blood pressure would be higher in those children who had been breast fed. Further evidence against the hypothesis of Leeson et al comes from the work of Taittonen et al, who found that breast feeding after 3 months of age was associated with an average reduction in blood pressure of 6.5 mm Hg.⁴

We were surprised that Leeson et al did not refer to their own related research published earlier this year in the Lancet, in which they concluded that consumption of breast milk was associated with lower blood pressure at age 13-16 years.⁵ This research was based on a unique opportunity afforded by a randomised trial to overcome some of the biases that are likely to be operating in observational studies, such as the one they report in your journal.

Jim Sikorski, general practice research fellow. 
Department of General Practice and Primary Care, Guy's, King's, and St Thomas's School of Medicine, London SE11 6SP jim.sikorski{at}kcl.ac.uk

Carol Dezateux, clinical senior lecturer in paediatric epidemiology. 
Department of Paediatric Epidemiology and Biostatistics, Institute of Child Health, London WC1N 1EH

Competing interests: None declared.

Does this study herald the return of national dried milk?

EDITOROn the day the article by Leeson et al was published, and because of the national furore it created, I was asked to generate a response for midwives to use to allay the worries of mothers telephoning the 24 hour advice line.¹ We work in the wards and departments of a busy obstetric unit in the north east of England, supporting mothers in the initiation of breast feeding, but not one mother or any family members queried the health benefits of breast feeding. People in the north east of England do watch television and read newspapers, so it seems they disregarded what they saw as another conflicting message from health professionals.

We promote breast feeding in areas where there has been a traditional bottle feeding culture. Articles such as the one by Leeson et al do not make our work any easier, but I agree with Holmes in her response that we should not be defensive and that all research should be scrutinised, even if it does threaten conventional wisdom. I also agree with other respondents that further, large scale research may lead to different conclusions. My own local response questions the statistical methods used and the effects of confounding variables (such as weaning patterns, definitions of exclusive or partial breast feeding, etc), which have already been raised by other respondents. But another issue that should be considered in this debate is the type of formulas in use during the period studied, between 1969 and 1975. Before 1974, most types of formula milk were still comparatively unmodified. Most contained 100% milk fats, which were difficult for young infants to digest and absorb. In this area, a large proportion of the population was fed evaporated milk and national dried milk during this period. Presumably, although Cambridge is a more affluent area, the formulas available were still comparatively unmodified.

The 1974 report, Present Day Practice in Infant Feeding (first report), led to the withdrawal of national dried milk and stated that all artificial milk should approximate the composition of breast milk as nearly as is practicable. Formula manufacturers have since spent many millions (or billions) trying to meet this objective.

Given the time scale, it seems that many of the respondents in the reported study would have been fed unmodified infant formula. Do the findings of this study herald the return of national dried milk, as it seems from this study that these types of formula have benefits over breast milk? I do not think so; other factors need to be considered.

Anne Holt, senior lecturer and practitioner. 
South Cleveland, Hospital, University of Teesside, Middlesbrough TS4 3BW anne.holt{at}tees.ac.uk

Competing interests: Unrepentant mother of four children, all of whom have been breast fed for over a year; two have been breastfed for over four years.

Breast feeding: distension or distortion?

EDITORIt seems that we live in an increasingly dangerous world: recent media scares have included calcium channel blockade, the oral contraceptive pill, and, latterly, long distance air travel. Now it seems that even breast feeding, promoted for its benefits by our grandparents, is not without risk, as described by Leeson et al in their paper.¹ But the scientific evidence on which such claims are based varies substantially.

Leeson et al set out to test the hypothesis that breast feeding is associated with a detrimental reduction in arterial distensibility. Why they sought to measure distensibility of the brachial artery as an early marker of cardiovascular disease is unclear. Although they say that arterial distensibility diminishes with age in relation to other risk factors, the references they cite concern changes in the carotid and femoral arteries and aorta, and not, as in their study, the brachial artery. This is an important distinction: although aortic distensibility does decrease with age, brachial distensibility does not change.² Moreover, despite careful application of the same methods employed by Leeson et al, others have shown that age and hypercholesterolaemia do not influence brachial distensibility. ^{3 4} Although we agree that aortic pulse wave velocity (a measure of distensibility) does predict cardiovascular outcome in hypertensive and normotensive people and those with renal disease, we are unaware of any data suggesting that the same is true of brachial distensibility.

Overall, Leeson et al could not show any difference in brachial distensibility between those who were breast fed and those who were not. Brachial pulse pressure, a surrogate measure of large artery stiffness that predicts outcome, did not differ significantly between the two groups. There was, however, an inverse association between the duration of breast feeding and distensibility, but this was significant only in women. This is surprising since their original hypothesis was based on the observation that boys who are breast fed up to 1 year of age have an increased risk of ischaemic heart disease in later life.⁵

As a result of the resulting media coverage, many mothers may choose not to breast feed their infants despite much evidence as to its benefits, including a reduction in cardiovascular disease in later life, as noted by Leeson et al. After the pill scare many women stopped taking the oral contraceptive pill, which resulted in a rise in unplanned pregnancies. Finally, there is the propensity for the infant food industry to use such data and media coverage out of context for commercial benefit.

Ian B Wilkinson, lecturer in clinical pharmacology. 
Clinical Pharmacology Unit, University of Cambridge, Addenbrooke's Hospital, Cambridge CB2 2QQ

John R Cockcroft, senior lecturer in cardiology. 
Department of Cardiology, Wales Heart Research Institute, University Hospital, Cardiff CF14 4XN cockcroftjr{at}cf.ac.uk

Competing interests: None declared.

Statistical analysis was unclear

EDITORLeeson et al, in the conclusions in their paper, seem to rely in part on the statistical treatment of their limited observational data,¹ in particular, the use of multiple regression analysis and t tests. Multiple regression was used to determine regression coefficients as a measure of association between length of breast feeding and non-invasive brachial artery distensibility. The t tests were used to test the null hypothesis (presumably) that there were no differences in brachial artery distensibility between those who were not breast fed, those who breast fed to age 4 months, and those who breast fed above age 4 months.

We are uncertain whether the regression analysis incorporated all subjects, but this is implied in the paper. Therefore, we are presented with a larger number of those adults who were either breast fed for a short period or not breast fed at all, and smaller numbers at longer periods of breast feeding (although numbers are not specifically given in the paper at each time grouping). We now need to interpret the regression coefficients (table 3), actually quite broad at the 95% confidence intervals. The paper tells us that the P values associated with these regressions are just significant, but no mention is made of the r² values that will tell us how much of the variability of arterial distensibility is explained by all variables, including duration of breast feeding. Furthermore, there seems to be no analysis of adults who had been breast fed alone and no r² value.

In addition, for the dichotomised groups, arterial distensibility is compared (t test) with the non-breastfed group, and a similar comparison is made between the dichotomised groups. Although two comparisons are reported, we suggest that these sort of comparisons should be conducted by using one way analysis of variance with appropriate testing afterwards (for example, Bonferroni), or if multiple t tests are used, then the level of significance (presumably set at P=0.05 here) should be reduced to account for multiple comparisons (we suggest three comparisons in this case). The low level of significance reported between the dichotomised groups (P=0.02) is unlikely to survive such conservative statistical treatment. These approaches are more conservative but give us greater confidence in the assertion that some arbitrarily determined time point could be important in determining future risk of cardiovascular disease.

We find little in this paper that will change our current personal habits or advice we give to other parents. We are delighted that the authors agree.

Paul Michael Dark, Medical Research Council clinical training fellow. 
Emergency and Intensive Care Medicine, University of Manchester, Manchester Paul.M.Dark{at}man.ac.uk

Marie-Josée Rölli, retainee, family medicine. 
Greenmount Medical Centre, Bury BL8 4DR

Competing interests: Parents of breastfed infants and children. M-JR is a member of the breastfeeding network.

Authors' reply

EDITOROur paper on the duration of breast feeding and later arterial distensibility evoked much comment, especially from those who promote breast feeding. Unfortunately the media coverage may have deflected attention from the cautious way we framed our findingsand our clear recommendation that they should not change breastfeeding practice.

Some cast doubt on our work by implying that it was motivated or influenced by formula milk companies. This is certainly not so. The study was funded by the Medical Research Council and the university, with no industrial connection whatsoever. Our childhood nutrition centre is core funded by the government (MRC). Our longstanding research includes some of the strongest scientific evidence available favouring breast feeding, in terms of its beneficial effects on cognitive development, blood pressure, bone health, atopic disease, infection, gut disease, and catch-up growth ^{1 2} evidence much used by professional organisations that support breast feeding. We have also researched the efficacy and safety of new advances in infant formula milks and, for transparency, cite this in our article. As an independent centre we publish what we find in the interests of public health, quite regardless of any pressures from either industry or advocacy groups.

Our findings have clearly seemed counterintuitive to many. Dettwyler cites an anthropological argument based on primate work, that humans were evolved to breast-feed for two and a half to seven years, as evidence that our results are biologically implausible. Life span was, however, much shorter when human lactation evolved, and we cannot assume that breast feeding, through past evolution, would now confer any advantage in terms of reduced adult degenerative disease or postreproductive survival. Nor can we assume that breast feeding evolved such that humans would necessarily be well adapted to a modern Western style post-weaning diet. Holmes affirms this view.

Our paper has stimulated comment on interpretation and methods. We agree with Greiner that it is difficult to interpret non-randomised outcome studies on breastfed infants, which of course also applies to the extensive and potentially confounded literature purporting to show benefits of breast feeding. This centre has been one of the only ones to conduct large scale randomised studies on breast milk versus formula in a circumstance in which this is ethicalin non-breastfed premature infants who can be assigned randomly to formula milk or donated banked breast milk. These few studies provide experimental evidence for long term effects of breast milk on health outcomes.¹ When randomisation is precluded (as, say, with smoking), however, causation must be established from a weight of epidemiological evidence, supported by animal experiments. We appraised the possible significance of our own data in such a context, although we accept that the research is at an early stage.

Some respondents imply that we were directly comparing formula feeding to breast feeding. This was not our intention. As Holt noted, formula milks used in the 1970s were different from those currently available, and study of formula fed subjects in our cohort would have had little contemporary relevance. In epidemiological and intervention studies, breast feeding seems to confer cardiovascular benefit over formula feeding.¹ Our interest focused solely on the duration of breast feeding in relation to vascular health in a Western population, in view of previous work we reference.

Our paper considers carefully our surrogate marker of arterial disease, brachial artery distensibility. Wilkinson and Cockroft note that much work on distensibility has been based on the widely used aortic pulse wave velocity. Oddly, their response entirely ignores more recent studies, including this paper, which consistently show an association between peripheral artery distensibility and concentrations of cholesteroland that the various methods for measuring distensibility in central and peripheral arteries are well intercorrelated.^3-6 Simple non-invasive vascular measures, as used in our study, provide unique opportunities to investigate early stages of disease development.

We used a statistically robust approach to data analysis and have been appropriately cautious in our interpretation, taking account of cohort size and significance level. We would reassure Dark and Rölli that the relation between breastfeeding duration and arterial distensibility persists whether analysis is performed on the entire cohort or solely on those breast fed. The r² for distensibility versus length of breast feeding is 0.22, suggesting the model accounts for around a quarter of the variability in distensibility.

Finally, we wish to re-emphasise why we would not suggest any current change in breast feeding practice. Firstly, our data are at too early a stage to be translated into health policy. Secondly, any risk-benefit analysis must include the many positive purported benefits of breast feeding on short and long term outcome.

If the hypothesis we raised proves correct, that more prolonged breastfeeding duration followed by a Western style diet explains our results, then future intervention policy might be better directed to our Western diet rather than breast feeding. We hope that the complex social issues that surround this subject will not cloud the need for dispassionate research to optimise infant nutrition in relation to long term health.

Paul Leeson, research fellow. 
Alan Lucas, professor. 
Medical Research Council Childhood Nutrition Research Centre, Institute of Child Health, London WC1N 1EH

Competing interests: The centre has collaborated with the infant food industry for its outcome studies on nutrition.

Summary of rapid responses

This paper and the accompanying editorial by Booth generated a great deal of heated argument. ^{1 2} We received a total of 51 responses for the two articles, all but eight within the first month of publication. We posted 10 rapid responses on the first day of publication on bmj.com, and over 60% (32) of responses were posted by the time the next issue of the printed journal was published, including a reply from the authors.

All but five of the responses were highly critical of the paper, largely for shortcomings in the methods and because it was funded by a manufacturer of formula milk. Others were concerned about the negative effects on breast feeding resulting from the media's treatment of the results.

Luis Gabriel Cuervo, a member of the BMJ's editorial board, roundly criticised the BMJ in its management of the paper:

"The BMJ has a responsibility not only to publish evidence. It also has to foresee the effect of the published paper on global health and clearly address it. The breach that allowed the media to manipulate the results and jump to the conclusion that breast feeding for more than four months causes cardiovascular disease is inadmissible and will surely be commercially exploited for unscrupulous purposes, here and in the developing world, with terrible consequences. Later explanatory letters may not have the same impact in the media and may not compensate for the damage that has been done."

Three lone voices joined the authors' in the wilderness.

Allan Astrup Jensen, research director of a company in Denmark, thought that "the many critical responses try to kill the messenger because the message is unpleasant and may hurt common health policies. No paper is perfect, including this one. There will always be questions raised and criticism of methods, execution, and reporting."

Andrew Mimnagh, a general practitioner, and Timothy James, a university senior lecturer, were concerned about the demand by some respondents to ban research sponsored by companies as unethical. Mimnagh added: "I agree the finding is counterintuitive but so are many `proven facts' in the natural world."

James was disturbed by the "low level of logic" in some of the responses: "[It seems that] the answer has been predetermined and only evidence that supports that answer is acceptable. This is contrary to the entire scientific approach to truth seeking, which demands that we go wherever the evidence takes us, whether it is where we wanted to go or not." He concludes that drawing "conclusions for our own environment is a complex multifactorial matter, which cannot be summed up in a simple slogan like `breast is best'however unethical the behaviour of sellers of breast milk substitutes."

Sharon Davies, letters editor. 
BMJ