Contents - Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
I recognize that the protocol for your study was rigorous, but I was a little confused and concerned because of so many interacting variables. You studied 40 variables, but I want to know how some were excluded and others were retained as important variables to make this statistical analysis conclusive. Thank you for your lecture and expert use of statistics.
I am not a statistician, but I think our study demonstrates the power of regression analysis techniques. They have made valid and robust statistical analyses of this kind possible, which has been a very important step forward in statistical methodology.
I would like to ask one question about what is generally regarded as a very elegant study. When you discussed the instance of gastrointestinal illness from 0 to 13 weeks, you said that receiving colostrum had no impact (or appeared to have no impact) on reducing the incidence of infection. Nevertheless, when you came to 40 to 52 weeks, the early weaners had comparable rates to the partial breastfeeders and seemed to be distinct as a group from both the bottlefeeders and the full breastfeeders. Have you any way of accounting for this disparity at 40 to 52 weeks? Might that be due to a failure to control for another confounding factor at that age?
It is possible that what you say is correct. Some differences appeared to emerge later in the first year, and it is possible that colostrum is having some effect. It does seem, however, to be a marginal effect rather than a dominant one. All the data were subjected to the correction of confounding variables throughout the study, so the differences that you see were the true, observed, and corrected differences.
I was just thinking about the additional confounders that might apply at a later stage, such as contact with other infants or toddlers from toddler groups.
Contact with other children was one of the points that we specifically included as a variable. If older children are coughing over the infant, this is obviously a very important variable and was a point on which we collected specific information that was included in the analysis.
Dr. Colombo said it is a very elegant study and very nicely done, but I am not sure if I know what the mother craft variable is, and that may answer my question. I would like to point out, however, that the nature of your confounders is not the same as the nature of the behaviour in your study. In other words, the confounders that you have chosen are primarily structural variables, whereas breastfeeding is really a process variable. It could very well be that there are confounders at the process level that could perhaps account for a little bit more of the variance of the differences between groups.
As to the first point: mother craft is parent craft. This was measured as attendance at classes during the antenatal period giving advice about preparation for motherhood and labour and so on. This variable was very strongly related to social class. If you correct for social class, you would at the same time correct for the mother craft variable. You are absolutely right that you can only do the best with what is quantifiable information when allowing for confounding variables. Studies of this kind will always be open to the possibility that we have not identified some confounding variable that may be important but not measurable. Short of being able to do a truly randomized controlled trial (which is not possible for breastfeeding), the best you can do is to collect and measure as many of the direct and indirect confounders as possible. You are right that despite all our efforts, we still may have failed to capture one or more of the crucially important confounding factors.
I am not trying to sound overly critical. The point I am trying to make and with which you agree is that in these kinds of studies you can never control for all the potential confounders. The best model would be a clinical trial. What I do want to emphasize is that unless we get information at the process level, particularly in interpersonal communication or relationships, we will be missing a very significant portion of the pie. I can give an example of the process variable. In the case of breastfeeding, this is the dynamic interaction between the mother and the child that involves certain kinds of behaviours that result in feeding. In the same way, there may be other interpersonal aspects of the relationship between the mother and the infant that do not have to do with feeding but could be correlates of it and account for the outcome variable. You are absolutely right that the ideal is a clinical trial, but we cannot always do that.
I think cigarette smoking is a process variable in the sense that you are talking about, and that was an important confounder. Many of the important variables are associated with social class and disappear when it is controlled for.
These are the normal difficulties of experimental surveys. You cannot always do experiments because of ethical problems.
With your early weaners, what did you do to correct for the possibility that illness might cause the end of breastfeeding? You had data on those early weeks.
Yes, we did look at illness causing early weaning. If we thought that was the case, then the illness was attributed to the feeding category they were in before they had the illness. We looked at that specifically, but in fact it was relatively uncommon. In the analysis we adjusted both ways, for if we assigned mothers who weaned during an illness to the breastfeeding or the weaned groups, this made no difference to the overall conclusions. That was a point raised in the Bauchner study  and we did try to take account of it.
I, too, want to congratulate Professor Howie for a very elegantly carried out study. The exchange with Dr. Pollitt reminded me of a plea that an American politician once made of a scientist. He said he was looking for a onearmed scientist to testify before Congress, because scientists always said, "On the one hand you have this, and on the other hand you have that." The politician wanted to know if we couldn't find the one-armed variety, so that we could be as unambiguous as possible.
The key word that Professor Howie used was breastfeeding, which does refer to a process, as opposed to a single biological variable such as human milk. My second point is that there are some historical studies that came to some very comparable conclusions, without the degree of rigour that your study had, particularly looking at the persistence effect. This suggests that maybe we are looking at something beyond breastfeeding to mother craft and other similar variables. On the other hand, there are some biologically plausible explanations that I think Professor Hanson will get to.
Another point I want to make concerns HIV. I think HIV is a terribly important issue, and I want to ask if there is anyone here who has more recent information. The most recent information I have is that the P24 antigen, which is a good indicator of replicating virus, has been found only in colostrum. Thus, if vertical transmission occurs, it may occur only in the colostral phase. This may be because there are so many cells in colostrum, and HIV is an intracellular virus. Although antigens have been identified in the cell-free medium, I don't know whether these have any significance for the nature of transmission.
If that's true, that is a very important point. I wasn't aware of it, and I am sorry I can't comment.
I don't have as many answers as I would like to have, but there are a number of risk factors that we know about. The main risk is that of infection of the infant via the mother's blood at delivery. There is clear evidence that this does occur. The second risk situation is when an infected lactating mother transmits the virus to the baby in the milk, despite the protective factors in human milk. The statement of WHO is partially based on the fact that if breastfeeding were not practiced in areas of Africa, there would be an estimated 2,000 deaths from diarrhoea for each death from AIDS. This is a difficult matter, and there are two sides to it. I wasn't aware of Dr. Garza's point, and it is very important. However, African studies show that acute infection during lactation may be transmitted later than the colostral phase. This remains an open question.
Coming back to your study, I am really impressed with the sustained effect of breastfeeding against infection. I am also curious about how you reached the conclusion that there weren't any further advantages to breastfeeding for more than 13 weeks. Surely if babies continue to breastfeed, they will be better protected.
From the hospital admission data it looked as though there was a step down in the advantages after 13 weeks. We reanalysed for breastfeeding lasting only six weeks, but this seemed to be too short a time to establish maximal benefit. On the other hand, it looked as though the benefits of feeding beyond 13 weeks fell for both disease incidence and hospital admission. The data do show that the feeding method in the early weeks has long-term effects. I don't know if Professor Hanson is in a position to comment on this, but I believe there is some evidence that the first 13 weeks is a period of immunologic vulnerability. I have heard of the concept of the gastrointestinal tract maturing at about 13 weeks, so that before this time the baby is particularly vulnerable to disease. It may be that protection during this time is particularly important, but I am not as expert in this as some other people here. We can only report the results as we found them.
We are doing a similar study. Infants were breastfed until an illness developed. Up to the age of six months, there was a real difference in the rate of infections between exclusively and non-exclusively breastfed infants.
I have a comment on the HIV problem. I was part of the WHO Working Group that drafted that resolution. As Professor Hanson mentioned, we had to calculate the number of deaths that would be prevented by avoiding breastfeeding and thereby reducing HIV transmission, versus the number of deaths that would be prevented by continuing breastfeeding. The initial problem is that in Africa you don't know which mothers are HlV-positive, because there is no widespread screening. This means that any kind of recommendation against breastfeeding would affect both HlV-positive and HlV-negative mothers. It is a major reason for concern, but with the current state of knowledge, the risks of HIV transmission through breastfeeding are much smaller than the risks of not breastfeeding, so that's why we arrived at the conclusion in favour of continuing breastfeeding.
I certainly support the conclusion that the WHO came to, but as I am sure you are also aware, the concerns about HIV are being used by the milk manufacturers, particularly in Africa, to introduce a significant fear factor to the mothers, advising them to abandon breastfeeding. As you also say, this may lead to greater detriment than any benefits they may achieve. We have to be fully aware of this fear factor and make our views clear to the mothers.
Were there enough breastfeeders among the low-income mothers to enable you to analyse separately the impact of breastfeeding on the occurrence of disease in this group? This is important, because they are so often our targets.
There was a deficiency in the number of breastfeeders in the lowest economic groups, but the benefits were striking in both the upper and lower socioeconomic groups. The main point of our study is that there has been a perception that advice in favour of breastfeeding is all right for the poor people of developing countries but is really irrelevant to the United States and Europe. What we found shows that this is not true at all. It puzzles me why we have an 18% vomiting and diarrhoea rate within the first three months in a place like Dundee, where we have clean water supplies. If we say that unclean water is the key issue, then Dundee should have far lower infection rates than Manila and other similar places. The fact that rates of gastrointestinal disturbance among bottle-feeders are so similar across cultures makes me wonder whether we are really observing infection. Alternatively, it may be some disorder of the gastrointestinal tract other than infection that is actually making the babies vomit and have diarrhoea. Maybe we haven't yet identified the true nature of the problem of early vomiting and diarrhoea.
You might find a lot of rotavirus infection, which is widespread in both developed and developing countries and has nothing to do with water supplies.
We did a study that looked particularly at lowincome populations in Houston about 15 years ago.
We never published it, because we weren't confident that we could deal with some of the issues that were raised, but we found very substantial differences in diarrhoeal and respiratory infections in babies of low-income women who were breast- or bottle-feeding. We controlled for socio-economic status in that we recruited only women who were delivering at our public assistance hospital. But when we looked at our two groups, we found that despite our efforts to control for socioeconomic status, the women who elected to breastfeed were taller and had infants with higher birthweights than the women who elected to bottlefeed. This suggests there is some inherent difference, even in a group of homogeneous socioeconomic status, between women who choose to bottle-feed and those who choose to breastfeed.
Contents - Previous - Next