When can we use BMI for lactating women again?

I find this to be a rather odd question? I do not understand why we would want to use BMI in our work other than as an act of venerating our ancestors.

BMI is a very old (c. 1830 and revived in 1972) indicator. It is also a flawed indicator in that it is strongly affected by body shape (requiring correction) and can identify very fit people as being very ill (e.g. marathon runners are starving, rugby forwards are morbidly obese). It is, at best, only a vague means of estimating adiposity in sedentary individuals.

The mathematical model behind BMI:

    BMI = weight / height^2

is odd since mass increases with the cube of linear dimensions. The result of using a poorly specified model is that taller individuals with the same body shape and composition will have higher BMIs. Short people will appear thinner and tall people will appear fatter. This is not what we want.

The relation between BMI and health outcomes is confused. For example, a recent (2006) meta-analysis showed that people with "normal" BMIs had higher risk of cardiovascular disease than people with "overweight" (25-30) BMIs. I "cherry picked" this study. Other studies have found the same results and other studies have found different results. When we see results from studies that go in both directions we suspect that we are seeing variation around no effect and / or we have a flawed indicator.

The use of BMI seems to have cerated what is called the "obesity paradox" in which person with "overweight" or "obese" BMI do better than those with "normal" BMI (as we have above). In many cases nobody is quite sure what is going on in these studies. Much can be explained by inadequate control for confounding (mostly smoking - smokers tend to be lean), reverse causation (i.e. weight loss associated with severe illness reduces BMI), and stratification collider bias. It is of note that this "obesity paradox" is usually found in studies in which BMI is the sole measure of adiposity and is seldom found in studies using other measures of adiposity. This suggest that the use of BMI is problematic.

This is not touching the issues with pregnancy and lactation. These have been discussed here.

The general scientific consensus is running against BMI. It has fallen out of favour for many applications. There are a number of alternatives to BMI ...

Waist circumference, waist to hip circumference ratio, and waist circumference to height ratio work well as measures of adiposity.

The Corpulence Index (AKA Ponderal Index):

    CI = weight / height^3

does not have the flaws of BMI regarding problems with short and tall people and has better PPV in athletes. Variants exist. This form:

    CI = 1000 * weight^(1 / 3) / height in cm

scales to have 20 to 25 as the normal range. Not the CI has a better specified mathematical model. Some workers have suggested modifications using fractional powers between 2 and 3 to better correlate CI with percentage fat mass.

A Body Adiposity Index has been proposed:

    BAI = 100 * [hip circumference / (height * height^(1 / 2))] - 18

is strongly correlated with percentage body fat from DEXA (at least in African-American populations).

Most of these indices address the issue of excess adiposity (this is the reason BMI was resurrected in the 1970s). We often want to look at the other problem (i.e. malnutrition). For this application we can (and do) use MUAC in the same way as we use it in children (i.e. we apply simple threshold to raw measurements). This is not strongly affected by height or body shape and is associated with poor health outcomes. There is also no paradox of high muscle mass being associate with survival. I think it does what we need cheaply, quickly, simply, and at high coverage.

Why use BMI? What is it that I am misunderstanding?

Dear Mija,
Dear Mark,

The key questions to address when choosing a criterion to identify malnourished patients are:

What is the associated risk?
What intervention you are planning for malnourished mothers?
Will detected women benefit from the intervention?

Mija, I assume that you want to detect women in need of food supplement, but I am not clear about the expected outcome. Do you want women who get the food supplement to put on weight? In this case, are you sure they will benefit from this weight gain? Do you want to increase breast milk production by food supplementation, although the evidence in favour of this is very weak? Do you want to improve the well being of supplemented mothers? In any case, not clear to me that BMI is the best index to select women who will benefit from supplementation based on any of these criteria. Do you have any evidence in its favour? If you don’t use BMI in the 6 months following delivery for the reasons you explain, maybe better not to use it as well for the next 6 months.

I do believe that there is a mathematical rationale behind the BMI. Contrarily to what Mark says, BMI is roughly independent of size. This can be shown by regressing log weight against log height in any adult population. Usually the slope of the regression is about 2, which means that weight is related to height squared (and not to its cube). This can be explained by different body shapes in short and tall people. And this has been known since the 19th century.

Despite this nice mathematical relationship, I agree with Mark that for many outcomes BMI has been found to be less effective than other indices. For example, for the assessment of the risk of heart disease, waist circumference, and its ratio with hip circumference, which better reflect central fat, has been found by some more relevant than BMI. See the comment below on this topic:

Kragelund C, Omland T. A farewell to body-mass index? Lancet. 2005 Nov 5;366(9497):1589-91.

Mija, my suggestion at this stage would be to clarify what intervention you have in mind for malnourished mothers and which outcome you are aiming for. And to check from available evidence that for these objectives BMI is really the best choice to identify women who will benefit from the intervention.

I hope this helps

Dear Mija,

Thanks for the clarification. If I understand correctly, you are planning a hospital based case control study. This is fine. However, if you want to extrapolate your findings to the community, better to take controls in the community as well, else you will have a bias and may find spurious associations due to different treatment seeking behaviours (Berksonian bias).

Case-control studies usually have small sample sizes and you can choose a complicated indicator of nutritional status such as BMI. However, for simplicity reasons, and also extrapolating what has been found for other outcomes, I would go for MUAC. This also has the advantage of being suitable for pregnant lactating, non pregnant, non lactating women. If you are not sure about what to do, you can also collect mother’s MUAC, weight, height and BMI and see which of these indices is the most closely associated with SAM in the child. You could even assess the association of mother’s weight, height and BMI with different descriptors of nutritional status of the child, ie MUAC, weight-for-age, weight-for-height and height-for-age. This would be an interesting small research project. I am not aware this has been done in the context of SAM and will answer the question you ask.

I see your intention. You could use BMI for this. It would, I think be best to use raw BMI rater than to classify using BMI cut-points. Two reasons for this are (i) sampel size (when you classify you throw away information) and (ii) functional meaning of BMI cut-points varies from place to place due to body shape issues. Thta last point is, I think, key. You cannot definitively say that a woman is malnourished using internationally agreed cut-points (noe exist). We faced this issue in Samalia in 1992. We found and treated men and women with BMIs classified as incompatible with life. Your question is difficult to answer using BMI. I think the best you can do is to see if BMI is similar (or different) in the two groups. Do not despair. If the BMI in one group is lower than in the other group then there will be (at any cut-point) more malnutrition in the first group. Why not also use MUAC? You could use MUAC and TSF to look at fat and muscle mass.

You need to beware of potential bias / confounding when using BMI. If (e.g.) there is a systematic difference in ethnicity or livelihood between the two groups than this could affect body-shape which would affect BMI. You could control for this in analysis (this would need a bigger sample size). A better correction might be to adjust BMI for body shape using the Cormic index.

BTW, I would never use BMI alone to diagnose adults in general because of the problems associated with it.

I hope this helps.

André,

See here, here, and here.

I also liked this IJE article which covers some of the issues.

This article does support the use of squared height.

I do not think the use of BMI is as uncontroversial and many in the nutrition community believe.

I would not have agreed to this in 1995. Our experiences with using BMI in the 1992 Somali Famine to decide admission to therapeutic feeding for adults were universally bad. BMI is known to be of limited use for individual diagnosis (the application in the 1992 Somali Famine) without using additional metrics but it may have some value for population studies. Using BMI for population studies (surveys) is problematic since prevalence estimates that do not translate into need / burden / caseload estimates.