Menu ENN Search
Language: English Français

Urgent: Weight for Height Z-scores: Unisex Z-score table used in West African countries vs. calculated Z score values

This question was posted the Assessment forum area and has 11 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Miriam Chang

Normal user

10 Jul 2015, 05:00

In the 2011 Guidelines for THE INTEGRATED MANAGEMENT OF SEVERE
ACUTE MALNUTRITION published by ACF, Annex 4 depicts a Weight-for-Height table. This same table has been adopted by several National CMAM protocols in West Africa, e.g. Mali and Chad.
The instructions for Weight-for-Height table say to use it for both boys and girls. The footnote explains that “These tables are derived from the WHO2006 standards for Boys. Because using separate tables for boys and girls may lead to many more boys being admitted to therapeutic programs than girls, the use of the boys table for both sexes is recommended to avoid discrimination against female children. …”

Our problem is when we calculate Weight for Height Z-scores using the algorithm and LMS values from WHO 2006 growth standard for boys, the calculated Z-score values do not match the cut-off values in the Weight-for-Height table. For example,
A male child whose length is 75.5cm, weighing 7.5kg, the calculated the Z-score is -3.1 (-3.106 using ENA) but the table says 7.5kg corresponds to Z-score >-3. So this could lead to the child being registered into a different treatment.
We would really appreciate insight into
1) how are the cut off values in the table derived? What are the assumptions behind the table? Could anyone please shed some light on the reasons behind the observed discrepancies between the table and the calculated Z-score?
2) For lengths 45cm to 56.5cm, why are the weights provided in 2 decimal places? In practice, many OTPs and health facilities use hanging Salter scales that only weigh to 1 decimal place.
Thank you very much

Mark Myatt

Consultant Epideomiologist

Frequent user

10 Jul 2015, 17:21

I cannot comment on how the ACF (2011) table was made. The forum administrator should ask ACF to comment. Have you checked the published clinic-use tables on the WHO site? I suggest you use the WHO tables (if you must use WHZ rather than MUAC).

That "discrimination against female children" is an odd thing since this should not be an issue with a well-constructed reference (the WGS has, IMO, limitations but it is well constructed reference). I think it reflects a belief (not necessarily proven) that prevalence of SAM should be the same for girls and boys and we must make this true by building in a pro-female bias. I do not think it a good thing to do.

As for (2) ... it is common practice to measure these young and small children using paediatric scales which tend to yield finer measures (e.g. to 20 g) than hanging scales (which struggle to give 100 g). I think you can simplify the table by removing rows with spurious accuracy.

I hope this is of some use.

Tamsin Walters

en-net moderator

Forum moderator

17 Jul 2015, 13:02

Hello,

We contacted ACF who have replied:
We (Health and Nutrition team in ACF France) checked the Unisex Z-score table used in ACF protocols as referred to in the question and this is exactly the same as the WHO WHZ unisex table - it would appear that this is not an ACF made table.

The question of the digit: this is because, we advise to weigh the very young children with a pediatric scale with a 100g precision.

If you would like to send the table to ACF to check, please send via post@en-net.org

This may be a question for WHO, so we can also try to contact them for input.

Best wishes,
Tamsin

Tamsin Walters

en-net moderator

Forum moderator

23 Jul 2015, 17:06

Hello,

WHO have confirmed that they have not produced a unisex WHZ table. They strongly recommend the use of separate standards for boys and girls and that is how the WHO Child Growth Standards were developed.

Since this problem with the calculations does not seem to be a common one related to the tables, could it be that the issue is with data entry or the ENA function?

Best wishes,
Tamsin

Juergen Erhardt

Normal user

23 Jul 2015, 17:56

Dear Tamsin, ENA always differentiates between girls and boys. Without entering the sex no z-scores are calculated. I hope the ones who prepared the ACF guidelines can solve this problem.

Tamsin Walters

en-net moderator

Forum moderator

23 Jul 2015, 18:10

Many thanks Juergen.

Dear Anon, please let us know whether you have checked the table you are using against the WHO tables, easily located on their website here http://www.who.int/childgrowth/standards/weight_for_height_field/en/

Best wishes,
Tamsin

Michael Golden

Normal user

24 Jul 2015, 00:54

Dear all,
I am responsible for formulating and advocating the UNISEX Weight-for-height table for the admission of children for therapeutic care of severe acute malnutrition in both the ACF protocol and the Generic protocol that has been adopted by several National Governments.

The main reason is that WHO actively discriminates against severely malnourished girls.

I say this because Yvonne Grellety and myself analysed data from children admitted for therapeutic feeding using the old UNISEX table (derived from the NCHS standards). Two critical points came from this analysis.

First, approximately the same number of females and males were admitted using the unisex criteria.

All age female 6,042 , all age male 6,303.
6-59months female 4,375, 6-59months male 4,700

Second, the mortality rate was slightly higher in the females admitted USING UNISEX criteria.
For the 6-59month children female deaths =492/4,375 (mortality 11.25%); male deaths 489/4,700 (mortality 10.40%).
The difference is not significant (chisq = 0.197).

These, to my knowledge, are the only empirical data for admissions using unisex criteria that show the risk of death of the admitted children.

These data give very strong empirical evidence that a child of any given height has the same risk of death at a given weight deficit whether the child is a girl or a boy!

However, the WHO standards -3Z weight-for-height cut-off for girls is at a lower weight (at any given height for children over 6 months) than for boys!

If we apply the WHO standards we mean that a girl has to lose MORE WEIGHT, or be at a lower weight at any given height, to become eligible for admission (using weight-for-height criteria) than a boy even though the girl has the same risk of death as a boy. This is discrimination against girls. I have shown these data to WHO and others - and they have just reiterated in this forum as well as their published recommendations that they will continue to advocate discrimination against girls! I am posting this because of this problem.

We should be clear that the WHO standards have been derived from elite groups of normal children - although the raw data has not been released for independent examination, it is clear that none of the children studied to derive the standards were actually severely malnourished! The derivation of the -3Z is a statistical extrapolation that goes far beyond the limits of the observed data; extrapolation far beyond the observed data is always prone to error. (note that the nominal Z-scores are not actually standard deviations but are derived from centiles [percentiles for our American readers] so that the width of the z-score bands varies and is entirely dependent upon the distribution and skewness of the observed data - thus, as I say, the -3Z cut off is not based upon any data from children within the weight-for-height bands that are used to define severe malnutrition, let alone their risk of death or need for treatment). What I have presented are empirical data that show that it is wrong and discrimatory to define those that should receive treatment for severe malnutrition and those for whom treatment should be denied based upon their gender.

The UNISEX table that I generated from the LMS WHO data is actually simply the boy's standards that are being used for both sexes. This prevents any discrimination against boys by reducing the weight they need to achieve to be admitted (which would occur if one used the mean value), and simply admits girls on the same basis as boys.

It is conservative and if there is ANY DOUBT then the principle of primum non-nocare (first do no harm) must apply, for to do anything else would be unethical.

This table was first used in the National Protocol for treatment of SAM in Yemen after detailed discussion with the Yemeni Ministry of Heath. It has since been adopted by many countries mainly in West Africa - a procedure that was agreed at a regional meeting and endorsed by the delegates of the regional National Government's Ministries of Health within West Africa.

Using the WHO recommendations it is noteworthy that, in general, more boys than girls are admitted when using weight-for-height criteria and more girls than boys are admitted when using the MUAC <115mm criterion.

A second and much less important reason for using a unisex table is that it is much simpler for the staff to use one table than to use two different tables (even with the possibility that some boys may be admitted using the wrong table particularly when it is photocopied in black and white).

I should emphasise that this is for admitting children for therapeutic feeding using weight-for-HEIGHT criteria. It is NOT for monitoring the growth of normally nourished children using weight-for-AGE criteria. IT is at present appropriate to use different growth standards for girls and boys ion growth-monitoring/ read to health charts, at least until we have empirical data on the outcome of those children who fall below the -3Z weight-for-age or height-for-age criteria; I am unaware of any data that addresses the other anthropometric standards.

I hereby urge WHO to change its recommendations and cease discriminating against severely malnourished girls.

Carlos Grijalva-Eternod

UCL Institute for Global Health

Normal user

24 Jul 2015, 10:00

Dear Mike,
Thank you for such detail comment.
Even when assessing the -3 z-score value againts the mean z-score value at any given length/height as percentage of the mean for each sex, there is a systematic difference between sexes. That is, girls wil have to reach slightly lower percentages of the mean that boys. See this link for a graph I quickly put together showing this syetmatic difference.

Intrestingly, your approch does not necesarily remove sex-bias in selection. By utilising the boys cut-off values for both sexes, at most times, you have only shifted the discrimination (see this link).

This brings me to my question on the assumptions that you are following. Why should we expect comparable number for malnutrition and mortality between sexes? In most difficult situations, even for dying, men/boys tend to do worse than women/girls do, for intance infant childhood mortality, homeleness, mental health, etc. Furthermore, even at this early age, there is sexual dimorphism in the accrecion of tissue masses, where boys will favour more lean mass accretion and girls more fat mass. This means that comparable low weights for any given length/height will likely confer a different risk of dying as girls will have more fat mass than boys for any given unit of weight.

Michael Golden

Normal user

24 Jul 2015, 16:59

Dear Carlos,
All you have done is to graph 70% of the median of the WHO standards - as the means are different for boys and girls it is obvious that the differences should be maintained. This analysis neither addresses the central problem of extrapolation beyond the observed data nor, more importantly, the empirical data that I present where there is the same mortality experience of girls and boys who are admitted using unisex criteria - these data are NOT dependent upon standards. They show that when exactly the same critieria are used for admitting boys and girls that the mortality risk is the same.

In the young child there is very little difference in body composition between girls and boys - and, to my knowlege, NO DATA on differences in body composition of severely malnourished children by gender - although such data may become available in the future from the research at Jimma, Ethiopia using air displacement plethesmography.

Carlos Grijalva-Eternod

UCL Institute for Global Health

Normal user

24 Jul 2015, 21:10

Dear Mike,
Thank you for your reply. Alas, it seems that my past comment was not well explained.
To clarify it, I did not graph the 70% of the median of the WHO GS as you state. What I did was to divide all the weight values at -3 z-scores by the mean values, values obtained from the LMS values provided by WHO. In agreement with your initial disclosure the results from this simple division suggested that to reach a -3-zscore value girls would have to present lower percentage of the mean values than boys. Your UNISEX approach, of applying the -3 z-score values of boys to girls just shifted that situation at most times, that is, girls would now have to present higher percentage of the mean values than boys. My graphs aimed at providing you with another way of looking at this issue and they were not aimed to address the issue of extrapolation beyond the observed data nor to present a challenge to the empirical data you presented. To my knowledge, the NHCS references also did not include data from children with acute malnutrition, so I think the problem of extrapolation beyond the observed data applies equally to NHCS unisex tables, as the cut-off values are statistically derived for both growth curves NHCS or WHO, and for unisex or sex-specific tables.

My previous post is more directed at getting more information about the assumptions you are using to interpret your empirical data. You state that using similar cut-off values for weight on any given height/length provides you with similar numbers of children that will be selected for intervention and it also provides you with similar mortaility rates. Going beyond the problem of equating the phenotype of acute malnutrition with that of mortality, I am curious as to why you are interpreting the data you have of using a similar cut-off value, when compared to using a sex-specific cut-off value using the WHO GS) as accounting for the latter discrimination against girls. I think your assumption remains of equality of risk at similar phenotypes and I am questioning that logic. If I were to assume that boys are expected to have greater mortality rates, I would be interpreting your findings as the unisex cut-off of weight for any given height discriminates against boys.

Regarding body composition, you are correct, we have little evidence of body composition at this age, although the one that exists suggest that sexual dimorphism on different tissue masses is manifested early in life; but to my knowledge I do not think that anyone has formally test it. For instance, the data your refer to from Jimma, Ethiopia already shows at an ages as early as the first 6 months of life indication that girls might have greater fat mass for any given height (see table 8). Again, Gregers did not formally tested for differences, but a systematic difference is observable at most centile curves.

Michael Golden

Normal user

25 Jul 2015, 00:16

Dear Carlos,
I am not making any assumptions at all - I did not, a priori, expect to find these results when I did the analysis. But the fact remains that when we admit children, regardless of sex using the same criteria, we end up with the same mortality rate statistically, although slightly higher in girls. I am not speculating about the reasons for this finding - but there are several possiblities. Nevertheless, the data are independent of standards.

You analysis of dividing the -3z by the mean value does not make any sense to me - the divisior for the girls is smaller than for the boys so the results you get are bound to move the lines relative to each other - but to me this division is meaningless/unintrepretable from either a physiological or an operational point of view. I am trying to understand why you would have performed such a mathematical manipulation!

You are quite right that the data that generated the NCHS standards did not include malnourished children - and to get the cut-offs they were extrapolated beyond the data - this is an arguement against using either NCHS or WHO distributions of normal children to derive cut-of points for determing whether a child should be admitted for treatment.

The most rational approach would be one that determined the risk of death for a child with SAM to define the cut-off points. But this will be confinded for ethical reasons to deaths after admission (which I do assume represent the severity of the condition). The analysis of Claudine Prudhon is the only analysis that I am aware of that attempted to do this. Prudhon C Am J Epidemiol 1996 144 116-23; Prudhon C Eur J Clin Nutr 1997 51 771-7

If you use Prudhon's formula (which is based upon mortality of children treated mainly with the WHO protocol) on a weight against height chart you will find that the -3Z WHO line (boys) is similar (but not identical) to the 2% mortality risk; this is far superior to the NCHS -3Z line where children less than 63cm have a >10% mortality. So there is independent justification for using the -3Z WHO criteria WHZ as an appropriate criterion for admission. In other fora I advocated to use the Prudhon's risk of death lines to determine cut-off points without any success (and with modern protocols we should do a lot better than those achieved with the children included in either Prudhon's analysis of Grellety's analysis (her thesis is on ENN archive). Neither author divided the risk of death by gender, the data I presented are a reanalysis of the raw data presented in Grellety's thesis.

I reiterate that I have made no assumptions as to the reasons, you are speculate about mortality risk of all children by gender and body composition - but that is beside the point - the data do not support your speculation or you contention that there should be a difference. given these data it is unethical to discriminate against girls and using the boys standards for both sexs certainly does not disadvantage boys as you seem to contend.

Mark Myatt

Consultant Epideomiologist

Frequent user

2 Aug 2015, 11:40

Just an aside ...

I agree with Mike that the most rational approach to selecting a SAM case-definition for SAM is one that is informed by mortality risk. This will need to look at mortality in both treated and untreated cases since we want to identify cases that would be likely to die if they remained untreated and likely to survive if they were treated. I think that the Prudhon Index was constructed using treated cases only and can, therefore, only provide a partial answer. There are a number of historical cohort studies that provide data regards untreated cases. Reviewing this data provides strong support for the use of MUAC-based case-definitions.

Back to top

» Post a reply