Influence of variation in birth weight within normal range and within sibships on IQ at age 7 years: cohort study
BMJ 2001; 323 doi: https://doi.org/10.1136/bmj.323.7308.310 (Published 11 August 2001) Cite this as: BMJ 2001;323:310All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
EDITOR, To say this study is ‘fatally flawed’ is to take a very black
and white view of research. I see this study as part of a wider body of
evidence; evidence which the authors review. Obviously it is difficult to
measure ‘intelligence’; any measure will be less than perfect. The
measure used could well have a racial bias, leading to the difference
noted by Mr Lwegaba.
However the authors concern themselves with the question of whether
the link found by others between low birth weight and IQ extends into the
normal range of birth weights. This question still seems relevant even if
the US population has changed since the data were collected. And in fact
it might be easier to test for a relationship between birth weight and IQ
using these data than using data collected today. I rather suspect that
the US population is now not only ethnically more diverse, but its family
composition is more diverse and families are more mobile than in the
fifties and sixties. It would, I suspect, be more difficult now to collect
high quality data and then adjust that data to a common family social
environment. Having said that, with this data the authors can only show
that there was a relationship between birth weight and IQ in the past;
whether there still is, is another question. It seems plausible to me but
isn’t proven.
The sample used may well be biased towards the white and well-off.
Analysis of covariance poses the question, if all covariates are held at
average values, does birth weight have any influence on IQ? Now if you
think that if these average values are not correct (because of sampling
bias) and that at the correct values there would be a different
relationship between birth weight and IQ then you have a problem. If you
knew more about the US population at the time the data were collected, you
could adjust for sampling bias. You would then estimate the relationship
between birth weight and IQ with covariates held at population (rather
than sample) average values. I’d be surprised if it makes much difference.
Jim Young
Basel Institute for Clinical Epidemiology,
University Hospital Basel,
Switzerland
Competing interests: No competing interests
Dear Editor,
The birthweight and IQ study of Matte, Breshnahar, Begg and Susser is
fatally flawed. No valid or reliable IQ instrument exists that accurately
measures cognition below age 6, the Bayley Scales most often
used for this purpose. Without an objective pretest, no conclusions can
be drawn linking cognitive measurement and birthweight. In addition, the
old WISC manual from 1949 was cited in the bibliography, which has not
been used since 1990. Much litigation has occurred concerning its racial
and ethnic bias.
I am also concerned about the sibship sample. Even though
multivariate adjustment was made, more white mothers from higher
socioeconomic groups were used, thereby suggesting a sampling bias. Given
the dramatic disparity between the sample used, the advances in
neuroscience since 1959-1966, and the changed racial and ethnic diversity
in America, it is unclear as to the basis used for publication of this
paper. The topic of low birthweight and future outcomes for children is
too important to be left only to those in public health, but must also
include educators.
I have worked with and examined the files of approximately 24,000
classified children since 1977.
There is no question that poverty and malnutrition are critical factors in
outcome, but often changed with early intervention. I strongly suggest
that these authors read From Neurons to Neighborhoods, Shonkoff and
Phillips, National Academy Press, 2001.
Sincerely,
Marilyn Arons
President & CEO
Melody Arons Center for Applied Preschool Research and Education,
210 Carlton Terrace,
Teaneck, New Jersey 07666 USA
Competing interests: No competing interests
It is possible that the perinatal factors considered in the
references 11 and 12 includes perinatal asphxia. It is not mentioned in
the methods for selection of participants, therefore I wish to know if it
was considered a possible confounder as a covariate of both birth weight
and IQ. The IQ being affected by brain damage even though there might not
be clinically neurological deficit?
Are there differences in the descriptive data,(not included in table
1) maternal age, maternal education, birth weight, etc, that can explain
why white race have IQ 102.4 and other races have 89.5?
Regards.
Anthony Lwegaba.
Competing interests: No competing interests
Use a single model for both sexes
EDITOR, In discussing the relationship between birth weight and IQ,
Matte and colleagues note that the ‘difference between boys and girls is
puzzling and needs replication’ [1]. I suggest that they first confirm
this difference between the sexes using a more orthodox statistical
analysis.
There is no need to fit separate models for each sex just because
‘the distribution of birth weight differs between boys and girls’. A
single model for both sexes will still give valid estimates of the
relationship between birth weight and IQ. Senn [2, section 3.9] explains
this and gives an example; in his example replace verum and placebo with
say low and high birth weight. A covariate for sex is needed of course, to
adjust for any differences in IQ at age 7 between the sexes.
Indeed it is not appropriate to test for an interaction between sex
and birth weight unless sex is already included in the model. It seems
quite plausible that average IQ at age 7 could differ between sexes. This
must be adjusted for, before one can test whether the relationship between
IQ at age 7 and birth weight differs between the sexes. The authors may
have done this; it’s hard to tell. The ambiguity would not arise if they
had used a single model for both sexes, with sex as a covariate.
Finally it should be possible to fit a single model to the full
sibship sample. A single model, rather than four separate models (for boys
and girls within both one and two sib samples), would have more power to
detect differences and would be easier to interpret. In addition families
that could contribute sibships of both sexes would not need to be dropped
from the analysis to preserve independence. An analysis of covariance with
mother as a random effect might be a reasonable model; observations on
children with the same mother are then correlated [3, section 2.1]. Some
additional covariates might be appropriate when modelling the full sibship
sample: perhaps the number of sibs in the family (two or more) and sib
composition (mixed or same sex). If there is a statistically significant
interaction between sex and birth weight, a plot of IQ against birth
weight showing the different slopes for the two sexes (over the observed
range of birth weights) would help the reader to see the nature of the
interaction and to judge whether it has any practical significance.
Jim Young
Basel Institute for Clinical Epidemiology,
University Hospital Basel,
Switzerland
[1] Matte TD, Bresnahan M, Begg MD, Susser E. Influence of variation
in birth weight within nromal range and within sibships on IQ at age 7:
cohort study. BMJ 2001;323:310-4.
[2] Senn S. Statistical issues in drug development. Wiley 1997.
[3] Longford NT. Random coefficient models. Oxford University Press, 1993.
Competing interests: No competing interests