Cook County Hospital, University of Illinois at Chicago, Chicago, IL, USA.
The relationship between an infant's birthweight and survival chances has been recognized for over a 100 years,1 but in this issue of the International Journal of Epidemiology Allen Wilcox has pulled together work from recent decades that brings a clearer understanding of what we know and do not know.2 His refined analysis of birthweight distributions and how they conform, or fail to conform, to the familiar bell-shaped normal distribution has made possible more informative comparisons between different populations than earlier approaches. Incorrect underlying assumptions or inadequate data weakened much prior work in this field. The point of research in birthweight is to reduce perinatal deaths, and a focus of population comparisons has been to eliminate the gross disparities in outcomes between ethnic groups. The Wilcox model of population birthweight analysis has been faulted by critics as, in effect, perpetuating disparities by lending credence to genetic determinism, an ideology advanced in particularly flagrant terms in Murray and Herrnstein's Bell Curve.3 Is that fair?
Birthweight became a part of the US national standard birth certificate in 1950.4 Many countries and US states began maintaining computerized birth files that included this important item in the 1950s and we now have archives going back nearly 50 years. Studies of entire populations across generations are becoming common.5,6 Over this same generation, major changes have taken place in the worlds of politics and science: massive uprisings against racism and colonialism, the transformation of socialism into state (or other) capitalism, the rapid development of scientific disciplines, including, among others, epidemiology and molecular genetics. The elegant mathematics of the normal distribution, first published by De Moivre in 1733, grew out of the study of errors in astronomical measurements, including the work of Galileo.7 However, in the 21st century as in the 17th and 18th, the models developed by scientists exist in a dialectical give-and-take with wider political and intellectual life in society. Then the central issues might have been whether the earth was the centre of the universe, with implications for the role of the king as centre of the ruling class. Today a debate exists over the genetic meaning or non-meaning of race, with implications for the role of relations of dominance on the national and international level. Despite the protests of scientists to the contrary,8,9 science remains part of the culture that spawns it. This implies social and political responsibilities for scientists.
How does the analysis of birthweight distributions relate to the social conflicts of the past 25 years? A central theme in Allen Wilcox's work over this period is the contradiction between birthweight as a powerful predictor of mortality in populations and birthweight as a variable lacking any logical causal link to life and death.10 Mere mass tells the biologist little about an organism's survival potential. Hummingbirds and blue whales have their own adaptations, and size alone, although defining one dimension of diversity, does not imply value. For a given species, however, size measurements can offer hints about past nutrition or present vigour. Size at birth offers additional and special information because it tells us not only the final endpoint of growth but may indicate an untimely interruption of that growth in uteroby preterm birth. This is the main reason why those who care for sick infants or make policy to promote perinatal health concern themselves with birthweight.
Obviously simply studying gestational age would be more straightforward. After all, most life-threatening newborn conditions result from shortened time in utero. For example, respiratory distress syndrome, the leading single pathological entity leading to early newborn death, results from birth prior to the biochemical maturation of the lung.11,12 However, since the timing of ovulation and conceptionor even of the last menses before conceptionis often in doubt, studies based on gestational age end up excluding or misclassifying around 5% of births.1315 Perhaps a missing data rate of a few per cent seems trivial in a dataset of millions, but the amount of missing and misleading gestational age data is about 50 times more than the amount of missing birthweight data, which is less than 0.1%.13 Both must be considered in the light of mortality rates in the range 0.52%. Clearly, missing data on the wrong births could have a major impact on the outcome measure.
Indeed, missing data on the wrong births is exactly what happens. That is because information on gestational age is not absent or erroneous in a random subset of the population. Those women who have fewer years of education or give birth in their teens or receive inadequate prenatal care are also more likely to have incorrect gestational ages (or none at all) recorded on their infants' birth certificates.13 These are exactly the babies at greatest risk for death. Having poor prenatal care and limited education are markers for social and economic stressors such as exposure to racism. Racism in turn is associated with preterm birth.16,17 The oppressed segments of the population are least well tracked for problematic variables like gestational age. Science fails again to escape the bonds of class society. But birthweight is collected accurately for almost every in-hospital birth. Could this offer the escape route for epidemiologists who want to do useful and unbiased analyses without having to first establish equity in health care?
The approach seemed promising, but investigators needed a way to use birthweight, with its the clean and complete data, to draw inferences concerning the rate of preterm birth in populations. The use of the 2500-g cut-point is crude at best. The logical uncoupling of distinct components of the birthweight distribution, in analyses first published by Wilcox and Ian Russell in this Journal in 1983,18 provided researchers with more powerful conceptual tools for understanding birthweight and mortality. The independence of mean birthweights and preterm birth rates (as reflected in the residual distribution) has been supported by the work of numerous investigators studying different populations in various countries15,1923 or the same population over time.6 Some have duplicated the Wilcox-Russell analysis while others modified it slightly, such as by use of z-scores for birthweight.
One aspect of this useful approach remains problematic. The association of mortality most strongly with the proportion of residual births could imply that the other births are normal in the physiological sense as well as the statistical (i.e. Gaussian) sense. At a population level this has been interpreted to mean that some groups of babies are just supposed to be small. As long as we are discussing individuals living at different altitudes, this is not a problem, but when we start talking about infants of different ethnic groups, concerns develop that real social problems may be defined away rather than vigorously opposed. This is the source of concerns about implied genetic determinism.3 In the context of a society whose dominant elements justify their positions by arguing the genetic inferiority of those they dominate, it is hard to be neutral. In the pursuit of pure science a well-meaning investigator may be perceived asand may beaiding and abetting a social order he abhors.
In fact, evidence indicates that normal birthweight can change for a given ethnic group over time.6 In the US, white and African American infants both showed increases in mean birthweight over a generation, but not in equal amounts. Black infants did worse, paralleling their disadvantaged position in the American racial hierarchy. Moreover, secular change in births at the extreme low weight end of the curve actually increased for black infants, a trend opposite to their improving mean birthweight, and a change not experienced by whites.6
Basically Wilcox is scientifically right: chasing the mean can be misleading, even if most of the time the mode and the tail of the birthweight curve move in tandem. Keeping these outcome parameters distinct will produce improved analyses. The most efficient approach to eliminating racial disparity in perinatal deaths will be targeting the causes for excess preterm births. Many useful studies remain to be done to assess the impact of the social, psychological and environmental factors that underlie preterm birth, and use of a z-score approach will contribute to their clarity and precision. However, in applying an analytical technique to real people, to people living in real societies, to people experiencing births and deaths in their families, the scientist must always proceed with an eye on the social, historical and political context in which those real peopleand the scientistfind themselves.
References
1 Cone TE. Perspectives in neonatology. In: Smith GF, Vidyasager D (eds). Historical Review and Recent Advances in Neonatal and Perinatal Medicine. Chicago: Mead Johnson Nutritional Division, 1983.
2
Wilcox AJ. Reformulations: on the importanceand the unimportanceof birthweight. Int J Epidemiol 2001;30:123341.
3 Muntaner C, Nieto FJ, O'Campo P. The Bell Curve: on race, social class, and epidemiologic research. Am J Epidemiol 1996;144:53136.[ISI][Medline]
4 US Department of Health, Education and Welfare. Vital Statistics of the United States, Vol. I. Washington: US Government Printing Office, 1954, p.20.
5
Barker D. Fetal origins of coronary heart disease. Br Med J 1995;311:17174.
6 Chike-Obi U, David RJ, Coutinho R, Wu S. Birth weight has increased over a generation. Am J Epidemiol 1996;144:56369.[Abstract]
7 Patel JK, Read CB. Handbook of the Normal Distribution. New York: Marcel Dekker, Inc., 1996.
8 Rothman KJ, Adami HO, Trichopoulos D. Should the mission of epidemiology include the eradication of poverty? Lancet 1998;352:81013.[ISI][Medline]
9
Zielhuis GA, Kiemeney LALM. Social epidemiology? No way. Int J Epidemiol 2001;30:4344.
10 Wilcox AJ. Birthweight and Perinatal Mortality [Dissertation, University of North Carolina]. Ann Arbor, MI: University Microfilms International, 1979.
11 Binkin NJ, Rust KR, Williams RL. Racial differences in neonatal mortality: what causes of death explain the gap? Am J Dis Child 1988; 142:43440.[Abstract]
12 Notter RH. Lung Surfactants: Basic Science and Clinical Applications. New York: Marcel Dekker, Inc., 2000.
13 David RJ. The quality and completeness of birthweight and gestational age data in computerized birth files. Am J Public Health 1980;70:96473.[Abstract]
14 Mustafa G, David RJ. Comparative accuracy of clinical estimates versus menstrual gestational age in computerized birth certificates. Public Health Rep 2001 (in press).
15 Buekens P, Wilcox AJ, Kiely J, Masuy-Stroobant G. Birthweight, preterm births and neonatal mortality in Belgium and the United States. Paediatr Perinat Epidemiol 1995;9:27380.[ISI][Medline]
16 David RJ, Collins JW. Bad outcomes in black babies: Race or racism? Ethnicity Dis 1991;1:23644.[Medline]
17 Collins JW, David RJ, Symons R, Handler A, Wall SN, Dwyer L. Low-income African-American mothers' perception of exposure to racial discrimination and infant birth weight. Epidemiology 2000; 11:33739.[ISI][Medline]
18 Wilcox AJ, Russell IT. Birthweight and perinatal mortality. I. On the frequency distribution of birthweight. Int J Epidemiol 1983; 12:31418.[Abstract]
19 Chen R, Wax Y, Lusky A, Toppleberg G, Barell V. A criterion for a standardized definition of low birthweight. Int J Epidemiol 1991;20:18086.[Abstract]
20 Carlson E, Hoem JM. Low-weight neonatal survival paradox in the Czech Republic. Am J Epidemiol 1999;149:44753.[Abstract]
21 Wilcox A, Skjaerven R, Buekens P, Kiely J. Birth weight and perinatal mortalitya comparison of the United States and Norway. JAMA 1995;273:70911.[Abstract]
22 Olsen SF, Olsen J. A birth-weight adjusted comparison of perinatal mortality in the Faroe Islands and Denmark. Scand J Soc Med 1994; 22:21924.[ISI][Medline]
23 Gruenwald P. Fetal growth as an indicator of socioeconomic change. Public Health Rep 1968;83:86772.[ISI][Medline]