AMERICAN JOURNAL OF HUMAN BIOLOGY 15:667–680 (2003) Timing Is Everything: A Reconsideration of Fetal Growth Velocity Patterns Identifies the Importance of Individual and Sex Differences MICHELLE LAMPL1* AND PHILIPPE JEANTY2 1 Department of Anthropology, Emory University, Atlanta, Georgia 2 Women’s Health Alliance, Nashville, Tennessee ABSTRACT Fetal growth has been posited to follow a ‘‘timing hypothesis’’ sequence in which the second trimester favors a single growth velocity peak in body length and the third trimester accommodates a single growth velocity peak in weight accrual. To our knowledge, this proposition has never been tested with high-frequency longitudinal ultrasound data from normally growing human fetuses. The present study examined whether fetal growth in leg length had its peak velocity at or about 20–26 gestational weeks and declined subsequently and whether estimated fetal weight velocity was maximal at or about 33 weeks and declined subsequently; if the greatest acquisition of leg length occurred in the second trimester and weight in the third trimester; and if birth outcomes reflected these relationships. The data in this study included approximately weekly longitudinal ultrasound data collected from 44 maternal/fetal pairs in Brussels, Belgium. Diaphyseal lengths of the femur and tibia provided information on leg growth and estimated fetal weight was assessed from the biparietal and occipital-frontal head diameters and transverse and anterior–posterior diameters of the abdomen. Growth patterns were investigated from individual growth curves derived from daily growth velocity z-scores. Paired t-tests compared individuals’ trimestral increments in leg length and fetal weight. Least-squares regression models employing the robust procedure for repeated measurements were used to test for relationships between trimester, size, growth rates, and birth outcome, controlling for day of measurement, sex, maternal smoking, and gestational age at birth. The normal fetuses in this study grew by pulsatile patterns of leg and estimated weight acquisition, not a single peak and decline process. Greater incremental growth in estimated fetal weight occurred during the second trimester and leg length in the third trimester. Individual and sex effects were significant in growth velocity patterns. Girls grew with greater synchrony between leg and weight growth and were accelerated by comparison with boys, with faster leg growth predicting lower ponderal index by the second trimester. Birth outcomes were sex-specific in timing effects and predictive variables. These results support the importance of sex-specific analyses, reemphasize the common notion that girls grow faster than boys, and direct attention to cross-talk between energy # 2003 Wiley-Liss, Inc. resources and growth. Am. J. Hum. Biol. 15:667–680, 2003. Common assumptions regarding the timing of fetal growth in body dimensions posit that the primary time for growth in body length and weight are sequential, with the second trimester focused on length and the third trimester devoted to fetal weight gain. This notion was formalized in 1982 as the ‘‘timing hypothesis,’’ presented as an interpretive schema for understanding causal relationships between prenatal insult and neonatal body size and proportion (Villar and Belizan, 1982). This model proposed that a better understanding of intrauterine growth retardation follows from an appreciation of developmental timing. Specifically, it was suggested that neonatal size and body proportions could themselves provide clues as to when in fetal development a growth ß 2003 Wiley-Liss, Inc. insult had occurred. First trimester insults were said to produce proportionately small neonates, second trimester insults resulted in short neonates, and third trimester insults were assumed to produce long and thin neonates. The evidentiary base cited to substantiate the fetal growth timing propositions were two graphs originally published by Tanner (1978), illustrating a peak in fetal growth in crown–heel length at about *Correspondence to: M. Lampl, 1557 Pierce Drive, Department of Anthropology, Emory University, Atlanta, GA 30322. E-mail: mlampl@emory.edu Received 4 March 2003; Accepted 5 May 2003 Published online in Wiley InterScience (www.interscience. wiley.com). DOI: 10.1002/ajhb.10204 668 M. LAMPL AND P. JEANTY 20 gestational weeks of age and a peak in growth in fetal weight at about 33–34 weeks, followed by a decline. These illustrative images of fetal growth rate are reproduced in Figure 1. The timing hypothesis assumptions have been questioned several times by empirical data: In a study with approximately monthly ultrasound assessments from 17 weeks, no distinctive growth patterns in femur length (used as a proxy for body length) or abdominal diameter/circumference (used as a proxy for fetal weight) were found by trimester, and it was concluded that growth retardation in both symmetrical (body and head proportionately reduced) and asymmetrical (body reduced proportionately greater than head circumference) intrauterine growth retardation (IUGR) started in the second trimester (Vik et al., 1997). Larger studies previously concluded that body proportions Permission to reproduce this figure electronically was not obtained from the original publisher Fig. 1. Fetal growth velocity in crown–heel length (above) and body weight (below) as per Tanner (1978, 1990). Reprinted by permission of the publisher from FETUS INTO MAN: PHYSICAL GROWTH FROM CONCEPTION TO MATURITY by J. M. Tanner, pp. 39, 41, Cambridge, Mass.: Harvard University Press, Copyright # 1978, 1989, by J. M. Tanner. at birth and growth retardation do not meet the timing hypothesis expectations as outlined above (Kramer et al., 1989) and are not informative about prenatal growth rate (Petersen et al., 1992). Investigations of the timing effects of maternal weight gain and risk for IUGR have resulted in conflicting reports, with some samples documenting that both trimesters carry risk for IUGR (Strauss and Dietz, 1999) and others identifying the second trimester as the most important for maternal weight gain effects (Abrams and Selvin, 1995), with some data more specifically suggesting a threshold effect in early to midpregnancy for birth outcome (Stein et al., 1995). Ecologically oriented studies have concluded that it is the third trimester that is most significant and body length is the variable most affected (Neufeld et al., 1999b). In spite of these critical inquiries, the timing hypothesis paradigm persists and is central to clinical interpretations and public health interventions aimed not only at fetal growth and well being, but postnatal growth and health as well (Barker et al., 1989). The fetal growth timing assumptions are part of a considerable theoretical debate focusing on the nature and outcomes of fetal programming, forging a theory of causal links between the timing of specific prenatal insults and resulting susceptibility to later disease by way of birth size and proportions (Barker et al., 1993; Law et al., 1992). Thus, it is of some interest to reconsider the evidentiary base of the timing hypothesis paradigm. An examination of the data sources giving rise to the graphs in Figure 1 is informative. The curve of growth velocity in crown–heel length is based on unpublished data from ‘‘several sources’’ (Tanner, 1990:39), and according to the author, the peak velocity may be up to 6 weeks later than shown, or between 20 and 26 gestational weeks. Birthweights of live-born premature male infants are cited as the basis for the graph describing a 33–34 week peak in fetal weight growth velocity (Tanner, 1990:41). The accompanying text notes that these graphs should not be taken to represent individual growth patterns, but serve only to describe the patterns of fetal growth in general. It should be noted that these general patterns derive from growth velocities calculated monthly for length and annually for weight. 669 FETAL GROWTH PATTERNS In spite of the careful caveats of the original author, the messages taken from the graphs have become the basis for both clinical interpretation and research rationale about fetal growth and development in individuals. To our knowledge, the explicit propositions of the timing hypothesis have not been previously tested with longitudinal fetal growth data collected with sufficient frequency to identify the details of growth velocity changes in normally growing fetuses. The present study employs serial fetal ultrasounds and investigates three aspects of the timing hypothesis paradigm: The patterns of fetal length and weight growth velocity in the individual, the temporal relationships between the two body parameter velocities, and the predictability of birth outcome. The first proposition tested is that fetal growth in body length has its peak velocity in the second trimester, at or about 20–26 gestational weeks (Tanner, 1978, 1990), and declines subsequently. This specific statement is not without contest in its original form. Longitudinal study of crown–heel length measurements are not possible throughout pregnancy due to the biological constraints on the position of the fetus, characterized by spinal curvature and limb flexion. Crown–rump measurements are, likewise, not highly accurate in a longitudinal study of the late-developing fetus due to poor positional replicability. Because of these limitations, the present investigation employed growth of the leg to assess growth in length. The variable ‘‘leg length’’ was the sum of individual diaphyseal measurements of the tibia and femur. Statistical analysis suggested that this parameter was useful: Weekly growth rates of the tibia and femur were significantly correlated across the sample (P < 0.01, pairwise with Bonferroni correction). Thus, leg growth velocity was used as a reasonable proxy for total body length growth velocity, in lieu of the unobtainable crown–heel length or unreliable measurement of crown–rump length to test the timing hypothesis propositions regarding fetal ‘‘growth in length.’’ The present approach is explicitly an approximation and cannot account for crown–rump growth rates and resulting upper to lower body proportionality contributions to total body length. None of these issues were addressed in the originally published propositions and are not commonly noted when employing the timing hypothesis assumptions. The investigation of the growth patterns of the fetal leg has merit in the context of recent reports of associations between fetal leg growth and childhood health outcomes. For example, third trimester femur length has been found to predict well childhood height (Cacciari et al., 2000) and fetal femur length, and growth rate between 18 and 38 weeks has been found to be inversely associated with childhood blood pressure (Blake et al., 2002). Moreover, strong predictive relationships have been identified between birth length and adult height (Sorensen et al., 1999), and the risk of coronary heart disease has been inversely related to leg length but poorly associated with trunk length in adults (Smith et al., 2001). Thus, the growth patterns of the fetal leg may have long-term implications. The second specific proposition addressed in the present study is that fetal weight velocity is maximal at or about 33–34 gestational weeks of age and declines subsequently (Tanner 1978, 1990). A corollary to the first two propositions is that a trimestral sequence occurs, such that the greatest growth rate and acquisition of length is during the second trimester, and the greatest growth rate and acquisition of weight is during the third trimester. This inference is also examined. Finally, the predictions that maximal effects of length and weight growth on birth outcome occur during the second and third trimesters, respectively, are explored in the growth patterns of this sample. SUBJECTS AND METHODS Sample Forty-four maternal/fetal pairs were participants in a longitudinal growth study between the ages of 13 weeks and birth, with a protocol designed to obtain weekly measurements. All participants were employees at the same health care institution in Brussels, Belgium, and volunteered under conditions of informed consent. Gestational ages were determined as the time from last menstrual period confirmed by measurements of crown–rump length between 8–10 weeks. All pregnancies were single births of healthy mothers with uncomplicated pregnancies. Thirty-three subjects had birth data on weight, length, and infant sex (52% males). 670 M. LAMPL AND P. JEANTY Ultrasound measurements and rationale Six measurements taken serially on each fetus contributed to the present analysis: the biparietal diameter (BPD) and occipitalfrontal diameter (OCF) of the head, the transverse and anteroposterior diameters of the abdomen (TAD, APAD), and the diaphyseal lengths of the two bones of the lower leg, the tibia, and femur. These measurements were obtained by a physician (Jeanty) using calipers and a Toshiba SAL 10A with a frame freeze and a 2.4 MHz transducer. Detailed descriptions of the measurement protocol have been previously published (Jeanty et al., 1982, 1984b,c). Estimations of fetal weight were a calculated composite of individual body measurements. In this study, the previously validated and published equation is used (Jeanty et al., 1984a) where estimated fetal weight (EFW) ¼ Log ½ð0:34  BPÞ þ ð2:7  OCFÞ  ðTADÞ þð2  APADފ=100 This formula was chosen because it does not rely on leg measurements, but derives from the two dimensions of the head and abdomen. The choice of this composite estimation permits unconfounded comparison between estimated fetal weight and leg growth. Analysis methods The object of this study was to identify the patterns and timing of fetal growth velocity in leg length and fetal weight at the level of the individual. Growth velocities were calculated in terms of change per day to accommodate unequal measurement intervals, as no subject met the weekly protocol for the duration of the study precisely each seventh day. The equality of distributions across subject, week, and trimester was investigated (ANOVA, Kruskall-Wallis). Both sample and individual z-scores were calculated for the growth velocities in leg and fetal weight in order to track individual velocity changes and to provide a basis for comparing the growth patterns of the two variables, measured by different scales. All analyses comparing growth rates between variables are based on standard scores. Cubic splines were employed to construct individual growth curves. Least-squares regression models employing the robust procedure for repeated measure- ments were used to test for age-specific and trimestral effects on leg and EFW growth velocities. Individuals’ incremental growth in leg length and estimated fetal weight was derived across the second (weeks 13–26) and third (weeks 26–38) trimesters from the growth curves to control for variability in day of ultrasound measurement. ShapiroWilks test for normality and two-tailed t-test procedures investigated the equality of means across trimesters (paired t-test for unequal variance). The effects of sex and maternal smoking on incremental growth were investigated by least-squares regression. Least-squares regression models employing the robust procedure for repeated measurements were used to explore trimestral relationships in size, growth rates, and birth outcome. Gestational age at birth was included in all regression models investigating birth outcomes. All statistical analyses were based on the computational assumptions of Stata 6 (Stata Corp., Release 6.0. College Station, TX) and significance was accepted at P  0.05. RESULTS Fetal size in terms of mean femur length, biparietal, and abdominal diameters were between the 40th and 60th centiles of reference standards for other similar European samples across the second and third trimesters (Crequat et al., 2000; Guihard-Costa and Larroche, 1995). The means (and standard deviations, SD) for birthweight and length were 3,349 g (SD 399 g) and 50.4 cm (SD 1.5 cm), respectively. All births were term, with an average gestational age of 275 days (39.2 weeks, SD 7.4 days) and there was no intrauterine growth retardation assessed by weight for gestational age. All subjects received pediatric care at birth and all neonates were clinically normal. Controlling for day of ultrasound measurement and maternal smoking, significant sex differences were found in terms of leg length, with girls exceeding boys by midsecond trimester (beta 3.1, P ¼ 0.06, model R2 ¼ 0.77, P ¼ 0.00). No significant sex differences were found in leg length and EFW in the third trimester. Fetal growth velocity patterns The distributions of growth velocities for estimated fetal weight and leg length by FETAL GROWTH PATTERNS Fig. 2. Distribution of growth velocities for estimated fetal weight (above) and leg length (below) by gestational weeks of age. Pooled data are shown as means (horizontal lines within boxes), the interquartile range (boxes), and the 95% range (bars). Within individuals, the velocities are normally distributed. The pooled sample’s ranges result from the pattern differences between children: each vertical range includes both peaks and troughs of individual subjects. gestational week of age are plotted in Figure 2. Significantly different distributions characterized both variables by week (P < 0.01) and EFW velocity by trimester (P < 0.01). Thus, pooling the data was inappropriate and individual-level analyses were undertaken to address the research question regarding the nature of individual growth velocity patterns. Neither the growth velocity of the leg nor that for estimated fetal weight was characterized by a single peak value followed by decline. Instead, the growth patterns of both variables followed a pulsatile pattern of peaks and troughs (Fig. 3), unassociated with the timing hypothesis peak ages (Fig. 4). The significant variance heterogeneity by week found in the pooled data reflected individual differences in the timing of the peaks and troughs. Combining individual growth 671 Fig. 3. Growth velocity in the leg (above) and estimated fetal weight (below) in an individual fetus (subject number 6). The leg growth velocity graph provides an overlay of diaphyseal length velocities of the tibia (circles) and femur (triangles), together with the calculated leg length velocity (squares). The y-axis is mm/day to correct for unequal measurement intervals actualized under the weekly protocol. The y-axis for the growth velocity of estimated fetal weight is grams per day. patterns would have superimposed one individual’s growth peak pulse on another’s trough and attenuated the pulsatile patterns by which the individuals grew. These data emphasize the importance of individual differences in growth patterns in fetal development. Fetal leg and weight growth relationships As no single-peak growth pattern characterized the growth velocities of either leg length or EFW in these individuals, other approaches were further explored to understand the relationship of the present sample’s growth to the timing paradigm. Were there some broader patterns associated with individuals’ growth velocities in leg length and EFW that might be compatible with the generalizations presented in Figure 1? The first approach tested the proposition that comparatively high velocities over the 672 M. LAMPL AND P. JEANTY Fig. 4. The multiple peaks and troughs by which individual growth in leg length (circles) and estimated fetal weight (triangles) velocity progressed is illustrated (Subject 7). The pattern of multiple peaks contrasts with the predicted fetal growth model in terms of overall single peak and decline process and timing. range of the respective predicted peak growth periods might have occurred. Leastsquares regression for repeated measurements investigated this by comparing the leg growth velocities during the putative timing hypothesis peak length growth interval (20–26 weeks’ gestational age) with all other weeks. The model controlled for potential confounding effects of day of measurement, maternal smoking, and sex, and was run separately for both raw velocities and velocity standard scores. No significant effects of ‘‘peak growth period’’ were identified (model P ¼ 0.00, R2 ¼ 0.08) in leg length growth velocity. Likewise, the growth velocities of EFW during weeks 33–34 were compared to all other weeks, controlling for the same potential confounders. A significant negative slope was identified for the EFW growth velocity during these weeks (beta of –0.43 for EFW velocity standard score, SE 0.12, range –0.66 to –0.19, model P ¼ 0.00, R2 ¼ 0.32). Least-squares regression for repeated measurements also investigated the effects of trimester (second vs. third) on individuals’ leg length and EFW growth velocities (separately for both raw velocities and z-scores), controlling for confounders. No significant linear effects of trimester were found for either leg velocity (model P ¼ 0.00, R2 ¼ 0.07) or EFW velocity (model P ¼ 0.00, R2 ¼ 0.24). Thus, the pooled sample data did not identify significant effects of either the timing hypothesis peak growth weeks or the associated trimesters on leg length or fetal weight growth velocities. Finally, individuals’ incremental growth in each trimester was assessed to investigate the proposition that greater growth in length occurs during the second trimester and greater growth in fetal weight during the third trimester. Increments were derived from individuals’ leg length and EFW growth curves to identify growth progress in the second trimester (between weeks 13–26) and third trimester (weeks 26–38). Total EFW incremental growth in individuals was significantly greater in the second trimester than incremental growth in leg length (t-tests, P < 0.00). This was followed by greater growth in leg length by comparison with EFW in the third trimester (t-tests, P ¼ 0.00) (Fig. 5). No significant effects of sex or maternal smoking were evident in comparisons investigating equivalence of means. However, least-squares regression identified significant negative effects of maternal smoking on individuals’ leg length increments during the third trimester, controlling for sex (beta coefficient, –0.34, SE 0.12, P ¼ 0.01, R2 ¼ 0.24, model P ¼ 0.03). On further investigation, there was no significant effect on females, but maternal smoking was significantly associated with a negative slope in leg length increment during the third trimester for males (beta coefficient, –0.54; SE 0.17; 95% CI –0.92 to –0.17; P < 0.01) with an R2 ¼ 0.43 and model P < 0.01. These analyses led to the question: ‘‘If there was not a simple sequential relationship of a single second trimester peak in leg growth followed by a single third trimester peak in weight growth, was there another relationship between the two aspects of body growth?’’ A comparison of the fetal weight and leg length velocity growth patterns in individuals identified three relationships. First, there was a pattern of ‘‘copulsatility,’’ when growth pulses in both leg length and fetal weight velocities occurred simultaneously. A second relationship was a ‘‘beat frequency pattern,’’ in which one variable pulsed at a multiple of the other. Third, the leg and weight growth pulses exhibited a ‘‘trade-off growth,’’ when the parameters were pulsing in opposition. For example, the subject illustrated in the top panel of Figure 6 (Subject 32) grew by a 3:2 leg length:EFW pulse pattern between FETAL GROWTH PATTERNS 673 Fig. 5. Individuals’ incremental growth in EFW was greater than leg length in the second trimester, and incremental growth in leg length exceeded EFW across the third trimester (t-test for correlated samples, P < 0.01). Mean incremental standard scores (dots), SE (boxes), and SD (bars) are illustrated. Fig. 6. Individual growth velocity patterns in leg length (darker pen, circles) and weight (lighter pen, squares) for Subjects 32 (female, above) and 31 (male, below). The growth curves were generated from cubic splines. The y-axis represents z-score change per day, and the x-axis is gestational age in weeks. Individuals’ leg length did not proceed by a single peak at 20–26 weeks, followed by decline, and EFW did not grow by a process with a peak weight velocity about 33 weeks. 18–28 weeks, when the timing shifted to copulsatility. The individual illustrated in the bottom panel of Figure 5 (Subject 31) grew by a trade-off pattern throughout gestation. The top panel was a female and the bottom panel was a male. These qualitative patterns were confirmed by several approaches. First, correlation analysis identified that females experienced significant correlations between leg length and EFW growth velocities in both the second and third trimesters (pairwise, Bonferroni correction P ¼ 0.05, 0.01, respectively); males did not. Controlling for day of ultrasound measurement and the potential confounding effect of maternal smoking, regression analysis identified the quantitative directional effects of this relationship for females (Table 1). In the second trimester, leg length and EFW growth velocities predicted reciprocal effects that were of roughly similar magnitude (on the order of 0.2 SDs). In the third trimester, leg growth velocity predicted a modest increase in EFW velocity slope, while EFW effects on leg growth remained relatively the same. Thus, females experienced coordinated growth velocity patterns throughout gestation. No significant associations were found for males in growth velocity between EFW and leg length. In addition, a sequential temporal relationship in size was identified in which the EFW during the second trimester (week 23) modestly predicted leg length during the third trimester (week 32) for females (beta 0.66, P ¼ 0.06, R2 0.85, model P ¼ 0.02). No 674 M. LAMPL AND P. JEANTY TABLE 1. Robust regression results by trimester for females’ growth velocities, controlling for gestational age and maternal smoking Leg velocity Beta SE t P 95% Cl R2, model P EFW velocity Females Trimester 2 Trimester 3 0.24 0.34 0.09 0.07 2.8 4.7 0.01 0.00 0.05, 0.42 0.19, 0.49 0.06, 0.05 0.19, 0.00 EFW velocity Beta SE t P 95% Cl R2, model P Leg velocity Females Trimester 2 Trimester 3 0.19 0.50 0.09 0.13 2.1 3.8 0.05 0.00 0.00, 0.38 0.22, 0.78 0.06, 0.01 0.19, 0.00 Analyses employed growth velocities’ standard scores. TABLE 2. Regression results by trimester for models predicting estimated fetal weight (z-score) from leg length (z-score) controlling for gestational age, sex, and maternal smoking (total sample) and gestational age, maternal smoking (by sex) Fetal weight (z-score) Trimester 2 Leg length (z-score) (Total sample) Males Females Trimester 3 Leg length (z-score) (Total sample) Males Females Beta SE t P 95% Cl R2, model P 0.68 0.65 0.73 0.10 0.15 0.15 6.5 4.2 4.9 0.00 0.00 0.00 0.47, 0.89 0.32, 0.97 0.41, 1.04 0.94, 0.0 0.94, 0.0 0.94, 0.0 0.30 0.28 0.24 0.06 0.07 0.10 4.7 4.1 2.3 0.00 0.00 0.04 0.17, 0.43 0.13, 0.42 0.01, 0.46 0.83, 0.0 0.85, 0.0 0.84, 0.0 TABLE 3. Regression results by trimester for models predicting leg length (z-score) from fetal weight (z-score) controlling for gestational age, sex, and maternal smoking (total sample) and gestational age, maternal smoking (by sex) Leg length (z-score) Trimester 2 EFW (z-score) (Total sample) Male Female Trimester 3 EFW (z-score) (Total sample) Male Female Beta SE t P 95% Cl R2, model P 0.29 0.26 0.33 0.04 0.05 0.06 7.9 5.2 5.7 0.00 0.00 0.00 0.22, 0.37 0.16, 0.37 0.21, 0.46 0.97, 0.0 0.97, 0.0 0.97, 0.0 0.45 0.60 0.28 0.11 0.19 0.11 3.9 3.2 2.6 0.00 0.01 0.02 0.22, 0.68 0.20, 1.00 0.05, 0.52 0.90, 0.0 0.89, 0.0 0.92, 0.0 significant reverse effects were identified and no such relationship was found for males. The contemporaneous size relationships between fetal leg length and weight were also investigated by regression analyses (Tables 2, 3). Controlling for day of ultrasound measurement and maternal smoking, the slopes of the regression lines for leg length on fetal weight were greater during the second trimester than in the third trimester (approximately two-fold for males, and three-fold for females). By contrast, for models predicting leg length from EFW, the slopes of the regression lines doubled for males in the third trimester by comparison with the second, while they dropped modestly for females. Thus, leg growth in the third trimester male was heavily influenced by third trimester EFW. 675 FETAL GROWTH PATTERNS attenuated in the third trimester, when the slope of leg length to birth length outcome dropped by one-half, after controlling for the positive effects of EFW and the negative effects of third trimester leg growth velocity. Thus, females with long legs in the second and third trimesters were longer at birth. However, controlling for these effects, girls who had high third trimester leg growth rates were significantly shorter at birth. No effects were identified in the second trimester for boys. In the third trimester, a model with both EFW and leg length best predicted birth length in boys, with EFW predicting twice the outcome effects of leg length. Predictability of birth outcome Regression analyses investigated the trimestral effects that leg length and estimated fetal weight had on predicting birth outcome in total body weight, length, and ponderal index (Table 4) controlling for gestational age at birth, age at ultrasound measurement, and maternal smoking. Significant sex differences were found in the predictive relationships between the timing of body growth and birth outcome. Birth length. Among girls, leg length in the second trimester significantly predicted birth length. This relationship continued TABLE 4. Birth outcomes predicted by robust regression models controlling for age at ultrasound measurement, gestational age at birth, and maternal smoking Birth outcome Birth length Trimester 2 Females Leg length Trimester 3 Males EFW Leg length Females Leg length Leg velocity EFW Birth weight Trimester 2 Males EFW Leg length Trimester 3 Males EFW Leg length Females Leg length EFW Ponderal index Trimester 2 Males EFW Leg length Females Leg length Trimester 3 Males Leg velocity Leg length EFW Females Leg velocity Leg length EFW P 95% Cl R2, model P 2.9 0.02 0.5, 4.1 0.34, 0.03 0.37 0.28 3.2 2.1 0.01 0.06 0.37, 2.0 0.03, 1.2 0.41, 0.00 1.3 0.1 0.9 0.49 0.57 0.04 2.7 2.8 1.7 0.02 0.02 0.12 0.27, 2.4 0.18, 0.02 0.30, 2.2 0.36, 0.03 1.1 1.6 0.3 1.1 3.4 1.5 0.01 0.16 0.4, 1.8 0.7, 3.8 0.40, 0.00 1.8 1.5 0.7 0.8 2.6 2.0 0.02 0.06 0.31, 3.3 0.10, 3.2 0.56, 0.00 2.0 1.8 0.7 0.8 2.8 2.1 0.02 0.06 0.47, 3.5 0.04, 3.6 0.32, 0.01 1.4 1.5 0.7 1.9 2.0 0.8 0.07 0.45 0.1, 2.9 2.6, 5.6 0.39, 0.00 3.8 1.4 2.7 0.02 6.7, 0.20, 0.10 0.2 2.2 0.1 0.1 1.4 1.2 2.1 1.6 0.1 0.05 0.13 0.90 0.4, 0.00 0.7, 5.2 2.5, 2.8 0.46, 0.00 0.2 1.0 2.0 0.1 1.0 1.8 3.1 1.0 1.1 0.01 0.32 0.28 0.05, 0.3 3.2, 1.1 1.8, 5.8 0.08, 0.04 Beta SE t 2.3 0.8 1.2 0.6 All variables were transformed to z-scores prior to analysis. 0.76 676 M. LAMPL AND P. JEANTY Birthweight (BW). No significant relationship was evident for females in the second trimester, but in the third trimester both leg length and EFW together best predicted BW. For males, EFW was the significant factor in predicting outcome, increasing across trimesters by nearly 60%. Ponderal index (PI). For females in the second trimester, only leg length in the second trimester significantly predicted PI at birth, such that a 1 z-score leg length increase predicted a negative 3.8 z-score change in PI at birth. Thus, long legs by the second trimester predicted a long and thin female neonate at birth. Controlling for leg length and fetal weight, leg growth velocity in the third trimester positively predicted birth PI. In boys, EFW in the second trimester modestly positively predicted PI. By the third trimester, only leg velocity growth rate was a significant predictor, with negative effects on PI at birth. DISCUSSION Contrary to the commonly cited notion regarding the pattern of fetal growth, there was no single peak in growth velocity of either leg length or estimated fetal weight in the patterns of individual growth in this sample. Specifically, leg growth velocity was not characterized by a single or maximal peak range between 20 and 26 weeks of age, followed by declining velocity, and estimated fetal weight growth did not reach a peak at about 33 weeks of age. These results are confined to a clarification of the contribution made by leg growth to the timing hypothesis propositions about total body length. The growth patterns of both body parameters in these healthy normally growing fetuses were nonlinear, with multiple peaks and troughs. Further, the timing hypothesis proposition that the second trimester is the interval of greatest length acquisition, followed by weight in the third trimester, was not supported by the present study. In this sample the growth patterns were the opposite, with estimated fetal weight growth accrual declining after the second trimester and leg growth in the third trimester continuing apace with, or exceeding, that of the second. Regression analyses of within- and between-term effects identified more com- plex relationships between growth in fetal weight and leg length, with significant sex effects. Growth of legs and fetal weight progressed by a more synchronous pattern in females than males, and females experienced earlier growth of lower limbs than boys, suggesting advanced developmental timing. While second trimester EFW predicted leg length in third trimester females, third trimester EFW predicted birth length in males: a similar relationship, offset in timing. It can by hypothesized that girls are more canalized, with their birth outcomes set by genetics and in utero conditions earlier than boys. This may be influenced by hormonal conditions as female ovarian development, peaking in the middle of the second trimester, contributes to growth and maturation. Whether the postnatally documented effect of estrogen on bone growth operates prenatally remains to be documented (Okazaki et al., 2002). The male delay in growth and development may result in increasing male vulnerability, as their growth is more dependent on third trimester conditions. Greater daily fetal weight growth rate in third trimester males has been previously reported in a high-risk population (DeJong et al., 1998). A time of increased strain on maternal resources in general (Martinez-Morales et al., 1999), this timing difference may underlie some of the documented sex differences in male risk during gestation (e.g., Zaren et al., 2000). This was seen in the present sample as the negative slopes of leg length increment that were predicted by maternal smoking in the third trimester. Overall, these results predict that males born prematurely are at greater risk for consequences of immaturity than females. This study suggests that size at birth has different biological meanings for each sex, reflecting a different developmental timing program by sex. This is relevant to ongoing debates about the health sequelae of birth size in later life (Barker, 1995). Many studies find the epidemiological relationships between low BW and/or low PI apply only to males (Kuzawa, 2001; Suzuki et al., 2000; Valdez et al., 1994; Ziegler et al., 2000). This makes sense if the relationships found here apply more generally. For example, in this sample low PI at birth accompanied a fast-growth pattern in girls, as those with longer legs were already on a trajectory for long and thin morphology at FETAL GROWTH PATTERNS birth by the second trimester. Previous work suggests that these observations may reflect fetal entrainment of maturational rate, as earlier menarcheal age occurred among adolescent girls who were long and thin at birth in a population-based study (Adair, 2001). Moreover, in the present sample high female third trimester leg growth rates predicted shorter length and higher ponderal index at birth. It can by hypothesized that these girls were comparatively later maturers, by reference to the previous observation identifying later menarcheal ages among adolescent girls who were relatively short and heavy at birth (Adair, 2001). By contrast, boys with high EFW in the second trimester were on a trajectory for higher PI at birth. By the third trimester, high leg growth rates predicted decreasing PI. Thus, boys with low PI at birth may have been poor growers overall by the second trimester as a result of first trimester factors, a finding in line with previous reports (Brown et al., 2002), or they may have been robust third trimester growers. The importance of maternal energy resources for third trimester leg growth has been previously reported (Neufeld et al., 1999a). Long and thin morphology at birth may reflect a much more demanding energetic stress in males whose delayed growth trajectory puts them at greater risk for managing diminishing maternal resources during the third trimester. Thus, low PI (or weight) at birth may have a very different meaning in terms of the contributory biological patterns for the sexes, reemphasizing the questionable meaning of birth size as a predictive variable (Wilcox, 2001). These data suggest that mixed sex analyses are likely to err in their description of fetal growth pattern effects and trimester-focused analyses of benefits and insults on birth outcome are not likely to reveal details of growth chronology. Mechanisms for sex-specific growth patterns have been described (Davey et al., 1999). The present study observations support previous reports identifying the importance of both second (Abrams and Selvin, 1995; Hediger et al., 1989; Li et al., 1998; Metcoff, 1980) and/or third (Strauss and Dietz, 1999; Neufeld et al., 1999a,b) trimester maternal nutrition on birthweight, length, and ponderal index. One of the questions arising from these observations is mechanistic. Does the timing 677 relationship found here, in which growth in fetal weight precedes growth in leg length, reflect an energetic relationship? This question bears on an ongoing debate regarding the temporal relationship between growth in height and weight (Waterlow, 1994) and is a proposition in line with reports of weight accrual preceding linear growth during recovery from malnutrition (Walker and Golden, 1988). The energy-driven-growth line of reasoning necessitates careful attention to the meaning of fetal weight. The common frame of reference for the concept of weight is as a monitor of changing body energy stores. The nomenclature belies the nature of the measurement. Any estimate of fetal weight presently employed is a composite of body measurements that is only predictive of true weight. The upper body measurements used here to predict weight focus on the head and abdominal circumference. Other studies have documented that the latter is likely to have been a good marker of adipose tissue accrual and, thus, a reasonable proxy for fetal weight (Kehl et al., 1996). But central to concerns surrounding the well being of the fetus, and articulated by all debates in progress concerning fetal programming, is neonatal body composition as partitioned into lean body mass and fat reserves. Longitudinal data documenting developmental timing in fetal fat formation are rare. Early histological studies identified the second trimester as a key period of adipogenesis, with no sex differences and a proximal/distal gradient (Berg, 1911; Poissonet et al., 1984). The pattern and pace of increase in fat cell number, by contrast to fat storage within existing cells, is only beginning to be carefully documented longitudinally in the normally growing human fetus (Enzi et al., 1981). From a physiological viewpoint, it is not likely that there are distinctive long-term, or linear, trimestral-based differences in bone and body weight growth, as predicted by the timing hypothesis. It is much more likely that there is cross-talk between the two processes during development, as suggested by changing growth rates in the longitudinal data presented here. In fact, the often discussed observation that there appears to be a trade-off between growth in body length and weight during early development has recently been given a scientific foundation and can now be 678 M. LAMPL AND P. JEANTY framed more precisely: Both bone and fat derive from the same stem cell (Pittenger et al., 1999). This deterministic moment in a stem cell’s life is prompted by signals from differentiation factors that influence the alternative genetic programs that contrast the functional aspects of chondrocytes, osteoblasts, and fat cells. This is followed by activation and proliferation signals, which interface the environment and the developing organism (Ahdjoudj et al., 2001). In summary, the longitudinal data in this study documented that fetal growth progresses by a pulsatile pattern of leg and estimated weight acquisition, not a single peak and decline process, according to a temporal clock that was neither accurately characterized by the timing hypothesis, nor universally paced across individuals. The healthy growing fetuses in this sample exhibited greater incremental growth in EFW during the second trimester, by comparison with leg length, which increased in the third trimester. As the fetal weight estimates are based on head and abdominal measurements, a careful interpretation of these results is that they redemonstrate the proximal/distal growth gradient, and girls were accelerated. The high predictability of leg length for fetal weight in the second trimester, by contrast with the reverse, identifies that if legs grew, it was likely that the upper body had grown; but upper body growth may not have been accompanied by lower leg growth. This developmental pattern is likely to reflect fetal blood flow characteristics and to express both developmental timing and energy shortfall that preferences the upper body (Lampl et al., 2003). By the third trimester, upper body growth was highly predictive of leg length in boys, but less so in girls, whose third trimester leg length was already predicted by second trimester upper body growth. The pulsatile patterns identified in these data corroborate previous reports of nonlinear and discontinuous fetal growth patterns based on time-intensive longitudinal data (Bernstein et al., 1996, 1999), clarify the specifics of individual fetal growth velocity variability (Guihard-Costa, 2000), and are in line with a saltatory growth process during infancy (Lampl et al., 1992). Individual variability in pulse pattern timing within and between leg growth and estimated fetal weight suggests a functional relationship in the temporal trajectories compatible with an underlying nonlinear dynamic process. This would be a highly adaptive strategy, one commonly found in biological systems designed for high flexibility in their function (Pincus, 1995). A dynamical growth process would permit coordinated development in size and body proportionality by pathways that translate available resources into changing growth rates. These observations suggest that growth pulse pattern changes are a central adaptive strategy and underlie sex differences in developmental timing. Differences in body proportionality at birth may reflect differences in the developmental timing of nonlinear growth patterns. The chronology of growth, and the precise point on the trajectory between pulses in weight and length that coincides with an individuals’ birth, may be very important in determining birth outcome in body proportion. This is a hypothesis-generating study and these propositions remain to be tested in future investigations. The present data document a pulsatile fetal growth process that accommodates sex and individual developmental differences in growth timing. Further research on fetal body composition is needed to clarify the physiology of energy storage and utilization as it applies to the process of fetal growth. Sex differences are likely to be found. Recent scientific evidence suggests that it is improbable that the linear sequence outlined by the timing hypothesis is physiologically accurate. It is more reasonable to posit that a synergistic, nonlinear, and dynamic paradigm will be the explanatory framework for fetal growth and development. LITERATURE CITED Abrams B, Selvin S. 1995. Maternal weight gain pattern and birth weight. Obstet Gynecol 86:163–169. Adair LS. 2001. Size at birth predicts age at menarche. Pediatrics 107:E59. Ahdjoudj S, Lasmoles F, Oyajobi BO, Lomri A, Delannoy P, Marie PJ. 2000. Reciprocal control of osteoblast/chondroblast and osteoblast/adipocyte differentiation of multipotential clonal human marrow stromal F/S. Pediatr Res 47:578–585. Barker DJP. 1995. Fetal origins of coronary heart disease. Br Med J 311:171–174. Barker DJP, Osmond C, Golding J, Kuh D, Wadsworth ME. 1989. Growth in utero, blood pressure in childhood and adult life, and mortality from cardiovascular disease. Br Med J 298:564–567. Barker DJ, Osmond C, Simmonds SJ, Wield GA. 1993. The relation of small head circumference and thinness FETAL GROWTH PATTERNS at birth to death from cardiovascular disease in adult life. Br Med J 306:422–426. Berg. 1911. Über die Anlage und Entwicklung des Fettgewebes beim Menschen. Z Morphol Anthropol 13:305–341. Bernstein IM, Badger GJ. 1999. The pattern of normal fetal growth. In: Lampl M, editor. Saltation and stasis in human growth and development. London: SmithGordon. p 27–32. Bernstein IM, Blake K, Wall B, Badger GH. 1996. Evidence that normal fetal growth can be noncontinuous. Obstet Gynecol Surv 51:213–214. Blake KV, Gurrin LC, Beilin LJ, Stanley FJ, Kendall GE, Landau LI, Newnham JP. 2002. Prenatal ultrasound biometry related to subsequent blood pressure in childhood. J Epidemiol Community Health 56:713–718. Brown JE, Murtaugh MA, Jacobs DR Jr, Margellos HC. 2002. Variation in newborn size according to pregnancy weight change by trimester. Am J Clin Nutr 76:205–209. Cacciari E, Salardi S, David C, Tassinari D, Dalla Casa C, Pilu GL, Mainetti B, Gualandi S, Bovicelli L. 2000. Is statural growth predictable in utero? Follow-up from the second trimester of gestation to the 8th year of life. J Pediatr Endocrinol Metab 13:381–386. Crequat J, Duyme M, Brodaty G. 2000. Biometry 2000. Fetal growth charts by the French College of fetal ultrasonography and the Inserm U 155. Gynecol Obstet Fertil 28:435–445. Davey HW, Wilkins RJ, Waxman DJ. 1999. STAT5 signaling in sexually dimorphic gene expression and growth patterns. Am J Hum Genet 65:959–965. DeJong CL, Gardosi J, Baldwin C, Francis A, Dekker GA, van Geijn HP. 1998. Fetal weight gain in a serially scanned high-risk population. Ultrasound Obstet Gynecol 11:39–43. Enzi G, Zanardo V, Caretta F, Inelmen EM, Rubaltelli F. 1981. Intrauterine growth and adipose tissue development. Am J Clin Nutr 34:1785–1790. Guihard-Costa AM, Larroche J-C. 1995. Fetal biometry. Fetal Diagn Ther 10:215–278. Guihard-Costa AM, Droullé P, Thiebaugeorges O, Hascoet JM. 2000. A longitudinal study of fetal growth variability. Biol Neonate 78:8–12. Hediger ML, Scholl TO, Belsky DH, Ances IG, Salmon RW. 1989. Patterns of weight gain in adolescent pregnancy: effects on birth weight and preterm delivery. Obstet Gynecol 74:6–12. Jeanty P, Dramaix-Wilmet M, van Kerkem J, Petroons P, Schwers J. 1982. Ultrasonic evaluation of fetal limb growth. Radiology 143:751–754. Jeanty P, Cantraine F, Romero R, Cousaert E, Hobbins JC. 1984a. A longitudinal study of fetal weight growth. J Ultrasound Med 3:321–328. Jeanty P, Cousaert E, Cantraine F. 1984b. Normal growth of the abdominal perimeter. Am J Perinatol 1:129–135. Jeanty P, Cousaert E, Hobbins JC, Tack B, Bracken M, Cantraine F. 1984c. A longitudinal study of fetal head biometry. Am J Perinatol 1:118–128. Kehl RJ, Krew MA, Thomas A, Catalano PM. 1996. Fetal growth and body composition in infants of women with diabetes mellitus during pregnancy. J Matern Fetal Med 5;273–280. Kramer MS, McLean FH, Olivier M, Willis DM, Usher RH. 1989. Body proportionality and head and length ‘sparing’ in growth-retarded neonates: a critical reappraisal. Pediatrics 84:717–723. Kuzawa CW. 2001. Maternal nutrition, fetal growth, and cardiovascular risk in Filipino adolescents. Ph.D. dissertation. Ann Arbor, MI: University Microfilms. 679 Lampl M, Veldhuis JD, Johnson ML. 1992. Saltation and stasis: a model of human growth. Science 258:801–803. Lampl M, Kuzawa CW, Jeanty P. 2003. Prenatal smoke exposure alters growth in limb proportions and head shape in the midgestation human fetus. Am J Human Biol 15:533–546. Law CM, Barker DJ, Osmond C, Fall CH, Simmonds SJ. 1992. Early growth and abdominal fatness in adult life. J Epidemiol Community Health 46:184–186. Li R, Haas JD, Habicht J-P. 1998. Timing of the influence of maternal nutritional status during pregnancy on fetal growth. Am J Hum Biol 10:529–539. Martinez-Morales S, Bonillo-Perales A, Munoz-Hoyos A, Puertas-Prieto A, Uberos-Fernandez J, MolinaCarballo A, Bonillo-Perales JC, Sabatel-Lopez R. 1999. The influence of maternal erythrocyte deformability on fetal growth, gestational age and birthweight. J Perinat Med 27:166–172. Metcoff J. 1980. Maternal nutrition and fetal development. Early Hum Dev 4:99–120. Neufeld L, Pelletier DL, Haas JD. 1999a. The timing of maternal weight gain during pregnancy and fetal growth. Am J Hum Biol 11:627–637. Neufeld L, Pelletier DL, Haas JD. 1999b. The timing hypothesis and body proportionality of the intrauterine growth retarded infant. Am J Hum Biol 11:638–646. Okazaki R, Inoue D, Shibata M, Saida M, Kido S, Ooka H, Tomiyama H, Sakamoto Y, Matsumoto T. 2002. Estrogen promotes early osteoblast differentiation and inhibits adipocyte differentiation in mouse bone marrow stromal cells lines that express estrogen receptor (ER) alpha or beta. Endocrinology 143: 2349–2356. Petersen S, Larsen T, Greisen G. 1992. Judging fetal growth from body proportions at birth. Early Hum Dev 30:139–146. Pincus SM. 1995. Quantifying complexity and regularity of neurobiological systems. Methods Neurosci 28: 336–363. Pittenger MF, Mackay AM, Beck SC, Jaiswal RK, Douglas R, Mosca JD, Moorman MA, Simonetti DW, Craig S, Marshak DR. 1999. Multilineage potential of adult human mesenchymal stem cells. Science 284:143–147. Poissonnet CM, Burdi AR, Garn SM. 1984. The chronology of adipose tissue appearance and distribution in the human fetus. Early Hum Dev 10:1–11. Smith GD, Greenwood R, Gunnell D, Sweetnam P, Yarnell J, Elwood P. 2001. Leg length, insulin resistance, and coronary heart disease risk: the Caerphilly Study. J Epidemiol Community Health 55:867–872. Sorensen HT, Sabroe S, Rothman KJ, Gillman M, Steffensen FH, Fischer P, Sorensen TI. 1999. Birth weight and length as predictors for adult height. Am J Epidemiol 149:726–729. Stein AD, Ravelli AC, Lumey LH. 1995. Famine, thirdtrimester pregnancy weight gain, and intrauterine growth: the Dutch Famine Birth Cohort Study. Hum Biol 67:135–150. Strauss RS, Dietz WH. 1999. Low maternal weight gain in the second or third trimester increases the risk for intrauterine growth retardation. J Nutr 5:988–993. Suzuki T, Minami J, Ohrui M, Ishimitsu T, Matsuoka H. 2000. Relationship between birth weight and cardiovascular risk factors in Japanese young adults. Am J Hyper 13:907–913. Tanner JM. 1978. Fetus into man, 1st ed. Cambridge, MA: Harvard University Press. Tanner JM. 1990. Fetus into man, 2nd ed. Cambridge, MA: Harvard University Press. p 41. 680 M. LAMPL AND P. JEANTY Valdez R, Athens MA, Thompson GH, Bradshaw BS, Stern MP. 1994. Birthweight and adult health outcomes in a biethnic population in the USA. Diabetologia 37:624–631. Vik T, Vatten L, Jacobsen G, Bakketeig LS. 1997. Prenatal growth in symmetric and asymmetric small-for-gestational-age infants. Early Hum Dev 25:167–176. Villar J, Belizan JM. 1982. The timing factor in the pathophysiology of the intrauterine growth retardation syndrome. Obstet Gynecol Surv 37:499–506. Walker SP, Golden MH. 1988. Growth in length of children recovering from severe malnutrition. Eur J Clin Nutr 42:395–404. Waterlow JC. 1994. Relationship of gain in height to gain in weight. Eur J Clin Nutr 48(Suppl 1): S72–S74. Wilcox AJ. 2001. On the importance—and the unimportance—of birthweight. Int J Epidemiol 30:1233–1241. Zaren B, Lindmark G, Bakketeig L. 2000. Maternal smoking affects fetal growth more in the male fetus. Paediatr Perinat Epidemiol 14:118–126. Ziegler B, Johnsen SP, Thulstrup AM, Engberg M, Lauritzen T, Sorensen HT. 2000. Inverse association between birth weight, birth length and serum total cholesterol in adulthood. Scand Cardiovasc J 34:584–588.