AMERICAN JOURNAL OF EPIDEMIOLOGY
Vol. 132, No. 3
Copyright © 1990 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved
Printed in U.S.A.
THE PONDERAL INDEX IN TERM NEWBORN SIBLINGS MUIN J. KHOURY, CYNTHIA J. BERG, AND EUGENIA E. CALLE Khoury, M. J. (CDC, Atlanta, GA 30333), C. J. Berg, and E. E. Calle. The ponderal index in term newborn siblings. Am J Epidemiol 1990;132:576-83. Rohrer's ponderal index in newborns (birth weight/height3 x 100) has been used as an indicator of fetal growth status, especially to assess asymmetrical intrauterine growth retardation. Because low birth weight and intrauterine growth retardation tend to recur in sibships, we examined patterns of sibling correlation in the ponderal index in 795 live term (>37 weeks) singleton sibling pairs without birth defects born between 1966 and 1986 and fathered by male US Army veterans participating in a nationwide health study. Data on birth weight, length, gestational age, and other maternal and infant health characteristics were abstracted from hospital-of-birth medical records. The correlation coefficient of ponderal index in sib pairs was 0.24 (p < 0.001). Compared with 627 infants who had a prior sib with a ponderal index between the 10th and 90th percentiles, 92 infants who had a prior sib with ponderal index < the 10th percentile had a lower mean ponderal index and a higher proportion with ponderal index < the 10th percentile (13.0% vs. 8.5%). On the other hand, 76 infants who had a prior sib with ponderal index > the 90th percentile had a higher mean ponderal index and higher proportion with ponderal index > the 90th percentile (17.1% vs. 10.2%). The clustering of ponderal index in siblings persisted after controlling for factors such as race, gender, maternal age, gravidity, year of birth, gestational age, pregnancy complications, and prior maternal illnesses. The findings point to the presence of genetic and/or maternal factors affecting the growth status of term newborn infants. The significance of the ponderal index needs to be examined in future genetic and epidemiologic studies of intrauterine growth. fetal growth retardation; genetics
Rohrer's ponderal index in newborns refers to the ratio of birth weight (g) to the cube of the length (cm3) X 100 and has gained increasing attention in studies of low birth weight and intrauterine growth
retardation (1-12). The ponderal index quantifies the relation of the amount of soft tissue mass to skeletal development (1). Increasingly, researchers have recognized the pathophysiologic heterogeneity of intrauterine growth retardation (1, 4, 12). Received for publication May 9,1989, and in final T h e P°nderal index is being used to cateform March 16,1990. gorize growth-retarded infants into those From the Division of Birth Defects and Develop- w jth asymmetric growth retardation (low mental Disabilities, Center for Environmental Health
,
i • j
. ,, ,
...
.
and Injury Control, Centers for Disease Control, POnderal index, weight-for-gestational age Atlanta, GA. proportionately more affected than height) Reprint requests to Dr. Muin J. Khoury, Birth a n ( j symmetric growth retardation (normal Defects and Genetic Diseases Branch (F37), Center , , . , , ,, . , , ,, . ,, for Environmental Health and Injury Control, Centers P°nderal index, both weight and height affor Disease Control, Atlanta, GA 30333. fected) (4). These two subtypes of intraData for this study were gathered under an inter- uterine growth retardation may have difagency agreement with the Veterans Administration. f/e r •e n ,t ,t l m• m• • i n .i_ • i_ The authors thank Essie Fuller for her help in g pathogenesis, may be preparing the manuscript. related to different underlying etiologic fac576
PONDERAL INDEX IN TERM NEWBORN SIBLINGS
tors (11), and may have different postnatal morbidity and growth characteristics (610, 12). More recently, fetal ultrasound studies have shown the usefulness of the fetal ponderal index as a predictor of intrauterine growth retardation (13, 14) and fetal obesity (15). In addition, fetal growth indices (ratios of fetal abdominal circumference to femur length, and thigh circumference to femur length) have been suggested to correlate better with neonatal ponderal index than with birth weight or skinfold thickness (16). The tendency of birth weight to correlate in successive siblings has been documented (17, 18). In addition, low birth weight, prematurity, and intrauterine growth retardation seem to cluster in sibships and families (19-21). Such familial clustering may be related to genetic and/or maternal factors affecting intrauterine growth and timing of delivery. Because of the etiologic heterogeneity between prematurity and intrauterine growth retardation (22) and among different types of intrauterine growth retardation (4, 12), it is important to study the behavior of the ponderal index in successive births. We present results of an analysis of the correlation of ponderal index in term newborn siblings ascertained as part of a nationwide health study of male US Army veterans. METHODS
Study population Details of the study design, population, and data collection are published elsewhere (23, 24). Briefly, infants in this study were fathered by 1,677 male US Army veterans who participated in a nationwide health study. Medical records from the hospital of birth were obtained for all children fathered by these men. All children were born between 1966 and 1986. Information abstracted from medical records included maternal data, such as age, gravidity, pregnancy, and medical complications, and infant data, including birth weight, gestational age, length, and other medical information.
577
Full sibships were reconstructed using a computer algorithm that used maternal and paternal identifying information. All paternal half-sibs (different last name of mother) were excluded from the analysis. The final study population included 2,568 live singleton term (>37 weeks) infants. Of 3.369 infants originally available for this study, the following were excluded: 40 (1.2 percent) with unrecorded birth weight; 273 (8.1 percent) with unrecorded length; eight (0.2 percent) stillbirths; 180 (5.3 percent) preterm infants ( the 90th percentile. These characteristics included year of birth (four groups: 1980), maternal age (five groups: 35, not recorded), gravidity (i.e., pregnancy order which is equal to the total number of prior pregnancies plus one; four groups: 1, 2, >3, not recorded), pregnancy complications (two groups: yes, no), prior maternal illness (two groups: yes, no), sex (two groups: male, female), and gestational age in weeks (four groups: 37-38, 39-41, 42-43, not recorded). Analysis of variance was used to examine differences in mean ponderal index, and the x 2 test for homogeneity was used to examine differences in the proportion of infants with ponderal index < the 10th percentile and ponderal index > the 90th percentile in these groups.
578
KHOURY ET AL.
To examine patterns of sibling recurrence in ponderal index, the proportion of infants with ponderal index < the 10th percentile, and ponderal index > the 90th percentile, we first examined the birth order composition of the 2,568 infants; 1,474 were first births, 795 were second births, 253 were third births, and 46 were fourth or higher births. Here, birth order refers to the order of the child among all full siblings (same father and mother). Because of lack of statistical independence in observations on siblings, we limited our analysis to the 795 second births for whom we had information on ponderal index and other factors in the prior sib. Linear regression was performed on the ponderal index in the second sib as a function of the ponderal index in the first sib. In addition, logistic regression analysis was used to predict the odds of the second sib having a ponderal index < the 10th percentile and ponderal index > the 90th percentile as a function of the ponderal index in the first sib. For both linear and logistic models, we proceeded from a crude analysis, in which ponderal index in the first sib was the only independent variable, to models that included other potentially confounding factors such as race, year of birth, maternal age, gravidity, maternal complications, maternal illness, sex, gestational age, and birth interval. RESULTS
ethnic groups (Hispanic and others) compared with that in whites (2.49 vs. 2.57), and lower in infants born at 37-38 weeks compared with that in those born between 39 and 41 weeks (2.52 vs. 2.56). The proportion of babies with ponderal index < the 10th percentile and ponderal index > the 90th percentile generally tended to follow the mean ponderal index. The ponderal index in second sibs by status of first sibs In table 2, we present the mean ponderal index and the proportion of infants with ponderal index < the 10th percentile and ponderal index > the 90th percentile in second sibs according to the ponderal index distribution of first sibs. Compared with 627 infants who had a prior sib with a ponderal index between the 10th and the 90th percentiles, the 92 infants who had a prior sib with ponderal index < the 10th percentile had a lower mean ponderal index (2.49 vs. 2.58), a higher risk of a ponderal index < the 10th percentile (13.0 percent vs. 8.5 percent), and a lower risk of a ponderal index > the 90th percentile (4.4 percent vs. 10.2 percent). On the other hand, 76 infants who had a prior sib with ponderal index > the 90th percentile had a higher mean ponderal index (2.69 vs. 2.58), a higher risk of a ponderal index > the 90th percentile (17.1 percent vs. 10.2 percent), and a lower risk of ponderal index < the 10th percentile (5.3 percent vs. 8.5 percent).
The ponderal index in the study population The 10th, 50th, and 90th percentiles of the ponderal index for the study population were 2.20, 2.55, and 2.95, respectively. In table 1, we present the effects of maternal and infant variables on the mean ponderal index and the proportion of babies with ponderal index < the 10th percentile and ponderal index > the 90th percentile. Because of issues of statistical independence, we use only the first sib from each sibship, but similar results were obtained for other sib orders. The mean ponderal index was lower in males compared with that in females (2.54 vs. 2.58), lower in other racial/
Linear and logistic regression analyses In table 3, we present the results of multiple linear regression of ponderal index and multiple logistic regression of the binary outcome (ponderal index < the 10th percentile) on a number of independent variables. Full models are shown with regression coefficients and standard errors for all independent variables. The most parsimonious model for the linear regression, however, included status of first sib, sex, gestational age, and gravidity, while the most parsimonious model for the logistic regression included status of first sib,
PONDERAL INDEX IN TERM NEWBORN SIBLINGS
579
TABLE l
Maternal and infant characteristics in relation to mean ponderal index (PI) and percentage of infants with PI < the 10th percentile and PI > the 90th percentile in first sibs (n = 1,474): veterans child health study, 1966-1986 Variable
Year of birth £1970 1971-1975 1976-1980 1981-1986
n
Mean PI
282 622 424 146
2.58 2.57 2.54 2.56
651 345 116 25 337
2.58 2.55 2.53 2.56 2.56
737 350 222 165
2.56 2.58 2.55 2.55
118
%OfPI the 90th percentile
9.2 9.7
12.1 10.0
13.0 12.3
7.3 8.9
Maternal age (years) £24
25-29 30-34 >35
NRt
8.6*'** 13.0 9.5
10.6 8.4 5.2
20.0 12.5
16.0
11.1 8.3
9.1 9.1
12.6 12.1
10.4 10.9
2.58 2.56
13.6 10.6
15.3**
2.56 2.56
12.0 10.7
9.8 9.5
2.54* 2.58
12.5**
8.3
9.0
10.7
9.5
Gravidity 1
2 >3 NR Pregnancy complications Yes No Maternal illness Yes
No Sex Male Female Race/ethnicity White Black Other NR
Gestational age (weeks) 37-38 39-41 42-43 NR
1,356
92
1,382
743 731
1,073 113 28 260
2.57 2.55 2.49*** 2.54
194 848 166 266
2.52****' 2.56 2.56 2.59
9.0
9.5
9.6
11.5 10.7 15.6
10.6
18.6**
8.3 9.1 9.0 12.0
9.0
12.7 9.8
7.1 8.9
2
*p< 0.05 using the Student t test; **p < 0.05 using the x test; ***p < 0.05 using the t test (with whites as the reference group); ****p < 0.05 using the Mantel-Haenszel test for trend (excluding nonrecordeds); ***** p < 0.05 using the Student t test (with 39-41 weeks as the reference group). t NR, not recorded.
gestational age, and gravidity. In both models, the adjusted regression coefficients of the ponderal index of the first sib were essentially unchanged from values of
regression coefficients in models in which the only independent variable was the ponderal index status of first sibs. In linear regression, a strong association was found
580
KHOURY ET AL. TABLE 2
Mean ponderal index and percentage of second sibs with PI < the 10th percentile and PI > the 90th percentile, by PI status of first sibs: veterans child health study, 1966-1986 Status of the second sib (%)**
PI status of first sibs (percentiles)
Total
PI*
90th percentile
„..».
90th
92 627 76
2.49 2.58 2.69
13.0 8.5 5.3
12 53 4
82.6 81.3 77.6
76 510 59
4.4 10.2 17.1
4 64 13
* Analysis of variance for homogeneity of means, p = 0.0001; simple linear regression of PI in the second sib over PI in the first sib, p < 0.0001; ** x 2 for homogeneity of the percentile distribution = 9.87, 4 df, p = 0.043; *** extended Mantel-Haenszel test for trend, X2MH = 3.29, p = 0.069; **** extended Mantel-Haenszel test for trend, X2MH = 7.56, p = 0.006.
TABLE 3
Multiple linear and logistic regression analyses of factors predicting the ponderal index (PI) in the second sib and the odds that the second sib has a PI < the 10th percentile: veterans child health study, 1966-1986
Variable
Linear regression (outcome — mean PI) Regression coefficient
Logistic regression (outcome = PI < the 10th percentile)
Standard error
Regression coefficient
Standard error
0.240*
0.035
-1.523*
0.477
Sex (male/female)
-0.045*
0.021
0.212
0.264
Race Black/white Other/white NRf/white
0.080 -0.045 -0.046
0.044 0.066 0.027
-0.038 1.093 0.405
0.568 0.617 0.315
Gestational age (weeks) 30/
Vol. 132, No. 3
Copyright © 1990 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved
Printed in U.S.A.
THE PONDERAL INDEX IN TERM NEWBORN SIBLINGS MUIN J. KHOURY, CYNTHIA J. BERG, AND EUGENIA E. CALLE Khoury, M. J. (CDC, Atlanta, GA 30333), C. J. Berg, and E. E. Calle. The ponderal index in term newborn siblings. Am J Epidemiol 1990;132:576-83. Rohrer's ponderal index in newborns (birth weight/height3 x 100) has been used as an indicator of fetal growth status, especially to assess asymmetrical intrauterine growth retardation. Because low birth weight and intrauterine growth retardation tend to recur in sibships, we examined patterns of sibling correlation in the ponderal index in 795 live term (>37 weeks) singleton sibling pairs without birth defects born between 1966 and 1986 and fathered by male US Army veterans participating in a nationwide health study. Data on birth weight, length, gestational age, and other maternal and infant health characteristics were abstracted from hospital-of-birth medical records. The correlation coefficient of ponderal index in sib pairs was 0.24 (p < 0.001). Compared with 627 infants who had a prior sib with a ponderal index between the 10th and 90th percentiles, 92 infants who had a prior sib with ponderal index < the 10th percentile had a lower mean ponderal index and a higher proportion with ponderal index < the 10th percentile (13.0% vs. 8.5%). On the other hand, 76 infants who had a prior sib with ponderal index > the 90th percentile had a higher mean ponderal index and higher proportion with ponderal index > the 90th percentile (17.1% vs. 10.2%). The clustering of ponderal index in siblings persisted after controlling for factors such as race, gender, maternal age, gravidity, year of birth, gestational age, pregnancy complications, and prior maternal illnesses. The findings point to the presence of genetic and/or maternal factors affecting the growth status of term newborn infants. The significance of the ponderal index needs to be examined in future genetic and epidemiologic studies of intrauterine growth. fetal growth retardation; genetics
Rohrer's ponderal index in newborns refers to the ratio of birth weight (g) to the cube of the length (cm3) X 100 and has gained increasing attention in studies of low birth weight and intrauterine growth
retardation (1-12). The ponderal index quantifies the relation of the amount of soft tissue mass to skeletal development (1). Increasingly, researchers have recognized the pathophysiologic heterogeneity of intrauterine growth retardation (1, 4, 12). Received for publication May 9,1989, and in final T h e P°nderal index is being used to cateform March 16,1990. gorize growth-retarded infants into those From the Division of Birth Defects and Develop- w jth asymmetric growth retardation (low mental Disabilities, Center for Environmental Health
,
i • j
. ,, ,
...
.
and Injury Control, Centers for Disease Control, POnderal index, weight-for-gestational age Atlanta, GA. proportionately more affected than height) Reprint requests to Dr. Muin J. Khoury, Birth a n ( j symmetric growth retardation (normal Defects and Genetic Diseases Branch (F37), Center , , . , , ,, . , , ,, . ,, for Environmental Health and Injury Control, Centers P°nderal index, both weight and height affor Disease Control, Atlanta, GA 30333. fected) (4). These two subtypes of intraData for this study were gathered under an inter- uterine growth retardation may have difagency agreement with the Veterans Administration. f/e r •e n ,t ,t l m• m• • i n .i_ • i_ The authors thank Essie Fuller for her help in g pathogenesis, may be preparing the manuscript. related to different underlying etiologic fac576
PONDERAL INDEX IN TERM NEWBORN SIBLINGS
tors (11), and may have different postnatal morbidity and growth characteristics (610, 12). More recently, fetal ultrasound studies have shown the usefulness of the fetal ponderal index as a predictor of intrauterine growth retardation (13, 14) and fetal obesity (15). In addition, fetal growth indices (ratios of fetal abdominal circumference to femur length, and thigh circumference to femur length) have been suggested to correlate better with neonatal ponderal index than with birth weight or skinfold thickness (16). The tendency of birth weight to correlate in successive siblings has been documented (17, 18). In addition, low birth weight, prematurity, and intrauterine growth retardation seem to cluster in sibships and families (19-21). Such familial clustering may be related to genetic and/or maternal factors affecting intrauterine growth and timing of delivery. Because of the etiologic heterogeneity between prematurity and intrauterine growth retardation (22) and among different types of intrauterine growth retardation (4, 12), it is important to study the behavior of the ponderal index in successive births. We present results of an analysis of the correlation of ponderal index in term newborn siblings ascertained as part of a nationwide health study of male US Army veterans. METHODS
Study population Details of the study design, population, and data collection are published elsewhere (23, 24). Briefly, infants in this study were fathered by 1,677 male US Army veterans who participated in a nationwide health study. Medical records from the hospital of birth were obtained for all children fathered by these men. All children were born between 1966 and 1986. Information abstracted from medical records included maternal data, such as age, gravidity, pregnancy, and medical complications, and infant data, including birth weight, gestational age, length, and other medical information.
577
Full sibships were reconstructed using a computer algorithm that used maternal and paternal identifying information. All paternal half-sibs (different last name of mother) were excluded from the analysis. The final study population included 2,568 live singleton term (>37 weeks) infants. Of 3.369 infants originally available for this study, the following were excluded: 40 (1.2 percent) with unrecorded birth weight; 273 (8.1 percent) with unrecorded length; eight (0.2 percent) stillbirths; 180 (5.3 percent) preterm infants ( the 90th percentile. These characteristics included year of birth (four groups: 1980), maternal age (five groups: 35, not recorded), gravidity (i.e., pregnancy order which is equal to the total number of prior pregnancies plus one; four groups: 1, 2, >3, not recorded), pregnancy complications (two groups: yes, no), prior maternal illness (two groups: yes, no), sex (two groups: male, female), and gestational age in weeks (four groups: 37-38, 39-41, 42-43, not recorded). Analysis of variance was used to examine differences in mean ponderal index, and the x 2 test for homogeneity was used to examine differences in the proportion of infants with ponderal index < the 10th percentile and ponderal index > the 90th percentile in these groups.
578
KHOURY ET AL.
To examine patterns of sibling recurrence in ponderal index, the proportion of infants with ponderal index < the 10th percentile, and ponderal index > the 90th percentile, we first examined the birth order composition of the 2,568 infants; 1,474 were first births, 795 were second births, 253 were third births, and 46 were fourth or higher births. Here, birth order refers to the order of the child among all full siblings (same father and mother). Because of lack of statistical independence in observations on siblings, we limited our analysis to the 795 second births for whom we had information on ponderal index and other factors in the prior sib. Linear regression was performed on the ponderal index in the second sib as a function of the ponderal index in the first sib. In addition, logistic regression analysis was used to predict the odds of the second sib having a ponderal index < the 10th percentile and ponderal index > the 90th percentile as a function of the ponderal index in the first sib. For both linear and logistic models, we proceeded from a crude analysis, in which ponderal index in the first sib was the only independent variable, to models that included other potentially confounding factors such as race, year of birth, maternal age, gravidity, maternal complications, maternal illness, sex, gestational age, and birth interval. RESULTS
ethnic groups (Hispanic and others) compared with that in whites (2.49 vs. 2.57), and lower in infants born at 37-38 weeks compared with that in those born between 39 and 41 weeks (2.52 vs. 2.56). The proportion of babies with ponderal index < the 10th percentile and ponderal index > the 90th percentile generally tended to follow the mean ponderal index. The ponderal index in second sibs by status of first sibs In table 2, we present the mean ponderal index and the proportion of infants with ponderal index < the 10th percentile and ponderal index > the 90th percentile in second sibs according to the ponderal index distribution of first sibs. Compared with 627 infants who had a prior sib with a ponderal index between the 10th and the 90th percentiles, the 92 infants who had a prior sib with ponderal index < the 10th percentile had a lower mean ponderal index (2.49 vs. 2.58), a higher risk of a ponderal index < the 10th percentile (13.0 percent vs. 8.5 percent), and a lower risk of a ponderal index > the 90th percentile (4.4 percent vs. 10.2 percent). On the other hand, 76 infants who had a prior sib with ponderal index > the 90th percentile had a higher mean ponderal index (2.69 vs. 2.58), a higher risk of a ponderal index > the 90th percentile (17.1 percent vs. 10.2 percent), and a lower risk of ponderal index < the 10th percentile (5.3 percent vs. 8.5 percent).
The ponderal index in the study population The 10th, 50th, and 90th percentiles of the ponderal index for the study population were 2.20, 2.55, and 2.95, respectively. In table 1, we present the effects of maternal and infant variables on the mean ponderal index and the proportion of babies with ponderal index < the 10th percentile and ponderal index > the 90th percentile. Because of issues of statistical independence, we use only the first sib from each sibship, but similar results were obtained for other sib orders. The mean ponderal index was lower in males compared with that in females (2.54 vs. 2.58), lower in other racial/
Linear and logistic regression analyses In table 3, we present the results of multiple linear regression of ponderal index and multiple logistic regression of the binary outcome (ponderal index < the 10th percentile) on a number of independent variables. Full models are shown with regression coefficients and standard errors for all independent variables. The most parsimonious model for the linear regression, however, included status of first sib, sex, gestational age, and gravidity, while the most parsimonious model for the logistic regression included status of first sib,
PONDERAL INDEX IN TERM NEWBORN SIBLINGS
579
TABLE l
Maternal and infant characteristics in relation to mean ponderal index (PI) and percentage of infants with PI < the 10th percentile and PI > the 90th percentile in first sibs (n = 1,474): veterans child health study, 1966-1986 Variable
Year of birth £1970 1971-1975 1976-1980 1981-1986
n
Mean PI
282 622 424 146
2.58 2.57 2.54 2.56
651 345 116 25 337
2.58 2.55 2.53 2.56 2.56
737 350 222 165
2.56 2.58 2.55 2.55
118
%OfPI the 90th percentile
9.2 9.7
12.1 10.0
13.0 12.3
7.3 8.9
Maternal age (years) £24
25-29 30-34 >35
NRt
8.6*'** 13.0 9.5
10.6 8.4 5.2
20.0 12.5
16.0
11.1 8.3
9.1 9.1
12.6 12.1
10.4 10.9
2.58 2.56
13.6 10.6
15.3**
2.56 2.56
12.0 10.7
9.8 9.5
2.54* 2.58
12.5**
8.3
9.0
10.7
9.5
Gravidity 1
2 >3 NR Pregnancy complications Yes No Maternal illness Yes
No Sex Male Female Race/ethnicity White Black Other NR
Gestational age (weeks) 37-38 39-41 42-43 NR
1,356
92
1,382
743 731
1,073 113 28 260
2.57 2.55 2.49*** 2.54
194 848 166 266
2.52****' 2.56 2.56 2.59
9.0
9.5
9.6
11.5 10.7 15.6
10.6
18.6**
8.3 9.1 9.0 12.0
9.0
12.7 9.8
7.1 8.9
2
*p< 0.05 using the Student t test; **p < 0.05 using the x test; ***p < 0.05 using the t test (with whites as the reference group); ****p < 0.05 using the Mantel-Haenszel test for trend (excluding nonrecordeds); ***** p < 0.05 using the Student t test (with 39-41 weeks as the reference group). t NR, not recorded.
gestational age, and gravidity. In both models, the adjusted regression coefficients of the ponderal index of the first sib were essentially unchanged from values of
regression coefficients in models in which the only independent variable was the ponderal index status of first sibs. In linear regression, a strong association was found
580
KHOURY ET AL. TABLE 2
Mean ponderal index and percentage of second sibs with PI < the 10th percentile and PI > the 90th percentile, by PI status of first sibs: veterans child health study, 1966-1986 Status of the second sib (%)**
PI status of first sibs (percentiles)
Total
PI*
90th percentile
„..».
90th
92 627 76
2.49 2.58 2.69
13.0 8.5 5.3
12 53 4
82.6 81.3 77.6
76 510 59
4.4 10.2 17.1
4 64 13
* Analysis of variance for homogeneity of means, p = 0.0001; simple linear regression of PI in the second sib over PI in the first sib, p < 0.0001; ** x 2 for homogeneity of the percentile distribution = 9.87, 4 df, p = 0.043; *** extended Mantel-Haenszel test for trend, X2MH = 3.29, p = 0.069; **** extended Mantel-Haenszel test for trend, X2MH = 7.56, p = 0.006.
TABLE 3
Multiple linear and logistic regression analyses of factors predicting the ponderal index (PI) in the second sib and the odds that the second sib has a PI < the 10th percentile: veterans child health study, 1966-1986
Variable
Linear regression (outcome — mean PI) Regression coefficient
Logistic regression (outcome = PI < the 10th percentile)
Standard error
Regression coefficient
Standard error
0.240*
0.035
-1.523*
0.477
Sex (male/female)
-0.045*
0.021
0.212
0.264
Race Black/white Other/white NRf/white
0.080 -0.045 -0.046
0.044 0.066 0.027
-0.038 1.093 0.405
0.568 0.617 0.315
Gestational age (weeks) 30/
