Iwama et al. Supplementary information Page 1
SUPPLEMENTARY INFORMATION
MATERNAL CLINIC AND HOME BLOOD PRESSURE MEASUREMENTS
DURING PREGNANCY AND INFANT BIRTH WEIGHT: THE BOSHI STUDY
Noriyuki Iwama1,4, Hirohito Metoki2,4, Takayoshi Ohkubo3, Mami Ishikuro2, Taku Obara2,
Masahiro Kikuya2, Katsuyo Yagihashi5, Hidekazu Nishigori4, Takashi Sugiyama4, Junichi
Sugawara2,4, Nobuo Yaegashi4, Kazuhiko Hoshi5, Masakuni Suzuki5, Shinichi Kuriyama6,
Yutaka Imai7, and The BOSHI Study Group
1
Tohoku University Graduate School of Medicine, Sendai, Japan
2
Tohoku Medical Megabank Organization, Tohoku University, Sendai, Japan
3
Department of Hygiene and Public Health, Teikyo University School of Medicine, Tokyo,
Japan
4
Obstetrics and Gynecology, Tohoku University, Sendai, Japan
Suzuki Memorial Hospital, Iwanuma, Japan
6
Division of Molecular Epidemiology, Tohoku University Graduate School of Medicine,
Sendai, Japan
7
Department of Planning for Drug Development and Clinical Evaluation, Tohoku University
Graduate School of Pharmaceutical Sciences, Sendai, Japan
5
Correspondence to: Hirohito Metoki MD, PhD
Tohoku Medical Megabank Organization, Tohoku University
1-1 Seiryomachi, Sendai 980-8574, Japan
Tel: +81-22-717-8104; Fax: +81-22-717-8106
E-mail: hmetoki@med.tohoku.ac.jp
Iwama et al. Supplementary information Page 2
Data collection
Maternal age at conception, parity, last menstrual period, expected date of confinement
(confirmed by fetal crown-rump length measured by ultrasonography before 12 weeks’
gestation), gestational age, body weight at every prenatal checkup, delivery week, maternal
disease that existed before pregnancy, pregnancy complications in prior and current
pregnancy, and information about infants including birth weight and sex were collected from
medical records. In addition, maternal pre-pregnancy height, body weight, smoking status (no
smoking before conception, smoking until conception was recognized, or smoking during
pregnancy), alcohol intake (no alcohol intake before conception, alcohol intake until
conception was recognized, or alcohol intake during pregnancy), and family history of
hypertension were collected by questionnaire and survey by midwives. Gestational weight
gain (GWG) was calculated by subtracting pre-pregnancy body weight from weight at the last
prenatal checkup before delivery. The diagnosis of hypertensive disorders in pregnancy
(HDP) was based on the criteria of the International Society for the Study of Hypertension in
Pregnancy (ISSHP).1 The season of conception was classified as follows: spring (March,
April, and May), summer (June, July, and August), autumn (September, October, and
November), and winter (December, January, and February). The date of conception was back
calculated and defined as 2 weeks and 0 days from the expected date of confinement.
Statistical analysis
We assessed differences in maternal and neonatal characteristics among study subjects and
women excluded due to missing CBP or HBP data. Differences in continuous variables
between both groups were tested by the student t-test. Differences in categorical variables
between both groups were tested by chi-square test or Fisher exact test, as appropriate.
In this study, associations of maternal CBP and HBP with infant’s birth weight were
examined using the proportional odds model with possible confounding factors, as opposed
to using the multivariate linear regression model.2 The reason this method was chosen was
because the partial regression coefficient derived from multivariate linear regression model
might not always reflect clinical importance, whereas the proportional odds model can be
applied while considering clinical significance.3 In terms of birth weight, several studies have
reported continuous inverse relationships between birth weight and future coronary risk
factors,4 type 2 diabetes,5 and preeclampsia;6 in these studies, birth weight categories were
divided by every 500 g.4-6 Therefore, ordinal categories of infant’s birth weight were as
follows: ≥3500 g, 3000-3499 g, 2500-2999 g, and <2500 g. These values were considered
clinically meaningful and were treated as the lower ordered dependent variable in this study.
Two infants whose birth weight was below 2000 g were included in the category of <2500 g.
Six infants whose birth weight was ≥4000 g were included in the category of ≥3500 g. To
avoid overfitting in the proportional odds model, we calculated the limit of the number of
Iwama et al. Supplementary information Page 3
explanatory variables using the following formula: (n – n-2
ni3 ) / (10 or 20).7 The n and
k indicate total sample size and the number of categories of ordinal dependent variables,
respectively. The ni also indicates the sample size in each category of ordinal dependent
variables. As a result, the limiting number of explanatory variables in this study ranged from
26 to 52. In all analyses, the number of explanatory variables was below 26.
First, CBP and HBP were included into the proportional odds model separately. After each
BP was divided into quartiles, the linearity of the above associations was tested. After that, 1
standard deviation (SD) of both CBP and HBP were included in the model as continuous
variables simultaneously to compare the effect size among CBP and HBP. In addition, the
likelihood ratio test was used to compare the goodness of fit between the models in which
CBP or HBP were included separately and in which both CBP and HBP were included
simultaneously.3 All analyses were adjusted by maternal age at conception (<20 years,
20-34.9 years, and ≥35 years), pre-pregnancy body-mass index (BMI), gestational weight
gain (GWG), parity (primipara or not), history of pregnancy-induced hypertension (PIH; yes
or no), smoking status (no smoking before conception, smoking until conception was
recognized, or smoking during pregnancy), alcohol intake (no alcohol intake before
conception, alcohol intake until conception was recognized, or alcohol intake during
pregnancy), family history of hypertension (yes or no), gestational diabetes mellitus (GDM;
yes or no), season of conception (winter was treated as the reference category), delivery week,
infant’s sex, and gestational week at BP measurement. In the analysis for the association of
HBP with infant birth weight, mean gestational week at HBP measurement was included as a
covariate. HDP was not included in the model as a covariate because HDP would be an
intervening variable.
The multicollinearity among explanatory variables was suspected when tolerance was
below 0.25 using multivariate linear regression analysis in which ordinal categories of birth
weight were included as continuous dependent variables and the same explanatory variables
were included as in the proportional odds model.8 The proportional odds assumption was
confirmed by the score test.8 Statistical significance was defined as a two-sided p-value <
0.05. Non-significant p-values in the score test indicated that the proportional odds
assumption was not rejected statistically.8 All statistical analyses were performed with SAS
version 9.4 (SAS Institute Inc., Cary, NC, USA).
Iwama et al. Supplementary information Page 4
Reason for exclusion from analysis
Among 752 women enrolled, miscarriage (6 women), still birth (3 women), and women
who declined the consent (3 women) were excluded. Ten women were transferred to other
hospitals due to premature threatened labor (3 women), diabetes mellitus (1 woman),
ulcerative colitis (1 woman), history of cardiac surgery for ventricular septal defect (1
woman), and intrauterine growth restriction (1 woman). Among women who developed
HDP, 3 women were transferred to other hospitals because obstetricians
determined that their condition was difficult to manage in their hospital. Those 3
women with HDP were excluded from this analysis because perinatal outcomes
could not be collected. Eight women returned to their parents’ home for delivery. Three
women whose CBP were missing were excluded. Furthermore, 114 women were excluded
because home BP was not measured within 7 days of CBP measurement.
Differences in maternal and neonatal characteristics among study subjects and women
excluded due to missing CBP or HBP data
Among study subjects and women excluded due to missing either CBP or HBP values, the
following characteristics differed significantly: maternal age (31.7 [4.7] years vs. 29.5 [5.5]
years, respectively, p<0.0001), the proportion of women whose age was <20 years (0.8% vs.
3.4%, respectively, p=0.04), the proportion of HDP (8.9% vs. 16.2%, respectively, p=0.02),
PIH (7.4% vs. 13.7%, respectively, p=0.03), gestational hypertension (5.8% vs. 11.1%,
respectively, p=0.03), infant’s birth weight (3057 [391] g vs. 3136 [396] g, respectively,
p=0.048), and the proportion of infant’s birth weight <2500g (7.4% vs. 1.7%, respectively,
p=0.02) and ≥3500g (12.4% vs. 20.5%, respectively, p=0.02). No other characteristics
differed significantly among these groups (Supplementary Table 1).
Comparison of effects of SBP versus DBP on infant birth weight
We also performed additional analyses to compare whether SBP or DBP was more strongly
associated with lower categories of infant birth weight. Both SBP and DBP were included in
the proportional odds model simultaneously. As a result, the statistical significance of DBP
remained although that of SBP did not (Supplementary Table 2). The likelihood ratio tests
that were conducted for CBP and HBP showed significant improvement of the goodness of fit
when DBP was included in addition to SBP, whereas the inverse was not true (Supplementary
Table 3). The comparison among MAP and SBP or DBP could not be performed because
multicollinearity was suspected.
Iwama et al. Supplementary information Page 5
References
1 Brown MA, Lindheimer MD, de Swiet M, Van Assche A, Moutquin JM. The classification
and diagnosis of the hypertensive disorders of pregnancy: statement from the International
Society for the Study of Hypertension in Pregnancy (ISSHP). Hypertens Pregnancy.
2001;20(1):Ix-xiv.
2 McCullagh. P. Regression models for ordinal data. Journal of the Royal Statistical Society
Series B (Methodological). 1980;42(2):109-142.
3 Norris CM, Ghali WA, Saunders LD, Brant R, Galbraith D, Faris P, Knudtson ML,
Investigators A. Ordinal regression model and the linear regression model were superior to
the logistic regression models. J Clin Epidemiol. 2006;59(5):448-456.
4 Barker DJ, Osmond C, Forsen TJ, Kajantie E, Eriksson JG. Trajectories of growth among
children who have coronary events as adults. N Engl J Med. 2005;353(17):1802-1809.
5 Tian JY, Cheng Q, Song XM, Li G, Jiang GX, Gu YY, Luo M. Birth weight and risk of type
2 diabetes, abdominal obesity and hypertension among Chinese adults. Eur J Endocrinol.
2006;155(4):601-607.
6 North RA, McCowan LM, Dekker GA, Poston L, Chan EH, Stewart AW, Black MA, Taylor
RS, Walker JJ, Baker PN, Kenny LC. Clinical risk prediction for pre-eclampsia in nulliparous
women: development of model in international prospective cohort. BMJ. 2011;342:d1875.
7 Frank E. Harrell Jr. Regression Modeling Strategies: With Applications to Linear Models,
Logistic Regression, and Survival Analysis. New York: Springer; 2010.
8 Katz MH. Multivariable Analysis: A Practical Guide for Clinicians and Public Health
Researchers Third Edition. Cambridge: Cambridge University Press; 2011.
Iwama et al. Supplementary information Page 6
Supplementary Table 1. Differences in maternal and neonatal characteristics among
study subjects and women excluded due to missing CBP or HBP measurements
Study
participants
Women
excluded
due to
missing BP
values
605
117
31.7 (4.7)
0.8
29.5 (5.5)
3.4
< 0.0001
0.04*
72.7
26.5
158.3 (5.2)
54.4 (9.1)
21.7 (3.5)
10.2 (3.9)
59.0
4.0
79.5
17.1
158.5 (5.2)
56.2 (9.9)
22.3 (3.7)
10.8 (4.8)
53.9
2.6
0.1
0.04
0.4
0.06
0.1
0.1
0.3
0.6*
No smoking before conception
Until conception was recognized
Smoking during pregnancy
Alcohol intake (%)
No alcohol intake before conception
Until conception was recognized
Alcohol intake during pregnancy
Season of conception (%)
Spring
Summer
84.3
11.7
4.0
82.9
12.8
4.3
0.7
0.7
0.8*
53.5
44.8
1.7
62.4
37.6
0.0
0.09
0.2
0.2*
22.5
19.8
16.2
23.1
0.1
0.4
Autumn
Winter
Family history of hypertension (%)
Hypertensive disorders in pregnancy (%)
Chronic hypertension†
PIH
Gestational hypertension
Preeclampsia
28.3
29.4
37.0
8.9
1.7
7.4
5.8
1.7
28.2
32.5
35.9
16.2
3.4
13.7
11.1
2.6
1.0
0.5
0.8
0.02
0.3*
0.03
0.03
0.09*
Characteristics
No. of subjects
Maternal characteristics
Age (years)
<20 years (%)
20-34.9 years (%)
≥35 years (%)
Height (cm)
Pre-pregnancy weight (kg)
Pre-pregnancy BMI (kg/m2)
Gestational weight gain (kg)
Primipara (%)
History of PIH (%)
Smoking (%)
p-value
Iwama et al. Supplementary information Page 7
Superimposed preeclampsia
Placental abruption
GDM (%)
SLE (%)
Neonatal characteristics
Sex (male/female) (%)
Gestational age at delivery, week
Preterm delivery (%)
SGA‡ (%)
Birth weight, g
<2500 g (%)
2500-2999 g (%)
3000-3499 g (%)
≥3500 g (%)
0.5
0.9
0.5*
0.5
0.7
0.002
0.0
2.6
0.0
1.0*
0.09*
1.0*
51.4 / 48.6
39.7 (1.3)
2.6
6.5
3057 (391)
7.4
52.1 / 47.9
39.7 (1.2)
1.7
7.7
3136 (396)
1.7
0.9
0.8
0.8*
0.6
0.048
0.02
35.9
44.3
12.4
37.6
40.2
20.5
0.7
0.4
0.02
Data were expressed as mean (standard deviation) or percentages.
CBP, clinic blood pressure; HBP, home blood pressure; BMI, body-mass index; PIH,
pregnancy-induced hypertension; GDM, gestational diabetes mellitus; SLE, systemic lupus
erythematosus; SGA, small for gestational age.
*Fisher’s exact test
†
Chronic hypertension without antihypertensive drugs before 20 weeks’ gestation
‡
SGA was defined as a birth weight below the 10th percentile of Japanese infants.
Iwama et al. Supplementary information Page 8
Supplementary Table 2. Comparison of effects of SBP and DBP on infant birth weight*†
Model
Clinic BP
(per 1 SD§)
Home BP
(per 1 SD§)
SBP
Adjusted OR (95%CI)
DBP
Adjusted OR (95%CI)
P-value of
the score test‡
0.85 (0.68 – 1.07)
1.36 (1.08 – 1.71)
0.21
0.92 (0.71 – 1.19)
1.41 (1.10 – 1.80)
0.28
*All models were adjusted by maternal age, pre-pregnancy BMI, GWG, parity, history of PIH,
family history of hypertension, smoking status, alcohol intake, GDM, deliver week, infant’s
sex, season of conception, and gestational week at each BP measurement.
†
The comparison between MAP and SBP or DBP could not be conducted because of
multicollinearity.
‡
The score test for the proportional odds assumption. P ≥ 0.05 indicates that the proportional
odds assumption is not rejected statistically.
§
1 SD=12 mmHg for clinic SBP, 9 mmHg for clinic DBP, 9 mmHg for home SBP, and 7
mmHg for home DBP.
SBP, systolic blood pressure; DBP, diastolic blood pressure; BP, blood pressure; OR, odds
ratio; CI, confidence interval; BMI, body-mass index; GWG, gestational weight gain; PIH,
pregnancy-induced hypertension; GDM, gestational diabetes mellitus; SD, standard
deviation.
Iwama et al. Supplementary information Page 9
Supplementary Table 3. Comparison of the effect of SBP and DBP on infant birth
weight (the Likelihood ratio test)*†
-2 log likelihood
Likelihood ratio‡
P-value
Clinic BP
Model 1 (Only SBP, per 1 SD§)
Model 2 (Only DBP, per 1 SD§)
Model 3 (SBP and DBP, per 1 SD§)
1157.38
1152.68
1150.89
6.49
1.79
-
0.01
0.18
-
Home BP
Model 1 (Only SBP, per 1 SD§)
Model 2 (Only DBP, per 1 SD§)
1154.67
1147.85
7.25
0.43
0.007
0.51
1147.42
-
-
Model
Model 3 (SBP and DBP, per 1 SD§)
*Please refer the adjustments in Supplementary Table 2.
†
The comparison between MAP and SBP or DBP could not be conducted because of
multicollinearity.
‡
The difference between -2 log likelihood of model 3 and that of each model, respectively.
§
1 SD=12 mmHg for clinic SBP, 9 mmHg for clinic DBP, 9 mmHg for home SBP, and 7
mmHg for home DBP.
SBP, systolic blood pressure: DBP, diastolic blood pressure; SD, standard deviation.