Accepted Manuscript

Title: MATERNAL SERUM URIC ACID LEVELS AND

BLOOD PRESSURE DURING PREGNANCY: A
COMMUNITY-BASED COHORT STUDY

Authors: Guoli Zhou, Claudia Holzman, Zhehui Luo, Claire

Margerison-Zilko

PII: S0301-2115(18)30008-3
DOI: https://doi.org/10.1016/j.ejogrb.2018.01.008
Reference: EURO 10191

To appear in: EURO

Received date: 30-8-2017

Accepted date: 8-1-2018

Please cite this article as: Zhou Guoli, Holzman Claudia, Luo Zhehui,
Margerison-Zilko Claire.MATERNAL SERUM URIC ACID LEVELS AND
BLOOD PRESSURE DURING PREGNANCY: A COMMUNITY-BASED COHORT
STUDY.European Journal of Obstetrics and Gynecology and Reproductive Biology
https://doi.org/10.1016/j.ejogrb.2018.01.008

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 MATERNAL SERUM URIC ACID LEVELS AND BLOOD PRESSURE DURING

PREGNANCY: A COMMUNITY-BASED COHORT STUDY

Guoli Zhou a, MPH, MD, PhD; Claudia Holzman,b*, DVM, MPH, PhD; Zhehui Luo b, MS,

PhD; Claire Margerison-Zilko b, MPH, PhD

T
IP
a
Biomedical Research Informatics Core, Clinical & Translational Sciences Institute,

R
Michigan State University, East Lansing, Michigan, USA; bDepartment of Epidemiology

SC
and Biostatistics, College of Human Medicine, Michigan State University, East Lansing,

Michigan, USA
U
N
*Corresponding Author:
A

Claudia Holzman, DVM, MPH, PhD, Department of Epidemiology and Biostatistics,

M

College of Human Medicine, Michigan State University, 909 Fee Road, Room B601,
D

East Lansing, Michigan 48824, USA. E-mail: holzman@msu.edu.

TE
EP
CC
A

1
Abstract

Objective Studies find both very low and high serum uric acid (UA) levels are related to

oxidative stress and to conditions such as cardiovascular diseases and chronic kidney

disease in the general population. Pregnancy studies have focused only on high

maternal UA. In present study, we tested whether unusually high and low levels of

T
maternal serum UA are associated with increases in blood pressure (BP) during

IP
pregnancy.

R
Study Design The Pregnancy Outcomes and Community Health Study enrolled 3,019

SC
pregnant women between their 16th-27th week of pregnancy from 52 clinics in 5

Michigan communities (1998-2004). UA levels were measured in maternal blood

U
collected at enrollment from a sub-cohort of 1,223 participants. BP was abstracted from
N
prenatal medical records; these analyses used highest recorded diastolic BP (DBP) and
A

its companion systolic BP (SBP). Mean arterial pressure (MAP) was calculated using
M

the formula of (2×DBP + SBP)/3. Covariates, including maternal race/ethnicity, age at

D

enrollment, education level, medical insurance status, body mass index before
TE

pregnancy, parity, smoking during pregnancy, alcohol use during pregnancy, and

gestational week at blood collection, were considered as potential confounding

EP

variables. Associations between UA levels and BP were evaluated with linear spline or

multiple linear regression models. Models’ robustness was examined with bootstrap
CC

estimation of variance, sensitivity analysis, and 10-fold cross-validation.

Results Both DBP and MAP had a J-shaped relationship with maternal UA; the
A

breakpoints (nadirs) were 0.153 and 0.161 mmol/L UA, respectively. For DBP versus

UA, adjusted regression coefficient (β)= ̶ 95.67 (standard error (SE)=37.67 and p=0.01)

2
for the left and adjusted β=48.95 (SE=9.56 and p<0.01) for the right; for MAP versus

UA, adjusted β= ̶ 58.48 (SE=31.42 and p=0.06) for the left and adjusted β=52.23

(SE=11.39 and p<0.01) for the right. Maternal SBP followed a positive linear trend with

UA levels (adjusted β=37.75, SE=12.93, and p<0.01). All results were robust.

Conclusion Extreme high and low maternal serum UA levels may be informative in

T
studying maternal blood pressure during pregnancy.

IP
Key Words: maternal blood pressure; serum uric acid; linear spline; pregnancy

R
SC
U
N
A
M
D
TE
EP
CC
A

3
Introduction

In humans, uric acid (UA) is an end product from the catabolism of purines.

Studies find both very low and high serum uric acid (UA) levels are related to oxidative

stress [1,2,3] and to conditions such as cardiovascular diseases4 and chronic kidney

disease5 in the general population. A positive relationship between hyperuricemia (i.e.,

T
high UA) and gestational hypertension has been reported [6,7,8] These pregnancy

IP
studies focused exclusively on high maternal UA, though there is reason to suspect that

R
unusually low maternal UA also may be clinically relevant. Such a speculation is

SC
supported by the facts that both high and low serum UA levels are related to increased

oxidative stress in in-vivo studies9 and that increased maternal oxidative stress is linked
U
to hypertensive disorders in pregnant women [10,11,12,13,14].
N
In the present study we hypothesized that maternal blood pressure (BP) would
A

be higher at both extremes of UA levels (a non-linear relationship). We used data from

M

the prospective Pregnancy Outcomes and Community Health (POUCH) Study which
D

measured serum UA levels in mid-pregnancy. Our goal was to examine, in greater

TE

detail, the relationship between maternal UA levels and BP, as a step towards

maximizing the usefulness of maternal UA measures in research and clinical settings.

EP

Materials and methods

CC

The POUCH Study is a prospective cohort study designed primarily to examine

pathways to preterm delivery. The cohort included 3,019 pregnant women recruited
A

between their 16th-27th week of pregnancy from 52 clinics in 5 Michigan communities

(1998-2004). Inclusion criteria included no known congenital anomaly, maternal age≥15

4
years, prenatal screening of maternal serum alpha-fetoprotein (AFP) at the 15th-22nd

week of pregnancy, no pre-pregnancy history of diabetes mellitus, and competency in

English [15].

A sub-cohort of 1,371 participants was sampled with the inclusion of all pre-term

deliveries (<37 gestational weeks), all term deliveries with high maternal serum AFP,

T
and a race-stratified random sample of term deliveries with normal maternal serum AFP

IP
including oversampling of African-Americans [15]. Within the sub-cohort, we excluded

R
women who had: 1) no available prenatal records for abstraction of diastolic blood

SC
pressure (DBP) and systolic blood pressure (SBP), n=30; 2) no measure of maternal

serum UA levels in blood collected at enrollment, n=61; 3) a diagnosis of chronic

U
hypertension, defined as having a preexisting diagnosis of chronic hypertension or
N
antihypertensive treatment or having a SBP >140 mm Hg or DBP >90 mm Hg before 20
A

weeks of gestation, n=42; or 4) a diagnosis of renal disease including pyelonephritis,

M

glomerulonephritis, and renal diseases secondary to systemic disease, n=15). The

D

remaining 1,223 pregnant women from the sub-cohort were included in the final
TE

analysis. Sampling weights were calculated to account for the cohort and sub-cohort

sampling protocol and missing measurements. The POUCH Study was approved by the
EP

institutional review board (IRB) of the Michigan State University and the participants

gave informed consent. All data used in the current study were de-identified based on
CC

the Health Insurance Portability and Accountability Act (HIPAA) Security Guidelines.

We abstracted prenatal records and, for these analyses, selected the highest
A

DBP and its companion systolic SBP in order to calculate MAP. These measures were

used to calculate mean arterial pressure (MAP), i.e., (2×DBP + SBP)/3. Prenatal

5
records also were used to identify women with pregnancy-related hypertension

disorders, i.e. SBP >140 mmHg or DBP >90 mmHg beginning at > 20 weeks’ gestation

without proteinuria (gestational hypertension) or with proteinuria (preeclampsia).

Maternal serum UA (mmol/L) was measured at the time of the enrollment, i.e., in the

16th-27th week of pregnancy, at the Heinz Chemistry and Nutrition Laboratory in the

T
Department of Epidemiology at the University of Pittsburgh using the method developed

IP
by Fossati et al [16].

R
Maternal covariates included: race/ethnicity (classified into black and non-black),

SC
age at enrollment (dichotomized at < and ≥30 years), education level (dichotomized at

< and ≥ 12th grade completed), medical insurance status (Medicaid vs. non-Medicaid),
U
body mass index (BMI) before pregnancy (dichotomized as obese (BMI≥30) and non-
N
obese (BMI<30)), parity (nulliparity vs. ≥1 parity), smoking during pregnancy (yes/no),
A

alcohol use during pregnancy (yes/no), and gestational week at blood collection
M

(dichotomized at < and ≥ 20 weeks) were considered as potential confounding

D

variables. Gestational age at enrollment was calculated using the date of the last
TE

menstrual period (LMP). If a gestational age estimate from early ultrasound differed

from the LMP-based estimate by more than two weeks, then the ultrasound estimated
EP

gestational age was used.

Statistics: Outcomes (DBP, SBP, and MAP) and exposure (maternal serum UA
CC

level) were treated as continuous variables. In order to simplify the analysis of non-

linear relationships between outcomes and exposure, all covariates were categorized as
A

binary and are presented as percentages (Table 1). For univariate analyses, a t-test

was used to examine whether the outcome and/or exposure were associated with each

6
potential confounding variable. For multivariate analyses, a linear spline or a multiple

linear regression model was applied based on locally weighted smoothing curve. For

the linear spline model, the breakpoint value was refined using a nonlinear least-

squares estimation [17]. The problem of model overfitting was investigated through a

10-fold cross-validation procedure [18]. The model robustness was further estimated

T
with a bootstrap procedure [19] and a sensitivity analysis that removedwomen with

IP
preeclampsia or gestational hypertension (including antihypertensive treatment after 20

R
weeks of gestation). The multivariate analyses are detailed in Online Supporting

SC
Information. The statistical significance level, α, was set as 0.05 for a two-sided test.

Data management was conducted with SAS v9.4 (SAS Institute, Cary, North Carolina)
U
and all statistical analyses were weighted for the POUCH sampling design and carried
N
out with STATA v13.0 (StataCorp LP, College Station, Texas).
A
M

Results
D

As shown in Table 1, our analytic sample included 1,223 pregnant women,

TE

24.6% of whom were African-American, 27.6% >30 years old, 18.7% with education

less than 12th grade, 25.8% with pre-pregnancy obesity, 48.6% with Medicaid
EP

insurance, 58.4% with ≥1 previous live birth, 17.0% and 17.5% with smoking and

alcohol drinking during pregnancy, respectively, and 15.1% with gestational age at
CC

enrollment less than 20 weeks. Univariate analysis indicated that there were significant

(p<0.05) or marginally significant (p<0.1) associations between outcomes (i.e., SBP,

A

DBP, and MAP) and all individual covariates except for alcohol drinking during

pregnancy and gestational age at enrollment (Table 1). In contrast, only three

covariates, maternal age, maternal obesity before pregnancy, and maternal Medicaid

7
insurance status, had a significant (p<0.05) or marginally significant (p<0.1) relationship

with maternal UA concentration (Table 1).

In the lowess curve comparing maternal DBP to UA levels we observed a

breakpoint (the lowest point) at about 0.140 mmol/L UA with a negative linear trend to

the left of the breakpoint and a positive linear trend to the right of the breakpoint (Figure

T
1A). Based on these findings of a ‘J shape’ relation between maternal DBP and UA

IP
levels, we fitted linear spline regression models. Our model selection process is detailed

R
in Table S1. In the unadjusted model, the negative linear trend (β= ̶ 95.29, SE=56.42)

SC
on the left side of the breakpoint, UA<0.140 mmol/L, was marginally significant

(p=0.09). The right side of the breakpoint, UA>0.140 mmol/L, had a significant positive
U
linear trend (β=46.32, SE=8.46, p<0.01). The full model, with all available potential
N
confounders, showed statistically significant trends for maternal DBP and UA levels on
A

both sides of the breakpoint, with a negative trend on the left (adjusted β= ̶150.78,
M

SE=58.01, p=0.01) and positive trend on the right (adjusted β=44.15, SE=8.79, p<0.01).
D

In a reduced model that removed non-significant covariates (alcohol use, maternal

TE

education, gestational week at blood collection), both significant trends were maintained

(adjusted β= ̶149.20, SE=56.83, p=0.01 for the left, and adjusted β=44.21, SE=8.90,
EP

p<0.01 for the right). Using a nonlinear least-squares estimation in STATA, we found

that the ‘optimal’ maternal UA breakpoint was 0.153 mmol/L (95% CI: 0.137 ~ 0.168
CC

mmol/L). We created the final adjusted model with the new optimal breakpoint (adjusted

β= ̶ 95.67, SE=37.67, p=0.01 for the left, and adjusted β=48.95, SE=9.56, p<0.01 for the
A

right) (Table 2, Figure 2A), which had a better fit (smaller RMSE) than the reduced

model with the initial breakpoint (0.140 mmol/L of UA) (data not shown).

8
 The lowess curve comparing maternal SBP to UA levels showed a positive linear

relation (Figure 1B). The unadjusted linear regression model had a significant slope of

β=48.91 (SE=11.84, and p<0.01); inclusion of all available potential confounders

attenuated the slope (adjusted β=37.71, SE=12.89, and p<0.01) (Table S1). In the

reduced model (also the final model) that retained only statistically significant covariates

T
(race/ethnicity, Medicaid Insurance status, maternal obesity before pregnancy, and

IP
maternal parity) the adjusted slope β=37.75 (SE=12.93 and p<0.01) (Table S1,Table 2,

R
Figure 2B).

SC
The shape of the lowess curve comparing maternal MAP to UA was most similar

to that of DBP and UA, i.e. a J-shaped curve (Figure 1C, Table S1). Using the nonlinear
U
least-squares estimation we found the ‘optimal’ maternal UA breakpoint, 0.161 mmol/L
N
(95% CI: 0.140 ~ 0.182 mmol/L), for comparing maternal MAP and UA levels. In the
A

final model the negative linear trend to the left of the breakpoint, UA<0.161 mmol/L, was
M

marginally significant (adjusted β= ̶ 58.48, SE=31.42, p=0.06) and the positive linear
D

trend to the right of the breakpoint, UA>0.161 mmol/L, was significant (adjusted
TE

β=52.23, SE=11.39, p<0.01) (Table 2, Figure 2C).

Examinations of assumptions, over-fitting, and sensitivity analysis of the final

EP

models indicated that our results were robust. Detailed information is present in Online

Supporting Information.
CC

Comment
A

We observed a J-shaped relation between maternal DBP or MAP and maternal

UA levels in mid-pregnancy. The breakpoint (lowest point in J shape) was 0.153 mmol/L

9
UA for DBP and 0.161 mmol/L UA for MAP, respectively. By contrast, the relation

between maternal SBP and UA levels followed a positive linear trend in unadjusted and

adjusted analyses. It must be noted that the strength of the J-shaped relation between

maternal DPB or MAP and UA in the present study persisted but was attenuated after

removing women with preeclampsia or gestational hypertension. We lacked sufficient

T
numbers of women with hypertensive disorders of pregnancy to test the J-shaped

IP
relation in this subgroup alone.

R
To our knowledge, the current study is the first to report a non-linear relation

SC
between DBP or MAP and UA levels in pregnant women. Paula et al studied BP and UA

levels among 58 pregnant women enrolled at ≥20 weeks’ gestation in a hospital setting
U
and reported higher UA levels in association with higher DBP, but not with SBP [6]. A
N
larger, retrospective cohort study (n=1880) by Hawkins et al measured maternal UA
A

levels near the time of delivery [8] They found that women with gestation-corrected
M

hyperuricemia had higher SBP and DBP. In these two studies, maternal UA levels were
D

categorized and nonlinear relationships were not investigated. Interestingly our UA

TE

breakpoints for DBP and MAP are close to the low end of the estimated ‘normal clinical

range’ of UA, i.e. 0.144 ~ 0.294 mmol/L in second trimester pregnant women according
EP

to a study by Abbassi-Ghanavati et al [20].

The physiology/pathology underlying our findings is, at this point, uncertain. MAP
CC

reflects Cardiac Output × Total Peripheral Resistance (TPR) [21], thus for a given

cardiac output, MAP is positively correlated with TPR. A recent cross-sectional study
A

using multiple linear regression models reported a significant positive relationship

between UA levels and TPR in women, but not in men [22]. The mechanism connecting

10
UA, TPR, and MAP may be related to blood viscosity, based on the Hagen-Poiseuille

Equation, i.e., Resistance = 8×Vessel Length×Blood Viscosity / (π×Vessel Radius4)

[23]. Blood viscosity is a sensitive marker for oxidative stress24 while oxidative stress is

associated with high UA levels [1,2,3,9]. In addition, reactive oxygen species (ROS, a

series of indicators of oxidative stress) may increase peripheral vascular resistance by

T
directly impairing endothelial function [25]. At physiological levels, UA is an important

IP
anti-oxidant [3,26]. Thus, low UA may reduce antioxidant activity in blood and increase

R
ROS, and finally increase MAP via an oxidative stress-induced increase of blood

SC
viscosity. On the other hand, high UA might increase oxidative stress via at least two

pathways: 1) high UA functions as pro-oxidant to increase oxidative stress [27]; and 2)

U
high UA may reflect a high level of xanthine oxidase activity [28] whereas increased
N
xanthine oxidase activity can increase oxidative stress [29].
A

Because MAP equals ⅔ × DBP + ⅓ × SBP, MAP is more sensitive to changes in

M

DBP compared to SBP. As discussed above, MAP is positively correlated with TPR.
D

Thus, low UA-induced oxidative stress may increase DBP via the oxidative-stress-
TE

induced TPR, whereas high UA-related increasing DBP would have the same

underlying mechanisms as that for high UA-related increasing MAP.

EP

Arterial stiffness is a major contributor to high SBP [30] and evidence has shown

that high UA is positively correlated with arterial stiffness [31]. Arterial stiffness involves
CC

a complex and dynamic interaction among structural proteins, extracellular matrix,

inflammatory molecules, and ROS in vessel walls [32]. Low UA-related ROS might
A

contribute less to arterial stiffness, and consequently not appear associated with SBP

increase, hence only the linear relation between UA and SBP and not a ‘J-shape’.

11
 Our analyses have several strengths. The data come from the prospective

POUCH Study, a large, diverse pregnancy cohort, thereby enhancing generalizability.

Blood samples collected in mid-pregnancy are accompanied by a rich data gathering

protocol allowing for the measurement of many potential confounding variables. The

application of a 10-fold cross-validation technique to address concerns of model

T
overfitting improves the internal validity. By applying non-linear regression models we

IP
increase statistical power and reduce the loss of information that occurs when variables

R
are categorized. Finally, the use of bootstrap estimation of variance and sensitivity

SC
analysis assure the robustness of our analyses and allows us to examine the influence

of subgroups (e.g. women with hypertensive disorders of pregnancy) on overall results.

U
There are also some study limitations that merit consideration. We used the
N
single highest DBP measurement abstracted from prenatal records, thus there is likely
A

measurement error and a large variance. In order to capture MAP, we used SBP at the
M

time of the highest DBP; it is possible that this approach misclassified some women in
D

the SBP analysis who had a higher SBP at a different time point in pregnancy. Maternal
TE

UA was measured only once, in mid-pregnancy, while blood pressure was abstracted

across pregnancy, consequently time order between changes in UA and blood pressure
EP

could not be evaluated. Also, we could not rule out some common underlying pathology

in pregnancy as an explanation for the association between maternal UA and blood

CC

pressure, therefore our findings should be interpreted cautiously. While we captured

many potential confounding variables, some (e.g. diet) were not available to test in our
A

models.

12
Conclusions

We conclude that maternal UA levels at mid-pregnancy have a non-linear

association with maternal DBP/MAP and a linear association with SBP. The proportion

of women with ‘low UA levels’, i.e. below the breakpoint in the DBP/MAP J-shape

curves, was relatively small and large studies may be needed to confirm our findings. In

T
addition, explanations for the observed J- shape relation would require more in-depth

IP
studies of vascular health among pregnant women along with basic science research on

R
the physiology/pathology associated with UA.

SC
Disclosure of Interest
U
N
The authors have no conflicts of interest to declare.
A
M

Funding
D

None
TE
EP

Acknowledgments

We are grateful to Dr. B. Bullen (College of Human Medicine, Michigan State

CC

University), project director, for preparing the POUCH cohort data.

A

13
References
[1] Corry DB, Eslami P, Yamanoto K, Nyby MD, Makino H, Tuck ML. Uric acid
stimulates vascular smooth muscle cell proliferation and oxidative stress via the
vascular renin-angiotensin system. J Hypertens 2008; 26:269-75.
[2] Lanaspa MA, Sanchez-Lozada LG, Choi YJ, et al. Uric acid induces hepatic
steatosis by generation of mitochondrial oxidative stress: potential role in fructose-

T
dependent and –independent fatty liver. J Biol Chem 2012; 287:40732-44.

IP
[3] Fabbrini E, Serafini M, Colic Baric I, Hazen SL, Klein S. Effect of plasma uric acid on
antioxidant capacity, oxidative stress, and insulin sensitivity in obese subjects. Diabetes

R
2014; 63:976-981.

SC
[4] Alderman MH, Cohen H, Madhavan S, Kivlighn S. Serum uric acid and
cardiovascular events in successfully treated hypertensive patients. Hypertens 1999;
34:144-50.
U
[5] Suliman ME, Johnson RJ, Garcia-Lopez E, et al. J-shaped mortality relationship for
N
uric acid in CKD. Am J Kidney Dis 2006; 48:761-71.
A
[6] Paula LG, da Costa BE, Poli-de-Figueiredo CE, Antonello IC. Does uric acid provide
information about maternal condition and fetal outcome in pregnant women with
M

hypertension? Hypertens Pregnancy. 2008; 27:413-420.

[7] Andrew L, Patel N. Uric acid levels in pregnancy induced hypertension (PIH) in
D

relation to maternal and perinatal outcomes. Int J Clin Biochem Res. 2016; 3:150-153.
TE

[8] Howkins TL, Roberts JM, Mangos GJ, Davis GK, Roberts LM, Brown MA. Plasma
uric acid remains a marker of poor outcome in hypertensive pregnancy: a retrospective
EP

cohort study. BJOG 2012; 119:484-92.

[9] Peraçoli MT, Bannwart CF, Cristofalo R, et al. Increased reactive oxygen species
and tumor necrosis factor-alpha production by monocytes are associated with elevated
CC

levels of uric acid in pre-eclamptic women. Am J Reprod Immunol 2011; 66:460-7.

[10] Rogers MS, Wang CCR, Tam WH, Li CY, Chu KO, Chu CY. Oxidative stress in
A

midpregnancy as a predictor of gestational hypertension and pre-eclampsia. BJOG

2006; 113:1053-9.
[11] Garcia-Benavides L, Guzman-Sanchez A, Hernandez-Mora FJ, Muro-Gomez AM,
Gomez-Martinez ML, Siller-Lopez FR. PP078. Total antioxidant capacity in patients with

14
pregnancy induced hypertension: its relation to maternal and/or perinatal complications
(abstract). Pregnancy Hypertens 2012; 2:282-283.
[12] Turpin CA, Sakyi SA, Owiredu WK, Ephraim RKD, Anto EO. Association between
adverse pregnancy outcome and imbalance in angiogenic regulators and oxidative
stress biomarkers in gestational hypertension and preeclampsia. BMC Pregnancy
Childbirth 2015; 15:189. doi: 10.1186/s12884-015-0624-y.

T
[13] Loverro G, Greco P, Capuano F, Carone D, Cormio G, Selvaggi L. Lipoperoxidation

IP
and antioxidant enzymes activity in pregnancy complicated with hypertension. Eur J
Obstet Gynecol Reprod Biol. 1996; 70:123-7.

R
[14] Hilali N, Kocyigit A, Demir M, et al. DNA damage and oxidative stress in patients

SC
with mild preeclampsia and offspring. Eur J Obstet Gynecol Reprod Biol. 2013;
170:377-80.
[15] Holzman C, Senagore PK, Wang J. Mononuclear leukocyte infiltrate in
U
extraplacental membranes and preterm. Am J Epidemiol. 2013; 177:1053-1064.
N
[16] Fossati P, Prencipe L, Berti G. Use of 3,5-dichloro-2-hydroxybenzenesulfonic
A
acid/4-aminophenazone chromogenic system in direct enzymic assay of uric acid in
serum and urine. Clin Chem. 1980; 26:227-231.
M

[17] Bruin J. Newtest: command to compute new test. UCLA: Statistical Consulting
Group. 2006; http://www.ats.ucla.edu/stat/stata/ado/analysis/.
D

[18] Hastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning: Data

TE

Mining, Inference, and Prediction. 2nd Ed. Springer-Verlag. 2001.

[19] Kolenikov S. Resampling variance estimation for complex survey data. Stata J.
EP

2010; 10:165-199.
[20] Abbassi-Ghanavati M, Greer LG, Cunningham FG. Pregnancy and laboratory
studies: a reference table for clinicians. Obstet Gynecol. 2009; 114:1326-31.
CC

[21] Klabunde RE. Cardiovascular Physiology Concepts. Lippincott Williams & Wilkins.
2011.
A

[22] Hsu PF, Chuang SY, Cheng HM, et al. Associations of serum uric acid levels with
arterial wave reflections and central systolic blood pressure. Int J Cardiol. 2013;
168:2057-2063.

15
[23] Thurston GB. Viscosity and viscoelasticity of blood in small diameter tubes.
Microvasular Res. 1976; 11:133-146.
[24] Richards RS, Nwose EU. Blood viscosity at different stages of diabetes
pathogenesis. Br. J Biomed Sci. 2010; 67:67-70.
[25] Wong WT, Wong SL, Tian XY, Huang Y. Endothelial dysfunction: the common
consequence in diabetes and hypertension. J Cardiovasc Pharmacol. 2010; 55:300-

T
307.

IP
[26] Weisskopf MG, O’Reilly E, Chen H, Schwarzschild MA, Ascherio A. Plasma urate
and risk of Parkinson’s disease. Am J Epidemiol. 2007; 166:561-7.

R
[27] Corry DB, Eslami P, Yamanoto K, Nyby MD, Makino H, Tuck ML. Uric acid

SC
stimulates vascular smooth muscle cell proliferation and oxidative stress via the
vascular renin-angiotensin system. J Hypertens. 2008; 26:269-275.
[28] Dawson J, Walters M. Uric acid and xanthine oxidase: future therapeutic targets in
U
the prevention of cardiovascular disease? Br J Clin Pharmacol. 2006; 62:633-44.
N
[29] Ryan MJ, Jackson JR, Hao Y, Leonard SS, Always SE. Inhibition of xanthine
A
oxidase reduces oxidative stress and improves skeletal muscle function in response to
electrically stimulated isometric contractions in aged mice. Free Radic Biol Med. 2011;
M

51:38-52.
[30] O'Rourke MF, Kelly RP, Avolio AP, Hayward CS. Effects of arterial dilator agents
D

on central aortic systolic pressure and on left ventricular hydraulic load. Am J Cardiol.
TE

1989; 63:381-441.
[31] Gomez-Marcos MA, Recio-Rodríguez JI, Patino-Alonso MC, et al. Relationship
EP

between uric acid and vascular structure and function in hypertensive patients and sex-
related differences. Am J Hypertens. 2013; 26:599-607.
[32] Zieman SJ, Melenovsky V, Kass DA. Mechanisms, pathophysiology, and therapy of
CC

arterial stiffness. Arterioscler Thromb Vasc Biol. 2005; 25:932-943.

A

16
Figure Legends:

Figure 1. Lowess smoothing plots for maternal DBP/SBP/MAP vs. maternal UA. Note:

panel A: DBP vs. UA; panel B: SBP vs. UA; panel C: MAP vs. UA. The solid lines

represent lowess curves and the dots indicate the observed raw data points.

T
Figure 2. Visualization of the final models for DBP/SBP/MAP vs. maternal UA. Note:

IP
panels A, B, C: linearized estimates of variance; panels D, E, F: Bootstrapped estimates

R
of variance. Panel A & D: DBP vs. UA; panel B & E: SBP vs. UA; panel C & F: MAP vs.

SC
UA. The solid lines represent point estimates and the upper and lower dashed lines

indicate 95% confidence bands by nonparametric smoothing with locally weighted

regression (lowess). U
N
A
M
D
TE
EP
CC
A

17