Selenium and maternal blood pressure during childbirth

Abstract

Evidence suggests selenium concentrations outside the nutritional range may worsen cardiovascular health. This paper examines the relationship between selenium and maternal blood pressure among 270 deliveries using umbilical cord serum as a proxy for maternal exposure levels. Multivariable models used linear splines for selenium and controlled for gestational age, maternal age, race, median household income, parity, smoking, and prepregnancy body mass index. Nonparametric analysis of this dataset was used to select spline knots for selenium at 70 and 90 μg/L. When selenium was <70 μg/L, increasing selenium levels were related to a non-statistically significant decrease in blood pressure. For selenium 70 – 90 μg/L, a 1-μg/L increase was related to a 0.37 mmHg (95% confidence interval (CI): 0.005, 0.73) change in systolic and a 0.35 mmHg (0.07, 0.64) change in diastolic blood pressure. There were very few selenium values >90 μg/L. Other studies indicate that the maternal/cord selenium ratio is 1.46 (95% CI: 1.28, 1.65). This u-shaped relationship between selenium and blood pressure is consistent with a dual role of selenium as an essential micronutrient that is nonetheless a toxicant at higher concentrations; however this needs to be studied further.

INTRODUCTION

Selenium is an antioxidant and promotes thyroid and immune system function (Brown and Arthur, 2001). The recommended daily allowance for selenium, currently 55 μg/day (60 μg/day during pregnancy), is needed to maximize glutathione peroxidase synthesis (Institute of Medicine, 2000). Either selenium deficiency or excess may lead to adverse health effects (Institute of Medicine, 2000); globally, deficiency is a larger concern than excess (Combs, 2001).

In North America, daily requirements for selenium are generally met or exceeded (Combs, 2001; Institute of Medicine, 2000). It has been questioned whether selenium intake in excess of recommended levels might contribute to development of cardiovascular disease (Navas-Acien et al, 2008). Additionally, increased selenium intake may contribute to diabetes and cardiovascular outcomes (Stranges et al, 2010).

Elevations in blood pressure (BP) during pregnancy are of concern due to possible chronic sequelae (Kaaja and Greer, 2005; Lykke et al, 2009). However, few studies have explored selenium and BP during pregnancy. Research in the United Kingdom suggested preeclamptic mothers have less selenium compared to healthy mothers (Mistry et al, 2008; Rayman et al, 2003); in contrast, a study among African mothers suggested the opposite (Mahomed et al, 2000). Other work implicated selenium supplementation with decreased risk of gestational hypertension among Chinese mothers (Han and Zhou, 1994) and supplementation in a clinical trial was related to a decrease in preeclampsia among Iranian mothers (Tara et al, 2010).

The relationship of selenium and BP may depend on whether nutritional needs for selenium have been met. This study evaluates the relationship between selenium exposure as measured in umbilical cord serum and maternal BP in a cross-sectional study of mothers giving birth in Baltimore, Maryland. Additionally, prior studies were used to create a maternal/cord selenium ratio in order to estimate maternal selenium in the current study.

MATERIALS AND METHODS

The Baltimore THREE (Tracking Health Related to Environmental Exposures) Study is a cross-sectional study of mothers and infants conducted with approval of the Maternal and Fetal Research Committee, Department of Gynecology and Obstetrics, and the Johns Hopkins School of Medicine Institutional Review Board (Apelberg et al, 2007). Informed consent was not required for this study because collected biological samples would otherwise have been discarded and sample collection constituted no more than minimal risk. Strict procedures were developed to protect subject confidentiality. Due to this design, maternal serum was not available.

Mothers giving birth between November 2004 and March 2005 in the Johns Hopkins Hospital were eligible for inclusion (n=603 deliveries). Deliveries with multiple births (n=12) and births where cord blood was unavailable or of insufficient quantity (n=291) were excluded. These analyses also excluded mothers with missing data on selenium (n=13), BP (n=2), prepregnancy body mass index (BMI) (n=11), and median household income (n=3). One mother with admission systolic BP <20 mmHg was also excluded. Data were collected from analyses of umbilical cord blood and maternal electronic medical records. Two study personnel extracted data from medical records; a random 10% sample was reviewed by study clinicians.

Hospital staff took maternal BP measurements as part of routine medical care at admission and continuously while hospitalized for labor and delivery. BP at hospital admission and maximum BP (based on systolic) were abstracted. Reliability of these measurements was assessed using other variables within the dataset (supplementary information). Measurements with systolic BP ≥ 140 mmHg or diastolic BP ≥ 90 mmHg were considered “elevated”, but not tantamount to a diagnosis of hypertension. Hypertension was classified as 1) all hypertension cases (pregnancy-related, chronic, or medication use) and 2) the subset of women with pregnancy-related hypertension diagnoses (supplementary information).

Trained clinical staff collected umbilical cord blood using standard procedures (Witter et al, 2001). Serum samples were analyzed at the United States Centers for Disease Control and Prevention (CDC). Selenium was determined using National Institute of Standards and Technology (NIST)-traceable, matrix-matched calibrators on inductively coupled plasma dynamic reaction cell mass spectrometry methodology adhering to Clinical Laboratory Improvement Amendments (CLIA) ‘88 standards (Xiao, 2006). Limit of detection (LOD) was 5 μg/L. Cotinine was measured using liquid chromatography in conjunction with atmospheric pressure ionization tandem mass spectrometry; LOD=0.015 ng/mL (Bernert et al, 1997). A mother was classified as a smoker if serum cotinine was ≥ 10 ng/mL (CDC, 2005) or if smoking during pregnancy was noted in the medical record.

BMI was calculated as prepregnancy weight (kg) / height (m²). Median household income was obtained from the United States 2000 Census. The median income was based on a small census area called a block group, which includes 600–3000 individuals (supplementary information).

Statistical analyses were performed using Stata 11 (College Station, TX). Our hypothesis was that selenium’s effect may be concentration-dependent; however, there are no guidelines for umbilical cord selenium concentrations. Therefore, we used our data to identify cutoff points for spline models. Inflection points (where the dose-response changes direction) were identified visually using lowess curves of selenium with each BP measurement (supplementary information). The exact inflection point differed slightly for each BP measurement; for consistency, knots were chosen to approximate inflection points across all BP measures.

Potential covariates were considered for inclusion based on prior publications (Mistry et al, 2008; Rayman et al, 2003; Schulpis et al, 2004) and bivariate comparisons of variables within this dataset (criteria of p<0.30 for one-way ANOVAS or t-tests). In addition to covariates included in our final models, we considered alcohol use, parity, medical insurance, anemia, cord blood lead levels and cord serum fatty acids; these were eliminated based on likelihood ratio tests and observation of whether the selenium coefficient changed >10% following their inclusion. A series of final models (including gestational age, maternal age, maternal race, primiparity, income, BMI, and smoking) are presented with increasing levels of adjustment for confounders. We also ran models with selenium as a linear term. Models were evaluated using Akaike’s information criterion, likelihood ratio tests, normal quantile plots, and residual vs. fitted value plots. Similar procedures were used for logistic models to evaluate hypertension and “elevated” BP.

In order to put our findings into perspective with other studies that measured maternal serum selenium levels and not cord serum levels, prior reports were used to calculate a maternal/umbilical cord serum selenium ratios. These were used to estimate maternal serum selenium levels in this population. Relevant articles were identified using PubMed. Summary data on maternal selenium and umbilical cord selenium were extracted and used to calculate article-specific as well as an overall mean ratio weighted by study sample size.

RESULTS

Population characteristics are shown in Table 1. Mean admission BP was 122.7 mmHg (systolic) and 72.0 mmHg (diastolic). Corresponding values for maximum BP were 144.3 mmHg and 77.1 mmHg. Mean selenium was 69.9 μg/L (95% confidence interval (CI): 68.5, 71.4). Knots for selenium splines were chosen at 70 and 90 μg/L. Roughly half had selenium concentrations ≤ 70 μg/L, and very few (n=12) had concentrations >90 μg/L (Table 1).

Table 1.

Population Characteristics, Baltimore THREE Study, 2004–2005.

Category	Characteristic	N	%
All		270	100.0
Maternal age	<20 years	55	20.4
	20 – 29 years	128	47.4
	≥30 years	87	32.2
Maternal race	Caucasian	57	21.1
	African American	190	70.4
	Asian	23	8.5
Smoking	Smoked during pregnancy	49	18.2
	Did not smoke	221	81.9
Prepregnancy BMI	<18.5 kg/m2	14	5.2
	18.5–24.9 kg/m2	126	46.7
	25.0–29.9 kg/m2	61	22.6
	≥30.0 kg/m2	69	25.6
Median household income (neighborhood)	<$25,000	82	30.4
	$25,000–50,000	144	53.3
	>$50,000	44	16.3
Parity	First childbirth	117	43.3
	Second or higher	153	56.7
Gestation	Preterm delivery (<37 wk)	34	12.6
	Not preterm delivery	236	87.4
“Elevated” admission BP	Yes	36	13.3
	No	234	86.7
“Elevated” maximum BP	Yes	115	42.6
	No	155	57.4
GH or preeclampsia	Yes	16	5.9
	No	254	94.1
Any hypertension	Yes	26	9.6
	No	244	90.4
Selenium categories	<70 μg/L	150	55.6
	71–90 μg/L	108	40.0
	>90 μg/L	12	4.4

BMI = body mass index; BP = blood pressure; GH = gestational hypertension; THREE = Tracking Health Related to Environmental Exposures.

Percents may not sum to 100 due to rounding. “Elevated” BP is systolic blood pressure >= 140 mmHg or diastolic BP>= 90 mmHg. Any hypertension includes gestational hypertension, preeclampsia, and chronic hypertension. Median household income is based on roughly 1,500 persons in close geographic proximity, in the same census block group.

There was a consistent pattern in the relationship between selenium concentrations and maternal BP (Table 2). When selenium concentrations were <70 μg/L, increasing selenium levels were related to decreasing BP. However, when selenium concentrations were 70–90 μg/L, increasing selenium levels were related to increasing BP; these were statistically significant (p <0.05) or of borderline significance (p <0.10) for all measures except maximum diastolic BP. For selenium concentrations >90 μg/L there was a decrease in BP with increased selenium; however, there was a large uncertainty in this range as this group only has only twelve individuals.. There was no association between selenium and BP in models where selenium was a linear term (Table 2).

Table 2.

Change and 95% Confidence Interval for Maternal Blood Pressure (in mmHg) During Childbirth Related to a 1 μg/L Increase in Umbilical Cord Serum Selenium, Baltimore THREE Study, 2004–2005, n=270.

Outcome	Linear model Selenium		Selenium < 70 μg/L		Spline model Selenium 71–90 μg/L		Selenium ≥ 91 μg/L
	Change	95% CI	Change	95% CI	Change	95% CI	Change	95% CI
Admission SBP
Model 1	0.07	−0.08, 0.23	−0.19	−0.51, 0.13	0.41**	0.04, 0.77	−0.25	−1.03, 0.13
Model 2	0.05	−0.10, 0.20	−0.21	−0.53, 0.11	0.36*	−0.01, 0.73	−0.16	−0.95, 0.63
Model 3	0.08	−0.07, 0.23	−0.17	−0.48, 0.14	0.37**	0.004, 0.73	−0.09	−0.86, 0.68
Admission DBP
Model 1	0.02	−0.11, 0.14	−0.15	−0.40, 0.11	0.26*	−0.03, 0.55	−0.33	−0.96, 0.26
Model 2	0.02	−0.10, 0.14	−0.13	−0.38, 0.13	0.27*	−0.02, 0.56	−0.38	−1.01, 0.25
Model 3	0.05	−0.07, 0.17	−0.11	−0.36, 0.14	0.30**	0.01, 0.60	−0.36	−0.98, 0.26
Maximum SBP
Model 1	0.07	−0.11, 0.25	−0.29	−0.67, 0.09	0.52**	0.09, 0.97	−0.34	−1.26, 0.58
Model 2	0.04	−0.14, 0.22	−0.33	−0.71, 0.05	0.46**	0.02, 0.90	−0.23	−1.16, 0.71
Model 3	0.07	−0.11, 0.25	−0.24	−0.61, 0.12	0.43**	0.003, 0.85	−0.13	−1.04, 0.77
Maximum DBP
Model 1	−0.18	−0.38, 0.02	−0.59**	−1.01, −0.17	0.17	−0.31, 0.65	−0.02	−1.05, 1.02
Model 2	−0.16	−0.37, 0.04	−0.56**	−0.99, −0.13	0.18	−0.31, 0.67	−0.03	−1.08, 1.01
Model 3	−0.14	−0.35, 0.07	−0.52**	−0.95, −0.10	0.19	−0.31, 0.68	0.02	−1.03, 1.08

THREE = Tracking Health Related to Environmental Exposures; CI = confidence interval; SBP = systolic blood pressure; DBP = diastolic blood pressure

^*P < 0.10;

^**P < 0.05

Model 1: Blood pressure ~ selenium + gestational age

Model 2: Model 1 + maternal age + maternal race

Model 3: Model 2 + primiparity + median household income + smoking + prepregnancy body mass index

Results from multivariable logistic regression models for “elevated” BP were similar to admission and maximum BP. We did not observe any relationship of selenium with hypertension outcomes (supplementary information).

Eleven studies with data on both maternal and umbilical cord selenium were identified (Butler Walker et al, 2006; Kantola et al, 2004; Lorenzo Alonso et al, 2005; Makhoul et al, 2004; Micetic-Turk et al, 2000; Mistry et al, 2008; Osman et al, 2000; Rudge et al, 2009; Sakamoto et al, 2010; Schulpis et al, 2004; Wasowicz et al, 1993). Study-specific and overall maternal/cord selenium ratios are presented (Table 3, Figure 1). Individual maternal:cord selenium ratios ranged from 0.85 to 1.84, with a weighted average of 1.46 (95% CI: 1.28, 1.65). Applying this ratio to selenium concentrations in our study suggests that there is decreasing maternal BP with increasing selenium when maternal selenium is <102 μg/L and increasing maternal BP with increasing selenium when maternal selenium is 102–131 μg/L.

Table 3.

Maternal and Umbilical Cord Selenium Concentrations Reported in Prior Studies.

Study information		Selenium		Maternal (M)		Cord (C)		M/C
Study	Location	Media	Method	N	Mean	N	Mean	Ratio
Wasowicz 1993	Poland	Plasma	FM	64	35	64	28.1	1.25
Micetic-Turk 2000	Slovenia	Serum	FIHG-AAS	20	62	20	34	1.82
Osman 2000	Sweden	Serum	GFAAS	74	71.89	74	52.93	1.36
Makhoul 2004	Israel (preec.)	Serum	AAS	32	86.64	38	52.9	1.64
Makhoul 2004	Israel (controls)	Serum	AAS	119	84.1	130	57.48	1.46
Schulpis 2004	Greece	Serum	GFAAS	1118	68.3	1118	37.02	1.84
Schulpis 2004	Albania	Serum	GFAAS	820	37.4	820	34.33	1.09
Kantola 2004	Finland (nonsmokers)	Serum	AAS	63	110	67	111	0.93
Kantola 2004	Finland (smokers)	Serum	AAS	15	111	15	124	0.90
Lorenzo Alonso 2005	Spain	Serum	AAS	48	90	48	76.3	1.18
Butler Walker 2006	Canada (Caucasians)	Plasma	GFAAS	132	124	125	87	1.43
Butler Walker 2006	Canada (Dene/Metis)	Plasma	GFAAS	92	119	81	75	1.59
Butler Walker 2006	Canada (Inuit)	Plasma	GFAAS	144	119	161	74	1.61
Mistry 2008	United Kingdom (preec.)	Serum	GFAAS	25	39.7	25	29	1.37
Mistry 2008	United Kingdom (controls)	Serum	GFAAS	27	58.4	27	42.1	1.39
Rudge 2009	South Africa	WB	ICP-MS	62	104	62	111	0.94
Sakamoto 2010	Japan	RBC	ICP-MS	81	192	81	227	0.85

Abbreviations: FM = LS-5 spectroflourometer; GFAAS = graphite furnace atomic absorption spectroscopy; FIHG-AAS = flow injection hydride generation atomic absorption spectrometry; AAS = atomic absorption spectroscopy; ICP-MS = inductively coupled plasma mass spectrometry; preec. = preeclampsia; M/C = maternal/cord; RBC = red blood cells; WB = whole blood. All maternal samples were collected around the time of delivery except for Osman 2000, where they were collected at gestational week 36. Selenium is measured in μg/L except for Sakamoto 2010 which was measured in ng/g. Arithmetic means are presented except for Rudge 2009, where the median was used. Maternal/fetal ratio is calculated for this paper using the individual authors’ summary data.

Figure 1.

Average maternal and umbilical cord selenium concentrations as reported previously (Table 3). Circles are scaled by the average of maternal and cord sample size of the individual studies; this ranges from n=15 to n=1118. The solid line indicates a maternal:cord ratio of 1; the dashed line represents the weighted average among maternal:cord ratios, 1.46.

DISCUSSION

We did not find any evidence of a linear relationship of selenium with BP across the entire range of blood concentrations. However, when modeling selenium as a spline, we saw both positive and negative relationships that were dependent upon the selenium concentration range. While these relationships were not all statistically significant at the p <0.05 level, they were consistent among different BP measurements. Maternal/cord selenium ratios calculated from prior work have a weighted average of 1.46. Variability between individual ratios could arise from differences in selenium intake, health status, selection criteria, or other genetic or environmental factors. Based on application of this ratio, this study reports a trend of lower maternal BP with increasing selenium when cord serum selenium is <70 μg/L (and maternal serum selenium approximately <105 μg/L) and higher maternal BP with increasing selenium when cord serum selenium is between 70–90 μg/L (and maternal serum selenium approximately 102–131 μg/L). There are too few observations over 90 μg/L to draw firm conclusions about effects in that range.

The Institute of Medicine notes that adult nutritional requirements of selenium are met where serum selenium is 70–90 μg/L (Institute of Medicine, 2000). Based on estimated maternal selenium values in our study, roughly 75% of the current population would meet the lower level of this requirement. This is noticeably lower than other reports of serum selenium among US adults (Laclaustra et al, 2009) yet higher than many other reports (Table 3).

Our results are consistent with prior observations of a similar biphasic relationship between selenium and cardiovascular mortality (Bleys et al, 2008) or hypertension (Laclaustra et al, 2009). Overall, however, studies that have directly assessed selenium and blood pressure-related health outcomes have had inconsistent results; it is possible that this variety may reflect variability in baseline selenium levels in different populations. For example, several studies reporting that increased selenium is protective of cardiovascular disease have been performed in regions where selenium deficiencies are more prevalent (Broadley et al, 2006; Rayman, 2008).

This study and Laclaustra et al both identified biphasic relationships of selenium and cardiovascular outcomes; however, unlike the Laclaustra study, we do not observe a relationship with hypertension (Laclaustra et al, 2009). There are several potential explanations for this, including that this study had lower selenium concentrations than Laclaustra (accounting for imputation to maternal values), smaller sample size and therefore less statistical power, and a younger population (with a small number of cases of diagnosed hypertension) compared to the Laclaustra et al study.

There are some potential limitations to this study. We assayed total selenium and did not quantify the forms of selenium that are found in serum, selenium-cysteine and selenium-methionine. Selenium-cysteine is immediately available to be incorporated into proteins, whereas selenium-methionine is a form of stored selenium that nonetheless can be made biologically available (Flores-Mateo et al, 2006). These different forms of selenium may have differential effects on BP. Also, selenium storage as selenium methionine occurs in tissues; the factors involved with selenium storage and release of selenium methionine into blood are not well understood. It is possible that such factors could confound the relationship between selenium and BP.

BP was measured automatically during labor and delivery and abstracted from electronic medical records. Potential limitations of this approach are that we did not use an average of repeated measurements but instead abstracted the BP at admission and the highest and lowest measurements. Also, measurements were taken during a potentially stressful time and may not reflect BP levels at other times. However, BP measurements were consistent for all participants, and there is no reason to assume that the process of labor and delivery would modify the relationship between selenium and BP. BP was recorded automatically rather than being charted by hand, eliminating one potential source of bias and error. Also, in comparing BP to diagnosed hypertension and other variables such as delivery mode (supplementary information), it seems that the measures produce consistent results. Therefore, it is reasonable to expect these measurements serve to describe relative BP within this population.

Prior studies that assumed a linear relationship between selenium and BP produced results that are consistent with our findings in that they reported small positive or no associations between selenium and BP (Jossa et al, 1991; Nawrot et al, 2007; Suadicani et al, 1992; Taittonen et al, 1997). A spline model is more appropriate given the nature of the statistical relationship and biological plausibility of that relationship because of the dual role of selenium as a nutrient at low levels and a toxicant at higher levels.

We were able to obtain high quality selenium measurements. Umbilical cord serum was used as a proxy for maternal serum selenium levels; these levels are highly correlated (Micetic-Turk et al, 2000; Rudge et al, 2009; Sakamoto et al, 2010). However, use of umbilical cord measures probably results in some misclassification of maternal exposure, which, if random, would attenuate the relationship between selenium and blood pressure.

Our exposure and outcome measurements reflect roughly the same time period; to the extent that BP is responding immediately to selenium levels this could be advantageous. However, it also is possible that selenium levels over time have chronic effects on BP. Prior work suggests that increasing maternal blood volume and fetal nutritional needs may affect selenium concentrations over the course of pregnancy (Kantola et al, 2004). It is currently unclear whether maternal selenium generally increases (Dawson et al, 2000), decreases (Kantola et al, 2004), or remains constant (Navarro et al, 1996) over the course of pregnancy. Since we do not have information about maternal selenium levels over the course of pregnancy we cannot evaluate whether these changes are important in this population. However, we are not aware of any evidence that changes in selenium over the course of pregnancy are differentially related to BP; therefore, we would expect any misclassification to either not affect or attenuate the relationships presented here.

Another consideration is whether mothers deficient in selenium might transfer selenium at a higher rate to the fetus, since these elements are critical for the process of growth and development. This is supported by Schulpis and colleagues (Schulpis et al, 2004), who reported much lower selenium levels in maternal serum among Albanians compared to Greeks, but only small differences in cord serum selenium concentrations from the two populations. However, concentrations of umbilical cord serum selenium in this study are roughly twice as high as those reported by Schulpis. This suggests that umbilical cord serum selenium concentrations in our population would not be affected by maternal selenium deficiency.

In this study, we employed flexible models to evaluate the relationship between selenium in umbilical cord serum, maternal BP during labor and delivery. Consistent with prior studies, our results show that there may be different effects of selenium exposure depending on the overall concentrations. This suggests that for some populations, selenium may be a factor in increased BP. Moreover, we stress the importance of applying flexible models when evaluating associations that are likely to have different relationships within different exposure ranges. These results should be verified in larger populations, ideally through the study of additional exposure and outcome biomarkers.