Keywords
labor dystocia - obesity - mechanisms - metabolomics - parturition
Cesarean birth increased steadily in the United States over the past several decades
across all gestational ages, maternal ages, and racial groups.[1]
[2]
[3] After a brief plateau from 2009 to 2016, the cesarean delivery rate rose again in
2017 to reach 32.0%.[4] Morbidity after primary cesarean and in subsequent pregnancies is increased due
to wound complications, deep vein thrombosis and pulmonary embolism, bladder injury,
obstetrical hemorrhage, placenta previa/accreta, and cesarean hysterectomy.[5]
[6]
[7] For this reason, a key strategy identified by the Centers for Disease Control and
Prevention to decrease pregnancy-related deaths is to avoid the use of cesarean delivery
unless medically indicated.[8]
Cesarean deliveries are most often performed for abnormally slow labor progression,
a complication known as labor dystocia.[9]
[10] Expectations for labor progress are based on findings from large studies showing
the median time required for changes in the uterine cervix during labor.[11]
[12] Secondary analyses of these studies showed that there is a dose-dependent slowing
of labor with elevations in maternal body mass index (BMI); for example, a woman having
her first baby who has a BMI ≥ 40 kg/m2 may take up to 7 hours longer than a normal weight women to progress from 4 to 10 cm
of cervical dilation.[13] Largely due to labor dystocia,[14] women with higher BMI ranges are at increased risk of cesarean delivery, even after
accounting for the influence of parity, pregnancy complications, and hospital characteristics.[15]
[16]
[17]
[18]
Basic science studies examining mechanisms for slow labor progress suggest that in
obese pregnant women, myometrial cells contract with reduced efficiency.[19]
[20]
[21] A theoretical cause of myometrial inefficiency in settings of maternal obesity is
lipotoxicity, whereby storage of excess free fatty acids in nonadipose tissues and
use of fatty acids instead of glucose for cellular energy requirements causes widespread
cellular damage and dysfunction.[22]
[23]
[24] High-throughput metabolomics, the comprehensive study of the end products and pathways
of metabolism, offer a new capacity for the study of labor dystocia by allowing for
the simultaneous measurement of trace biomarkers from several different mechanistic
pathways.[25]
[26]
[27]
[28]
[29] Thus, we conducted a case–control study to identify the metabolic profiles activated
in the late second trimester/early third trimester serum of pregnant, African-American
women who would go on to experience labor dystocia at term gestation. African-American
women are nearly twice as likely as white women to be obese,[30] and 15 to 30% are more likely to have cesarean delivery.[31]
[32] We hypothesized that African-American women with labor dystocia would have distinct
metabolic profiles in late pregnancy compared with similar women who experienced normal
active labor progression, and that metabolic changes consistent with lipotoxicity
would predict labor dystocia in women with obesity.
Methods
Study Design and Ethical Approvals
Women from an ongoing prospective investigation (parent study) were included in this
project. We obtained IRB approval from Emory University for this work. We conducted
a matched case–control study, with a planned enrollment of 62 women (31 cases, 31
controls) based on an a priori power analysis. We used the MD Anderson Microarray
Sample Size tool, appropriate for metabolomics data, to estimate power. Given that
metabolomics data after log transformation are observed to follow an approximately
normal distribution, 27 samples per group are required to reach 90% power (assuming
a standard deviation of 0.7, controlling for a fold change of 2 and false positive
under 1, which is consistent with prior analyses).[33]
[34]
[35]
[36]
[37]
Study Participants
To select cases and comparison controls for these analyses, all participants from
the parent study[38] (African-American women, age 18–45 years) were first assessed for eligibility. Women
meeting inclusion criteria (term delivery [gestational age ≥ 37 0/7 weeks], nondiabetic
(no gestational diabetes mellitus nor pre-existing diabetes], who completed active
phase labor [4–10 cm of cervical dilation with regular contractions] with a nonanomalous
singleton fetus in vertex presentation) had their medical records abstracted. Next,
cases of labor dystocia and controls with normal labor progress were selected, matching
individually on BMI at hospital admission (±3 kg/m2) and maternal age (±5 years).
Medical chart abstractions were conducted by two experienced clinicians using a standardized
chart abstraction tool to ascertain maternal demographic factors, prenatal complications,
gestational age at time of delivery, and details of delivery. Maternal delivery BMI
was calculated from height at first prenatal visit and weight at the time of hospital
admission for term labor (or at last prenatal visit occurring at 36 weeks gestation
or later, if hospital admission weight was not available). Information on cervical
progress was collected from each examination during labor. Repeat rater reliability
testing was performed to confirm chart abstraction accuracy, showing greater than
98.0% accuracy on 10% of the abstracted charts. Clinical data was entered into REDCap
and downloaded into R Statistical Software 3.5.2 (R Foundation Vienna, Austria) for
data cleaning and statistical analyses.
Active phase labor durations were calculated for each woman who had regular, progressive
contractions from 4 to 10 cm dilation. The slowest labors were chosen as cases, with
controls selected from among the fastest labors while matching for maternal hospital
admission BMI and age. Both BMI and age were also included as covariates in multivariate
analyses to control for possible confounding caused these matching factors.[39] Investigators using metabolomics have shown that obesity causes profound changes
in pregnant women that are distinguishable in a variety of body fluids and tissue
samples and suggest multiple ways in which obesity changes metabolism.[40]
[41] For this reason, we examined the within-group influence of maternal delivery BMI
by creating obese and nonobese groups (BMI at hospital admission < or ≥ 30 kg/m2) of cases and controls for metabolomics analyses. Independent risk factors for labor
dystocia were either used as exclusion criteria, balanced in case–control groups,
used as covariates in multivariate analyses, or not controlled in analyses if they
occurred during active phase labor as part of the causal pathway ([Supplementary Table 1]).[42]
[43]
Frequencies and percentages were calculated for categorical variables describing maternal
demographic, pregnancy, and labor characteristics across case–control groups; continuous
data were assessed using median and interquartile range (non-normal distribution)
or mean and standard deviation (normal distributions). Mann–Whitney U test or t-tests were used for continuous data and likelihood ratio tests for categorical data.
A 2-tailed α of 0.05 was used for determining statistical significance.
High-Resolution Untargeted Metabolomics
Serum samples (26–30 weeks gestation) from participants were transported on dry ice
and maintained at –80°C until analysis. Thawed samples were analyzed using high-resolution
metabolomics platform, as described.[35]
[37]
[44]
[45]
[46] Briefly, the samples were treated with acetonitrile (2:1, v/v), spiked with internal
standard mix, and centrifuged at 14,000× g for 5 minutes at 4°C to remove proteins.
Samples were maintained at 4°C in an autosampler until injection. A Thermo Scientific
Q Exactive HF Quadruple-Orbitrap mass spectrometer (Thermo Fisher, San Diego, CA)
was used, while an Orbitrap Fusion Tribrid mass spectrometer and an LTQ-Velos Orbitrap
were used as backup as necessary. Dual chromatography (anion exchange and C18 reverse-phage
columns) was used, and the coverage was enhanced by allowing positive and negative
electrospray analysis of the same samples. Samples were analyzed in triplicate to
improve reliable detection of low abundance metabolites. Samples were batch randomized,
with case–control pairs randomized and analyzed within a single batch. For quality
control purposes, six technical replicates were run at the beginning, middle, and
end of each batch of 40 samples. Raw data files were extracted and aligned using apLCMS[47] and xMSanalyzer.[48] ComBat[49] was used prior to data analysis to correct for batch effects.
Metabolomics Data Extraction
The liquid chromatography mass-spectrometry (LC-MS/MS) data were processed by computational
methods designed for high-resolution mass spectrometry[47]
[48] to extract and quantify metabolite features. The resulting data contained 16,481
ions (defined by mass to charge ratio [m/z] and retention time), which were filtered by detection quality metrics. For categorical
variables, we checked for sparse cells and regrouped categories if necessary. For
missing data, sensitivity analyses were performed to evaluate the impact of missing
data on results, with consideration for implementing multiple imputation of missing
values. All differential analyses were performed using xmsPANDA package v. 1.0.7.5.[50] The comparison between groups with or without labor dystocia produced a statistic
score to rank metabolic features, according to how likely they differentiated between
the two groups. Features were filtered if they were present in less than 80% across
all samples or were present in less than 80% in one group, log2 transformed, and quantile
normalized. Missing features were replaced with median values. We selected different
metabolite features from the comparison groups using linear regression with a p-value < 0.05. Multiple hypothesis correction was performed using the Benjamini–Hochberg
false discovery rate (FDR) correction method with threshold q-value of 0.20.[51] Because the goal of this study was to explore metabolic differences among women
with and without labor dystocia in pregnancy for future hypothesis-testing, we also
used raw p-values for visualization methods and metabolic pathway enrichment analysis.
Analyses were also conducted after adjustment for maternal BMI and age. In building
multivariable models, we checked linearity assumptions for continuous predictors and
considered higher-order terms as appropriate. Type 1 (-log10p vs m/z) and Type 2 (-log10p vs. retention time) Manhattan plots were created to visualize
patterns of differentially expressed metabolites across all features with respect
to molecular mass and chemical properties. Two-way hierarchical clustering analysis
(HCA) was used to visualize relationship between women with and without labor dystocia
in each group.
Metabolic Pathway Analysis
Metabolic pathways analysis was conducted with the top-ranked (p < 0.05) metabolite features using Mummichog 2.0 (http://mummichog.org/).[44] Mummichog combines metabolite identification and metabolic pathway/network analysis.
Significant metabolic patterns, even if not represented by conventional pathways,
can be identified by Mummichog as clustered metabolic reactions. Pathway enrichment
analysis thereby provides a two-step approach, which protects against both type I
and type II errors.[52] By this analysis, we identified distinct metabolic profiles/networks that were associated
with labor dystocia at raw p < 0.05.[29]
[53]
[54]
[55] Metabolic pathway enrichment analyses were conducted on all samples with adjustment
for maternal BMI and age, then again within obese and nonobese groups. Significant
metabolic pathways with a minimum overlap size of 4 (e.g., number of metabolites with
significant discrimination between cases and controls within a particular metabolic
pathway) were evaluated for further analyses.
Results
A total of 287 medical record abstractions were completed. After excluding women who
did not meet inclusion criteria or were missing key information on delivery BMI or
cervical progress, a total of 165 labors remained. From these, 48 cases of labor dystocia
and 49 normal labor progression controls were selected, matching on BMI at hospital
admission and maternal age. Maternal demographic, pregnancy/labor characteristics,
and labor outcome variables were similar between cases and controls (p > 0.05) within the obese and nonobese groups, with the exception of epidural anesthesia
(higher in nonobese cases) ([Table 1]). Median active phase labor duration (4–10 cm) was 10.9 hours in cases and 1.9 hours
in controls across all maternal BMIs.
Table 1
Maternal demographic, pregnancy and labor characteristics, and labor outcomes by comparison
groups
|
Total sample (n = 97)
|
Labor dystocia case
|
Normal labor progression control
|
p-Value difference between cases and controls
|
|
BMI < 30 kg/m2
(n = 19)
|
BMI ≥ 30 kg/m2
(n = 29)
|
BMI < 30 kg/m2
(n = 20)
|
BMI ≥ 30 kg/m2
(n = 29)
|
Nonobese[a]
|
Obese[b]
|
|
Maternal demographic
|
|
Race: African-American (n, %)
|
97 (100)
|
19 (100)
|
|
20 (100)
|
|
1
|
1
|
|
Insurance status (n, %)
|
|
|
|
|
|
0.49
|
0.17
|
|
Private
|
19 (19.6)
|
2 (10.5)
|
8 (27.6)
|
4 (20.0)
|
5 (17.2)
|
|
|
|
Medicaid
|
78 (80.4)
|
17 (89.5)
|
21 (72.4)
|
16 (80.0)
|
24 (82.8)
|
|
|
|
Maternal age (years, median ± IQR)
|
24.0 (7)
|
22.0 (3)
|
25.0 (10)
|
24.0 (6)
|
25.0 (8)
|
0.27
|
0.43
|
|
Pregnancy characteristics
|
|
Prepregnancy BMI (kg/m2, median ± IQR)
|
28.3 (11.4)
|
21.9 (5.2)
|
31.7 (11.7)
|
22.7 (2.9)
|
33.7 (8.6)
|
0.80
|
0.93
|
|
Gestational weight gain (kg, median ± IQR)
|
11.3 (8.8)
|
10.4 (7.7)
|
14.0 (14.0)
|
10.5 (6.3)
|
9.5 (9.6)
|
0.71
|
0.70
|
|
Delivery BMI (kg/m2, median ± IQR)
|
32.6 (10.3)
|
26.9 (4.7)
|
36.7 (10.1)
|
26.7 (2.6)
|
37.0 (7.5)
|
0.95
|
0.35
|
|
Parity (n, %)
|
|
|
|
|
|
0.62
|
0.43
|
|
Nulliparous
|
47 (48.5)
|
8 (42.1)
|
16 (55.2)
|
10 (50.0)
|
13 (44.8)
|
|
|
|
Multiparous
|
50 (51.5)
|
11 (57.9)
|
13 (44.8)
|
10 (50.0)
|
16 (55.2)
|
|
|
|
Labor characteristics
|
|
|
|
|
|
|
|
|
Gestational age at labor admission (weeks, days, median ± IQR)
|
39.2 (1.9)
|
38.6 (2.0)
|
39.4 (1.4)
|
38.6 (1.6)
|
39.3 (2.0)
|
0.67
|
0.59
|
|
Mode of labor onset (n, %)
|
|
|
|
|
|
0.64
|
1
|
|
Induced labor
|
42 (43.3)
|
5 (26.3)
|
16 (55.2)
|
5 (25.0)
|
16 (55.2)
|
|
|
|
Spontaneous labor
|
55 (56.7)
|
14 (78.7)
|
13 (44.8)
|
15 (75.0)
|
13 (44.8)
|
|
|
|
Cervical ripening in labor (n, %)
|
29 (29.9)
|
2 (10.5)
|
12 (41.4)
|
5 (25.0)
|
10 (34.5)
|
0.23
|
0.59
|
|
Synthetic oxytocin in labor (n, %)
|
59 (60.8)
|
11 (57.9)
|
23 (82.1)
|
9 (45.0)
|
16 (55.2)
|
0.42
|
0.03
|
|
Type of membrane rupture (n, %)[c]
|
|
|
|
|
|
0.27
|
0.16
|
|
Spontaneous
|
47 (48.5)
|
9 (47.4)
|
9 (31.0)
|
14 (70.0)
|
15 (51.7)
|
|
|
|
Artificial
|
49 (50.5)
|
10 (52.6)
|
19 (65.5)
|
6 (30.0)
|
14 (48.3)
|
|
|
|
Epidural in labor (n, %)
|
73 (75.3)
|
16 (84.2)
|
25 (86.2)
|
11 (55.0)
|
21 (72.4)
|
0.04
|
0.19
|
|
Active labor duration
(hours, median ± IQR)
|
6.5 (8.9)
|
8.25 (6.8)
|
11.4 (4.9)
|
1.35 (2.0)
|
2.5 (2.3)
|
<0.001
|
<0.001
|
|
Labor outcomes
|
|
Type of delivery (n, %)
|
|
|
|
|
|
0.08
|
0.09
|
|
Spontaneous vaginal
|
84 (86.6)
|
17 (89.5)
|
23 (79.3)
|
20 (100)
|
24 (82.8)
|
|
|
|
Vacuum-assisted vaginal
|
11 (11.3)
|
2 (10.5)
|
4 (13.8)
|
0
|
5 (17.2)
|
|
|
|
Cesarean
|
2 (2.1)
|
0
|
2 (6.9)
|
0
|
0
|
|
|
|
Postpartum hemorrhage (>500 mL, n %)
|
6 (6.2)
|
0
|
4 (13.8)
|
0
|
2 (6.9)
|
–
|
0.38
|
|
Newborn gender (n, %)
|
|
|
|
|
|
0.87
|
1
|
|
Male
|
50 (51.5)
|
10 (52.6)
|
15 (51.7)
|
10 (50)
|
15 (51.7)
|
|
|
|
Female
|
47 (48.5)
|
9 (47.4)
|
14 (48.3)
|
10 (50)
|
14 (48.3)
|
|
|
|
Newborn NICU admission (n, %)
|
5 (5.3)
|
0
|
2 (6.9)
|
2 (10.5)
|
1 (3.6)
|
0.09
|
0.57
|
|
Neonatal birthweight (grams, mean ± SD)
|
3119.3 (394.4)
|
2965 (571.0)
|
3060.0 (533.0)
|
3070.0 (446.5)
|
3175.0 (465.0)
|
0.83
|
0.42
|
Abbreviations: BMI, body mass index; IQR, interquartile range; NICU, neonatal intensive
care unit; SD, standard deviation.
a,b
p-Value for comparison of cases and controls within nonobese[a] and obeseb groups. Likelihood ratio significance testing for categorical variables and Mann–Whitney
U test used for continuous variables. p-Values < 0.05 are shown in bold font.
c One case of “uncertain” method for rupture of membranes in labor dystocia case in
obese group.
After quality filtering, a total of 8,792/16,481 (53.3%) of detected ions (defined
by m/z and retention time) were identified in mass spectral data from LC-MS/MS analysis
of serum from participants for further analyses. Metabolic features that differentiated
between labor dystocia cases and controls were selected in subgroups of obese women
(140 features) and nonobese women (109 features) from the total sample. Using two-way
HCA with these features, we observed separation between cases and controls that was
also visible among significant features using principal components analysis ([Fig. 1], Ai-ii and Bi-ii). No individual metabolites were significant by FDR q-value of 0.20.
Fig. 1 Two-way hierarchical cluster analysis (HCA) (i), principal components analysis (PCA)
(ii), and metabolic pathway analysis (iii) showing metabolites and metabolic pathways
differentially expressed between women with labor dystocia and normal labor controls
among (A) obese and (B) nonobese women caption: Each column in HCA represents a participant
and each row represents a metabolite feature. Red hues indicate metabolites with enhanced
levels, and blue hues represent metabolites with lower metabolite concentrations.
Labor dystocia is represented in green and normal labor in red across x-axis. Red
triangles in PCA represent normal labor and green triangles represent labor dystocia.
(A) Obese women, (i) HCA with top 140/8813 metabolites with a raw p < 0.05 is shown and (ii) PCA and (iii) Mummichog enriched metabolic pathways differentiating
labor dystocia in matched groups (B) Nonobese women, (i) HCA with top 109/8704 metabolites with a raw p < 0.05 is shown and (ii) PCA and (iii) Mummichog enriched metabolic pathways differentiating
labor dystocia in matched groups of (A) obese and (B) nonobese women shown with dotted green line represents pathways significant at p < 0.05. Metabolic pathway analyses include adjustment within each group for maternal
body mass index (continuous, at hospital admission for labor) and maternal age.
Next, we examined metabolic pathways enriched in cases of labor dystocia within (a)
the entire sample, (b) obese women, and (c) nonobese women ([Fig. 1], A-iii and B-iii and [Table 2]). Using the pathway enrichment program Mummichog, we determined which of the m/z ions differing at raw p < 0.05 were also enriched in specific metabolic pathways after adjusting for the
influence of maternal BMI and age. Across all three comparison groups, cases of labor
dystocia differed from normal labor controls in androgen and estrogen biosynthesis
pathway metabolites (total sample, p = 0.028, overlap size 6 of 45; nonobese women, p = 0.036, overlap size 5 of 42; obese women, p = 0.035, overlap size 5 of 45).
Table 2
Metabolite pathways significantly dysregulated in labor dystocia in multivariate analyses
|
Pathways
|
Total sample
|
BMI < 30 kg/m2
|
BMI ≥ 30 kg/m2
|
|
Overlap size
|
Pathway size
|
p-Value
|
Overlap Size
|
Pathway size
|
p-Value
|
Overlap Size
|
Pathways size
|
p-Value
|
|
C21-steroid hormone biosynthesis and metabolism
|
11
|
64
|
0.001
|
9
|
60
|
0.003
|
|
|
|
|
Glycosphingolipid metabolism
|
5
|
30
|
0.017
|
5
|
30
|
0.010
|
|
|
|
|
Tryptophan metabolism
|
7
|
53
|
0.021
|
|
|
|
|
|
|
|
Fatty acid activation
|
|
|
|
6
|
33
|
0.005
|
|
|
|
|
Androgen and estrogen biosynthesis and metabolism
|
6
|
45
|
0.028
|
5
|
42
|
0.036
|
5
|
45
|
0.035
|
|
Bile acid biosynthesis
|
|
|
|
4
|
31
|
0.041
|
|
|
|
Abbreviations: BMI, body mass index; C21, 21 carbon.
Significant metabolite pathways predicting labor dystocia shown for total sample and
in subgroups defined by maternal obesity using raw p-value multivariate analysis with adjustment within each group for maternal BMI (continuous,
at hospital admission for labor) and maternal age. Pathways with an overlap size of
at least 4 are shown (i.e., pathway included at least 4 metabolites that significantly
differentiated the case–control groups).
Among women in the nonobese group, C21-steroid hormone biosynthesis and metabolism
pathway metabolites significantly differentiated labor dystocia cases and controls.
This metabolic pathway also differentiated women with labor dystocia in the total
sample after adjusting for the influence of maternal BMI and age, but did not differentiate
women in the obese group who experienced labor dystocia. As measured in the untargeted
LC-MS/MS, serum levels of metabolites in the C21-steroid pathway were lower among
women who went on to experience labor dystocia compared with control samples.
Similar to the C21-steroid pathway findings, glycosphingolipid pathway metabolites
were also overrepresented among the features discriminating women with labor dystocia
from controls in both the total sample and nonobese group, but not the obese group
([Table 2]). By contrast, only women in the nonobese group with labor dystocia were differentiated
from normal labor controls by metabolites mapped to the bile acid biosynthesis and
fatty acid activation pathways. Among m/z features linked to the bile acid biosynthesis and fatty acid activation pathways,
all were relatively higher in serum from women with normal labor progression, compared
with nonobese women who experienced labor dystocia (0.23–3.59-fold increase in controls
vs. cases, p = 0.04–0.47). Neither of these pathways differentiated labor dystocia case–controls
in the obese group nor in the total sample once maternal BMI and age were included
in multivariate models.
When the influence of maternal BMI and age was held constant, tryptophan metabolism
significantly differentiated labor dystocia cases and controls across all women ([Table 2]). Among metabolites tentatively matched from this pathway to untargeted LC-MS/MS
ions, women with normal labor progression had relatively higher circulating serum
levels than women with labor dystocia (1.32–2.28-fold increase in controls vs. cases,
p = 0.03–0.004).
Discussion
In this case–control study, we present evidence that labor dystocia was predicted
by different metabolic pathways in late second/early third trimester serum among obese
versus nonobese African-American women, and only androgen and estrogen biosynthesis
pathway metabolites predicted labor dystocia across both obese and nonobese groups.
A range of factors can slow labor progress, thereby increasing women's risk for unplanned
cesarean birth. However, there is growing recognition that in some women, there is
a phenotype of labor dystocia that complicates the labor process and can worsen labor
outcomes.[17]
[24] Maternal BMI above normal ranges appears to be one part of this labor dystocia phenotype,
as evidenced by clinical investigations showing increases in labor length[13]
[56] and risk for cesarean delivery[57] among women with higher BMIs. However, labor dystocia is not confined to obese women.[21]
[38]
[39] Additionally, not all pregnant women with obesity go on to experience labor dystocia
and cesarean birth.[58] Thus, there is a need to identify biomarkers of labor dystocia.[12]
Sex hormone metabolism was associated in this study with labor dystocia in both the
obese and nonobese groups, and in the total sample after adjustment for maternal age
and BMI. We repeated metabolomics analyses while adjusting for fetal gender, but saw
no difference in associations between labor dystocia and sex hormone metabolism pathways
(analyses not shown). In other investigations, women with obesity were noted to have
increased levels of estrogen and testosterone both within[59] and outside of pregnancy.[60] In addition, lower serum levels of estrogen and progesterone[61] and single nucleotide polymorphisms of sex hormone-related genes[62] were associated with preeclampsia in two recent investigations. We did not locate
published studies reporting links between sex hormone metabolites and labor dystocia.
However, given close links between maternal obesity and both hypercholesterolemia[63] and preeclampsia,[64] further research with targeted measures of specific sex and steroid hormones in
women with active labor slowing is warranted.
Contrary to our prediction that metabolic changes consistent with lipotoxicity would
predict labor dystocia in women with obesity, we found more evidence for associations
between lipotoxicity and labor dystocia among nonobese women in our sample. Both fatty acid activation and glycosphingolipid metabolism
pathways were increased in nonobese women who had fast active labor progression (controls).
These observations hint that more efficient mobilization of lipids from the diet and
adipose tissue for use by energy-requiring tissues may protect pregnant women from
cellular damage and dysregulation associated with lipotoxicity, thereby leading to
faster labor progression. We also observed decreases in C21-steroid hormone metabolites
among nonobese women who went on to experience labor dystocia. The C21 steroid hormone
metabolic pathway involves production of both progestins and adrenal corticosteroids,
and is closely related to the C-19 and C-18 pathways, which produce androgens and
estrogens, respectively.[65]
There is evidence from mouse and human models that increased fatty acid oxidation
decreases systemic free fatty acid and dyslipidemia[66]
[67]—both landmarks of lipotoxicity. In addition, glycosphingolipid metabolism—the other
pathway associated with labor dystocia among nonobese women in this study—is also
related to lipotoxicity. Glycosphingolipids are a group of bioactive glycolipids that
are involved in signal transduction, cholesterol lipid rafts, and cellular communication.[68] Our observation of relatively lower levels of metabolites tentatively mapped to
glycosphingolipid metabolism in nonobese women with labor dystocia is interesting,
given evidence that cholesterol lipid rafts may be involved with an unstable, asynchronous
uterine contractile pattern in a mouse model.[69] Myometrial lipotoxicity involving accumulation of lipids and reduction in muscle
mass in uterine tissue from obese pregnant women were recently demonstrated by Gam
et al.[22] Building on these findings, we theorize that labor dystocia cases among nonobese
women in our investigation were more likely than normal-labor comparison subjects
to have lipotoxicity-related changes in their uterus that ultimately slowed their
labor progression. For example, as a result of problems with fatty acid oxidation,
their myometrial tissue may have accumulated lipid deposits, thereby reducing contractile
force and frequency during labor and slowing cervical dilation. In addition, glycosphingolipid
metabolite alterations may have decreased these women's capacity to maintain a stable,
synchronous uterine contractile pattern—thereby further impeding labor progress.
As mentioned above, we did not find evidence for associations between lipotoxicity
and labor dystocia among obese women in our sample. It is possible that lipotoxicity may be a more important differentiator
of the labor dystocia phenotype among women who are not yet obese, but show evidence
of less effective metabolism of fats. It is not known how circulating levels of metabolites
from the glycosphingolipid or fatty acid activation pathways in obese women compared
with nonobese women in our sample, as no m/z features from these pathways were also annotated at a high confidence level. However,
lower fatty acid oxidation and lipid accumulation have been correlated with higher
maternal BMIs in umbilical stem cells of neonates born to obese versus normal weight
women.[70] These data suggest that reduced lipid oxidation during pregnancy is an important
facet of obesity-related changes in the maternal–fetal pair. It is possible that perturbed
fatty acid activation and glycosphingolipid pathways were therefore present in most
or all of the women from our sample of obese women, thereby reducing heterogeneity
of these pathways for the purposes of our metabolic pathway analysis. Future targeted
metabolomics investigations are needed to quantify levels of metabolites involved
in lipotoxicity among women of different BMIs and compare lipotoxicity metabolite
exposure to labor durations.
Also supported in the literature was our unexpected finding of relatively higher levels
of features linked to the tryptophan metabolism pathway in women with normal labor
progress after adjustment for the influence of maternal BMI and age. The tryptophan
metabolic pathway is important during pregnancy for fetal development, neuronal protection,
and serotonin signaling.[71] Serotonin signaling deficiencies may be linked to depression or anxiety in pregnancy,
and both maternal trauma history[72] and prenatal depression and anxiety[73]
[74]
[75] were associated with labor dysfunction in previous clinical investigations. In addition,
serotonin stimulates uterine smooth muscle contractions in pregnant women.[76] Thus, relative upregulation of tryptophan metabolism in women with normal labor
progress could suggest possible links between maternal mood, uterine contractility,
and labor duration. However, further investigation is needed to target metabolites
of the tryptophan metabolic pathway, including serotonin, in pregnant women who experience
labor dystocia and other difficulties of parturition.
The application of metabolomics to understand pregnancy risks is still a young field.
In 2009, the first review on this topic was published based on limited data.[77] Since then, emerging studies demonstrate the promising possibilities of this technology
in the complex field of pregnancy where a mother–baby pair are interacting for mutual
benefit.[78]
[79]
[80]
[81]
[82] Most recently, metabolomics was used to identify specific metabolites that were
predictive of women's eventual development of preeclampsia[83]
[84] and of their neonate's level of adiposity.[85]
Metabolomics is also useful for understanding labor processes. Several teams of investigators
differentiated between women in different phases of labor using metabolomics.[86]
[87] As mentioned, metabolic profiles can be generated from a variety of biological fluids;
this methodologic advantage was recently utilized to demonstrate biomarkers present
in amniotic fluid and maternal serum collected during women's second trimester that
predicted later preterm delivery.[88] Metabolomics has also been used to identify biomarkers produced in mice uterine
tissue following pharmacologic stimulation versus inhibition of contractions, revealing
new detail on the mechanisms of uterine contractility in rodents.[89]
Our study has several strengths. First, potential differences in labor duration that
could potentially be caused by differences in labor management were offset in this
study by our focus on participants who were managed by the same team of physicians
and midwives following the same protocols across two hospitals.[42] Another strength of this study is our careful consideration of variables known to
associate with the duration of labor in our choice of inclusion/exclusion criteria,
case–control group assignment, and statistical covariates. A range of factors can
affect labor progress, including maternal age,[90] BMI,[13] maternal parity,[11] and mode of labor onset.[91] In this study, we adjusted for the influence of maternal age, and matched cases
with controls on mode of labor onset and BMI. In bivariate comparisons, we saw no
difference after controlling for the above-mentioned factors in maternal parity. An
additional strength was our use of high-resolution LC-MS/MS combined with powerful
analytic tools to provide broad coverage of both low- and high-abundance metabolites
that are potentially relevant to labor dystocia pathophysiology.
Our study also has several limitations. First, the retrospective nature of this analysis
means that we cannot differentiate between cause and effect, but only on associations
between metabolic pathways present in late pregnancy serum and term labor dystocia.
Second, although our sample size exceeded a priori calculations, larger numbers of
participants—especially in the nonobese group—would strengthen these analyses. In
addition, our metabolomics samples were collected in the late second/early third trimester
of pregnancy. Future investigations might better capture the metabolic signature of
a labor dystocia phenotype by also including a first trimester sample and a late third
trimester sample. With these repeated measures, a longitudinal analysis might be conducted
where intraindividual might be filtered out, leaving a clearer signal for significant
m/z features and metabolic pathways distinguishing women with labor progression difficulty.
We were also not able to consider whether the timing of late-pregnancy serum influenced
the strength of metabolic signals in this analysis due to small sample size. Future
longitudinal analysis with larger groups of samples would allow us to consider how
metabolic levels in these critical pathways might vary across pregnancy. In our analyses
of individual metabolites, we were unable to correct for multiple comparisons, as
no metabolite reached the FDR q-value < 0.20. Possibly, in a future analysis involving longitudinal measures or standardization
in sample collection by time of day and dietary intake, this threshold would be reached.
However, our use of metabolic pathway analysis using features significant at raw p-value provides a 2 step approach that protects against both type I and type II errors
while revealing networks of perturbed metabolites.[52] This approach allowed us to explore metabolic differences among women with and without
labor dystocia in pregnancy for future hypothesis-testing using targeted metabolomics
approaches while employing orthogonal methods to confirm findings. Finally, our study
is limited by our inclusion of only one race of women for all analyses. However, given
the disparities experienced by African-American women in rates of cesarean delivery[92] and maternal morbidity/mortality[93]
[94] and the benefits of looking intrarace for differences in high-throughput methodologies
like metabolomics,[95] this study's focus on one race is a logical choice. Metabolic pathways associated
with labor dystocia in African-American women from this study should be further investigated
in future studies that might include samples with a range of maternal races, standardization
in sample collection by time of day and dietary intake, and targeted analysis of associated
metabolites.
We need new approaches to improve birth outcomes in women with a labor dystocia phenotype.[58]
[96]
[97] As a methodology that covers the spectrum from hypothesis development to hypothesis
testing, metabolomics offers investigators the ability to quantify multiple metabolites
in a single sample and flexibility in the type and amount of sample required for analyses.
Thus, metabolomics offers considerable promise in the investigation of mechanisms
contributing to labor dystocia among women of different race/ethnicities, with or
without obesity. With a fuller understanding of the unique metabolic mechanisms contributing
to labor dystocia, we step closer to breaking the cascade of health complications
resulting from poor labor outcomes.
