Objective
The objective of the study was to determine whether pregnancies resulting in early preterm birth (PTB) (<30 weeks) were more likely than term pregnancies to have elevated midtrimester serum tumor necrosis factor alpha (TNF-α) levels combined with lipid patterns suggestive of hyperlipidemia.
Study Design
In 2 nested case-control samples drawn from California and Iowa cohorts, we examined the frequency of elevated midpregnancy serum TNF-α levels (in the fourth quartile [4Q]) and lipid patterns suggestive of hyperlipidemia (eg, total cholesterol, low-density-lipoproteins, or triglycerides in the 4Q, high-density lipoproteins in the first quartile) (considered independently and by co-occurrence) in pregnancies resulting in early PTB compared with those resulting in term birth (n = 108 in California and n = 734 in Iowa). Odds ratios (ORs) and 95% confidence intervals (CIs) estimated in logistic regression models were used for comparisons.
Results
Early preterm pregnancies were 2-4 times more likely than term pregnancies to have a TNF-α level in the 4Q co-occurring with indicators of hyperlipidemia (37.5% vs 13.9% in the California sample (adjusted OR, 4.0; 95% CI, 1.1–16.3) and 26.3% vs 14.9% in the Iowa sample (adjusted OR, 2.7; 95% CI, 1.1–6.3). No differences between early preterm and term pregnancies were observed when TNF-α or target lipid abnormalities occurred in isolation. Observed differences were not explicable to any maternal or infant characteristics.
Conclusion
Pregnancies resulting in early PTB were more likely than term pregnancies to have elevated midpregnancy TNF-α levels in combination with lipid patterns suggestive of hyperlipidemia.
Studies have reported a link between preterm birth (PTB) and increased midpregnancy levels of tumor necrosis factor alpha (TNF-α) as well as PTB and midpregnancy lipid levels ; the combined influence of these factors on the risk of PTB has not been explored.
Given the established link between TNF-α and lipid release from adipocytes (TNF-α induced lipolysis) as well as the association between TNF-α and lipid metabolism, we hypothesized that in some instances PTB risk could be associated with the co-occurrence of increased TNF-α levels and lipid levels. Most specifically when elevated TNF-α occurs in combination with hyperlipidemia (eg, suggested by increased total cholesterol [TC], low-density lipoproteins [LDLs], or triglycerides [TGs] or low high-density-lipoproteins [HDLs]). The importance of these combined influences on risk have been investigated in studies focused on gestational diabetes and preeclampsia as well as studies focused on obesity, cardiovascular disease, insulin sensitivity, and metabolic syndrome.
Here we used second-trimester serum samples collected as part of routine second-trimester screening for aneuploidies and neural tube defects to investigate the potential interrelationships between TNF-α and lipid levels for their contribution to the risk in pregnancies resulting in early PTB (<30 weeks) compared with term pregnancies. We examined these associations in 2 nested case-control samples drawn from California and Iowa cohorts.
Materials and Methods
The relationships between midpregnancy TNF-α, lipids patterns, and early PTB (less than 30 weeks) were examined in 2 independent nested case-control samples of pregnancies from California (n = 108) and Iowa (n = 734). Included from California were 72 case pregnancies resulting in early PTB and 36 singleton control pregnancies resulting in term birth (≥37 completed weeks of gestation). Case and control pregnancies were drawn from a larger cohort of more than 1.4 million pregnancies who underwent routine prenatal screening in 2005-2008, of whom 662,889 had no indication of chromosomal or structural defects, smoking, diabetes, or amniotic fluid abnormalities present in prenatal screening, had a birth certificate, or for cases, neonatal intensive care unit (NICU) records.
As described in earlier work, 1928 pregnancies in this cohort delivered before 30 weeks. PTB less than 30 weeks was our focus for the original study, given that in California, there is routine data collection in NICUs for greater than 90% of newborns with birth before 30 weeks. Our focus in the source study was bronchopulmonary dysplasia (BPD) in PTBs less than 30 weeks, which relied on these records.
The present study used data from this source study and associated prenatal screening specimens that had been stored for a portion of these pregnancies as part of regular banking protocols in the state (wherein there is routine banking of specimens in 3 geographic areas [the California Central Valley, Los Angeles and Orange County, and San Diego County]). Thirty-nine of the 246 pregnancies resulting in PTB less than 30 weeks with BPD in the source study had stored midpregnancy samples (18 females and 21 males).
For biomarker testing, we implemented a balanced design wherein 18 of the 21 males with PTB less than 30 weeks and BPD were randomly selected and grouped with the 18 females (n = 36 total). We then randomly selected 18 females and males from stored samples for PTBs less than 30 weeks and no BPD (from a total stored sample of 115 of 1682 in the original set [62 females and 53 males]) and 18 females and males from stored samples from the term births [from a total stored sample of 63,312 of 575,387 in the original set (32,318 females and 30,994 males]).
To best mirror California analyses with respect to PTB less than 30 weeks but without any stratification by BPD or other exclusions, included from Iowa was a complete sample of 57 case pregnancies resulting in PTB less than 30 weeks drawn from a total sample of 12,057 singleton pregnancies with second-trimester screening through the Iowa prenatal screening program in May 2009 through November 2010. Also included were 677 singleton control pregnancies resulting in term birth selected at an approximate 1:1 ratio to all PTBs (<37 weeks, n = 657). Additional details regarding the cohort from which Iowa cases and controls were drawn have been described elsewhere.
Information on maternal characteristics and related data
In both samples, information on maternal characteristics, measurements on routine second-trimester serum screening tests, and information on infant characteristics were abstracted from state prenatal screening program data (the California Prenatal Screening Program within the Genetic Disease Screening Program and the Iowa Prenatal Screening Program) and from state birth certificate records. Included from screening records was maternal race/ethnicity, weight at testing, age at testing, and gestational weeks at the time of second-trimester screening. Other data extracted from birth certificates included parity, birthweight, and gestational age.
For Iowa, information on body mass index (BMI), diabetes, congenital defects (chromosomal or other), information on whether the PTB was spontaneous or medically indicated, and the presence of other factors closely linked with PTB (hypertension, diabetes, previous PTB) were also abstracted from birth certificates. In California, prenatal screening records, hospital discharge records (for cases and controls), and NICU records (for cases) were used along with birth certificates to make exclusions for smoking history, diabetes, structural or chromosomal defects to determine whether the PTB was spontaneous or medically indicated and to determine whether there was any presence of hypertension, diabetes, or previous PTB.
In California, no data on maternal height were present in any record for these birth years. As such, we relied on a diagnostic code for BMI of 30 kg/m 2 or greater to determine obesity. In California, NICU records were also used to identify cases with and without BPD and other indicators of case morbidity including intraventricular hemorrhage and retinopathy of prematurity. In both California and Iowa samples, case pregnancies resulting in spontaneous PTB were considered to be those in which vital statistic birth certificate records (California and Iowa) or hospital discharge records (California) included a flag for premature labor, premature rupture of membranes, or tocolytic administration.
Case pregnancies resulting in medically indicated premature birth were considered to be those without premature labor or premature rupture of membranes for whom there was a flag for medical induction or artificial rupture of membranes or for whom there was a cesarean section given birth less than 37 weeks and none of the aforementioned flags. Case pregnancies without any of these specific codes were considered as having an unknown PTB subtype.
Serum samples and biochemical testing
In both California and Iowa samples, nonfasting maternal serum was drawn between 15 and 20 weeks of gestation. Serum was stored in 1 mL tubes at –80°C until thawed for testing. TNF-α testing in the California serum samples was performed by the Human Immune Monitoring Center at Stanford University using a human 51-plex kit that was purchased from Affymetrix Inc (Santa Clara, CA) and read using a Luminex 200 instrument (Austin, TX) in accordance with the manufacturer’s recommendations. Details regarding the use of this kit by the Human Immune Monitoring Center have been described elsewhere.
Median fluorescence intensity values were reported for TNF-α using Masterplex software (Hitashi Corp, San Bruno, CA) wherein for TNF-α, sensitivity and lower level of detect (LLOD) was 4.8 pg/mL and interassay coefficient of variation was 8.81%. TNF-α at the University of Iowa was measured using a Millipore high-sensitivity kit (HSCYTO-60SPMX13; Millipore, Billerica, MA), read using a Bio-Rad Bioplex 200 instrument (Bio-Rad, Hercules, CA), and reported using Bio-Rad Bioplex Manager software (sensitivity and LLOD, 0.05 pg/mL and interassay coefficient of variation, 3.49%) (all used in accordance with manufacturer recommendations).
All cases and controls in California and Iowa had TNF-α levels above the assay-specific LLOD thresholds. Lipids in both samples were measured at the State Hygienic Laboratory at the University of Iowa using a Roche Diagnostics c111 Cobas Analyzer (Basel, Switzerland) (TC, calibrated range, 9.70–800.00 mg/dL; LDL, calibrated range, 3.86-548.00 mg/dL; HDL, calibrated range, 3.00–120.00 mg/dL; and TGs, calibrated range, 8.85–885.00 mg/dL).
Analyses
To test whether there is evidence of a link between elevated TNF-α and hyperlipidemia in pregnancies resulting in early PTB (as has been demonstrated in studies focused on other related outcomes [eg, gestational diabetes, preeclampsia, obesity, cardiovascular disease, insulin sensitivity]), elevated serum TNF-α levels (defined as in the fourth quartile [4Q]) and lipid patterns suggestive of hyperlipidemia (including elevated TC, LDL, TGs, elevated ratios of TG:HDL and LDL:HDL [defined as in 4Q], low HDL levels, or a low ratio of HDL:TC (defined as in the first quartile [1Q]) were compared in case and control pregnancies using logistic regression models. Elevated TNF-α and lipid patterns suggestive of hyperlipidemia were evaluated as independent predictors as well as combined predictors for risk of PTB. Quartiles were generated for California and Iowa samples separately based on the distribution of each analyte in control pregnancies.
Models were adjusted for maternal weight and gestational week at serum draw to account for differences in blood volume and normal variations in lipid levels during pregnancy. A χ 2 and the Wilcoxon rank-sum test were used to examine whether pregnancies ending in PTB less than 30 weeks with vs without co-occurring elevated TNF-α and lipid pattern suggestive of hyperlipidemia (TC, LDL, TGs, TG:HDL, or LDL:HDL in 4Q; HDL or HDL:TC in 1Q) differed significantly on maternal and infant characteristics.
All analyses were done using Statistical Analysis Software (SAS) version 9.3 (Cary, NC). Methods and protocols for the study were approved by the Committee for the Protection of Human Subjects within the Health and Human Services Agency of the State of California, the Institutional Review Board of Stanford University, and the Institutional Review Board of the University of Iowa.
Results
Early preterm cases and term controls in the California sample were primarily of white or Hispanic race/ethnicity (30.6% and 62.5% in cases and 27.8% and 52.8% in controls). All cases and controls in the Iowa sample were white. In California, early preterm cases weighed more at midpregnancy serum draw than term controls (161.9 compared to 140.6 pounds, z = [minus]2.8, P < .01). In Iowa, early preterm cases were less likely than controls to have given birth previously (40.3% vs 58.6%, P < .05) ( Table 1 ).
| Characteristic | California | Iowa | ||
|---|---|---|---|---|
| Cases <30 wks | Controls ≥37 wks | Cases <30 wks | Controls ≥37 wks | |
| 100.0 (n = 72) | 100.0 (n =36) | 100.0 (n = 57) | 100.0 (n = 677) | |
| Race/ethnicity, % | ||||
| White | 30.6 | 27.8 | 100.0 | 100.0 |
| Hispanic | 62.5 | 52.8 | 0 | 0 |
| Black | 1.4 | 8.3 | 0 | 0 |
| Asian | 4.2 | 8.3 | 0 | 0 |
| Other | 1.4 | 2.8 | 0 | 0 |
| Maternal age (y), % | ||||
| Younger than 18 | 1.4 | 2.8 | 0 | 1.6 |
| 18-34 | 62.2 | 86.1 | 89.5 | 86.6 |
| ≥35 | 27.8 | 11.1 | 10.5 | 11.8 |
| Parity, % | ||||
| 1 | 41.7 | 38.9 | 59.7 | 41.4 |
| ≥2 | 58.3 | 61.1 | 40.3 | 58.6 a |
| Weight at test (lbs), mean (SD) | 161.9 (42.2) a | 140.6 (32.4) | 177.4 (48.9) | 171.4 (41.7) |
| Week at test (15-20), mean (SD) | 16.8 (1.3) | 16.9 (1.4) | 16.7 (1.3) | 16.8 (1.3) |
When evaluated without consideration of co-occurring indicators of hyperlipidemia, California cases, but not Iowa cases, were nearly 3 times more likely than term controls to have TNF-α levels in the 4Q (odds ratio [OR], 2.8; 95% confidence interval (CI), 1.1–7.0). When lipids were evaluated without consideration of co-occurrence with TNF-α levels, in both samples case pregnancies were more than twice as likely as controls to have TGs in the 4Q (OR, 3.4, 95% CI, 1.4–8.4 in California, and OR, 2.1; 95% CI, 1.2–3.7 in Iowa) or HDL:TC in the 1Q (OR, 3.2; 95% CI 1.3–8.3 in California and OR, 2.0; 95% CI, 1.1–3.5 in Iowa). These risks for lipids were attenuated among cases and controls with TNF-α levels in the 4Q.
Risks associated with lipids remained and were often increased when cases and controls were examined in combination with TNF-α levels in the 4Q ( Table 2 ). The most notable risk across samples was for co-occurring TNF-α and TGs in the 4Q. Pregnancies ending in PTB less than 30 weeks were more than 4 times as likely to exhibit this pattern in California (OR, 4.4; 95% CI, 1.2–16.8) and more than twice as likely to exhibit this pattern in Iowa (OR, 2.7; 95% CI, 1.2–6.0).
| Variable | California | Iowa | ||
|---|---|---|---|---|
| Cases <30 wks | Controls ≥37 wks | Cases <30 wks | Controls ≥37 wks | |
| n (%) OR (95% CI) a | n (%) Referent | n (%) OR (95% CI) a | n (%) Referent | |
| Sample | 72 (100.0) | 36 (100.0) | 57 (100.0) | 677 (100.0) |
| Independent models | ||||
| TNF-α 4Q b | 33 (45.8) 2.8 (1.1–7.0) c | 9 (25.0) | 18 (31.6) 1.3 (0.7–2.3) | 173 (25.6) |
| TC 4Q b | 18 (25.0) 0.9 (0.3–2.4) | 9 (25.0) | 19 (33.3) 1.5 (0.8–2.7) | 171 (25.3) |
| HDL 1Q b | 33 (45.8) 2.4 (1.0–5.9) | 9 (25.0) | 25 (43.9) 2.2 (1.3–3.8) d | 175 (25.9) |
| LDL 4Q b | 26 (36.1) 1.6 (0.6–4.2) | 9 (25.0) | 16 (28.1) 1.2 (0.6–2.1) | 171 (25.3) |
| TGs 4Q b | 38 (52.8) 3.4 (1.4–8.4) | 9 (25.0) | 24 (42.1) 2.1 (1.2–3.7) d | 172 (25.4) |
| HDL:TC 1Q b | 40 (55.6) 3.2 (1.3–8.3) c | 9 (25.0) | 24 (42.1) 2.0 (1.1–3.5) c | 179 (26.4) |
| TGs:HDL 4Q b | 43 (59.7) 4.2 (1.7–10.7) d | 9 (25.0) | 21 (36.8) 1.7 (1.0–3.1) | 170 (25.1) |
| LDL:HDL 4Q b | 34 (47.2) 2.3 (0.9–6.1) | 9 (25.0) | 25 (43.9) 2.3 (1.3–4.0) d | 172 (25.4) |
| Lipids in which TNF-α is <4Q | ||||
| TC 4Q | 7 (9.7) 0.5 (0.1–1.5) | 6 (16.7) | 14 (24.6) 1.4 (0.7–2.6) | 131 (19.4) |
| HDL 1Q | 8 (22.2) 1.3 (0.5–3.5) | 21 (29.2) | 13 (22.8) 1.3 (0.7–2.5) | 124 (18.3) |
| LDL 4Q | 8 (22.2) 0.6 (0.2–1.8) | 13 (18.1) | 12 (21.1) 1.1 (0.6–2.2) | 132 (19.5) |
| TGs 4Q | 19 (26.4) 1.7 (0.6–4.7) | 6 (16.7) | 15 (26.3) 1.5 (0.8–2.8) | 130 (19.2) |
| HDL:TC 1Q | 23 (31.9) 1.2 (0.4–3.2) | 8 (22.2) | 14 (24.6) 1.3 (0.7–2.5) | 134 (19.8) |
| TGs:HDL 4Q | 21 (29.2) 1.4 (0.5–3.6) | 8 (22.2) | 13 (22.8) 1.3 (0.7–2.4) | 126 (18.6) |
| LDL:HDL 4Q | 17 (23.6) 0.8 (0.3–2.3) | 8 (22.2) | 14 (24.6) 1.4 (0.7–2.6) | 132 (19.5) |
| Lipids in which TNF-α 4Q | ||||
| TC 4Q | 11 (15.3) 2.0 (0.5–8.4) | 3 (8.3) | 5 (8.8) 1.5 (0.6–3.9) | 40 (5.9) |
| HDL 1Q | 12 (16.7) e | 1 (2.8) | 12 (21.1) 3.1 (1.5–6.3) d | 52 (7.7) |
| LDL 4Q | 13 (18.1) e | 1 (2.8) | 4 (7.0) 1.2 (0.4–3.5) | 39 (5.8) |
| TGs 4Q | 19 (26.4) 4.4 (1.2–16.8) c | 3 (8.3) | 9 (15.8) 2.7 (1.2–6.0) c | 43 (6.4) |
| HDL:TC 1Q | 17 (23.6) e | 1 (2.8) | 10 (17.5) 2.8 (1.3–6.1) d | 46 (6.8) |
| TGs:HDL 4Q | 22 (30.6) e | 1 (2.8) | 8 (14.0) 2.3 (1.0–5.1) | 45 (6.7) |
| LDL:HDL 4Q | 17 (23.6) e | 1 (2.8) | 11 (19.3) 3.7 (1.8–7.7) d | 40 (5.9) |
a Adjusted for gestational week at serum draw and maternal weight
b Wherein 1Q and 4 Q cut points were derived from the term controls
e Not calculated because of frequency in referent sample 1 or less.
When the frequency of co-occurring elevated TNF-α and any lipid patterns suggestive of hyperlipidemia were compared for cases and controls, 37.5% of early preterm cases in California and 26.3% of early preterm cases in Iowa (compared with 13.9% of term controls in California and 14.9% in Iowa) were found to have a TNF-α level in the 4Q and 1 or more at-risk lipid pattern suggestive of hyperlipidemia resulting in ORs of 4.0 (95% CI, 1.1–16.3) and 2.7 (95% CI, 1.1–6.3), respectively, after adjustment for gestational week and maternal weight at serum draw. No significance between group differences were noted when elevated TNF-α or at-risk lipid patterns occurred in isolation ( Table 3 and Figure ).
| Variable | California | Iowa | ||
|---|---|---|---|---|
| Cases | Controls | Cases | Controls | |
| n (%) OR (95% CI) a | n (%) | n (%) OR (95% CI) a | n (%) Referent | |
| Sample | 72 (100.0) | 36 (100.0) | 57 (100.0) | 677 (100.0) |
| No TNF-α 4Q b | 10 (13.9) | 9 (25.0) | 10 (17.5) | 206 (30.4) |
| No at-risk lipid c | Referent | Referent | ||
| TNF-α 4Q only | 6 (8.3) 1.4 (0.3–6.9) | 4 (11.1) | 3 (5.3) 0.8 (0.2–2.9) | 72 (10.6) |
| At-risk lipid only c | 29 (40.3) 1.2 (0.4–3.8) | 18 (50.0) | 29 (50.9) 2.0 (0.9–4.2) | 298 (44.0) |
| TNF-α 4Q and at-risk lipid c | 27 (37.5) 4.0 (1.1–16.3) d | 5 (13.9) | 15 (26.3) 2.7 (1.1–6.3) d | 101 (14.9) |
a Adjusted for gestational week at serum draw and maternal weight
b Wherein 1Q and 4Q cut points were derived from the term controls
c TC, LDL, TGs, TG:HDL, or LDL:HDL in the 4Q and/or HDL or HDL:TC ratio in the 1Q
