Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
There is increasing evidence that infection and the inflammatory response contribute to the etiology of spontaneous preterm birth.1–4 Genetic susceptibility factors in immune response genes have been investigated, particularly cytokine polymorphisms known to be proinflammatory.5–14 To date, the most frequently investigated single nucleotide polymorphisms (SNPs) include tumor necrosis factor (TNF)(−488), interleukin (IL)1β(+3954), and IL1β(−511), each of which appears to be associated with alterations in cytokine expression. Although previous studies have suggested associations with preterm birth, replication of results has been controversial, in part due to issues of study design (ie, insufficient power, bias in selection of cases or controls, and population stratification).
The genomic revolution has generated a catalog of common genetic variants and provided an opportunity to directly examine their contribution to complex conditions, such as spontaneous preterm and small-for-gestational age (SGA) birth.15 Many researchers advocate conducting primary investigations with haplotypes rather than SNPs because haplotypes can span larger regions of a potential gene of interest.16 Haplotypes are derived from blocks of DNA known to be in linkage disequilibrium. Because a haplotype is an extended region of linked variants that are inherited as a unit from a single chromosome, allelic variants that travel together on the same haplotype can be analyzed simultaneously (known as cis interactions).16 Additionally, a critical benefit of haplotypes is in examining haplotype diversity and frequency across populations (known as “transracial mapping”). In the case of common conditions, such as preterm and SGA births, this approach may help identify variants common across populations that are likely to be of etiological importance.16
We examined the relationships of common polymorphisms in proinflammatory cytokines with spontaneous preterm birth and SGA in a cohort of women enrolled in a prospective pregnancy study. In this report, we will describe the relationship among 10 polymorphisms on proinflammatory cytokine genes IL1α, IL1β, IL2, IL6, TNF, and lymphotoxin α (LTA), bacterial vaginosis infection, and spontaneous preterm birth and SGA. The relationship of anti-inflammatory cytokine polymorphisms with spontaneous preterm birth and SGA is described in a companion report.17
The Pregnancy, Infection, and Nutrition cohort study enrolled women between 24 and 29 weeks' gestation at Wake County Human Services Department and the Wake Medical Center/Wake AHEC from February 1996 through June 1998 and at the University of North Carolina, Chapel Hill obstetric clinics from August 1995 through June 2000.18 We collected genital tract swabs and blood from women at the intake visit. Study exclusions are detailed elsewhere.18 From the larger cohort, a nested preterm case–control sample was selected for DNA extraction, genotyping, and analysis, to yield an approximate 1:1.5 case:control ratio. An additional 100 women who were neither preterm cases nor randomly selected controls delivered an SGA infant and consented to genetic analysis. These women were added to the SGA analysis (Fig. 1). Characteristics of our study population can be found in Appendices 1 and 2. (All appendices are available with the electronic version of this article.)
Gestational age was assigned by early ultrasound measurements completed before 22 weeks' gestation. In the absence of early ultrasonography, last menstrual period date was used. Preterm birth was dichotomized as birth before 37 completed weeks of gestation. Spontaneous preterm birth included those with spontaneous membrane rupture occurring at least 4 hours before the onset of labor, and births after spontaneous labor within 4 hours of membrane rupture. Study obstetricians reviewed records of cases to determine preterm clinical presentation. For births with ambiguous information in the charts and for a 10% random sample of all preterm births, further review was conducted by an obstetrician who was unaware of the initial categorization. Infants were classified as SGA if they fell below the 10th percentile of weight for gestational age, stratified by race, sex, and parity.
Bacterial Vaginosis Assessment
Participants had genital tract swabs collected during a routine pelvic examination between 24 and 29 weeks' gestation. Dacron swabs were wiped against the lateral vaginal side wall and then rolled onto a glass microscope slide and allowed to air dry. All slides were gram-stained, counter-stained with safranin and evaluated for 3 morphotypes of bacteria. A detailed description of this assessment can be found elsewhere.19 Briefly, a score was determined by summing points assigned to each morphotype identified. Designation of bacterial vaginosis was made for vaginal flora scores of 7–10, whose vaginal pH was greater than 4.5.
DNA Extraction and Genetic Analyses
DNA was extracted from peripheral blood samples collected at the intake visit using the ABI automated DNA extractor. Genotyping for selected polymorphisms was performed using TaqMan for allelic discrimination (Table 1).20 A 10% random sample, distributed randomly across plates, was assayed in duplicate for quality control.
Deviations from Hardy–Weinberg equilibrium were examined in the control series stratified by race. Loci that violated Hardy–Weinberg in control subjects were excluded from haplotype analyses. Linkage disequilibrium, within strata of race and case-status, was assessed for SNPs that occurred on the same chromosome; pairwise comparison measures were calculated in Arlequin.21 Linkage disequilibrium was assessed for polymorphisms assuming known gametic phase using D′. D′ with an absolute value of 1 indicates loci are in complete linkage. Linkage disequilibrium also was calculated assuming unknown gametic phase using likelihood ratio tests that compared the likelihood of the sample under the hypothesis of no association between loci to the likelihood of the sample allowing the loci to be associated.22 Population haplotype frequencies in case and control groups separately and combined were estimated using the Excoffier and Slatkin EM algorithm.23 Frequency distributions were then compared using an omnibus likelihood ratio test.24 This test computes a haplotype frequency likelihood ratio statistic comparing haplotype frequency likelihoods for cases and controls separately versus combined. A null distribution of this statistic is then approximated by randomization tests to obtain empirical P values.24 We implemented 10,000 permutations of case–control status to calculate empirical P values. Within haplotype configurations that had significantly different case–control haplotype frequency distributions, we examined χ2 statistics comparing the frequencies of individual haplotypes between cases and controls, referenced against all other haplotypes in that configuration. The null distributions of these statistics were also approximated by randomization tests using 10,000 permutations of case-control status to derive empirical P values.24
Subject-level haplotypes were reconstructed using PHASE.25 PHASE assumes a coalescent prior and implements Gibbs sampling (a Bayesian methodology) to infer haplotypes using unphased genotype data. The prior expectation included in this model is that unresolved haplotypes (subjects who are ambiguously phased) tend to be similar to known haplotypes. This method has been shown to reduce error rates by greater than 50% in some circumstances.25 We excluded subjects if 50% or more of the markers had missing genotype data, or if any of the loci-specific phase probabilities dropped below 90%. Odds ratio (OR) estimates and 95% confidence intervals (CIs) compared subjects who carried the index haplotype with all other subjects. These were estimated in 2-by-2 contingency tables within strata of race.
Semi-Bayesian hierarchical logistic regression was used to obtain the OR estimate and 95% CI for the main effect of each cytokine polymorphism for SGA and spontaneous preterm birth, with separate models for African-American and white subjects. Hierarchical regression greatly improves the accuracy of unstable estimates, especially when multiple exposures (in this case, polymorphisms) are under study, and individual cell sizes are small.26 Our first-stage logistic regression model regressed case-status on the individual cytokine polymorphisms. Dichotomous versions of polymorphisms were created comparing carriers to noncarriers, and all models were adjusted for an ever/never version of smoking during pregnancy. Our second-stage linear regression model regressed the SNP β-coefficients on covariates thought to predict their magnitude. We included 2 second-stage covariates that grouped polymorphisms by immunologic pathways. The first covariate assigned polymorphisms in proinflammatory cytokines a “1” and polymorphisms in anti-inflammatory cytokines “0”; the second covariate did the opposite. Consequently, we assumed the target parameters for cytokine polymorphisms within the same immunologic pathway were randomly sampled from a common underlying distribution with an unknown mean.27 Residual variation unexplained by the second-stage covariates was captured in the independent random variable, Δ, which has mean 0 and variance τ.2 We assumed with 95% certainty that the OR for each SNP, after adjusting for the second-stage covariates, would fall within a 10-fold range (τ2 of 0.35). We conducted sensitivity analyses to determine the degree to which our assumptions about residual variation affected our estimates (data not shown). All models were adjusted for smoking during pregnancy. Single locus effects were also estimated in 2-by-2 contingency tables, stratified by race.
We implemented a 3-tiered analysis approach: (1) we applied omnibus likelihood ratio tests to identify regions that may contain one or more disease-predisposing alleles; (2) we reconstructed subject-level haplotypes and obtained effect estimates comparing subjects who carried at least one copy of the index haplotype with all other subjects; (3) we examined SNP associations within strata of race for haplotypes that were strongly associated with the odds of the outcome in a hierarchical framework that adjusted for all polymorphisms, their immunologic pathway, and smoking during pregnancy. We also analyzed SNPs in 2-by-2 contingency tables, stratified by race.
Finally, we examined a potential gene–environment interaction between bacterial vaginosis infection in the 24th-29th gestational week and the presence of a proinflammatory high-risk allele on spontaneous preterm birth. Breslow-Day tests of homogeneity were examined for statistical evidence of deviation from mutliplicativity. The Interaction Contrast Ratio was examined for evidence of deviation from additivity.28,29
Spontaneous Preterm Birth
IL1α and IL1β
There were no differences in the overall IL1β haplotype frequency distribution between cases and control subjects among either whites or African-Americans. Similarly, there were no differences in African-Americans for the IL1α haplotype configuration. (Whites were excluded from haplotype analyses of IL1α due to a modest departure from Hardy-Weinberg equilibrium; Appendices 3 and 5). White women who carried the +3954C/−581C/−1061T haplotype had 1.7 (0.9–3.2) times the risk of spontaneous preterm birth, and subjects who carried the +3954C/−581T/−1061C haplotype had 2.1 (0.9–5.2) times the risk of spontaneous preterm birth. In hierarchical regression, white carriers of the IL1β(−1061)T or IL1β(−581)C alleles were not at substantially increased risk of spontaneous preterm birth (for IL1β(−1061), 1.4 [0.6–3.5]; for IL1β(−581), 1.2 [0.5–3.0]). White carriers of IL1α(+4845)T were at higher risk (1.8; 0.9–3.7; Fig. 2 and Appendix 4). The risk associated with carrying any of the IL1 polymorphisms did not appear elevated in hierarchical regression among African-Americans (Fig. 2 and Appendix 6).
Neither white nor African-American carriers of IL2(−385)G had increased risk of spontaneous preterm birth in hierarchical regression (for white women, 1.0 [0.5–3.0]; for African-American women, 0.7 [0.3–1.6]; Fig. 2 and Appendices 4 and 6).
Both white and African-American carriers of IL6 (−237)C had similarly increased risk of spontaneous preterm birth (for white women, 1.8 [0.8–3.6]; for African-American women, 1.6 [0.7–3.8]) (Fig. 2 and Appendices 4 and 6). The frequency of the IL6 variant allele in African-American women is low (6%; Table 1).
TNF and LTA
Polymorphisms in the TNF and LTA genes were strongly associated with one another (for all pairwise comparisons in African-Americans and whites, D′ ≥ 0.98), and were therefore considered as part of a common haplotype configuration. Among white mothers, haplotype frequency distributions differed between spontaneous preterm cases and control subjects (Table 2). Differences in the overall frequency distribution appeared to be driven by the AGG and GAG haplotypes (for the AGG haplotype, 1.5 [0.8–2.6]; for the GAG haplotype, 0.6 [0.3–1.0]). Carriers of the GAC haplotype were at increased risk of spontaneous preterm birth (1.6; 0.9–2.9), although the haplotype frequency was not substantially different between cases (0.42) and control subjects (0.38). White subjects who carry the TNF(−488)A or LTA(IVS1−82)C variants also had increased risk of spontaneous preterm birth in hierarchical regression (for TNF(−488)A, 2.0 [0.9–4.0]; for LTA(IVS1−82)C, 2.6 [1.3–5.5]; Fig. 2 and Appendix 4). Additionally, subjects who carry the haplotype that contains none of the variant alleles, GAG, had 40% reduced risk of spontaneous preterm birth (0.6; 0.3–1.0).
There was no difference in haplotype frequency distributions between spontaneous preterm cases and control subjects in African-American women (Appendix 5). However, the pattern of single locus effect estimates was similar in direction to whites, but with smaller magnitude (Fig. 2 and Appendix 6).
Gene–Infection Interaction Between Proinflammatory Polymorphisms and Bacterial Vaginosis Infection
In white women, there were no significant multiplicative interactions between inflammatory cytokine polymorphisms and bacterial vaginosis infection; TNF(−488)A and LTA(IVS1+90)G both deviated from additivity, although in an unexpected direction (interaction contrast ratios of −2.8 and −3.7 respectively). In African-American women, each of the joint effects of IL1β(+3954)T, IL6(−237)C, and TNF(−488)A and bacterial vaginosis infection were both greater than multiplicative and greater than additive, yielding interaction contrast ratios ≥3.0 and Breslow-Day P values ≤ 0.10. In particular, carriers of IL6(−237)C with bacterial vaginosis had greater than 4-fold the risk of spontaneous preterm birth compared with women who carried the variant but were not infected (4.4; 1.2–16.4). Similarly strong effects on preterm birth were found among carriers of TNF(−488)A (2.8; 0.9–9.1) and IL1β(+3954)T (2.8; 0.9–8.6). (Complete interaction tables available from the authors by request.)
IL1α and IL1β
There were no differences in IL1β or IL1α haplotype frequency distributions between cases and controls for either white or African-American women (Appendices 7 and 8). White women who carried the IL1β(+3954)C/IL1β (−581)C/IL1β(−1061)T haplotype had 40% reduced risk of SGA (0.6; 0.3–1.0). Both the IL1β(−581)C and IL1β(−1061)T variants, contained in the CCT haplotype, have hierarchical regression estimates below 1.0 (for IL1β(−581)C, 0.7 [0.3–1.7]; for IL1β(−1061)T, 0.6 [0.2–1.3]; Fig. 3 and Appendix 4). There were no notable haplotype or hierarchical SNP effects among African-American women.
White women who carried the IL2(−385)G allele had modestly increased risk of SGA (1.4; 0.8–2.5) (Fig. 3 and Appendix 4). This association was more pronounced in 2-by-2 contingency tables (1.6; 1.0–2.6). There was no association in African-American women (Fig. 3 and Appendix 6).
IL6(−237)C was not associated with SGA in either whites (0.8; 0.4–1.4) or African-American women (1.0; 0.4–2.5; Fig. 3 and Appendices 4 and 6).
TNF and LTA
There were no differences in haplotype frequency distributions in either African-Americans or white women, and no individual haplotypes were strongly associated with SGA (Appendices 7 and 8).
To assess the contribution of genetic variation in proinflammatory cytokines to the etiology of spontaneous preterm birth and SGA, it is critical to study gene pathways because cytokines have pleotropic effects on host response. In this regard, the study of complex conditions, such as spontaneous preterm birth and SGA, should focus on critical sets of genes, both upstream and downstream responders. The temporal alteration of cytokine levels in response to multifactorial challenges (eg, infection and toxins) could prove to be critical in the development of spontaneous preterm birth or SGA. Likewise, small changes in expression over time could be accentuated in the setting of gestation, because pregnancy is characterized by a delicate balance between normal immune function and continual antigen challenge.
We undertook an investigation of 10 polymorphisms in proinflammatory cytokines (IL1α, IL1β, IL2, IL6, TNF, and LTA) and risk of spontaneous preterm birth and SGA in a case–control study, nested within a well-defined prospective cohort of pregnant women. A companion report contains the second part of our investigation, in which we describe the relationship of polymorphisms in anti-inflammatory cytokines with spontaneous preterm birth and SGA.17 The polymorphisms we included were either identified from previous studies or included because their minor allele frequency was greater than 5% and prior in vitro data suggested a functional effect. Because allele and haplotype frequencies, pairwise linkage disequilibrium estimates, and risk of spontaneous preterm birth differ between races, all of our analyses were stratified by maternally self-identified race; however, we cannot exclude the possibility of residual confounding by race within these groupings. This study was limited by a small sample size once stratified by race.
It has been postulated that perturbations of the inflammatory response could contribute to the etiology of spontaneous preterm birth.1,3 Amniotic fluid concentrations of TNF-α and IL-6 have been found to be elevated in association with preterm labor and birth.1,30,31 Inflammatory cytokines TNF-α, IL-1β, and IL-6 stimulate prostaglandin synthesis in the amnion, chorion, and decidual tissue, which contributes to the initiation of labor. Furthermore, TNF-α is known to induce apoptosis, and high levels of TNF-α could damage the placenta either directly, or by activating natural killer cells, lymphokine activated killer cells, and macrophages,32 which in turn weaken fetal membranes. TNF-α also stimulates the production of matrix metalloproteinase enzymes, which degrade collagen,31,33 a major constituent of membranes. However, increased concentrations of inflammatory cytokines are also linked to term deliveries and may be a normal component of the body's preparation for parturition.31,34,35 Some cases of preterm birth could reflect an early activation of these components normally associated with delivery, and some could result from infection and genetic predisposition combined.
Previous studies have examined the relationship between TNF(−488)A and mother's susceptibility to spontaneous preterm birth.6,7,11,12,14 The polymorphisms we included in the TNF region have functional consequences on gene transcription; specifically, the TNF(−488)A/LTA(IVS1+90) G haplotype has been linked to increased production of TNF-α and LTA relative to all other haplotypes.36,37 This haplotype has been associated with autoimmune, allergic, and infectious diseases.38–41 We found that the AGG, GAC, and GAG haplotypes were strongly associated with increased or decreased odds of spontaneous preterm birth in whites. Currently, there are no data indicating that increased TNF or LTA expression can be attributed to the LTA(IVS1−82)C allele. However, carriers of the GAC haplotype or the LTA(IVS1−82)C allele also had increased risk of spontaneous preterm birth, which may point to linkage disequilibrium between LTA(IVS1−82)C and another causal variant. We have accounted for only a small proportion of the variation within TNF and LTA, which may explain the lack of consistency in haplotype results between white and African-American women—the latter possessing less linkage disequilibrium across markers and therefore greater haplotype diversity. Nonetheless whites and African-Americans had similar patterns in hierarchical regression effect estimates.
Both African-American and white carriers of IL6(−237)C had increased risk of spontaneous preterm birth. The hierarchical regression estimates were of approximately the same magnitude and precision even though the IL6(−237)C variant occurs at substantially lower frequency in the African-American subgroup. Circulating IL-6 levels vary inversely with progesterone during a normal menstrual cycle.42 Additionally, a recent clinical trial in a high-risk population found that weekly intramuscular injections of 17 alpha-hydroxyprogesterone caproate reduced risk of preterm birth by approximately one third.43 Not surprisingly, the levels of IL-6 in maternal serum, amniotic fluid, vaginal fluid, and placenta have been reported to increase during normal labor, with even higher levels observed in women undergoing preterm labor.44 IL-6 secretion is regulated by a complex haplotype, including an AnTn site upstream of −237.45 However, in the IL6 promoter, the predominant haplotype including the IL6(−237)G variant has been shown to account for higher IL6 expression. We found that carriers of the IL6(−237)C variant were at increased risk of spontaneous preterm birth, which, counterintuitively, suggests that those subjects who produce less IL-6 (possibly leading to higher levels of progesterone) have higher risk of preterm birth. This finding should be validated in a larger study.
Finally, polymorphisms in IL1α and IL1β were modestly associated with spontaneous preterm birth in white women. Although carriers of IL1α(+4845)T were at increased risk, this polymorphism violated Hardy–Weinberg in our white subgroup. However, IL1α(+4845)T is believed to be functional; the G to T substitution results in an amino acid change from alanine to serine at residue 114. It has been hypothesized that this may affect proteolytic processing of proIL-1α due to its proximity to a protease cleavage site, causing proIL-1α to build up inside the cell and leading to less circulating IL-1α.46
We found preliminary evidence for gene–environment interactions with bacterial vaginosis infection in African-American women. African-American women who carried any proinflammatory SNP and were negative for bacterial vaginosis at 24–29 weeks' gestation had very low risk of spontaneous preterm birth, whereas those who both carried IL1β(+3954)T, IL6(−237)C, and TNF(−488)A and were positive for bacterial vaginosis had substantially increased risk. This increase was greater than the additive and the multiplicative benchmarks. Our results lend support to the hypothesis that inflammation may be a critical factor in the etiology of preterm birth; however, the nature of this relationship is still unclear.
The evidence linking cytokines to SGA has been mixed. Both pro- and anti-inflammatory cytokine-favored environments have been implicated.47–50 We examined the role that genetic variants in proinflammatory cytokines play in the etiology of SGA and found only a borderline association for IL2(−385)G limited to whites. This finding suggests that this pathway might not be critical for SGA development. Although we chose frequent SNPs and haplotypes to cover the most likely candidate SNPs in the genes of interest, it is still possible that additional variants in these genes could be associated with SGA.
The relationship between proinflammatory cytokine SNPs and spontaneous preterm birth is complex. Only the TNF/LTA haplotypes were associated with spontaneous preterm birth in both African-Americans and whites. The relationship between proinflammatory cytokines and spontaneous preterm birth may well be modified by the presence of bacterial vaginosis infection during pregnancy, but more detailed infection data are needed to examine this hypothesis thoroughly.
We gratefully acknowledge the contributions of the PIN study participants and staff, particularly Nancy Dole Runkle and Laura Henderson, and Amy Herring, for her critical review of the manuscript.
1. Dudley DJ. Pre-term labor: an intra-uterine inflammatory response syndrome? J Reprod Immunol
2. Goldenberg RL, Hauth JC, Andrews WW. Intrauterine infection and preterm delivery. N Engl J Med
3. Gomez R, Ghezzi F, Romero R, et al. Premature labor and intra-amniotic infection. Clinical aspects and role of the cytokines in diagnosis and pathophysiology. Clin Perinatol
4. Kniss DA, Iams JD. Regulation of parturition update. Endocrine and paracrine effectors of term and preterm labor. Clin Perinatol
5. Aidoo M, McElroy PD, Kolczak MS, et al. Tumor necrosis factor-alpha promoter variant 2 (TNF2) is associated with pre-term delivery, infant mortality, and malaria morbidity in western Kenya: Asembo Bay Cohort Project IX. Genet Epidemiol
6. Dizon-Townson DS, Kinney S, Lu J, et al. A promoter mutation in the tumor necrosis factor-alpha a possible association with intra-amniotic infection and preterm delivery. Am J Obstet Gynecol
7. Dizon-Townson DS, Major H, Varner M, et al. A promoter mutation that increases transcription of the tumor necrosis factor-alpha gene is not associated with preterm delivery. Am J Obstet Gynecol
8. Ferriman E, Simpson N, Reid J, et al. Fetal carriage of IL-1Beta*2 is not associated with premature membrane rupture and preterm delivery [abstract]. Am J Obstet Gynecol
9. Genc MR, Gerber S, Nesin M, et al. Polymorphism in the interleukin-1 gene complex and spontaneous preterm delivery. Am J Obstet Gynecol
10. Murtha A, Herbert W, Lieff S, et al. Preterm delivery and maternal interleukin-1 polymorphisms. Am J Obstet Gynecol
11. Roberts AK, Monzon-Bordonaba F, Van Deerlin PG, et al. Association of polymorphism within the promoter of the tumor necrosis factor alpha gene with increased risk of preterm premature rupture of the fetal membranes. Am J Obstet Gynecol
12. Macones GA, Parry S, Elkousy M, et al. A polymorphism in the promoter region of TNF and bacterial vaginosis: preliminary evidence of gene-environment interaction in the etiology of spontaneous preterm birth. Am J Obstet Gynecol
13. Moore S, Ide M, Randhawa M, et al. An investigation into the association among preterm birth, cytokine gene polymorphisms and periodontal disease. BJOG
14. Annells MF, Hart PH, Mullighan CG, et al. Interleukins-1, -4, -6, -10, tumor necrosis factor, transforming growth factor-beta, FAS, and mannose-binding protein C gene polymorphisms in Australian women: risk of preterm birth. Am J Obstet Gynecol
15. Risch NJ. Searching for genetic determinants in the new millennium. Nature
16. Johnson GC, Esposito L, Barratt BJ, et al. Haplotype tagging for the identification of common disease genes. Nat Genet
17. Engel SM, Olshan AF, Savitz DA, et al. Risk of small-for-gestational age is associated with common anti-inflammatory cytokine polymorphisms. Epidemiology
18. Savitz DA, Dole N, Williams J, et al. Determinants of participation in an epidemiological study of preterm delivery. Paediatr Perinat Epidemiol
19. Royce RA, Jackson TP, Thorp JM Jr, et al. Race/ethnicity, vaginal flora patterns, and pH during pregnancy. Sex Transm Dis
20. Packer BR, Yeager M, Staats B, et al. SNP500Cancer: a public resource for sequence validation and assay development for genetic variation in candidate genes. Nucleic Acids Res
. 2004;32 Database issue:D528–D532. Available at: http://snp500cancer.nci.nih.gov
. Accessed April 6, 2005.
21. Schneider S, Roessli D, Excoffier L. Arlequin ver. 2. 000: A software for population genetics data analysis. Geneva: Genetics and Biometry Laboratory, University of Geneva, Switzerland; 2000.
22. Slatkin M, Excoffier L. Testing for linkage disequilibrium in genotypic data using the Expectation-Maximization algorithm. Heredity
23. Excoffier L, Slatkin M. Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population. Mol Biol Evol
24. Fallin D, Cohen A, Essioux L, et al. Genetic analysis of case/control data using estimated haplotype frequencies: application to APOE locus variation and Alzheimer's disease. Genome Res
26. Greenland S, Poole C. Empirical-Bayes and semi-Bayes approaches to occupational and environmental hazard surveillance. Arch Environ Health
27. De Roos AJ, Poole C, Teschke K, et al. An application of hierarchical regression in the investigation of multiple paternal occupational exposures and neuroblastoma in offspring. Am J Ind Med
28. Hosmer DW, Lemeshow S. Confidence interval estimation of interaction. Epidemiology
29. Greenland S, Poole C. Invariants and noninvariants in the concept of interdependent effects. Scand J Work Environ Health
30. Laham N, Brennecke SP, Bendtzen K, et al. Tumour necrosis factor alpha during human pregnancy and labour: maternal plasma and amniotic fluid concentrations and release from intrauterine tissues. Eur J Endocrinol
31. Maymon E, Ghezzi F, Edwin SS, et al. The tumor necrosis factor alpha and its soluble receptor profile in term and preterm parturition. Am J Obstet Gynecol
32. Raghupathy R. TH1-type immunity is incompatible with successful pregnancy. Immunology Today
33. Watari M, Watari H, DiSanto ME, et al. Pro-inflammatory cytokines induce expression of matrix-metabolizing enzymes in human cervical smooth muscle cells. Am J Pathol
34. Dealtry GB, O'Farrell MK, Fernandez N. The Th2 cytokine environment of the placenta. Int Arch Allergy Immunol
35. Romero R, Mazor M, Sepulveda W, et al. Tumor necrosis factor in preterm and term labor. Am J Obstet Gynecol
36. Wilson AG, Symons JA, McDowell TL, et al. Effects of a polymorphism in the human tumor necrosis factor alpha promoter on transcriptional activation. Proc Natl Acad Sci USA
37. Bouma G, Crusius JB, Oudkerk Pool M, et al. Secretion of tumour necrosis factor alpha and lymphotoxin alpha in relation to polymorphisms in the TNF genes and HLA-DR alleles. Relevance for inflammatory bowel disease. Scand J Immunol
38. Knight JC, Udalova I, Hill AV, et al. A polymorphism that affects OCT-1 binding to the TNF promoter region is associated with severe malaria. Nat Genet
39. McGuire W, Hill AV, Allsopp CE, et al. Variation in the TNF-alpha promoter region associated with susceptibility to cerebral malaria. Nature
40. Li Kam, Wa TC, Mansur AH, et al. Association between −308 tumour necrosis factor promoter polymorphism and bronchial hyperreactivity in asthma. Clin Exp Allergy
41. Nuntayanuwat S, Dharakul T, Chaowagul W, et al. Polymorphism in the promoter region of tumor necrosis factor-alpha gene is associated with severe meliodosis. Hum Immunol
42. Angstwurm MW, Gartner R, Ziegler-Heitbrock HW. Cyclic plasma IL-6 levels during normal menstrual cycle. Cytokine
43. Meis PJ, Klebanoff M, Thom E, et al. Prevention of recurrent preterm delivery by 17 alpha-hydroxyprogesterone caproate. N Engl J Med
44. Raghupathy R, Makhseed M, Azizieh F, et al. Maternal TH1- and TH2-type reactivity to placental antigens in normal human pregnancy and unexplained recurrent spontaneous abortions. Cell Immunol
45. Terry CF, Loukaci V, Green FR. Cooperative influence of genetic polymorphisms on interleukin 6 transcriptional regulation. J Biol Chem
46. Kawaguchi Y, Tochimoto A, Ichikawa N, et al. Association of IL1A gene polymorphisms with susceptibility to and severity of systemic sclerosis in the Japanese population. Arthritis Rheum
47. Heyborne KD, McGregor JA, Henry G, et al. Interleukin-10 in amniotic fluid at midtrimester: immune activation and suppression in relation to fetal growth. Am J Obstet Gynecol
48. Omu AE, Al-Qattan F, Diejomaoh ME, et al. Differential levels of T helper cytokines in preeclampsia: pregnancy, labor and puerperium. Acta Obstet Gynecol Scand
49. Spong CY, Sherer DM, Ghidini A, et al. Second-trimester amniotic fluid or maternal serum interleukin-10 levels and small for gestational age neonates. Obstet Gynecol
50. Rijhsinghani AG, Thompson K, Tygrette L, et al. Inhibition of Interleukin-10 during pregnancy results in neonatal growth retardation. Am J Reprod Immunol