The second half of the 20th century witnessed a substantial decrease in the perinatal mortality rate in the United States. Although the U.S. stillbirth rate also gradually decreased during this epoch, from 18 per 1,000 births in 1950 to 6.05 per 1,000 births in 2006,1 this decrease has been substantially less in comparison to neonatal mortality and the stillbirth rate remains higher than that of many other developed countries. In fact, the U.S. stillbirth rate is similar to the neonatal death rate (6.51/1,000 births) and affects almost 26,000 neonates per year.1
Smoking and drug abuse during pregnancy are potential modifiable risk factors for stillbirth.2–12 However, the association between smoking and illicit drugs and stillbirth is primarily based on studies relying on self-reporting of smoking and drug abuse. Our objective was to determine the association of smoking and illicit drug use to stillbirth by measurement of metabolites in maternal serum and umbilical cord homogenate in deliveries complicated by stillbirth compared with live births.
PATIENTS AND METHODS
The Stillbirth Collaborative Research Network of the Eunice Kennedy Shriver National Institute of Child Health and Human Development conducted a population-based case–control study of stillbirth (fetal death 20 weeks of gestation or greater) in five a priori–defined geographically diverse regions with screening and enrollment at the time of delivery between March 2006 and September 2008. Details of methods and study design13 and sample size considerations14 have been published previously. Attempts were made to enroll all eligible women whose delivery resulted in one or more stillborn fetuses and a representative sample of eligible women whose delivery resulted in only liveborn neonates supplemented by oversampling of women with live births delivering at less than 32 weeks of gestation and those of African descent delivering at 32 weeks of gestation or greater.13 Approval was obtained from the institutional review board of each clinical site and the data coordinating center. An advisory board reviewed the progress and safety of the study. All participants gave written informed consent.
A stillborn fetus was defined by Apgar scores of 0 at 1 and 5 minutes and no signs of life by direct observation. Deliveries resulting from the intentional termination of a live fetus were excluded. Gestational age was determined by the best clinical estimate using multiple sources including assisted reproduction (if applicable), first day of the last menstrual period, and obstetric ultrasonograms as previously described.15 Stillbirths and live births were classified as small for gestational age (SGA) if the birth weight was less than the tenth percentile for gestational age based on population norms.16
Study components included a comprehensive standardized fetal postmortem examination and uniform placental pathology evaluation performed by a perinatal pathologist.17,18 A standardized maternal interview during the delivery hospitalization and detailed chart abstraction of prenatal office visits, antepartum hospitalizations, and the delivery hospitalization were conducted. Biospecimens collected included maternal blood for serum and DNA, fetal blood from the umbilical cord (when available), placental tissue, and, in cases, fetal tissue. The consent process provided participants the option to decline consent to one or more components of the study: interview, chart abstraction, blood draw, placental examination, autopsy, genetic studies, storage and future use of biospecimens, and future contact for additional research. The consent form discussed planned testing of the afterbirth for legal and illegal drugs, the deidentification of results, and the protections afforded by the Certificate of Confidentiality that had been obtained for the study. No special consent was obtained for cotinine or toxicology testing.
Umbilical cord segments from cases and controls were collected in sterile containers and frozen at −80°C until assay. Cords were homogenized before batch enzyme-linked immunosorbent analyses for amphetamine, methamphetamine, cocaine (benzoylecgonine), pethidine, meperidine, hydrocodone, and tetrahydrocannabinolic acid. All samples were initially tested by enzyme-linked immunosorbent analysis and presumptive positives were confirmed using appropriate mass spectrometric assays using established and validated procedures.19
Maternal blood for serum samples was collected at delivery and centrifuged for 15 minutes at 1,300 g at room temperature at all participating clinical sites. Serum samples were then frozen at −80°C until assay. After completion of the study enrollment, serum aliquots were shipped to the University of Utah Center for Human Toxicology and batch-analyzed for cotinine using solid-phase extraction and liquid chromatography. The personnel performing the assays were blinded to clinical outcomes.
Medical records from all deliveries with positive cord homogenate narcotic results were reviewed for evidence of prescribed narcotic administration for any reason before delivery. Only those with positive cord homogenate testing and medical records with no evidence of narcotic administration before delivery were considered positive for illicit narcotic use.
Nicotine and cotinine metabolism is accelerated in pregnancy20 and the maternal serum cotinine per cigarette ratio is typically less in pregnant compared with nonpregnant women.21 Thus, the threshold for defining exposure may be different in pregnant and nonpregnant women. We addressed this issue by using quartiles, established in our controls, in addition to a 3-ng/mL threshold to assess cotinine exposure.22 A positive serum cotinine less than 3 ng/mL in women who denied smoking was used as a proxy for passive exposure among nonsmokers.22
The delivery, defined as a case if there were any stillbirths delivered and as a control if all live births were delivered, was the unit of analysis. The analyses were weighted for the oversampling of live births and other aspects of the study design as well as for differential consent among the women with stillbirth and among the women with live birth using SUDAAN 11.0 software.23 The construction of the weights has been previously described.13 The weighted samples of live births and stillbirths are intended to approximate random selections of live births and stillbirths in the catchment areas over the enrollment period. Crude and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were calculated from univariate and multivariable logistic regression models, respectively. Predictor variables in the models were treated as categorical. However, for ordered categories on smoking history in the trimester, the neonate was born, and cotinine levels, tests for linear and quadratic trends in the log odds of stillbirth were also conducted using orthogonal contrasts. All tests were performed at a nominal significance level of α=0.05. All single degree-of-freedom tests were two-sided without correction for multiple comparisons.
Adjusted ORs were computed to account for stillbirth risk factors known at pregnancy confirmation (baseline) using a modification to a risk factor score for stillbirth that was developed on the logit scale using the coefficients from a logistic regression model. Variables contributing to the baseline risk factor score were those described previously,14 specifically the following maternal characteristics: age, race or ethnicity, marital status, education, pregnancy history, body mass index, smoking status, alcohol use, illicit drug use, hypertension, diabetes, seizure disorder, blood type, Rh factor, and multiple gestation in current pregnancy as well as paternal age, family income, insurance or method of payment, and clinical site. All variables included in the score were categorical and an “average” of the regression coefficients associated with the categories was used when a variable was missing for an observation. The average was based on the sample-weighted proportion of live births by category. The modification to the risk factor score for this analysis was to exclude coefficients associated with smoking status and illicit drug use.
The relationships among cotinine levels (negative, 50th percentile or less, greater than 50th percentile), tetrahydrocannabinolic acid, and SGA fetus on pregnancy outcome were studied by comparing the stillbirth ORs for one of the factors with and without accounting for another in logistic regression models. A commonly used threshold of 10% reduction (or increase) in the OR was taken as a measure of confounding. In addition, the interactions of high levels of cotinine (greater than 50th percentile) with an SGA fetus and with preeclampsia were studied using logistic regression models with an interaction term and computing stillbirth ORs for high cotinine levels stratified by whether the fetus was SGA and by whether preeclampsia was a condition noted in the chart at delivery.
Enrollment to the Stillbirth Collaborative Research Network study and inclusion in the serum cotinine and toxicology analyses are shown in Figure 1. For 663 stillbirth deliveries (cases), 418 (63%) had a cord segment collected for subsequent toxicology studies and 579 (87%) had maternal serum analyzed for cotinine. More than half (380 [57%]) had both maternal serum and cord segments for analysis. For 1,932 live birth deliveries (controls), 1,050 (54%) had cord segments collected for subsequent toxicology studies and 1,545 (80%) had maternal serum analyzed for cotinine. Approximately half (891 [46%]) had both maternal serum and cord segments for analysis. Cotinine and toxicology testing was done on virtually all women with adequate blood or cord collected. Absence or insufficient sample was the result of the participant declining sample collection, inconvenient timing, administrative error, and in the vast majority of cases for umbilical cord, discarding of the placenta before it could be retrieved for examination.
Table 1 shows characteristics of cases and controls that did, and did not, undergo cotinine testing and toxicology screening. For both groups, those with cotinine testing, toxicology screening, or both were more likely to be non-Hispanic white and less likely to be non-Hispanic black than those without testing. Participants in the case group and those in the control group with both cotinine testing and toxicology screening were more likely to have commercial insurance and deliver at later gestational ages. Also, a disproportionate number of participants in the control group with testing were between 20 and 39 years of age compared with those without testing.
Women who self-reported smoking were more likely than those who did not to be non-Hispanic white, 20–34 years of age, of low education, unmarried, and low income. Those who self-reported drug use were more likely than women who did not to be non-Hispanic white and unmarried (data not shown).
Self-reported smoking and drug use, cotinine levels, and cord homogenate findings in all stillbirth and live birth deliveries are depicted in Table 2. There was an increase in the stillbirth OR with increasing amounts of self-reported smoking in the trimester the neonate was born (linear trend P<.004). Compared with women who never smoked, women who reported smoking one to nine cigarettes per day had a 1.77 OR for stillbirth (95% CI 1.13–2.80); and those smoking 10 or more cigarettes per day had an OR for stillbirth of 2.17 (95% CI 1.25–3.78). Similar results were noted with serum cotinine levels. Compared with women testing negative, those with positive cotinine concentrations 50th percentile or less had an OR of 2.04 (95% CI 1.39–3.01); and those with cotinine levels greater than 50th percentile had an OR of 2.39 (95% CI 1.62–3.52) (linear trend P<.001). Similar results were noted if cotinine concentrations between 0.25 and 2.99 and 3.00+ were used (linear trend P<.001). Women who denied smoking but had elevated cotinine levels had increased odds for stillbirth using either the 3-ng/mL cut point or percentiles (eg, positive cotinine less than 3 ng/mL OR 2.06, 95% CI 1.24–3.41; positive cotinine greater than 3 ng/mL OR 2.61, 95% CI 1.39–4.88).
Women with stillbirth were twice as likely as those with a live birth to report having been addicted to an illicit drug (OR 2.30, 95% CI 1.37–3.86). A positive test for any drug in the cord homogenate was associated with an OR for stillbirth of 1.94 (95% CI 1.16–3.27). The OR was higher in women having a positive toxicology screen who also reported ever using illicit drugs (OR 3.30, 95% CI 1.54–7.03). The most common individual drug, tetrahydrocannabinolic acid, was positive in 3.9% of participants in the case group and 1.7% of participants in the control group (OR for stillbirth 2.34, 95% CI 1.13–4.81). Among women with testing for cotinine and illicit drugs, women who were positive for cotinine and not illicit drugs had an OR of 1.70 (95% CI 1.13–2.56) compared with those who were negative for both; and women who were positive for both had an OR of 3.86 (95% CI 1.61–9.24). However, the ORs for positive for both compared with positive for cotinine only were not significantly different and there was evidence of confounding of the relationship between illicit drugs and stillbirth by cotinine.
Because they were already at higher risk for complications, we anticipated that smoking and illicit drugs would have less influence on pregnancies complicated by multiple gestation, obstetric complications, or fetal aneuploidy. We therefore repeated these analyses in nonanomalous, singleton pregnancies excluding intrapartum stillbirths, as shown in Table 3. The OR for stillbirth in women with positive cotinine levels 50th percentile or less was 1.88 (95% CI 1.19–2.97) and for those with levels greater than the 50th percentile was 2.67 (95% CI 1.75–4.07). Women with any positive toxicology screen had an increased odds of stillbirth of 2.23 (95% CI 1.29–3.88). Positive cord homogenate tetrahydrocannabinolic acid was associated with an increased odds of stillbirth of 2.83 (95% CI 1.34–5.99).
Selected ORs adjusted for prepregnancy risk factors for stillbirth are shown in Table 4. Self-reported smoking and elevated levels of cotinine were associated with stillbirth even after adjustment for other known risk factors. The adjusted OR for stillbirth with positive cotinine levels 50th percentile or less was 2.05 (95% CI 1.33–3.17) and for cotinine levels greater than 50th percentile was 2.56 (95% CI 1.66–3.93). A positive test for drug use also was associated with stillbirth after adjustment. The adjusted results were also significant in the subgroup of nonanomalous, singleton pregnancies excluding intrapartum stillbirths. There were too few cases of positive results to assess adjusted ORs for each individual illicit drug.
Adjusting for whether the fetus was SGA reduced the stillbirth OR for cotinine (50th percentile or less compared with negative and greater than the 50th percentile compared with negative) by greater than 10%. Thus, at least part of the association between smoking and stillbirth is mediated through fetal growth restriction. Furthermore, the interaction between high cotinine levels and fetal SGA was significant (P<.02) and the stillbirth ORs for high cotinine levels among SGA and non-SGA fetuses were 2.43 (95% CI 1.53–3.86) and 0.81 (95% CI 0.36–1.82), respectively. In contrast, there was no significant interaction between high cotinine levels and preeclampsia in association with a stillbirth outcome of pregnancy.
Adjusting for cotinine level reduced the stillbirth OR for tetrahydrocannabinolic acid by greater than 10%, but adjusting for tetrahydrocannabinolic acid did not reduce the stillbirth ORs for cotinine level. Thus, we cannot exclude the possibility that the association between cannabis and stillbirth is partially the result of confounding by tobacco smoke. There was no evidence of confounding of the relationship between tetrahydrocannabinolic acid and stillbirth by SGA fetus.
Among the 1,271 deliveries with both serum cotinine and drug testing, one woman was human immunocompromised virus-positive, seven were positive for hepatitis B, and four were positive for hepatitis C. Only two of these women (both positive for hepatitis C) had either a positive cotinine or drug test. These small numbers preclude further analyses of the relationship between substance abuse and viral infection.
In this population-based study of stillbirth, we noted a twofold increase in stillbirth in women with positive umbilical cord homogenate screening. The most common drug detected was tetrahydrocannabinolic acid, which was significantly associated with stillbirth (OR 2.34, 95% CI 1.13–4.81). The effect was at least partially confounded with the effects of cotinine. Cannabis remains the most commonly used illicit drug in the United States. In 2009, 16.7 million persons reported using marijuana within the previous 30 days, a 2.3 million per month increase from 2007.24 Previous studies of cannabis use in pregnancy have been based on self-report and either showed no association with adverse pregnancy outcomes or were associated with decreased fetal growth.25–28
Although numbers were small, hydrocodone and morphine trended toward an association with an increased odds of stillbirth, which is important given the epidemic of prescription opioid drug abuse.29 Approximately 1 in 20 of the U.S. population aged 12 years or older has used opioid pain relievers nonmedically24 and the potential exists that this could involve substantial numbers of pregnant women.
We also demonstrated a strong association between maternal smoking and stillbirth. Both self-reported smoking and maternal serum cotinine levels were associated with an increased stillbirth risk. Moreover, there was a general dose–response effect, strengthening the biological plausibility of the association. These data are similar to other reports associating self-reported maternal smoking with stillbirth.9–11 Prior studies also have noted a dose-dependent relationship between smoking and stillbirth and have demonstrated ORs in the range of 2.0.10,11,30 In this study, we used cotinine levels to objectively verify and quantitate smoking.
We also identified an increased risk of stillbirth among women exposed to second-hand smoke. We acknowledge that some of these women may have actually smoked but that number is likely small.31–33 Although recent studies have reported a relationship between second-hand smoke and stillbirth,34–36 none used cotinine levels to verify and quantify the degree of exposure.
Our study had several limitations. First, participants who did not have cotinine and toxicology testing differed in race or ethnicity and gestational age from those whom samples were available for testing, which may bias our findings. Second, drug use during pregnancy declines at term, which may have been another source of bias. Third, it is unclear whether exposure occurred before or after the stillbirth. Finally, despite the large number of women with stillbirth, we had a relatively small number of women testing positive for individual drugs. Thus, we lacked a sample size to make definitive conclusions regarding the relationship between some individual drugs and stillbirth and among cannabis use, smoking, and stillbirth.
There were also several strengths of our study. The study was population-based and racially and ethnically diverse. In addition, all participants were evaluated with a thorough standardized protocol that minimized variability in data and sample collection. Our study also included a maternal interview and medical record abstraction to allow for in-depth questions about smoking and drug use. Finally, in addition to self-reported substance abuse, exposure to tobacco and illicit drugs was confirmed by analyses that were blinded to the clinical outcome.
In summary, positive toxicology screen for illicit drugs was associated with a two- to threefold increase in stillbirth risk. Documentation of tetrahydrocannabinolic acid indicating cannabis use increased the odds of stillbirth twofold. Cannabis users often smoke as well, and more research is needed to investigate the interaction of tetrahydrocannabinolic acid and cigarette smoking. In addition, positive cotinine levels and smoking were associated with a twofold to two- to 2.5-fold increase in the risk of stillbirth. Furthermore, even apparent passive smoking exposure was associated with stillbirth. Between 10% and 30% of pregnant women in developed countries continue to smoke during pregnancy.11 Women who quit smoking from their first to second pregnancy have been shown to reduce their risk of stillbirth to the same level as nonsmokers in the second pregnancy.37 In addition, cannabis use remains common during pregnancy with 2% of the women in this study with a positive cord homogenate (among live birth control participants). Smoking and illicit drugs continue to be common and important modifiable risk factors for stillbirth. Because cannabis use may be increasing with increased legalization, the relevance of our study's findings may increase as well. Clinicians should be alert to these risks and should educate women regarding dangers associated with marijuana use and active and passive smoke exposure during pregnancy.
1. MacDorman MF, Kirmeyer S, Wilson EC. Fetal and perinatal mortality, United States, 2006. National Vital Statistics Reports. Vol. 60. Hyattsville (MD): National Center for Health Statistics; 2012.
2. Fretts RC. Etiology and prevention of stillbirth. Am J Obstet Gynecol 2005;193:1923–35.
3. Ananth CV, Liu S, Kinzler WL, Kramer SM. Stillbirths in the United States, 1981–2000: an age, period, and cohort analysis. Am J Public Health 2005;95:2213–7.
4. Ludlow JP, Evans SF, Hulse G. Obstetric and perinatal outcomes in pregnancies associated with illicit substance abuse. Aust N Z J Obstet Gynaecol 2004;44:302–6.
5. Plessinger MA. Prenatal exposure to amphetamines. Obstet Gynecol Clin North Am 1998;25:119–38.
6. Addis A, Moretti ME, Syed FA, Einarson TR, Koren G. Fetal effects of cocaine: an updated meta-analysis. Reprod Toxicol 2001;15:341–69.
7. Bauer CR, Shankaran S, Bada HS, Lester B, Wright LL, Krause-Steinrauf H, et al.. Maternal Lifestyles Study (MLS): effects of substance exposure during pregnancy on acute maternal outcomes. Pediatr Res 1996;39:257A.
8. Fretts R. Stillbirth epidemiology, risk factors, and opportunities for stillbirth prevention. Clin Obstet Gynecol 2010;53:588–96.
9. Cnattingius S, Stephansson O. The epidemiology of stillbirth. Semin Perinatol 2002;26:25–30.
10. Salihu HM, Wilson RE. Epidemiology of prenatal smoking and perinatal outcomes. Early Hum Dev 2007;83:713–20.
11. Wisborg K, Kesmodel U, Henriksen TB, Olsen SF, Secher NJ. Exposure to tobacco smoke in utero and the risk of stillbirth and death in the first year of life. Am J Epidemiol 2001;154:322–7.
12. Kennare R, Heard A, Chan A. Substance use during pregnancy: risk factors and obstetric and perinatal outcomes in South Australia. Aust N Z J Obstet Gynaecol 2005;45:220–5.
13. Parker CB, Hogue CJR, Koch MA, Willinger M, Reddy U, Thorsten VR, et al.; Stillbirth Collaborative Research Network. Stillbirth Collaborative Research Network: design, methods and recruitment experience. Pediatr Perinatal Epidemiol 2011;25:425–35.
14. Stillbirth Collaborative Research Network Writing Group. Association between stillbirth and risk factors known at pregnancy confirmation. JAMA 2011;306:2469–79.
15. Carey JC, Klebanoff MA, Hauth JC, Hillier SL, Thom EA, Ernest JM, et al.. Metronidazole to prevent preterm delivery in pregnant women with asymptomatic bacterial vaginosis. National Institute of Child Health and Human Development Network of Maternal-Fetal Medicine Units. N Engl J Med 2000;342:534–40.
16. Alexander GR, Himes JH, Kaufman RG, Mor J, Kogan M. A United States national reference for fetal growth. Obstet Gynecol 1996;87:163–8.
17. Pinar H, Koch MA, Hawkins H, Heim-Hall J, Abramowsky CR, Thorsten VR, et al.: Stillbirth Collaborative Research Network. The stillbirth collaborative research network postmortem examination protocol. Am J Perinatol 2012;29:187–202.
18. Pinar H, Koch MA, Hawkins H, Heim-Hall J, Shehata B, Thorsten VR, et al.. The Stillbirth Collaborative Research Network (SCRN) placental and umbilical cord examination protocol. Am J Perinatol 2011;28:781–92.
19. Montgomery D, Plate C, Alder SC, Jones M, Jones J, Christensen RD. Testing for fetal exposure to illicit drugs using umbilical cord tissue vs meconium. J Perinatol 2006;26:11–4.
20. Dempsey D, Jacob P 3rd, Benowitz NL. Accelerated metabolism of nicotine and cotinine in pregnant smokers. J Pharmacol Exp Ther 2002;301:594–8.
21. Rebagliato M, Bolumar F, Florey Cdu V, Jarvis MJ, Pérez-Hoyos S, Hernández-Aguado I, et al.. Variations in cotinine levels in smokers during and after pregnancy. Am J Obstet Gynecol 1998;178:568–71.
22. Benowitz NL, Bernert JT, Caraballo RS, Holiday DB, Wang J. Optimal serum cotinine levels for distinguishing cigarette smokers and nonsmokers within different racial/ethnic groups in the United States between 1999 and 2004. Am J Epidemiol 2009;169:236–48.
23. Research Triangle Institute. SUDAAN language manual, volumes 1 and 2, release 11. Research Triangle Park (NC): Research Triangle Institute; 2012.
24. Substance Abuse and Mental Health Services Administration. Results from the 2009 National Survey on Drug Use and Health: volume I. Summary of national findings (Office of Applied Studies, NSDUH Series H-38A, HHS Publication No. SMA 10-4586 Findings). Rockville (MD): NSDUH Series; 2010.
25. Linn S, Schoenbaum SC, Monson RR, Rosner R, Stubblefield PC, Ryan KJ. The association of marijuana use with outcome of pregnancy. Am J Public Health 1983;73:1161–4.
26. Hatch EE, Bracken MB. Effect of marijuana use in pregnancy on fetal growth. Am J Epidemiol 1986;124:986–93.
27. Fergusson DM, Horwood LJ, Northstone K; ALSPAC Study Team. Avon Longitudinal Study of Pregnancy and Childhood. Maternal use of cannabis and pregnancy outcome. BJOG 2002;109:21–7.
28. El Marroun H, Tiemeier H, Steegers EA, Jaddoe VW, Hofman A, Verhulst FC, et al.. Intrauterine cannabis exposure affects fetal growth trajectories: the Generation R Study. J Am Acad Child Adolesc Psychiatry 2009;48:1173–81.
29. Centers for Disease Control and Prevention (CDC). Vital signs: overdoses of prescription opioid pain relievers—United States, 1999–2008. MMWR Morb Mortal Wkly Rep 2011;60:1487–92.
30. Stephansson O, Dickman PW, Johansson A, Cnattingius S. Maternal weight, pregnancy weight gain, and the risk of antepartum stillbirth. Am J Obstet Gynecol 2001;184:463–9.
31. Klebanoff MA, Levine RJ, Clemens JD, DerSimonian R, Wilkins DG. Serum cotinine concentrations and self-reported smoking during pregnancy. Am J Epidemiol 1998;148:259–62.
32. DeLorenze GN, Kharrazi M, Kaufman FL, Eskenazi B, Bernert JT. Exposure to environmental tobacco smoke in pregnant women: the association between self-report and serum cotinine. Environ Res 2002;90:21–32.
33. Yeager DS, Krosnick JA. The validity of self-reported nicotine product use in the 2001-2008 National Health and Nutrition Examination Survey. Med Care 2010;48:1128–32.
34. Crane JM, Keough M, Murphy P, Burrage L, Hutchens D. Effects of environmental tobacco smoke on perinatal outcomes: a retrospective cohort study. BJOG 2011;118:865–71.
35. Subramoney S, d’Espaignet ET, Gupta PC. Higher risk of stillbirth among lower and middle income women who do not use tobacco, but live with smokers. Acta Obstet Gynecol Scand 2010;89:572–7.
36. Leonardi-Bee J, Britton J, Venn A. Secondhand smoke and adverse fetal outcomes in nonsmoking pregnant women: a meta-analysis. Pediatrics 2011;127:734–41.
© 2014 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
37. Högbert L, Cnattingius S. The influence of maternal smoking habits on the risk of subsequent stillbirth: is there a causal relation? BJOG 2007;114:699–704.