Secondary Logo

Journal Logo

Original Article

Environmental Tobacco Smoke and Pregnancy Outcome

Kharrazi, Martin*; DeLorenze, Gerald N.; Kaufman, Farla L.; Eskenazi, Brenda§; Bernert, John T. Jr; Graham, Steve; Pearl, Michelle; Pirkle, James

Author Information
doi: 10.1097/01.ede.0000142137.39619.60
  • Free


Exposure to second-hand tobacco smoke, also known as environmental tobacco smoke (ETS), is widespread. However, levels vary greatly both nationally and internationally as a result of differences in smoking regulations and cultural tolerance for tobacco.1–3 Recommendations to pregnant women to reduce their ETS exposure also vary among medical practitioners. Through education campaigns and smoking restrictions in work and public places over the past 10 years, California has succeeded in reducing smoking and ETS exposure.4 Accurate information about the impact of low-level ETS exposure on human health is needed to help policymakers, health practitioners, and individuals make informed decisions.

Three recent reviews of more than 25 epidemiologic studies of ETS and pregnancy outcomes concluded that ETS leads to diminished birth weight5–7; however, the magnitude and shape of that association is still uncertain because of poor exposure assessment, especially at low levels of exposure. Questionnaire studies have been hindered by incomprehensive questions and the inability of most people to quantify their ETS exposure.8–11 Previous biomarker studies, although offering an objective assessment of exposure, have been unable to accurately quantify exposure at low levels because of the high assay detection limits.12–14 Both types of studies have likely misclassified sizable portions of their study populations as unexposed to ETS and thus have underestimated the effects of ETS on birth weight and other less well-studied pregnancy outcomes. It is still unclear to what extent the relation between ETS and birth weight is the result of shortened gestation. Reviewers recommend that additional research be done to address these problems.5–7

In this investigation, we used a sensitive cotinine assay, capable of measuring exposure to very low ETS levels, in more than 3000 women who enrolled in California's prenatal screening program. Because this study is the largest of its kind, we aimed to quantify the magnitude and describe the shape of the relation of ETS exposure with several measures of gestational duration and with fetal growth and mortality.


Study Population and Data Collection Procedures

The overall population included pregnant women from 11 counties of central California who enrolled in the state's maternal serum alpha-fetoprotein prenatal screening program in April 1992 (n = 3669). Statewide, approximately 60% of women delivering live births enrolled in the program in 1992. Blood was collected at 15 to 19 weeks gestation in serum separator tubes and sent by mail to a laboratory where serum was analyzed for alpha-fetoprotein. The remainder of each specimen was stored for up to 5 years at −20°C before cotinine analysis at the Centers for Disease Control and Prevention (CDC). Written informed consent about possible use of the specimen for research purposes was obtained from participants before prenatal screening. Both the California Health and Human Services Agency and the CDC Committees for the Protection of Human Subjects approved this investigation.

Prenatal screening program data were linked with live birth and fetal death certificate data files by maternal identifiers (first, last, and maiden names; birth date; and last menstrual period date) using probabilistic record linkage software.15 We excluded the following subjects from analyses: 324 who could not be linked to certificate data, 109 for whom a specimen was not found or the quantity was insufficient for laboratory analysis, 40 who were carrying multiple fetuses, 4 with missing or out-of-range (ie, <500 g or >6000 g) birth weights, 19 with unusable gestational ages (described subsequently), and 377 with cotinine levels in the active smoking range (ie, >10 ng/mL16). This left 2777 woman–live birth pairs and 19 woman–fetal death pairs from the 11-county region.

A 4-county subgroup of the population completed a 1-page questionnaire at the time of birth certification, which included maternal height, prepregnancy weight, weight gain, and tobacco exposure information. Spanish and English versions of the questions were provided. For this subgroup, infant's head circumference and length were abstracted from the hospital chart and recorded on the questionnaire by the birth recorder. Further analyses were conducted on this subgroup (n = 1071 for birth weight, n = 842 for infant length and body mass index, n = 705 for head circumference and for the ratio of brain weight to body weight).

Exposure Measure

Cotinine was measured at the CDC using a sensitive isotope-dilution high-performance liquid chromatographic/atmospheric-pressure ionization tandem mass spectrometric procedure.17 All serum samples were coded before analysis so as not to reveal identifiers or identify health or covariate information to laboratory staff. Processing and quality control procedures have been described previously.10 Limits of detection were calculated by the limiting standard deviation method18 and remained below the nominal method detection limit of 0.050 ng/mL.

The cotinine distribution of 10 ng/mL or less was log-normally distributed (Fig. 1). We considered serum cotinine as a categorical variable and as a continuous log10-transformed variable. Approximately one third of the specimens had values below the nominal method detection limit. To preserve the integrity of the log normal distribution (ie, to avoid placing an unnaturally large proportion of subjects at a single value at the lower end of the distribution), actual values below the detectable limit were used for these specimens (n = 1046). In addition, to retain 66 specimens with values of zero in the analyses after log transformation, we assigned the value of the lowest measured specimen (0.001 ng/mL) to these specimens. Assay results below the limit of detection served as the best, nonbiased estimates of exposure available. When the statistical analyses were rerun 1) including only samples above the detection limit or 2) by assigning a value equal to the limit of detection divided by the square root of 2 to specimens below the detectable limit, similar and sometimes stronger results were obtained.

Serum cotinine distribution for an 11-county live birth study population. Cotinine levels less than or equal to 10 ng/mL are likely nonsmokers (n = 2777); more than 10 ng/mL cotinine are likely smokers (n = 373).

Outcome Measures

A number of outcomes were investigated to characterize the range of effects on fetal growth and mortality, and on pregnancy duration. These included measures of gestational age, birth weight, and fetal death (20+ weeks gestation). For the 4-county study population, we also studied infant head circumference and length, as well as infant body proportionality measured by body mass index at birth (BMI = weight [kg]/length [m2]) and by the ratio of brain weight to body weight (100 × [0.037 × head circumference (cm)2.57]/birth weight [g]).19 To summarize the impact of ETS across the latter half of pregnancy, we created an index of 3 adverse pregnancy outcomes combining all fetal deaths with live births that were term-low birth weight or born preterm.

Initially, gestational age was calculated from date of last normal menses listed on the live birth certificate (87% of study population). When this value was missing, incompatible with birth weight (ie, short gestational age and large birth weight), or out-of-range (<140 or >314 days), data from the prenatal screening record was used in the following order of preference: date of last normal menses (6%), ultrasound findings (5%), and clinical examination (2%). Birth weight was taken from the live birth certificate. Other live birth outcome data came from the hospital record, as recorded on the questionnaire.

The overall distributions of gestational age and birth weight were measured by the mean and standard deviation (in days and grams, respectively). The size of the lower tail of the gestational age distribution—an indicator of babies born too early—was measured by the percent preterm (<37 weeks gestation). For birth weight, the lower tail was measured by the percent term-low birth weight (37 or more weeks and <2500 g), an indicator of babies that grew slowly in utero.

Data Analysis

Multiple linear and logistic regression techniques were used to quantify the relations between cotinine, categorized into quintiles, and study outcomes. Similar techniques were used with continuous log cotinine. Nonlinearity across the range of cotinine values in the relations between log cotinine and continuous study outcomes was explored using nonparametric LOESS regression20 in S-Plus and by entering log cotinine into the models as quadratic and cubic terms. Five covariates (categorized as in Table 1) were included in most models: mother's ethnicity, age, parity, source of payment for prenatal care, and infant sex. We derived these covariates from a larger list of 8 covariates; we chose these 5 because they were risk factors for 1 or more of the study outcomes according to the literature and because they consistently had a sizable influence on the cotinine regression coefficients (ie, the value changed by 10% or more when added to unadjusted models). Infant sex did not meet the second condition but was retained in adjusted models nonetheless because it was the covariate most strongly related to birth weight. Three risk factors—marital status, adequacy of prenatal care, and maternal education—were not included in adjusted models because they did not meet the second condition. In adjusted models of mean birth weight, BMI, and brain–body ratio, gestational duration was controlled by inclusion of linear and quadratic gestational age terms; for models of head circumference and infant length, gestational duration was controlled by inclusion of a linear gestational age term only. In the 4-county subgroup, additional covariates were included to test for their explanatory effects: maternal height in meters (continuous), BMI before pregnancy (continuous), and weight gain in kilograms (continuous).

Distribution of Selected Demographic Characteristics and Geometric Mean Cotinine Level (ng/mL) for Demographic Subgroups in the 11-County and 4-County Live Birth Study Populations

To quantify the impact of ETS on fetal growth rates apart from those of shortened gestation, we compared the percentage change in the log cotinine regression coefficients when gestational age terms were in and out of the previously described birth weight models. We used the Wilcox and Russell approach21 to describe and quantify the birth weight distribution for births in 4 cotinine categories. A publicly available program22 was used to determine the parameters around the predominant part of the distribution (defined by the mean and standard deviation) and the residual part of the distribution (defined by the percent of the population lying outside the lower tail of the predominant distribution). Raw frequencies and adjusted frequencies, based on multivariate regression results (calculated by summing the predicted means and residuals for each subject, as described by English and Eskenazi23), were entered into the program. Birth weight and residual distributions were compared across the 4 cotinine categories.

Cumulative fetal death rates by completed weeks of gestation (using a gestational age-specific hazard model)24 were calculated for 4 cotinine categories to determine the timing of these events. The population attributable risk of adverse pregnancy outcome, adjusting for previously described covariates, was calculated for ETS exposure at or above 0.05 ng/mL cotinine using an algorithm adapted for cohort studies.25 All other statistical analyses were performed using SAS procedures (SAS Institute, Cary, NC).


The study population was largely of Mexican origin or non-Hispanic white, and the mothers were generally having their second or higher live birth. The mean maternal age was 25.4 years, and the median years of schooling was 12 (mean = 11.3). More than half used government payment for prenatal care (Table 1).

The distribution of midtrimester cotinine in the live birth study population, both at or below 10 ng/mL (likely nonsmokers) and above 10 ng/mL (likely smokers), is displayed in Figure 1. A bimodal distribution was observed. For purposes of the remaining analyses, women with levels above 10 ng/mL were excluded to focus on the effects of ETS. The geometric mean cotinine concentration of the remaining study population was 0.08 ng/mL (log10 value = −1.11) and ranged from 0.001 to 10.0 ng/mL (range in log values = −3 to 1). As a group, non-Hispanic black women had the highest cotinine levels (Table 1). Other high-exposure groups were teens, women with 9 to 11 years of schooling, and women using government programs for payment of prenatal care. ETS exposure was similar in the 11- and 4-county study populations (Table 1). For context, geometric mean cotinine values for women who reported 0, 1, and 2 or more household smokers in the 4-county study population were 0.06, 0.17, and 0.29 ng/mL, respectively.10

The associations between cotinine levels (in quintile categories) and study outcomes can be found in Tables 2 and 3. A comparison of study outcomes shows elevated odds ratios or lower means at the highest quintile level (0.236–10 ng/mL) compared with the lowest quintile (<0.026 ng/mL), in both unadjusted and adjusted analyses; these outcomes included preterm delivery, term-low birth weight, and the summary adverse pregnancy outcome measure, as well as mean birth weight and infant length. Some of these outcomes (including mean birth weight and infant length) showed evidence of a dose-dependent relation with cotinine.

Odds Ratios of Selected Pregnancy Outcomes by Cotinine Quintiles in Unadjusted and Adjusted Logistic Regression Models, 11-County Study Population
Mean Differences in Gestational Age and Infant Anthropometric Measurements by Cotinine Quintiles in Unadjusted and Adjusted Linear Regression Models, A. 11-County Study Population and B. 4-County Study Population

The associations between continuous log cotinine and study outcomes are presented in Tables 4 and 5. The regression coefficients refer to the proportional change in the odds ratio or in the mean difference with each unit increase in log cotinine, assuming a linear effect across all log cotinine values. Results were either similar to or stronger than those seen in the quintile analysis, but confidence intervals tended to be wide for most of the categorical outcomes. The odds ratio of adverse pregnancy outcomes was 3.4 (95% CI = 1.3–8.6) across the exposure range {total range OR = exp[4*ln(1.36)] = 3.4} after controlling for covariates.

Odds Ratios for Selected Pregnancy Outcomes for Each Unit Increase in Log Cotinine Level in Unadjusted and Adjusted Logistic Regression Models, 11-County Study Population
Proportion of Variance Explained (R 2) and Change (Beta) in Mean Value of Gestational Age and Infant Anthropometric Measurements for Each Unit Increase in Log Cotinine Level in Unadjusted and Adjusted Linear Regression Models, A. 11-County Study Population and B. 4-County Study Population

Inverse linear relations were found between log cotinine and mean birth weight and between log cotinine and infant length (Table 5). Over the total 4-unit range of log cotinine values, birth weight dropped 109 g (4 × −27.2 g) and length shortened by 0.84 cm (4 × −0.21 cm) in adjusted analyses. To determine the shape of the relation between cotinine and birth weight and between cotinine and infant length, quadratic and cubic log cotinine terms were added to the previously described models, but the fit of these models was not improved over that of the linear models (data not shown). In addition, various nonparametric regression analyses were attempted, but none provided evidence of nonlinearity or of a threshold in the relation between log cotinine and birth weight or infant length.

An inverse linear relation was observed for log cotinine and head circumference in unadjusted analyses, but no such association appears after controlling for covariates (Table 5). The addition of higher-order polynomial cotinine terms did not improve the fit of the head circumference models. Log cotinine did not predict the child's body mass index or the ratio of brain weight to body weight in a linear fashion across the range of ETS exposures (Table 5). However, the fit improved when quadratic and cubic log cotinine terms were added to the adjusted model predicting infant's body mass index [R2 increased 0.6% from 9.0–9.6%; partial F test = 3.78 (df 2,823), P value = 0.023; mean BMI = 13.2 − 0.50(log cotinine) − 0.52(log cotinine)2 − 0.12(log cotinine)3]. Infant BMI appeared steady across lower levels of cotinine through approximately 0.5 ng/mL. Above this level, infant body mass index declined in an accelerating fashion through the highest cotinine values. A similar relation was found for ponderal index. The addition of higher-order polynomial cotinine terms did not improve the fit of the model predicting the ratio of brain weight to body weight.

Figure 2 presents the cumulative distribution of deliveries by gestational age in 4 cotinine groups. Women in the highest cotinine groups (>0.1 ng/mL) delivered earlier than those in lower cotinine groups. To assess the role of shortened gestation on mean birth weight, we compared the values of the cotinine regression coefficient in the adjusted model in Table 5 with and without terms for gestational age and gestational age squared. Ten percent of the cotinine–birth weight relation was attributable to shortened gestation ([−30.3 g to −27.2 g]/−30.3 g). A similar result was obtained when the adjusted model included a binary term for preterm birth (in addition to the 2 gestational age terms).

Cumulative delivery rates (%) by completed week of gestation through week 37 in 4 cotinine subgroups up to 10.0 ng/mL, 11-county live births.

The birth weight findings noted here also were observed in the 4-county subgroup (Table 5). Additional covariates were available for 748 subjects to control for possible differences in maternal height and prepregnant BMI. When these covariates were added to birth weight models, the value of the regression coefficient changed little (from −44 g to −51 g). Maternal weight gain during pregnancy was added to the model to examine whether the relation with cotinine was mediated through weight gain. Again, there was little change in the coefficient (from −51 g to −46 g). To determine whether previous or occasional smokers influenced these results, we excluded 64 self-reported smokers or quitters and obtained a similar decrement in birth weight (−44 g).

The birth weight distributions for 4 cotinine groups in the 11-county population are presented in Figure 3. As cotinine levels increased, mean birth weight declined in a dose-dependent fashion, and there was a greater proportion of births in the lower tail of the distribution. Using the unadjusted and modified Wilcox and Russell approach23 to examine the lower tail, we found that birth weight residuals (unadjusted = 8.0%; adjusted = 11.5%) were highest for the highest cotinine group (0.5–10 ng/mL). The 3 lower cotinine groups had similar birth weight residual values (unadjusted = 2.2–3.3%; adjusted = 1.7–2.0%).

Birth weight distributions in 4 cotinine subgroups up to 10.0 ng/mL, 11-county live births. From lowest to highest cotinine levels, mean ± standard deviation birth weights were 3444 ± 531 g (n = 1033), 3424 ± 551 g (n = 654), 3383 ± 525 g (n = 812), and 3333 ± 574 g (n = 278).

An examination of the 19 fetal deaths in our study indicated much higher rates for the highest cotinine group (0.50–10.0 ng/mL) beginning early in pregnancy (ie, weeks 21–23). Week differences were subtler in the lower cotinine groups but consistent with a dose-response effect. As cotinine levels increased, fetal deaths occurred at earlier gestational ages, and over the course of pregnancy, higher cumulative fetal death rates resulted (0.6%, 0.9%, 1.1%, and 1.8% for the 4 levels of cotinine).

As seen earlier, the odds of adverse pregnancy outcome (fetal death, preterm delivery or term-low birth weight) doubled across cotinine quintiles (Table 2) and tripled across the range of log cotinine values (Table 4). The predicted probability of adverse pregnancy outcomes increased from 5% to 12% in a linear fashion between 0.05 ng/mL and 4.0 ng/mL cotinine in models that included linear, quadratic, and cubic log cotinine terms (data not shown). In multivariate attributable risk analyses, ETS levels at or above 0.05 ng/mL (experienced by 62% of the study population) accounted for 12% of all adverse pregnancy outcomes.


ETS exposure in pregnant women is associated with several adverse outcomes across the latter half of pregnancy among mothers who are presumed to be nonsmokers. Evidence of gestational shortening was found at cotinine levels between 0.1 ng/mL and 10.0 ng/mL. We found that the relation of ETS and birth weight was in part (approximately 10%) the result of an increased rate of preterm delivery, but was mainly due to a slowed rate of growth in utero. This slowed rate of growth spared the head and brain and was proportional in terms of length and weight across the lower ETS exposure range. At higher levels of maternal ETS exposure (above approximately 0.5 ng/mL cotinine), babies showed signs of thinning. In addition, at these higher levels of exposure a larger proportion of births was observed in the lower tail of the birth weight distribution, which, according to Wilcox and Russell,22 has greater public health consequence on subsequent morbidity and mortality than a shift in the main part of the distribution toward lower birth weight values alone. These results did not appear to operate through diminished maternal weight gain during pregnancy. There was no threshold level of ETS exposure below which birth weight and infant length was not reduced. Despite small numbers, there was evidence that fetal deaths increased as ETS levels increased, but most dramatically at or above 0.5 ng/mL cotinine.

Cotinine, a primary metabolite of nicotine, is a useful biomarker of exposure to tobacco products. Although there are also dietary sources of nicotine, these contribute little to the serum cotinine levels found in most individuals with concentrations above 0.02 ng/mL.26–28 The half-life of cotinine is approximately 20 hours,29 but may be shorter in pregnant women.30 Nicotine exposure to the fetus has been documented through measurement of elevated nicotine and cotinine levels in hair and meconium in neonates born to pregnant women exposed to ETS.31–33

Sidestream tobacco smoke is a mixture of approximately 4000 chemicals, including nicotine, cadmium, polycyclic aromatic hydrocarbons, and carbon monoxide.4,29,34 The role of nicotine on placental vasoconstriction and function has been previously described,35–37 suggesting mechanisms for effects on gestational duration, fetal growth, and mortality. The role of carbon monoxide in the creation of carboxyhemoglobin and the resulting slowed release of oxygen to fetal tissues may provide another mechanism for slowed fetal growth.37–39 There are likely to be other fetal and maternal toxins in tobacco smoke and mechanisms of action yet to be identified.

Others have posited that socioeconomic status and occupational exposures can possibly explain the health effects observed in ETS studies.40 In this study, we controlled for a number of sociodemographic variables, and in the 4-county subset, we were able to control for additional factors. Although the values of regression coefficients and odds ratios were often reduced after adjustment for these factors, substantial associations remained. Because there is undoubtedly misclassification in the measurement of covariates and because there are other possible explanatory factors outside the realm of this study, residual or uncontrolled confounding cannot be entirely eliminated as an explanation for the study findings. However, we cannot find any factors that are related to cotinine, as well as birth weight and length, in a linear dose-dependent fashion.

The results of this study are generally stronger than in previous studies that estimated ETS exposure from cotinine measurement, in part because of the greater sensitivity of the assay used. We found that adjusting our detection level upwards to those of the 3 largest prior cotinine-based studies12–14 resulted in comparable birth weight findings (±20 g), despite different populations, exposure distributions, time periods, and designs (Table 6). When the actual assay detection limit is used (last column), the current study findings (shown categorically here) are generally stronger than those of the cited studies. For example, a decrement of –101 g is found for the highest cotinine group (1.1–10 ng/mL) when compared with <0.05 ng/mL, but reduces to –61 g when a more heavily exposed comparison group (<2.0 ng/mL) is used. The cited studies included women as “unexposed” who would have had detectable levels in our study. The past practice of including low-level ETS-exposed women in the “unexposed” population would also serve to underestimate the magnitude of effect on pregnancy outcomes of women's total tobacco exposure resulting from active smoking, which includes passive exposure. The commonly quoted 200-g decrement in birth weight attributable to active smoking5,41,42 may be underestimated by approximately 100 g, which is the ETS effect. When smokers are added to our current study population, we see a decrement of 327 g (95% CI = –413 to –240 g) in mean birth weight of infants of smoking women with midtrimester cotinine values of 100 ng/mL or higher (vs. <0.05 ng/mL) in a regression model similar to that in Table 6.

Adjusted Differences in Mean Birth Weight (Grams) From the Current Study, Using Cotinine Categories From 3 Other Studies, Compared With Multivariate Results From Those Studies

The strengths of the current study lie in its large, diverse population (unselected for tobacco exposure) and its use of a highly sensitive, objective ETS exposure measure. We examined the shape of the relation of ETS across a broad exposure range with several study outcomes across the latter half of pregnancy to view the findings in context. Weaknesses include use of a single opportunistic cotinine measurement at midpregnancy, which may not reflect the subject's actual ETS exposure during a yet-to-be-established sensitive time of gestation. The result of this is misclassification of ETS exposure; thus, reported effect sizes are likely to be underestimated.

Although an historically conservative cotinine cutoff of 10 ng/mL was used to exclude smokers from these analyses,43 because of accelerated metabolism of nicotine and cotinine in pregnancy,30 some women who were occasional smokers or who quit smoking before the 15th to 19th week of gestation were nonetheless included in our study. Exclusion of such women and reanalysis in a subgroup with available self-reported smoking information did not alter the results, nor did reanalysis restricting the population to women with less than 2 ng/mL cotinine. The observed linear relations with birth weight and infant length spanned the entire cotinine distribution and were not limited to the upper distribution where these women would likely be found.

The distribution of cotinine in this 1992 California population (geometric mean = 0.08 ng/mL; 95% CI = 0.07–0.08 ng/mL) was much lower than that found nationally with the same assay in 1988–1994 among girls and women aged 4 years or older who were not tobacco-users (mean = 0.20 ng/mL; 95% CI = 0.17–0.22 ng/mL).27 This could reflect tighter restrictions on smoking in work and public places in California but may also be the result of higher nicotine metabolism in pregnant compared with nonpregnant women.30 With cotinine levels dropping nationally by more than 70% during a 10-year period,44 ETS exposure in our 1992 study population is probably reflective of average ETS exposure in the United States a decade later.

Evidence suggests that ETS exposure endangers the fetus. The public health impact of this exposure is not small. If we accept that the associations reported here are causal, more than 10% of adverse pregnancy outcomes (comprising fetal deaths, preterm deliveries, and term-low-birth-weight babies) could be eliminated by reducing levels of ETS exposure in the population.


George Cunningham, Robert Haas, Lynn Goldman, and Richard Kreutzer were Coinvestigators on these projects. Enid Satariano assisted with hospital enrollment and training, preparation of the questionnaire, and initial record linkage and data analysis. Susan Hurley conducted subsequent record linkages. Birth recorders and supervisors in 20 of 21 hospitals in 4 counties (Fresno, Kern, Kings, and Tulare counties) collected questionnaire data. County registrar offices linked and routed questionnaires to state investigators. Charles Chan from the Office of Health Information and Research provided the live birth and fetal death data, and Lynn Palmer from the Genetic Disease Branch provided the prenatal screening data. Betsy Noth provided assistance with maintenance and shipping of serum specimens. Connie Sosnoff, James McGuffey, and Melissa Morrison conducted serum cotinine analyses. Gayle Windham provided useful comments on an earlier draft of this paper. Judy Bolstad assisted with manuscript preparation.


1.Serxner S, Catalano R, Dooley D, et al. Tobacco use: selection, stress, or culture? J Occup Med. 1991;33:1035–1039.
2.Shopland DR, Niemcryk SJ, Marconi KM. Geographic and gender variations in total tobacco use. Am J Public Health. 1992;82:103–106.
3.Giovino GA, Henningfield JE, Tomar SL, et al. Epidemiology of tobacco use and dependence. Epidemiol Rev. 1995;17:48–65.
4.Pierce JP, Gilpin EA, Emery SL, et al. Has the California tobacco control program reduced smoking? JAMA. 1998;280:893–899.
5.National Cancer Institute. Health Effects of Exposure to Environmental Tobacco Smoke: The Report of the California Environmental Protection Agency. Smoking and Tobacco Control Monograph no. 10. Bethesda, MD: US Department of Health and Human Services, National Institutes of Health, National Cancer Institute; 1999. NIH Pub. No. 99–4645.
6.World Health Organization. Tobacco Free Initiative: International Consultation on Environmental Tobacco Smoke (ETS) and Child Health. Consultation Report. Jan 11–14, 1999; Geneva, Switzerland: 1–26. [serial online] Available at:
7.Windham GC, Eaton A, Hopkins B. Evidence for an association between environmental tobacco smoke exposure and birth weight: a meta-analysis and new data. Paediatr Perinatal Epidemiol. 1999;13:35–57.
8.Emmons KM, Abrams DB, Marshall R, et al. An evaluation of the relationship between self-report and biochemical measures of environmental tobacco smoke exposure. Prev Med. 1994;23:35–39.
9.O'Connor TZ, Holfor TR, Leaderer BP, et al. Measurement of exposure to environmental tobacco smoke in pregnancy women. Am J Epidemiol. 1995;142:1315–1321.
10.Kaufman FL, Kharrazi M, Delorenze GN, et al. Estimation of environmental tobacco smoke exposure during pregnancy using a single question on household smokers versus serum cotinine. J Exposure Analysis Environ Epidemiol. 2002;12:1–10.
11.DeLorenze GN, Kharrazi M, Kaufman FL, et al. Exposure to environmental tobacco smoke in pregnant women: the association between self-report and serum cotinine. Environ Research Section A. 2002;90:21–32.
12.Haddow JE, Knight GJ, Palomaki GE, et al. Estimating fetal morbidity and mortality resulting from cigarette smoke exposure by measuring cotinine levels in maternal serum. Prog Clin Biol Res. 1988;281:289–300.
13.Rebagliato M, Florey Cdu V, Bolumar F. Exposure to environmental tobacco smoke in nonsmoking pregnant women in relation to birth weight. Am J Epidemiol. 1995;142:531–537.
14.Eskenazi B, Prehn AW, Christianson RE. Passive and active maternal smoking as measured by serum cotinine: the effect on birthweight. Am J Public Health. 1995;85:395–398.
15.Jaro M. Probabilistic linkage of large public health data files. Stat Med. 1995;14:491–498.
16.Benowitz NL, Kuyt F, Jacob P III, et al. Cotinine disposition and effects. Clin Pharmacol Therap. 1983;34:604–611.
17.Bernert JTJr, Turner WE, Pirkle JL, et al. Development and validation of sensitive method for determination of serum cotinine in smokers and nonsmokers by liquid chromatography/atmospheric pressure ionization tandem mass spectrometry. Clin Chem. 1997;43:2281–2291.
18.Taylor JK. Principles of measurement. In: Quality Assurance of Chemical Measurements. Chelsea, MI: Lewis Publishers; 1987:75–93.
19.McLennan JE, Gilles FH, Neff RK. A model of growth of the human fetal brain. In: Gilles FH, Leviton A, Dooling EC, eds. The Developing Human Brain: Growth and Epidemiologic Neuropathy. Boston: Wright PG; 1983:43–58.
20.Cleveland WS. Robust locally-weighted regression and smoothing scatterplots. J Am Stat Assoc. 1979;74:829–836.
21.Wilcox AJ, Russell IT. Birthweight and perinatal mortality: I. On the frequency distribution of birthweight. Int J Epidemiol. 1983;12:314–318.
22.Wilcox AJ. The Analysis of Birth Weight and Infant Mortality. An Alternative Hypothesis. February 1, 2002. Available at:
23.English PB, Eskenazi B. Reinterpreting the effects of maternal smoking on infant birthweight and perinatal mortality: a multivariate approach to birthweight standardization. Int J Epidemiol. 1992;6:1097–1105.
24.Kramer MS, Liu S, Luo Z, et al. Analysis of perinatal mortality and its components: time for a change? Am J Epidemiol. 2002;156:493–497.
25.Carinci F. ATTRISK. A SAS macro for the computation of attributable risks according to the adjusted parameterization of a logistic regression. Unit of Statistics and Information Systems, Consorzio Mario Negri Sud, Italy. Adapted from: Bruzzi P, Green SB, Byar DP, et al. Estimating the population attributable risk for multiple risk factors using case–control data. Am J Epidemiol. 1985;122:904–914.
26.Repace JL. Dietary nicotine. Won't mislead on passive smoking. BMJ. 1994;308:61–62.
27.Pirkle JL, Flegal KM, Bernert JT, et al. Exposure of the US population to environmental tobacco smoke: the Third National Health and Nutrition Examination Survey, 1988 to 1991. JAMA. 1996;275:1233–1240.
28.Siegmund G, Leitner E, Pfannhauser W. Determination of the nicotine content of various edible nightshades (Solanaceae) and their products and estimation of the associated dietary intake. J Agric Food Chem. 1999;47:3113–3120.
29.Benowitz NL. Cotinine as a biomarker of environmental tobacco smoke exposure. Epidemiol Rev. 1996;18:188–204.
30.Dempsey D, Peyton J, Benowitz NL. Accelerated metabolism of nicotine and cotinine in pregnant smokers. J Pharmacol Exp Therap. 2002;301:594–598.
31.Eliopoulos C, Klein J, Phan MK, et al. Hair concentrations of nicotine and cotinine women and their newborn infants. JAMA. 1994;271:621–623.
32.Ostrea EM, Knapp K, Romero A, et al. Meconium analysis to assess fetal exposure to nicotine by active and passive maternal smoking. J Pediatr. 1994;124:471–476.
33.Jaakkola JJ, Jaakkola N, Zahlsen K. Fetal growth and length of gestation in relation to prenatal exposure to environmental tobacco smoke assessed by hair nicotine concentration. Environ Health Perspect. 2001;109:557–561.
34.Taylor AE, Johnson DC, Kazemi H. Environmental tobacco smoke and cardiovascular disease. Circulation. 1992;86:699–702.
35.Hood RD. An assessment of potential effects of environmental tobacco smoke on prenatal development and reproductive capacity. Environmental Tobacco Smoke. In: Ecobichon DJ, Wu JM, eds. Proceedings of the International Symposium at McGill University, 1989. Lexington, MA: Lexington Books; 1990:241–268.
36.Guerin MR, Jenkins RA, Tomkins BA. The Chemistry of Environmental Tobacco Smoke. Composition and Measurement. Boca Raton, FL: Lewis Publishers; 1992.
37.Lambers DS, Clark KE. The maternal and fetal physiologic effects of nicotine. Semin Perinatol. 1996;20:115–126.
38.Longo LD. The biological effects of carbon monoxide on the pregnant woman, fetus, and newborn infant. Am J Obstet Gynecol. 1977;129:69–103.
39.Garvey DJ, Longo LD. Chronic low level maternal carbon monoxide exposure and fetal growth and development. Biol Reprod. 1978;19:8–14.
40.Sterling TD, Glicksman A, Perry H, et al. An alternative explanation for the apparent elevated relative mortality and morbidity risks associated with exposure to environmental tobacco smoke. J Clin Epidemiol. 1996;49:803–808.
41.US Public Health Service. The Health Consequences of Smoking. A Report of the Surgeon General. US Department of Health, Education and Welfare. DHEW Publication No. (HSM) 1971:71–7513.
42.Abel EL. Smoking during pregnancy: a review of effects on growth and development of offspring. Hum Biol. 1980;52:593–625.
43.Jarvis MJ, Tunstall-Pedoe H, Feyerabend C, et al. Comparison of tests used to distinguish smokers from non smokers. Am J Public Health. 1987;77:1435–1438.
44.Centers for Disease Control and Prevention (CDC). Second National Report on Human Exposure to Environmental Chemicals. NCEH Publication No. 02–0716. January 2003:79–80.
© 2004 Lippincott Williams & Wilkins, Inc.