Share this article on:

Indicators of Cocaine Exposure and Preterm Birth

Savitz, David A. PhD; Henderson, Laura MA; Dole, Nancy PhD; Herring, Amy ScD; Wilkins, Diana G. PhD; Rollins, Douglas MD, PhD; Thorp, John M. Jr MD

Original Research

OBJECTIVE To identify predictors of cocaine exposure during pregnancy, using hair and urine assays and self-report, and the association with preterm birth.

METHODS A nested case-control study was conducted in a cohort of 2611 black and white women enrolled in prenatal care in central North Carolina. Cocaine exposure was ascertained by self-report (263 cases, 612 controls), urine assays at 24–29 weeks' gestation (226 cases, 564 controls) and postpartum (160 cases, 408 controls), and postpartum hair assays (169 cases, 435 controls). The major metabolite of cocaine, benzoylecgonine, was measured in urine. Cocaine and benzoylecgonine were measured in hair.

RESULTS Cocaine exposure was identified in 2% based on self-report, 5–6% based on urine assays, and 13–15% based on hair assays. Black ethnicity, lower education, and poverty were strongly predictive of positive hair assays. Hair cocaine and benzoylecgonine were not associated with preterm birth, with the possible exception of higher levels of cocaine and benzoylecgonine and birth before 34 weeks' completed gestation. The urine screen at 24–29 weeks' gestation also gave some indication of a possible association (odds ratio 1.7, 95% confidence interval 0.9, 3.5).

CONCLUSION The study corroborates the incompleteness of cocaine exposure assessment by self-report and urine screens relative to hair assays. The strong demographic predictors of exposure suggest where intervention efforts should be targeted. The most sensitive markers of exposure, hair cocaine and benzoylecgonine, are not associated with preterm birth, perhaps because they reflect different patterns of cocaine exposure than the other measures.

Cocaine exposure during pregnancy is common based on cocaine metabolites in hair but is not strongly associated with risk of preterm birth.

Carolina Population Center, Departments of Epidemiology and Biostatistics, School of Public Health, and Department of Obstetrics and Gynecology, School of Medicine, University of North Carolina, Chapel Hill, North Carolina; and Center for Human Toxicology, University of Utah School of Medicine, Salt Lake City, Utah.

Address reprint requests to: David A. Savitz, PhD, University of North Carolina, Department of Epidemiology, CB #7400, Chapel Hill, NC 27599–7400; E-mail:

This study was supported by grants HD28684 and DA09096 from the National Institutes of Health, cooperative agreements S455–16/17, S0807–18/18, and S1099–19/19 through the Association of Schools of Public Health/Centers for Disease Control and Prevention, U64/CCU412273 through the Centers for Disease Control and Prevention, and funds from the Wake Area Health Education Center in Raleigh, North Carolina.

We would like to thank project manager, Jude F. Williams, and the clinic site coordinators, Barbara Eucker and Anne Carter. We greatly appreciate the cooperation of the obstetrics clinics and their staff, particularly Dr. Peter Morris, Ida Dawson, Dr. Cathi Weatherly-Jones, Dr. Juan Granados, Dr. Thad McDonald, and Sara Caviness, and the many obstetric care providers in these clinics who assisted in the collection of specimens.

Received August 1, 2001. Received in revised form October 15, 2001. Accepted November 9, 2001.

The known biologic effects of cocaine make an adverse effect on pregnancy highly plausible.1 Candidate mechanisms include placental vasoconstriction and decreased fetal oxygen, sudden increases in maternal blood pressure, and increased uterine activity.2 Given these pathways, the repeated observation of increased risk of preterm birth and fetal growth restriction associated with cocaine use1,3,4 has been interpreted as causal, though studies finding little or no independent association between cocaine use and preterm birth or fetal growth restriction have also been reported.5–7

There are two principal challenges to isolating an etiologic effect of cocaine use on adverse pregnancy outcome. First, use of cocaine is associated with markers of increased risk for preterm birth, including lower socioeconomic status, being unmarried, and black ethnicity,8,9 as well as with higher levels of alcohol and tobacco use.5 Though less extensively investigated, inadequate nutrition, sexually transmitted infections, and physical abuse may also be correlated with cocaine use. Thus, it is difficult to isolate the effects of cocaine from the conditions that gave rise to its use and the many consequences of heavy use.

Second, accurate information on cocaine use is extremely difficult to obtain. Given that use of cocaine is generally prohibited, self-report is notoriously incomplete. Only a small proportion of women who test positive based on biologic indicators admit to use.5,9 Urine screens for the cocaine metabolite, benzoylecgonine, are quite sensitive but only reflect exposure in the past 1–3 days.10 Analyses of cocaine and benzoylecgonine in maternal and fetal hair and in meconium may integrate information on use over longer periods of time, up to several months,11–13 but there remain some logistic challenges to the application of these technologies.2 Assays of maternal hair are three to four times as sensitive in detecting cocaine exposure as urine tests or self-report.14 However, they have been used to only a very limited extent in larger surveys.8,15

We conducted a prospective cohort study of pregnant women in central North Carolina, and included measures of cocaine exposure based on self-report, urine collected during pregnancy and after delivery, and hair specimens collected at the time of delivery for preterm cases and controls. Combining this unusually extensive array of cocaine indicators with detailed information on potential confounding factors enabled us to extend our understanding of the potential effect of cocaine exposure on preterm birth.

Back to Top | Article Outline


The Pregnancy, Infection, and Nutrition Study was conducted at prenatal care clinics affiliated with University of North Carolina Hospitals and at Wake County Human Services and Wake Area Health Education Center. The protocol was reviewed and approved by University of North Carolina School of Medicine and Wake Medical Center Institutional Review Boards. As described in detail elsewhere,16 we recruited women for a prospective cohort study, with enrollment at 24–29 weeks' gestation. We asked women to provide blood, urine, and genital tract specimens and participate in a telephone interview in the subsequent 2 weeks to collect information on a number of potential determinants of preterm birth, including education and income, use of tobacco, alcohol, and illicit drugs, symptoms of infection during pregnancy, physical exertion, and employment. Poverty index, which incorporates the number of adults and children supported by the household income, was calculated17 and expressed as a percent of the level that defines poverty. Given the clinic locations, most patients resided in Raleigh, Durham, Chapel Hill, Burlington, and surrounding smaller communities in central North Carolina. For all preterm cases and a random sample of the cohort identified at the time of recruitment (regardless of gestational age at delivery), we collected hair and urine specimens after delivery at the hospital.

Recruitment began in August 1995, and the analyses presented here include women whose last menstrual periods were between February 18, 1996, and March 8, 2000. During that period, 4293 women were identified as eligible, having come to the participating clinic before 30 weeks' gestation with a singleton pregnancy, access to a telephone, able to communicate in English, and planning to continue care and deliver at a study hospital. Among those eligible, 2611 (61%) were successfully recruited, based on willingness to provide genital tract specimens, with approximately 30% lost because of patient refusal, 3% because of inability to make contact at the time of the clinic visit, 4% because of physician refusal, and 1% for other reasons. Patterns of participation were analyzed in detail and suggest that those recruited were generally similar to those eligible, particularly with respect to risk for adverse pregnancy outcome.16

Among the 2611 recruited women, 320 were preterm cases, 722 who delivered at term were selected as controls, and 14 were lost to follow-up. Among the cases and controls, 953 (91%) completed the telephone interview, which included information on cocaine use, with a mean gestational age of 26 weeks at the time of interview. The urine collection at the 24–29 week visit was completed by 931 (89%) of the women. Among the 1042 women eligible to participate in the postpartum hair and urine collection, hair specimens were obtained from 200 (63%) preterm cases and 493 (68%) non-cases. Urine was collected at delivery from 188 (59%) preterm cases and 467 (65%) non-cases.

Gestational age was estimated based on an algorithm that combined last menstrual period with ultrasound dating. If both were available and the two agreed within 14 days, the last menstrual period date was used to assign gestational age, whereas if the disparity was over 14 days, ultrasound dating was used. If a last menstrual period date was not available, we used the earliest available ultrasound dating. For this cohort, 77% of women had both last menstrual period and ultrasound dates, 15% ultrasound only, and 8% last menstrual period date only. Most of the ultrasounds were taken before the 20th week of gestation (89%). Where both were available, gestational age was assigned based on last menstrual period date in 87% and based on ultrasound in 13%.

While in the hospital at the time of delivery, preterm delivery cases and controls were asked to provide a hair specimen. A sample of approximately 75 hairs was closely cut from the vertex region of the scalp, placed in an aluminum foil packet with the scalp end identified, and stored at room temperature until shipment to the Center for Human Toxicology at the University of Utah. Immunoassay (STC Technologies, Bethlehem, PA) was initially used as a method of screening hair specimens, with those found to be positive further analyzed by liquid chromatography/mass spectrometry. As the study proceeded, we eliminated this screening step and subjected all specimens to liquid chromatography/mass spectrometry analysis.

Hair specimens were analyzed according to the methods described by Paulsen et al.18 For the assay of cocaine and benzoylecgonine, hair strands from each specimen were individually aligned (root to tip) and segmented to ensure that a consistent hair length was analyzed for each specimen. Specimens (10–20 mg) were digested overnight in 0.1 N HCl, buffered and extracted by solid-phase extraction (Bond Elute Certify columns, Varian Corporation, Harbor City, CA). Atmospheric pressure electrospray ionization analysis of sample extracts was performed using a Hewlett-Packard series 1100 LC-MSD (Hewlett Packard Corporation, Palo Alto, CA). The limits of quantitation of the liquid chromatography/mass spectrometry assay were 0.05 ng/mg (cocaine) and 0.02 ng/mg (benzoylecgonine). The coefficients of variation for intra- and interassay precision were determined to be less than 4% and 7%, respectively, for cocaine and benzoylecgonine in fortified hair.

Urine was collected at both the 24–29 week visit and postpartum. For preterm delivery cases and controls only, the stored urine was shipped to the Center for Human Toxicology at the University of Utah, thawed, and assayed for benzoylecgonine and other illicit drugs, including opiates and amphetamine/methamphetamine, using microplates from STC Technologies for immunoassays. The results indicate the presence of drug metabolites but not the quantity, with a detection limit of 50 ng/mL for benzoylecgonine.

Indices of cocaine use based on self-report, urine assays, and hair assays were analyzed. In the questionnaire, all women who reported that they had used cocaine during the first 6 months of pregnancy were considered positive, and all others were considered negative. The urine screen identified the presence or absence of benzoylecgonine. Hair assays evaluated both cocaine and benzoylecgonine separately and quantitatively. During the early part of the study, when specimens were first screened by immunoassay, we considered women whose hair either screened negative or screened positive and was found to be negative on liquid chromatography/mass spectrometry as negative. Only those who had cocaine or benzoylecgonine detected by liquid chromatography/mass spectrometry were considered positive, both during the period when we were prescreening with immunoassay and after we began assaying all specimens by liquid chromatography/mass spectrometry. In addition to considering the dichotomy of present or absent, we evaluated women with more than 0 to less than 1 ng/mg separately from women with more than 1 ng/mg of cocaine and benzoylecgonine.

We first examined the proportion positive on the various indicators of cocaine exposure, namely the questionnaire, urine at 24–29 weeks' gestation, postpartum urine, and hair cocaine or benzoylecgonine. Concordance was calculated by defining positive as detection of hair cocaine or benzoylecgonine and negative as detection of neither hair cocaine nor benzoylecgonine. Results from other measures were compared with this standard. Hair cocaine in the absence of benzoylecgonine may be a result of environmental contamination, but because environmental exposure is likely to be strongly predictive of exposure and only 16 of 103 women (16%) who were positive for either were positive for cocaine alone, we accepted hair cocaine or benzoylecgonine as a marker of exposure. The questionnaire and urine measures were each compared with the hair indicator. The sensitivity is calculated as the proportion identified as positive among those who have exposure according to the standard, and the specificity is the proportion identified as negative among those who do not have the exposure according to the standard. A summary measure of agreement is the κ coefficient,19 which quantifies the agreement taking into account the level of agreement expected by chance alone.

We also examined social and demographic predictors of cocaine exposure, and used logistic regression to adjust the predictors for one another in analyses of hair cocaine/benzoylecgonine detection. Logistic regression was used to analyze the relation between the measures of cocaine exposure and the risk of preterm birth, comparing preterm cases (less than 37 weeks' gestation versus less than 34 weeks' gestation) and controls. We considered potential confounding by mother's age, race, marital status, education, poverty status, parity, tobacco use, alcohol use, height, prepregnancy body mass index, and bacterial vaginosis. By first conducting a logistic regression analysis with only the potential confounders in the model, we retained only for those factors associated with odds ratios (ORs) of less than 0.8 or greater than 1.2, leading us to include mother's age, race, education, poverty level, parity, tobacco use, height, and body mass index.

Back to Top | Article Outline


To evaluate patterns of cocaine exposure (Table 1), we considered only the randomly selected subcohort of women chosen at the time of enrollment. The number of participants available for evaluation of the different indicators of cocaine exposure varied, with the greatest number from self-report (n = 763) and the fewest for postpartum urine assays (n = 531) depending on questionnaire completion and specimen collection. The proportion identified as positive varied markedly across indicators of exposure, with only 2% identifying themselves as having used cocaine during the first 6 months of pregnancy in the questionnaire, 5–6% positive on urine assays, and 13–15% positive on hair assays (Table 1). Clearly, the measures yield different information as indicated by the κ coefficients (Table 2). Except for the two hair measures, cocaine and benzoylecgonine, none of the κ coefficients exceeded 0.24, indicating very poor agreement. Another perspective is based on considering hair cocaine and benzoylecgonine as the referent (positive defined as positive on either, negative defined as negative on both) and calculating measures of agreement for the other assays. With hair treated as the referent, sensitivity for the other measures is very poor, 9% for the questionnaire and 16% for the urine assays, whereas specificity is excellent, 100% for the questionnaire and 96–98% for the urine assays.

Table 1

Table 1

Table 2

Table 2

Analysis of patterns of cocaine exposure indicates markedly higher prevalence among younger, black, less educated, and poorer women (Table 1). Gradients were clearest for hair cocaine in terms of both absolute and relative differences, but despite small numbers, the same trends were generally seen for self-report and more clearly for urine assays. Because these social and demographic predictors are highly correlated with one another, it was of interest to assess their independent predictive effects. We developed a logistic regression model to predict being positive on hair cocaine or benzoylecgonine, with other indicators of exposure too rare to study in this manner. After adjustment for other factors, predictors of exposure were older maternal age (OR 0.7 for age less than 20, 95% confidence interval [CI] 0.3, 1.5), black ethnicity (OR 6.0, 95% CI 3.1, 11.7), lower education (for 12 and less than 12 years of education, ORs 1.7 [95% CI 0.8, 3.7] and 6.0 [95% CI 2.4, 14.9], respectively, versus more than 12 years of education), and poverty (for less than 100% and 100–250% of poverty, ORs 2.8 [95% CI 0.9, 8.5] and 3.0 [95% CI 1.1, 8.7], respectively, versus more than 250% of poverty). Thus, except for the reversal of the maternal age pattern, each of the other predictors (ie, race, education, and income), was strongly associated with a higher probability of having positive hair assay results, somewhat more for race and education than for income.

Adjusted measures of the relative risk of preterm birth (less than 37 weeks' gestation) for cocaine exposure provided no evidence of a positive association based on hair cocaine, hair benzoylecgonine, postpartum urine, or self-report (Table 3). All ORs were between 0.6 and 1.2, with limited precision, and the unadjusted measures (not shown) were only slightly higher, ranging from 0.8 to 1.4. Only the urine screen at 24–29 weeks' gestation showed some indication of a possible association (OR 1.7, 95% CI 0.9, 3.5), again reduced slightly from the unadjusted value (OR 1.9, 95% CI 1.0, 3.5). Restricting preterm cases to those born before the completion of 34 weeks' gestation exacerbated the problems with imprecision, but there were suggestions of a positive association with high levels of hair cocaine, low and high levels of hair benzoylecgonine, detection of benzoylecgonine in 24–29 week urine, and self-reported use in the questionnaire. The evidence is somewhat stronger for an association with births before 34 weeks' completed gestation as compared with births before 37 weeks' completed gestation.

Table 3

Table 3

Back to Top | Article Outline


The prevalence of cocaine exposure depends to a great extent on the demographic composition of the population and the methods of ascertainment. Our estimate of 2–6% prevalence based on self-report and urine tests is somewhat higher than was found in the Vaginal Infection and Prematurity Study5 and in a statewide survey in California,20 but lower than the prevalence using comparable methods for a poor, urban population seen at Grady Memorial Hospital in Atlanta6 and for patients at Boston City Hospital clinics and Pinellas County, Florida, public health clinics, where the prevalences were 17% and 15%, respectively.21,22 The prevalence of exposure based on hair assays, around 15%, was substantially lower than was found in a clinic population in New York City in the early 1990s.14 Given the social determinants of drug use, variability over time and location in the prevalence of exposure during pregnancy would be expected.

The social and demographic patterns of cocaine exposure in our study show consistency with previous studies, citing increased use based on urine assays among black women, women with lower levels of education, poor women, and those who are older.8,9,23 Behavioral risk factors for adverse pregnancy outcome, including tobacco and alcohol use, poor nutrition, and use of other illegal drugs, are also found more commonly among women who use cocaine.21,24 The magnitude of demographic differences in cocaine exposure in our study was striking. Race, education, and income were associated with adjusted risk ratios for cocaine exposure as high as 3–6, so that black women who had low education and were poor had an estimated prevalence of cocaine exposure of around 52% based on hair assays. The magnitude of the disparity between different measures of cocaine use depends on pattern of use as well as methods of detection. Urine may be more effective at detecting regular use, whereas hair analyses may be capable of detecting less frequent users. The modest detection rates based on self-report and urine and the high prevalence of exposure based on hair assays suggest that either there are many women who are environmentally exposed to cocaine or they are using cocaine intermittently. The latter explanation would be consistent with evidence that most women who used cocaine during pregnancy did so less than twice per month.25

Because of differing ability to detect occasional use, our analyses of hair cocaine and pregnancy outcome may reflect a different exposure than the many studies of cocaine exposure based on self-report or urine detection. In one sense, the hair assays are just doing a “better” job of identifying what is reflected in urine assays and self-report with much more complete identification of exposure. However, it can also be argued that the hair assays reflect sporadic, less intense exposure, and the other methods are more effective in isolating regular, more intense users.15 Our failure to observe strong associations between urine-detected cocaine exposure and preterm birth may well be due in part to imprecision, given that most prior studies found sizable associations.3,4,26–32 However, two of the methodologically strongest, largest studies detected either a very modest association, with relative risks on the order of 1.3,5 or no association.6 It seems that when “obvious” cocaine users (ie, those who are identified through routine clinical care) are compared with other women, the cocaine users show marked increases in risk whereas more systematic, comprehensive screening to identify exposure is less consistent in showing elevated risk of preterm birth.

Application of this technology to evaluate fetal growth15 indicated that the detection of cocaine metabolites in hair was less strongly associated with intrauterine growth restriction than detection of cocaine metabolites in urine, but there was a dose-response gradient between concentration of cocaine in hair and restricted growth. Those findings are generally parallel to our observations on preterm birth: there was some suggestion that positive urine assays at 24–29 weeks' gestation were associated with preterm birth, evidence against hair assays being associated with preterm birth, but some weak indications that higher concentrations of metabolites in hair were predictive of birth before 34 weeks' gestation. If, in fact, more sensitive methods of detecting exposure, whether by systematic urine screens or even more effectively by hair assays, are less strongly associated with adverse pregnancy outcome than selective screening and less sensitive methods, then two explanations are possible. First, there may be a true effect of cocaine exposure, but occasional or sporadic use may not be associated with a measurable increase in risk, whereas intense use (larger quantities, more frequently) is. Alternatively, there may be no causal effect of cocaine on preterm birth, but those women who are most easily detected in clinical settings because of poverty and correlated lifestyle factors, such as sexually transmitted diseases and poor nutrition, suffer increased risks of preterm birth for reasons other than the cocaine use per se.

Despite advantages in the sensitivity of methods for detecting cocaine exposure in our study, there are a number of potentially important limitations. Foremost is the imprecision in our measurement of the association between cocaine and preterm birth. Confidence intervals on the risk ratios for hair cocaine and even more so for urine cocaine are quite wide. Our ability to make an informed assessment and comparison of modestly elevated or reduced risk ratios is limited as a result. Selection bias could arise at the point of recruitment into the cohort, which is unlikely because the outcome is not known, or more plausibly at the time of recruitment to the case-control study of cocaine exposure. Women who had a history of cocaine use may have been less willing to participate, reducing the overall prevalence of cocaine exposure in our study. However, for this to bias the association between cocaine exposure and preterm birth, the losses would have to differ in relation to pregnancy outcome as well, which seems unlikely. A detailed analysis of characteristics of women who enrolled compared with eligible women who did not enroll16 indicates some tendency for women who did not enroll to come from the Wake clinics, be black, and have lower educational level. Most importantly, the groups were quite similar with respect to risk of preterm birth: 12.3% of enrolled women, 12.5% of women who refused, and 17.1% of women who could not be recruited because of logistic problems delivered preterm. We were able to consider a range of potential confounding factors, but undoubtedly there are uncontrolled differences between cocaine users and nonusers. Most such influences would be expected to favor nonusers, such that more complete adjustment would be expected to reduce, not elevate, measures of association between cocaine and preterm birth.

The hair assays for cocaine exposure can yield false-positives, in which women who truly did not use cocaine are identified as users, based on environmental contamination of the hair. The metabolite, benzoylecgonine, is believed to be more reflective of systemic exposure, as compared with cocaine itself, although even that is uncertain because incorporated cocaine can hydrolyze to form benzoylecgonine.33 Hair specimens were not washed before analysis, largely because of the risk of inadvertently removing incorporated cocaine and benzoylecgonine and the uncertain benefit in removing environmental contamination,34 but this does allow for positive assays as a result of environmental contamination.

Another potentially important limitation is the tendency for highly pigmented hair to more readily incorporate certain drugs than lighter colored hair. Although there are basic structural similarities among all hair types regardless of hair color and ethnicity, there do appear to be some differences in the chemical and physical characteristics of ethnic hair types together with considerable intraethnic variation.35 In theory, black women and other women with dark hair would test positive for cocaine and benzoylecgonine more often, given equal doses, than women with lighter hair colors. However, in our data, there was no tendency for brown or black hair to be positive more than other hair colors, adjusting for age, education, and income.

Given the high prevalence of cocaine exposure, clear evidence of physiologic effects, and almost certainly an influence on placental abruption,1,5 concerns with cocaine and pregnancy should and likely will persist. However, for purposes of policy and personal decision making, there is questionable value in attempts to identify the precise threshold, if any, for safety. With the continued advancement of the technology of hair assays, there will be increasingly affordable opportunities to learn more about the timing and amount of cocaine use in pregnancy to determine more subtle patterns for the time windows and dose levels of cocaine that may be associated with pregnancy outcome.

Back to Top | Article Outline


1. Holzman C, Paneth N. Maternal cocaine use during pregnancy and perinatal outcomes. Epidemiol Rev 1994;16:315–34.
2. Rizk B, Atterbury JL, Groome LJ. Reproductive risks of cocaine. Hum Reprod Update 1996;2:43–55.
3. Feldman JG, Howard LM, McCalla S, Salwen M. A cohort study of the impact of perinatal drug use on prematurity in an inner-city population. Am J Public Health 1992;82:726–8.
4. Kliegman RM, Madura D, Kiwi R, Eisenberg I, Yamashita T. Relation of maternal cocaine use to the risks of prematurity and low birth weight. J Pediatr 1994;124:751–6.
5. Shiono PH, Klebanoff MA, Nugent RP, Cotch MF, Wilkins DG, Rollins DE, et al. The impact of cocaine and marijuana use on low birth weight and preterm birth: A multicenter study. Am J Obstet Gynecol 1999;172:19–27.
6. Sprauve ME, Lindsay MK, Herbert S, Graves W. Adverse perinatal outcome in parturients who use crack cocaine. Obstet Gynecol 1997;89:674–8.
7. Eyler FD, Behnke M, Conlon M, Woods NS, Wobie K. Birth outcome from a prospective, matched study of prenatal crack/cocaine use: Interactive and dose effects on health and growth. Pediatrics 1998;101:229–37.
8. Bendersky M, Alessandri S, Gilbert P, Lewis M. Characteristics of pregnant substance abusers in two cities in the Northeast. Am J Drug Alcohol Abuse 1996;22:349–62.
9. Vega WA, Kolody B, Porter P, Noble A. Effects of age on perinatal substance abuse among Whites and African Americans. Am J Drug Alcohol Abuse 1997;23:431–51.
10. Hamilton HE, Wallace JE, Shimek EL, Land P, Harris SC, Christenson JG. Cocaine and benzoylecgonine excretion in humans. J Forensic Sci 1977;22:697–707.
11. Koren G, Graham K, Shear H, Einarson T. Bias against the null hypothesis: The reproductive hazards of cocaine. Lancet 1989;2:1440–2.
12. Callahan CM, Grant TM, Phipps P, Clark G, Novack AH, Streissguth AP, et al. Measurement of gestational cocaine exposure: Sensitivity of infants' hair, meconium, and urine. J Pediatr 1992;120:763–8.
13. Grant T, Brown Z, Callahan C, Barr H, Streissguth AP. Cocaine exposure during pregnancy improving assessment with radioimmunoassay of maternal hair. Obstet Gynecol 1994;83:524–31.
14. Kline J, Ng SKC, Schitini M, Levin B, Susser M. Cocaine use during pregnancy: Sensitive detection by hair assay. Am J Public Health 1997;87:352–8.
15. Kuhn L, Kine J, Ng S, Levin B, Susser M. Cocaine use during pregnancy and intrauterine growth retardation: New insights based on maternal hair tests. Am J Epidemiol 2000;152:112–9.
16. Savitz DA, Dole N, Williams J, Thorp JM, McDonald T, Carter AC, et al. Study design and determinants of participation in an epidemiologic study of preterm delivery. Paediatr Perinat Epidemiol 1999;13:114–25.
17. US Bureau of the Census. Current population reports, poverty in the United States: 1996. Series P60-198. Washington, DC: US Government Printing Office, 1997.
18. Paulsen R, Wilkins D, Slawson M, Shaw K, Rollins D. Effects of four laboratory decontamination procedures on the quantitative determination of cocaine and metabolites into hair. J Analytic Toxicol 2001; 27:490–6.
19. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas 1960;20:37–46.
20. Vega WA, Kolody B, Hwang J, Noble A. Prevalence and magnitude of perinatal substance exposures in California. N Engl J Med 1993;329:850–4.
21. Frank DA, Zuckerman BS, Amaro H, Aboagye K, Bauchner H, Cabral H, et al. Cocaine use during pregnancy: Prevalence and correlates. Pediatrics 1988;82:888–95.
22. Chasnoff IJ, Landress JH, Barrett ME. The prevalence of illicit drug or alcohol use during pregnancy and discrepancies in mandatory reporting in Pinellas County, Florida. N Engl J Med 1990;322:1202–6.
23. Brantley M, Rochat R, Floyd V, Norris D, Franko E, Blake P, et al. Population-based prevalence of perinatal exposure to cocaine — Georgia, 1994. MMWR 1996;45:887–91.
24. Streissguth AP, Grant TM, Barr HM, Brown ZA, Martin JC, Mayock DE, et al. Cocaine and the use of alcohol and other drugs during pregnancy. Am J Obstet Gynecol 1991; 164:1239–43.
25. Faden VB, Graubard BI. The effect of positive and negative health behavior during gestation on pregnancy outcome. J Subst Abuse 1997;9:63–76.
26. Cherukuri R, Minkoff H, Feldman J, Parekh A, Glass L. A cohort study of alkaloid cocaine (“crack”) in pregnancy. Obstet Gynecol 1988;72:147–51.
27. MacGregor SN, Kith LG, Chasnoff IJ, Rosner MA, Chisum GM, Shaw P, et al. Cocaine use during pregnancy: Adverse perinatal outcome. Am J Obstet Gynecol 1987; 157:686–90.
28. Little BB, Snell LM, Klein VR, Gilstrap LC III. Cocaine abuse during pregnancy: Maternal and fetal implications. Obstet Gynecol 1989;73:157–60.
29. Mastrogiannis DS, Decavalas GO, Verman U, Tejani N. Perinatal outcome after recent cocaine usage. Obstet Gynecol 1990;76:8–11.
30. Ney JA, Dooley SL, Keith LG, Chasnoff IJ, Socol ML. The prevalence of substance abuse in patients with suspected preterm labor. Am J Obstet Gynecol 1990;162:1562–7.
31. Handler A, Kistin N, Davis F, Ferre C. Cocaine use during pregnancy: Perinatal outcomes. Am J Epidemiol 1991; 133:818–25.
32. Bateman DA, Ng SKC, Hansen CA, Heagarty MC. The effects of intrauterine cocaine exposure in newborns. Am J Public Health 1993;83:190–3.
33. Nakahara Y, Kikura R. Hair analysis for drugs of abuse. VII. The incorporation rates of cocaine, benzoylecgonine and ecgonine methyl ester into rat hair and hydrolysis of cocaine in rat hair. Arch Toxicol 1994;68:54–9.
34. Welch MJ, Sniegoski LT, Allgood CC, Habram M. Hair analysis for drugs of abuse: Evaluation of analytical methods, environmental issues, and development of reference materials. J Anal Toxicol 1993;17:389–98.
35. Cone E, Josephs T. The potential for bias in hair testing for drugs of abuse. In: Kintz P, ed. Drug testing in hair. Boca Raton, FL: CRC Press, 1996:17–68.
© 2002 The American College of Obstetricians and Gynecologists