Objective. Evidence that mothers report higher levels of drinking retrospectively than during pregnancy has led some investigators to suggest that women systematically underreport alcohol antenatally and that alcohol-related deficits may actually reflect heavier prenatal exposure. This study is the first to compare the validity of antenatal and retrospective reports of pregnancy drinking, drug use, and smoking in relation to effects on infant neurobehavioral outcomes.
Methodology. Three hundred fifty-four inner-city mothers were interviewed regarding their alcohol, drug use, and smoking during pregnancy and retrospectively at 13 months’ postpartum. Their infants were assessed at 6.5, 12, and 13 months on a large battery of neurobehavioral assessments.
Results. Although higher levels of alcohol were reported retrospectively, the correlations of prenatal alcohol exposure with infant outcome were as strong or stronger for the antenatal measures and only the antenatal reports predicted poorer cognitive performance on the Bayley Scales and symbolic play, slower processing speed on the Fagan Test of Infant Intelligence and cross-modal transfer, and slower infant reaction time. Women also reported higher levels of cocaine and marijuana but not cigarette smoking retrospectively. Relations between cocaine use and smoking on birth size and gestational age were as strong for either report. No effects were detected in relation to either report of marijuana use during pregnancy.
Conclusions. These findings suggest that antenatal alcohol interviews provide the most valid information and demonstrate the importance of assessing prenatal alcohol use during pregnancy to minimize the risk of failing to detect neurobehavioral deficits. Adverse effects were consistently seen at levels as low as 0.5 oz absolute alcohol/day (the equivalent of 7 drinks per week) based on maternal antenatal report. These data suggest that alcohol-related deficits do not reflect heavier prenatal exposure than that reported during pregnancy and that threshold values derived from antenatal reports are reasonably accurate.
The growth in research on the effects of prenatal exposure to alcohol and drugs on infant and child development has led to an increased concern with how accurately alcohol and drug exposure can be ascertained by maternal report. Even when biological tests are available to assay exposure, they often provide no information regarding quantities or timing of exposure and are most useful in conjunction with a maternal report.1,2 The question as to how best and when to interview the mother regarding prenatal alcohol and drug use, therefore, remains of critical importance.
Ernhart and colleagues3,4 found that mothers reported higher levels of pregnancy drinking retrospectively at 5 years’ postpartum than when interviewed during pregnancy. Moreover, the 5-year recall was found to be as valid as the mother’s contemporaneous report when assessed in relation to craniofacial anomalies associated with drinking during pregnancy. The authors suggested that the antenatal values reflected “underreporting” because of the stigma associated with pregnancy drinking and that alcohol-related deficits may actually reflect heavier prenatal exposure. Other studies have also found that mothers report higher levels of pregnancy drinking retrospectively than when interviewed antenatally.5 In a previous report,6 we found that although higher levels of pregnancy drinking are reported retrospectively, the validity coefficients for the retrospective reports are smaller than for the reports obtained during pregnancy. In that report, we only compared the relation of antenatal and retrospective report to neonatal size and gestational age (GA). To our knowledge, this study is the first to examine the validity of antenatal and retrospective reports of pregnancy drinking in relation to effects on infant neurocognitive outcomes and the first to compare smoking reported antenatally and retrospectively.
The implications of this research are important for several clinical and methodological reasons. Based on previous studies, it is not clear whether memory after the fact or stigma during pregnancy is likely to bias the maternal report more. 1) Where there is a known threshold below which the infant is not affected by alcohol exposure, the antenatal and retrospective reports will usually indicate very different thresholds. If the maternal antenatal report, which is generally lower, is more accurate, then observed deficits are related to lower levels of alcohol than those reported retrospectively. 2) Many studies do not obtain information regarding alcohol or drug exposure until after the infant is born. If these data are less valid, these studies may fail to detect true alcohol- or drug-related effects. 3) If antenatal reporting is more valid, studies designed to assess drug effects may undercontrol for alcohol exposure if only a retrospective report is used and inadvertently attribute alcohol effects to other prenatal drug exposures.
The cohort consists of 354 black mothers interviewed both antenatally and at 13 months’ postpartum regarding their consumption of alcohol, cocaine, and other drugs during pregnancy. The women were recruited at their first visit (M = 23.3 weeks of pregnancy, standard deviation [SD]= 7.9) to Hutzel Hospital, a large, inner-city maternity hospital serving primarily (92%) black women. Each woman was interviewed individually at each prenatal clinic visit using a timeline follow-back approach to determine incidence and amount of drinking on a day-by-day basis during the preceding 2 weeks.7 Recall was linked to specific times of day and activities. Volume was recorded for each type of beverage consumed each day, converted to absolute alcohol (AA) using multipliers proposed by Bowman et al,8 and averaged across the clinic visits to provide a contemporaneous report of “oz AA/d during pregnancy.” Each oz of AA is equivalent to 2 standard drinks. Because most of the mothers did not begin prenatal care before the second trimester, the “during pregnancy” measure primarily reflects drinking during the latter part of pregnancy. At the first visit, the mother was also asked to recall her day-by-day drinking during a typical week around the time she became pregnant, ie, at conception. Using the protocol developed by Sokol, Martier, and Ernhart,7 each gravida was interviewed regarding her alcohol consumption both currently and at conception. Moderate and heavy drinking women were overrepresented in the sample by inviting all women reporting alcohol consumption at conception averaging at least 0.5 oz absolute alcohol per day (AA/d or the equivalent of 1 standard drink per day) to participate as well as a random sample of about 5% low-level drinkers and abstainers. To reduce the risk that alcohol would be confounded with cocaine exposure, a group of heavy cocaine (at least 2 days/wk), low alcohol (<0.5 oz AA/d) users were also recruited, 44 of whom are included in this sample.
Twenty-two mothers (6.2%) were between 15 and 17 years of age, and one-third (35%) were primiparous. Infant exclusionary criteria included birth weight <1500 g, GA <32 weeks, major chromosomal anomalies or neural tube defects, or multiple births. Age of testing was corrected for 14 infants born between 34 and 38 weeks’ gestation, and t tests revealed no significant differences for these infants on any of the test measures (all P values >.25). Informed consent was obtained from all participants.
Alcohol and Drug Consumption Measures
At each prenatal clinic visit (M = 5.4, SD = 3.2), the mother was interviewed regarding her alcohol use during the previous 2 weeks with drinking linked to specific times of day and activities. Day-by-day recall of a “typical week” at the time of conception was obtained at the first clinic visit. The data obtained included number of drinking days, volume consumed, and specific beverages for each drinking day. Quantities were converted to ounces of AA using the following multipliers: liquor, 0.40; beer, 0.04; wine, 0.20; and wine coolers, 0.05.8 The AA data obtained at each clinic visit were averaged across visits to provide a summary measure of drinking during pregnancy.
Detailed drug use data during pregnancy were obtained at each prenatal visit except the first; data on postpartum drug use were obtained at the 13-month visit. Because of wide variability in the dosage and degree of purity of commonly used substances, exposure was summarized in terms of the average number of days per month each of 5 categories of drugs was used: cocaine, marijuana, opiates (primarily heroin, methadone, or codeine), depressants (eg, barbiturates, tranquilizers), and other stimulants. As with alcohol, the antenatal drug data were averaged across prenatal visits. In addition, urine samples obtained from the mothers at the first prenatal visit were screened by enzyme-mediated immunoassay technique for benzoylecgonine, the primary metabolite of cocaine, cannabinoid metabolites, morphine, and methadone. Any sample that tested positive was retested before being designated positive in the medical record. Maternal report data were compared with results from the urine screens. Fifteen women in the cohort (4.2%) denied using cocaine yet tested positive; 34 denied using marijuana (9.6%) and tested positive; and 2 denied using opiates (0.6%) and tested positive. The somewhat higher percentage of denial of marijuana relative to cocaine and opiates may be artifactual because the metabolites of marijuana persist in the urine for a longer period than the other drugs. The urine-positive women who denied use of a drug were omitted from the analyses for that drug. Antenatal drug data were also missing for 45 women (12.7%) who only had 1 prenatal clinic. Those women for whom there was no evidence of a given category of drug use—from hospital medical records, the urine screen, or a 13-month retrospective pregnancy drug use interview—were assigned a value of 0 for that category. The remaining women were not included in the drug analyses.
Psychosocial problems relating to alcohol abuse were assessed at the first prenatal clinic visit using the Michigan Alcoholism Screening Test (MAST).9 Each of 25 questions in the MAST is scored as 0, 1, 2, or 5, yielding a total from 0 to 53, with a score of 5 or more indicating a history of alcohol-related problems. Antenatal MAST data are missing for 1 woman who was inadvertently not given the interview in the prenatal clinic.
Mothers and infants were seen in the hospital after delivery and were transported to a University-based laboratory for assessment at 6.5, 12, and 13 months’ postpartum by a community-based outreach worker whose primary responsibility was locating and transporting subjects, many of whom did not have telephones or moved between visits. After testing at the 13-month visit, the mother was interviewed retrospectively regarding her alcohol and drug use during pregnancy by a member of the postnatal assessment staff who was blind with respect to her antenatal interview. A day-by-day account for a “typical week” during pregnancy was used instead of the 2-week history. Data on postnatal drinking levels and the MAST were also collected at the 13-month follow-up visit using the same procedure. The mother was asked to give a day-by-day account of her alcohol and drug use and smoking for a typical week during the past year.
Because AA/d, and cocaine and marijuana/days per month were positively skewed (skew >3.0), they were normalized by means of log X + 1 transformation. Means and SDs are reported in terms of the original distributions; log transforms are used in correlational and regression analyses. The antenatal and retrospective measures of smoking during pregnancy, each of which had a few outliers, were transformed by recoding values greater than 3 SD above the mean to 1 point greater than the highest observed value, as recommended by Winer.10
All maternal and infant assessments were conducted by a postnatal research team member blind with respect to maternal alcohol and drug use. Infants received a toy donated by Johnson and Johnson, Inc (New Brunswick, NJ), and mothers received a small remuneration and a photo of their infant at each visit. The infant tests and procedures used in the study and their relation to prenatal alcohol and cocaine exposure are described in detail in several publications.11–14
The Fagan Test of Infant Intelligence (FTII)15 was administered at 6.5 and 12 months. The infant, seated on the mother’s lap, is first shown 2 identical photos and then a novel photo paired with the familiar one. The normative response at these ages is to look longer at the novel stimulus. Preference for the novel stimulus indicates the ability to recall the familiar stimulus and discriminate it from the novel one. Infant fixation was recorded on a computer, and preference for novelty was computed by dividing duration of time looking at the novel stimulus by total time looking at the paired familiar and novel stimuli for each of the 10 problems, averaged across the 2 ages to enhance reliability.16–18 In addition, following Colombo et al,16 mean fixation duration was computed for each problem by dividing the total duration looking time by the number of looks and averaging these across the 10 problems at each age and then across the 2 ages. A pattern of short looks is believed to reflect more rapid processing of information and has been found to relate to superior performance on several infant tests and to predict higher childhood IQ.14,16,19–25 FTII data were used only if available for both visits and were omitted for 34 infants, who completed fewer than 3 problems at 6.5 or 12 months, because of maternal interference with the testing procedure, illness during the visit, refusal to look at the photos, or computer failure. Mean FTII interobserver reliability based on 9 assessments was r = 0.95 (range: 0.90–0.98).26
Cross-modal transfer, another infant information processing task with good predictive validity, was assessed at 12 months using the procedure developed by Rose and associates.27,28 The stimuli consisted of 6 pairs of acrylic objects, 1 member of each pair shielded from view and presented to the infant initially for tactile familiarization. Visual fixation to each pair of stimuli was then recorded on the Zenith PC computer (Glenview, IL) by a second examiner. Preference for the novel object was computed separately for each problem by dividing fixation duration to the novel stimulus by total time fixating either stimulus and then averaged across the 6 problems. Mean fixation duration was computed by dividing total fixation duration by number of looks during each 20-s visual test period and averaging across the 6 problems. Cross-modal data were missing for the first 2 infants in the study who were tested before the cross-modal apparatus was available and for 2 attributable to technical failure.
Infant processing speed was evaluated using the Visual Expectancy Paradigm (VExP),29 a measure which for the first time can assess infant reaction time (RT). This test has been shown to have moderate internal consistency, adequate test-retest reliability at 3 and 6.5 months, and moderate predictive validity in relation to 4-year IQ.30–32 Originally designed to be used with 3-month-olds,33 we adapted the VExP for use with 6.5-month-olds.12,24 The infant, seated on the mother’s lap, views a videotape consisting of a repeating sequence of 8 geometric designs, schematic faces, and revolving or erupting designs. The stimuli are displayed for 700 ms to the left or right of the infant’s visual center, separated by an interstimulus interval of 1 s. After a brief baseline of random presentations, 60 stimuli appear in a predictable left-right alternating sequence with the entire series lasting 120.5 seconds. The infant’s eye movements are recorded on videotape, which are subsequently coded to determine speed of response (RT), defined as latency between the onset of the stimulus and the time the infant’s eye began to move toward the stimulus. Responses faster than 200 ms are assumed to have been initiated before the stimulus appeared on the monitor and are, therefore, considered anticipations rather than RTs.29 Percent “fast” (201–300 ms) RTs is tabulated by dividing number of fast postbaseline RTs by total number of postbaseline RTs. Percent “quick” responses is tabulated by combining number of fast postbaseline RTs with the number of anticipations and responses faster than 200 ms and dividing by total number of events fixated. Mean interobserver reliability for RT based on 20 assessments was r = 0.93 (range: 0.63–0.99). This assessment was added to the infant assessment battery after data collection had begun; 82 infants for whom we have both antenatal and retrospective maternal reports were administered the test.
Complexity of play was assessed at 12 months using the procedure developed by Belsky and colleagues.34 Ten minutes of free play were described simultaneously by a trained observer on audiotape. Suggestion and modeling were then used to elicit higher levels of play than those spontaneously exhibited by the infant. The audiotapes were coded on a 14-level complexity of play scale by scorers whose mean interobserver reliability based on 33 assessments was 94.5% (range: 83%–100%). Following Hrncir, Speller, and West,35 spontaneous play level was defined as the mean of the highest level of play exhibited during each 15-s period. Elicited play level was defined as the highest level elicited by the examiner.34 The audiotapes were also coded using a tempo of play measure, sustained directed activity (SDA)36 by 2 coders with a mean interobserver reliability based on 15 subjects of r = 0.99. SDA is measured in terms of the amount of time spent in visually directed manipulation of a toy or toy group and was defined here as the mean of the 2 longest periods of directed attention. Spontaneous play and SDA data were missing for 84 infants seen before the play assessment procedure was instituted, who refused to play or fell asleep, and for whom equipment failed or the tapes were accidentally erased. Spontaneous play was missing for three infants whose tapes were erased after being scored for SDA. Elicited play was missing for 78 infants tested before the elicited play procedure was instituted or who refused, were too fussy, or fell asleep.
The Bayley Scales of Infant Development,37 which are among the most frequently used standardized test of infant function, consist of the Mental Development Index (MDI) and Psychomotor Development Index (PDI). The Bayley Scales were administered at 13 months to all but 9 infants who refused to cooperate on the MDI and/or PDI. Mean interobserver reliability based on 36 assessments was 95% (range: 76%–100%).
In addition to demographic and alcohol, drug, and smoking information, data were collected from the mothers on a broad range of background characteristics known to influence child cognitive and neurobehavioral outcome. The Peabody Picture Vocabulary Test-Revised (PPVT-R), administered at the 6.5-month visit, is a brief vocabulary recognition test, which is strongly correlated with more comprehensive measures of adult IQ, such as the Wechsler Adult Intelligence Scale, median r = 0.72, and is highly stable over a 1-year period, median r = 0.72.38 Quality of parenting was evaluated at 12 months on the Home Observation for Measurement of the Environment (HOME),39 which combines a semi-structured maternal interview with observation of mother-infant interaction. Its validity and sensitivity as an indicator of socioenvironmental influences on development have been demonstrated by evidence that it explains significant variance in childhood cognitive functioning over and above that attributable to socioeconomic status (SES) and maternal education.40,41 Its validity for use with black samples has also been established.42 Because of safety considerations, the HOME was administered in the laboratory. Barnard, Bee, and Hammond43 have found that the predictive validity of a laboratory-administered 24-month HOME is as good as that reported by others using in-home assessments. Moreover, the correlations of Bayley MDI with our 12-month HOME was midway between those reported by Siegel44 and Barnard et al,43 both of whom used in-home administration at 1 year. Mean interobserver reliability in the present study was 95% (range: 88%–100%). The HOME was missing for 5 infants, who missed the 12-month visit and for whom the interview was not made up at the subsequent visit.
Maternal depression was assessed at 6.5 and 12 months on the Beck Depression Inventory (BDI),45,46 a 21-item self-report measure, which has strong internal consistency reliability, r = 0.93,47 high test-retest reliability over an 18-month period, r = 0.68,48 and is highly correlated with in-depth clinical assessments of depression, median r = 0.66.46 The average of the 2 BDI scores was used except in those cases where only one score was obtained. The BDI was missing for 10 women, whose infants were not accompanied to the laboratory by the mother or who did not complete the BDI.
After describing the sample characteristics, 4 sets of analyses are presented. We first compared the levels of alcohol and drug use reported antenatally and retrospectively and examined the correlations between the reports obtained at these 2 time points. We then grouped the consumption levels and examined the number of women whose consumption reports were consistent versus those which differed by level of consumption. Next, we compared the validity of the antenatal and retrospective measures in relation first to maternal characteristics and then to infant growth and neurobehavioral outcomes. Finally, we compared the relation of these reports to postpartum maternal drinking and examined other factors that might explain differences in the levels reported at the time periods.
The cohort was predominantly lower class, with a large majority receiving welfare (Table 1). At recruitment, only 10% of the mothers were married. They were poorly educated; 24.3% had not completed high school, 34.2% had graduated from high school, and only 1.1% had completed college. Although the mean PPVT-R scores were very low, correlational analyses indicated that this test was valid for this sample in relation to maternal education, ego maturity, depression, and the HOME.49 The mean for the total HOME scores was similar to other lower SES samples assessed at 12 months in North Carolina, Texas, and Arkansas.50 Although the majority (80.2%) of the women were classified as minimally (scores of 0–9) or mildly (10–16) depressed, 17.5% were moderately depressed (17–29), and 2.0% severely depressed (30 or higher), proportions comparable to the incidence of depression in the control group of another low SES sample.51
There were 204 males (57.6%) and 150 females (42.4%; Table 2). Only 3 (0.8%) newborns weighed <1800 g and 41 (12.7%) <2500 g. Performance on the FTII was somewhat lower than that in Fagan and Singer’s normative sample15 and in a middle-class, low-risk sample18 and similar to that reported for at-risk infants52 and other lower social class full terms.17 Median RTs were about 150 ms faster than those reported by Haith and McCarty32 at 3.0 months, and percentage of fast responses was 3 times higher.24 Mean spontaneous play level was between “functional relational” and “enactive naming,” indicating that much of the infant’s play involved bringing together toys that belonged together, with only some very simple pretend behavior. The infants demonstrated more complex play in response to elicitation by the examiner, the mean response consisting of imitation of a series of single pretense acts, such as drinking from the cup and giving the doll a drink from the cup. The means and ranges for spontaneous play and SDA were very similar to those reported for comparable measures for a middle class sample at 1 year.53 Bayley scores were similar to those of other 1-year-old infants from lower and middle class samples.50,51
Alcohol and Drug Use During Pregnancy
The main drugs used by the women were alcohol, cocaine, and marijuana. Use of alcohol and cocaine were weakly related (r ranged from 0.14–0.27; P values <.001), according to both the antenatal and retrospective reports, as were the correlations between cocaine and marijuana (r ranged from 0.12–0.27; P values <.001). The antenatal reports of alcohol and marijuana were unrelated (r = 0.01); but there was a weak relation between the retrospective alcohol and marijuana reports (r = 0.13; P < .01). Virtually the same correlations were found between alcohol and smoking and between cocaine and smoking for both the antenatal and retrospective reports (r ranged from 0.28–0.34; P values <.001).
As expected, AA/d at conception and during pregnancy were strongly related (r = 0.77; P < .0001). Women reported drinking an average of 0.9 oz of AA/d (SD = 1.9; range = 0.01–24.8), the equivalent of almost 2 standard drinks per day at conception, but these levels declined to 0.2 oz AA/d (SD = 0.5; range = 0.001–6.5) or about a half of a standard drink per day during the second and third trimesters, a decline similar to that found in a sample of white, predominantly middle class women.52
As shown in Table 3, women reported higher levels of drinking during pregnancy when interviewed retrospectively at 13 months’ postpartum. They also reported more frequent use of cocaine and marijuana retrospectively. By contrast, women reported smoking more cigarettes during pregnancy when interviewed antenatally. Despite the significant differences between the levels reported antenatally and retrospectively, the correlations between the 2 reports were all moderate to strong. As expected, MAST scores, which represent a lifelong measure of drinking problems, were virtually unchanged for the antenatal and retrospective reports, and the test-retest correlations were strong.
Table 4 presents the antenatal and retrospective AA/d data organized into 6 categories of alcohol use, ranging from abstainers to very heavy users. In this sample, 36.4% of the women were consistent for the antenatal and retrospective reports. Among those who were inconsistent, 139 (61.8%) reported higher levels retrospectively, whereas only 86 (38.2%) report higher levels antenatally. A large proportion of the women (37.1%) who reported drinking very low levels (<0.1 oz AA/d) antenatally appeared to have forgotten this drinking when interviewed retrospectively. By contrast, 54.0% of the women who reported at least 0.1 oz antenatally report higher levels retrospectively, a pattern that was particularly pronounced among the heavier drinking women. During pregnancy, only 2 women (0.6%) reporteddrinking 3.0 or more oz AA/d, and only 16 (4.5%) reported 1.0 oz or more, the usual criterion for “risk drinking.” When interviewed retrospectively, 27 women (7.6%) reported drinking 3.0 oz AA/d or more, and 80 (22.6%), at least 1.0 oz AA/d. Only 10 of the women (2.8%) reporting <0.1 oz antenatally appeared retrospectively to have been drinking at risk levels (>1.0 oz AA/d), as contrasted to 70 women (19.8%) of those who reported at least 0.1 oz AA/d antenatally. Women reporting AA/d >0.5 reported an average of 0.9 oz retrospectively, compared with 0.3 oz for the AA/d <0.5 oz group, t (353) = −5.70, P < .001.
Cocaine and marijuana data are similarly arrayed in Table 5. A higher proportion of the women (86.3% for cocaine and 75.3% for marijuana) were consistent across visits, but in both instances most of the consistent women were abstainers (89.7% and 90.0%, respectively). Among those who were inconsistent, 27 (62.8%) of the cocaine users and 47 (68.1%) of the marijuana users reported higher drug use levels retrospectively than during pregnancy. By contrast to alcohol use, most women (62.9%) reported the same level of smoking antenatally and retrospectively (Table 6). Of those who were inconsistent, only 23 (26.7%) reported smoking more retrospectively, compared with 63 (73.3%) who report smoking more antenatally, the opposite pattern shown for the other substances.
Validity of the Antenatal and Retrospective Measures
Each of the maternal antenatal and retrospective reports of alcohol, cocaine, and marijuana use and smoking during pregnancy was examined in relation to the MAST and other maternal background characteristics. Antenatal and retrospective reports were unrelated to maternal verbal IQ on the PPVT-R or quality of parenting on the HOME, which are, therefore, not included in the data presented in Table 7. Except for maternal depression, the correlations with alcohol and cocaine were similar in magnitude for the antenatal and retrospective reports, and those that differed were generally stronger for the antenatal measure. Maternal depression was more highly correlated with the retrospective reports, possibly because it was assessed closely in time to the 13-month visit and may have influenced the 13-month recall. For smoking, the correlations with the retrospective report tend to be stronger.
As shown in Table 8, all the correlations of prenatal alcohol exposure with infant outcome were as strong or stronger for the antenatal measures. In fact, almost none of the alcohol-related neurobehavioral deficits found during infancy, including poorer performance on the Bayley MDI, level of play, processing speed on the FTII, and reaction time on the VExP, would have been detected using the maternal retrospective report of alcohol use.
The relation of cocaine with the birth size measures and GA were detected using both maternal reports; the associations of birth weight and head circumference with the retrospective report were somewhat stronger, r = −0.18, P = .001 and −0.13, P < .05, respectively. No neurobehavioral effects were detected for either the antenatal or retrospective measures using continuous measures of prenatal cocaine exposure. In light of our previous findings that infant neurobehavioral deficits are associated only with heavy cocaine use during pregnancy,13 we compared dichotomous measures of heavy cocaine use reported antenatally and retrospectively. The antenatal and retrospective reports were equally sensitive to recognition memory on the FTII at 12 months, but only the antenatal report was sensitive to VExP reaction time measures. Virtually no effects were detected for either the antenatal or retrospective reports of marijuana use during pregnancy. The relation of smoking to birth size and GA was virtually the same for both reports. Whereas poorer performance on the Bayley PDI was more strongly related to the retrospective report on smoking, r = −0.19, P < .01, the correlations of smoking with processing speed on both the FTII, r = 0.14, P < .05, and cross-modal test, r = 0.11, P < .05, were detected antenatally and would not have been detected by the retrospective smoking report.
As has been found in other studies, women reported reducing their drinking during pregnancy as compared with the levels they consumed at time of conception and resumed drinking at higher levels during the postpartum period (Fig 1). Whereas the levels of drinking reported at conception were significantly higher than those reported during pregnancy, t (353) = −8.2, P < .001, there was no difference between the alcohol reported at conception and retrospectively for the pregnancy period, t (353) = −0.03. The alcohol consumption levels reported for the postpartum period were markedly higher than those reported antenatally, t (353) = −9.7, P < .001, but not significantly higher than those reported retrospectively, t (353) = −1.8, P < .10.
By contrast, depending on which report was used, a comparison with postpartum smoking (M = 7.5, SD = 9.2) indicated that women did not cut down their smoking as compared with their smoking reported antenatally (M = 8.5, SD = 10.4; t (231) = 1.79, P < .10) or that they thought they had cut down slightly (M = 5.9, SD = 8.5; t (231) = −4.08, P < .001), when they provided the retrospective report.
Single Visit Versus Aggregate Data
It could be argued that the greater validity of the antenatal reports may be attributable to their being an aggregate of an average of 5 maternal visits across pregnancy (M = 5.4); whereas the retrospective report is comprised of a single report of a typical week during pregnancy. The relative validity of the alcohol levels reported at the first prenatal visit (M = 23.3 weeks, SD = 7.9) was, therefore, compared with the antenatal and retrospective reports (Fig 1). The single antenatal report was not as strong a predictor as the composite antenatal alcohol measure of birth size, GA, and the Bayley MDI. However, by contrast to the retrospective report, the single antenatal visit was as strongly predictive as the aggregate antenatal measure for most of the neurobehavioral endpoints.
Other Determinants of Increase in Level Reported Retrospectively
Multiple regression analyses were conducted to examine the maternal characteristics that predicted greater increases in retrospective reporting relative to the antenatal reports. The principal predictors of reporting drinking more alcohol retrospectively are maternal age, depression, and MAST score (Table 9). Among women reporting <0.1 oz AA/d antenatally who reported >1.0 oz AA/d retrospectively, 54.5% were moderately to severely depressed. Higher retrospective reports of cocaine and marijuana use were also related to maternal depression.
Although mothers report higher levels of pregnancy drinking when interviewed retrospectively, our findings suggest that the antenatal data are more accurate and that interviews conducted during pregnancy provide the most valid information. The correlations of prenatal alcohol exposure with infant outcome were as strong or stronger for the antenatal measures and only the antenatal reports predicted poorer cognitive performance on the Bayley Scales and symbolic play, slower processing speed on the Fagan Test of Infant Intelligence and cross-modal transfer, and slower infant reaction time.
These findings demonstrate the importance of ascertaining prenatal alcohol exposure during pregnancy both to minimize the risk of failing to detect neurobehavioral effects and to ensure that the effects of exposure to alcohol are adequately controlled in studies examining the impact of other drugs, such as, cocaine on infant and child outcome. Given this evidence that the levels of alcohol consumption reported antenatally are more accurate, these data suggest that threshold values derived from antenatal reports are also the most valid. In this data set, adverse effects are consistently seen at levels as low as 0.5 oz AA/d (the equivalent of 7 drinks per week) based on maternal antenatal report.12,14,54 Although the thresholds in our analyses are characterized in terms of oz of AA/d, only 1 of the 480 mothers in the original Detroit cohort actually drank every day and only 3 drank >4 days/wk.55 A mean of 0.5 oz AA/d exposure, therefore, typically represents higher doses of alcohol on those days on which drinking occurs. For the mothers who drank at least 0.5 oz AA/d, the median drinking pattern was 3.0 oz AA/drinking day or 6 standard drinks on 2.3 days of the week.
Ernhart et al3 suggested that higher levels of alcohol consumption are reported retrospectively because the stigma associated with drinking during pregnancy prevents many women from admitting that they are drinking while pregnant. By contrast, the evidence reported here indicating that the antenatal reports are more valid in relation to infant outcome suggests that most women actually reduce their alcohol intake during pregnancy and that adverse effects on development are associated with these reduced levels of exposure. The high levels of alcohol reported retrospectively are more similar in magnitude to the levels reported for the postpartum period. These data, therefore, suggest that, when interviewed retrospectively about a typical week during pregnancy, many of these mothers do not recognize or have forgotten the degree to which they reduced their alcohol intake during pregnancy. Thus, their postpartum drinking patterns seem to influence their retrospective recall.
Because prenatal and postnatal drinking are moderately related, distinguishing between the effects of prenatal exposure and the postnatal environment is often a major methodological challenge.56 Given that alcohol data collected retrospectively seem to be influenced by postpartum drinking levels, it is more difficult to distinguish between prenatal and postnatal effects in studies that collect data on pregnancy alcohol use retrospectively and to attribute observed deficits in the infant or child to prenatal exposure rather than postnatal environmental effects.
In the comparison of the alcohol data from a single antenatal visit, the average antenatal report, and the retrospective typical week, the measure from a single antenatal visit was not as strong a predictor of birth size, GA, and performance on the Bayley MDI. Researchers have long known that timing of exposure can be extremely important in determining the impact on a specific domain or organ system. This finding, therefore, could be attributable to the fact that for many mothers the single antenatal visit occurred relatively early in pregnancy, whereas birth size and GA, and Bayley performance may be related more strongly to later pregnancy drinking.14,57,58 It is impressive that data from a single visit during pregnancy were as strongly predictive of the neurobehavioral deficits as the aggregate antenatal data and detected deficits that would otherwise have been missed by the retrospective report of a typical week. Moreover, given that no difference was found between the levels of alcohol reported for a typical week at conception and the retrospective report of a typical week (Fig 1), it is possible that in a retrospective interview women may generalize to their more typical drinking pattern and underestimate the changes and fluctuations in their drinking across pregnancy, unless asked to recall specific trimesters or periods during pregnancy. These data further suggest that for research purposes the approach developed by Sokol et al,7 using a day-by-day recall and averaging these reports across prenatal visits, generates the most valid information regarding the mother’s drinking during pregnancy.
We have previously reported that when narrow band assessments are administered, the relative dissociability of different components of human cognition and the effects of different substances on these components become more apparent55,59 These data suggest that prenatal exposure to a range of toxicants does not generally result in a common pattern of impairment but instead leads to a specific neurobehavioral profile or behavioral signature. By contrast to the alcohol effects, our findings and those of other studies have consistently failed to indicate growth or neurobehavioral deficits in relation to prenatal marijuana use. Although the data presented here indicate that women report more marijuana use retrospectively, the infant outcomes cannot be used to determine which of the 2 maternal reports is more valid. Most studies of prenatal cocaine exposure have found gestational age or intrauterine growth deficits, but few, if any, cognitive effects in relation to continuous or dichotomous (yes/no) measures of prenatal cocaine exposure.13 As we have shown here, the birth size and GA effects can be detected using either antenatal or retrospective reports. We have previously reported that prenatal exposure to cocaine at sufficiently high doses early in pregnancy has the potential to produce neurobehavioral changes in infants.13 Inclusion of large numbers of light and moderately exposed infants among the exposed can obscure effects of heavy exposure. A relation to neurobehavioral outcome begins to emerge when the mother admits to heavy cocaine use, either antenatally or retrospectively, but this disclosure is not as consistent across these 2 time periods, perhaps because of fear of stigma or loss of the infant. Our finding of different thresholds for the gestational age and cognitive effects of cocaine exposure lends support to the hypothesis that these effects reflect impairment attributable to different mechanisms of action. The central nervous system deficits, poorer cognitive performance, and faster reactivity, which are believed to be mediated by direct action of cocaine on neurotransmitters, were seen only in relation to a measure of heavy exposure reported at the initial prenatal clinic visit; whereas shortened gestation, which seems to be mediated primarily through vasoconstriction, was detected at lower levels of exposure using a dichotomous measure of cocaine use collected across pregnancy. Our findings underscore the importance of examining neurobehavioral effects of cocaine exposure at heavy doses and the need to develop new interview techniques for obtaining more accurate information from mothers regarding their cocaine use in pregnancy, which can be used to determine threshold effects. A recent report indicated that greater sensitivity of a biological measure may result in detecting a greater number of lighter users but that infant deficits were related to levels detected by maternal report and urine toxicology rather than to the lower levels of exposure detected by maternal hair analysis.2
When compared with alcohol recall, maternal retrospective reports of smoking during pregnancy are more consistent with antenatal reports, probably because mothers cut down their smoking less or not at all during pregnancy. In a predominantly middle-class, white study on the effects of prenatal environmental contaminants on infant and child development, we compared maternal reports of alcohol and smoking during pregnancy assessed in the hospital immediately after delivery with retrospective reports when the children were 4 years of age.60 Maternal recall at 4 years’ postpartum for smoking was markedly more reliable (r = 0.81) than for alcohol consumption (r = 0.53), presumably because smoking is more habitual and, therefore, easier to recall. The findings on smoking in both of these studies suggest that, unlike alcohol, data on smoking during pregnancy are valid, even when collected retrospectively. These findings have important implications for recent reports suggesting that smoking during pregnancy is related to attention-deficit/hyperactivity disorder61,62 and conduct disorder63 in the offspring. Given that the validity of the retrospective smoking measure is greater than for alcohol, analyses controlling for alcohol may over attribute deficits to smoking and undercontrol for the effect of prenatal alcohol exposure.
It is important to distinguish between the phenomenon of denial among alcohol abusing individuals and the hypothesis that women generally underreport their alcohol consumption during pregnancy. The assumption that a certain proportion of very heavy drinkers will deny alcohol and drug use is generally accepted in the literature, based on data obtained via collateral report from significant others. Women may deny cocaine or other illicit drug use to avoid stigma or loss of their child. By contrast, the data reported here show that, although some denial among the heavy drinkers is to be expected, there is no systematic tendency for women generally to understate the amount they drink during pregnancy. These data are especially pertinent to studies examining the quantities of alcohol intake associated with specific developmental sequelae. If underreporting were the norm, quantities of alcohol consumption reported during pregnancy that are found to be correlated with specific deficits would have to be adjusted upward to provide an accurate index of the actual quantities of alcohol intake associated with those deficits. Our data supporting the validity of the quantities reported during pregnancy suggest that the developmental deficits that have been found to be associated with prenatal alcohol exposure11,12,14,64,65 are indeed associated with relatively low doses of alcohol intake during pregnancy.
This research was supported by the National Institute on Alcohol Abuse and Alcoholism grants RO1-AA06966 and P50-AA0706, with supplemental support from a National Institutes of Health Minority Biomedical Research Support grant SO6-RR08167; and by a grant from the Joseph Young, Sr, Fund from the State of Michigan.
We thank Joel Ager, Susan Martier, and Melissa Estrin-Kaplan, who collaborated on portions of this research, and our staff for their contributions.
- Received March 20, 2001.
- Accepted November 2, 2001.
This paper was presented, in part, at the National Institute on Alcohol Abuse and Alcoholism 25th Anniversary Celebration; May 1995; Bethesda, MD.
Reprint requests to (S.W.J.) Department of Psychiatry and Behavioral Neurosciences, Wayne State University School of Medicine, 2751 E Jefferson, Room 460, Detroit, MI 48207. E-mail:
- ↵Kuhn L, Kline 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– 119
- ↵Sokol RJ, Martier S, Ernhart C. Identification of alcohol abuse in the prenatal clinic. In: Chang NC, Chao M, eds. Early Identification of Alcohol Abuse. Rockville, MD: Alcohol, Drug Abuse, and Mental Health Administration Research; 1983. Monograph No. 17
- ↵Bowman RS, Stein LI, Newton JR. Measurement and interpretation of drinking behavior. Q J Stud Alcohol.1975;36 :1154– 1172
- ↵Winer BJ. Statistical Principles in Experimental Design. 2nd ed. New York, NY: McGraw-Hill;1971
- ↵Fagan JF, Singer LT. Infant recognition memory as a measure of intelligence. In: Lipsitt LP, ed. Advances in Infancy Research. Norwood, NJ: Ablex;1983:31– 72
- ↵Thompson LA, Fagan JF, Fulkner DW. Longitudinal prediction of specific cognitive abilities from infant novelty preference. Child Dev.1991;67 :530– 538
- Colombo J, Mitchell DW. Individual differences in early visual attention: Fixation time and information processing. In: Colombo J, Fagen J, eds. Individual Differences in Infancy: Reliability, Stability, Prediction. Hillsdale, NJ: Lawrence Erlbaum;1990:193– 227
- ↵Rose SA, Gottfried AW, Bridger WH. Cross-modal transfer in infants: relationship to prematurity and socioeconomic background. Dev Psychol.1978;14 :643– 652
- ↵Canfield RL. Stability of RT and visual expectancies from 4 to 6 months of age. Abstr Society Res Child Dev.1991;8 :199
- ↵Kagan J. Change and Continuity Infancy. New York, NY: John Wiley;1971
- ↵Bayley N. Bayley Scales of Infant Development. New York, NY: Psychological Corporation;1969
- ↵Dunn LM, Dunn LM. PPVT Manual for Forms L and M. Circle Pines, MN: American Guidance Service;1981
- ↵Caldwell BM, Bradley RH. Home Observation for Measurement of the Environment. Little Rock, AR: University of Arkansas Press;1979
- ↵Bradley RH, Caldwell BM. 174 children: a study of the relationship between home environment and cognitive development during the first 5 years. In: Gottfried AW, ed. Home Environment and Early Cognitive Development: Longitudinal Research. New York, NY: Academic Press;1984:5– 56
- ↵Brumitt GA, Jacobson JL. The influence of home environment and socio-economic status on cognitive performance at 4 years. Abstr Soc Res Child Dev.1989;6 :203
- ↵Barnard KE, Bee HL, Hammond MA. Home environment and cognitive development in a healthy, low-risk sample: the Seattle Study. In: Gottfried AW, ed. Home Environment and Early Cognitive Development: Longitudinal Research. New York, NY: Academic Press;1984
- ↵Siegel LS. Home environmental influences on cognitive development in pre-term and full-term children during the first 5 years. In: Gottfried AW, ed. Home Environment and Early Cognitive Development: Longitudinal Research. New York, NY: Academic Press;1984
- ↵Beck AT, Rush AJ, Shaw BF, Emery G. Cognitive Therapy of Depression. New York, NY: Guilford;1979
- ↵Jacobson SW, Settles DV, Lucio L, et al. Test-retest reliability of the Beck Depression Inventory over an 18-month period. NIH MARC/MBRS Symposium; 1992; Dorado Beach, Puerto Rico
- ↵Jacobson SW, Jacobson JL, Frye KF. Incidence and correlates of breastfeeding in socioeconomically disadvantaged women. Pediatrics.1991;88 :728– 736
- ↵Jacobson JL, Jacobson SW. Prenatal alcohol exposure and neurobehavioral development: where is the threshold? Alcohol Health Res World.1994;18 :30– 36
- ↵Jacobson SW, Jacobson, JL. Teratogenic insult and neurobehavioral function in infancy and childhood. In: Nelson CA, ed. The Effects of Early Adversity on Neurobehavioral Development: The Minnesota Symposia on Child Psychology. Mahwah, NJ: Erlbaum;2000:61– 112
- ↵Jacobson JL, Jacobson SW. Methodological issues in human behavioral teratology. In: Lipsitt LP, Rovee-Collier C, eds. Advances in Infancy Research. Norwood, NJ: Ablex;1990:111– 148
- Copyright © 2002 by the American Academy of Pediatrics