- Email: [email protected]
Increased mortality among women who drank alcohol during pregnancy
Alcohol 42 (2008) 603e610
Increased mortality among women who drank alcohol during pregnancy John Peder Berga, Mary Ellen Lynchb, Claire D. Colesb,* a Emory University, Department of Psychology, Atlanta, GA 30332, USA Emory University School of Medicine, Department of Psychiatry and Behavioral Sciences, Atlanta, GA 30332, USA Received 29 January 2008; received in revised form 30 May 2008; accepted 2 June 2008
b
Abstract Women giving birth to children with fetal alcohol syndrome have a higher risk of early mortality. However, the risk of increased mortality associated with drinking at lower levels during pregnancy has not been evaluated previously. Accordingly, mortality at 20 years post recruitment was examined in a sample (N 5 570) of women recruited between 1980 and 1986, who drank more than 1 ounce of absolute alcohol per week (oz AA/wk) during pregnancy and compared to that in abstainers from the same low socioeconomic, African-American population. Using data from archival information and state mortality records, Cox proportional hazards survival models were constructed to determine whether alcohol use, cigarette use, infant birth weight, infant dysmorphia, and alcohol use and abuse by the woman’s own mother (family history positive) were associated with increased risk for mortality. At follow-up (in 2003), 9.5% of the sample had died, with the rate for controls being 3.6%, for those alcohol users who stopped during pregnancy, 12.7%, and for the alcohol users who continued drinking throughout pregnancy, 12.5%. Thus, women using alcohol in pregnancy, whether they stopped or continued to use, were significantly (c2 (2) 5 12.1, P ! .01) more likely than abstainers to have died before follow-up. Nondrinkers’ mortality rate was lower than that of other women from this low-income, high-risk population, whereas the drinkers’ rate was 2.7 times higher. In a multivariate analysis, factors contributing to mortality risk included alcohol use and cigarette smoking, but not infant birth weight. Drinking at any level during pregnancy should be regarded as a risk factor for the mother as well as for offspring. Health care professionals working with such women should provide counseling and support for abstinence. Ó 2008 Elsevier Inc. All rights reserved. Keywords: Alcohol use in pregnancy; Maternal mortality; Cigarette smoking
Introduction Examination of the effects of alcohol use during pregnancy usually focuses on the risk of fetal alcohol exposure, but such use may signal excessive risk to the mother’s health as well (Stratton et al., 1996). In clinical samples, it has been noted that women giving birth to a child with fetal alcohol syndrome (FAS) often die before the child is grown (unpublished observation, Hymbaugh et al., 1995; Spohr et al., 1993). In their sample of Native American children diagnosed between birth and 17 years, May et al. (1983) reported a maternal mortality rate of 23.1%. This finding is consistent with the 30% rate reported in a Peruvian clinical group (Mena et al., 1986). These previous studies focused on mortality in women giving birth to children with the FAS and most of these women could be
* Corresponding author. Maternal Substance Abuse and Child Development Project, Department of Psychiatry and Behavioral Sciences, Emory University, 1256 Briarcliff Road, Suite 321 West, Atlanta, GA 30306, USA. Tel: þ1-404-712-9814; fax: þ1-404-712-9809. E-mail address: [email protected] (C.D. Coles). 0741-8329/08/$ e see front matter Ó 2008 Elsevier Inc. All rights reserved. doi: 10.1016/j.alcohol.2008.06.004
characterized as alcoholic; however, inability to stop drinking during pregnancy despite known risks to the fetus may be a marker for risk factors associated with premature mortality in women who have not been diagnosed as alcohol abusers or whose drinking is not yet heavy enough to warrant treatment. Such a pattern may signal a need for intervention for the mother herself as well as for the child even when drinking is moderate and the child does not have FAS (Smith et al., 1987). To evaluate this possibility, the present study examines the relationship between documented alcohol consumption during pregnancy and mortality among mothers over an interval of approximately 20 years (Mean 20.3 years [Standard deviation 6 1.7], range, 17e25 years) following a target pregnancy. There is a well-established relationship between higher levels of alcohol consumption and increased mortality in both men and women (e.g., Rehm et al., 2001, 2003). Studies using survival analysis techniques find that alcohol abuse is related to premature age at death as well as increased mortality risk (e.g., Neumark et al., 2000) due both to the development of chronic diseases, including cirrhosis and specific forms of cancer, and to negative outcomes
604
J.P. Berg et al. / Alcohol 42 (2008) 603e610
resulting from accidents, interpersonal violence, and unintentional injuries associated with substance abuse (e.g., Rivara et al., 2004; Stinson & DeBakey, 1992). However, women are at risk for detrimental consequences of alcohol use at lower levels of consumption than men even when differences in body size are taken into account (e.g., Frezza et al., 1990; Lieber, 1997). Lieber (1997) reports that women have (1) higher blood alcohol concentration than men after consuming similar amounts of alcohol and (2) are more susceptible than men to health problems associated with alcohol, such as liver damage. He attributed these effects, at least in part, to lower body water content in women and to gender differences in the metabolism of alcohol. Frezza et al. (1990) reported that, in comparison to men, women experience lower levels of activity for the enzyme gastric alcohol dehydrogenase, resulting in less alcohol being metabolized in the stomach and more being passed on through the blood to the rest of the system. This places women more at risk for health problems related to alcohol use. This difference is reflected in calculated health and mortality risks showing similar levels of risk reached for women at lower doses of alcohol than for men (Rehm et al., 2003; Rivara et al., 2004). Using the etiologic-fraction method, Rivara et al. (2004) determined the gender specific proportion of causes of death diagnoses that could be attributed to alcohol use, the alcohol etiological fraction. Overall, 1.5% of all female deaths in the United States in 2000 could be attributed to alcohol use using this method (Rivara et al., 2004). Other predictors of mortality Although the relation between alcohol consumption and premature mortality seems to be clear, other factors also may contribute to or predict premature mortality, including use of tobacco and other drugs, family history of alcohol problems, and biological vulnerabilities that may be associated with perinatal risk indicators, like lower birth weight in the infants born to the pregnant women. Use of tobacco, which is often associated with alcohol use, is a leading cause of premature death in the United States (e.g., Fellows et al., 2002; Mokdad et al., 2004). According to Armour et al. (2005), between 1997 and 2001, approximately 438,000 premature deaths that occurred per year were attributable to tobacco use. Family history of alcoholism has been related to alcohol dependence problems in women in a number of recent studies (e.g., Curran et al., 1999; Eng et al., 2005; Gogineni et al., 2006). According to Gogineni et al. (2006), women in a community sample with a family history of alcoholism in either parent were more likely to show alcohol dependence symptoms and experience negative consequences related to alcohol use than women without such history. Curran et al. (1999) reported that family history of alcoholism was a stronger predictor of alcohol problems in women than in men and that, for women, family history was
especially important for those of low socioeconomic background. Finally, not all women who use alcohol in pregnancy have affected children, suggesting that there may be increased biological vulnerability to the effects of alcohol in some mothereinfant dyads. Perinatal indicators of newborn health in the target pregnancy (e.g., birth weight or physical dysmorphia) may be related to the mother’s alcohol consumption and overall health. Thus, they may be associated with an increased risk for mortality. In summary, abusive alcohol use is clearly related to mortality and the impact of alcohol use can be quantified. Drinking during pregnancy may be a marker for women especially at risk for alcohol-related mortality. Based on this previous research, we offer the following hypotheses: 1 Women who drink alcohol during pregnancy will be more likely to die prematurely than those who do not. 2 Women who drink during pregnancy will be more likely to die from alcohol-related causes than women who do not. 3 Other predictors of premature death will include other substance use, family history of substance use/abuse, and perinatal risk indicators. Materials and methods Sample This study was carried out using archival information. This information was obtained using procedures in accordance with the ethical standards of the Medical School Institutional Review Board and with the Helsinki Declaration of 1975, as revised in 1983, and consistent with the requirements of the Health Information Portability and Accountability Act of 1996. In addition, approval of the protocols was given by the Georgia Department of Human Resources, Division of Public Health, Office of Health Information and Policy. The sample included 570 women in three groups: 196 women who abstained from alcohol use, 110 participants who drank early in their pregnancy but stopped by the second trimester, and 264 participants who drank throughout pregnancy. Recruitment occurred at a prenatal clinic in a large, urban public hospital in Atlanta, GA, between 1980 and 1986 and, at recruitment women completed an informed consent procedure approved by the Emory University School of Medicine’s Institutional Review Board. Pregnant women were screened at a prenatal visit for current alcohol use. The majority of patients applying for care at this hospital were low income and African-American. Women who reported drinking at least 1 ounce of absolute alcohol or more each week (oz AA/wk) and a control group of women who did not drink were asked to participate in a study on the effects of drinking during pregnancy. During the recruitment visit, participants were advised about the negative effects on the fetus of alcohol use during pregnancy and, when appropriate, referred for substance abuse services.
J.P. Berg et al. / Alcohol 42 (2008) 603e610
Measures At ascertainment, participants were interviewed to complete a questionnaire about age, ethnicity, current and past alcohol use, family and partner alcohol use, other drug use, and cigarette smoking. At the time of delivery, offspring birth weight was recorded and dysmorphia scores were obtained by a pediatric geneticist who completed a physical examination of infants looking for the characteristic features associated with prenatal alcohol exposure. Based on the observations of Jones and Smith (1973), the Dysmorphia Checklist used to record the results of the physical examination provides a summary score with higher scores indicating more signs associated with prenatal alcohol exposure (Coles et al., 1991). For the present study, marital status, monthly income, offspring birth weight, dysmorphia scores, drug use, and family and partner drug use were not available for all participants. There were several causes for this missing data: some participants did not volunteer this information at recruitment, some participants moved away from the area or did not give birth at the recruitment hospital, not all participants or their offspring were able to be seen by the pediatric geneticist, and some data were lost in the 27 years since recruitment. See Tables 1 and 2 for information on sample demographics, birth outcomes, and alcohol and drug use. Archival information For this study, all data were obtained from existing recordsdeither those collected previously as part of this
605
longitudinal study or from public health records. For 114 participants, offspring birth weight was not available as they had not returned to the hospital for delivery or were otherwise missed. Consequently, state birth records data from 1980 to 1985 were searched to obtain these missing birth weights. Matches were determined by the mother’s first name, mother’s last name, mother’s date of birth, child’s last name, and child’s date of birth. Birth weights were recovered for 66 offspring. It is probable that many of the remaining 48 missing birth weights were associated with fetal loss event such as miscarriage, still birth, or pregnancy termination. There was no significant difference among drinking groups in the proportion of missing birth weights. To determine mortality, state mortality archives were searched for records matching participant data. For the time period 1980e1993, hardcopy records were searched manually (unpublished manuscript, Bahna, 1995; unpublished master’s thesis, De Marco, 2000). For the time period 1994e2003, a state database of all deaths during this period was searched using a series of procedures in SPSS 13.0 (SPSS Inc., 2004). First, participant data, with a unique marker for identification, was merged with the state mortality database. Then, duplicate records were identified in the new merged database using the SPSS function ‘‘Identify Duplicate Cases.’’ Cases were initially identified as duplicate when participant last name, first name initial, and date of birth matched a state death record. Participant mortality was confirmed when the full first name, last name, date of birth, and ethnicity matched state records. Minor name
Table 1 Sample at ascertainmentddemographic and birth outcomes of women who abstained and women who drank alcohol during pregnancy Condition Variable
N
Controls
Stopped drinking
Continued drinking
Group comparison
Age (years)
570
24.2 6 4.3
25.7 6 5.2
26.3 6 4.6
F (2,567) 5 11.1***
Ethnicity (%) African American Caucasian Asian Unknown
570 82.7 9.2 1.5 6.6
89.1 8.2 0 2.7
84.5 8.3 0.4 6.8
Marital status (%) Single Married Divorced Separated Widowed
553 57.5 31.6 3.1 7.3 0.5
65.7 15.2 3.8 13.3 1.9
64.3 18.4 5.1 11.8 0.4
Monthly income (%) $0e300 $301e600 $601e900 $901þ
356
Birth weight of child (g) M (SD) Child dysmorphia score M (SD)
522 327
*P ! .05; **P ! .01; ***P ! .001.
c2 (6) 5 6.0
c2 (8) 5 19.3*
50.9 33.3 8.8 7.0
62.1 22.7 13.6 1.5
60.2 23.9 11.4 4.5
3,230.2 6 575.5 2.2 6 2.7
3,072.7 6 557.9 4.1 6 4.4
2,813.2 6 744.7 5.1 6 6.3
Correlation with mortality 0.08 0.04
0.01
c2 (6) 5 7.8
0.10
F (2,519) 5 21.5*** F (2,324) 5 9.5***
0.13** 0.12*
606
J.P. Berg et al. / Alcohol 42 (2008) 603e610
Table 2 Sample at ascertainmentdalcohol and drug use of women who abstained and women who drank alcohol during pregnancy Condition Variable
N
Controls
Stopped Drinking
Continued Drinking
Group comparison
Correlation with mortality
Smoking (%) Oz/Alcohol per week (M 6 SD) Cocaine ever? (%) Marijunana ever? (%) Opiates ever? (%) ‘‘Downers’’ever? (%) ‘‘Uppers’’ ever? (%) PCP ever? (%) LSD ever? (%)
570 570 459 515 449 510 510 504 488
29.6 0.1 6 0.1 2.0 50.0 1.4 2.9 8.7 1.2 0.6
69.1 8.6 6 10.9 20.2 80.8 4.5 10.7 16.3 4.0 6.3
77.1 10.5 6 13.6 23.3 75.1 2.3 11.5 23.6 5.2 4.0
c2 (2) 5 111.4*** F (2,567) 5 59.7*** c2 (2) 5 31.5*** c2 (2) 5 39.0*** c2 (2) 5 2.4 c2 (2) 5 10.5** c2 (2) 5 15.7*** c2 (2) 5 4.6 c2 (2) 5 7.0*
0.16*** 0.17*** 0.02 0.003 0.001 0.06 0.02 0.02 0.06
*P ! .05; **P ! .01; ***P ! .001.
spelling differences were tolerated. This search method was validated by comparing the outcomes so obtained to previous results found using manual searching techniques (see Coles et al., 2003). In two instances, records with matching first and last names but slightly different birth dates (e.g., 1 year difference) were found. Subsequently, the search method was expanded to identify records that matched first initial, last name, and date of birth plus or minus 2 years. Using this modified method, the search routine located all previous matches and identified a new match between participant data and state records. To reduce omission errors due to changed last names, marriage records were searched in the two counties in which the majority of the participants resided during recruitment. The search resulted in a list of alternate last names for each participant, which were used in the 1994e2003 SPSS searches. Using the last names obtained from the marriage records, an additional five matches were discovered. Data analysis Group differences in demographic factors and drug use were analyzed using analysis of variance and c2 procedures appropriate for the type of data. To evaluate the hypothesis that women drinking in pregnancy have a higher mortality risk than those who did not, a Cox proportional hazards survival regression (Cox, 1972) was performed comparing women who drank at anytime during pregnancy and those who abstained. The women who drank were further analyzed by comparing the mortality risk for those women who drank throughout their pregnancy to those who stopped drinking before their second trimester using Cox regression analyses. The hypothesis that other predictors of increased mortality risk would include other substance use, family history of substance use, and perinatal risk factors was evaluated by first correlating related variables to mortality. Variables significantly correlated to mortality were then analyzed in individual Cox regression analyses to determine if they predicted an increase in mortality risk. Because data were not available for all participants for some variables, results from analyses with these variables should be interpreted with caution. Finally, alcohol use status,
smoking status, and birth weight were compared for mortality risk in multiple stepwise Cox regression analyses. Results Sample characteristics Participants were primarily African-American (84.7%), single (62.2%), and had monthly incomes less than $300/ month (1980e1986 dollars or $661/month in 2008 dollars) (Consumer Price Index Calculator, 2008). Drinking status groups differed significantly on age and marital status but not income. Offspring birth weight was lower and dysmorphia scores higher among participants who drank during pregnancy (see Table 1). Participants in the stopped drinking and continued drinking groups smoked more, drank more, and reported a higher lifetime prevalence of illicit drug use (cocaine, marijuana, downers, uppers, and LSD, but not opiates or PCP) compared to participants who abstained (see Table 2). Women who drank were also more likely to have mothers, fathers, and partners who drank than were abstainers, but did not report having parents who drank heavily or had trouble due to drinking more often than did nondrinkers (see Table 3). Women who drank throughout pregnancy were not significantly different from women who stopped drinking on age at delivery, amount of alcohol consumption (oz AA/ wk), offspring dysmorphia score, marital status, ethnicity, income, lifetime prevalence of illicit drug use, having partners who drank, having fathers who drank, were heavy drinkers, or had trouble due to drinking, and having mothers who drank. Those participants who continued to drink throughout their pregnancy were more likely to report that their own mothers were heavy drinkers (c2 (1) 5 9.3, P ! .01) and had trouble due to drinking (c2 (1) 5 4.3, P ! .05), and to have infants with lower birth weights (F (1,338) 5 10.1, P ! .01). Mortality analyses Of the 570 participants recruited, 54 or 9.5% had died before follow-up in 2003. Separated by group status, 3.6% of controls, 12.7% of the stopped drinking group,
J.P. Berg et al. / Alcohol 42 (2008) 603e610
607
Table 3 Sample at ascertainmentdfamily drug and alcohol use of women who abstained and women who drank alcohol during pregnancy Condition Variable (% responding, ‘‘yes’’)
N
Controls
Stopped drinking
Continued drinking
Group comparison
Correlation with mortality
Does your mother drink? Is your mother a heavy drinker? Does your mother have trouble due to drinking? Does your father drink? Is your father a heavy drinker? Does your father have trouble due to drinking? Does the father of the baby drink?
391 378 363 391 386 361 540
17.4 9.1 2.4 22.8 11.1 5.7 11.9
33.3 2.4 1.2 37.9 18.6 11.5 34.5
39.2 15.0 7.6 42.5 17.6 9.2 41.6
c2 c2 c2 c2 c2 c2 c2
0.12* 0.12* 0.16** 0.03 0.05 0.02 0.03
(2) 5 13.8*** (2) 5 10.0** (2) 5 6.5* (2) 5 10.7** (2) 5 2.4 (2) 5 1.8 (2) 5 43.1***
*P ! .05; **P ! .01; ***P ! .001.
and 12.5% of the continued drinking group had died (c2 (2) 5 12.1, P ! .01). Fig. 1 compares the survival plots of the three drinking groups. Table 4 shows the cause of death information for all deceased participants. Also indicated is the female-specific etiological fraction (percentage of deaths attributable to alcohol) as designated by Rivara et al. (2004). Among controls, 14.3% of causes of deaths were consistent with effects of alcohol, whereas in the alcohol-using groups, 34.0% were consistent. However, a c2 analysis did not reach significance (c2 (1) 5 1.1, P 5 .29). For the purposes of the present study, an etiological fraction greater than .05 was considered attributable to alcohol. The sample death rate was compared to the average death rate for African-American women in Georgia from 1986 to 2002. During that period, the average death rate for ages 25e45 was 203.6 deaths per 100,000 people (WONDER, 2004). Extrapolating the sample death rate by alcohol use group to the same units revealed a death rate of 155.2, 553.5, and 543.5 per 100,000 for the controls, women who stopped drinking, and women who continued drinking, respectively, thus indicating a death rate lower
Fig. 1. Cumulative survival curve for women who abstained (controls), drank but stopped, and drank throughout pregnancy.
than the general population for controls but an elevated mortality rate for both alcohol user groups. Age at delivery, demographics, child outcome, smoking, alcohol and drug use, and family and partner drug use variables were evaluated for their correlations with mortality. Tables 1e3 list the variables and their correlations with mortality. Alcohol use, cigarette use, infant birth weight, infant dysmorphia score, and grandmothers’ (the mother’s own mother) alcohol use/abuse variables correlated significantly with mortality. No other variables were significantly correlated to mortality. Univariate Cox proportional hazard regression analyses were conducted on variables significantly correlated with mortality (see Table 5 for results). Drinking anytime during pregnancy predicted mortality (odds ratio [OR]: 3.7, P 5 .001). Stopping drinking by the second semester did not differ from continued drinking throughout pregnancy in predicting mortality (OR: 0.96, P 5 .908). Cigarette use increased the risk of mortality similarly to alcohol use (OR: 3.6, P ! .001). Infant birth weight was statistically significant in predicting mortality (OR: 0.9995, P 5 .002) indicating that for every gram increase in birth weight, the odds of mortality decreased 0.05%. All grandmothers’ drinking variables were positively associated with mortality. Of these familial drinking variables, ‘‘trouble due to drinking’’ presented the highest risk (OR: 4.0, P 5 .002) whereas alcohol use (OR: 2.1, P 5 .018) and heavy use (OR: 2.3, P 5 .026) had lower but significant associated risks. Finally, infant dysmorphia score was statistically significant in predicting mortality (OR: 1.06, P 5 .027) indicating a 6% increase in the odds of mortality for every unit increase in the dysmorphia score. Because dysmorphia score and grandmothers’ drinking data were not available for all participants, analyses with these variables were exploratory in nature and the results should be interpreted with caution. A multivariate Cox regression was performed using the birth weight, smoking status, and drinking status variables. Table 5 presents the results of this analysis. Grandmother’s drinking and child dysmorphia score were excluded because only 40% of the full sample had data for these variables after excluding cases missing data listwise. Both smoking (OR: 2.5, P 5 .026) and drinking (OR: 2.5, P 5 044) status remained significant in the multivariate
608
J.P. Berg et al. / Alcohol 42 (2008) 603e610
Table 4 Causes of death of controls and alcohol users Age at death Cause of death
Etiological fraction*
Controls 66 99 246 417 490 526 596
42 35 45 29 37 38 44
Breast cancer Obesity Obesity Homicide Breast cancer HIV Chronic renal failure
0.007
Alcohol 33 Users 44 224 389 481 595 298 324 393 444 52 211 551 564 112 395 545 129 299 32 512 178 279
28 47 46 41 40 46 34 21 31 43 45 36 25 43 53 47 43 58 38 42 32 41 57
1.000 1.000 1.000 1.000 1.000 1.000 0.315 0.315 0.315 0.315 0.315 0.109 0.109 0.041 0.058 0.227 0.270 0.0015
292 100 123 128 150 160 192 219 233 235 274
41 34 35 50 35 49 39 32 51 37 46
293 322 434 435 437 456 458 476 499 524 533 540 563
37 52 30 39 47 36 45 32 31 39 41 44 34
Alcohol abuse Mental disorder due to alcohol use Alcoholic hepatic failure Alcohol dependence syndrome Alcohol dependence syndrome Alcoholic cirrhosis of liver Homicide Homicide Homicide Homicide Homicide Car accident Car accident Larynx cancer Cancer on palate Suicide Accidental poisoning by narcotics Atherosclerotic heart disease Inhalation of object Other disorders of the intestine Cerebrovascular disease Diabetes mellitus End-stage renal disease pneumonia Pneumonia Renal failure HIV Pulmonary embolism Hepatic failure Congestive heart failure HIV HIV HIV HIV Unspecified disorder of kidney and urethra Atherosclerotic heart disease CancerdUnspecified End-stage renal disease Unknown Pulmonary embolism Septic emboli Chronic renal failure Cervical cancer HIV HIV HIV HIV Heart failure
ID
*Rivara, et al. (2004)
0.315 0.007
regression whereas birth weight was not (OR: 0.9995, P 5 .085) despite being statistically significant in the univariate analyses. Discussion These results confirm that drinking during pregnancy may be a marker for premature mortality among women. More than 12% of the women who drank during pregnancy were deceased at follow-up in 2003 compared to 3.6% of those who abstained and the mortality rate in the alcoholusing group was higher than the population rate for women of this age. Although drinking during pregnancy is recognized as a risk factor for offspring, these results suggest that the mothers’ own health is at risk and that those who drink during pregnancy are in need of intervention both to assure a healthy birth and to support the mother’s own long-term well being. The results are consistent with prior literature in that heavier drinking is associated with higher mortality rates (e.g., Rehm et al., 2001, 2003). Other important predictive factors included smoking during pregnancy, which is probably associated with continued smoking, and a family history of problem drinking by the participant’s own mother (grandmother). Although a family history of alcohol problems is established as a predictor of alcohol problems in offspring, the literature is conflicted concerning the relative impact of paternal versus maternal drinking patterns (see Gogineni et al., 2006). It is possible that maternal influence is more salient or influential than paternal in daughters’ socialization in this low-income sample and that this explains the impact of the grandmother’s drinking behavior. However, there may also be a biological explanation. The results suggest some association between biological markers of offspring health (birth weight and dysmorphia score) and an increased risk of mortality for the mother. Both birth weight and dysmorphia score were correlated with mortality and both significantly predicted mortality in univariate Cox regression models. Biological markers of alcohol exposure in offspring may indicate a reduced ability to cope with alcohol’s damaging effects in the mother. However, in multivariate analyses birth weight was no longer significant in predicting mortality in the presence of alcohol and cigarette use. This suggests that the association between birth weight and mortality was strongly mediated by alcohol and cigarette use. The factors that were not associated with mortality are notable. Variables normally associated with mortality such as age at delivery were not a factor in sample mortality over the follow-up despite the fact some women were over 40 years of age at delivery. Ethnicity, income, marital status, lifetime illicit drug use, partner alcohol use, and alcohol use/abuse by the woman’s own father (grandfather) were also not significantly correlated with mortality. These findings indicate the marked, detrimental effect of alcohol and cigarette use in an otherwise healthy sample.
J.P. Berg et al. / Alcohol 42 (2008) 603e610
609
Table 5 Predictors of mortality: results of Cox proportional hazard regression analyses (univariate and multivariate) Predictor
N
Univariate model; HR (95% CI)
Drank alcohol during pregnancy Drinking throughout vs. stopping by second trimester Smoked during pregnancy Offspring birth weight Maternal drinkingb Maternal heavy drinkingb Maternal trouble due to drinkingb Offspring dysmorphia scoreb
570 374 570 522 391 378 363 327
3.7** (1.7e8.2) 0.96 (0.52e1.80) 3.6***(1.8e7.4) 0.9995** (0.9992e0.9998) 2.1* (1.1e3.7) 2.3* (1.1e4.8) 4.0** (1.7e9.4) 1.06* (1.01e1.11)
Multivariate modela; HR (95% CI) 2.5* (1.0e6.3) 2.5* (1.1e5.5) 0.9995 (0.9992e0.9999)
HR 5 Hazard ratio, CI 5 Confidence interval. *P ! .05; **P ! .01; ***P ! .001. a In the multivariate model, only significant (P ! .05), nonexploratory predictors from the univariate analyses were included. b Exploratory variable due to missing data.
There are some limitations to this study. Mortality data include only deaths recorded in the state of Georgia. If a participant moved to another state and her death occurred there, it would not be recorded in the Georgia state death records. Thus, these results may be an underestimate of mortality. Access to national death records would make the record search more comprehensive. Another limitation results from the way in which the original data were collected. We were limited in our analysis to archival information and could only select variables for analysis from those collected at recruitment and when the infant was born. Thus, we would like to know more about family history of alcoholism and the participants’ medical history subsequent to this target pregnancy but the information is not available. Similarly, we would like to have current medical histories for the women who are still living to evaluate morbidity as well as mortality but cannot access that information. Finally, we believe that there are significant limitations in the way in which the alcohol etiological fraction is calculated and that the method used underestimates the contribution of alcohol use and abuse to mortality. For instance, HIV is listed as the primary cause of death for 10 women in the sample. It is likely that substance abuse is associated with a higher risk for HIV but we cannot calculate this likelihood. Similarly, subject number 458, whose cause of death is listed as ‘‘chronic renal failure’’ that is not attributable to alcohol using this system (Rivara et al., 2004), was known to us as an end-stage alcoholic. As previously noted, the mortality rate for this sample is quite high as compared to that for African-American women in Georgia. Participants in this sample were drawn from a low-income, urban sample and, therefore, compose a high-risk group. Incidence of deaths due to violence, HIV, and other causes associated with risks of living in an impoverished environment are probably elevated in this sample. In summary, the results of this 20-year follow-up show that use of alcohol during pregnancy is associated with premature mortality among women. Although it had been observed previously (Stratton et al., 1996) that mothers of children with FAS were at risk for premature mortality, the present study suggests that this risk may be more
general. Such behavior should be regarded by health care providers as a signal that intervention is needed to assist the mother in dealing with her drinking problems as well as other behaviors that may be placing her at risk. Intervention at this point could have long-term positive effects on the mother’s health. Acknowledgments We would like to thank Colin Smith, Sheila Butler, and Robert Attaway at the Georgia Department of Human Resources, Division of Public Health, Office of Health Information and Policy for their assistance in obtaining Georgia mortality and birth records. We would also like to thank the Fulton County and Dekalb County probate courts for their assistance in obtaining marriage records. References Armour, B. S., Woolery, T., Malarcher, A., Pechacek, T. F., and Husten, C. (2005). Annual smoking-attributable mortality, years of potential life lost, and productivity lossesdUnited States, 1997-2001. Morb. Mortal. Wkly. Rep. 54, 625–628. Coles, C. D., Brown, R. T., Smith, I. E., Platzman, K. A., Erickson, S., and Falek, A. (1991). Effects of prenatal alcohol exposure at school age: I. Physical and cognitive development. Neurotoxicol. Teratol. 13, 357– 367. Coles, C. D., Lynch, M. E., DeMarco, M., and Corley, T. C. (2003). Alcohol use during pregnancy and maternal mortality: an update. Alcohol. Clin. Exp. Res. 27, 375. Consumer Price Index Calculator (2008). Online calculator. Federal Reserve Bank of Minneapolis. Available at: http://woodrow.mpls.frb. fed.us/Research/data/us/calc. Accessed May 15, 2008. Cox, D. R. (1972). Regression models and life tables. J. R. Stat. Soc. Ser. B 34, 187–220. Curran, G. M., Stoltenberg, S. F., Hill, E. M., Mudd, S. A., Blow, F. C., and Zucker, R. A. (1999). Gender differences in the relationships among SES, family history of alcohol disorders and alcohol dependence. J. Stud. Alcohol. 60, 825–832. Eng, M. Y., Schuckit, M. A., and Smith, T. L. (2005). The level of response to alcohol in daughters of alcoholics and controls. Drug. Alcohol. Depend. 79, 83–93. Fellows, J. L., Trosclair, A., Adams, E. K., and Rivera, C. C. (2002). Annual smoking-attributable mortality, years of potential life lost, and economic costsdUnited States, 1995-1999. Morb. Mortal. Wkly. Rep. 51, 300–303.
610
J.P. Berg et al. / Alcohol 42 (2008) 603e610
Frezza, M., Di Padova, C., Pozzato, G., Terpin, M., Baraona, E., and Lieber, C. S. (1990). High blood alcohol levels in women: the role of decreased gastric alcohol dehydrogenase activity and first-pass metabolism. N. Engl. J. Med. 322, 95–99. Gogineni, A., King, S., Jackson, K., Kramer, J., Bucholz, K., Chan, G., et al. (2006). Female offspring of alcoholic individuals: recent findings on alcoholism and psychopathology risks: Symposium presented at the Research Society on Alcoholism, 2004, Vancouver, Aruna Gogineni, Chair. Alcohol. Clin. Exp. Res. 30, 377–387. Jones, K. L., and Smith, D. W. (1973). Recognition of the fetal alcohol syndrome in early infancy. Lancet 2, 999–1001. Lieber, C. S. (1997). Gender differences in alcohol metabolism and susceptibility. In R. W. Wilsnack, & S. C. Wilsnack (Eds.), Gender and Alcohol: Individual and Social Perspectives (pp. 77–89). New Brunswick, NJ: Rutgers Center of Alcohol Studies. May, P. A., Hymbaugh, K. J., Aase, J. M., and Samet, J. M. (1983). Epidemiology of fetal alcohol syndrome among American Indians of the Southwest. Soc. Biol. 30, 374–387. Mena, M., Casanueva, V., Fernandez, E., Carraso, R., and Perez, H. (1986). Fetal alcohol syndrome at schools for mentally handicapped children in Concepcio´n. Chile. PAHO Bull. 20, 157–169. Mokdad, A. H., Marks, J. S., Stroup, D. F., and Gerberding, J. L. (2004). Actual causes of death in the United States, 2000. J. Am. Med. Assoc. 291, 1238–1245.
Neumark, Y. D., Van Etten, M. L., and Anthony, J. C. (2000). ‘‘Alcohol dependence’’ and death: survival analysis of the Baltimore ECA sample from 1981 to 1995. Subst. Use. Misuse. 35, 533–549. Rehm, J., Gmel, G., Sempos, C. T., and Trevisan, M. (2003). Alcoholrelated morbidity and mortality. Alcohol. Res. Health. 27, 39–51. Rehm, J., Gutjahr, E., and Gmel, G. (2001). Alcohol and all-cause mortality: a pooled analysis. Contemp. Drug. Probl. 28, 337–361. Rivara, F. P., Garrison, M. M., Ebel, B., McCarty, C. A., and Christakis, D. A. (2004). Mortality attributable to harmful drinking in the United States, 2000. J. Stud. Alcohol. 65, 530–536. Smith, I. E., Lancaster, J. S., Moss-Wells, S., Coles, C. D., and Falek, A. (1987). Identifying high-risk pregnant drinkers: biological and behavioral correlates of continuous heavy drinking during pregnancy. J. Stud. Alcohol. 48, 304–309. Spohr, H. L., Willms, J., and Steinhausen, H. C. (1993). Prenatal alcohol exposure and long-term developmental consequences. Lancet 341, 907–910. SPSS (2004). Computer Program. Version 13.0. Chicago, IL: SPSS Corp. Stinson, F. S., and DeBakey, S. F. (1992). Alcohol-related mortality in the United States, 1979-1988. Br. J. Addict. 87, 777–783. Stratton, K., Howe, C., and Battaglia, F. (1996). Fetal Alcohol Syndrome: Diagnosis, Epidemiology, Prevention, and Treatment. Washington, DC: National Academy Press. WONDER [online database] (2004). (Atlanta, GA: Center for Disease Control). Available at: http://wonder.cdc.gov. Updated March 2004.
Increased mortality among women who drank alcohol during pregnancy John Peder Berga, Mary Ellen Lynchb, Claire D. Colesb,* a Emory University, Department of Psychology, Atlanta, GA 30332, USA Emory University School of Medicine, Department of Psychiatry and Behavioral Sciences, Atlanta, GA 30332, USA Received 29 January 2008; received in revised form 30 May 2008; accepted 2 June 2008
b
Abstract Women giving birth to children with fetal alcohol syndrome have a higher risk of early mortality. However, the risk of increased mortality associated with drinking at lower levels during pregnancy has not been evaluated previously. Accordingly, mortality at 20 years post recruitment was examined in a sample (N 5 570) of women recruited between 1980 and 1986, who drank more than 1 ounce of absolute alcohol per week (oz AA/wk) during pregnancy and compared to that in abstainers from the same low socioeconomic, African-American population. Using data from archival information and state mortality records, Cox proportional hazards survival models were constructed to determine whether alcohol use, cigarette use, infant birth weight, infant dysmorphia, and alcohol use and abuse by the woman’s own mother (family history positive) were associated with increased risk for mortality. At follow-up (in 2003), 9.5% of the sample had died, with the rate for controls being 3.6%, for those alcohol users who stopped during pregnancy, 12.7%, and for the alcohol users who continued drinking throughout pregnancy, 12.5%. Thus, women using alcohol in pregnancy, whether they stopped or continued to use, were significantly (c2 (2) 5 12.1, P ! .01) more likely than abstainers to have died before follow-up. Nondrinkers’ mortality rate was lower than that of other women from this low-income, high-risk population, whereas the drinkers’ rate was 2.7 times higher. In a multivariate analysis, factors contributing to mortality risk included alcohol use and cigarette smoking, but not infant birth weight. Drinking at any level during pregnancy should be regarded as a risk factor for the mother as well as for offspring. Health care professionals working with such women should provide counseling and support for abstinence. Ó 2008 Elsevier Inc. All rights reserved. Keywords: Alcohol use in pregnancy; Maternal mortality; Cigarette smoking
Introduction Examination of the effects of alcohol use during pregnancy usually focuses on the risk of fetal alcohol exposure, but such use may signal excessive risk to the mother’s health as well (Stratton et al., 1996). In clinical samples, it has been noted that women giving birth to a child with fetal alcohol syndrome (FAS) often die before the child is grown (unpublished observation, Hymbaugh et al., 1995; Spohr et al., 1993). In their sample of Native American children diagnosed between birth and 17 years, May et al. (1983) reported a maternal mortality rate of 23.1%. This finding is consistent with the 30% rate reported in a Peruvian clinical group (Mena et al., 1986). These previous studies focused on mortality in women giving birth to children with the FAS and most of these women could be
* Corresponding author. Maternal Substance Abuse and Child Development Project, Department of Psychiatry and Behavioral Sciences, Emory University, 1256 Briarcliff Road, Suite 321 West, Atlanta, GA 30306, USA. Tel: þ1-404-712-9814; fax: þ1-404-712-9809. E-mail address: [email protected] (C.D. Coles). 0741-8329/08/$ e see front matter Ó 2008 Elsevier Inc. All rights reserved. doi: 10.1016/j.alcohol.2008.06.004
characterized as alcoholic; however, inability to stop drinking during pregnancy despite known risks to the fetus may be a marker for risk factors associated with premature mortality in women who have not been diagnosed as alcohol abusers or whose drinking is not yet heavy enough to warrant treatment. Such a pattern may signal a need for intervention for the mother herself as well as for the child even when drinking is moderate and the child does not have FAS (Smith et al., 1987). To evaluate this possibility, the present study examines the relationship between documented alcohol consumption during pregnancy and mortality among mothers over an interval of approximately 20 years (Mean 20.3 years [Standard deviation 6 1.7], range, 17e25 years) following a target pregnancy. There is a well-established relationship between higher levels of alcohol consumption and increased mortality in both men and women (e.g., Rehm et al., 2001, 2003). Studies using survival analysis techniques find that alcohol abuse is related to premature age at death as well as increased mortality risk (e.g., Neumark et al., 2000) due both to the development of chronic diseases, including cirrhosis and specific forms of cancer, and to negative outcomes
604
J.P. Berg et al. / Alcohol 42 (2008) 603e610
resulting from accidents, interpersonal violence, and unintentional injuries associated with substance abuse (e.g., Rivara et al., 2004; Stinson & DeBakey, 1992). However, women are at risk for detrimental consequences of alcohol use at lower levels of consumption than men even when differences in body size are taken into account (e.g., Frezza et al., 1990; Lieber, 1997). Lieber (1997) reports that women have (1) higher blood alcohol concentration than men after consuming similar amounts of alcohol and (2) are more susceptible than men to health problems associated with alcohol, such as liver damage. He attributed these effects, at least in part, to lower body water content in women and to gender differences in the metabolism of alcohol. Frezza et al. (1990) reported that, in comparison to men, women experience lower levels of activity for the enzyme gastric alcohol dehydrogenase, resulting in less alcohol being metabolized in the stomach and more being passed on through the blood to the rest of the system. This places women more at risk for health problems related to alcohol use. This difference is reflected in calculated health and mortality risks showing similar levels of risk reached for women at lower doses of alcohol than for men (Rehm et al., 2003; Rivara et al., 2004). Using the etiologic-fraction method, Rivara et al. (2004) determined the gender specific proportion of causes of death diagnoses that could be attributed to alcohol use, the alcohol etiological fraction. Overall, 1.5% of all female deaths in the United States in 2000 could be attributed to alcohol use using this method (Rivara et al., 2004). Other predictors of mortality Although the relation between alcohol consumption and premature mortality seems to be clear, other factors also may contribute to or predict premature mortality, including use of tobacco and other drugs, family history of alcohol problems, and biological vulnerabilities that may be associated with perinatal risk indicators, like lower birth weight in the infants born to the pregnant women. Use of tobacco, which is often associated with alcohol use, is a leading cause of premature death in the United States (e.g., Fellows et al., 2002; Mokdad et al., 2004). According to Armour et al. (2005), between 1997 and 2001, approximately 438,000 premature deaths that occurred per year were attributable to tobacco use. Family history of alcoholism has been related to alcohol dependence problems in women in a number of recent studies (e.g., Curran et al., 1999; Eng et al., 2005; Gogineni et al., 2006). According to Gogineni et al. (2006), women in a community sample with a family history of alcoholism in either parent were more likely to show alcohol dependence symptoms and experience negative consequences related to alcohol use than women without such history. Curran et al. (1999) reported that family history of alcoholism was a stronger predictor of alcohol problems in women than in men and that, for women, family history was
especially important for those of low socioeconomic background. Finally, not all women who use alcohol in pregnancy have affected children, suggesting that there may be increased biological vulnerability to the effects of alcohol in some mothereinfant dyads. Perinatal indicators of newborn health in the target pregnancy (e.g., birth weight or physical dysmorphia) may be related to the mother’s alcohol consumption and overall health. Thus, they may be associated with an increased risk for mortality. In summary, abusive alcohol use is clearly related to mortality and the impact of alcohol use can be quantified. Drinking during pregnancy may be a marker for women especially at risk for alcohol-related mortality. Based on this previous research, we offer the following hypotheses: 1 Women who drink alcohol during pregnancy will be more likely to die prematurely than those who do not. 2 Women who drink during pregnancy will be more likely to die from alcohol-related causes than women who do not. 3 Other predictors of premature death will include other substance use, family history of substance use/abuse, and perinatal risk indicators. Materials and methods Sample This study was carried out using archival information. This information was obtained using procedures in accordance with the ethical standards of the Medical School Institutional Review Board and with the Helsinki Declaration of 1975, as revised in 1983, and consistent with the requirements of the Health Information Portability and Accountability Act of 1996. In addition, approval of the protocols was given by the Georgia Department of Human Resources, Division of Public Health, Office of Health Information and Policy. The sample included 570 women in three groups: 196 women who abstained from alcohol use, 110 participants who drank early in their pregnancy but stopped by the second trimester, and 264 participants who drank throughout pregnancy. Recruitment occurred at a prenatal clinic in a large, urban public hospital in Atlanta, GA, between 1980 and 1986 and, at recruitment women completed an informed consent procedure approved by the Emory University School of Medicine’s Institutional Review Board. Pregnant women were screened at a prenatal visit for current alcohol use. The majority of patients applying for care at this hospital were low income and African-American. Women who reported drinking at least 1 ounce of absolute alcohol or more each week (oz AA/wk) and a control group of women who did not drink were asked to participate in a study on the effects of drinking during pregnancy. During the recruitment visit, participants were advised about the negative effects on the fetus of alcohol use during pregnancy and, when appropriate, referred for substance abuse services.
J.P. Berg et al. / Alcohol 42 (2008) 603e610
Measures At ascertainment, participants were interviewed to complete a questionnaire about age, ethnicity, current and past alcohol use, family and partner alcohol use, other drug use, and cigarette smoking. At the time of delivery, offspring birth weight was recorded and dysmorphia scores were obtained by a pediatric geneticist who completed a physical examination of infants looking for the characteristic features associated with prenatal alcohol exposure. Based on the observations of Jones and Smith (1973), the Dysmorphia Checklist used to record the results of the physical examination provides a summary score with higher scores indicating more signs associated with prenatal alcohol exposure (Coles et al., 1991). For the present study, marital status, monthly income, offspring birth weight, dysmorphia scores, drug use, and family and partner drug use were not available for all participants. There were several causes for this missing data: some participants did not volunteer this information at recruitment, some participants moved away from the area or did not give birth at the recruitment hospital, not all participants or their offspring were able to be seen by the pediatric geneticist, and some data were lost in the 27 years since recruitment. See Tables 1 and 2 for information on sample demographics, birth outcomes, and alcohol and drug use. Archival information For this study, all data were obtained from existing recordsdeither those collected previously as part of this
605
longitudinal study or from public health records. For 114 participants, offspring birth weight was not available as they had not returned to the hospital for delivery or were otherwise missed. Consequently, state birth records data from 1980 to 1985 were searched to obtain these missing birth weights. Matches were determined by the mother’s first name, mother’s last name, mother’s date of birth, child’s last name, and child’s date of birth. Birth weights were recovered for 66 offspring. It is probable that many of the remaining 48 missing birth weights were associated with fetal loss event such as miscarriage, still birth, or pregnancy termination. There was no significant difference among drinking groups in the proportion of missing birth weights. To determine mortality, state mortality archives were searched for records matching participant data. For the time period 1980e1993, hardcopy records were searched manually (unpublished manuscript, Bahna, 1995; unpublished master’s thesis, De Marco, 2000). For the time period 1994e2003, a state database of all deaths during this period was searched using a series of procedures in SPSS 13.0 (SPSS Inc., 2004). First, participant data, with a unique marker for identification, was merged with the state mortality database. Then, duplicate records were identified in the new merged database using the SPSS function ‘‘Identify Duplicate Cases.’’ Cases were initially identified as duplicate when participant last name, first name initial, and date of birth matched a state death record. Participant mortality was confirmed when the full first name, last name, date of birth, and ethnicity matched state records. Minor name
Table 1 Sample at ascertainmentddemographic and birth outcomes of women who abstained and women who drank alcohol during pregnancy Condition Variable
N
Controls
Stopped drinking
Continued drinking
Group comparison
Age (years)
570
24.2 6 4.3
25.7 6 5.2
26.3 6 4.6
F (2,567) 5 11.1***
Ethnicity (%) African American Caucasian Asian Unknown
570 82.7 9.2 1.5 6.6
89.1 8.2 0 2.7
84.5 8.3 0.4 6.8
Marital status (%) Single Married Divorced Separated Widowed
553 57.5 31.6 3.1 7.3 0.5
65.7 15.2 3.8 13.3 1.9
64.3 18.4 5.1 11.8 0.4
Monthly income (%) $0e300 $301e600 $601e900 $901þ
356
Birth weight of child (g) M (SD) Child dysmorphia score M (SD)
522 327
*P ! .05; **P ! .01; ***P ! .001.
c2 (6) 5 6.0
c2 (8) 5 19.3*
50.9 33.3 8.8 7.0
62.1 22.7 13.6 1.5
60.2 23.9 11.4 4.5
3,230.2 6 575.5 2.2 6 2.7
3,072.7 6 557.9 4.1 6 4.4
2,813.2 6 744.7 5.1 6 6.3
Correlation with mortality 0.08 0.04
0.01
c2 (6) 5 7.8
0.10
F (2,519) 5 21.5*** F (2,324) 5 9.5***
0.13** 0.12*
606
J.P. Berg et al. / Alcohol 42 (2008) 603e610
Table 2 Sample at ascertainmentdalcohol and drug use of women who abstained and women who drank alcohol during pregnancy Condition Variable
N
Controls
Stopped Drinking
Continued Drinking
Group comparison
Correlation with mortality
Smoking (%) Oz/Alcohol per week (M 6 SD) Cocaine ever? (%) Marijunana ever? (%) Opiates ever? (%) ‘‘Downers’’ever? (%) ‘‘Uppers’’ ever? (%) PCP ever? (%) LSD ever? (%)
570 570 459 515 449 510 510 504 488
29.6 0.1 6 0.1 2.0 50.0 1.4 2.9 8.7 1.2 0.6
69.1 8.6 6 10.9 20.2 80.8 4.5 10.7 16.3 4.0 6.3
77.1 10.5 6 13.6 23.3 75.1 2.3 11.5 23.6 5.2 4.0
c2 (2) 5 111.4*** F (2,567) 5 59.7*** c2 (2) 5 31.5*** c2 (2) 5 39.0*** c2 (2) 5 2.4 c2 (2) 5 10.5** c2 (2) 5 15.7*** c2 (2) 5 4.6 c2 (2) 5 7.0*
0.16*** 0.17*** 0.02 0.003 0.001 0.06 0.02 0.02 0.06
*P ! .05; **P ! .01; ***P ! .001.
spelling differences were tolerated. This search method was validated by comparing the outcomes so obtained to previous results found using manual searching techniques (see Coles et al., 2003). In two instances, records with matching first and last names but slightly different birth dates (e.g., 1 year difference) were found. Subsequently, the search method was expanded to identify records that matched first initial, last name, and date of birth plus or minus 2 years. Using this modified method, the search routine located all previous matches and identified a new match between participant data and state records. To reduce omission errors due to changed last names, marriage records were searched in the two counties in which the majority of the participants resided during recruitment. The search resulted in a list of alternate last names for each participant, which were used in the 1994e2003 SPSS searches. Using the last names obtained from the marriage records, an additional five matches were discovered. Data analysis Group differences in demographic factors and drug use were analyzed using analysis of variance and c2 procedures appropriate for the type of data. To evaluate the hypothesis that women drinking in pregnancy have a higher mortality risk than those who did not, a Cox proportional hazards survival regression (Cox, 1972) was performed comparing women who drank at anytime during pregnancy and those who abstained. The women who drank were further analyzed by comparing the mortality risk for those women who drank throughout their pregnancy to those who stopped drinking before their second trimester using Cox regression analyses. The hypothesis that other predictors of increased mortality risk would include other substance use, family history of substance use, and perinatal risk factors was evaluated by first correlating related variables to mortality. Variables significantly correlated to mortality were then analyzed in individual Cox regression analyses to determine if they predicted an increase in mortality risk. Because data were not available for all participants for some variables, results from analyses with these variables should be interpreted with caution. Finally, alcohol use status,
smoking status, and birth weight were compared for mortality risk in multiple stepwise Cox regression analyses. Results Sample characteristics Participants were primarily African-American (84.7%), single (62.2%), and had monthly incomes less than $300/ month (1980e1986 dollars or $661/month in 2008 dollars) (Consumer Price Index Calculator, 2008). Drinking status groups differed significantly on age and marital status but not income. Offspring birth weight was lower and dysmorphia scores higher among participants who drank during pregnancy (see Table 1). Participants in the stopped drinking and continued drinking groups smoked more, drank more, and reported a higher lifetime prevalence of illicit drug use (cocaine, marijuana, downers, uppers, and LSD, but not opiates or PCP) compared to participants who abstained (see Table 2). Women who drank were also more likely to have mothers, fathers, and partners who drank than were abstainers, but did not report having parents who drank heavily or had trouble due to drinking more often than did nondrinkers (see Table 3). Women who drank throughout pregnancy were not significantly different from women who stopped drinking on age at delivery, amount of alcohol consumption (oz AA/ wk), offspring dysmorphia score, marital status, ethnicity, income, lifetime prevalence of illicit drug use, having partners who drank, having fathers who drank, were heavy drinkers, or had trouble due to drinking, and having mothers who drank. Those participants who continued to drink throughout their pregnancy were more likely to report that their own mothers were heavy drinkers (c2 (1) 5 9.3, P ! .01) and had trouble due to drinking (c2 (1) 5 4.3, P ! .05), and to have infants with lower birth weights (F (1,338) 5 10.1, P ! .01). Mortality analyses Of the 570 participants recruited, 54 or 9.5% had died before follow-up in 2003. Separated by group status, 3.6% of controls, 12.7% of the stopped drinking group,
J.P. Berg et al. / Alcohol 42 (2008) 603e610
607
Table 3 Sample at ascertainmentdfamily drug and alcohol use of women who abstained and women who drank alcohol during pregnancy Condition Variable (% responding, ‘‘yes’’)
N
Controls
Stopped drinking
Continued drinking
Group comparison
Correlation with mortality
Does your mother drink? Is your mother a heavy drinker? Does your mother have trouble due to drinking? Does your father drink? Is your father a heavy drinker? Does your father have trouble due to drinking? Does the father of the baby drink?
391 378 363 391 386 361 540
17.4 9.1 2.4 22.8 11.1 5.7 11.9
33.3 2.4 1.2 37.9 18.6 11.5 34.5
39.2 15.0 7.6 42.5 17.6 9.2 41.6
c2 c2 c2 c2 c2 c2 c2
0.12* 0.12* 0.16** 0.03 0.05 0.02 0.03
(2) 5 13.8*** (2) 5 10.0** (2) 5 6.5* (2) 5 10.7** (2) 5 2.4 (2) 5 1.8 (2) 5 43.1***
*P ! .05; **P ! .01; ***P ! .001.
and 12.5% of the continued drinking group had died (c2 (2) 5 12.1, P ! .01). Fig. 1 compares the survival plots of the three drinking groups. Table 4 shows the cause of death information for all deceased participants. Also indicated is the female-specific etiological fraction (percentage of deaths attributable to alcohol) as designated by Rivara et al. (2004). Among controls, 14.3% of causes of deaths were consistent with effects of alcohol, whereas in the alcohol-using groups, 34.0% were consistent. However, a c2 analysis did not reach significance (c2 (1) 5 1.1, P 5 .29). For the purposes of the present study, an etiological fraction greater than .05 was considered attributable to alcohol. The sample death rate was compared to the average death rate for African-American women in Georgia from 1986 to 2002. During that period, the average death rate for ages 25e45 was 203.6 deaths per 100,000 people (WONDER, 2004). Extrapolating the sample death rate by alcohol use group to the same units revealed a death rate of 155.2, 553.5, and 543.5 per 100,000 for the controls, women who stopped drinking, and women who continued drinking, respectively, thus indicating a death rate lower
Fig. 1. Cumulative survival curve for women who abstained (controls), drank but stopped, and drank throughout pregnancy.
than the general population for controls but an elevated mortality rate for both alcohol user groups. Age at delivery, demographics, child outcome, smoking, alcohol and drug use, and family and partner drug use variables were evaluated for their correlations with mortality. Tables 1e3 list the variables and their correlations with mortality. Alcohol use, cigarette use, infant birth weight, infant dysmorphia score, and grandmothers’ (the mother’s own mother) alcohol use/abuse variables correlated significantly with mortality. No other variables were significantly correlated to mortality. Univariate Cox proportional hazard regression analyses were conducted on variables significantly correlated with mortality (see Table 5 for results). Drinking anytime during pregnancy predicted mortality (odds ratio [OR]: 3.7, P 5 .001). Stopping drinking by the second semester did not differ from continued drinking throughout pregnancy in predicting mortality (OR: 0.96, P 5 .908). Cigarette use increased the risk of mortality similarly to alcohol use (OR: 3.6, P ! .001). Infant birth weight was statistically significant in predicting mortality (OR: 0.9995, P 5 .002) indicating that for every gram increase in birth weight, the odds of mortality decreased 0.05%. All grandmothers’ drinking variables were positively associated with mortality. Of these familial drinking variables, ‘‘trouble due to drinking’’ presented the highest risk (OR: 4.0, P 5 .002) whereas alcohol use (OR: 2.1, P 5 .018) and heavy use (OR: 2.3, P 5 .026) had lower but significant associated risks. Finally, infant dysmorphia score was statistically significant in predicting mortality (OR: 1.06, P 5 .027) indicating a 6% increase in the odds of mortality for every unit increase in the dysmorphia score. Because dysmorphia score and grandmothers’ drinking data were not available for all participants, analyses with these variables were exploratory in nature and the results should be interpreted with caution. A multivariate Cox regression was performed using the birth weight, smoking status, and drinking status variables. Table 5 presents the results of this analysis. Grandmother’s drinking and child dysmorphia score were excluded because only 40% of the full sample had data for these variables after excluding cases missing data listwise. Both smoking (OR: 2.5, P 5 .026) and drinking (OR: 2.5, P 5 044) status remained significant in the multivariate
608
J.P. Berg et al. / Alcohol 42 (2008) 603e610
Table 4 Causes of death of controls and alcohol users Age at death Cause of death
Etiological fraction*
Controls 66 99 246 417 490 526 596
42 35 45 29 37 38 44
Breast cancer Obesity Obesity Homicide Breast cancer HIV Chronic renal failure
0.007
Alcohol 33 Users 44 224 389 481 595 298 324 393 444 52 211 551 564 112 395 545 129 299 32 512 178 279
28 47 46 41 40 46 34 21 31 43 45 36 25 43 53 47 43 58 38 42 32 41 57
1.000 1.000 1.000 1.000 1.000 1.000 0.315 0.315 0.315 0.315 0.315 0.109 0.109 0.041 0.058 0.227 0.270 0.0015
292 100 123 128 150 160 192 219 233 235 274
41 34 35 50 35 49 39 32 51 37 46
293 322 434 435 437 456 458 476 499 524 533 540 563
37 52 30 39 47 36 45 32 31 39 41 44 34
Alcohol abuse Mental disorder due to alcohol use Alcoholic hepatic failure Alcohol dependence syndrome Alcohol dependence syndrome Alcoholic cirrhosis of liver Homicide Homicide Homicide Homicide Homicide Car accident Car accident Larynx cancer Cancer on palate Suicide Accidental poisoning by narcotics Atherosclerotic heart disease Inhalation of object Other disorders of the intestine Cerebrovascular disease Diabetes mellitus End-stage renal disease pneumonia Pneumonia Renal failure HIV Pulmonary embolism Hepatic failure Congestive heart failure HIV HIV HIV HIV Unspecified disorder of kidney and urethra Atherosclerotic heart disease CancerdUnspecified End-stage renal disease Unknown Pulmonary embolism Septic emboli Chronic renal failure Cervical cancer HIV HIV HIV HIV Heart failure
ID
*Rivara, et al. (2004)
0.315 0.007
regression whereas birth weight was not (OR: 0.9995, P 5 .085) despite being statistically significant in the univariate analyses. Discussion These results confirm that drinking during pregnancy may be a marker for premature mortality among women. More than 12% of the women who drank during pregnancy were deceased at follow-up in 2003 compared to 3.6% of those who abstained and the mortality rate in the alcoholusing group was higher than the population rate for women of this age. Although drinking during pregnancy is recognized as a risk factor for offspring, these results suggest that the mothers’ own health is at risk and that those who drink during pregnancy are in need of intervention both to assure a healthy birth and to support the mother’s own long-term well being. The results are consistent with prior literature in that heavier drinking is associated with higher mortality rates (e.g., Rehm et al., 2001, 2003). Other important predictive factors included smoking during pregnancy, which is probably associated with continued smoking, and a family history of problem drinking by the participant’s own mother (grandmother). Although a family history of alcohol problems is established as a predictor of alcohol problems in offspring, the literature is conflicted concerning the relative impact of paternal versus maternal drinking patterns (see Gogineni et al., 2006). It is possible that maternal influence is more salient or influential than paternal in daughters’ socialization in this low-income sample and that this explains the impact of the grandmother’s drinking behavior. However, there may also be a biological explanation. The results suggest some association between biological markers of offspring health (birth weight and dysmorphia score) and an increased risk of mortality for the mother. Both birth weight and dysmorphia score were correlated with mortality and both significantly predicted mortality in univariate Cox regression models. Biological markers of alcohol exposure in offspring may indicate a reduced ability to cope with alcohol’s damaging effects in the mother. However, in multivariate analyses birth weight was no longer significant in predicting mortality in the presence of alcohol and cigarette use. This suggests that the association between birth weight and mortality was strongly mediated by alcohol and cigarette use. The factors that were not associated with mortality are notable. Variables normally associated with mortality such as age at delivery were not a factor in sample mortality over the follow-up despite the fact some women were over 40 years of age at delivery. Ethnicity, income, marital status, lifetime illicit drug use, partner alcohol use, and alcohol use/abuse by the woman’s own father (grandfather) were also not significantly correlated with mortality. These findings indicate the marked, detrimental effect of alcohol and cigarette use in an otherwise healthy sample.
J.P. Berg et al. / Alcohol 42 (2008) 603e610
609
Table 5 Predictors of mortality: results of Cox proportional hazard regression analyses (univariate and multivariate) Predictor
N
Univariate model; HR (95% CI)
Drank alcohol during pregnancy Drinking throughout vs. stopping by second trimester Smoked during pregnancy Offspring birth weight Maternal drinkingb Maternal heavy drinkingb Maternal trouble due to drinkingb Offspring dysmorphia scoreb
570 374 570 522 391 378 363 327
3.7** (1.7e8.2) 0.96 (0.52e1.80) 3.6***(1.8e7.4) 0.9995** (0.9992e0.9998) 2.1* (1.1e3.7) 2.3* (1.1e4.8) 4.0** (1.7e9.4) 1.06* (1.01e1.11)
Multivariate modela; HR (95% CI) 2.5* (1.0e6.3) 2.5* (1.1e5.5) 0.9995 (0.9992e0.9999)
HR 5 Hazard ratio, CI 5 Confidence interval. *P ! .05; **P ! .01; ***P ! .001. a In the multivariate model, only significant (P ! .05), nonexploratory predictors from the univariate analyses were included. b Exploratory variable due to missing data.
There are some limitations to this study. Mortality data include only deaths recorded in the state of Georgia. If a participant moved to another state and her death occurred there, it would not be recorded in the Georgia state death records. Thus, these results may be an underestimate of mortality. Access to national death records would make the record search more comprehensive. Another limitation results from the way in which the original data were collected. We were limited in our analysis to archival information and could only select variables for analysis from those collected at recruitment and when the infant was born. Thus, we would like to know more about family history of alcoholism and the participants’ medical history subsequent to this target pregnancy but the information is not available. Similarly, we would like to have current medical histories for the women who are still living to evaluate morbidity as well as mortality but cannot access that information. Finally, we believe that there are significant limitations in the way in which the alcohol etiological fraction is calculated and that the method used underestimates the contribution of alcohol use and abuse to mortality. For instance, HIV is listed as the primary cause of death for 10 women in the sample. It is likely that substance abuse is associated with a higher risk for HIV but we cannot calculate this likelihood. Similarly, subject number 458, whose cause of death is listed as ‘‘chronic renal failure’’ that is not attributable to alcohol using this system (Rivara et al., 2004), was known to us as an end-stage alcoholic. As previously noted, the mortality rate for this sample is quite high as compared to that for African-American women in Georgia. Participants in this sample were drawn from a low-income, urban sample and, therefore, compose a high-risk group. Incidence of deaths due to violence, HIV, and other causes associated with risks of living in an impoverished environment are probably elevated in this sample. In summary, the results of this 20-year follow-up show that use of alcohol during pregnancy is associated with premature mortality among women. Although it had been observed previously (Stratton et al., 1996) that mothers of children with FAS were at risk for premature mortality, the present study suggests that this risk may be more
general. Such behavior should be regarded by health care providers as a signal that intervention is needed to assist the mother in dealing with her drinking problems as well as other behaviors that may be placing her at risk. Intervention at this point could have long-term positive effects on the mother’s health. Acknowledgments We would like to thank Colin Smith, Sheila Butler, and Robert Attaway at the Georgia Department of Human Resources, Division of Public Health, Office of Health Information and Policy for their assistance in obtaining Georgia mortality and birth records. We would also like to thank the Fulton County and Dekalb County probate courts for their assistance in obtaining marriage records. References Armour, B. S., Woolery, T., Malarcher, A., Pechacek, T. F., and Husten, C. (2005). Annual smoking-attributable mortality, years of potential life lost, and productivity lossesdUnited States, 1997-2001. Morb. Mortal. Wkly. Rep. 54, 625–628. Coles, C. D., Brown, R. T., Smith, I. E., Platzman, K. A., Erickson, S., and Falek, A. (1991). Effects of prenatal alcohol exposure at school age: I. Physical and cognitive development. Neurotoxicol. Teratol. 13, 357– 367. Coles, C. D., Lynch, M. E., DeMarco, M., and Corley, T. C. (2003). Alcohol use during pregnancy and maternal mortality: an update. Alcohol. Clin. Exp. Res. 27, 375. Consumer Price Index Calculator (2008). Online calculator. Federal Reserve Bank of Minneapolis. Available at: http://woodrow.mpls.frb. fed.us/Research/data/us/calc. Accessed May 15, 2008. Cox, D. R. (1972). Regression models and life tables. J. R. Stat. Soc. Ser. B 34, 187–220. Curran, G. M., Stoltenberg, S. F., Hill, E. M., Mudd, S. A., Blow, F. C., and Zucker, R. A. (1999). Gender differences in the relationships among SES, family history of alcohol disorders and alcohol dependence. J. Stud. Alcohol. 60, 825–832. Eng, M. Y., Schuckit, M. A., and Smith, T. L. (2005). The level of response to alcohol in daughters of alcoholics and controls. Drug. Alcohol. Depend. 79, 83–93. Fellows, J. L., Trosclair, A., Adams, E. K., and Rivera, C. C. (2002). Annual smoking-attributable mortality, years of potential life lost, and economic costsdUnited States, 1995-1999. Morb. Mortal. Wkly. Rep. 51, 300–303.
610
J.P. Berg et al. / Alcohol 42 (2008) 603e610
Frezza, M., Di Padova, C., Pozzato, G., Terpin, M., Baraona, E., and Lieber, C. S. (1990). High blood alcohol levels in women: the role of decreased gastric alcohol dehydrogenase activity and first-pass metabolism. N. Engl. J. Med. 322, 95–99. Gogineni, A., King, S., Jackson, K., Kramer, J., Bucholz, K., Chan, G., et al. (2006). Female offspring of alcoholic individuals: recent findings on alcoholism and psychopathology risks: Symposium presented at the Research Society on Alcoholism, 2004, Vancouver, Aruna Gogineni, Chair. Alcohol. Clin. Exp. Res. 30, 377–387. Jones, K. L., and Smith, D. W. (1973). Recognition of the fetal alcohol syndrome in early infancy. Lancet 2, 999–1001. Lieber, C. S. (1997). Gender differences in alcohol metabolism and susceptibility. In R. W. Wilsnack, & S. C. Wilsnack (Eds.), Gender and Alcohol: Individual and Social Perspectives (pp. 77–89). New Brunswick, NJ: Rutgers Center of Alcohol Studies. May, P. A., Hymbaugh, K. J., Aase, J. M., and Samet, J. M. (1983). Epidemiology of fetal alcohol syndrome among American Indians of the Southwest. Soc. Biol. 30, 374–387. Mena, M., Casanueva, V., Fernandez, E., Carraso, R., and Perez, H. (1986). Fetal alcohol syndrome at schools for mentally handicapped children in Concepcio´n. Chile. PAHO Bull. 20, 157–169. Mokdad, A. H., Marks, J. S., Stroup, D. F., and Gerberding, J. L. (2004). Actual causes of death in the United States, 2000. J. Am. Med. Assoc. 291, 1238–1245.
Neumark, Y. D., Van Etten, M. L., and Anthony, J. C. (2000). ‘‘Alcohol dependence’’ and death: survival analysis of the Baltimore ECA sample from 1981 to 1995. Subst. Use. Misuse. 35, 533–549. Rehm, J., Gmel, G., Sempos, C. T., and Trevisan, M. (2003). Alcoholrelated morbidity and mortality. Alcohol. Res. Health. 27, 39–51. Rehm, J., Gutjahr, E., and Gmel, G. (2001). Alcohol and all-cause mortality: a pooled analysis. Contemp. Drug. Probl. 28, 337–361. Rivara, F. P., Garrison, M. M., Ebel, B., McCarty, C. A., and Christakis, D. A. (2004). Mortality attributable to harmful drinking in the United States, 2000. J. Stud. Alcohol. 65, 530–536. Smith, I. E., Lancaster, J. S., Moss-Wells, S., Coles, C. D., and Falek, A. (1987). Identifying high-risk pregnant drinkers: biological and behavioral correlates of continuous heavy drinking during pregnancy. J. Stud. Alcohol. 48, 304–309. Spohr, H. L., Willms, J., and Steinhausen, H. C. (1993). Prenatal alcohol exposure and long-term developmental consequences. Lancet 341, 907–910. SPSS (2004). Computer Program. Version 13.0. Chicago, IL: SPSS Corp. Stinson, F. S., and DeBakey, S. F. (1992). Alcohol-related mortality in the United States, 1979-1988. Br. J. Addict. 87, 777–783. Stratton, K., Howe, C., and Battaglia, F. (1996). Fetal Alcohol Syndrome: Diagnosis, Epidemiology, Prevention, and Treatment. Washington, DC: National Academy Press. WONDER [online database] (2004). (Atlanta, GA: Center for Disease Control). Available at: http://wonder.cdc.gov. Updated March 2004.