American Journal of Epidemiology Vol. 159, No.

2
Copyright © 2004 by the Johns Hopkins Bloomberg School of Public Health Printed in U.S.A.
All rights reserved DOI: 10.1093/aje/kwh034

Alcohol Consumption during Pregnancy and the Risk of Preterm Delivery

Katrine Albertsen1,2, Anne-Marie Nybo Andersen3, Jørn Olsen4, and Morten Grønbæk1

1 Centre for Alcohol Research, National Institute of Public Health, Copenhagen, Denmark.
2 Danish Epidemiology Science Centre, Institute of Preventive Medicine, Copenhagen University Hospital, Copenhagen,
Denmark.
3 Department of Social Medicine, Institute of Public Health, University of Copenhagen, Copenhagen, Denmark.
4 Danish Epidemiology Science Centre, Department of Epidemiology and Social Medicine, University of Aarhus, Aarhus,

Denmark.

Received for publication April 15, 2003; accepted for publication July 31, 2003.

The authors evaluated the association between amount and type of alcohol consumed during pregnancy and
the risk of preterm delivery and whether the relation differs among very (<32 completed weeks) and moderate
(from 32 to <37 completed weeks) preterm delivery. The study is based on data of 40,892 pregnant women
included in the first part of the Danish National Birth Cohort. The women completed a computer-assisted
telephone interview between December 12, 1997, and December 31, 2000, and delivered a liveborn singleton.
Of these women, 1,880 gave birth preterm. Compared with those who abstained during pregnancy, the relative
risks for preterm delivery among women who consumed from four to less than seven drinks and seven or more
drinks per week during pregnancy were 1.15 (95% confidence interval: 0.84, 1.57) and 1.77 (95% confidence
interval: 0.94, 3.31), respectively. Below these intake levels of alcohol, no increased risk of preterm delivery was
found. Among women who consumed seven or more drinks per week, the relative risk of very preterm delivery
was 3.26 (95% confidence interval: 0.80, 13.24) compared with that of nondrinkers. There were no differences in
the associations between type of beverage and preterm delivery.

alcohol drinking; alcoholic beverages; delivery, obstetric; labor, premature; pregnancy

Abbreviation: CI, confidence interval.

Alcohol in high doses is known to have a teratogenic
effect. Whether a low intake of alcohol is teratogenic is not
known, and it remains controversial whether there exists a
safe level of drinking during pregnancy (1). Preterm birth is
one of the main causes of neonatal mortality, neonatal
morbidity, and functional impairments (2, 3). Previous
studies on alcohol consumption during pregnancy and
preterm delivery have revealed conflicting results. An
increased risk has been reported even at an intake of less than
two drinks per week in late pregnancy (4), while others
report increased risk at intake levels around 10–14 drinks per
week (5, 6). A J-shaped relation between alcohol consump-
tion and risk of preterm delivery has been reported from two
prospective studies (6, 7), but only one of these studies found
a threshold for adverse effect at 10 drinks per week, while
the other found no significant increased risk even at a weekly
intake of 21 or more drinks of alcohol. Other studies have
had too small a sample size or used too broad exposure cate-
gories to find any association between alcohol consumption
and preterm delivery (8, 9). Possible explanations for the
lack of agreement between studies include low statistical
power and inadequate or biased information on maternal
alcohol consumption.
If a low consumption of alcohol has an impact on the risk
of preterm delivery, it has public health importance, since a
low level of alcohol consumption during pregnancy is wide-
spread in many countries. It is also of importance to evaluate
if alcohol consumption is related to the degree of preterm
delivery, since very preterm deliveries (those occurring
before 32 weeks) account for the majority of early neonatal
deaths (10). We further find it of interest to investigate the
differences in associations between type of alcohol and

Reprint requests to Katrine Albertsen, Centre for Alcohol Research, National Institute of Public Health, Svanemøllevej 25, 2100 Copenhagen
Ø, Denmark (e-mail: kal@niph.dk).

155 Am J Epidemiol 2004;159:155–161

156 Albertsen et al.
preterm delivery. A difference would indicate that the
ethanol part of alcohol is not the only exposure of interest.
Other substances in beer, wine, or spirits may be involved in
the causal pathway to preterm delivery.
The aims of this study are therefore to examine the associ-
ation between the amount and type of alcohol consumption
and the risk of preterm delivery (before 37 weeks), moderate
preterm delivery (32 to before 37 weeks), and very preterm
delivery (before 32 weeks).
MATERIALS AND METHODS
Study design and population
The study was carried out within the Danish National Birth
Cohort, which is an ongoing nationwide study of pregnant
women and their offspring. The pregnant women received
written information about the cohort at the first antenatal
visit to the general practitioner in gestational weeks 6–10
and were included when an informed consent form was
registered. About half of all general practitioners in
Denmark participated in the study, and it is estimated that
around 60 percent of all women invited chose to participate.
Approximately 35 percent of all pregnant women in
Denmark are recruited into the cohort. Besides being preg-
nant, the criteria for inclusion in the cohort were as follows:
1) a permanent address in Denmark, 2) intention to carry the
pregnancy to term, and 3) the woman spoke Danish well
enough to participate in four telephone interviews during
pregnancy and early motherhood. Participants contribute
with information on exposures during the pregnancy by
means of computer-assisted telephone interviews and blood
samples. The first pregnancy interview is scheduled to take
place between 12 and 16 weeks of gestation (11).
In the present study, we used data from 41,847 pregnant
women who had given their first pregnancy telephone inter-
view within the time period from December 12, 1997, to
December 31, 2000, and whose pregnancy resulted in live-
born singletons. A total of 955 pregnancies were excluded
from the analysis for the following reasons: missing or
invalid information on gestational age at delivery (n = 60),
missing data on alcohol consumption during or before preg-
nancy (n = 249), or missing data on the covariates included
in multivariate models (n = 82). Women who participated
with more than one pregnancy (n = 558) were included only
for their first pregnancy, fulfilling the above-mentioned
inclusion criteria. Women who were interviewed later than
37 weeks of gestation were excluded (n = 6), since they were
not at risk of preterm delivery. Altogether, 40,892 (97.7
percent) women were eligible for analyses.
Exposures measurement
All women were asked to report their average weekly
intake of beer, wine, and spirits during their pregnancy. The
question was as follows: “How many ordinary beers do you
drink per week?” The same question was asked for “glasses
of wine” and “glasses of spirits.” We assumed that the
alcohol content per drink was roughly equal for the three
types of drinks; one bottle of beer contains 11.6 g of alcohol,
and 12 g of alcohol are equal to one-sixth bottle of wine or
40 ml of spirits, which are approximate averages for one
serving in Denmark. The total weekly alcohol consumption
was calculated by adding the intakes of beer, wine, and
spirits. Women who reported some intake of beer, wine, and/
or spirits but less than one drink per week are assigned a
numeric value of 0.5 drink. Alcohol consumption during
pregnancy is categorized into nondrinkers, one-half drink
per week, 1–1.5 drinks per week, 2–3.5 drinks per week, 4–
6.5 drinks per week, and seven or more drinks per week. For
the subsequent analysis of an association between type of
alcohol and preterm delivery, we categorized the women
into preference for type of alcohol. Preference for one type
of alcohol was defined as beer, wine, or spirits if the intake
of this type exceeded 50 percent of the woman’s total alcohol
intake. Women who drank alcohol but did not have a
preferred alcoholic beverage were classified as mixed
drinkers. Women who abstained from drinking during preg-
nancy were classified in a separate category of nondrinkers.
Measurement of potential confounders
Information on the main part of potential confounders was
obtained by self-reports in the telephone-based interview.
Information on diabetes status was obtained from the National
Discharge Register. The following known or suspected deter-
minants of preterm delivery were assumed to potentially
confound the analyses: number of spontaneous abortions (n =
0, 1, ≥2), prior preterm delivery (nulliparous, no, yes), type 1
diabetes (no, yes), hypertension (present, present only during
pregnancy, no hypertension), bleeding episodes during preg-
nancy (yes or no), occupational status in the household
defined as the highest attained occupational status of the
woman or partner (higher-grade professionals, lower-grade
professionals, skilled workers, unskilled workers, students,
economically inactive, unclassifiable), maternal age (<20, 20–
24, 25–29, 30–34, ≥35 years), parity (n = 0, 1, ≥2 births),
maternal height (<160, 160–164, 165–169, 170–174, ≥175
cm), prepregnancy weight (<50, 50–59, 60–69, 70–79, ≥80
kg), gender of the infant (male, female), coffee consumption
(0, 1–5, ≥6 cups per day; 1 cup = approximately 239 ml,
depending on cup size and coffee strength), and smoking
habits. Smoking habits were reported as the number of ciga-
rettes (1 g), small cigars (3 g), cigars (5 g), or pipes (3 g)
smoked a day, which were categorized into nonsmoking,
stopped during pregnancy, 1–10 g, and ≥11 g per day.
Outcome measurement
The outcome measure of interest was gestational age at
delivery based on information from the National Discharge
Register, which comprises information on all deliveries in
Denmark. Gestational age in the National Discharge Register
is recorded in terms of completed weeks and days. National
guidelines regarding the assessment of gestational age are as
follows: either gestational age estimated from the first day of
the last menstrual period (preconditioned, a regular bleeding
pattern during the last 6 months, and no use of oral contracep-
tives during the last 3 months before pregnancy) or estimation
from ultrasound examination before 24 weeks of gestation
Am J Epidemiol 2004;159:155–161
 Alcohol, Pregnancy, and Preterm Delivery 157

TABLE 1. Study population characteristics according to alcohol intake during pregnancy, Denmark, 1997–2000

Very preterm Moderate preterm Prior Mean

Amount No Higher- Diabetes
delivery delivery Above Non- Nulliparous preterm alcohol
(drinks/week) Subjects coffee grade type 1
35 years smoker women delivery intake
and type (no.) intake professionals (per
No. % No. % (%) (%) (%) (per (drinks/
of alcohol (%) (%) thousand)
thousand) week)
Amount of
alcohol
Nondrinkers 22,333 115 0.5 970 4.3 9 73 64 48 18 6.9 2.2 0.0
0.5 6,537 30 0.5 259 4.0 10 76 52 44 22 6.3 3.5 0.5
1–1.5 7,130 43 0.6 259 3.6 12 75 43 42 25 4.6 2.1 1.1
2–3.5 4,004 21 0.5 131 3.3 18 71 35 39 30 6.7 1.7 2.3
4–6.5 769 6 0.8 36 4.7 28 62 25 32 31 10.4 1.3 4.6
≥7 119 2 1.7 8 6.7 33 47 24 29 22 0.0 0.0 8.7
Preference of
type
Beer 2,136 17 0.8 84 3.9 16 65 37 36 19 3.3 3.3 1.5
Wine 13,152 66 0.5 487 3.7 13 75 45 43 27 6.2 2.3 1.3
Spirits 175 0 0.0 9 5.1 10 66 67 51 11 0.0 5.7 0.8
Mixed 3,096 19 0.6 113 3.7 14 74 38 38 22 6.5 2.6 1.6

(12). Preterm delivery was defined as a gestational age of less
than 37 weeks (259 days) of gestation and subdivided into
degree of preterm: moderate preterm (32 to before 37 weeks)
and very preterm delivery (before 32 weeks).
Statistical analyses
The relative risks of preterm delivery according to the
amount of alcohol consumed and the preference for type of
alcohol during pregnancy were estimated using the Cox
regression model. Since the time of interview during preg-
nancy and the risk of preterm delivery vary during gestation,
careful adjustment for gestational age was needed. There-
fore, gestational age in days was used as an underlying time
variable. We used the model with delayed entry, so women
entered the cohort on the day of interview, and follow-up
ended at delivery or after 258 days (36 weeks and 6 days) of
gestation, whichever came first. Deliveries occurring after
258 days were censored at that time.
We conducted two sets of analyses: one estimating the rela-
tive risk estimates of preterm delivery according to total
alcohol intake without considering type of beverage, and
another estimating the relative risks of preterm delivery
according to preference for type of alcohol adjusting for the
total consumption. The reference group in the analyses of total
alcohol consumption was nondrinkers. The reference group in
the analyses of type of alcohol was women who preferred
wine and had a total alcohol consumption of 1–1.5 drink(s) per
week. Since the risk of repeating a preterm birth is high and
since a previous preterm birth or other recognized reproduc-
tive problems might lead to reduced alcohol consumption, we
furthermore analyzed data for nulliparous women only.
Subsequently, we estimated the influence of alcohol on very
and moderate preterm delivery, respectively, by including an
interaction between alcohol and a variable indicating the
degree of preterm delivery. This procedure corresponds to
making separate analyses of very and moderate preterm
delivery, except that common estimates of the influences of
each of the covariates other than alcohol are obtained.
The change-in-estimate method was used to assess which
of the potential confounders actually did confound the anal-
yses (13). Variables that were equally distributed according
to consumption levels of the amount and type of alcohol and
that were not evident from the literature to be associated with
alcohol consumption were removed one by one from the full
model, if they did not change the estimates between alcohol
intake and preterm birth by more than 5 percent.
RESULTS
The mean gestational age at delivery was 279 days (range,
158–315 days), and the overall proportion of preterm delivery
was 4.6 percent (1,880/40,892) and of the preterm deliveries
the proportion of very preterm delivery was 12 percent (217/
1,880). Fifty-five percent of the women reported that they
abstained from alcohol consumption in their first trimester.
Women who had a high alcohol intake during pregnancy were
more often above 35 years of age, smokers, and coffee
consumers and less often nulliparous than were women who
reported abstaining from alcohol (table 1). Among women
who consumed any amount of alcohol, most (70.9 percent)
preferred wine. Less than 1 percent of the drinking women
preferred spirits, while 11.5 percent of the drinking women
preferred beer. Women who preferred wine were more often
nonsmokers and from higher-grade professional households
than were women who preferred beer or spirits. Women who
preferred beer were more often above 35 years of age and
more often coffee consumers than were women who preferred
wine or spirits. The few women who preferred spirits were
less often above 35 years of age and more often nulliparous,
and they had a lower mean alcohol consumption than did
women who preferred beer or spirits (table 1).

Am J Epidemiol 2004;159:155–161
158 Albertsen et al.

TABLE 2. Unadjusted and adjusted relative risks and 95% confidence intervals for preterm
delivery, according to weekly alcohol consumption during pregnancy, Denmark, 1997–2000

Delivery before 37 weeks

Alcohol Nulliparous women (n = 18,349) All women (n = 40,892)

(drinks/week) Adjusted† Adjusted‡
Unadjusted Unadjusted
RR* RR 95% CI* RR RR 95% CI
Nondrinkers§ 1.00 1.00 1.00 1.00
0.5 0.98 0.97 0.82, 1.15 0.91 0.92 0.81, 1.05
1–1.5 0.95 0.94 0.79, 1.12 0.87 0.91 0.80, 1.04
2–3.5 0.95 0.94 0.75, 1.18 0.78 0.80 0.68, 0.96
4–6.5 1.44 1.41 0.92, 2.19 1.13 1.15 0.84, 1.57
≥7 3.09 2.91 1.29, 6.55 1.79 1.77 0.94, 3.31

* RR, relative risk; CI, confidence interval.

† Adjusted for type 1 diabetes, age, smoking during pregnancy, coffee consumption during
pregnancy, and occupational status in the household.
‡ Adjusted for type 1 diabetes, age, previous preterm delivery, smoking during pregnancy, coffee
consumption during pregnancy, occupational status in the household, and parity.
§ Reference group.

Amount and type of alcohol and risk of preterm delivery
Alcohol consumption below four drinks per week was
associated with risk estimates below unity. Women who had
an intake between two and four drinks per week had a lower
risk of preterm delivery, a relative risk of 0.80 (95 percent
confidence interval (CI): 0.68, 0.96) compared with
nondrinkers. Women who had alcohol consumption of
between four and seven drinks per week had a relative risk of
preterm delivery of 1.15 (95 percent CI: 0.84, 1.57) compared
with nondrinkers. Likewise, women who had a consumption
of seven or more drinks per week had a relative risk of 1.77
(95 percent CI: 0.94, 3.31) compared with nondrinkers (table
2). Adjustment for diabetes status, age, previous preterm
delivery, parity, smoking, coffee consumption, and occupa-
tional status did not affect the results (table 2).
The estimates from the analysis restricted to nulliparous
women only were in line with the estimates based on the entire
population. However, the apparent protective effect of an
alcohol consumption between two and four drinks per week in
the analysis based on the entire population diminished and was
no longer statistically significant, when the analysis is restricted
to nulliparous women only (table 2). No significant interaction
between alcohol consumption and parity was detectable.
In the analysis of preference of type of alcohol, no differ-
ences in risk were seen. All the estimates were close to 1.00.
Only very few women preferred spirits, so the estimate for
spirits is based on only 175 women. As seen from table 3,

TABLE 3. Unadjusted and adjusted relative risks and 95 percent confidence

intervals for preterm delivery, according to preferred type of alcohol during
pregnancy, Denmark, 1997–2000

Delivery before 37 weeks

Preference for Nulliparous women (n = 18,349) All women (n = 40,892)

type of alcohol Adjusted† Adjusted‡
Unadjusted Unadjusted
RR* RR 95% CI* RR RR 95% CI
Nondrinkers 1.10 1.10 0.91, 1.33 1.18 1.12 0.97, 1.30
Wine§ 1.00 1.00 1.00 1.00
Beer 1.13 1.06 0.78, 1.43 1.12 1.10 0.89, 1.36
Spirits 1.15 1.09 0.48, 2.45 1.19 1.10 0.57, 2.13
Mixed 1.13 1.10 0.84, 1.44 1.03 1.03 0.84, 1.26

* RR, relative risk; CI, Confidence interval.

† Adjusted for type 1 diabetes, age, smoking during pregnancy, coffee consumption
during pregnancy, occupational status in the household, and total alcohol consumption
during pregnancy.
‡ Adjusted for type 1 diabetes, age, previous preterm delivery, smoking during
pregnancy, coffee consumption during pregnancy, occupational status in the household,
parity, and total alcohol consumption during pregnancy.
§ Reference group.

Am J Epidemiol 2004;159:155–161
 Alcohol, Pregnancy, and Preterm Delivery 159

TABLE 4. Unadjusted and adjusted relative risks and 95 percent confidence intervals for
very preterm and moderate preterm delivery, according to weekly alcohol consumption
during pregnancy, Denmark, 1997–2000

Before 32 weeks From 32 to before 37 weeks

Alcohol
Unadjusted Adjusted† Unadjusted Adjusted†
(drinks/week)
RR* RR 95% CI* RR RR 95% CI
Nondrinkers‡ 1.00 1.00 1.00 1.00
0.5 0.89 0.91 0.61, 1.35 0.91 0.93 0.81, 1.06
1–1.5 1.17 1.24 0.87, 1.76 0.83 0.87 0.76, 1.00
2–3.5 1.02 1.06 0.66, 1.69 0.75 0.77 0.64, 0.93
4–6.5 1.52 1.53 0.67, 3.49 1.08 1.10 0.79, 1.54
≥7 3.32 3.26 0.80, 13.24 1.60 1.58 0.79, 3.19

* RR, relative risk; CI, confidence interval.

† Adjusted for type 1 diabetes, age, previous preterm delivery, parity, smoking during pregnancy,
coffee consumption during pregnancy, and occupational status in the household.
‡ Reference group.

adjustments for diabetes status, age, previous preterm delivery,
parity, smoking, coffee consumption, and occupational status
did not meaningfully alter the results. The estimates were
changed negligibly when the analysis was restricted to nullipa-
rous women only (table 3).
Amount of alcohol and risk of very and moderate
preterm delivery
An alcohol intake of between four and less than seven and
of seven or more drinks per week was associated with a rela-
tive risk for very preterm delivery of 1.53 (95 percent CI:
0.67, 3.49) and 3.26 (95 percent CI: 0.80, 13.24), respec-
tively, compared with that of nondrinkers (table 4). In
general, the relative risks, but not risk difference, for very
preterm delivery were higher than the relative risks for
moderate preterm delivery. Adjustment for diabetes status,
age, previous preterm delivery, parity, smoking, coffee
consumption, and occupational status did not affect the
results (table 4). Restricting the analysis to nulliparous
women only did not alter the results (results not shown), but
this analysis was hampered by an even lower number of very
preterm deliveries.
DISCUSSION
This study on birth outcomes of more than 40,000 preg-
nancies showed that an alcohol consumption of seven or
more drinks per week during pregnancy was associated with
a slightly increased risk of preterm delivery. We found no
increased risk of preterm delivery among women with
alcohol consumption less than four drinks per week. There
was no obvious difference in the effects of intakes of beer,
wine, or spirits on preterm delivery. Examining the influence
of alcohol on risk of very preterm and moderate preterm
delivery revealed results similar to those for preterm
delivery, the risk estimates being even higher for very
preterm delivery.
A J-shaped relation between alcohol consumption and
preterm delivery has been found in two other cohort studies
(6, 7). However, the increase in risk was found at different
intake levels. Kesmodel et al. (6) reported a threshold for
adverse effect at 10 or more drinks per week, while
McDonald et al. (7) reported no significant increase in the
risk of preterm delivery at a weekly intake of 21 or more
drinks. Our results also indicate a J-shaped relation between
alcohol consumption during pregnancy and the risk of
preterm delivery, since women drinking between two and
four drinks per week have a significantly decreased risk of
preterm delivery. However, when the population was
restricted to nulliparous women, the apparent protective
effect of alcohol consumption below four drinks per week
diminished, indicating that the apparent protective effect
could be caused by uncontrolled confounding, arising from
behavioral changes regarding alcohol consumption related to
previous reproductive experience.
In addition, our results show that the risk of preterm
delivery increases at even lower levels of alcohol consump-
tion. This is in agreement with the findings by Shiono et al.
(14), who reported an increased risk at an intake level of one
or more drinks per day in a large prospective study. Lunds-
berg et al. (4) reported an increased risk of preterm delivery
at an intake level equivalent to less than two drinks per week
in late pregnancy; however, their study showed no associa-
tion with drinking early in pregnancy. However, other
studies have not found any association between alcohol
consumption during pregnancy and risk of preterm delivery
(8, 9). Our findings of higher relative risk estimates for very
preterm than moderate preterm delivery corroborate two
smaller studies subdividing preterm delivery. Both of these
studies categorized alcohol consumption into only drinkers
and nondrinkers (15, 16). Shiono et al. (14) also investigated
the association between alcohol consumption and very
preterm delivery, defined as birth before 33 weeks. In this
study, preterm delivery was not subdivided into very and
moderate preterm delivery but was defined as very preterm
(before 33 weeks) or preterm (before 37 weeks). The relative

Am J Epidemiol 2004;159:155–161
160 Albertsen et al.
risk for very preterm delivery is lower than what we found in
this study. We are aware of only two smaller studies having
investigated the association between type of beverage and
preterm delivery (4, 17). Our findings of no obvious differ-
ences in associations between the type of beverage and
preterm delivery are in accordance with their results.
Possible explanations for the lack of agreement across
studies investigating the association between alcohol
consumption and preterm delivery include unequal charac-
terization and timing of alcohol consumption, low statistical
power for addressing small effects, differences in methods of
obtaining the measure of exposure, and differences in adjust-
ments of the confounding variables.
This study was based on a large population with almost
complete follow-up. The number of women with missing
information on alcohol consumption and covariates was low,
because of the use of computer-assisted telephone interviews
in data collection. This study is the largest so far regarding
alcohol consumption and preterm delivery, and it is one of
the few investigating the risk of very preterm delivery
according to different consumption levels of alcohol.
Although this study included a large number of participants,
few events hampered the analysis of an association between
alcohol consumption and very preterm delivery.
A possible limitation of this study is the participation rate
of only 60 percent of those women invited to the Danish
National Birth Cohort. In order to introduce bias, the deci-
sion to participate should be selective regarding both the
alcohol consumption and the risk of preterm delivery. It is
unlikely that the risk of preterm delivery could play a role
since women did not report this information. It is unlikely
that the decision to participate is especially influenced by
alcohol consumption, since the Danish National Birth
Cohort is not directly aimed at examining health conse-
quences of alcohol intake. We therefore find it unlikely that
the small percentage of participants biased the results
considerably.
All studies on alcohol consumption in humans, including
this study, rely on self-reported data, which are susceptible
to information bias, almost certainly in the form of under-
reporting. However, a methodological study showed that, in
the absence of overreporting, even considerable under-
reporting seems to have little impact on the association
between exposure and outcome, especially when the number
of nonexposed is large (18). For these data, the interviewer
effect (arising from variations in interviewers’ health beliefs
and personal habits) was found to be negligible in this meth-
odological study (19). Information on alcohol intake during
pregnancy was obtained prospectively, so that differential
recall bias is thus unlikely. Information on alcohol intake
during pregnancy was obtained only once in the first half of
pregnancy, and the reported value was used as an indicator
for the general level of exposure during pregnancy. This may
imply misclassification for women who change their alcohol
habits during pregnancy, which would, if anything, affect the
results toward the null value. The exposure measure is not
time specific, because the range in time of interviewing
varied; however, it is not known how the timing of alcohol
might play a role in the induction of preterm delivery. The
use of average measures of alcohol exposure masks varia-
tions in the drinking pattern, and it is possible that periodic
high alcohol intakes (binge episode) have a different effect
on the fetus than an alcohol intake that does not reach the
same blood alcohol levels. There may be different explana-
tions for the apparent protective effect of alcohol consump-
tion of between two and four drinks per week. First, it may
be due to a healthy drinker effect, for example, that healthy
women generally tend to drink more than women with
recognized reproductive problems. The lack of a protective
effect in the analysis restricted to nulliparous women
supports this explanation. Second, the reference group of
nondrinkers consists of two distinct groups of women:
women who decrease alcohol intake during pregnancy and
women who abstain both prior to and during pregnancy. In
accordance with the results from Passaro et al. (20), the
results from our study found that women who abstained from
alcohol prior to pregnancy had a higher risk of preterm
delivery compared with women drinking some amounts of
alcohol (data not shown). In addition, the group of women
who abstained from alcohol prior to and during pregnancy
had the highest risk of preterm delivery (data not shown).
The gestational age at delivery was obtained through
register linkage, and most of these measures are based upon
ultrasound examinations. The use of gestational age deter-
mined from the first day of the last menstrual period as the
only source of calculation of gestational age tends to over-
estimate the gestational age for those with long menstrual
cycles, thereby underestimating the proportion of preterm
delivery (21). The proportion of preterm delivery in this
study is lower than those in many other studies, but it is
comparable with the proportion of preterm delivery in
Denmark (22).
The suggested pathophysiologic mechanism for alcohol to
induce preterm delivery is related to an increase in the
production of prostaglandins in relation to alcohol intake.
Increased secretion of prostaglandins has been found in alco-
holic mothers and their offspring (23). Alcohol has, on the
other hand, been used clinically to avoid premature labor in
the late 1960s and during the 1970s. This alleged tocolytic
effect of alcohol has, however, not been supported by
double-blinded clinical trials (24).
No difference in the associations between types of alcohol
and preterm delivery implies that ethanol probably is
accountable in the causal pathway between alcohol
consumption and preterm delivery.
In conclusion, alcohol consumption below four drinks per
week does not increase the risk of preterm delivery, while a
daily alcohol intake during pregnancy seems to increase the
risk of preterm delivery. The alcohol-related risk of preterm
delivery is independent of type of alcohol consumed.
ACKNOWLEDGMENTS
This particular study is funded by grants from the Danish
National Board of Health and the Health Insurance Founda-
tion. The Danish National Research Foundation has estab-
lished the Danish Epidemiology Science Centre that initiated
and created the Danish National Birth Cohort. The cohort is
Am J Epidemiol 2004;159:155–161

furthermore a result of a major grant from this foundation.
Additional support for the Danish National Birth Cohort is
obtained from the Pharmacy Foundation, the Egmont Foun-
dation, the March of Dimes Birth Defects Foundation, and
the Augustinus Foundation.
The authors thank Ditte Johansen and Per Kragh Andersen
at the Danish Epidemiology Science Centre for their valu-
able comments on the methodology used in this study.
REFERENCES
1. Kesmodel U. Alcohol in pregnancy. (In Danish). Ugeskr
Laeger 1999;161:4989–94.
2. Kaushik SL, Parmar VR, Grover N, et al. Neonatal mortality
rate: relationship to birth weight and gestational age. Indian J
Pediatr 1998;65:429–33.
3. Robertson PA, Sniderman SH, Laros RK Jr, et al. Neonatal
morbidity according to gestational age and birth weight from
five tertiary care centers in the United States, 1983 through
1986. Am J Obstet Gynecol 1992;166:1629–41.
4. Lundsberg LS, Bracken MB, Saftlas AF. Low-to-moderate ges-
tational alcohol use and intrauterine growth retardation, low
birthweight, and preterm delivery. Ann Epidemiol 1997;7:498–
508.
5. Berkowitz GS, Holford TR, Berkowitz RL. Effects of cigarette
smoking, alcohol, coffee and tea consumption on preterm deliv-
ery. Early Hum Dev 1982;7:239–50.
6. Kesmodel U, Olsen SF, Secher NJ. Does alcohol increase the
risk of preterm delivery? Epidemiology 2000;11:512–18.
7. McDonald AD, Armstrong BG, Sloan M. Cigarette, alcohol,
and coffee consumption and prematurity. Am J Public Health
1992;82:87–90.
8. Verkerk PH, van Noord Zaadstra BM, Florey CD, et al. The
effect of moderate maternal alcohol consumption on birth
weight and gestational age in a low risk population. Early Hum
Dev 1993;32:121–9.
9. Marbury MC, Linn S, Monson R, et al. The association of alco-
hol consumption with outcome of pregnancy. Am J Public
Health 1983;73:1165–8.
10. Berkowitz GS, Papiernik E. Epidemiology of preterm birth.
Epidemiol Rev 1993;15:414–43.
11. Olsen J, Melbye M, Olsen SF, et al. The Danish National Birth
Am J Epidemiol 2004;159:155–161
Alcohol, Pregnancy, and Preterm Delivery 161
Cohort—its background, structure and aim. Scand J Public
Health 2001;29:300–7.
12. Danish Association of Obstetrics and Gynecology (DSOG).
Obstetric diagnoses and operations. Guide to classification and
definitions. 1st ed. (In Danish). Copenhagen, Denmark: DSOG,
1996.
13. Greenland S. Modeling and variable selection in epidemiologic
analysis. Am J Public Health 1989;79:340–9.
14. Shiono PH, Klebanoff MA, Rhoads GG. Smoking and drinking
during pregnancy. Their effects on preterm birth. JAMA 1986;
255:82–4.
15. Kramer MS, McLean FH, Eason EL, et al. Maternal nutrition
and spontaneous preterm birth. Am J Epidemiol 1992;136:574–
83.
16. Adams MM, Sarno AP, Harlass FE, et al. Risk factors for pre-
term delivery in a healthy cohort. Epidemiology 1995;6:525–
32.
17. Kaminski M, Franc M, Lebouvier M, et al. Moderate alcohol
use and pregnancy outcome. Neurobehav Toxicol Teratol 1981;
3:173–81.
18. Verkerk PH. The impact of alcohol misclassification on the
relationship between alcohol and pregnancy outcome. Int J Epi-
demiol 1992;21(suppl 1):33–7.
19. Andersen AMN, Olsen J. Do interviewers’ health beliefs and
habits modify responses to sensitive questions? A study using
data collected from pregnant women by means of computer-
assisted telephone interviews. Am J Epidemiol 2002;155:95–
100.
20. Passaro KT, Little RE, Savitz DA, et al. The effect of maternal
drinking before conception and in early pregnancy on infant
birthweight. Epidemiology 1996;7:377–83.
21. Hall MH, Carr-Hill RA, Fraser C, et al. The extent and anteced-
ents of uncertain gestation. Br J Obstet Gynecol 1985;92:445–
51.
22. Danish National Board of Health. Medical birth and malforma-
tion statistics 1994 and 1995. Copenhagen, Denmark: Vital Sta-
tistics, 1997:6.
23. Ylikorkala O, Halmesmaki E, Viinikka L. Urinary prostacyclin
and thromboxane metabolites in drinking pregnant women and
in their infants: relations to the fetal alcohol effects. Obstet
Gynecol 1988;71:61–6.
24. Castren O, Gummerus M, Saarikoski S. Treatment of imminent
premature labour. Acta Obstet Gynecol Scand 1975;54:95–100.