ORIGINAL ARTICLE

Caffeine and Miscarriage Risk

David A. Savitz,* Ronna L. Chan, Amy H. Herring,
Penelope P. Howards, and Katherine E. Hartmann
Background: Coffee and caffeine have been inconsistently found to
be associated with increased risk of clinical miscarriagea potentially important association given the high prevalence of exposure.
Methods: Women were recruited before or early in pregnancy and
interviewed regarding sources of caffeine, including assessment of
changes over the perinatal period. We identified 2407 clinicallyrecognized pregnancies resulting in 258 pregnancy losses. We examined the relationship of coffee and caffeine intake with clinicallyrecognized pregnancy loss prior to 20 weeks completed gestation,
using a discrete-time continuation ratio logistic survival model.
Results: Coffee and caffeine consumption at all 3 time points were
unrelated to total miscarriage risk and the risk of loss after the
interview. Reported exposure at the time of the interview was
associated with increased risk among those with losses before the
interview.
Conclusions: There is little indication of possible harmful effects of
caffeine on miscarriage risk within the range of coffee and caffeine
consumption reported, with a suggested reporting bias among
women with losses before the interview. The results may reflect
exposure misclassification and unmeasured heterogeneity of pregnancy losses.
(Epidemiology 2008;19: 55 62)

he effect of caffeine on fetal survival has been of interest

since the 1980s.1 The biologic rationale for this concern
includes caffeines ability to cross the placenta,2 the mothers
decreased ability to metabolize caffeine later but not earlier in

Submitted 24 January 2007; accepted 14 August 2007.

From the *Department of Community and Preventive Medicine, Mount Sinai
School of Medicine, New York, New York; Departments of Epidemiology and Biostatistics, University of North Carolina School of Public
Health, Chapel Hill, North Carolina; Division of Epidemiology, Statistics and Prevention Research, National Institute of Child Health and
Human Development, Bethesda, Maryland; and Department of Obstetrics and Gynecology, Vanderbilt University School of Medicine, Nashville, Tennessee.
Supported by the American Water Works Association Research Foundation
under Contract 2579 and in part by the Intramural Research Program of
the NIH, National Institute of Child Health and Human Development,
and the National Institute of Environmental Health Sciences (training
grant ES07018).
Correspondence: David A. Savitz, Department of Community and Preventive
Medicine, Mount Sinai School of Medicine, One Gustave L. Levy Place,
Box 1057, New York, NY 10029. E-mail: david.savitz@mssm.edu.
Copyright 2008 by Lippincott Williams & Wilkins
ISSN: 1044-3983/08/1901-0055
DOI: 10.1097/EDE.0b013e31815c09b9

Epidemiology Volume 19, Number 1, January 2008

pregnancy,3,4 and the diverse pharmacologic effects of caffeine including increased levels of catecholamines and cellular cyclic adenosine monophosphate (AMP).5,6
Despite methodologic challenges,1 the association between caffeine and risk of miscarriage merits closer evaluation given the widespread consumption of caffeinated beverages and the frequency (and largely unknown causes) of
pregnancy loss. We ascertained caffeine consumption in
some detail as part of a study of pregnancy loss and disinfection by-products in water. Unlike most prior studies, we
recorded information on caffeine consumption and pregnancy-related changes in consumption early in gestation and
were able to stratify results by timing of losses relative to the
timing of exposure assessment.

METHODS
Participant Recruitment and Data Collection
Between 2000 and 2004, we recruited women from
Galveston (Texas), Memphis (Tennessee), and Raleigh
(North Carolina) through public and private prenatal care
providers as well as directly from the communities. Pregnant
women at 12 weeks gestation who were 18 years of age or
older were enrolled; in addition, women aged 18 to 45 who
were trying to conceive for 6 months were identified and
enrolled if they became pregnant. Although women were
recruited when they were pregnant, some miscarriages had
occurred by the time of interview. Details of the recruitment
process are described elsewhere.7,8 Once pregnant, women
were interviewed by telephone before 16 weeks gestation.
Use of alcohol, drugs, and cigarettes; caffeine consumption;
water exposure; medical history; and reproductive history
were queried. An endovaginal ultrasound was performed if
possible before 8 weeks gestation to date the pregnancy and
assess fetal viability.
Miscarriage was defined as loss of a clinically recognized pregnancy at or before 20 completed weeks gestation
from last menstrual period (LMP). We relied on self-reported
LMP to define gestational age because the loss often occurred
prior to the ultrasound and because for viable pregnancies,
the estimated dates of conception from early ultrasound and
LMP were highly concordant. Pregnancy outcomes were
identified through self-report during follow-up and were confirmed by medical records or by vital records for live births
and fetal deaths after 20 weeks gestation.

55

Epidemiology Volume 19, Number 1, January 2008

Savitz et al

TABLE 1. Demographic, Health Behavior, and Pregnancy

History of Women Included in Analysis, by Timing of
Interview and Pregnancy Outcome

Study site
Raleigh
Memphis
Galveston
Maternal age (yrs)
25
2529
3034
35
Race/ethnicity
White, non-Hispanic
Black, non-Hispanic
Hispanic
Asian
Other
Missing
Education
12 yrs
12 to 16 yrs
16 yrs
Missing
Marital status
Married
Other
Missing
Annual income ($)
40,000
40,00180,000
80,001
Missing
Smoking
Nonsmoker
10 cigarettes/d
10 cigarettes/d
Alcohol use
Yes
No
Missing
Body mass index
Underweight
Normal weight
Overweight
Obese
Missing
Age at menarche (yrs)
11
1213
14
Missing

56

Loss Before
Interview
(n 74)
No. (%)*

Loss After
Interview
(n 184)
No. (%)*

Live Birth
(n 2112)
No. (%)*

22 (30)
52 (70)

92 (50)
92 (50)

1047 (50)
1065 (50)

52 (70)
22 (30)
0 (0)

130 (71)
54 (29)
0 (0)

1488 (70)
623 (30)
1 (1)

70 (95)
2 (3)
2 (3)

175 (95)
2 (1)
7 (4)

2046 (97)
16 (1)
50 (2)

17 (23)
34 (46)

47 (26)
94 (51)

646 (31)
1025 (49)

23 (31)

43 (23)

441 (21)

17 (23)
57 (77)

37 (20)
147 (80)

373 (18)
1739 (82)

30 (41)
34 (46)
10 (14)
0 (0)
11.3 2.7

46 (25)
80 (44)
58 (32)
0 (0)
8.2 2.1

200 (10)
722 (34)
1188 (56)
2 (1)
9.4 2.4

9.0 2.2

11.7 3.2

NA

Loss Before
Interview
(n 74)
No. (%)*

Loss After
Interview
(n 184)
No. (%)*

Live Birth
(n 2112)
No. (%)*

34 (46)
31 (42)
9 (12)

86 (47)
60 (33)
38 (21)

953 (45)
789 (37)
370 (18)

14 (19)
16 (22)
29 (39)
15 (20)

52 (29)
44 (24)
42 (23)
46 (25)

638 (30)
681 (32)
573 (27)
220 (10)

47 (64)
23 (31)
1 (1)
1 (1)
2 (3)
0 (0)

95 (51)
76 (41)
7 (4)
2 (1)
4 (2)
0 (0)

1196 (57)
642 (30)
196 (9)
43 (2)
33 (2)
2 (1)

18 (24)
11 (15)
45 (61)
0 (0)

62 (34)
30 (16)
92 (50)
0 (0)

615 (29)
464 (22)
1032 (49)
1 (1)

52 (70)
22 (30)
0 (0)

112 (61)
72 (39)
0 (0)

1411 (67)
700 (33)
1 (1)

23 (31)
23 (31)
25 (34)
3 (4)

75 (41)
56 (30)
47 (26)
6 (3)

879 (42)
680 (32)
466 (22)
87 (4)

*Unless otherwise indicated (last 2 rows of table).

NA indicates not applicable.

70 (95)
2 (3)
2 (3)

175 (95)
7 (4)
2 (1)

2003 (95)
72 (3)
37 (2)

Assessment of Caffeine Intake

19 (26)
54 (73)
1 (1.4)

2 (1)
182 (98.9)
0 (0.0)

36 (2)
2075 (98.2)
1 (0.1)

2 (3)
36 (49)
18 (24)
16 (22)
2 (3)

3 (2)
90 (49)
46 (25)
41 (22)
4 (2)

80 (4)
1049 (50)
494 (23)
434 (21)
55 (3)

7 (10)
44 (60)
23 (31)
0 (0)

39 (21)
93 (51)
50 (27)
2 (1)

474 (22)
1107 (52)
514 (24)
17 (1)

Vitamin use
Yes
No
Employment status
Yes
No
Missing
Diabetes
No
Chronic diabetes
Gestational diabetes
Miscarriage history
No prior pregnancy
1 prior pregnancy
with no miscarriage
1 prior pregnancy
with 1 miscarriage
Induced abortion history
Any
None
Nausea and vomiting
during early pregnancy
No symptoms
Nausea only
Nausea and vomiting
Missing
Mean SD gestational
age at interview
Mean SD gestational
age at loss

We assessed daily consumption of caffeinated beverages in a typical week, referred to as current consumption.
For women who were still pregnant at the baseline interview,
we asked this question at that time. We asked about consumption during pregnancy. For women who had already
experienced a loss, Current consumption reflects exposure
late in the first trimester. We inquired about caffeinated
(brewed and instant) coffee, caffeinated (iced and hot) tea,
and caffeinated soda consumption, including the number and
size of cups consumed per day. We assigned cups of coffee
and hot tea as small (4 10 oz), medium (1214 oz), and large
(16 24 oz). Iced tea was categorized as small (4 10 oz),
medium (1220 oz), and large (2234 oz), and sodas as small
(8 12 oz), medium (14 22 oz), and large (24 34 oz). Reported daily consumption of less than one cup or glass was set
to half of a small cup. We assigned caffeine content to each
caffeinated beverage9 using midpoints of the cup-size inter 2008 Lippincott Williams & Wilkins

Epidemiology Volume 19, Number 1, January 2008

Caffeine and Miscarriage Risk

TABLE 2. Coffee and Total Caffeine Consumption Among

Women Included in the Analysis, by Timing of Interview and
Pregnancy Outcome

Coffee Only
Prior to pregnancy
None
Median
Median
Missing
Median among
consumers (mg/d)
4 wk post-LMP
None
Median
Median
Missing
Median among
consumers (mg/d)
During pregnancy
None
Median
Median
Missing
Median among
consumers (mg/d)
Total Caffeine
Prior to pregnancy
None
Median
Median
Missing
Median among
all women (mg/d)
Median among
consumers (mg/d)
4 wk post-LMP
None
Median
Median
Median among
all women (mg/d)
Median among
consumers (mg/d)
During pregnancy
None
Median
Median
Missing
Median among
all women (mg/d)
Median among
consumers (mg/d)

Loss Before
Interview
No. (%)*

Loss After
Interview
No. (%)*

Live Birth
No. (%)*

45 (61)
11 (15)
17 (23)
1 (1)
360.4

131 (71)
21 (11)
32 (17)
0 (0)
348.0

1414 (67)
277 (13)
419 (20)
2 (1)
348.0

50 (68)
11 (15)
13 (18)
0 (0)
360.4

139 (76)
16 (9)
29 (16)
0 (0)
372.9

1536 (73)
255 (12)
319 (15)
2 (1)
348.0

53 (72)
7 (10)
13 (18)
1 (1)
200.8

158 (86)
12 (7)
14 (8)
0 (0)
200.8

1798 (85)
143 (7)
170 (8)
1 (1)
200.8

9 (12)
33 (45)
31 (43)
1 (1)
192.2

43 (23)
71 (39)
70 (38)
0 (0)
154.9

423 (20)
844 (40)
844 (40)
1 (1)
181.5

227.5

233.0

244.3

18 (24)
28 (38)
28 (38)
123.7

52 (28)
65 (35)
67 (36)
139.8

599 (28)
757 (36)
756 (36)
115.3

215.7

210.3

207.9

24 (32)
18 (24)
31 (42)
1 (1)
76.9

78 (42)
53 (29)
53 (29)
0 (0)
66.2

950 (45)
624 (30)
537 (25)
1 (1)
38.4

200.8

150.1

144.2

*Unless otherwise indicated.

Median coffee consumption among all women was 0.

Median calculated among those who consumed coffee or caffeine, based on the
distribution of values in the specific time interval.

2008 Lippincott Williams & Wilkins

vals to generate exposure indices. We evaluated total caffeine

exposure from all sources and from coffee alone.
In addition to collecting data on current consumption,
we also inquired about changes in consumption, including the
amount and timing of any changes. Most women (92%)
reported a change during pregnancy. When the day of the
change was not provided (approximately 25% of those who
reported changing) but the week could be estimated (ie, first,
second, etc.), we assigned the day as the midpoint of that
week; when only the month could be recalled, we assigned
the midpoint of the month.
We considered 3 time points of caffeine exposure:
Prepregnancy exposure; 4 weeks after LMP (after any changes
associated with planning pregnancy) and current consumption at
the time of the interview (or when still pregnant, for women who
experienced losses). The third time period was presumably
reflective of changes that occurred early in pregnancy, after the
onset of any nausea and vomiting.

Statistical Methods
We used a discrete-time continuation ratio logistic survival model to estimate week-specific odds ratios for the probability of having a miscarriage in a given week, conditional on
a woman still being pregnant at the beginning of that week.
Women who were planning pregnancy were not included in the
risk set for analysis until they enrolled following a positive
pregnancy test. Women were followed from the day of enrollment up to 20 weeks gestation; possible outcomes were pregnancy survival to 20 weeks, miscarriage, or loss to follow-up.
Coffee and caffeine consumption at each of the 3 time points
was divided into 3 groups: none, less than or equal to the median
among consumers at that time point, and above the median. In
separate models, we compared those 75th percentile among
consumers with those who consumed none.
Potential confounders included maternal age, race/ethnicity, education, marital status, income, smoking, alcohol
use, body mass index, age at menarche, employment status,
diabetes, miscarriage history, induced abortion history, vitamin use, and nausea and vomiting in early pregnancy. Covariates were retained in the final model if they were predictive of the outcome based on a P value of 0.20, or if they
changed effect estimates for the exposure of interest by
10% when excluded from the model. Increased risk was
found for women who were older, non-Hispanic, highly
educated, unmarried, alcohol users, and nonusers of vitamins,
as well as those lacking nausea and vomiting; those covariates were included in the final models.
Of the 2766 women enrolled in the study, 32 withdrew
and 227 were excluded for other reasons, including being
greater than 12 weeks gestation at enrollment based on study
ultrasound, being unreachable by telephone for more than 7
weeks, or moving out of the study area. Additionally, second
or third study pregnancies (n 69), multiple gestations (n
23), and women with inconsistent or invalid essential data

57

Epidemiology Volume 19, Number 1, January 2008

Savitz et al

TABLE 3. All Pregnancy Loss (n 258). Unadjusted and Adjusted* Odds Ratios
Relating Coffee and Caffeine Consumption at 3 Time Points to the Risk of
Miscarriage (Before or After Interview), Contrasting the Following Consumption
Categories: None, Below or Equal to the Median, Above the Median, and Above
the 75th Percentile. All Study Participants

Coffee Intake (mg/d)

Prepregnancy
0
0 to 348.0
348.0
696.0
4 wk post-LMP
0
0 to 348.0
348.0
602.3
Time of telephone interview
0
0 to 200.8
200.8
372.9
Total Caffeine Intake (mg/d)
Prepregnancy
0
0 to 243.7
243.7
513.2
4 wk post-LMP
0
0 to 210.3
210.3
463.1
Time of telephone interview
0
0 to 144.3
144.3
273.2

No. Losses

Unadjusted OR

Adjusted OR (95% CI)*

176
32
49
18

1.0
0.8
0.9
1.0

1.0
0.9 (0.61.3)
0.8 (0.51.1)
0.9 (0.51.5)

189
27
42
15

1.0
0.8
1.0
0.8

1.0
0.8 (0.51.2)
0.9 (0.61.3)
0.7 (0.41.2)

211
19
27
5

1.0
1.1
1.2
0.8

1.0
1.2 (0.71.9)
1.0 (0.61.6)
0.7 (0.31.8)

52
104
101
48

1.0
1.0
0.9
0.9

1.0
1.0 (0.71.4)
0.9 (0.61.2)
0.8 (0.51.2)

70
93
95
47

1.0
1.1
1.1
1.0

1.0
1.0 (0.71.4)
1.0 (0.71.3)
0.9 (0.61.3)

102
71
84
46

1.0
1.1
1.4
1.5

1.0
1.0 (0.71.4)
1.2 (0.91.7)
1.3 (0.91.9)

*Model adjusted for maternal age, race/ethnicity, maternal education, marital status, alcohol use, vitamin use, and
symptoms of nausea and vomiting during early pregnancy.

Data missing for 1 woman with pregnancy loss.

Reference category.

(n 8) were excluded from the analysis, leaving a total of

2407 pregnancies. There were 258 miscarriages (74 29%
before the interview and 184 71% after the interview),
2112 live births, and 37 women who contributed persontime of observation but had unknown pregnancy outcomes.

RESULTS
Study participants were predominantly from Raleigh or
Memphis (83%), included a substantial proportion of black
women (32%), represented a range of education and income
levels, and contained few smokers (5%) or alcohol users (2%)
(Table 1). Sizable proportions were overweight (24%) or

58

obese (21%). Prior miscarriage was reported by 21%, and

prior induced abortion by 18%. A higher proportion of the
pregnancy losses occurred before the interview, as opposed to
after, among women from Memphis, older mothers, women
with higher education and income, alcohol users, nonusers of
vitamins, and those without nausea or vomiting.
Among women who had live births, 33% of women
reported that they drank coffee prior to pregnancy. The
proportion of daily coffee drinkers declined to 15% at the
time of the interview; overall caffeine intake diminished as
well (Table 2). Caffeinated soda, iced tea, and coffee all
contributed substantially to caffeine intake. Coffee con 2008 Lippincott Williams & Wilkins

Epidemiology Volume 19, Number 1, January 2008

Caffeine and Miscarriage Risk

TABLE 4. Pregnancy Loss Before Interview (n 74). Unadjusted and Adjusted*

Odds Ratios Relating Coffee and Caffeine Consumption at 3 Time Points to the
Risk of Miscarriage Before Interview, Contrasting the Following Consumption
Categories: None, Below or Equal to the Median, Above the Median, and Above
the 75th Percentile

Coffee Intake (mg/d)

Prepregnancy
0
0 to 348.0
348.0
696.0
4 wk post-LMP
0
0 to 348.0
348.0
602.3
Time of telephone interview
0
0 to 200.8
200.8
372.9
Total Caffeine Intake (mg/d)
Prepregnancy
0
0 to 243.7
243.7
513.2
4 wk post-LMP
0
0 to 210.3
210.3
463.1
Time of telephone interview
0
0 to 144.3
144.3
273.2

No. Losses

Unadjusted OR

Adjusted OR (95% CI)*

45
11
17
7

1.0
0.8
1.2
1.8

1.0
0.8 (0.32.0)
1.0 (0.61.9)
1.5 (0.73.5)

50
11
13
4

1.0
0.9
1.2
0.9

1.0
1.0 (0.42.2)
1.0 (0.51.9)
0.8 (0.32.3)

53
7
13
2

1.0
1.7
2.3
1.4

1.0
1.9 (0.84.4)
1.8 (0.93.7)
1.1 (0.34.5)

9
33
31
14

1.0
1.6
1.3
1.2

1.0
1.6 (0.73.4)
1.2 (0.62.6)
1.1 (0.42.6)

18
28
28
11

1.0
1.2
1.1
0.6

1.0
1.2 (0.62.2)
1.0 (0.52.0)
0.5 (0.21.4)

24
18
31
18

1.0
1.3
2.1
2.6

1.0
1.1 (0.62.2)
1.9 (1.13.5)
2.3 (1.24.5)

*Model adjusted for maternal age, race/ethnicity, maternal education, marital status, alcohol use, and vitamin use.

Data missing for women with pregnancy loss.

Reference category.

sumption did not decrease more notably among women

reporting nausea and vomiting in early pregnancy compared with those who did not have nausea and vomiting
(data not shown).
Coffee consumption before LMP was highly correlated
with consumption 4 weeks after LMP (r 0.80), but the
correlation was lower with consumption at the time of the
interview (r 0.53). Total caffeine showed similar patterns
(r 0.82 and r 0.53, respectively). Coffee and total caffeine
consumption were highly correlated with one another prepregnancy (r 0.86) and at 4 weeks after LMP (r 0.82), but
less so at the time of the interview (r 0.64).
2008 Lippincott Williams & Wilkins

Among all women, coffee and caffeine consumption at

all 3 time points were unrelated to the overall risk of miscarriage (Table 3), with all adjusted odds ratios between 0.7 and
1.3. The most suggestively elevated odds ratios were for total
caffeine above the median and above the 75th percentile at
the time of the interview, with odds ratios of 1.2 and 1.3,
respectively. Analyses of losses before the interview (Table
4), for whom there is greater susceptibility to biased reporting
of exposure, yielded evidence of a positive association with
coffee and caffeine exposure around the time of the interview, but the results were imprecise. When analyses were
restricted to losses after the interview (Table 5), the results

59

Epidemiology Volume 19, Number 1, January 2008

Savitz et al

TABLE 5. Pregnancy Loss After Interview (n 184). Unadjusted and Adjusted*

Odds Ratios Relating Coffee and Caffeine Consumption to the Risk of Miscarriage
After Interview, Contrasting None, Below or Equal to the Median, Above the
Median, and Above the 75th Percentile

Coffee Intake (mg/d)

Prepregnancy
0
0 to 348.0
348.0
696.0
4 wk post-LMP
0
0 to 348.0
348.0
602.3
Time of telephone interview
0
0 to 200.8
200.8
372.9
Total Caffeine Intake (mg/d)
Prepregnancy
0
0 to 243.7
243.7
513.2
4 wk post-LMP
0
0 to 210.3
210.3
463.1
Time of telephone interview
0
0 to 144.3
144.3
273.2

No. Losses

Unadjusted OR

Adjusted OR (95% CI)*

131
21
32
11

1.0
1.0
1.0
1.0

1.0
1.2 (0.72.1)
0.8 (0.51.3)
1.1 (0.52.3)

139
16
29
11

1.0
0.9
1.0
1.0

1.0
1.1 (0.62.1)
0.8 (0.51.2)
0.6 (0.31.4)

158
17
9
3

1.0
1.0
1.0
2.9

1.0
0.7 (0.31.6)
1.0 (0.51.8)
3.0 (0.713.2)

43
71
70
34

1.0
1.0
1.0
1.1

1.0
1.1 (0.71.7)
1.0 (0.61.6)
0.8 (0.51.5)

52
65
67
36

1.0
0.9
0.9
0.9

1.0
1.1 (0.71.7)
0.8 (0.51.3)
0.8 (0.51.3)

78
53
53
28

1.0
0.8
1.1
1.1

1.0
0.9 (0.61.4)
1.1 (0.71.7)
1.1 (0.61.8)

*Model adjusted for maternal age, race/ethnicity, maternal education, marital status, alcohol use, vitamin use, and
symptoms of nausea and vomiting during early pregnancy.

Reference category.

were uniformly close to the null except for a single, imprecise

measure.

DISCUSSION
These data show little evidence for associations between coffee or caffeine consumption prior to or early in
pregnancy and the risk of miscarriage. Relative to previous
studies, our population had modest levels of coffee and
caffeine intake, with median coffee intake of 350 mg/d prior
to and early in pregnancy, and 200 mg/d at the time of the
interview among those who consumed any caffeine. The
median levels of caffeine consumption before and during
pregnancy among those who had some exposure were modest

60

relative to previous studies, equivalent to 1.1 to 1.7 seven-oz

cups of brewed coffee per day prior to pregnancy and 4 weeks
post-LMP. This restricted our ability to examine possible
effects above the 300 to 400 mg/d range addressed in a
number of previous studies.1 Our study population may be
more health-conscious than those in previous studies because
we enrolled volunteers who sought prenatal care early or
were planning a pregnancy. This selectivity or the geographic
location of the study may account for the relatively low level
of coffee and caffeine consumption. Whereas some studies
report elevated risks at levels in the range that we were able
to examine (300 mg/d),10 13 most did not.14 18 Classification of previous studies based on timing of caffeine assess 2008 Lippincott Williams & Wilkins

Epidemiology Volume 19, Number 1, January 2008

ment (during vs. after pregnancy) does not provide clear

evidence that timing of assessment is predictive of results,
with both suggestively positive12,17 and negative16,18 studies
among those with early assessment.
Given our observed null association, a key question
concerns potential biases that may have masked an underlying causal effect. One potential source of error is misclassification of coffee and caffeine intake. We did not
account for caffeine from chocolate or medications, and
there is substantial variability in the dose of caffeine
resulting from different beverage preparation methods.19
Thus, even if self-reported consumption of cups of coffee
per day and the cup sizes were accurate, the inferred
dose of caffeine using standard conversion tables9 would
be subject to error that is nondifferential by outcome.
The exposure window of interest falls shortly before
conception and extends into the earliest weeks of pregnancy, a time of marked change in behavior, making a
single assessment of limited value. We assessed caffeine
use in some detail earlier in pregnancy than most previous
studies, but our assessment could be subject to erroneous
recall of the timing and nature of changes. Caffeine assessment preceded 67% of miscarriages, limiting the potential for recall bias. Women who were interviewed after
miscarriage reported elevated levels of coffee and caffeine
intake during pregnancy compared with women who had
losses after the interview. The elevated consumption could
reflect recall bias or true differences, perhaps due to
decreased nausea and thus increased consumption after
fetal demise but before recognition of the loss, or erroneously reporting on their intake after the loss was known to
have occurred. In addition to concerns about exposure
assessment, our inability to isolate subsets of pregnancy
losses based on karyotypes may mask a modest effect
among chromosomally distinctive subsets.14
Nausea in early pregnancy is strongly predictive of fetal
survival,20 23 presumably indicative of hormonal changes
that support continued pregnancy. Nausea also affects dietary
habits, often resulting in aversion to coffee. Thus, there has
been a concern that caffeine consumption would decline
among women who have nausea (and a more favorable
prognosis) relative to those who do not have nausea, resulting
in an artifactual positive association between caffeine and
pregnancy loss.24 In our study, we did not find a differential
decline in caffeine in relation to reported nausea and vomiting, with all women reporting markedly decreasing coffee
and caffeine from prior to pregnancy through the course of
pregnancy. Adjustment for nausea and vomiting did not
affect the measures of association between coffee or caffeine
and miscarriage risk.
The assessment of coffee and caffeine use early in
pregnancy, before most losses had occurred, should minimize the potential for outcome-related differences in ac 2008 Lippincott Williams & Wilkins

Caffeine and Miscarriage Risk

tual or reported exposure. When we restricted losses to

those that occurred after the interview, presumably eliminating recall bias resulting from having had a loss, the
results remained null. It should be noted, however, that all
women were reporting retrospectively with regard to caffeine intake prior to pregnancy and at 4 weeks post LMP.
While studies in populations that combine early assessment of caffeine with a higher exposure range would yield
more informative results pertaining to higher doses, these
data provide evidence to suggest that, within the lower
range examined, caffeine intake is not associated with risk
of miscarriage.

ACKNOWLEDGMENTS
We thank Christina Makarushka, Project Manager, and
Yanfang Jiang, Staff Programmer, for their contributions.
REFERENCES
1. Signorello LB, McLaughlin JK. Maternal caffeine consumption and
spontaneous abortion. A review of the epidemiologic evidence. Epidemiology. 2004;15:229 239.
2. Goldstein A, Warren R. Passage of caffeine into human gonadal and
fetal tissue. Biochem Pharmacol. 1962;11:166 168.
3. Aldridge A, Bailey J, Neims AH. The disposition of caffeine during and
after pregnancy. Semin Perinatol. 1981;5:310 314.
4. Brazier JL, Ritter J, Berland M, et al. Pharmacokinetics of caffeine
during and after pregnancy. Dev Pharmacol Ther. 1983;6:315322.
5. Weathersbee PS, Lodge JR. Caffeine: its direct and indirect influence on
reproduction. J Reprod Med. 1977;19:55 63.
6. Kirkinen P, Jouppila P, Koivula A, et al. The effect of caffeine on
placental and fetal blood flow in human pregnancy. Am J Obstet
Gynecol. 1983;147:939 942.
7. Promislow JH, Makarushka CM, Gorman JR, et al. Recruitment for a
community-based study of early pregnancy: the Right From The Start
study. Paediatr Perinat Epidemiol. 2004;18:143152.
8. Savitz DA, Singer PS, Hartmann KE, et al. Drinking Water Disinfection
By-Products and Pregnancy Outcome. Final Report for AwwaRF
Project 2579. Denver, CO: AwwaRF; 2005.
9. Bunker ML, McWilliams M. Caffeine content of common beverages.
J Am Diet Assoc. 1979;74:28 32.
10. Parazzini F, Chaternoud L, Di Cintio E, et al. Coffee consumption and
risk of hospitalized miscarriage before 12 weeks of gestation. Hum
Reprod. 1998;13:1286 1291.
11. Infante-Rivard C, Fernandez A, Gauthier R, et al. Fetal loss associated
with caffeine intake before and during pregnancy. JAMA. 1993;270:
2040 2043.
12. Wen W, Shu XO, Jacobs DR Jr, et al. The associations of maternal
caffeine consumption and nausea with spontaneous abortion. Epidemiology. 2001;12:38 42.
13. Cnattingius S, Signorello LB, Anneren G, et al. Caffeine intake and the
risk of first-trimester spontaneous abortion. N Engl J Med. 2000;343:
1839 1845.
14. Kline J, Levin B, Silverman J, et al. Caffeine and spontaneous abortion
of known karyotype. Epidemiology. 1991;2:409 417.
15. Fenster L, Eskenazi B, Windham GC, et al. Caffeine consumption
during pregnancy and spontaneous abortion. Epidemiology. 1991;2:
168 174.
16. Fenster L, Hubbard AE, Swan SH, et al. Caffeinated beverages, decaffeinated coffee, and spontaneous abortion. Epidemiology. 1997;8:515
523.
17. Dlugosz L, Belanger K, Hellenbrand K, et al. Maternal caffeine consumption and spontaneous abortion: a prospective cohort study. Epidemiology. 1996;7:250 255.
18. Mills JL, Holmes LB, Aarons JH, et al. Moderate caffeine use and the
risk of spontaneous abortion and intrauterine growth retardation. JAMA.
1993;269:593597.

61

Savitz et al

19. Bracken MB, Triche E, Grosso L, et al. Heterogeneity in assessing

self-reports of caffeine exposure: implications for studies of health
effects. Epidemiology. 2002;13:165171.
20. Jarnfelt-Samsioe A, Samsioe G, Velinder GM. Nausea and vomiting in
pregnancy-a contribution to its epidemiology. Gynecol Obstet Invest.
1983;16:221229.
21. Klebanoff MA, Koslowe PA, Kaslow R, et al. Epidemiology of vomiting
in early pregnancy. Obstet Gynecol. 1985;66:612 616.

62

Epidemiology Volume 19, Number 1, January 2008

22. Tierson FD, Olsen CL, Hook EB. Nausea and vomiting of pregnancy
and association with pregnancy outcome. Am J Obstet Gynecol. 1986;
155:10171022.
23. Weigel RM, Weigel MM. Nausea and vomiting of early pregnancy and
pregnancy outcome. A meta-analytical review. Br J Obstet Gynaecol.
1989;96:13121318.
24. Stein Z, Susser M. Miscarriage, caffeine, and the epiphenomena of
pregnancy: the causal model. Epidemiology. 1991;2:163167.

2008 Lippincott Williams & Wilkins