American Journal of Epidemiology
Copyright O 1997 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 145, No. 4
Printed in U.SA.
Caffeine Intake and Delayed Conception: A European Multicenter Study on
Infertility and Subfecundity
F. Bolumar,1 J. Olsen,2 M. Rebagliato,1 L. Bisanti,3 and the European Study Group on Infertility and
Subfecundity 4
The effects of caffeine consumption on delayed conception were evaluated in a European multicenter study
on risk factors of infertility. Information was collected retrospectively on time of unprotected intercourse for the
first pregnancy and the most recent waiting time episode in a randomly selected sample of 3,187 women aged
25-44 years from five European countries (Denmark, Germany, Italy, Poland, and Spain) between August 1991
and February 1993. The consumption of caffeinated beverages at the beginning of the waiting time was used
to estimate daily caffeine intake, which was categorized as 0-100, 101-300, 301-500, and >501 mg. Risk of
subfecundity (>9.5 months) and the fecundability ratio, respectively, were assessed by logistic regression and
Cox proportional hazard analyses, adjusting for age, parity, smoking, alcohol consumption, frequency of
intercourse, educational level, working status, use of oral contraceptives, and country. A significantly increased odds ratio (OR) of 1.45 (95% confidence interval (Cl) 1.03-2.04) for subfecundity in the first pregnancy
was observed for women drinking more than 500 mg of caffeine per day, the effect being relatively stronger
in smokers (OR = 1.56, 95% Cl 0.92-2.63) than in nonsmokers (OR = 1.38, 95% Cl 0.85-2.23). Women in the
highest level of consumption had an increase in the time leading to the first pregnancy of 11 % (hazard ratio =
0.90, 95% Cl 0.78-1.03). These associations were observed consistently in all countries as well as for the most
recent waiting time episode. The authors conclude that high levels of caffeine intake may delay conception
among fertile women. Am J Epidemiol 1997; 145:324-34.
caffeine; infertility; risk factors
Caffeine is one of the most widely and routinely
consumed pharmacologically active substances and is,
in general, considered quite harmless to human beings.
It can be detected in all body fluids shortly after
intake, and it passes through all biologic membranes,
including the placental barrier (1).
Although the effects of caffeine are very much related to the mode of administration (2, 3) and the
mechanisms underlying the pharmacologic effects of
caffeine remain controversial (4), animal research
(mainly in rodents) has consistently found caffeine to
be teratogenic (5-7). However, the doses used in those
studies were too high to be compared with human
doses. In general, the effects of caffeine on fertility
and reproduction in rodents are limited and appear to
be restricted mainly to an increased time to pregnancy
(TTP) and lower birth weight (3, 8, 9). Nevertheless,
differences in the metabolism of caffeine between
rodents and humans, particularly in the uneven proportion of caffeine metabolites that are trimethyl derivatives (1, 10), make the extrapolation of caffeine
effects from animals to humans difficult (11-13).
Several studies on human beings have reported an
association between caffeine intake and delayed TTP
(14-20), whereas others have shown either no association or association only at very high levels of consumption (20-23). These discrepancies in study results have been attributed to recruitment of selected
groups of subjects, different precision in measuring
coffee consumption and intake of other sources of
caffeine, differences in brewing methods, and lack of
control for potentially confounding variables (24-27).
In each country, coffee consumption patterns are
influenced by lifestyle, culture, tradition, and behavior
(28). Within Europe, coffee brewing methods and
drinking habits differ widely between the northern and
Received for publication March 11, 1996, and accepted for publication October 2, 1996.
Abbreviations: OR, odds ratio; TTP, time to pregnancy.
1
Department of Public Health, Alicante University, Spain.
2
Danish Epidemiology Science Centre and Department of Epidemiology and Social Medicine, Aarhus University, Denmark.
3
Regione Lombardia, Servizio Epidemiologia e Sistema Informativo, Milano, Italy.
4
Members of the European Study Group on Infertility and Subfecundity: Svend Juul (Denmark), Jom Olsen (Denmark), Patrick
Thonneau (France), Wilfred Karmaus (Germany), Irene FigaTalamanca (Italy), Luigi Bisanti (Italy), Francisco Bolumar (Spain).
Reprint requests to Prof. Francisco Bolumar, Departamento de
Salud Publica, Universidad de Alicante, Campus de San Juan, Ap.
Correos 374.03080, Alicante, Spain.
324
Caffeine Intake and Infertility
southern countries. Whereas filtering coffee is the
most popular brewing method in the north, mocha and
espresso coffee are used predominantly in Italy and
Spain. These differences in brewing methods result in
different efficiency in caffeine extraction (28), which
is greater (97-100 percent) for filter coffee. The different types of coffee (e.g., robusta, arabica) and cup
sizes (small cups, mugs, and so forth) generally used
in each country also account for differences in caffeine
content per cup of coffee. The stronger coffee is used
in the south, but in small cups; and the weaker coffee
is used in the north in mugs. As a result of these
variables acting in different directions, the estimated
average caffeine content per cup of coffee does not
differ much among the different countries. However,
in spite of the small variation in the average content
per cup, the frequency of consumption is very different, resulting in a marked downward gradient in the
average daily caffeine intake from northern to southern Europe. This wide range of exposure permits comparisons that would prove very difficult to carry out in
any single country.
Taking into account these peculiarities, the relation
between self-reported coffee/caffeine consumption
and delayed TIP was evaluated, among women who
were trying to become pregnant, in the context of a
population-based European study on putative risk factors for infertility and subfecundity.
MATERIALS AND METHODS
A detailed description of subjects and methods has
been reported elsewhere (29). Briefly, women aged
25-44 years from five European countries (Denmark,
Germany, Italy, Poland, and Spain) were randomly
selected from population registers, census registers,
and electoral rolls. Between August 1991 and February 1993, a total of 6,630 women were interviewed.
Participation rates ranged from 54 to 88 percent, with
a 68 percent overall participation rate.
Information was collected retrospectively by personal interview on sociodemographic factors, contraceptive use, sexual activity, reproductive history, cigarette smoking, caffeinated beverages, alcohol
consumption, and TIP. The questionnaire contained
questions on exposures at the starting time of unprotected sexual intercourse covering both the first and
the most recent waiting time. Questions on exposures
also included the partner when appropriate.
All women were asked how many cups of coffee
and tea they drank per day as well as how many
glasses/bottles of cola. It was assumed that there were
50 mg of caffeine per cup of tea, and 40 mg in cola
drinks. Each person's caffeine intake from coffee was
estimated after taking into account the average cup
Am J Epidemiol
Vol. 145, No. 4, 1997
325
size, the different coffee mixtures, and the different
coffee brewing methods used in each of the participating countries (24, 28). Accordingly, it was estimated that a cup of coffee had an average caffeine
content of 130 mg in Denmark, based on a 3.5:1 ratio
of consumption of arabica (1.1 percent caffeine) versus robusta coffee (2.2 percent caffeine), the use of
filter coffee (97-100 percent extraction efficiency) as
the most common brewing method, and a 190-ml cup
size (28). In southern Europe, the estimated average
caffeine content per cup of 115 mg was based on a 1:1
arabica to robusta ratio, the use of espresso (80 percent
extraction efficiency) and mocha coffee (92-98 percent extraction efficiency), and a usual cup size of
35-50 ml (28). In Germany, although the most common brewing method and cup size were similar to
those in Denmark, the 8:1 ratio between arabica and
robusta coffee (28) yielded an estimated average caffeine content of 115 mg per cup.
Total caffeine intake in milligrams per day was
categorized into four levels: 0-100, 101-300, 301500, and >501. Coffee consumption was measured as
cups of coffee per day and categorized into four levels:
none, one to two, three to four, and five or more.
Alcohol consumption was added up from specific
questions on each type of alcohol and measured in
number of drinks per week. Cigarette smoking was
measured as number of cigarettes per day and later
grouped into the following categories: none, one to 10,
and 11 or more.
The relations between fertility and coffee and caffeine consumption were assessed using the reported
I I P . TTP was measured in months and calculated as
the time between stopping a birth control method and
time of conception. In fact, TTP was estimated from
the following question: "How many months did it take
you to become pregnant? That is, how many months
were you having sexual intercourse without doing
anything to avoid pregnancy?"
The variable TTP was categorized in months as
0-3.4, 3.5-9.4, 9.5-15.4 and >15.5. Although several
analyses were carried out using these different cutoff
points, only the results based on the cutoff point of 9.5
months are presented in the tables. The threshold was
set at 9.5 months to avoid digit preference clustering
and to remain below the eligibility criterion for treatment (>12 months), which would censor the observed
TTP.
Although the analysis was based mainly on the first
pregnancy, the relations between fertility and coffee
and caffeine consumption were assessed using the
reported TTP for both the first waiting time episode,
which included only women who finally became pregnant, and the most recent waiting time. Data from the
326
Bolumar et a).
most recent waiting time were censored in the sense
that unsuccessful waiting times were also included in
the analyses. Both analyses were restricted to those
women who had planned their pregnancies and had not
been treated for infertility, since only those women
had well-defined TTPs.
Use of oral contraceptives within 12 months prior to
the starting time, mother's education, working status
(not working, part-time job, full-time job), alcohol
consumption, cigarette smoking, frequency of sexual
intercourse, history of pelvic inflammatory diseases,
parity, and age were considered as potential confounders. To calculate the risk of delayed conception in
relation to caffeine and coffee consumption, adjusted
odds ratios were estimated through a multivariate logistic regression analysis using TTP as a dichotomous
variable (>9.5 months). The analysis was performed
for all countries together (stratified by smoking status)
and for each individual country using the most appropriate model based on a 10 percent "change-inestimate" confounder-selection strategy (30). Adjusted
fecundity ratios were calculated for risk factors and
confounding variables by means of a Cox proportional
hazards model for continuous data, following a similar
selection strategy as described for the multivariate
logistic models. The proportional risk assumption was
verified for all the variables included in the model.
The statistical analysis was done using an SPSS software package (31).
RESULTS
Of the 6,630 women who were interviewed, 1,567
had never been pregnant; 3,187 reported that their first
pregnancy was planned and were therefore included in
the first pregnancy analysis. For the analysis of the
most recent waiting time, 3,092 women were finally
included.
Concerning the exposure, approximately 80 percent
of all women reported drinking at least one cup of
coffee per day at the start of the waiting time leading
to the first pregnancy. Consumption was greatest in
Denmark, where more than 40 percent of women
reported drinking five or more cups of coffee per day,
and least in Poland, where only 2.1 percent reported
that level of coffee drinking. The average daily caffeine intake was 707 mg in Denmark (83.8 percent
from coffee, 15.5 percent from tea, and 0.7 percent
from cola); 353 mg in Germany (86.7 percent from
coffee, 10.5 percent from tea, and 2.8 percent from
cola); 278 mg in Poland (41.7 percent from coffee,
55.7 percent from tea, and 2.6 percent from cola); 286
mg in northern Italy (90.6 percent from coffee, 6.2
percent from tea, and 3.2 percent from cola); 256 mg
in southern Italy (93.7 percent from coffee, 3.5 percent
from tea, and 2.8 percent from cola); 199 mg in Spain
(94.6 percent from coffee, 1.9 percent from tea, and
3.5 percent from cola).
In table 1, the frequency distribution of TTP in
months according to the women's coffee consumption
and caffeine intake is shown, together with the mean
and median TTP values for each level. The distribution was very similar for coffee and caffeine based on
the levels of exposure adopted; and significantly
longer periods of waiting time for increasing levels of
coffee and caffeine intake were found.
In table 2, the concordance between the different
levels of exposure as measured by coffee or caffeine is
shown. Although the categories were not exactly comparable, 22.5 percent of women in the highest caffeine
intake category were classified as low to moderate
consumers when coffee was the measure of exposure.
The distribution of potential confounding variables
by levels of caffeine intake is shown in table 3.
Women who used oral contraceptives within the last
TABLE 1. Distribution of waiting time to the first pregnancy (TTP) In months according to the coffee
drinking habits and caffeine Intake of women from Denmark, Germany, Haly, Poland, and Spain, August
1991 to February 1993
TTP (months)
No.
of
women
0-3.4
3.5-9.4
Coffee (cups/day)
None
1-2
3-4
;>5
707
1,257
738
475
64
59
62
55
Caffeine (mg/day)
0-100
101-300
301-500
£501
522
1,230
802
601
63
62
61
55
'
(%]
Mean
Medan
ff
20
21
20
22
16
20
18
23
6.5
7.4
7.3
8.2
2.0
3.0
2.0
3.0
0.003
21
20
20
21
16
18
19
24
6.5
6.9
7.3
8.9
2.0
3.0
2.0
3.0
0.001
• p values obtained from Knjskal-WaJlis one-way analysis of variance.
Am J Epidemiol
Vol. 145, No. 4, 1997
Caffeine Intake and Infertility
327
TABLE 2. Concordance between different Ievets of exposure as measured by coffee consumption and
caffeine Intake during the waiting time to the first pregnancy of women from Denmark, Germany, Italy,
Poland, and Spain, August 1991 to February 1993
Coffee
(cups/day)
None
1-2
Caffeine (mg/day)
0-100
%
101-300
%
301-500
%
£501
%
523
100.0
142
1,094
11.5
88.5
32
153
618
4.0
19.0
77.0
8
14
114
467
1.3
2.3
18.9
77.5
£5
12 months prior to the starting time had a higher
intake, as did those who were employed full time.
There were variations by educational level, but no
coherent pattern emerged. Those who drank more caffeinated beverages also had a greater alcohol consumption and smoked more, but they tended to have a
lower level of sexual activity. Finally, the women's
age was weakly associated with their caffeine intake.
The distribution of potential confounding variables by
levels of coffee consumption was similar to that of
caffeine.
As summarized in table 4, after adjustment was
made for potential confounders, a significantly increased odds ratio (odds ratio (OR) = 1.45) for subfecundity was observed among women drinking more
than 500 mg of caffeine per day. When data were
broken down by smoking status, the effect of drinking
more than 500 mg was relatively stronger in smokers
(OR = 1.56) than in nonsmokers (OR = 1.38). The
effect varied among the different countries and was
higher for Denmark and Germany and lower for southern Europe.
All analyses were repeated by using cutoff points
for TTPs at 3.5, 9.5 (only data shown), 12.5, and 15.5
months; comparable results were obtained with those
shown in table 4 for the first three cutoff points, but
not for the last one (OR = 1.02).
When the exposure was ascertained from coffee
consumption, a similar result was obtained after controlling for the same set of confounding variables.
Women at high levels of exposure (five or more cups/
day) carried a significant risk (OR = 1.40, 95 percent
confidence interval 1.01-1.95) of delayed conception
(TCP >9.5 months).
In table 5, the same analysis focused on the most
recent waiting time episode can be seen. The emerging
pattern is consistent with that observed in the first
pregnancy, although the association was weaker.
Women drinking more than 500 mg of caffeine per
day had an increased risk of subfecundity (OR =
1.32), although the association was not statistically
significant. A comparable result (OR = 1.26) was
observed for the highest level of coffee consumption.
Am J Epidemiol
Vol. 145, No. 4, 1997
Using Cox's regression (table 6), the risk ratio for
the women who consumed more than 500 mg per day
of caffeine at the start of the waiting time leading to
their first pregnancy, after adjustment for main confounding factors, was 0.90, which represents an increase in the TCP of 11 percent. A similar result was
observed when using coffee as the exposure variable.
In relation to the most recent waiting time, the increase
in the TCP was less than 10 percent for both coffee and
caffeine intake.
In figures 1 and 2 are shown the distributions of the
waiting time to the first pregnancy and the most recent
waiting time according to the caffeine intake levels,
after adjustment for main confounders. Overall within
the first 3 months, 59 percent conceived; and 12 percent took more than 12 months to get pregnant in the
first pregnancy. With regard to the most recent waiting
time, 60 percent conceived within the first 3 months;
and after 12 months, 16 percent were still not pregnant.
DISCUSSION
This study revealed an association of caffeine, at the
upper level of intake, with reduced fecundity in a
sample of fertile women who planned their pregnancy.
This association was consistently observed in all countries, when using either coffee consumption or total
caffeine intake as the exposure measure, and for waiting times to pregnancy related to the first and the most
recent waiting time episodes.
Our data do not allow an assessment of individual
effects of all three caffeinated beverages due to the
small numbers of tea and cola drinkers. Given the high
correlation between coffee and caffeine and since coffee was the main source of caffeine intake in our
sample, it was not possible to test whether the detected
effect was specifically related to coffee consumption
or to the overall caffeine intake. We decided, therefore, to use the latter as the exposure measure for the
main analysis since the use of coffee consumption
might, to some extent, underestimate the exposure of
CO
CO
14.00 (7.96)
23.73(4.01)
Coitus/month
Mother's age (years)
22
12
13
21
16
13
12
13
18
%
3.55 (6.12)
13.68(7.61)
23.78 (3.98)
2.84 (5.88)
16.28(9.54)
23.24(3.91)
Maan (SD)
22
29
27
24
29
24
28
24
26
%
Caffeine (mo/day)
2.98 (4.80)
42
40
36
38
45
33
45
39
39
%
101-300
2.02(6.51)
Mean (SO)
0-100
* p values obtained from chi-square test and analysis of variance,
t SD, standard deviation.
4.87(7.21)
Cigarettes/day
1,316
282
1,567
Work
No paid work
Part time
Full time
3.02 (5.25)
1,203
645
841
476
Mother's education
Primary school
Secondary school
Technical training
University
Mean (SDt)
Alcohol (drinks/week)
971
2,194
women
No.
of
Oral contraception
Yes
No
Confounding
variables
24.03(4.10)
13.69 (7.79)
5.28 (7.15)
3.33 (5.00)
Mean (SD)
301-500
14
19
24
18
10
30
15
23
17
%
23.66(4.21)
13.21 (7.12)
8.79 (8.68)
3.58(5.18)
Mean (SD)
£501
0.05
0.006
<0.001
<0.001
<0.001
<0.001
<0.001
P*
TABLE 3. Distribution of potential confounding variables by level of caffeine intake during the waiting time to the first pregnancy of women from Denmark, Germany,
Italy, Poland, and Spain, August 1991 to February 1993
a
0)
8
Caffeine Intake and Infertility
TABLE 4. Results of a logistic regression analysis of
subfecundlty (reported time to pregnancy £9.5 months)
according to caffeine Intake during the waiting time to the
first pregnancy of European women, August 1991 to February
1993
No. of
women
Caffeine
(mgAJay)
All countrlest
All subjects
0-100
101-300
301-500
£501
Adjusted
OR*
799
599
Nons moke rs
0-100
101-300
301-500
£501
373
796
431
220
1.00
1.03
0.91
1.38
0.72-1.46
0.61-1.35
0.85-2.23
Smokers
0-100
101-300
301-500
£501
148
431
368
379
1.00
1.04
1.13
1.56
0.63-1.71
0.68-1.87
0.92-2.63
1,227
Single countries*
Denmark
0-100
101-300
301-500
£501
28
67
80
276
1.00
0.58
1.58
1.60
0.77-1.36
0.74-1.37
1.03-2.04
0.13-2.66
0.41-6.09
0.45-5.70
Germany
0-100
101-300
301-500
£501
90
239
178
128
1.00
1.15
1.14
1.73
0.63-2.07
0.61-2.14
0.89-3.36
Poland
0-100
101-300
301-500
£501
18
93
66
10
1.00
1.40
2.09
1.20
0.41-4.75
0.59-7.46
0.18-7.81
Italy, north
0-100
101-300
301-500
£501
110
368
237
100
1.00
0.99
0.98
1.32
0.58-1.71
0.55-1.74
0.68-2.56
100
235
1.00
0.91
0.60
1.42
0.44-1.89
0.25-1.44
0.50-3.99
1.00
1.04
0.52
1.49
0.56-1.93
0.21-1.25
0.57-3.89
Italy, south
0-100
101-300
301-500
£501
Spain
0-100
101-300
301-500
£501
132
45
175
225
106
40
• OR, odds ratio; Cl, confidence Interval.
t Odds ratios were adjusted by women's age in years (£20,
21-25, 26-30, £31), use of oral contraceptives In the 12 months
prior to the starting time (yes/no), cigarette smoking (no, 0-10, £11)
and country.
t Odds ratios were adjusted by the same variables listed above
except country.
J Epidemiol
Vol. 145, No. 4, 1997
TABLE 5. Results of a logistic regression analysis of
subfecundity (reported time to pregnancy £9.5 months)
according to caffeine intake and coffee consumption during
the most recent waiting time episode of European women,
August 1991 to February 1993
95% Cl'
1.00
1.02
1.01
1.45
521
329
Caffeine (mg/day)
0-100
101-300
301-500
£501
Coffee (cups/day)
None
1-2
3-^4
£5
No. of
women
Adjusted
OR'.t
95%CI«
472
1,166
1.00
1.03
1.05
1.32
0.77-1.37
0.77-1.43
0.94-1.86
1.00
1.17
1.15
1.26
0.90-1.51
0.86-1.33
0.91-1.74
780
635
631
1,236
717
480
• OR, odds ratio; Cl, confidence interval.
t Odds ratios were adjusted in both models by women's age in
years (£20,21-25, 26-30, £31), use of oral contraceptives in the 12
months prior to the starting time (yes/no), cigarette smoking (no,
0-10, £11), parity (0, £1), working status, and country.
TABLE 6. Results of a Cox regression analysts* of the effect
of caffeine Intake on time to pregnancy for the first pregnancy
and the most recent waiting time in a representative sample
of 25- to 44-year-old women from Denmark, Germany, Italy,
Poland, and Spain, August 1991 to February 1993
Ffrstpregnancy
MRWTt
95%Clt
Hazard
ratio
95%Clt
1.00
1.01
1.00
0.90
0.91-1.13
0.89-1.13
0.78-1.03
1.00
1.01
0.98
0.91
0.90-1.14
0.86-1.11
0.79-1.06
1.00
0.93
0.96
0.90
0.85-1.03
0.86-1.07
0.79-1.03
1.00
0.95
0.93
0.92
0.85-1.05
0.82-1.04
0.80-1.06
Hazard
ratio
Caffeine (mg/day)
0-100
101-300
301-500
£501
Coffee (cups/day)
None
1-2
3-4
£5
* The Cox regression models included women's age in years
(520, 21-25, 26-30, £31), use of oral contraceptives in the 12
months prior to the starting time (yes/no), cigarette smoking (no,
0-10, £11), parity (0, £1), working status, and country.
t MRWT, most recent waiting time (includes waiting times that
might have ended in pregnancy or been censored); Cl, confidence
interval.
some of the women in the highest level of caffeine
intake (table 2).
Measuring exposure accurately is particularly difficult in studies dealing with caffeine consumption. Although several urine and serum biomarkers for caffeine intake are available (32) and have been used to
validate self-reported consumption as assessed by
questionnaire (33, 34), they cannot replace interview
data at their present stage of development given their
low sensitivity (27, 33). In any case, the design of this
study precluded the use of biomarkers to assess con-
330
Bolumar et al.
1.0
08
>
0.6 -
Caffeine (mg/day)
CD
I
E
0.4 -
o
•
501 +
O
301-500
x
0-300
0.2 -
0.0
12
18
24
30
36
Waiting time (months)
FIGURE 1. Adjusted distribution of waiting time to the first pregnancy (estimated from the model shown in table 6) by levels of caffeine
intake at the beginning of the waiting time.
sumption at the starting time since they measure only
recent consumption.
Calculation of the exposure took into account all
caffeinated beverages and was categorized into four
levels that differed from and were higher than those
adopted in previously published US papers on caffeine
and reproduction (14, 17-19, 21), since coffee consumption in some of the participating European countries is much higher than in the United States (28).
Our measure of total caffeine intake followed suggested recommendations for epidemiologic studies
(24) but was based on ecologic data on average cup
size, coffee mixtures, and brewing methods from each
country since no individual data were available. Although our measurement constitutes a moderate improvement over other measurements used in previous
studies, some misclassification is still present due to
interindividual variations within each country. In addition, serum levels of caffeine metabolites differ for
subjects reporting a similar caffeine intake due to
differences in metabolic degradation rates (26), further
contributing to misclassification of exposure. A possible explanation for the apparent absence of effect
when the cutoff point for subfecundity is 15.5 months
can be found in the different nature of subfecundity in
the population groups that conceive within 1 year of
attempting compared with the population groups that
conceive later. Healthy subjects belong to the former
groups, having a normal fecundity apart from a relatively short delay for conception influenced by their
exposure to external factors such as smoke, caffeine,
and shift work. The latter groups of population experience a pathologic fecundity that, regardless of the
exposure to external agents (or as a consequence of the
exposure to very hazardous external agents), is affected by internal, organic alterations like tubal constriction, hormonal alteration, uterine dimorphism,
psychological disorders, and so forth. Caffeine, among
others, can be considered a weak risk factor that probably reduces fecundity by a certain fraction, but withAm J Epidemiol
Vol. 145, No. 4, 1997
Caffeine Intake and Infertility
331
1.0
0.8
CO
I
0.6
Caffeine (mg/day)
CO
CD
3
E
0.4 -
o
•
501 +
°
301-500
X
0-300
0.2-
0.0
12
18
24
30
36
Waiting time (months)
FIGURE 2. Adjusted distribution of the most recent waiting time (estimated from the model shown in table 6) by levels of caffeine intake
at the beginning of the waiting time.
out being a sufficient cause of infertility. It is unlikely
that a couple with severe internal alterations can conceive earlier than 1 year of attempting even though
they do not drink any caffeinated beverages. An analogous reasoning may explain the weaker effect observed for the most recent waiting time since this
episode includes women who never succeed in becoming pregnant.
The effect was relatively stronger in smokers than in
nonsmokers, a result that agrees with a previous Danish study (20) but contradicts a very recent American
study (19). Although our results should be taken with
caution given the low magnitude of the difference in
ratios, it is possible that cigarette smoking interacts
moderately with the effect of caffeine. Several studies
have, however, suggested that cigarette smoking increases xanthine oxidase activity as reflected by the
pattern of caffeine metabolites excreted in urine (1,
35). If this is the case, the acceleration in the clearance
Am J Epidemiol
Vol. 145, No. 4, 1997
of caffeine produced by smoking would act in the
opposite direction, diminishing instead of exacerbating the effect. This is the mechanism suggested by
Stanton and Gray (19) to explain their findings. In our
case, an alternative interpretation may be advanced.
Heavy coffee drinkers differ in. several ways from
moderate drinkers and from those who do not drink at
all (21, 24), particularly with respect to cigarette
smoking. Heavy coffee drinkers tend to smoke more.
If we consider the possibility that some people who
accurately report their coffee consumption underreport
their tobacco consumption, the relatively greater effect
of caffeine seen in smokers might be due to residual
confounding as a consequence of misclassification in
the level of smoking, which biases the effect away
from the null, rather than an interaction between caffeine and tobacco.
The possibility that the observed association was
partly due to unmeasured confounders cannot be ex-
332
Bolumar et al.
eluded. However, an important aspect of this study is
that we controlled for the major, known risk factors for
infertility (20, 25, 27). In addition, other potential
confounders such as working status, frequency of intercourse, and prevalence of pelvic inflammatory diseases were also included in multivariate analysis.
Changes in trends of coffee consumption over time
could also bias the results, but the magnitude of the
association was not modified after including in the
model the calendar year of the starting time period.
Nutritional status, stress, or exercise may be potential
confounders on which we had no data.
If the waiting TTP is long, a woman might modify
spontaneously her consumption of caffeinated beverages, and therefore the recording of caffeine intake at
the time of conception might be biased as a measurement of exposure (36). This was avoided in our study,
however, since caffeine intake was estimated at the
starting time of the first episode.
Differential recall, due to marked differences in the
time elapsed from the starting time period up to the
interview date, might be a source of misclassification
bias. In the first pregnancy, this time ranged between
1 month and 31 years; but the mean time was approximately the same (11.9 years) for the different caffeine
categories, and it differed only in 1.5 years between
the fertile and subfertile women. However, since coffee consumption is not considered a reproductive hazard in Europe, any misclassification bias due to misreporting of caffeine consumption is most likely
nondifferential and of a random type. Such a misclassification will, in most situations, bias results toward
no effect values.
The exclusion of women with unplanned pregnancies might introduce bias since these women may be
more fertile and also have higher exposure to caffeine
and other lifestyle-related factors. Because the main
objective of this study was to determine whether caffeine intake could delay conception in women who
were trying to become pregnant, the definition of our
outcome measure, TTP, does not allow the proper
assessment of the effects of caffeine among women
with accidental pregnancies. Nevertheless, the average
daily caffeine consumption of women with planned
pregnancies (339.8 mg/day) did not significantly differ
from that of women with unplanned pregnancies
(346.2 mg/day), and both samples showed a similar
proportion of women with high caffeine intake (^500
mg/day). When the same multivariate analysis was
repeated adding those unplanned pregnancies in which
it was possible to estimate their TTP, the association
between caffeine consumption and infertility showed a
similar pattern (OR = 1.37, p = 0.035 for the highest
caffeine intake category).
Our results are consistent with those reported in a
previous Danish population-based survey (20), which
showed a negative effect (OR = 1.35) of caffeine on
fertility (TTP > 12 months), but only in women who
consumed eight or more cups of coffee/tea daily and
also smoked. We found a similar effect in smokers, but
at a lower level of caffeine consumption. In contrast to
our study, planned and unplanned pregnancies were
included, and the exposure measures were reported as
average intake before pregnancy instead of at the
starting time.
The results reported in two North American prospective studies (14, 18) and one retrospective study
(19) also found a positive association but of a larger
magnitude and at a lower level of consumption (from
about one to three cups of coffee per day). The different study design and the select groups of subjects
enrolled (volunteers, middle- and upper-class women,
or employees in semiconductor manufacturing facilities vs. population-based samples in our study) limit
the comparability with our results.
A large retrospective study (21) in pregnant women
reported no association between those who consumed
>7,000 mg of caffeine per month compared with
those who consumed ^500 mg per month. Another
study (22) concluded that no decrease in fertility occurred among coffee consumers of more than three
cups per day. The categorizations used in both studies
neither would have revealed an association in our own
study population nor would have allowed the possibility of detecting a risk related to a high caffeine intake.
Our study therefore has several strengths over previously published studies. The population was large and
selected at random from the general population. This
also allowed us to look for consistency of the effect in
different countries with varying levels of exposure.
Moreover, the analyses concerned the first pregnancy
and the most recent waiting time, which was not
restricted to women with proven fecundity. Frequency
of sexual intercourse was controlled in the analysis,
and our measurement of caffeine intake followed suggested recommendations for epidemiologic studies.
Finally, the exposure was determined at the beginning
of the waiting time and not during pregnancy.
Although some evidence has been gathered on an
association between coffee consumption and delayed
conception, a basic question still to be answered concerns the mechanism through which this association
occurs. Information on potential ethiopathogenic
mechanisms is scanty. It has been reported that caffeine decreases plasma levels of prolactin in nonpregnant healthy women (37). Infertility may be associated
with low serum prolactin concentration through
hypoprolactinemia-induced inhibition of corpus luAm J Epidemiol
Vol. 145, No. 4, 1997
Caffeine Intake and Infertility
teum function or by direct ovarian inhibition. Coffee
intake may also reduce levels of estrogens (38, 39), but
this pathway would operate in an opposite direction by
increasing gonadotropins and consequently ovarian
stimulation. Regardless of the nature of the specific
biologic mechanism involved, reduction of caffeine
intake would prevent its associated risk of delayed
conception if such a direct effect exists. However, if
instead of operating through a biologic effect caffeine
intake acts as a proxy for a stressful pattern of life, the
potential benefits of reducing caffeine intake by itself
are more doubtful. A more specific study aimed at
ruling out the effect of caffeine intake in relation to
other job- and life-related stress variables is needed.
10.
11.
12.
13.
14.
15.
16.
17.
ACKNOWLEDGMENTS
The European Study of Infertility and Subfecundity is a
European Community/Cooperation in Scientific and Technical Research (COST) Action Research Programme. Members of the project management group were S. Juul (Project
Leader), W. Karmaus, J. Olsen, T. Fletcher, F. Bolumar, I.
Figa-Talamanca, P. Thonneau, and S. Pautelakis.
Those responsible for collection of national data were W.
Karmaus (Germany), L. Bisanti and A. Spinelly (Italy), S.
Juul (Denmark), F. Bolumar (Spain), and R. Biczysko (Poland). The studies included in the Research Action were
funded by European Community contracts MR4/0205/DK
and MR4/0343/DK. Additional support was provided by
funds from each participating country (a Danish National
Research Foundation grant and grant FISS 92/0891E).
18.
19.
20.
21.
22.
23.
24.
25.
26.
REFERENCES
1. Arnaud MJ. Metabolism of caffeine and other components of
coffee. In: Garattini S, ed. Caffeine, coffee, and health. New
York, NY: Raven Press, 1993:43-95.
2. Fujii T, Sasaki N, Nishimura H. Teratogenicity of caffeine in
mice related to its mode of administration. Jpn J Pharmacol
1969;19:134-9.
3. Purves D, Sullivan FM. Reproductive effects of caffeine:
experimental studies in animals. In: Garattini S, ed. Caffeine,
coffee, and health. New York, NY: Raven Press, 1993:
317-42.
4. Daly JW. Mechanism of action of caffeine. In: Garattini S, ed.
Caffeine, coffee, and health. New York, NY: Raven Press,
1993:97-150.
5. Nishimura H, Nakai K. Congenital malformations in offspring
of mice treated with caffeine. Proc Soc Exp Biol Med 1960;
104:140-2.
6. Fujii T, Nishimura H. Adverse effects of prolonged administration of caffeine on rat fetus. Toxicol Appl Pharmacol 1972;
22:449-57.
7. Elmazar MMA, Me Elhatton PR, Sullivan FM. Studies on the
teratogenic effects of different oral preparations of caffeine in
mice. Toxicology 1982;23:57-71.
8. Palm PE, Arnold EP, Rachwall PC, et al. Evaluation of the
teratogenic potential of fresh-brewed coffee and caffeine in
the rat. Toxicol Appl Pharmacol 1978;44:1-16.
9. Nagasawa H, Sakurai N. Effects of chronic ingestion of cafAm J Epidemiol
Vol. 145, No. 4, 1997
27.
28.
29.
30.
31.
32.
33.
34.
35.
333
feine on mammary growth and reproduction in mice. Life Sci
1986;39:351-7.
Arnaud MJ. Comparative metabolic disposition of l-Me' 4 C
caffeine in rats, mice, and Chinese hamsters. Drug Metab
Dispos 1985;13:471-8.
Berger A. Effects of caffeine consumption on pregnancy
outcome: a review. J Reprod Med 1988;33:945-56.
Bonati M, Latini R, Tognoni G, et al. Interspecies comparison
of in vivo caffeine pharmacokinetics in man, monkey, rabbit,
rat, and mouse. Drug Metab Rev 1984-85;15:1355-83.
Oser BL, Ford RA. Caffeine: an update. Drug Chem Toxicol
1981,4:311-29.
Wilcox A, Weinberg CR, Baird D. Caffeinated beverages and
decreased fertility. (Letter). Lancet 1988;2:1453-5.
Christiansen RE, Oechsli FW, Van den Berg BJ. Caffeinated
beverages and decreased fertility. (Letter). Lancet 1989; 1:378.
Williams MA, Monson RR, Goldman MB, et al. Coffee and
delayed conception. Lancet 199O;335:16O3.
Grodstein F, Goldman MB, Ryan L, et al. Relation of female
infertility to consumption of caffeinated beverages. Am J
Epidemiol 1993;137:1353-60.
Hatch EE, Bracken MB. Association of delayed conception
with caffeine consumption. Am J Epidemiol 1993,138:
1082-92.
Stan ton CK, Gray RH. Effects of caffeine consumption on
delayed conception. Am J Epidemiol 1995; 142:1322-9.
Olsen J. Cigarette smoking, tea and coffee drinking, and
subfecundity. Am J Epidemiol 1991 ;133:734-9.
Joesoef MR, Beral V, Rolfs RT, et al. Are caffeinated beverages risk factors for delayed conception? Lancet 199O;335:
136-7.
Alderete E, Eskenazi B, Sholtz R. Effect of cigarette smoking
and coffee drinking on time to conception. Epidemiology
1995;6:403-8.
Florack EIM, Zielhuis GA, Rolland R. Cigarette smoking,
alcohol consumption, and caffeine intake and fecundability.
Prev Med 1994;23:175-80.
Schreiber GB, Maffeo CE, Robins M, et al. Measurement of
coffee and caffeine intake: implications for epidemiologic
research. Prev Med 1988; 17:280-94.
Schreiber GB, Robins M, Maffeo CE, et al. Confounders
contributing to the reported associations of coffee or caffeine
with disease. Prev Med 1988;17:295-309.
Leviton A. Coffee, caffeine and reproductive hazards in humans. In: Garattini S, ed. Caffeine, coffee, and health. New
York, NY: Raven Press, 1993:343-58.
Dlugosz J, Bracken MB. Reproductive effects of caffeine: a
review and theoretical analysis. Epidemiol Rev 1992; 14:
83-100.
D'Amicis A, Viani R. The consumption of coffee. In: Garattini S, ed. Caffeine, coffee, and health. New York, NY: Raven
Press, 1993:1-16.
Bolumar F, Olsen J, Boldsen J, et al. Smoking reduces
fecundity: a European multicenter study on infertility and
subfecundity. Am J Epidemiol 1996;143:578-87.
Maldonado G, Greenland S. Simulation study of confounderselection strategies. Am J Epidemiol 1993; 138:923-36.
Norusis MJ. SPSS/PC+ V2.0. Chicago, IL: SPSS Inc., 1988.
Yesair DW, Branfman AR, Callahan MM. Human disposition
and some biochemical aspects of methylxanthines. In: Spiller
GA, ed. The methylxanthine beverages and foods: chemistry,
consumption and health effects. New York, NY: Alan R Liss,
1984.
Little RE, Uhl CN, Labbe RF, et al. Agreement between
laboratory tests and self-reports of alcohol, tobacco, caffeine,
marijuana and other drug use in post-partum women. Soc Sci
Med 1986;22:9I-8.
Kennedy JS, von Moltke LL, Harmatz JS, et al. Validity of
self-reports of caffeine use. J Clin Pharmacol 1991 ;31:
677-80.
Brown CR, Jacob P, Wilson M, et al. Changes in rate and
334
Bolumar et al.
pattern of caffeine metabolism after cigarette abstinence. Clin
Pharmacol Ther 1988;43:488-91.
36. Weinberg CR, Baird DD, Wilcox AJ. Sources of bias in
studies of time to pregnancy. Stat Med 1994;13:671-81.
37. CasasM, Ferrer S, Calaf J, etal. Dopaminergic mechanism for
caffeine-induced decrease in fertility? Lancet 1989;1:731.
38. Petridou E, Katsouyanni K, Spanos E, et al. Pregnancy estrogens in relation to coffee and alcohol intake. Ann Epidemiol
1992;2:241-7.
39. Kapidaki M, Roupa Z, Sparos L, et al. Coffee intake and other
factors in relation to multiple deliveries: a study in Greece,
Epidemiology 1995;294-8.
Am J Epidemiol
Vol. 145, No. 4, 1997