in pregnancy and risk of spontaneous abortion, a case- controlled study was conducted in ... the causality is difficult and still open to debate. Key words: coffee ...
Human Reproduction vol.13 no.8 pp.2286–2291, 1998
Coffee consumption and risk of hospitalized miscarriage before 12 weeks of gestation
Fabio Parazzini1,2,4, Liliane Chatenoud1, Elisabetta Di Cintio1, Raffaella Mezzopane1, Matteo Surace1, Giovanni Zanconato3, Luigi Fedele3 and Guido Benzi2 1Istituto
di Ricerche Farmacologiche ‘Mario Negri’, via Eritrea, 62-20157 Milan, 2I Clinica Ostetrico Ginecologica, Universita` di Milano, Milan and 3Clinica Ostetrico Ginecologica, Universita` di Verona, Verona, Italy
4To
whom correspondence should be addressed at: Istituto di Ricerche Farmacologiche, ‘Mario Negri’, via Eritrea, 62-20157 Milan, Italy
In order to analyse the association between drinking coffee in pregnancy and risk of spontaneous abortion, a casecontrolled study was conducted in Milan, Northern Italy. Cases were 782 women with spontaneous abortion within the 12th week of gestation. The control group was recruited from women who gave birth at term (.37 weeks gestation) to healthy infants on randomly selected days at the same hospitals where cases had been identified: 1543 controls were interviewed. A total of 561 (72%) cases of spontaneous abortion and 877 (57%) controls reported coffee drinking during the first trimester of the index pregnancy. The corresponding multivariate odds ratios of spontaneous abortion, in comparison with non-drinkers, were 1.2, 1.8 and 4.0, respectively, for drinkers of 1, 2 or 3, and 4 or more cups of coffee per day. No relationship emerged between maternal decaffeinated coffee, tea and cola drinking in pregnancy, as well as paternal coffee consumption, and risk of spontaneous abortion. With regard to duration in years of coffee drinking, the estimated multivariate odds ratios of spontaneous abortion were, in comparison with non-coffee drinkers, 1.1 (95% confidence interval (CI) 0.9–1.4) and 1.9 (95% CI 1.5–2.6) for women reporting a duration of coffee consumption ø10 or .10 years. In conclusion, coffee drinking early in pregnancy was associated with an increased risk of abortion. This has biological implications, but epidemiological inference on the causality is difficult and still open to debate. Key words: coffee consumption/spontaneous abortion/risk factors
Introduction It has been suggested that drinking coffee during pregnancy increases the risk of spontaneous abortions. For example, an elevated late spontaneous abortion risk of about 70% was observed in heavy caffeine users in a prospective study of more than 3000 women in the USA (Srisuphan and Brachen, 2286
1986). Similar results emerged from at least one other cohort study conducted in Japan (Furuhashi et al., 1985). However, potential confounding factors, chiefly smoking and alcohol consumption, were not adequately taken into account in all studies. Furthermore, no relationship between caffeine consumption and spontaneous abortions emerged in a recent case-control study conducted in the USA on 607 cases and 1284 controls, and an apparent association between heavy caffeine consumption (.300 mg/day) disappeared after controlling for potential covariates (Fenster et al., 1991). To offer further information on the issue, data were analysed from a case-controlled study on risk factors for spontaneous abortion conducted in Milan, Northern Italy. Materials and methods Between 1990 and 1995, trained interviewers identified and questioned cases and controls using a structured questionnaire. The general design of the study has been published (Parazzini et al., 1994a,b, 1997). Cases were 782 women (median age 32 years) admitted for spontaneous abortion (within the 12th week of gestation) to the Clinica Luigi Mangiagalli (the largest obstetric hospital in Milan) and a network of obstetric departments in the greater Milan area. All miscarriages were confirmed by uterine curettage and pathological examination. Of these, 298 (38%) were between the 4th and the 8th week of gestation and 484 (62%) between the 9th and the 12th. The control group was recruited from women who gave birth at term (.37 weeks gestation) to healthy infants on randomly selected days at the same hospitals where cases had been identified. The control subjects were chosen within 1 month of case ascertainment. A total of 1543 control women (median age 30 years) was interviewed. They were not formally matched for age with cases, but the interviewers were aware of the requirement to interview cases and controls of comparable quinquennia of age. Information was obtained on general sociodemographic habits, personal characteristics and habits (including smoking and alcohol drinking), gynaecological and obstetrical history. Cases and controls were also asked about their consumption of coffee and other methylxanthine-containing beverages (tea, cola and also decaffeinated coffee) before pregnancy and during the first trimester of gestation. Drinkers who declared to have stopped coffee consumption at any time between last menstruation and spontaneous abortion, for cases, and last menstruation and 12 weeks of gestation, for controls were defined as women who quit drinking coffee during the first trimester of pregnancy. The usual paternal coffee consumption (number of cups per day) was collected by questioning the women only. Data on the epiphenomena of nausea were collected for each trimester of pregnancy in terms of intensity (none, low, moderate or serious) since June 1993. Therefore, only 354 (45%) cases and 743 (48%) controls had this information. Specific attention was paid to obtaining information on coffee (and other beverages) drinking in the first trimester of © European Society for Human Reproduction and Embryology
Coffee and risk of miscarriage
gestation in controls, too, who were interviewed after delivery. Less than 2% of eligible women refused to be interviewed. Data analysis The odds ratios were computed as estimators of the relative risks of spontaneous abortion for maternal and paternal coffee and other beverage drinking, together with their 95% approximate confidence intervals (CI), from data stratified for age by the Mantel–Haenszel procedure (Mantel and Haenszel, 1959). When a factor could be classified in more than two levels, the significance of the linear trend in risk was assessed by the Mantel test (Mantel, 1963). To account simultaneously for the effects of several potential confounding factors, unconditional multiple logistic regression was used, with maximum likelihood fitting to obtain the relative risks, their corresponding 95% CI and, when appropriate, tests for trend (Baker and Nelder, 1978). Included in the regression equations were terms for age, education, previous livebirths and spontaneous abortions, alcohol consumption, smoking and nausea intensity in the first trimester of pregnancy.
Results The distribution of cases and controls according to age, education, reproductive history, smoking habits and alcohol consumption in the first trimester of pregnancy is shown in Table I. Cases tended to be less educated, more frequently parae and more frequently reported previous spontaneous abortions than controls. Smokers during the first trimester of pregnancy were at a higher risk of miscarriage (odds ratio 1.7, 95% CI 1.3–2.1). In comparison with teetotallers, the ageadjusted odds ratio of miscarriage was 1.2 (95% CI 1.0–1.4) for women reporting one or more alcoholic drinks per day (chiefly wine) during the first trimester of pregnancy. As regards women with moderate or severe degree of nausea, in comparison with women without nausea, the odds ratios were 0.7 (95% CI 0.5–1.0) and 0.5 (95% CI 0.4–0.8), respectively. Cases tended to drink coffee more frequently than controls (Table II). In comparison with non-drinkers, the corresponding odds ratios for spontaneous abortion were 1.5 and 4.0 for consumption of four or more cups of coffee per day before and during the first trimester of pregnancy, respectively. In comparison with no coffee drinkers before and during pregnancy (122 cases and 323 controls), the multivariate odds ratio of miscarriage was 0.8 (95% CI 0.6–1.2) in the women who were coffee drinkers before pregnancy, but quitted coffee drinking during pregnancy. The relative risks of spontaneous abortions according to coffee consumption during the first trimester of the index pregnancy were further analysed in strata of age, education, reproductive history, smoking or alcohol drinking and nausea (Table III). There was no noteworthy interaction with any of the variables considered, as most of the estimated odds ratios were around 2 for drinkers of two or three cups, and 5 for drinkers of four or more cups of coffee per day. Maternal consumption of other methylxanthine-containing beverages, paternal coffee drinking and risk of abortion is considered in Table IV. No relationship emerged between decaffeinated coffee, tea and cola drinking and risk of spontaneous abortion. The odds ratio estimate for a father who drank
Table I. Distribution of 782a cases of spontaneous abortion and 1543 controls according to age, education, previous livebirths and spontaneous abortions, smoking and alcohol consumption during the first trimester of pregnancy (Milan, Italy 1990–1995) No. of spontaneous abortions
OR (95% CI) No. of controls
Age (years) ø20 20–24 25–29 30–34 35–39 ù40
16 86 239 285 133 23
20 164 452 610 258 39
– – – – – –
Education (years) ø7 8–12 ù13 χ21 trend
43 347 392
63 547 933
1d 1.02 (0.66–1.55) 0.68 (0.44–1.03) 15.62 P 5 0.0001
460 322
867 676
1d 0.86 (0.71–1.04)
Previous early miscarriages No 588 Yes 194
1284 259
1d 1.69 (1.36–2.11)
Previous livebirths 0 ù1
Smoking in the first trimester of current pregnancy No 615 1337 Yes 162 204
1d 1.65 (1.31–2.07)
Alcohol consumption in the first trimester of current pregnancy No 463 1019 1d Yes 319 524 1.19 (0.99–1.44) Nausea intensity in the first trimester of current pregnancyb None 164 274 1d Light 99 202 0.83 (0.61–1.14) Moderate 58 153 0.69 (0.48–1.00) Serious 33 114 0.54 (0.35–0.84) Missing 428 800 – aIn
some cases the sum does not add up to the total because of missing values. bBased on subjective evaluation of cases and controls. cMultiple logistic regression estimates including age, education, previous livebirths and miscarriages, maternal alcohol drinking and smoking in the first trimester of pregnancy and nausea intensity. dReference category. OR 5 odds ratio; CI 5 confidence interval.
coffee compared with a non-drinker was 1.3 (95% CI 0.9–1.9) and 1.1 (95% CI 0.8–1.5), respectively, for one and two or more cups per day. Discussion In this study women drinking coffee before and during the first trimester of pregnancy were at a higher risk of spontaneous abortion, and the risk increased with number of cups drunk per day. Likewise, duration of consumption longer than 10 years was associated with the risk of miscarriage. No relationship emerged between paternal coffee drinking and risk of miscarriage. Maternal decaffeinated coffee, tea and cola drinking was not significantly associated with the risk of miscarriages. In Italy, however, tea, cola and decaffeinated coffee consumption is limited. Consequently this study includes very few 2287
F.Parazzini et al.
Table II. Distribution of 782 cases of spontaneous abortion and 1543 controls, and corresponding relative risks according to indicators of maternal coffee consumption before and during the index pregnancy (Milan, Italy 1990–1995) Spontaneous abortions
Controls
OR (95% CI) M-Ha
MLRb
Coffee consumption before conception (cups/day) 0 122 323 1 143 307 2–3 385 705 ù4 132 208 χ21 trend
1c 1.25 1.48 1.74 15.25
Duration of coffee consumption (years)e ø10 326 .10 334 χ21 trend
1.21 (0.94–1.55) 2.14 (1.64–2.81) 33.67 P 5 0.0001
718 501
(0.93–1.66) (1.16–1.89) (1.28–2.36) P 5 0.0001
Coffee consumption during the first trimester of current pregnancy (cups/day) Non-drinkers before conception and during first trimester of pregnancy 122 323 1c Women who quit drinking in pregnancy 99 343 0.82 (0.60–1.10) 1 201 431 1.26 (4.00–1.65) 2–3 284 403 1.93 (1.49–2.51) ù4 76 43 4.74 (9.07–7.30) χ21 trendd 89.43 P 5 0.0001
1c 1.23 1.34 1.47 6.75
(0.91–1.64) (1.04–1.73) (1.07–2.02) P 5 0.0094
1.12 (0.87–1.44) 1.95 (1.48–2.58) 24.93 P 5 0.0001
1c 0.80 1.22 1.75 3.98 63.24
(0.59–1.09) (0.93–1.60) (1.23–2.29) (2.55–6.21) P 5 0.0001
aMantel–Haenszel estimates adjusted for age. bMultiple logistic regression estimates including
terms for age, education, previous livebirths and miscarriages, maternal alcohol drinking and smoking in the first trimester of pregnancy and nausea intensity. cReference category. dNon-drinkers before conception and women who quit drinking in pregnancy are considered together. eOne missing value in control group. OR5 odds ratio; CI5 confidence interval.
women reporting more than one serving per day of these beverages, and thus it is difficult to draw any conclusions on the effects of tea, cola and decaffeinated coffee on the risk of spontaneous abortion from the results of this study. Potential biases should be discussed. First of all, the time of data collection in cases and controls differed, cases being interviewed while in hospital for dilatation and curettage during the first trimester and controls while in hospital for delivery. However, a specific attempt was made to collect information on coffee consumption during the first trimester of pregnancy also from controls. This assessment of coffee consumption was based on self-reporting, so some underestimate could have occurred. In Italy, however, coffee consumption is normal and socially accepted (La Vecchia et al., 1987), and recommendations to avoid coffee in pregnancy have not received widespread attention and are not routinely advocated by gynaecologists. Information was not collected on the size of cups of coffee, but this should not be a problem in Italy, since the majority of women drink ‘espresso’ or ‘mocha’ (i.e. small cup unfiltered coffee). The usual paternal coffee consumption (number of cups per day) was collected by interviewing the women only; this information is probably acceptable for the purpose of this study. The choice as controls of women who delivered healthy infants at term may have introduced some potential bias. Coffee drinkers may give birth more frequently to low-birthweight infants (Narod et al., 1991). Since these women were 2288
not included in our control group, an artificially elevated coffee-related odds ratio of miscarriage may result. However, an analysis conducted considering separately as control group mothers delivering children weighing ,3500 g and ù3500 g did not show any difference in the odds ratio estimates. In fact, the odds ratios of spontaneous abortions, in comparison with non-coffee drinkers in pregnancy, for women drinking during the first trimester of pregnancy 1, 2 or 3, and 4 or more cups of coffee were 1.2, 2.2 and 4.4, respectively, and 1.2, 1.6 and 3.3, respectively, when the comparison group included mothers of children weighing ,3500 g and ù3500 g. In general, the interest of this analysis was to define the effect of coffee on the risk of miscarriage in comparison with normal outcome pregnancies, and the inclusion in the control group of women with adverse pregnancy outcomes related to coffee drinking would have spuriously underestimated the potential relationship between coffee and spontaneous abortion. In Italy, practically all clinical detectable abortions require hospital admission. However, a limitation of any hospitalbased case-control study on miscarriage is that subclinical abortions are excluded. In general, however, any definition and hence assessment of risk factors for subclinical miscarriage is difficult and, to date, no factor has been associated consistently with early, but not clinically diagnosed, abortion. The public health relevance of subclinical miscarriage is also unclear. Cases and controls were drawn from the same hospitals,
Coffee and risk of miscarriage
Table III. Odds ratiosa of spontaneous abortion according to maternal coffee consumption during the first trimester of pregnancy in strata of selected covariates (Milan, Italy 1990–1995) χ21 trend
Coffee cups per dayb,c 1
2–3
ù4
Age (years) ø25 26–30 .30
1.39 (31) 1.50 (79) 1.36 (91)
2.35 (42) 2.40 (101) 1.97 (141)
6.59 (12) 3.15 (16) 6.69 (48)
17.55 P 5 0.0001 26.16 P 5 0.0001 45.92 P 5 0.0001
Education (years) ,12 ù12
1.26 (84) 1.62 (117)
2.36 (144) 2.06 (140)
6.70 (47) 4.23 (29)
54.18 P 5 0.0001 34.76 P 5 0.0001
1.50 (126) 1.30 (75)
2.34 (158) 1.99 (126)
4.97 (37) 6.04 (39)
52.27 P 5 0.0001 38.31 P 5 0.0001
Previous early miscarriages No 1.37 (147) Yes 1.63 (54)
2.29 (218) 1.85 (66)
4.96 (54) 9.24 (22)
71.49 P 5 0.0001 18.46 P 5 0.0001
Alcohol consumption in the first trimester of current pregnancy No 1.43 (121) 2.00 (141) 5.15 (40) Yes 1.33 (80) 2.22 (143) 5.55 (36)
46.88 P 5 0.0001 34.87 P 5 0.0001
Smoking in the first trimester of current pregnancy No 1.38 (161) 2.25 (213) Yes 1.26 (40) 1.25 (67)
4.44 (42) 5.02 (34)
64.12 P 5 0.0001 9.30 P 5 0.0023
Nausea intensity in the first trimester of current pregnancy None 1.49 (50) 1.65 (62) 2.86 (16) Light 1.34 (34) 1.20 (33) 2.40 (6) Moderate/serious 1.49 (27) 1.48 (22) 4.35 (7)
6.71 P 5 0.0096 1.10 P 5 0.2933 8.98 P 5 0.003
Previous livebirths 0 ù1
aAdjusted for age by bReference category: cNumbers
the Mantel–Haenszel procedure. non-drinkers during the first trimester of pregnancy. of cases are given in parentheses.
Table IV. Distribution of 782 cases of spontaneous abortion and 1543 controls and corresponding relative risks according to maternal consumption during the first trimester of the index pregnancy of selected methylxanthinecontaining beverages and paternal coffee drinking (Milan, Italy 1990–1995) Spontaneous Controls abortions
Consumption during the first trimester of pregnancy Decaffeinated coffee (cups/day) 0 755 1485 ù1 27 58 Tea (cups/day)
OR (95% CI)a M-Hb
MLRc
1d 0.92 (0.58–1.47)
1d 1.02 (0.63–1.64)
0 ù1 Cola (drinks/day)
544 238
1045 498
1d 0.91 (0.76–1.10)
1d 0.92 (0.76–1.12)
0 ù1
721 61
1425 118
1d 1.01 (0.73–1.39)
1d 0.97 (0.69–1.35)
Paternal coffee drinking (cups/day) 0 1 ù2 χ21 trend
87 122 531
198 222 1078
1d 1.26 (0.90–1.76) 1.13 (0.86–1.49) 0.23 P 5 0.6299
1d 1.33 (0.95–1.88) 1.10 (0.83–1.46) 0.0019 P 5 0.9649
aOR5 odds ratio; CI5 confidence interval. bMantel–Haenszel estimates adjusted for age. cMultiple logistic regression estimates including terms for age, education, previous livebirths and miscarriages, maternal alcohol drinking, smoking in the first trimester of pregnancy and nausea intensity. dReference category.
2289
F.Parazzini et al.
and the participation rate was practically complete for cases and controls. Finally, allowance for potential distorting factors, including socioeconomic status and other major known or potential determinants of miscarriages, did not markedly change the estimated relative risk. Residual confounding by smoking is possible (Savitz and Baron, 1989), but is probably not of substantial importance in this analysis. In fact, the effect of coffee on the risk of miscarriage was present both in smokers and non-smokers. This gives support to the relationship between coffee and miscarriage risk. No information was collected regarding coffee consumption in previous pregnancies, thus the relationship between coffee consumption and outcome of previous pregnancies in subjects included in this study could not be analysed. Nausea in early months of pregnancy may cause a reduction in coffee consumption (Fenster et al., 1991; Stein and Susser, 1991), and nausea is less frequent in pregnancies which miscarry (Medalie, 1957). Thus, it has been suggested that coffee consumption could be considered an indicator, rather than a cause, of poor pregnancy outcome. Furthermore, the relationship between coffee consumption, nausea and risk of miscarriage could be complicated by the effect of coffee in reducing steroid concentrations in pregnancy (Petridou et al., 1992), since low oestrogen concentrations have been associated with decreased nausea and increased risk of poor outcome of gestation (Fairweather, 1968). In spite of these considerations, no interaction between nausea and coffee drinking emerged in this study even if the history of nausea had been timed within the same period as the history of coffee consumption (Stein and Susser, 1991). Available evidence on the relationship between coffee consumption in pregnancy and the risk of spontaneous abortions is not completely consistent. No association emerged in a prospective study conducted in Canada (Watkinson and Fried, 1985), and no difference in caffeine intake in pregnancy was observed in women who miscarried chromosomally normal or aberrant embryos (Kline et al., 1991). Furthermore, a higher risk of abortion in coffee drinkers disappeared after taking into account potential confounding factors in a case-control study conducted in California (Fenster et al., 1991). Conversely, an increased risk of miscarriage in coffee drinkers emerged in three studies (Weathersbee et al., 1977; Furuhashi et al., 1985; Srisuphan and Brachen, 1986) and in a case-control study on recurrent abortions, although in the latter study the finding was not statistically significant (Parazzini et al., 1991). In these studies the odds ratio estimates for higher level of exposure, defined by intake of two coffee servings per day to five or more, were about 2 to 3, i.e. largely consistent with the present results. Another case-control study found a positive association between coffee intake, both before and during pregnancy, and risk of miscarriage (Infante-Rivard et al., 1993). It is difficult to interpret these differences in terms of type of preparation of coffee in different studies. In fact, differences also emerged in studies conducted in the same countries, where differences in strength of coffee preparation should, if any, be limited. This study shows an association between coffee drinking before and during pregnancy. However, the association 2290
observed between coffee drinking before conception and risk of spontaneous abortion seems largely driven by women who did not stop coffee drinking in pregnancy. In fact the estimated odds ratio of spontaneous abortion, in comparison with women reporting no coffee drinking before conception, was 1.2 (not significant), for women who ceased coffee drinking during the first trimester of pregnancy. This suggests that the negative effect of coffee on the pregnancy outcome acts during pregnancy. In biological terms, coffee is mutagenic in human cell cultures (Ostertag et al., 1965) and has been associated with increased rates of malformation and resorptions in rodents (Collins, 1981; Fenster et al., 1991) and miscarriages in monkeys (Gilbert et al., 1988). Coffee raises cyclic 39,59adenosine monophosphate in cells and consequently interferes with embryological development (Morris and Weinstein, 1981), besides decreasing oestrogen concentrations in pregnancy (Petridou et al., 1992). In conclusion, in this study, coffee drinking in pregnancy was associated with increased abortion risk, and this has possible biological implications (Purves and Sullivan, 1993), but epidemiological inference on causality is difficult and still open to debate.
Acknowledgements This work was conducted within the framework of the CNR (Italian National Research Council) Applied Project ‘Fattori di Rischio di Malattia’ (subproject ‘Fattori di Malattia nella Patologia Materno Infantile’). The generous contribution of Mrs Angela Marchegiano Borgomainerio is gratefully acknowledged. The authors thank Mr Flavio Mari and Marina Parazzini for their help in data management and analysis. Ms Ivana Garimoldi, Judy Baggott and the G.A. Pfeiffer Memorial Library Staff provided helpful editorial assistance.
References Baker, R.J. and Nelder, J.A. (1978) The GLIM system. Release 3. Numerical Algorithms Group, Oxford. Collins, T.F.X. (1981) Update to review of caffeine studies. FDA By-Lines, 1, 19–24. Fairweather, D.V.I. (1968) Nausea and vomiting during pregnancy. Obstet. Gynecol. Ann., 7, 91–105. Fenster, L., Eskenazi, B., Windham, G.C. et al. (1991) Caffeine consumption during pregnancy and spontaneous abortion. Epidemiology, 2, 168–174. Furuhashi, N., Sato, S., Suzuki, M. et al. (1985) Effects of caffeine ingestion during pregnancy. Gynecol. Obstet. Invest., 19, 187–191. Gilbert, S.C., Rice, D.C., Reuhl, K.R. et al. (1988) Adverse pregnancy outcome in the monkey (Macaca fascicularis) after chronic caffeine exposure. J. Pharmacol. Exp. Ther., 245, 1048–1053. Infante-Rivard, C., Fernandez, A., Gauthier, R. et al. (1993) Fetal loss associated with caffeine intake before and during pregnancy. JAMA, 270, 2940–2943. Kline, J., Levin, B., Silverman, J. et al. (1991) Caffeine and spontaneous abortion of known karyotype. Epidemiology, 2, 409–417. La Vecchia, C., Pagano, R., Negri, E. et al. (1987) Determinants of alcohol consumption in Italy. Int. J. Epidemiol., 16, 295–296. Mantel, N. (1963) Chi-square tests with one degree of freedom: Extensions of the Mantel–Haenszel procedure. J. Am. Stat. Ass., 58, 690–700. Mantel, N. and Haenszel, W. (1959) Statistical aspects of the analysis of data from retrospective studies of disease. J. Natl. Cancer Inst., 22, 719–748. Medalie, J.H. (1957) Relationship between nausea and/or vomiting in early pregnancy and abortion. Lancet, 29, 117–119. Morris, M.B. and Weinstein, L. (1981) Caffeine and the fetus: is trouble brewing? Am. J. Obstet. Gynecol., 140, 6.
Coffee and risk of miscarriage Narod, S.A., de Sanjose`, S. and Victora, C. (1991) Coffee during pregnancy: A reproductive hazard? Am. J. Obstet. Gynecol., 164, 1109–1114. Ostertag, W.E., Duisberg, E. and Sturmann, M. (1965) The mutagenic activity of caffeine in man. Mutat. Res., 2, 293–296. Parazzini, F., Bocciolone, L., Fedele, L. et al. (1991) Risk factors for spontaneous abortion. Int. J. Epidemiol., 20, 157–161. Parazzini, F., Chatenoud, L. and La Vecchia, C. (1994a) Fetal loss and caffeine intake. JAMA, 272, 28. Parazzini, F., Tozzi, L., Chatenoud, L. et al. (1994b) Alcohol and risk of spontaneous abortion. Hum. Reprod., 9, 1950–1953. Parazzini, F., Chatenoud, L., Tozzi, L. et al. (1997) Determinants of risk of spontaneous abortions in the first trimester of pregnancy. Epidemiology, 8, 681–683. Petridou, E., Katsouyanni, K., Spanos, E. et al. (1992) Pregnancy estrogens in relation to coffee and alcohol intake. Ann. Epidemiol., 2, 241–247. Purves, D. and Sullivan, F.M. (1993) Reproductive effects of caffeine. Experimental studies in animals. In Garattini, S. (ed.) Caffeine, Coffee, and Health. Raven Press, New York, pp. 317–342. Savitz, D.A. and Baron, A.E. (1989) Estimating and correcting for confounder misclassification. Am. J. Epidemiol., 129, 1062–1071. Srisuphan, W. and Brachen, M.B. (1986) Caffeine consumption during pregnancy and association with late spontaneous abortion. Am. J. Obstet. Gynecol., 154, 14–20. Stein, Z. and Susser, M. (1991) Miscarriage, caffeine, and the epiphenomena of pregnancy: The causal model. Epidemiology, 2, 163–167. Watkinson, B. and Fried, P.A. (1985) Maternal caffeine use before, during and after pregnancy and effects upon offspring. Neurobehav. Toxicol. Teratol., 7, 9–17. Weathersbee, P., Osen, L. and Lodge, J.R. (1977) Caffeine and pregnancy: A retrospective study. Postgrad. Med., 62, 64–69. Received on October 31, 1997; accepted on April 28, 1998
2291