Our results indicate that the ingestion of caffeine during early pregnancy is associated with an increased risk of first-trimester spontaneous abortion of a fetus with a normal karyotype and that the increase in risk associated with caffeine is consistently present only among nonsmokers. Our results also suggest that the finding of the association between caffeine ingestion and spontaneous abortion is not entirely a product of bias introduced by pregnancy-induced symptoms.6,7,14
Measuring plasma cotinine allowed us to adjust for the misclassification of women according to smoking status. We found, in agreement with most previous studies,20,21 that smoking was associated with an increased risk of spontaneous abortion, but we were unable to detect an effect of caffeine among smokers. The effect of maternal smoking may conceal an effect of caffeine on the risk of spontaneous abortion. Alternatively, because smoking increases the rate at which caffeine is eliminated,22,23 nonsmokers may be particularly susceptible to its effects. A possible interaction between caffeine and smoking with regard to the risk of spontaneous abortion has been suggested previously.13 A similar interaction between caffeine and smoking, with caffeine affecting primarily nonsmokers, has been reported with regard to the risk of delayed conception.24 In contrast, caffeine has usually,25,26 but not always,27 been associated with reduced fetal growth, primarily among smokers. The metabolism of caffeine is markedly slower in late pregnancy than in early pregnancy,26 but the reasons for these differences in interactions between caffeine and smoking with respect to reproductive outcomes in early and late pregnancy remain speculative.
We probably limited both errors in the measurement of exposure by using face-to-face interviews close to the time of the spontaneous abortion in women who had such abortions and by matching the control subjects with the women with spontaneous abortion according to week of gestation. However, data on the ingestion of caffeine close to the time of spontaneous abortion could unintentionally include caffeine ingestion that occurred after fetal death, and such exposure should not be considered in a causal context.14 The most recent period for which caffeine ingestion was reported was the last completed week of gestation, thus there was an average of 3.5 days between the last measurement of caffeine ingestion and the time of spontaneous abortion.
The proportion of women who increased their intake of caffeine by at least 20 percent during the last two weeks of gestation was small (4 percent) in both the case and control groups, but we found that when the intake of caffeine during the last two weeks was not included in the analysis, the risk of spontaneous abortion of fetuses with normal karyotypes was reduced among nonsmokers. This analysis, hampered by limited statistical power, does not allow us to rule out the possibility of spurious positive results due to increased ingestion of caffeine among the case patients in response to lessening of the severity of symptoms. It is therefore reassuring that the ingestion of caffeine was unrelated to the spontaneous abortion of fetuses with abnormal karyotypes. The results of karyotyping were not known at the time of the interview, and the effect of systematic reporting errors would have applied equally to both fetal-karyotype groups. That the ingestion of caffeine did not influence the risk of spontaneous abortion of a fetus with an abnormal karyotype is in agreement with the lack of evidence regarding the potential mutagenicity of caffeine in humans. However, it is in contrast to the report that caffeine ingestion increased the risk of spontaneous abortions of fetuses with normal and abnormal karyotypes during the second trimester.10
Coffees brewed in Sweden have high concentrations of caffeine as well as many other constituents,28 and it is not known whether these constituents are possible risk factors for spontaneous abortion. The few studies, including our own, that have attempted to estimate the relative risks associated with caffeine from specific sources have had limited power.5,7,12,13 Nonetheless, our data suggest that caffeine from coffee has the same general effect as caffeine from other sources.
Our study corroborates many earlier reports that found an increased risk of spontaneous abortion among pregnant women who ingested caffeine.4,8,10,12,13 Recently, Klebanoff et al.9 reported that serum paraxanthine, a caffeine metabolite, was associated with an increased risk of spontaneous abortion only at very high concentrations (at or above the 95th percentile among the women they studied). Problems relating serum paraxanthine concentrations to caffeine intake, especially given the considerable variation among people in the rate of caffeine metabolism, make direct comparison of their results with ours impossible. However, their case group consisted mostly of women who had spontaneous abortions in the second trimester, and caffeine may be more detrimental in the first than in the second trimester, because the exposure of the fetus can be assumed to be greater in that period than later in the pregnancy.25,29
Given the limitations of the current study, and the mixed results of previous studies, our findings should be interpreted cautiously. However, reducing caffeine intake during early pregnancy may be prudent.
Supported by the International Epidemiology Institute through a grant from the National Soft Drink Association.