Relation of caffeine intake and blood caffeine concentrations during pregnancy to fetal growth: prospective population based study

BMJ 1996; 313 doi: (Published 30 November 1996) Cite this as: BMJ 1996;313:1358
  1. Derek G Cook, reader in epidemiologya,
  2. Janet L Peacock, lecturer in medical statisticsa,
  3. Colin Feyerabend, principal biochemistb,
  4. Iain M Carey, statisticiana,
  5. Martin J Jarvis, reader in health psychologyc,
  6. H Ross Anderson, professor of epidemiology and public healtha,
  7. J Martin Bland, professor of medical statisticsa
  1. a Department of Public Health Sciences, St George's Hospital Medical School, London SW17 0RE
  2. b Medical Toxicology Unit, New Cross Hospital, London
  3. c Imperial Cancer Research Fund Health Behaviour Unit, Department of Epidemiology and Public Health, University College, London
  1. Correspondence to: Dr Cook.
  • Accepted 24 September 1996


Objectives: To examine the association of plasma caffeine concentrations during pregnancy with fetal growth and to compare this with relations with reported caffeine intake.

Design: Prospective population based study.

Setting: District general hospital, inner London.

Subjects: Women booking for delivery between 1982 and 1984. Stored plasma was available for 1500 women who had provided a blood sample on at least one occasion and for 640 women who had provided a sample on all three occasions (at booking, 28 weeks, and 36 weeks).

Main outcome measure: Birth weight adjusted for gestational age, maternal height, parity, and sex of infant. The exposures of interest were reported caffeine consumption and blood caffeine concentration. Cigarette smoking was assessed by blood cotinine concentration.

Results: Caffeine intake showed no changes during pregnancy, but blood caffeine concentrations rose by 75%. Although caffeine intake increased steadily with increasing cotinine concentration above 15 ng/ml, blood caffeine concentrations fell. Caffeine consumption was inversely related to adjusted birth weight, the estimated effect being a 1.3% fall in birth weight for a 1000 mg per week increase in intake (95% confidence interval 0.5% to 2.1%). The apparent caffeine effect was confined to cigarette smokers, among whom the estimated effect was −1.6%/1000 mg a week (−2.9% to −0.2%) after adjustment for cotinine and −1.3% (−2.7% to 0.1%) after further adjustment for social class and alcohol intake. Adjusted birth weight was unrelated to blood caffeine concentrations overall (P = 0.09, but a positive coefficient), after adjustment for cotinine (P = 0.73), or among current smokers (P = 0.45).

Conclusions: Smokers consume more caffeine than non-smokers. Blood caffeine concentrations during pregnancy are not related to fetal growth, but caffeine intake is negatively associated with birth weight, with this effect being apparent only in smokers. The effect remains of borderline significance after adjustment for other factors. Prudent advice for pregnant women would be to reduce caffeine intake in conjunction with stopping smoking.

Key messages

  • Blood caffeine concentrations during pregnancy are not related to birth weight

  • Caffeine intake assessed by questionnaire is negatively associated with birth weight, with evidence that this effect is apparent only in smokers

  • As most studies have limited power to detect an interaction between smoking and effects of caffeine intake on birth weight, a meta-analysis of existing studies is recommended

  • It seems reasonable to advise women who smoke to reduce their caffeine intake as well as to stop smoking during pregnancy


Caffeine is commonly consumed during pregnancy, but elimination from the blood is slowed.1 Fetal concentrations are believed to be in equilibrium with maternal concentrations.1 Although it is biologically plausible that caffeine consumption could adversely affect the outcome of pregnancy, the epidemiological evidence is inconsistent, though the strongest evidence is for an effect on intrauterine growth.2

The lack of consistency may reflect methodological differences.3 A major weakness of all previous studies is that they have relied on reported consumption of caffeine. Such questionnaire measures are unreliable, and intake may be a poor reflection of blood concentrations. A second weakness is a reliance on self reported smoking status. Smoking is known to increase caffeine metabolism appreciably,4 and recent reports have emphasised that cotinine is a better predictor of reduced birth weight than questionnaire measures of smoking,4 5 6 raising the possibility of residual confounding.7 We previously reported that caffeine intake, estimated by questionnaire, was inversely related to fetal growth, with some evidence that the effect was present only in smokers.8 We have now measured cotinine and caffeine concentrations in stored plasma to address two issues: (a) whether blood caffeine concentrations are related to fetal growth and (b) whether the inverse association between caffeine intake and fetal growth are due to inadequate adjustment for cigarette smoking.


The study has been described in detail elsewhere.9 At St George's Hospital, a teaching hospital in south London, 1860 white women booking for antenatal care were invited to participate. These women represented consecutive bookings between August 1982 and March 1984, excluding those who spoke insufficient English, booked after 24 weeks, had insulin dependent diabetes, or had a multiple pregnancy. Women were interviewed at booking (mean 14 weeks) and at 28 and 36 weeks by trained researchers using a structured questionnaire. Whenever possible a blood sample was taken and plasma stored at −80°C until assayed for cotinine and caffeine in 1994. Blood samples were not collected at standard times. The assays were sensitive and specific, using gas chromatography with nitrogen and phosphorous detection respectively (C Feyerabend et al, unpublished data).10 The detection limits of the assays were 0.1 ng/ml for cotinine and 0.01 μg/ml for caffeine. Nondetectable concentrations were coded as 0.05 ng/ml and 0.005 μg/ml respectively.

Caffeine intake at the three points in pregnancy was defined as the number of cups of tea, coffee, cocoa, and cola drunk in the previous week. These were converted to milligrams of caffeine by using estimates for tea and coffee of 70 mg/cup and 92 mg/cup respectively11 and for cocoa and cola of 5 mg/cup and 40 mg/serving respectively.12 On each occasion women were also asked about their alcohol consumption in the preceding week.8 Social class of head of household was coded according to the Registrar General's classification.13

Obstetric data and fetal outcome were obtained from the structured obstetric record.9 Gestational age at delivery was calculated from date of delivery, dates of menstruation, and early ultrasound examination. The outcome measure for this analysis was a birth weight ratio adjusted for gestational age, maternal height, sex of infant, and parity of mother. The adjustment was carried out in two stages.14 Firstly, the birth weight was adjusted for gestational age by taking the ratio of the observed birth weight to the expected birth weight for that week of gestational age at birth using an external standard. The resultant birth weight ratio was then adjusted for the other biological factors with multiple regression. This gave an adjusted birth weight ratio suitable for use as the outcome variable in a linear regression model. As all the mean adjusted birth weight ratios presented are close to 1.0, differences between the ratios are equivalent to percentage differences—for example, the difference between the birth weight ratios 1.04 and 1.01 is 0.03, which is about a 3% difference in the two mean birth weights.

Because of the relatively short half life of caffeine, a single measurement of the blood concentration will not provide an accurate measure of average exposure. Most analyses in this report are thus based on the women for whom measurements were available on all three occasions. Average caffeine intake was defined as the mean intake estimated from the three questionnaires administered at booking, 28 weeks, and 36 weeks. Average blood concentrations of caffeine during pregnancy were defined as the geometric mean blood caffeine concentration; the geometric mean was used because the dispersion of blood caffeine increased with increasing concentration.

Statistical analyses were carried out using the SAS statistical package (SAS Institute, North Carolina). The GLM procedure was used to fit multiple regression models with the adjusted birth weight ratio as the outcome variable. A cut off of about 15 ng/ml cotinine seems optimal for distinguishing smokers from nonsmokers15 16 and was used to divide women into two exposure groups, smokers and non-smokers. To test formally for trends in the birth weight ratio with caffeine intake and blood caffeine concentration we regressed the ratio on caffeine intake or blood caffeine concentration, with intake and concentrations treated as continuous variables. Adjustments for the effect of smoking were made by including both log(cotinine) and its square in a regression model as there was evidence that the relation with log(cotinine) was not linear. Additional adjustment included alcohol as a continuous variable and social class as a factor with seven levels.



Of 1860 women who were invited, 1724 (93%) took part in the study. At least one blood sample was available for 1500 women. Complete data, including birth weight and questionnaire, were available on all three occasions (at booking, 28 weeks, and 36 weeks) for 640 women.


Blood caffeine concentrations at booking and at 28 weeks were moderately correlated (r = 0.50) as were the measurements at 28 and 36 weeks (r = 0.58). The corresponding correlations for the intake data were both 0.67. At each time point the distribution of blood caffeine measurements was highly positively skewed and showed increasing variability with increasing intake.

Table 1 presents the mean blood caffeine and the mean caffeine intake by occasion during pregnancy for those women for whom measurements were available at all three occasions. Although intake showed no particular pattern, the blood levels increased by 75% from 2.35 μg/l at booking to 4.12 μg/l at 36 weeks.

Table 1

Blood caffeine concentrations (μg/ml) and estimated caffeine intake (mg/week) for 640 women at three time points during pregnancy

View this table:


A clear positive relation existed between average blood caffeine concentrations and average caffeine intake both in smokers and non-smokers (fig 1), but with a wide range of blood concentrations at each intake level. Moreover, at each intake level, blood concentrations were lower in smokers than in non-smokers. Figure 2 shows the effect of cigarette smoking on caffeine metabolism. While caffeine intake increased steadily with increasing cotinine above 15 ng/ml, blood caffeine concentrations fell.

Fig 1
Fig 1

Median, 10th, and 90th centiles of blood caffeine concentration by caffeine intake

Fig 2
Fig 2

Mean blood caffeine concentrations and caffeine intake (with 95% confidence intervals) by blood cotinine concentration


Overall, the birth weight ratio decreased with increasing caffeine intake (table 2). Regressing the birth weight ratio on caffeine intake yielded a regression coefficient of −1.29%/g per week (95% confidence interval −2.05% to −0.53%), which was halved to −0.60%/g per week when cotinine was adjusted for (table 3). The inverse relation seemed, however, to be present only among smokers. Table 3 shows the effect of adjusting for confounding variables. The pattern remained unchanged after cotinine was adjusted for; the regression slope in non-smokers was −0.06%/g per week (−1.04% to 0.92%, P = 0.90), whereas in smokers it was −1.55%/g per week (−2.86% to −0.24%, P = 0.02). A formal test of the equivalence of the two slopes provided some evidence that they differed even after adjustment (P = 0.08). Further adjustment for alcohol and social class on slightly reduced numbers (n = 617) left a regression slope in smokers of −1.33%/g per week (−2.72% to 0.06%, P = 0.06).

Table 2

Birth weight ratio by caffeine consumption and blood caffeine concentration in all women and in non-smokers and smokers separately

View this table:
Table 3

Summary of regression coefficients of birth weight ratio on caffeine intake and blood caffeine concentration in all women and in non-smokers and smokers separately and with adjustment for confounding variables. Values are percentage changes in birth weight ratio/g caffeine intake/week and in birth weight/μg caffeine/ml blood

View this table:


In contrast, adjusted birth weight showed no relation with blood caffeine either overall or in current cigarette smokers (table 2). When the birth weight ratio was regressed on average blood caffeine concentrations while cotinine was controlled for, there was no evidence that the slopes differed in smokers and non-smokers (P = 0.35) (table 3), though the regression slope among smokers was weakly negative after adjustment for cotinine at −0.57%/μg/ml (P = 0.47). Using the log of blood caffeine in the regression analyses did not alter these findings.


We examined the effect of restricting our analyses to women for whom we had blood measurements on all three occasions by: (a) carrying out regressions on women with blood samples available at booking (n = 1138); (b) carrying out three separate regressions that included women for whom we had all three, only two, or only one blood measurement (n = 1500). Data (not shown) did not suggest that looking more widely at women for whom we had fewer than three measurements (including preterm births) in any way altered our conclusions on blood caffeine concentration or that blood caffeine had different effects at different times in pregnancy. For caffeine intake the relations with birth weight also remained closely similar. Based on subjects with data on intake at booking, and after adjustment for cotinine, social class, and alcohol intake at booking, the evidence for a greater effect of caffeine intake in smokers than in non-smokers was of borderline significance (P = 0.07).


We found no relation between blood caffeine concentrations during pregnancy and birth weight. This contrasts with the negative association that we and others have found between reported intake of caffeine and birth weight. A 1992 review noted that 10 out of 13 studies had reported a negative association, though not all were significant.2


A key element in understanding these apparently contradictory findings comes from recognising the importance of factors other than caffeine intake in determining blood concentrations. The metabolism of caffeine is known to slow during the course of pregnancy,1 and the effect of this is clearly seen in our data with a rise in blood concentrations from 2.35 μg/ml at booking to 4.12 μg/ml at 36 weeks, during which time intake changed little. More important is that cigarette smoking increases caffeine metabolism.17 We found that caffeine intake rose steadily with cotinine concentrations above 15 ng/ml whereas blood caffeine concentrations fell.


That blood caffeine concentrations are not associated with reduced fetal growth seems therefore to reinforce the view that the negative relation between caffeine intake and birth weight in previous studies might be due to inadequate control for the confounding effects of cigarette smoking.7 We used cotinine concentration rather than self reported smoking status to control for the effect of cigarette smoking because recent reports have suggested that cotinine is a better predictor of birth weight.4 5 6 In our study, however, the relation between reported caffeine intake and birth weight, although much reduced, remained of borderline significance despite adjustment for cotinine concentrations based on three measurements. Moreover, our previous suggestion that the effect of caffeine intake is stronger in or restricted to smokers8 remains and is supported by three other reports.18 19 20 Any factors not considered are unlikely to explain the relation of caffeine intake and reduced birth weight in smokers; in our study a wide range of social, psychological, and obstetric factors had little or no direct effect on fetal growth.9

Another explanation of why intake but not blood concentrations are related to birth weight arises if fetal blood concentrations are not in equilibrium with those in the mother, as is commonly supposed.1 The disposal of caffeine in newborn infants or by a fetus is very slow. High caffeine intake might result in raised fetal exposure, despite the high rate of metabolism in smokers. Alternatively it may be raised concentrations of some metabolite that are important. Only a trivial amount of caffeine is excreted unchanged, and the major metabolites of caffeine are pharmacologically active. It is known that smoking influences the demethylation processes involved in producing and eliminating these metabolites,1 and we know too little about these processes to rule out some biological interaction. Others have suggested that by blocking adenosine receptors, caffeine interferes with the normal physiological response to the raised carboxyhaemoglobin concentrations in smokers and thus exaggerates the effect of smoking on oxygen uptake.20 Such an explanation depends on the effect of smoking being due to its effect on carboxyhaemoglobin concentrations rather than being due to nicotine. In fact, studies of the effect of chewing tobacco during pregnancy suggest that nicotine has a direct effect.21 22

Given the limited power of individual studies to examine interactions between smoking and cotinine in their effect on fetal growth, data from all previous studies should be reviewed to establish whether such an interaction exists. Future studies should be designed with sufficient power to examine any biological interaction with the effect of cigarette smoking and should also include measurement of blood caffeine and its active metabolites.


In the absence of definitive evidence and given the widespread consumption of caffeinated drinks during pregnancy, two points need wider appreciation: (a) the metabolism of caffeine is appreciably slowed during pregnancy, leading to a pronounced rise in blood concentrations with no change in intake; and (b) smokers have a higher caffeine intake but a faster metabolism, resulting in lower blood concentrations. Thus anyone stopping smoking will show a pronounced rise in blood caffeine concentrations if their caffeine consumption remains unchanged.4 Prudent advice would seem to be to reduce caffeine intake in conjunction with stopping smoking.


  • Funding The cotinine and caffeine assays were funded by the Tobacco Products Research Trust. The original data collection was funded by a consortium of American tobacco companies.

  • Conflict of interest None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
View Abstract