Subjects and samples

All pregnant women aged 18 years old or over, attending the antenatal clinic at Leeds General Infirmary (Leeds, UK) were eligible. Pregnant women younger than 18 years of age or who were receiving their maternity care elsewhere were not eligible to take part. Maternity records of women attending the clinic were checked for suitable women. Women were approached by the research assistant, informed about the study, and given an information sheet. Women who agreed to take part were given a background questionnaire, CAT, 3 d food and drink diary, and nine Salivettes^® (Sarstedt Aktiengesellschaft and Co., Nümbrecht, Germany) to take home, complete and return. The study protocol was approved by the local research ethics committee and written informed consent was obtained from all subjects.

Using the CAT, the present study assessed caffeine intakes from weeks 5 to 12 of pregnancy only since this was the most appropriate time frame relative to the stage of pregnancy when most women attend the antenatal clinic for the first time, when women are called to the clinic at about 16 weeks of pregnancy. The food and drink diary was completed at the same time as the CAT although the time frames did not overlap. The diary asked about 3 d of intake at the time of completion whereas the CAT requested recall of intake during weeks 5–12 of pregnancy. While completing the food and drink diary, the women collected a total of nine saliva samples over two consecutive days. Women noted both the time of consumption of foods and drinks and of the saliva collection in the diary to ensure that the saliva measurements provided biomarkers of actual consumption for comparison with the recorded dose (food and drink diary). The saliva samples were also used to assess variation in salivary caffeine and paraxanthine concentrations at different times during the day, and between days. Each saliva sample was collected in a Salivette^® (Sarstedt Aktiengesellschaft and Co.). The Salivette^® consists of an outer centrifuge vessel containing a suspended insert and cotton wool swab. Women were required to keep the Salivette^® swab in their mouth for 10 min to ensure adequate saliva collection. A sample interval of 90 min was chosen between collections to cause minimal disruption to normal daily activity. On the first day, each woman provided saliva samples every 90 min over a 9 h period, involving a total of seven saliva samples. To avoid the presence of caffeine in the saliva due to recent consumption (rather than following absorption), the women were asked to avoid caffeine-containing foods and drinks, listed on a sheet provided, for 1 h before collecting the first sample, and for 15 min before taking each of the following six samples. The women were also asked to rinse their mouth with tap water before collection. On the second day, the same women were asked to provide a further two saliva samples at approximately mid-morning and mid-afternoon to reflect likely time of sample collection in the larger study. Again the women were asked to avoid caffeine-containing foods and drinks for 1 h before sample collection and to rinse their mouth with tap water before the samples were taken. The women were told to refrigerate the samples until they were returned by post to the research team along with the background questionnaire, CAT and diary.

Analysis of caffeine intakes

The CAT was developed to assess caffeine intakes from all possible sources of caffeine in a FFQ style, taking into account specific brand, preparation and portion size information. Since the CAT was developed to assess caffeine intake during pregnancy, questions on consumption of caffeine-containing foods and beverages were repeated to assess intakes during different weeks of pregnancy. Brand information was collected on coffee, tea, hot chocolate, cola and energy drinks, and was categorised into types of drink, for example, instant, filter, iced, and place of consumption to remind women of beverage consumption outside the home or workplace. Further questions requested portion sizes of tea, coffee and cola, changes in intake of tea and coffee during pregnancy, methods of preparation, and intakes of foods which may affect caffeine metabolism, for example, cruciferous vegetables, grapefruit and barbecued foods⁽Reference Murray, Lake, Gray, Edwards, Springall, Bowey, Williamson, Boobis and Gooderham^18–Reference Fontana, Lown, Paine, Fortlage, Santella, Felton, Knize, Greenberg and Watkins²⁰⁾. The caffeine content of foods and beverages were obtained from a UK government report by the Ministry of Agriculture, Fisheries and Food⁽²¹⁾ and also from manufacturers and coffee-houses, providing caffeine values for twenty-nine instant coffees, nine filter coffees, three coffee-house filter coffees, a standard espresso shot and decaffeinated shot, eight instant beverage mixtures, seven espresso-based drinks, eighteen hot chocolates, twenty-five teas prepared from tea bags, seventeen teas prepared from tea leaves, one iced-tea, three instant teas, thirty-six colas, thirty-three energy drinks, eleven soft drinks and two alcoholic drinks. Each of the above beverages had a standard caffeinated and decaffeinated value assigned based on the average caffeinated and decaffeinated content of each drink. For each of the chocolate items a value was assigned based on the average caffeine content from various brands. The caffeine content of fifty-nine over-the-counter drugs was accessed from manufacturers' websites. The CAT also assesses possible confounders, for example, smoking habits, alcohol intake, use of medication and symptoms of pregnancy. The CAT is thus the most detailed and comprehensive tool to assess caffeine intake during pregnancy which is currently available. A detailed computer algorithm was developed which assigned a value for caffeine content (mg) to all caffeinated products assessed in the CAT while taking into account portion sizes, brand information and frequency of intakes. Daily caffeine intake during weeks 5–12 of pregnancy could be calculated for each participant.

The food and drink diaries were analysed manually for each woman's daily caffeine intake. All caffeine-containing foods and drinks recorded in the diaries were allocated a caffeine value depending on amount and type of food or drink consumed. A mean caffeine value was calculated for each food and beverage source and assigned to any unbranded sources recorded in the diaries.

Analysis of saliva

Caffeine and metabolites were extracted and quantified using liquid:liquid extraction and reversed-phase HPLC with UV detection using a modification of the method of Butler et al. ⁽Reference Butler, Lang, Young, Caporaso, Vineis, Hayes, Teitel, Massengill, Lawsen and Kadlubar²²⁾. Salivettes^® were thawed and saliva reclaimed by centrifugation at 756 g for 5 min at room temperature. Saliva (180 μl) was added to a 5 ml screw-cap tube containing 50 mg ammonium sulfate following which 20 μl of a stock solution of β-hydroxyethyltheophylline (20 ug/ml) was added as internal standard. The tube was shaken vigorously for 10 s using a Baxter multi tube vortex (Alpha Laboratories, Eastleigh, Hants, UK) on setting 4, to thoroughly mix the contents and precipitate any protein. This was followed by the addition of chloroform (2·0 ml) and isopropanol (0·5 ml) and the tube shaken again for 4 min on setting 4. After centrifuging for 5 min (84 g) the aqueous top layer was discarded and the organic layer transferred to a thick-walled glass tube and dried down under N₂ at 45°C. The residue was reconstituted in 30 % methanol in water (200 μl), mixed vigorously for 3 s and centrifuged for 2 min (756 g) to give a final concentration of β-hydroxyethyltheophylline of 2 μg/ml. The reconstituted solution was transferred to an autosampler vial and 30 μl injected onto a Beckman Ultrasphere column (ODS; 4·6 mm × 25 cm) with a short guard column (5 cm × 0·4 cm) packed in-house with Bondapak C18/corasil (Waters, Elstree, Herts, UK). The solvent delivery system, autosampler and UV detector were manufactured by Gilson (Paris, France). UV wavelength was set at 280 nm. Solvents used for elution were 0·045 % acetic acid containing 9 % methanol (A) and 100 % methanol (B). Starting with solvent A, elution was a linear gradient over 5 min to a 2 % solution of solvent B, held at this for 10 min, over the next 5 min increased in a linear gradient to 5 % solvent B, followed by a linear increase over 5 min to 8 % B, changed to 15 % B and maintained for 15 min, raised to 75 % B and held at this for 10 min, followed by reversion to 100 % A (the starting solvent) which was maintained for 10 min to equilibrate the column before injection of the next sample. The flow rate was 1·2 ml/min and the retention time, in min, for each compound was approximately: theobromine, 10·3; 1,7-dimethyluric acid, 14·8; paraxanthine, 16·2; theophylline, 17·7; β-hydroxyethyltheophylline, 22·4; caffeine (1,3,7-trimethylxanthine), 29·5. Retention times varied slightly from day to day and were adjusted accordingly to produce optimum identification of each analyte when processing results. A single standard containing all the above compounds at the same concentration (5 μg/ml) was made up in 30 % methanol; this was run after every five samples. Two ‘in-house’ quality-control samples were also extracted and run with each batch of eighteen samples. Quality-control samples were made by spiking ‘blanked saliva’ with 5 μg/ml of each compound. The saliva, donated by volunteers, was collected in Salivettes^® and ‘blanked’ by gentle mixing with charcoal (0·1 g/ml) for 24 h, centrifuged (728 g) for 10 min and the supernatant fraction filtered through a 0·20 μm filter and stored at − 20°C until spiked. When in use, quality-control material was stored at 4°C. Within-batch CV (%) were: theobromine, 2·3; 1,7-dimethyluric acid, 2·4; paraxanthine, 2·2; theophylline, 2·5; β-hydroxyethyltheophylline, 3·1; 1,3,7-trimethylxanthine, 2·7. Between-batch CV (%) were: theobromine, 2·8; 1,7-dimethyluric acid, 2·5; paraxanthine, 2·2; theophylline, 3·0; β-hydroxyethyltheophylline, 2·4; 1,3,7-trimethylxanthine, 3·8. The limit of quantification was 50 ng/ml for all compounds, calculated from standards made up in 30 % methanol with no extraction. The assay was linear between 50–10 000 ng/ml for all the above compounds.

Caffeine, other methylxanthine metabolites and ammonium sulfate were purchased from Sigma (Poole, Dorset, UK) and all extraction and eluant reagents were supplied by Rathburn Chemicals (Walkerburn, Peebleshire, UK).

Statistical analysis

Statistical analysis was carried out using SPSS (version 10.1; SPSS Inc., Chicago, IL, USA), Stata (version 8.2; StataCorp LP, College Station, TX, USA) and MlwiN (University of Bristol, Bristol, UK)⁽Reference Rasbash, Browne and Goldstein²³⁾. Caffeine intakes from the CAT and 3 d diary were compared. κ Statistics were carried out to test the agreement between the caffeine intakes estimated either by the CAT or the diary (split at the median), and the mean of the seven saliva caffeine and paraxanthine measures on day 1, and the mean of the two saliva caffeine and paraxanthine measures on day 2. The total 3 d diary and the individual diary days which related to the day of saliva sample collection were used for this analysis. Time of gestation (weeks) was categorised into two groups (see Table 1 for categorisation), and agreement between the CAT and 3 d diary was estimated within each group as well as for the group as a whole. A variance components model was used to investigate the variance structure of the data and estimate the intra-class correlation coefficient (ICC) of agreement taking all sources of error into account.

Table 1Caffeine intake (mg/d) and time of caffeine assessment tool (CAT) completion

(Mean values and standard deviations)

* κ Statistic comparing CAT and diary.