STUDY OBJECTIVE To assess the agreement between four different measures of alcohol intake in pregnancy.
DESIGN AND SETTING Danish speaking pregnant women referred to the Midwife Centre in Aarhus, Denmark, for routine antenatal care were contacted at their first visit at approximately 15–16 weeks gestation from October to December 1998. The women were interviewed about current average alcohol intake and intake within the previous week, and subsequently filled in a two week diary on alcohol intake. When booking for delivery at the end of the first trimester the women were also asked to complete a questionnaire including a one item question on current average alcohol intake.
PARTICIPANTS Participants were 441 pregnant women.
MAIN RESULTS Per cent agreement ± 1 category ranged between 73 and 82. Mean (SD) intake ranged between 1.09 (1.35) drinks/week for diaries, and 0.69 (0.85) for questionnaires. Mean differences between methods were all close to zero. Three of the four measures yielded comparable distributions of average alcohol intake, but reports of intake within the past seven days seemed to be an inappropriate measure of average intake, yielding three times as many abstainers as expected when combining the methods.
CONCLUSIONS When assessing the distribution of alcohol intake in pregnancy or when studying adverse pregnancy outcomes that are probably caused mainly by sustained exposure it seems that for pregnant women with low to moderate alcohol intake diaries or an average measure from interviews or a simple one item questionnaire may be applied. A measure of intake for the previous week seems to be a relevant measure only when studying adverse pregnancy outcomes that are most probably caused by binge-like exposure.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The objective of measuring alcohol intake in pregnancy may be to assess either the distribution of intake in a population or the association between alcohol in pregnancy and adverse pregnancy outcomes. It remains controversial whether there is a lower safe level of drinking during pregnancy,1 2 and further research is therefore needed. An important element in this research is to develop reliable methods to assess alcohol intake. Information on alcohol intake in pregnancy is mostly collected through the use of self administered questionnaires3 or interviews, either face to face4 or by telephone.5 Although the majority of studies among pregnant women use prospectively collected information on alcohol intake few studies have compared prospective information obtained from different methods. Only a few studies have compared interviews with questionnaires,6 7 and the results have been inconsistent. Among men and non-pregnant women diaries have been shown to yield higher reports of intake than questionnaires,8-10 but to our knowledge diaries have never been used among pregnant women.
In a representative sample of pregnant Danish women with low to moderate alcohol consumption we assessed the agreement between four different measures of alcohol intake using diaries, two measures obtained from interviews, and questionnaires.
All pregnant women in Denmark are offered free antenatal care including a number of visits in the antenatal care centre run by midwives. Almost all women in Denmark participate in the antenatal care programme. Danish speaking pregnant women (n=478) referred to the Midwife Centre in Aarhus, Denmark, for routine antenatal care were contacted at their first visit during a two month period from 15 October until 17 December 1998.
The study consisted of an interview mainly concerned with drinking habits, knowledge about and attitudes towards alcohol intake during pregnancy. Questions were asked about current average weekly intake of different types of alcohol containing beverages (interviewa) (see appendix 1 for details). Questions included non-alcoholic wine and beer in order to improve information retrieval and avoid misclassification of non-alcoholic beverages as alcohol containing beverages. Questions on current intake were followed by questions on intake during the week before the interview (interview7) (appendix 1). Interviews were performed face to face by two specially trained midwives either immediately after the first antenatal care visit or, if this was not possible, on a day of the woman's own choice as soon as possible after the visit.
The women were subsequently asked to fill in a diary on alcohol intake during two consecutive weeks starting immediately after the interview. Each page in the diary represented one day, and for each day the woman was to note the number of drinks for each alcohol category and the time of day in a preprinted table (see appendix 2 for details). At the end of the diary the woman was asked if she believed she had changed her drinking habits since the interview. A stamped envelope was handed out together with the diary. If the diary had not been returned within four weeks after the interview, the woman was contacted by telephone and reminded to return the diary. Results from the diary are presented as an average of the two week period where not otherwise stated.
The Midwife Centre in Aarhus is associated with the Department of Obstetrics and Gynaecology, Aarhus University Hospital. When booking for delivery at the end of the first trimester women are asked to fill in a self administered questionnaire for the medical record and a research questionnaire. The self administered questionnaire asked a single question about average current maternal alcohol intake without specifying subtypes of alcohol (see appendix 1), smoking habits before pregnancy and currently, maternal height, prepregnant weight, parity, chronic diseases, and the use of medication within the last three months before pregnancy. The question on alcohol is the standard measurement used for clinical purposes in the department. For calculating mean deviations from the methods described above the mean of each category was used for the questionnaire (0, 0.5, 1, 2, 3.5, 7, etc.). The research questionnaire provided information on marital status and occupational status.
Median gestational age when filling in the questionnaire, at the time of interview, and on the first day of filling in the diary were as shown in figure 1.
The definition of a drink followed the definition from the Danish National Board of Health, one drink containing 12 g or 15 ml of pure alcohol, the equivalent of one normal beer, one glass of wine, 8 cl of fortified wine, or 4 cl of spirits. For comparison, one drink in Britain contains 8 g of pure alcohol,1 and the definition varies between countries.11 In recent years the strength of beer (as measured both in volume% and number of drinks: 0.5, 1, 1.5, etc) has been written on the label of each bottle in Denmark. Non-alcoholic beverages were coded as zero, light beer as half a drink, and strong beer as 1.5 drinks.
Of the 478 women invited to participate in the study 441 agreed to do so. Characteristics of participants and non-participants can be seen in table 1. Subsequently two interview questionnaires were lost before data entry and could not be recovered leaving 439 interviews for analysis. Eighty nine per cent of the interviews were performed before 19 completed weeks of gestation. A total of 434 women returned the diary. Eighty five per cent of diaries were returned within four weeks following the interview. Information from both interview and diary was available for 432 women (90%). Of the 441 women alcohol information from the routine questionnaire was available for 414. For the participants who were interviewed but did not return the questionnaire values for the variables in table 1 did not differ appreciably on inspection (information on smoking before pregnancy not available).
Various statistical measurements have been recommended for assessing validity, reliability and reproducibility of epidemiological tools.12-14 Following Bellach14 and our own previous assessment of self reported smoking habits,15 we present the results of various statistics: Agreement between methods was evaluated by calculating per cent complete agreement (P0) and per cent agreement ± 1 category (P1) as is often used in nutritional epidemiology,16 17 and mean differences between methods. We subsequently performed multiple linear regression analyses to see if the differences between methods could be explained by a number of maternal characteristics: We used the difference between methods as dependent variable, and included as independent variables current smoking habits, age, prepregnancy body mass index, parity, marital status (married/cohabiting; single), and occupational status (employed; unemployed; rehabilitation/social security; student; other). Regression analyses were performed first for all women together, and subsequently for each type of alcohol. Agreement was visualised as recommended by Bland and Altman.13 We evaluated the association between methods by calculating correlation coefficients (Spearman's ρ, rs).
The study was approved by the regional ethics committee, and the Danish Data Protection Agency.
Information on current alcohol intake from diaries, interviews, and questionnaires yielded comparable distributions of intake.
Information on alcohol intake within the past week yielded more abstainers than other methods.
When studying adverse pregnancy outcomes caused by sustained alcohol exposure average measures from diaries, interviews, or questionnaires may be applied.
When studying adverse pregnancy outcomes caused by binge-like exposure diaries or intake within the past week might be considered.
For diary, interviewa, interview7, and questionnaire overall mean (SD) intake was 1.09 (1.35), 0.96 (1.03), 0.74 (1.12), and 0.69 (0.85) drinks/week, respectively, and for women claiming not to be abstainers mean (SD) intake was 1.55 (1.36), 1.36 (0.98), 1.56 (1.17), and 0.96 (0.86) drinks/week, respectively. Total alcohol intake was comprised of 70% wine, 23% beer, 4% fortified wine, and 3% spirits in the diary; 59% wine, 26% beer, 8% fortified wine, and 8% spirits in interviewa; and 73% wine, 19% beer, 4% fortified wine, and 3% spirits in interview7. For total alcohol intake per cent agreement ± 1 category ranged between 73 and 82 (tables 2 and 3).
Alcohol intake seemed to be systematically underreported in interviewa, interview7, and the questionnaire compared with the diary (fig 2), whereas this was not the case for comparisons between the other methods.
For subtypes of alcohol agreement was good for beer and wine (table 3). For fortified wine and spirits agreement measured by mean difference and per cent agreement was good due to many abstainers, but poor as measured by correlation coefficients due to substantial disagreement between methods for non-abstainers (table 3).
Mean differences were all close to zero (table 3). Restricting analyses to women who were not abstainers, or to women who claimed not to have changed their drinking habits after the interview did not change this conclusion. In multiple linear regression analyses there was a tendency towards increasing difference in total intake between the diary and the other measures with increasing number of cigarettes smoked (p values for the regression coefficients for smoking were 0.06 for diary versus interviewa, 0.02 for diary versus interview7, and 0.04 for diary versus questionnaire). This was attributable to increasing difference in beer intake between the diary and the other measures with increasing number of cigarettes smoked (p values for the regression coefficients for smoking were 0.001 for diary versus interviewa, and <0.001 for diary versus interview7). No other covariates were associated with difference between measures.
The proportion of total abstainers was 29% for the diary and interviewa, 28% for the questionnaire, and 52% for interview7. In week one of the diary 49% were abstainers, in week two 41% were abstainers (table 4). Only 17% reported to be abstainers in all four instruments.
Differences between methods, including differences in the proportion of abstainers, may reflect real changes in alcohol intake during pregnancy. Comparing intake during week one of the diary with intake during week 2 (table 4), we found that only 44% of the women reported the same intake during the two consecutive weeks. Comparing interview7, which covered the week before starting the diary, with week one and week two of the diary 47% and 40%, respectively, reported similar intake.
In these data pregnant Danish women reported the highest alcohol intake in a two week diary, slightly lower intake in the interview, and lowest intake in a self administered questionnaire. Mean differences were all less than half a drink/week.
No gold standard has been described for the collection of information on alcohol intake, but it is generally believed that alcohol intake is underreported to some extent. Previous studies among men and non-pregnant women have shown that diaries generally yield higher estimates of alcohol intake than other methods.8-10 With respect to prospective interviews and questionnaires the study by Olsen and Frische7 showed no substantial differences between methods, whereas Pereira et al found some underreporting in interviews compared with questionnaires.6
Some of the differences may be explained by differences in the time interval studied and the specific gestational age at data collection. Previous studies have shown that 70%–80% of pregnant Danish women drink alcohol during pregnancy,18 19 but most women drink at very low levels and are likely to drink at irregular intervals. As can be seen when comparing week one of the diary with week two, more than half of the women changed their consumption from one week to another. Comparing each week in the diary with interview7 the same pattern emerged. This seems to suggest that information about intake during one specific week yields information about that particular week, but may not be representative for the overall average intake for a longer period of time because of the week to week variation. Interview7 thus yielded a very high proportion of abstainers compared with the other methods, which tended to reduce the overall mean intake; but at the same time the group of women claiming not to be abstainers in interview7drank more on average than in the other instruments. Compared with interview7 week one and week two of the diary each yielded a similar distribution of intake (many abstainers). Yet, the week to week variation reflected in interview7, and each of the weeks in the diary was reduced when using the average measure from the two week diary.
The objective of measuring alcohol intake may be to assess either the distribution of intake in a population or the association between alcohol in pregnancy and adverse pregnancy outcomes. On the basis of a hypothetical study group Verkerk suggested that the best measure for assessing the distribution of alcohol intake in a population would be detailed information on the drinking pattern of the previous week.20 In our data interview7 yielded the least precise measure of the average distribution. The “true” proportion of total abstainers is likely to be approximately 17% (those reporting to be abstainers in all methods), and interview7 thus yielded three times as many abstainers as expected as did each of the weeks in the diary. The two week diary and interviewa both yielded very similar distributions. The consumption pattern derived from the questionnaire was more similar to that of the diary and interviewa, although a tendency towards underreporting was evident.
This would suggest that in studies focusing on the distribution of alcohol intake among pregnant women, questions on intake should focus either on average intake in general (as in the questionnaire or interviewa), or on a time period of at least two weeks. This would also be the case for studies on adverse pregnancy outcomes that are probably caused mainly by sustained exposure to alcohol such as low birth weight, intrauterine growth retardation, and preterm delivery, as for most women intake during a specific week is not representative for the average intake. On the other hand, for some adverse pregnancy outcomes (for example, malformations and spontaneous abortions) the level of damage is likely to be determined by the peak blood alcohol concentration rather than the average intake.21-23 High blood alcohol concentrations are achieved by intake of large volumes of alcohol on a single occasion (binge drinking). In these cases average measures would be insufficient, and ideally specific measures of intake for each week or day in pregnancy would be necessary (interview7, or each week or day of the diary).
Avoiding misclassification of drinkers as abstainers may generally increase the likelihood of detecting an effect of alcohol on birth weight.20 It is probable that women reporting to be abstainers in the diary or interview7 were indeed abstainers during those particular weeks. So for any specific week in pregnancy getting a good measure for the proportion of abstainers may be possible. It seems, however, that such information may not be generalised to cover a longer period of time. This may have implications when studying the association between alcohol and adverse pregnancy outcomes that are most probable caused by sustained alcohol intake over a longer period of time. Hence, one should be cautious when interpreting apparent differences in pregnancy outcome between abstainers and women with a very low alcohol intake.4
The context in which the questions were asked may be of significance. In the questionnaire the question on alcohol consumption was mixed with a large number of other questions. Alcohol has thus not appeared to be the target issue, which may have reduced the tendency towards “deliberate” misreporting. The interview on the other hand dealt mostly with various aspects of alcohol in pregnancy, but the main focus was on the women's own attitudes towards drinking rather than level of consumption.
It has been suggested that the respondent's rapport with the interviewer and the extent of confidentiality implied in the procedure are important aspects in an interview situation.24 No substantial nor systematic differences were seen between the interviewers with respect to any of the measures. We considered, however, that the interview might influence the information given in the diary. We therefore instructed the interviewers that if during the interview a woman asked the interviewer specific questions on the possible hazards of drinking during pregnancy, such questions should be answered by referring to the routine antenatal care or to special information meetings arranged once a month for the participants, who had returned their diaries. By arranging these meetings we attempted to minimise the influence of the interviewer on the information given in the diary.
Still, when repeatedly asking questions on the same issue with a few weeks interval it is possible that the women become more aware of their drinking behaviour. If this were to affect the intake level we would expect the women to decrease their consumption. Yet, the diaries yielded the highest average intake, and only 5% of the women themselves believed that they had changed their drinking habits after the interview, half of whom had increased their consumption.
Some studies have assessed whether there might be any difference between the effects of different types of alcohol in pregnancy, and a few of these studies have suggested that beer may be the more harmful.25-27 In this study the contribution of each type of alcohol to the total intake was of a similar magnitude for diaries, interviewa , and interview7. Any of the methods may therefore be applied when investigating subtypes of alcohol.
Different measures of alcohol intake in pregnancy reflect different ways of collecting information. No gold standard has been described, but we found that among women with low and moderate alcohol consumption in pregnancy two week diaries were easy to fill in and not time consuming (approximately 0.5–1 minute needed per day). They yielded a high overall mean intake, a low proportion of abstainers compared with other methods, and a high mean intake among non-abstainers, probably making it the best method of collecting alcohol information in pregnancy, as the distribution of intake may be assessed, studies on adverse pregnancy outcomes that are probably caused mainly by sustained exposure may be carried out using the overall measure, and studies on adverse pregnancy outcomes that are probably caused mainly by binge-like exposure may be carried out using intake during a specific week or even on specific days during gestation. However, when assessing the distribution of intake in a population or when studying adverse pregnancy outcomes that are probably caused mainly by sustained exposure it seems that for pregnant women with low to moderate alcohol intake an average measure from interviews or a simple questionnaire may be applied. A measure of intake for the previous week on the other hand did not seem to be an appropriate measure for the overall distribution of alcohol intake and seems to be a relevant measure only when studying adverse pregnancy outcomes that are most probably caused by binge-like exposure.
Funding: this study was supported by the Faculty of Health Sciences, University of Aarhus, The Danish National Board of Health, and The Danish Research Foundation.
Conflicts of interest: none.