b Based on The Physicians Desk Reference, 1983, and Package Information. former coffee ..... Coffee drinking and myocardial infarction in young women. Amer.
PREVENTIVE
MEDICINE
Measurement
17, 280-294 (1988)
of Coffee and Caffeine Intake: Epidemiologic Research’
Implications
for
GEORGEB. SCHREIBER,D.Sc., CARLA E. MAFFEO, PH.D. ,* MORTONROBINS, M.S.P.H., MARY N. MASTERS,M.S.P.H., AND ANNELL P. BOND Westat, Inc., 1650 Research Blvd., Rockville,
Maryland
20850
Reportedassociations between coffee or caffeine intake and benign breast disease, cancers, and cardiovascular diseases have generally been weak and inconsistent. The apparent discrepancies in these studies might be attributable to imprecision in the measurement of coffee and caffeine intake. A study of a random sample of 2,714 U.S. adults disclosed considerable misclassification of total caffeine intake and, to a lesser extent, coffee intake when the estimates were limited to only the number of cups of coffee consumed. Adjustment for the following factors is recommended: amount of calfeinated and decaffeinated coffee consumed both on weekdays and on weekends; the size of the container used; the method used to brew caffeinated coffee; and the amount of caffeine imbibed from tea and soft drinks. Intake of coffee varied markedly between seasons of the year and over time. Random misclassification of coffee and caffeine intake would have the effect of obscuring dose-response relationships to disease incidence. 8 1988 Academic press, Inc. INTRODUCTION
Recently coffee or caffeine intake have been implicated as risk factors for a number of diseases, including benign breast disease (4, 11, 12), cancer (1, 2, 6,9, 10, 14, 18), and cardiovascular disease (3, 7, 8, 13, 15). While the reported associations have invariably been small and inconsistent, the precise role of caffeine, or any of the other constituents of coffee, remains to be defined. The inconsistencies in the relationship between coffee or caffeine intake and the occurrence of various diseases might be attributable to methodologic differences among the studies. One important difference may be in the precision of the various instruments used to assesscoffee and caffeine intake. The estimation of coffee intake only requires precise measurement of the volume of coffee consumed in a given time period, whereas caffeine intake measurements are much more complex since they involve not only volume of coffee consumed, but also the caffeine concentration of the coffee, and the consumption of caffeine from sources other than coffee. Another possible reason for discrepancies in study findings could be the presence of certain disease risk factors that are associated both with coffee or caffeine intake and with the disease under study. The importance of these confounding risk factors is explored in another article (16).
’ Survey was funded by the National Coffee Association of U.S.A., Inc. * To whom reprint requests should be addressed 280 0091-7435/88$3.00 Copyright 0 1988 by Academic Press, Inc. AU rights of reproduction in any form reserved.
MEASUREMENT
OF COFFEE
AND
CAFFEINE
INTAKE
281
METHODS
A probability sample of white, civilian, noninstitutionalized adults 25 to 74 years of age and residing in the coterminous United States was selected using random digit dialing (RDD) telephone procedures developed by Waksberg (17). Interviewing was a two-stage process. After subjects were screened to ensure eligibility with respect to age and race criteria, those eligible were interviewed. The screening response rate was 85%. Eighty-one percent of the eligible subjects completed the interview. Interviews were completed with 2,714 subjects who were interviewed from a central location, using standardized interview procedures, from July to early November 1983. To minimize both the response burden to the households and the standard error of the estimates, only one man or woman was interviewed per household. A random half of the sample was selected and only adult men were interviewed in that portion of the sample. In the remaining half, only adult women were interviewed. Respondents who were coffee consumers were asked to characterize their coffee-drinking behavior with regard to number of cups, size, and container used (i.e., cup or mug), and the portion of the cup usually drunk. Inquiry was also made about the kind of coffee consumed (regular or decaffeinated) and the brewing method used (instant, drip, or perked). Estimates were made of the average daily cups of coffee drunk by each respondent and adjusted for the volume consumed. Estimates were also calculated for the average daily intake of caffeine from coffee and the amount of caffeine intake derived from other sources, such as tea, chocolate, soft drinks, and certain medications as shown in Table 1. To assess the reliability of reporting coffee-drinking behavior, we selected a random subsample of 240 respondents to be reinterviewed 3 months after the initial interview. These subjects were then asked only about items pertaining to their coffee-drinking behavior. A number of variables were derived from a combination of several items in the questionnaire. Estimates of daily coffee consumption were obtained from frequency questions about usual weekday and weekend coffee use. The weighted average of these two items is henceforth referred to as “weighted coffee consumption.” The “estimated coffee consumption” was, in turn, derived from the weighted consumption by adjusting for the size of the container used and the quantity of coffee actually consumed. Container size and amount consumed are referred to as volume adjustments. “Crude estimated coffee-caffeine intake” is obtained when weighted coffee consumption is weighted by an assumed average caffeine content of 87 mg/5 oz of caffeinated coffee and 3 mg/Soz of decaffeinated coffee. “Refined coffee-caffeine intake” is calculated by applying caffeine estimates associated with various brewing methods to the estimated amount of caffeinated coffee consumed and adding the amount of coffee-caffeine derived from decaffeinated coffee consumed. RESULTS Coffee Consumption
Of the 2,714 subjects interviewed, 78% were current coffee drinkers, 7% were
282
SCHREIBER
ET AL.
TABLE 1 CAFFEINE CONTENT OF VARIOUS SOURCES OF CAFFEINE INTAKE
Coffee by method of brewing” (mg/5 oz) Instant Drip Perked Other Instant, drip, and perked Drip and perked Decaffeinated Tea by type (mg/5 oz) Caffeinated Caffeinated and decaffeinated Caffeinated soft drinks (mg/oz) Chocolate: Estimated caffeine (mg) for frequency of use Daily use Almost daily use Sometimes, almost never Medicationsb Anacin, Excedrin, Vanquish NoDoz, Vivarin, Cafedrine Darvon compound, Fiorinol Midol, Easy-Mens, Cope Prolamine, Appedrine Pre-Mens Forte, Aqua-Ban Cafergot, Wigraine, Migral Caffeine content is multiplied by the recommended frequency of use for each medication to estimate caffeine intake
am5 115 mg 85 mg 87 mg 87 mg 100 mg 3 mg 40 mg 30 mg 3 mg 20 mg 10 mg Omg 65 mg 200 mg 75 mg 65 mg 140 mg 200 mg 200 mg
LIBased on FDA Talk Paper, October 1983, and other publications providing information on coffee caffeine content. b Based on The Physicians Desk Reference, 1983, and Package Information.
coffee drinkers, and 15% never drank coffee. Among the coffee drinkers, the reported average daily intake was 2.6 cups per day, while the median was 2.8 cups per day. A higher percentage of men than women were current coffee drinkers (82% vs 78%). Of the total coffee drinkers, 42.2% drank only caffeinated coffee, 16.4% drank decaffeinated coffee exclusively, and the remaining 41.5% consumed both types of coffee.
former
Weekday vs Weekend There were somewhat fewer coffee consumers on weekends than on weekdays. Of the 2,714 adults surveyed, 1,658 (61%) reported drinking caffeinated coffee on weekdays compared with only 1,514 (56%) on weekends. The number of decaffeinated coffee drinkers was only slightly greater on weekdays than on weekends (707 vs 679). Of greater impact are the differences in the relative amounts of coffee consumed on weekdays vs weekends. The percentage of caffeinated coffee drinkers who drink 5 or more cups of coffee on weekdays was 23.1% compared with 18.3% on weekends. For decaffeinated coffee drinkers, the percentage drinking 5 or more cups was 9.5 and 7.7 on weekdays and weekends, respectively.
MEASUREMENT
OF COFFEE
AND
CAFFEINE
INTAKE
283
The effect of these weekday-weekend differences in coffee consumption on the overall categorization of coffee-drinking level is summarized in Table 2. Estimates of the average number of cups of coffee consumed daily for a full 7-day week are compared with reported average daily consumption during the 5 weekdays. Approximately 70% of coffee drinkers would have been classified in the same consumption category, using a cup containing 5 oz of coffee as a categorical unit. Therefore, if weekday consumption only is used to estimate the average daily intake, about 30% would have been misclassified to some extent. Most of those misclassified (86%) would have had their category of consumption 1 or more cups higher than their “true” consumption and the remaining 14% of all coffee drinkers would have had their average daily consumption of coffee underestimated by 1 or more cups of coffee. Volume Adjustment An estimate of coffee consumption is dependent on two volume factors: the container size and the amount of the coffee actually consumed. Respondents were asked whether they usually drank caffeinated coffee out of a cup or mug and whether they drank all or most of its contents, or left part of it unconsumed. For estimating purposes, a mug was calculated to contain 1.5 times the volume of a cup. To adjust for the amount of coffee consumed, those stating that they usually drank only part of their coffee-about 15% of respondents-had their consumption estimates reduced by the amount discarded, which was estimated to be 20%. These values were arrived at through observation of the coffee drinking habits of a small group of employees of Westat Research. Table 3 shows the percentage distribution of respondents according to their use of a cup or mug for drinking caffeinated coffee on weekdays. Of the respondents, 63% drank from cups exclusively and 35.0% used mugs only. A small percentage of individuals, 1.7%, used both cups and mugs. Mugs were used more frequently by heavy coffee drinkers. Of those who reported drinking 5 or more cups of caffeinated coffee, 43.4% used mugs compared with 33.4% of those who drank less than 3 cups daily. The distribution for type of container used for weekend coffee drinking presents essentially the same pattern. In Table 4 the reported average daily intake of caffeinated coffee is contrasted with the estimated intake, taking container size into account. Adjusting for container size shifts the distribution toward a higher intake of coffee. While the percentage of people whose intake is underestimated is fairly similar for each consumption category over 1 cup of coffee, the consumption is markedly underestimated by 2 or more cups for those individuals reporting coffee consumption of 4 or more cups daily. The number of individuals classified as drinking 10 or more daily cups of coffee, for example, would increase from 137 to 180, a 32% increase. In considering dose-response relationships, underestimates of this order could have a significant bearing on the findings. The impact of the second volume component, the amount of beverage consumed, on classification of coffee consumption was also examined. Respondents were asked whether they usually drank “all or most” or “only part” of their cup of coffee. The reported frequency of usual consumption was then compared with
0 Estimated from tabulated data.
Percentage misclassified by at least 1 cup Percentage with higher average daily coffee consumption on weekdays than that for a full week Percentage with lower average daily coffee consumption on weekdays than that for a full week
Less than 1 1.0-1.9 2.0-2.9 3.0-3.9 4.M.9 5.0-5.9 6.0-6.9 1.0-7.9 a.o+ Total
Weighted average daily consumption of coffee during a full week (No. cups)
16.3
13.3
14.4
0.0
11.2
2.8
19.1
386
27.7
436
143
63 312 10 1
2
11.2
58 369 63
1
127 15 1
Less than 1
1.9
27.6
29.5
369
102 260 7
3
2.6
32.8
35.4
1 232
6 70 150 4 1
4
6.4
44.0
50.4
109
8 40 54 5 2
5
5.8
42.8
48.6
25 34 71 7 1 138
6
Reported average daily consumption of coffee on weekdays (No. cups)
9.5
48.8
58.3
19 22 35 8 84
7
8.2”
52.8”
61.0”
2 13 28 174 218
8+
4.2”
26.6”
30.8“
185 447 429 348 224 113 112 73 184 2,115
Total
NUMBER OF COFFEE DRINKERS BY REPORTED CUPS OF COFFEE CONSUMED DAILY ON WEEKDAYS COMPARED TO THEIR WEIGHTED AVERAGE DAILY CONSUMPTION DURING A FULL WEEK
TABLE 2
3 g
x x g ; $
E
83.3 16.7
0.0 100.0
6
Cup and mug Total (%)
Number of respondents
