Of particular clinical interest are the risks and benefits conferred by different beverages. Beverages differ in the amount of potentially important constituents, such as caffeine, oxalate, and alcohol, that they contain. Therefore, they may influence stone formation differently. Several studies have examined the effects of particular beverages on changes in urine composition [6, 7], but few have examined the effect of particular beverages on kidney stone formation.

In a retrospective study [8], beer and coffee consumption were inversely associated with a history of kidney stones, and carbonated beverage (soda) consumption was positively associated with such a history. No significant associations were found with milk, tea, or water [8]. However, the authors did not adjust for other risk factors for kidney stone formation, such as dietary intake of calcium [9]. In a prospective study of the intake of 21 different beverages and incident kidney stone formation in men [10], we adjusted for nutrient intake and found an increased risk associated with consumption of apple juice and grapefruit juice and a decreased risk associated with consumption of coffee, tea, and alcoholic beverages. These results may not necessarily apply to women because the rate of kidney stone formation in women is one third that in men and the determinants of stone formation may differ by sex. No previous prospective data are available on the relation between beverage use and risk for stone formation in women.

We previously demonstrated an inverse association between total fluid intake and risk for incident stone formation in a large prospective study of women [11]. To investigate whether the individual type of fluid ingested is important in women, we examined the relation between 17 different beverages and risk for symptomatic kidney stones in a cohort of 81 093 women in the United States.

Methods

Top Methods Results Discussion Author & Article Info References

Study Sample

In 1976, 121 700 female nurses 30 to 55 years of age completed and returned the initial questionnaire; these women make up the Nurses' Health Study cohort [12]. The cohort is followed through biennial mailed questionnaires that ask about lifestyle practices, other exposures of interest, and the incidence of newly diagnosed disease. The sample for our current analysis was limited to the women who answered the 1992 questionnaire (which included a question about lifetime history of kidney stones) or the 1994 questionnaire (which asked about kidney stones in the previous 2 years). In addition, it included only participants who had also answered either the 1986 or the 1990 dietary questionnaire, which elicited information on beverage use. Although information on some beverages was obtained on the 1980 and 1984 questionnaires, water had not been included. Because water is the most commonly used beverage, we used only the beverage information from the 1986 and 1990 questionnaires, which included water in the beverage list.

We considered only cases of kidney stones that occurred during the 8 years between the return of the 1986 questionnaire and 31 May 1994 in women with no previous kidney stones. After further exclusion of women for whom the date of a kidney stone could not be confirmed, 81 093 women with no history of kidney stones at baseline remained for analysis.

Assessment of Diet

In 1986 and 1990, participants were asked to complete semiquantitative food-frequency questionnaires on which they reported their average use of 131 specified foods and beverages during the previous year. Nutrient intake was computed from the reported frequency of consumption of each specified unit of food or beverage and from published data on the nutrient content of the specified portions [13]. In addition, information was collected on the amount of supplemental calcium (such as calcium carbonate) ingested, either alone or as part of multivitamin preparations. The reproducibility and validity of the questionnaires in this cohort were documented previously [13-15]. A similar questionnaire was shown to be valid and reproducible in men [16, 17].

Individual beverage use was reported as the number of times that a standard serving size of a given beverage was consumed; nine frequency categories were used, ranging from less than once per month to six or more times per day. Eighteen beverages were included on both the 1986 and the 1990 questionnaires.

To calculate the average daily intake of each beverage, we multiplied the reported frequency of used by the serving size for the beverage. Because we had no a priori hypothesis that the various types of wine would differ for risk for stone formation, red and white wine were combined. A food-based validation study showed reasonable levels of reproducibility and validity for reported intake of the 11 individual beverages that were included in the 1980 version of the semiquantitative food-frequency questionnaire [14]. Pearson correlations between the food records and the questionnaire ranged from 0.36 for noncola carbonated beverages to 0.94 for beer, with a mean correlation for the beverages of 0.77.

We calculated total daily fluid intake by using data on frequency of use and serving size of individual beverages. In a validation study in men, a high correlation (r = 0.59) between reported fluid intake and 24-hour urinary volume was reported [10].

We adjusted nutrient values for total energy intake by taking the residuals from a linear regression model with total caloric intake as the independent variable and nutrient intake as the dependent variable [18, 19]. Energy-adjusted values reflect the nutrient composition of the diet independent of the total amount of food consumed. In addition, energy adjustment reduces any variation introduced by overall underreporting or overreporting of intake on the food-frequency questionnaire, thus improving the accuracy of nutrient measurements [18, 19]. Calculation of beverage intake was not adjusted for total energy intake.

Follow-up and Ascertainment of Cases

On the 1992 questionnaire, we asked participants whether they had ever had a kidney stone diagnosed and the date of the first occurrence. The 1994 biennial questionnaire asked about a new diagnosis of a kidney stone since 1992. If a kidney stone was reported, we mailed a supplementary questionnaire to the respondent to confirm the diagnosis and to ascertain the date of occurrence; symptoms; other relevant medical conditions; and the type of stone, if known. The response rate to the supplementary questionnaire was 92%. To confirm the validity of the responders' reports, we obtained the medical records for a random sample of 90 of the women who reported having a kidney stone. The records confirmed the diagnosis in all but one case.

Statistical Analysis

In this prospective study, information on diet was collected before any diagnosis of kidney stone was made. For each participant, person-months of follow-up were counted from the date of return of the 1986 questionnaire to the date of a diagnosis of kidney stone, death, or 31 May 1994, whichever occurred first. Information on exposures of interest from the 1986 questionnaire was updated in 1990. We allocated person-months of follow-up according to exposure status (the category of individual beverage use) at the start of each follow-up period. If dietary information was missing at the start of a time period, the participant was excluded for that time period. For example, if a participant answered the 1986 questionnaire but not the 1990 questionnaire, she would contribute person-time only for the 1986 to 1989 period. If a participant answered the 1990 questionnaire but not the 1986 questionnaire, she would contribute person-time only for the 1990 to 1994 period. The proportion of participants with information missing for any individual beverage was 5% or less for both the 1986 questionnaire (Table 1) and the 1990 questionnaire (data not shown). A validation study in men showed that missing items were not consumed in most instances; thus, participants with information missing for an individual beverage were assigned to the lowest category of intake [17]. The categories for the beverage items were selected before the analyses on the basis of the frequency response for each individual item. The goal was to have enough participants in each category (particularly the extreme categories) to provide sufficient power to examine associations comparing extremes of intake. For each beverage, we assessed the nine individual response categories (after examining the grouped categories) to determine whether substantial variation was being masked. Because the results were very similar, the a priori selected grouped categories were used to provide more stable estimates.

To formally test the null hypothesis that beverage type does not influence risk for kidney stones, we added intake of all specific beverages (except water) to a multiple logistic model that contained the total fluid intake variable and evaluated the change in total model deviance (likelihood ratio test) with 16 degrees of freedom. To make the results of the multivariate models more clinically useful and to have some standard for direct comparisons of different beverages, we converted the daily use of each beverage into the average number (or fraction) of 240-mL (8-oz) servings consumed per day. This provided an estimate of the effect on risk for stone formation of increasing the consumption of an individual beverage by one unit (240 mL).

The relative risk was used as the measure of association [20]. The Mantel extension test was used to evaluate linear trends across categories of intake [21]. We used a proportional hazards model to simultaneously adjust for multiple risk factors [22]. The variables included in these models were age (in 5-year categories); body mass index (six categories); level of supplemental calcium intake (0 mg/d, 1 to 100 mg/d, 101 to 500 mg/d, and >500 mg/d); and dietary intake of calcium, animal protein, potassium, sodium, sucrose, and fluid (quintile groups). We previously reported that these dietary variables were related to risk for kidney stones [11]. For all relative risks, we calculated 95% CIs. All P values are two-tailed.

Results

Top Methods Results Discussion Author & Article Info References

Table 1 shows the frequency of use of the individual beverages as reported in 1986. The beverages most often consumed were water (median, 2 to 3 glasses per day), caffeinated coffee (median, 1 cup per day), skim milk (median, 2 to 4 glasses per week), and orange juice (median, 2 to 4 glasses per week).

We confirmed 719 cases of incident symptomatic kidney stones during 553 081 person-years of follow-up over an 8-year period. After we controlled for potential risk factors, the relative risk for stone formation among women in the highest quintile of total fluid intake (>2592 mL/d) compared with women in the lowest quintile (<1412 mL/d) was 0.62 (95% CI, 0.48 to 0.80).

After we simultaneously controlled for possible confounding by age and nutrient intake, the addition of specific beverages contributed significantly to prediction of the formation of kidney stones (change in the –2 log likelihood, 84 with 16 degrees of freedom; P < 0.001). An additional model that included all specific beverages but not total fluid intake showed that caffeinated coffee, decaffeinated coffee, tea, and wine were significantly and inversely associated with risk and that grapefruit juice was directly associated with risk (Table 2). Specifically, each 240-mL (8-oz) serving of caffeinated coffee per day was associated with a 10% (CI, 5% to 15%) decrease in risk. Similarly, each 240-mL serving of decaffeinated coffee per day was associated with a 9% (CI, 2% to 15%) decrease in risk. For each 240-mL serving per day, an 8% (CI, 1% to 15%) decrease in risk was seen with tea and a 59% (CI, 32% to 75%) decrease was seen with wine; in contrast, a 44% (CI, 9% to 92%) increase in risk was seen with grapefruit juice.

We further examined risk for stone formation according to category of consumption of those beverages for which an overall association was seen (Table 3). The risk for stone formation decreased with increasing frequency of intake of caffeinated coffee, decaffeinated coffee, and wine. In addition, risk was lower in the highest category of tea consumption. The risk increased with increasing consumption of grapefruit juice.

We then compared the effect of individual beverages with that of water, presumably a neutral fluid. Caffeinated coffee (P = 0.008) and wine (P = 0.001) were inversely associated with risk and were significantly more protective than water. The differences between water and decaffeinated coffee (P = 0.09) and water and tea (P = 0.12) were marginally significant. Grapefruit juice was significantly worse than water and was directly associated with risk (P = 0.008).

Discussion

Top Methods Results Discussion Author & Article Info References

Our data support the hypothesis that total fluid intake is inversely associated with risk for kidney stones. In addition, individual beverages differ in their effect on risk. Caffeinated and decaffeinated coffee, tea, and wine were significantly associated with a lower risk for incident stone formation, and grapefruit juice was associated with a higher risk.

We examined 17 beverages. Had we found an unexpected significant association, use of a stricter level of significance would have been considered. Because all of the beverages that were significantly associated with the risk for stone formation were hypothesized a priori, altering interpretation of the P value is unnecessary.

With the exception of our prospective study in men [10], previous studies of the association between fluid intake and kidney stones have been small and have evaluated only a few beverages. Moreover, these studies have focused predominantly on physiologic changes in the urine rather than on actual stone formation and have included men almost exclusively. No prospective studies have included women.

Coffee [8, 23] and tea [23] have been reported to be inversely associated with risk for stone formation in retrospective studies in men. In a national, cross-sectional survey of men and women, coffee was inversely associated with risk but tea was positively associated with risk [24]. None of these studies adjusted for intake of other important dietary factors. In a large prospective study in men that did adjust for dietary intake, caffeinated and decaffeinated coffee and tea were also inversely associated with risk [10]. The reduction in risk was similar in magnitude to that seen in our study of women.

Beer [8, 23] and alcohol [24] consumption were inversely associated with risk in retrospective studies. In a prospective study in men, beer and wine were inversely associated with risk [10]; the reduction in risk with wine in men was similar to that seen in our current study.

The observed protective effect of caffeinated coffee, tea, and alcoholic beverages may be mediated through their effect on urinary concentration. Caffeine increases flow of a more dilute urine by interfering with the action of antidiuretic hormone on the distal nephron [25], thus decreasing risk for crystal formation. However, this beneficial effect may be partly offset by the increase in urinary calcium excretion caused by caffeine [26]. Similarly, alcohol inhibits secretion of antidiuretic hormone, leading to increased urine flow and decreased urinary concentration [25]. The larger magnitude of the decrease in risk associated with wine compared with beer in women and men is probably due to the higher alcohol content of wine. Although it is unclear why liquor was not associated with decreased risk, it may have been because of the small proportion of women (<16%) who reported consuming more than one shot per week (Table 1).

Because coffee and particularly tea were believed to be high in oxalate, the inverse association with these beverages is unexpected. However, little information exists on the oxalate content of foods. In addition, the oxalate content of coffee and tea may differ according to type, brand, and brewing method. Kasidas and Rose [27] used an enzymatic method to measure the oxalate content of foods and found that a 240-mL portion of tea contained 17 mg of oxalate and a 240-mL serving of instant coffee contained 8 mg of oxalate. A recent study [28] reported that only black tea contained a substantial amount of oxalate (11 mg to 21 mg in a 250-mL portion), but apparently very little of the oxalate is bioavailable [29]. Indian and herbal teas contained very little oxalate [28]. Thus, if the amount of oxalate contributed to the diet by black tea is relatively small, the minor increase in urine oxalate levels would be more than offset by the effect of the caffeine on urinary dilution and flow. Although the specific type of tea in our study was not elicited, we suspect that most participants were consuming the preparation most common in the United States, black tea.

The decrease in risk associated with decaffeinated coffee is also intriguing. Because we examined many beverages, some associations may have occurred by chance. The data are internally consistent for major sources of caffeine and alcohol, but the findings for decaffeinated coffee merit further study. Of note, these findings corroborate the inverse association seen with decaffeinated coffee in our male cohort [10]. Conceivably, a component of coffee other than caffeine confers protection.

The positive association between grapefruit juice and risk for stone formation was similar in magnitude among men [10]. Although this finding was consistent in two independent cohorts, its mechanism remains unclear. One misconception is that because citrus juices are acidic [30], their consumption leads to lower urinary citrate. In fact, the actual amount of free acid in citrus juices is small. The high content of citrate salts, predominantly potassium citrate, results in an overall alkaline load and a subsequent increase in urinary citrate [6]. This should decrease the risk for kidney stone formation.

The oxalate content of grapefruit juice remains controversial; some analyses report very little [27] and others as much as 11 mg in a 100-mL serving (Thomas WC Jr. Personal communication). If the oxalate content of grapefruit juice is indeed high, this could be a potential explanation for the increased risk. Of note, grapefruit juice has been shown to affect serum levels of several drugs by markedly inhibiting intestinal-wall first-pass oxidation by cytochrome P450 [31] and may influence the metabolism of other potentially important dietary factors. Still, the mechanism for the association between grapefruit juice and kidney stone formation remains unknown.

Unlike the positive association seen for apple juice in the prospective study in men [10], we found no significant association in women. Additional data from other studies are needed to determine the role of apple juice in stone formation.

Sugared cola has been associated with increased risk for stone formation in some studies [8, 24, 32] but not others [10, 23]. Although sugared soda was associated with an increased risk in the age-adjusted models in the current study, no significant association was seen in the multivariate models; this is similar to what we observed in men [10]. The intake of sugared soda in this cohort was inversely associated with the intake of dietary calcium (r = –0.18;P < 0.001) and potassium (r = –0.26;P < 0.001) and positively associated with intake of sucrose (r = 0.12; P < 0.001). Dietary calcium and potassium were inversely associated with risk, and sucrose was directly associated with risk, [11]; thus, the age-adjusted results appear to be due to lack of adjustment for these other important nutrients. Most previous studies [8, 23, 24, 32] did not adjust for other important dietary factors in the analyses, thus limiting the interpretation of the results.

After adjustment for age, skim milk was inversely associated with risk. However, this association disappeared after adjustment for dietary calcium intake. Because skim milk was the major contributor to dietary calcium intake, which we previously reported to be inversely associated with risk of stone formation [11], it is correct to conclude that skim milk use was also associated with decreased risk for stone formation. The decrease in risk may be due to binding of oxalate in the gastrointestinal tract by dietary calcium, resulting in decreased urinary oxalate excretion.

Beverage intake data were collected before kidney stones were diagnosed; thus, recall bias was avoided. It is unlikely, but if certain nondietary risk factors for kidney stones (such as family history of kidney stones) were strongly associated with the use of certain beverages, these factors may have influenced our results.

These findings were seen among women 40 years of age and older with no history of kidney stones. Only 4.0% of our cohort had a history of stones before 1986, and these persons were excluded from the analysis. We have no reason to believe that the relations we observed would be different in younger women or women with a history of calcium oxalate kidney stones. The observed associations were for incident stone formation, but the physiologic principles of stone formation are unchanged by the fact that a person has already had one stone. The results were consistent with those in men of similar age [10].

Our findings suggest that an increase in total fluid intake can reduce the risk for kidney stones and that specific beverages may be influential. In particular, intakes of caffeinated and decaffeinated coffee, tea, and wine were associated with decreased risk, whereas intake of grapefruit juice was associated with increased risk. Clinicians should consider using these results to counsel women who have already had a kidney stone.

Dr. Willett: Department of Nutrition, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115.