An incomplete random block design was set up so that 5 beverages were tested on each woman: water, both of the cola beverages, one of the milks, and one of the 2 citric acid–containing beverages. Within each subject the beverage sequence was randomized by using the random number function of EXCEL (Microsoft, Redmond, WA).
The sequence of tests in each woman was arranged on a 1-wk cycle, so that the entire suite of 5 tests was completed, in most subjects, within one calendar month. The subjects reported to the research unit fasting in the morning, after having voided at home, noting the time. They stayed in the unit, fasting, until a 2-h second specimen was collected. They were then fed a breakfast consisting of the test beverage and 2 pieces of specially baked, low-calcium, Italian style white bread, toasted, with butter. The milk serving was 340 mL (12 oz) and the carbonated beverage servings were 567 mL (20 oz). The difference in fluid volume between the milks and the other beverages was compensated for by having the subjects ingest an additional 227 mL deionized water. The differences in carbohydrate intake between the beverages were compensated for by providing the subjects with the requisite number of jelly beans (consisting mainly of sucrose and corn syrup) to equalize total sugar intake for all test meals (1.256 MJ total). All urine was collected during the next 5-h period, during which time subjects consumed an additional 567 mL deionized water; no further food was consumed until the 5-h urine collection had been completed.

Women were recruited who habitually consumed at least two 340-mL (12-oz) cans of carbonated beverages daily. The total sample size was 32. Two women dropped out of the study early and their data are not included in this report. All subjects were instructed to maintain their usual calcium intakes throughout the study and to refrain from high-sodium foods for 2 d before each test. The mean (±SD) age of the subjects was 31.4 ± 5.6 y and their mean body mass index (in kg/m2) was 25.2 ± 3.75. The subjects' habitual intake of carbonated beverages, by history, ranged from two to seven 340-mL (12-oz) servings/d (: 3.16). Twenty-seven of the 30 subjects habitually consumed predominantly colas, and 3 consumed predominantly the citric acid–containing beverages. Median calcium intake, as determined by a food-frequency questionnaire modified from the Block food-frequency questionnaire (14), was 19.6 mmol/d (interquartile range: 15.8–30.4 mmol/d).

Analytic methods
Urine samples were analyzed for calcium, creatinine, sodium, titratable acidity, and total acidity. Colas were analyzed for acidity, sodium, and phosphorus. Calcium was determined by atomic absorption and sodium by flame emission spectrophotometry (Perkin Elmer AAnalyst, Norwalk, CT), creatinine by an autoanalyzer method based on the Jaffe reaction (15), and both titratable and total acidity by the method of Chan (16). (This method equates total acidity to titratable acidity plus ammonium ion, less carbonic acid.) Phosphorus was analyzed by an autoanalyzer method based on Fiske and Subba Row (17). pH was measured by using a Fisher Accumet pH meter (model 915; Fisher Scientific, St Louis). The caffeine content of the beverages was attained from The Food Processor database (ESHA, Salem, OR).

Statistical methods
The data were analyzed in various ways, depending on the a priori hypotheses, ie, some by ANOVA, testing for treatment and order, and some by simple paired t tests. When the data were not normally distributed, the summary descriptive statistics were the median and interquartile range. Differences were tested against a null hypothesis of zero. In addition to the raw data, various derived variables were calculated, principally focused on estimating the excess urinary calcium (Cau) after beverage consumption. The increments produced by carbonated beverage ingestion were not much greater than the changes observed with water alone or the usual day-to-day variability in calciuria. (The within-subject CV for calcium content of the 2-h fasting specimens was 10.9%; that for the calcium-to-creatinine ratio was 9.1%.) For that reason and to minimize bias in estimating effects, we used 4 methods of estimating excess calciuria. The resulting data were analyzed statistically as above. The 4 methods, based on the analyte content in the 5-h urine sample after the test breakfast, except where otherwise indicated, were as follows: