Consumption of caffeinated beverages and kidney function decline in an elderly Mediterranean population with metabolic syndrome

It remains unclear whether caffeinated beverages could have deleterious renal effects in elderly population with underlying comorbid conditions. We investigated the associations between coffee, tea, or caffeine intake and 1-year changes in glomerular filtration rate (eGFR) in a large Spanish cohort of overweight/obese elderly with metabolic syndrome (MetS). This prospective analysis includes 5851 overweight/obese adults (55–75 years) with MetS from the PREDIMED-Plus study. We assessed coffee, tea, and caffeine consumption from a validated food-frequency questionnaire and creatinine-based eGFR using the Chronic Kidney Disease Epidemiology Collaboration equation. Multivariate-adjusted regression models were applied to test associations between baseline coffee, tea, or caffeine intake and 1-year eGFR changes. Caffeinated coffee (> 2 cups/day) and tea (at least 1 cup/day) drinkers had 0.88 and 0.93 mL/min/1.73 m2 greater eGFR decrease respectively, compared to those with less than 1 cup/day of coffee consumption or non-tea drinkers. Furthermore, caffeinated coffee consumption of > 2 cups/day was associated with 1.19-fold increased risk of rapid eGFR decline > 3 mL/min/1.73 m2 (95% CI 1.01–1.41). Similarly, individuals in the highest (median, 51.2 mg/day) tertile of caffeine intake had a 0.87 mL/min/1.73 m2 greater eGFR decrease. Decaffeinated coffee was not associated with eGFR changes. In conclusion, higher consumption of caffeinated coffee, tea, and caffeine was associated with a greater 1-year eGFR decline in overweight/obese adults with MetS.

Several limitations of these studies need to be addressed; (1) most previous studies are cross-sectional rather than longitudinal studies, (2) they only reported information on total coffee consumption and not its subtypes (caffeinated and decaffeinated coffee), (3) stratified analyses were often not reported, and (4) they were performed in general healthy adults, and not in vulnerable elderly individuals. In addition, the impact of coffee or tea on eGFR could be different among countries with different lifestyle and tea and coffee-drinking habits. Consequently, it is unclear whether caffeinated beverages have deleterious effects on kidney function in elderly population with underlying comorbid conditions. Therefore, the purpose of this study was to address the associations of caffeinated and decaffeinated coffee, tea, and caffeine consumption with 1-year changes in eGFR in a large Spanish cohort of overweight/obese older individuals with metabolic syndrome (MetS).

Methods
Study design and participants. The present data was analysed using an observational prospective design conducted within the frame of the PREDIMED-PLUS study, which included 6874 older adults enrolled between 2013 and 2016 by 23 Spanish centers working in collaboration with 208 National Health System primary care clinics. Briefly, PREDIMED-Plus is an ongoing, 6-year, multicenter, parallel randomized clinical trial (RCT) evaluating the long-term effect of a weight-loss intervention based on an energy-reduced Mediterranean diet (MedDiet), physical activity promotion and behavioral support (intervention), in comparison with usual-care recommending an energy-unreduced MedDiet (control), on primary cardiovascular prevention. Eligible participants were overweight/obese (BMI 27-40 kg/m 2 ) men and women aged 55-75 years harboring the MetS 25 , but free of cardiovascular disease at enrollment. More specific details of the study cohort and inclusion/exclusion criteria have been reported 26,27 and the protocol is available at http:// predi medpl us. com/. Local Ethics Committee approved the study protocol and all participants signed a written informed consent.
For the current study, participants who did not complete the semiquantitative food-frequency questionnaire (SFFQ) at baseline (n = 53) or those whose total energy intake was outside pre-defined limits (women < 500 and > 3500 kcal/day, and men < 800 and > 4000 kcal/day) (n = 188) were excluded. We also excluded subjects who died (n = 11) or were lost to follow-up (n = 16) within first year of follow-up, and who had missing data on eGFR at baseline or at the 1-year assessment. The remaining 5851 participants comprised the final sample.
Assessment of coffee, tea, and caffeine. The exposures of interest were baseline consumption of coffee (caffeinated and decaffeinated coffee), tea, and total caffeine intake. Trained dietitians collected participant information about dietary intake through face-to-face interviews using a Spanish version of the validated 143-item SFFQ 28 . Participants reported their average frequency of consumption over the preceding year for a specified serving size of each item. The nine possible answers ranged from "never or less than once per month" to "six or more per day", which were transformed into grams or milliliters per day using the standard portion size of each food and beverage. Two and one items on the FFQ were specifically related to coffee consumption (one for caffeinated coffee and one for decaffeinated coffee) and tea consumption, respectively. The standard serving size of 1 cup of coffee or tea was assigned as 50 mL in the questionnaire. Total coffee consumption was considered as the sum of caffeinated and decaffeinated coffee. Spanish food composition tables were used to estimate daily energy and nutrient intakes 29 .
We modeled three categories of coffee consumption (0 to < 1cup/day, 1-2 cups/day, and > 2 cups/day) and tea consumption (0 cup/day, < 1 cup/day, and ≥ 1 cups/day) with the lowest category as the referent group. Total caffeine intake was computed from the SFFQ using the caffeine contained in caffeinated coffee (400 mg/L), decaffeinated coffee (10.7 mg/L), tea (100 mg/L), regular sodas (79.2 mg/L), artificially sweetened soda (128 mg/L) and chocolate (180 mg/kg). Reference values from the European Food Safety Authority 30 were used to calculate caffeine intake. For data analyses, caffeine intake (mg/day) was categorized into tertiles. We did not collect any information on the types of tea, soda, and chocolate. Assessment of the outcome and covariates. The main outcome of interest for the present study was 1-year changes in kidney function, assessed by changes in eGFR (1-year data minus baseline data) estimated indirectly from serum creatinine (SCr) using the Chronic Kidney Disease Epidemiology Collaboration equation for Caucasian individuals 31 . SCr levels were determined by the enzymatic creatinine assay method (coefficient of variation < 4.0%). The secondary outcome was a rapid decline of kidney function, defined as 1-year eGFR decline rate > 3 mL/min/1.73 m 2 . This threshold represents a magnitude of change that is 3 times the decline rate imposed by natural ageing.
At baseline, trained staff collected participant's information through questionnaires regarding sociodemographic characteristics (sex, age, education, and employment status), lifestyles factors (smoking, physical activity 32,33 ), history of illnesses and medication use. Additionally, we assessed adherence to the MedDiet using a 17-item questionnaire, adapted from a previously validated questionnaire 34 . This tool was used to evaluate compliance with the intervention and as a key element to guide the motivational interviews during the followup. Compliance with each of the 17 items relating to characteristic food habits was scored with 1 or 0 points, when the criterion was met or not met, respectively. Therefore, the total MedDiet score range was 0-17, with 0 meaning no adherence and 17 meaning maximum adherence to MedDiet. Weight, height and waist circumference were determined in duplicate by trained personnel following a pre-established protocol. BMI was calculated by dividing the weight (kg) by the height squared (m 2 ). Blood pressure was measured at rest in triplicate by an automated digital device (Omron-HEM297705C). At baseline and 1-year, blood samples were collected after an overnight fast, and routine biochemical analyses including fasting glucose, lipid profile, and SCr was performed.
Statistical analyses. All  www.nature.com/scientificreports/ Descriptive data are presented as mean ± SD for quantitative variables or percentages for qualitative variables. Baseline characteristics of the study population across categories of total coffee and tea consumption were compared using ANOVA and chi-square tests, as appropriate.
Multivariate linear regression models were applied to test associations between categories (lowest category as reference) of coffee, tea, and caffeine intake and changes in eGFR at 1 year. For these associations, coffee, tea, (both for each cup/day increase) and caffeine (for each 50 mg/day increase) were also analyzed as continuous variables. The results are presented as β-coefficients and 95% confidence intervals (CIs). Furthermore, ORs were calculated for the association between categories of coffee and tea consumption and rapid eGFR decline (1-year changes in eGFR > 3 mL/min/1.73 m 2 ). Three models with additional adjustment for potential confounders were fitted: model 1, adjusted for sex (men/women), age (years, continuous), and baseline eGFR (mL/min/1.73 m 2 , continuous); model 2, additionally adjusted for BMI (kg/m 2 , continuous), smoking (never, current or former smoker), education (primary, secondary education or academic/graduate), physical activity (MET-min/day, continuous), type 2 diabetes (yes/no), dyslipidemia (yes/no), hypertension (yes/no), center (four categories by number of recruited participants: < 200, 200-250, 250-300, or ≥ 300), treatment group (intervention/control), MedDiet score (17-points, continuous), and total energy intake (kcal/day, quartiles); to minimize confounding by other dietary factors, a third model (model 3) was additionally adjusted for dietary intakes of total protein, saturated fat, alcohol, fiber, sodium, magnesium, potassium (each categorized at energy-adjusted quartiles by using the residual method 35 ), 1-year changes in systolic blood pressure (mmHg, continuous), and 1-year changes body weight (kg, continuous). Model 3 was further adjusted (cup/day, continuous) for coffee (for tea), tea (for each coffee), caffeinated coffee (for decaffeinated coffee), or decaffeinated coffee (for caffeinated coffee). All models were conducted with robust variance estimators to account for intra-cluster correlations, considering as clusters the members of the same household (n = 354 couples). Tests for linear trends across categories were conducted by modeling the median value for coffee, tea, or caffeine intake categories as continuous variables.
Finally, we performed a sensitivity analysis, in which we repeated the primary analysis but using percentage of eGFR change ((1-year eGFR − baseline eGFR)/(baseline eGFR) * 100) instead of eGFR change, adjusting for baseline eGFR. Statistical significance was set at p < 0.05.
Ethics approval and consent to participate. The

Results
At baseline, the mean age of participants was 65 years, the mean BMI 32.5 kg/m 2 , and 48% were women. The mean baseline eGFR was 84.2 mL/min/1.73 m 2 and mean change at 1 year was − 0.96 mL/min/1.73 m 2 . Mean daily consumption of caffeinated coffee, decaffeinated coffee, and tea were 42 mL/day (0.8 cup/day), 36 mL/day (0.7 cup/day), and 11 mL/day (0.2 cup/day) respectively. The mean intake of caffeine was 24.8 ± 25.7 mg/day. Table 1 shows baseline characteristics of study population by categories of coffee and tea consumption. Overall, 91.7% of participants were daily coffee consumers, and 36.4% drank > 2 cups of coffee/day. In contrast, only 32.5% were tea consumers and 11.3% drank at least 1 cup of tea/day. Compared with participants who consumed coffee less than daily (< 1 cup/day), those who consumed > 2 cups/day were more likely to be men, whereas tea consumers were more frequently women. Individuals who drank > 2 cups/day of coffee were younger and had higher BMI and baseline eGFR. They were more likely to have obesity, diabetes, and dyslipidemia and more frequently used oral hypoglycemic drugs, smoked, had higher waist circumference, fasting glucose and educational levels, and lower MedDiet adherence. For tea consumption, the associations were mostly in the opposite direction, except for education. High coffee consumption was also associated with a high intake of saturated fat, protein, sodium, magnesium, and potassium and with low fiber intake. Similar trends were observed for tea, except for saturated fat, fiber and alcohol intake, which showed an opposite pattern.
Similar results were observed when all of these analyses were repeated using percentage of change of eGFR instead of eGFR change, adjusting for baseline eGFR values (data not shown). Overall, we found no evidence that the observed associations varied across subgroups of sex, age, eGFR, obesity, diabetes, hypertension, dyslipidemia, smoking, MedDiet score at baseline, and intervention group ( Fig. 1 and Supplementary Fig. S1 online).

Discussion
In this study, we examined the association between coffee, tea, and caffeine intake and 1-year changes in eGFR in older Mediterranean individuals with overweight/obesity and MetS. We found no association between total coffee consumption and eGFR changes. However, when we explored the association by coffee subtypes, individuals who consumed > 2 cups/day of caffeinated coffee disclosed a greater eGFR decline and higher risk of rapid kidney function decline compared to those who consumed none to < 1 cup/day. Similar results were observed for drinkers of one or more cups of tea/day compared to non-consumers. Besides, participants in the highest tertile of total dietary caffeine intake showed a higher decline in eGFR than those in the reference tertile. No association of decaffeinated coffee consumption with eGFR changes was observed.
Coffee is a seed that contains a complex matrix of vitamins, minerals, bioactive phytochemicals, caffeine, diterpenes, melanoidins, and trigonelline with potential beneficial effects on human health 8,10,36 . Therefore, the association of coffee with health outcomes, especially non-communicable chronic diseases such as CKD has been of research interest in the last decade 10,11 . Unlike our results, epidemiological studies suggest that regular coffee consumption is either unassociated with kidney function 6,7,20 or has a protective effects against CKD in healthy middle-aged adults 21,24 . Cross-sectionally, the association between coffee and kidney function has been mainly explored in Asian populations. In two studies conducted in healthy Japanese adults 14,15 , higher eGFR was observed in usual coffee drinkers (≥ 1 cups/ day) compared to non-drinkers. Similarly, in Korean women 18 , habitual coffee consumption (≥ 2 cups/day) was associated with a lower risk of CKD only in diabetic women. However, other studies in Japanese middle-aged adults reported no association between coffee and eGFR 16,17 . Notwithstanding, higher cross-sectional eGFRs values do not necessarily reflect better kidney function; therefore, these results should be taken with caution since certain kidney diseases may also manifest hyperfiltration in early stages 37 . Finally, the results of a recent meta-analysis of cross-sectional studies conducted by Wijarnpreecha et al. 7 reported no significant association between coffee consumption and CKD risk in males, and a trend to a lower risk of CKD in women. Contrary, in our study, the associations between caffeinated coffee or tea consumption and eGFR were most apparently among men, despite interactions were not significant. Further research is warranted to clarify the differences observed by sex, and if they are proven to understand the plausible mechanisms involved.
Recently, prospective studies have reported inconsistent results. Three studies, one conducted in US adults 21 and two in Asian populations 19,24 , reported a decreased risk of incident CKD in healthy individuals with greater coffee consumption. In contrast, a study conducted in an Iranian population revealed no association between coffee and CKD incidence after 6 years of follow-up 22 . The only epidemiological study conducted in an European population reported a cross-sectional positive association between coffee consumption and eGFR at baseline, but no prospective association with eGFR changes after 15 year follow-up 20 . Until now, the only RCT in 19 young Japanese revealed a beneficial effect of coffee consumption for 2 weeks, but not green tea, on cystatin-C-based eGFR 23 .
The main discrepancies between previous studies and our observations might be attributable to the populations examined, follow-up time, and the assessment tools used. Most studies were conducted in apparently healthy middle-aged populations, and none of them included overweight/obese older adults at high cardiovascular risk. Additionally, previous studies have focused on exploring the association between total coffee and kidney function, but none of them have explored differences by coffee subtypes. Further, we also determined whether the observed detrimental effects of caffeinated coffee and tea consumption on kidney function differed www.nature.com/scientificreports/ across subgroups according to various baseline factors and similar trends were seen; this type of analysis was seldom performed in previous studies. The association between tea and kidney dysfunction has been less explored and the few studies in the field had reported no significant associations 15,19,20,22 . We observed that the consumption of ≥ 1cup tea/day was associated with higher eGFR decline. Despite, in our study tea consumption was lower than caffeinated coffee consumption, our findings regarding caffeine intake suggest that this nutrient could be the causal for the deleterious renal effect observed.
Additionally, we found no association between decaffeinated coffee and kidney function, that might reinforce the potentially harmful effect of caffeine 13 . However, the mechanism that confers this negative effect of Table 2. Crude and multivariate linear regression models of 1-year changes in eGFR according to categories of coffee and tea consumption at baseline. Values are means (95% CI) unless otherwise indicated. eGFR estimated glomerular filtration rate, Ref. referent value. P value for trend based on multivariate linear regression treating the median values of each category as an ordinal variable. Model 1: adjusted for sex (men or women), age (years, continuous), and baseline eGFR (mL/min/1.73 m 2 , continuous). Model 2: additionally adjusted for BMI (kg/m 2 , continuous), smoking (never, current, or former smoker), educational level (primary, secondary education or academic/graduate), physical activity (MET-min/day, continuous), type 2 diabetes prevalence (yes or no), dyslipidemia (yes or no), hypertension (yes or no), center (four categories by number of recruited participants: < 200, 200-250, 250-300, or ≥ 300), intervention group (intervention or control group), MedDiet score (17-points, continuous) and total energy intake (kcal/day, quartiles). Model 3: additional adjusted for dietary intakes of total protein, saturated fat, alcohol, fiber, sodium, magnesium, potassium (each categorized by energy-adjusted quartiles), 1-year changes in systolic blood pressure (mm Hg, continuous) and 1-year changes body weight (kg, continuous). Model 3 was further adjusted (cup/day, continuous) for coffee (for tea analysis), tea (for each coffee analysis), caffeinated coffee (for decaffeinated coffee analysis), or decaffeinated coffee (for caffeinated coffee analysis). All analyses were conducted with robust estimates of the variance to correct for intra-cluster correlation. a P < 0.05 (compared to the reference group). b P < 0.05. The standard serving size of 1 cup was defined as 50 mL in the questionnaire.  www.nature.com/scientificreports/ Figure 1. Mean multivariable-adjusted differences (95% CI) for 1-year changes in eGFR between the two highest and the lowest categories of caffeinated coffee consumption at baseline stratified by group. eGFR, estimated glomerular filtration rate; MedDiet, Mediterranean diet. Model adjusted (except the stratification factor itself) for sex (men or women), age (years, continuous), baseline eGFR (mL/min/1.73 m 2 , continuous), BMI (kg/m 2 , continuous), smoking (never, current, or former smoker), educational level (primary, secondary education or academic/graduate), physical activity (MET-min/day, continuous), type 2 diabetes prevalence (yes or no), dyslipidemia (yes or no), hypertension (yes or no), center (four categories by number of recruited participants: < 200, 200-250, 250-300, or ≥ 300), intervention group (intervention or control group), MedDiet score (17-points, continuous), total energy intake (kcal/day, quartiles), dietary intakes of total protein, saturated fat, alcohol, fiber, sodium, magnesium, potassium (each categorized at energy-adjusted quartiles), 1-year changes in systolic blood pressure (mmHg, continuous), and 1-year changes in body weight (kg, continuous), decaffeinated coffee, and tea consumption (cup/day, continuous). All analyses were conducted with robust estimates of the variance to correct for intracluster correlation. a P < 0.05 (compared to the reference group www.nature.com/scientificreports/ caffeine intake is uncertain 36 . A possible explanation could be related with the structural similarity of caffeine and adenosine acting as a nonselective antagonist of A1 adenosine receptors, which may be a potential mechanism for the impairment of renal function 11 . On distal afferent arterioles, A1 adenosine receptor activation causes vasoconstriction and decreases eGFR 38 . By contrast, adenosine antagonism by caffeine may prevent afferent arteriolar constriction or induce vasodilation, increasing renal plasma flow and acute elevation of eGFR 11 . Probably this increase in eGFR could lead to acute compensatory hyperfiltration and subsequent accelerated decline in GFR 37 . This hypothesis may be partly supported by the observed relationship in our study between caffeinated coffee and a slightly higher baseline eGFR and, subsequently, a higher 1-year eGFR decline. In this context, it has been reported that, compared to decaffeinated coffee, caffeinated coffee could cause a short-lasting increase in blood pressure in healthy young subjects 39 . Also, in young-to-middle-age hypertensive adults, caffeinated coffee drinking has been linked with glomerular hyperfiltration 40 , which may be exaggerated in older individuals with pre-existing comorbidities 13 (e.g., hypertension) as our studied population, but warrants further studies. In fact, in experimental studies conducted in obese diabetic rats, administration of caffeine for 2 weeks induced early renal injury characterized by proteinuria, increased renal vascular resistance, and increased heart rate 41 . Moreover, the same authors reported that long-term caffeine consumption exacerbated renal failure and induced more severe tubule-interstitial and glomerular damage 42 . Nevertheless, this has not been explored in humans and RCTs are needed. Our study has some strengths. First, the relatively large sample size, its prospective design, and adjustment for many potential confounding factors. Secondly, we examined the association between eGFR and coffee subtypes, Table 3. Odds ratio (95% confidence intervals) of rapid eGFR decline (1-year changes in eGFR > 3 mL/ min/1.73 m 2 ) according to categories of coffee and tea consumption at baseline. Values are means (95% CI) unless otherwise indicated. eGFR estimated glomerular filtration rate, Ref. referent value. P value for trend based on multivariate logistic regression treating the median values of each category as an ordinal variable. Model 1: adjusted for sex (men or women), age (years, continuous), and baseline eGFR (mL/min/1.73 m 2 , continuous). Model 2: additionally adjusted for BMI (kg/m 2 , continuous), smoking (never, current, or former smoker), educational level (primary, secondary education or academic/graduate), physical activity (MET-min/ day, continuous), type 2 diabetes prevalence (yes or no), dyslipidemia (yes or no), hypertension (yes or no), center (four categories by number of recruited participants: < 200, 200-250, 250-300, or ≥ 300), intervention group (intervention or control group), MedDiet score (17-points, continuous) and total energy intake (kcal/ day, quartiles). Model 3: additional adjusted for dietary intakes of total protein, saturated fat, alcohol, fiber, sodium, magnesium, potassium (each categorized by energy-adjusted quartiles), 1-year changes in systolic blood pressure (mm Hg, continuous) and 1-year changes body weight (kg, continuous). Model 3 was further adjusted (cup/day, continuous) for coffee (for tea analysis), tea (for each coffee analysis), caffeinated coffee (for decaffeinated coffee analysis), or decaffeinated coffee (for caffeinated coffee analysis). All analyses were conducted with robust estimates of the variance to correct for intra-cluster correlation. a P < 0.05 (compared to the reference group). b P < 0.05. The standard serving size of 1 cup was defined as 50 mL in the questionnaire. www.nature.com/scientificreports/ which gives greater insight into the potential effect of coffee and caffeine on kidney function. Finally, we conducted sub-group analyses to examine whether kidney function decline was accelerated by unhealthy behaviors (e.g., smoking or a poor diet) or pre-existing comorbid conditions. Nevertheless, we also acknowledge some limitations. It should be underlined that the caffeine content in coffee, tea, and other beverages (e.g., soft drinks) and foods varies greatly, which may lead to misclassification, and also that coffee composition can depend on the type of coffee bean and the brewing process used, which may influence the biological effects it has on the human organism. Also, SCr-based eGFR, a common biomarker of kidney function used in almost all epidemiologic studies, may overestimate true GFR values. Finally, our study was conducted in aged Mediterranean individuals with MetS, therefore our findings cannot be extrapolated to other populations. In summary, caffeinated coffee and tea consumption and caffeine intake were associated with a greater 1-year decline of renal function in older individuals at high cardiovascular risk. Further studies are needed to clarify these associations; if proven, advice on coffee and tea consumption along with caffeine intake should be included in nutritional strategies for kidney disease prevention, particularly in individuals at high risk.

Data availability
The dataset (including data dictionaries) of PREDIMED-Plus is available to external investigators in order to make possible the replication of the main analyses used for the published article. However, due to the restrictions imposed by the Informed Consent and the Institutional Review Boards (IRB), bona fide investigators interested in analyzing the PREDIMED-Plus dataset may submit a brief proposal and statistical analysis plan to the corresponding author (JS-S) at jordi.salas@urv.cat. Upon approval from the Steering Committee and IRBs, the data will be made available to them using an onsite secure access data enclave. The study protocol is available at http:// predi medpl us. com/.  Table 4. Crude and multivariate linear regression models of 1-year changes in eGFR according to tertiles of caffeine intake at baseline. Values are means (95% CI) unless otherwise indicated. eGFR estimated glomerular filtration rate, Ref. referent value. P value for trend based on multivariate linear regression treating the median values of each category as an ordinal variable. Model 1: adjusted for sex (men or women), age (years, continuous), and baseline eGFR (mL/min/1.73 m 2 , continuous). Model 2: additionally adjusted for BMI (kg/m 2 , continuous), smoking (never, current, or former smoker), educational level (primary, secondary education or academic/graduate), physical activity (MET-min/day, continuous), type 2 diabetes prevalence (yes or no), dyslipidemia (yes or no), hypertension (yes or no), center (four categories by number of recruited participants: < 200, 200-250, 250-300, or ≥ 300), intervention group (intervention or control group), MedDiet score (17-points, continuous) and total energy intake (kcal/day, quartiles). Model 3: additionally adjusted for dietary intakes of total protein, saturated fat, alcohol, fiber, sodium, magnesium, potassium (each categorized at energy-adjusted quartiles), 1-year changes in systolic blood pressure (mm Hg, continuous), and 1-year changes in body weight (kg, continuous). All analyses were conducted with robust estimates of the variance to correct for intra-cluster correlation. a P < 0.05 (compared to the reference group). b P < 0.05. Total caffeine consumption was computed from FFQs taking into account the caffeine contained in caffeinated coffee, decaffeinated coffee, tea, regular sodas, artificially sweetened soda, and chocolate.  www.nature.com/scientificreports/