eTable 1. Age- and multivariate-adjusted hazard ratios for rosacea by tea, soda, and chocolate in the Nurses’ Health Study II (1991-2005)
eTable 2. Association between caffeine intake and rosacea stratified by smoking status in the Nurses’ Health Study II (1991–2005)
eTable 3. Association between caffeine intake and rosacea stratified by alcohol intake in the Nurses’ Health Study II (1991–2005)
eTable 4. Association between caffeine intake and rosacea stratified by body mass index in the Nurses’ Health Study II (1991–2005)
eTable 5. Association between caffeine intake and rosacea stratified by physical activity in the Nurses’ Health Study II (1991–2005)
eFigure 1. Dose-response hazard ratio (with 95% CI) of incident rosacea by caffeine intake
Customize your JAMA Network experience by selecting one or more topics from the list below.
Li S, Chen ML, Drucker AM, et al. Association of Caffeine Intake and Caffeinated Coffee Consumption With Risk of Incident Rosacea in Women. JAMA Dermatol. 2018;154(12):1394–1400. doi:10.1001/jamadermatol.2018.3301
Is there an association between risk of incident rosacea and caffeine intake, including from coffee consumption?
In this cohort of 82 737 participants in the Nurses’ Health Study II, we identified 4945 incident cases of rosacea, and found a significant inverse association between risk of rosacea and increased caffeine intake, particularly that from coffee. This association was not found for caffeine intake from other food sources (tea, soda, and chocolate).
Our findings do not support limiting caffeine intake as a means to prevent rosacea and may have implications for the causes of and clinical approach to rosacea.
Caffeine is known to decrease vasodilation and have immunosuppressant effects, which may potentially decrease the risk of rosacea. However, the heat from coffee may be a trigger for rosacea flares. The relationship between the risk of rosacea and caffeine intake, including coffee consumption, is poorly understood.
To determine the association between the risk of incident rosacea and caffeine intake, including coffee consumption.
Design, Setting, and Participants
This cohort study included 82 737 women in the Nurses’ Health Study II (NHS II), a prospective cohort established in 1989, with follow-up conducted biennially between 1991 and 2005. All analysis took place between June 2017 and June 2018.
Data on coffee, tea, soda, and chocolate consumption were collected every 4 years during follow-up.
Main Outcomes and Measures
Information on history of clinician-diagnosed rosacea and year of diagnosis was collected in 2005.
A total of 82 737 women responded to the question regarding a diagnosis of rosacea in 2005 in NHS II and were included in the final analysis (mean [SD] age at study entry, 50.5 [4.6] years). During 1 120 051 person-years of follow-up, we identified 4945 incident cases of rosacea. After adjustment for other risk factors, we found an inverse association between increased caffeine intake and risk of rosacea (hazard ratio for the highest quintile of caffeine intake vs the lowest, 0.76; 95% CI, 0.69-0.84; P < .001 for trend). A significant inverse association with risk of rosacea was also observed for caffeinated coffee consumption (HR, 0.77 for those who consumed ≥4 servings/d vs those who consumed <1/mo; 95% CI, 0.69-0.87; P < .001 for trend), but not for decaffeinated coffee (HR, 0.80; 95% CI, 0.56-1.14; P = .39 for trend). Further analyses found that increased caffeine intake from foods other than coffee (tea, soda, and chocolate) was not significantly associated with decreased risk of rosacea.
Conclusions and Relevance
Increased caffeine intake from coffee was inversely associated with the risk of incident rosacea. Our findings do not support limiting caffeine intake as a means to prevent rosacea. Further studies are required to explain the mechanisms of action of these associations, to replicate our findings in other populations, and to explore the relationship of caffeine with different rosacea subtypes.
Rosacea is a common chronic inflammatory skin disease.1,2 Many triggers for rosacea have been postulated, including caffeine, hot beverages, sunlight, spicy foods, strenuous exercise, and hormonal factors.3-8
The reported direction and magnitude of the association between the risk of rosacea and caffeine and coffee intake in prior epidemiologic studies have been inconsistent.8-11 Previous studies did not differentiate between caffeinated and decaffeinated coffee and other beverages, and the distinction between amounts of caffeine and coffee consumed was made in only 1 study.8 We conducted the first large cohort study to our knowledge to evaluate the association between caffeine intake, coffee consumption, and risk of incident rosacea in a large cohort of women from the Nurses’ Health Study II (NHS II).
Details of NHS II have been described previously.12,13 The study was approved by the institutional review board of Brigham and Women’s Hospital and Harvard School of Public Health, Boston, Massachusetts. Participants’ completion and return of the questionnaires were considered informed consent.
In the cohort, participants were asked about their intake of food and beverages every 4 years. Participants could report the number of servings by selecting from 9 frequency responses (never, 1-3/mo, 1/wk, 2-4/wk, 5-6/week, 1/d, 2-3/d, 4-5/d, and 6/d) for caffeinated coffee, decaffeinated coffee, tea, soft drinks, and chocolate. The total caffeine intake was calculated by summing the caffeine content for a determined amount of each item and multiplying that by its frequency of intake, for which the methods have been detailed previously.14 Food frequency questionnaires have high reproducibility and validity, with strong correlations with multiple dietary records for coffee (r = 0.78) and other caffeine-rich beverage intake (tea, r = 0.93; caffeinated carbonated beverages, r = 0.84).15,16
In 2005, NHS II participants were asked if they had clinician-diagnosed rosacea, and, if so, the year of diagnosis in 5 time intervals (before 1991, 1991-1994, 1995-1998, 1999-2002, or 2003-2005). We recently reviewed the medical records of 16 participants who reported a diagnosis of rosacea in 2005, 14 of whom had information on rosacea in their medical records. The diagnosis of rosacea was confirmed in 12 of 14 cases. In a separate clinic-based validation study (unpublished data, W.-Q.L., June 1, 2018), the rosacea assessment question used in NHS II was validated (“Have you ever had clinician-diagnosed rosacea?”); in 37 patients with confirmed rosacea and 18 patients without rosacea, comparison between medical record diagnosis and self-reported rosacea yielded a sensitivity of 89%, a specificity of 92%, and a positive predictive value of 94%, which, though obtained from a different population, lends support to the validity of rosacea diagnosis in NHS II.
A total of 82 737 participants responded to the question regarding a diagnosis of rosacea in 2005 and were included in the final analysis. Person-years were calculated from the return date of the 1991 questionnaire to the date of diagnosis of rosacea or end of follow-up in June 2005, whichever came first. To estimate long-term intake of caffeine, caffeinated coffee, decaffeinated coffee, and other food and drinks containing caffeine, updated cumulative averages for intake were used from all available frequency questionnaires from different years.
We conducted Cox proportional hazards analyses to estimate age- and multivariate-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for the risk of incident rosacea associated with caffeine intake and coffee consumption. For these analyses, caffeine intake was categorized as quintiles, with cutoffs derived from caffeine intake in 1991 in NHS II. Coffee consumption was categorized a priori into 5 serving groups (<1/mo, 1/mo to 4/wk, 5-7/wk, 2-3/d, or ≥4/d). Caffeine intake from coffee and caffeine intake from other food sources were assessed as continuous variables. Trend tests were carried out using continuous measures by assigning the median to each category. The adjusted Cox proportional hazard analyses were fitted to a restricted cubic spline model to obtain the HR of rosacea as a function of caffeine intake with adjustment for covariates.
We conducted stratified analyses by smoking status, alcohol intake, body mass index (BMI, calculated as weight in kilograms divided by height in meters squared), and physical activity, as well as interaction analyses between these factors and the main exposure because they have been associated with rosacea in prior studies.11,17-19
A 4-year lag analysis excluding rosacea cases documented within the first 4 years of each updated assessment of caffeine and coffee intake was performed to address potential reverse-causation bias. To address the concern of potential residual confounding, we conducted a sensitivity analysis additionally adjusting for oral contraceptive use, cumulative UV (UV) flux, personal history of major chronic diseases, antidepressant medication use, and phobic anxiety. Because rosacea epidemiology varies with respect to race,20 a secondary sensitivity analysis was performed excluding nonwhite participants.
All statistical analyses were conducted using SAS software, version 9.4. All statistical tests were 2-tailed with a significance level of P < .05.
Table 1 summarizes the participant characteristics in 1991 stratified by quintile of caffeine intake. The proportion of current and past smokers and users of oral contraceptives were increased with increasing caffeine intake. In addition, participants with higher caffeine intake tended to be older and have higher alcohol intake. Other characteristics were similar among the 5 groups of caffeine intake. The biggest source of caffeine was caffeinated coffee, which showed the greatest difference in intake across quintiles of overall caffeine intake among the investigated food and drink groups.
During the 1 120 051 person-years of follow-up, we identified 4945 incident cases of rosacea. As supported by the data reported in Table 2, We found a significant inverse association between increased caffeine intake and risk of rosacea. From the lowest to highest quintile of caffeine intake, the cohort age-adjusted incidence rates (AAIRs) of rosacea were 504, 502, 501, 478, and 372 per 100 000 person-years, respectively. Compared with the lowest quintile, the absolute risks of rosacea were decreased by 2, 3, 26, and 132 per 100 000 person-years, respectively, for the second to fifth quintile. The multivariate-adjusted HRs (95% CIs) for rosacea from the lowest to highest quintiles of caffeine intake were 1 [reference], 0.91 (0.84-1.00), 0.92 (0.84-1.00), 0.85 (0.77-0.93), and 0.76 (0.69-0.84) (P < .001 for trend). The approximately linear downward-sloping restricted cubic spline curve (eFigure 1 in the Supplement) demonstrates the inverse association between caffeine intake and the risk of rosacea.
As supported by the data reported in Table 3, a significant inverse association was also observed between caffeinated coffee consumption and risk of incident rosacea (P < .001 for trend). Compared with individuals who consumed less than 1 serving of caffeinated coffee per month (AAIR, 495/100 000 person-years), participants who consumed 4 servings per day or more had the lowest risk of rosacea (AAIR, 364/100 000 person-years; HR = 0.77; 95% CI, 0.69-0.87). Decaffeinated coffee consumption was not associated with a decreased risk of rosacea (P = .39 for trend).
The data reported in Table 4 supports the association of caffeine intake with incident rosacea based on the caffeine source. Caffeine intake from coffee was inversely associated with risk of incident rosacea (P < .001 for trend), whereas caffeine intake from other sources (tea, soda, and chocolate) showed no association with risk of rosacea (P = .58 for trend).
We further examined the risk of rosacea associated with servings of caffeinated tea, caffeinated soda, and chocolate, and these data are reported in eTable 1 in the Supplement. We did not find a significant association between caffeinated tea or soda and risk of rosacea (P = .30 for trend and P = .08 for trend, respectively). Results suggested chocolate as a potential risk factor for rosacea (P = .04 for trend) (eTable 1 in the Supplement).
Analyses stratified by smoking, alcohol intake, physical activity, and BMI showed generally similar associations between caffeine intake and risk of rosacea as detailed with supporting data in eTables 2 through 5 in the Supplement. We did not find effect modification by smoking status (P = .37 for interaction), alcohol intake (P = .13 for interaction), physical activity (P = .33 for interaction), or BMI (P = .61 for interaction) on the association between caffeine intake and risk of rosacea.
The 4-year lag analysis and sensitivity analyses did not change the results materially (data not shown).
In the present study, we found that caffeine intake from coffee but not from other foods (tea, soda, and chocolate) was associated with a decreased risk of incident rosacea in a dose-dependent manner. Although the relative risk of rosacea associated with caffeine intake and caffeinated coffee consumption was moderate, the absolute risk of rosacea was decreased remarkably by 132 per 100 000 person-years for the highest vs lowest quintile of caffeine intake and 131 per 100 000 person-years for caffeinated coffee consumption of 4 servings per day or more vs less than 1 serving per month.
Previous case-control studies and review articles have asserted differing effects of caffeine intake or coffee consumption on the risk of rosacea.8-11 One case-control study from Estonia9 and a literature review from France10 found no significant difference in risk of rosacea between groups consuming different amounts of caffeine. A case-control study from Poland reported increased risk of rosacea among coffee drinkers.11 In addition, a clinical narrative review and randomized clinical study analyzed caffeine as a potential trigger for rosacea flares.3,8 In the randomized clinical trial,8 participants with rosacea consumed caffeinated coffee and water at different temperatures. Caffeinated coffee was shown to have no effect on flushing in patients with rosacea, whereas heat led to flushing reactions.8
A variety of specific agents and mechanisms may be responsible for caffeine’s influence on rosacea. One possibility is its effect on vascular contractility. Specifically, vasodilation from neurovascular dysfunction has been documented in the pathogenesis of rosacea, particularly papulopustular rosacea,21,22 and caffeine is known to lead to vasoconstriction23,24 through its effect on the renin-angiotensin-aldosterone system.25 Increased caffeine intake may decrease vasodilation and consequently lead to diminution of rosacea symptoms. Second, caffeine has been documented to contain antioxidant agents and to have immunosuppressant effects,26-30 which may result in decreased inflammation in rosacea. Third, hormonal factors have been implicated in the development of rosacea,5-7 and caffeine can modulate hormone levels, including levels of adrenaline, noradrenaline, and cortisol.25
Heat has been shown to be a trigger factor in patients with rosacea.8 That decaffeinated coffee showed no association suggests that ingredients other than caffeine may have counteracted the effects of heat. One potential ingredient is polyphenol present in coffee. Polyphenols have antioxidant, anti-inflammatory, and vascular effects,31 and they have been shown benefit in rosacea treatment, especially for facial erythema, papules, and pustules.31 Further studies are needed to elucidate determinants for risk of rosacea in decaffeinated coffee.
We hypothesize that the lack of association with rosacea found for caffeinated food and drinks other than coffee is due to the low absolute intake of caffeine from sources other than coffee. For the different quintiles of total caffeine intake, the tea, soda, and chocolate consumption amounts did not show an increasing trend, illustrating the dominance of coffee as a source of caffeine. An alternative explanation is that coffee may contain other compounds that lower the risk of rosacea. However, since no association was found with decaffeinated coffee consumption, caffeine is the putative component of coffee responsible for the inverse association between coffee and risk of rosacea.
The positive association found between chocolate consumption and risk of rosacea has a few possible explanations. First, the amount of caffeine per serving of chocolate varies widely.32,33 Second, chocolate itself may be a risk factor for rosacea.34 Because the caffeine content in chocolate is low, other compounds may be responsible for the observed association.
We acknowledge some limitations. First, data on lifetime diagnosis of rosacea and diagnosis year were self-reported in 2005 by participants, leaving our study prone to recall bias. However, misclassification of rosacea would be expected to be nondifferential with respect to caffeine and coffee intake. Our lag analysis limits the impact of misclassification of year of diagnosis. The validation study based on medical record review and the clinic-based validation study provide some support for the validity of the rosacea self-reports (unpublished data, W.-Q.L., June 1, 2018). However, we were only able to review the medical records for a small subset of cases to verify the accuracy of self-reported rosacea in the cohort. Efforts are warranted to better assess the accuracy of self-reported rosacea in the cohort.
Second, caffeine intake, consumption of coffee and other beverages, was assessed in 4-year intervals. Third, etiologic heterogeneity may underlie different types of rosacea,1 but we did not have data on rosacea subtypes. Fourth, although we had detailed data on many covariates, we cannot rule out the possibility of residual confounding from unmeasured confounders (such as family history, stress, heat, and hot beverages) or imperfectly measured confounders (as were adjusted for in our analyses). Compounds other than caffeine in the food and drinks investigated may influence risk of rosacea. Fifth, all participants were well-educated women, and most were white, which limits the generalizability of our findings.
In summary, we provide evidence that caffeine intake and caffeinated coffee consumption are associated with a decreased risk of incident rosacea. Our study may have implications for the causes of and clinical approach to rosacea. Our findings do not support limiting caffeine intake as a preventive strategy for rosacea. Further studies are required to explain the underlying mechanisms of observed associations and to explore the relationship of caffeine with rosacea subtypes.
Corresponding Author: Wen-Qing Li, PhD, Department of Dermatology, Warren Alpert Medical School, Brown University, 339 Eddy St, Providence, RI 02912 (email@example.com).
Accepted for Publication: August 30, 2018.
Published Online: October 17, 2018. doi:10.1001/jamadermatol.2018.3301
Author Contributions: Dr Wen-Qing Li had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Dr Suyun Li and Mr Chen are co–first authors.
Study concept and design: Drucker, Cho, Qureshi, W.-Q. Li.
Acquisition, analysis, or interpretation of data: S. Li, Chen, Drucker, Cho, Geng, W.-Q. Li.
Drafting of the manuscript: Chen, W.-Q. Li.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: S. Li.
Obtained funding: W.-Q. Li.
Administrative, technical, or material support: Cho.
Study supervision: Qureshi, W.-Q. Li.
Conflict of Interest Disclosure: Dr Drucker has served as an investigator and has received research funding from Sanofi and Regeneron and has been a consultant for Canadian Agency for Drugs and Technologies in Health, Sanofi, RTI Health Solutions, and Eczema Society of Canada. He has received honoraria from Astellas Canada, Prime Inc, Spire Learning, CME Outfitters, and Eczema Society of Canada. Dr Qureshi is Consultant for Abbvie, Amgen, Centers for Disease Control and Prevention, Janssen, Merck, Novartis, and Pfizer. No other disclosures are reported.
Funding/Support: This study was supported in part by the Research Career Development Award of Dermatology Foundation, Richard B. Salomon Faculty Research Award of Brown University, and the National Institute of Health grants for Nurses’ Health Study II (UM1 CA176726).
Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We are indebted to the participants and staff of the NHS II and Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital (as home of the NHS II) for their valuable contributions as well as the following state cancer registries for their help: Alabama, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Idaho, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Nebraska, New Hampshire, New Jersey, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, Tennessee, Texas, Virginia, Washington, and Wyoming.