The Nurses’ Health Study (NHS) began in 1976 when 121,700 US registered female nurses aged 30 to 55 years replied to a mailed health questionnaire (70% response rate).15
The Health Professionals Follow-Up Study (HPFS) started in 1986 with 51,529 US male health professionals (dentists, veterinarians, pharmacists, optometrists, osteopaths, and podiatrists) aged 40 to 75 years who responded to a similar mailed health questionnaire (33% response rate).16
Participants have been followed with biennial questionnaires on numerous lifestyle habits including caffeine consumption and newly diagnosed illnesses such as glaucoma. Follow-up rates were high (> 95% of the total possible person-time through 2004). This study was approved by the Human Research Committees of Brigham & Women’s Hospital, Massachusetts Eye and Ear Infirmary and the Harvard School of Public Health. Our research adhered to the tenets of the Declaration of Helsinki.
The first dietary assessments including caffeine consumption occurred in 1980 for NHS and 1986 for HPFS and thus these are “baseline” years for this study and the study period was restricted to 1980 – 2004 in the NHS and 1986 – 2004 in the HPFS. Generally, a participant contributed person-time if they were at least aged 40 years (as glaucoma risk increases after age 40) and reported having had an eye exam in the period at risk (to minimize possible detection bias). Participants contributed person-time in approximate 2-year units from the return date of the first questionnaire until the earliest occurrence of either a report of glaucoma, cancer, death, loss to follow-up, or 2004.
At baseline, the following participants were excluded: 1) 23,239 women who did not respond to the 1980 semiquantitative food frequency questionnaire assessment (FFQ), 2) 5,994 women and 1,596 men with inadequate diet information on the FFQ (“adequate” information for women was fewer than 10 out of 61 items blank and 500–3500 kcal/day while fewer than 70 out of 131 items blank in the FFQ, with a total caloric intake range of 800–4200 kcal/day was considered “adequate” for men), 3) 3,624 women and 1,927 men with prevalent cancers aside from nonmelanoma skin cancer (this exclusion was applied because cancer diagnoses cause profound changes in lifestyle habits), 4) 801 women and 818 men with a prevalent diagnosis of glaucoma or glaucoma suspect, 5) 739 women and 973 men lost to follow-up immediately after baseline, and 6) 6,472 women and 3,658 men who never reported an eye exam during follow-up. After these exclusions, 80,831 women and 42,557 men remained. In addition, for each two-year period, participants who were under age 40 or who did not report an eye exam were also considered ineligible. After excluding those temporarily ineligible because they were under age 40 (17,045 women and 236 men) or did not report receiving an eye exam when first asked (see below; 19,046 women and 12,512 men), 44,740 women and 29,809 men respectively contributed person time in the first 2 years from the NHS (1980–82) and the HPFS (1986–88). At later periods, these ineligible participants were allowed to contribute person-time, if they reached 40 years of age, and reported receiving eye exams. Hence, by 2004, a total of 79,120 women and 42,052 men contributed person-time. Follow-up rates through 2004 were high (> 95% of the total possible person-time).
Eligibility for the eye exam criterion was determined by selecting those who responded positively to the question of whether an eye exam was received in the previous two years. For example, if a NHS participant answered positively only in 1994 and 1996, then she contributed person-time only during 1992–94 and 1994–96. Because this question was first asked in 1990 in both cohorts, eye exam eligibility was determined this way from the risk period 1988–1990 and onwards. For the initial periods 1980–88 in NHS and 1986–88 in HPFS, eye exam eligibility was based on responses to the first 1990 questions.
Measurement of caffeine consumption
In both cohorts, dietary intake data were collected repeatedly over follow-up with the use of a validated self-administered semiquantitative food frequency questionnaire (FFQ). FFQ’s were administered in 1980, ’84, ’86, ’90, ’94 and ‘98 for the NHS and in 1986, ’90,’94 and ‘98 for the HPFS. The NHS 1980 FFQ included 61 food / beverage items; the 1984 NHS FFQ was expanded to 116 items, and similar versions of it were used from 1986 on in both the NHS (126 items) and HPFS (131 items).
In the FFQ, participants were asked to report their average intake of a serving of a food or beverage over the preceding year. In the NHS, the 1980 FFQ asked about consumption of coffee with caffeine (in cups), tea with caffeine (in cups), and chocolate (in 1-ounce servings). From 1984 on in the NHS and from 1986 in the HPFS, the FFQ was expanded to include intakes of decaffeinated coffee (in cups) and separate items for caffeinated soda and caffeine-free sodas. The questionnaire provided nine response possibilities for intake frequency for each item ranging from “never or less than once per month” to “6 or more times per day”. To convert participants’ average intake of one serving of a caffeinated beverage over the preceding year into average daily intakes of caffeine, we used information obtained from U.S. Department of Agriculture food-composition sources. The average caffeine contents used for these calculations were 137 mg caffeine per cup of coffee, 47 mg caffeine per cup of tea, 46 mg caffeine per can or bottle of cola beverage, and 7 mg caffeine per serving of chocolate.
Validity of food frequency questionnaire assessment of caffeine
The reproducibility and validity of the NHS and HPFS FFQs have been reported previously.17, 18
Validation studies revealed a high correlation between self-reported intake of caffeinated beverages (cups / day) according to the FFQ and diet records over 4 weeks: in the NHS, the correlations were 0.78 for coffee, 0.93 for tea and 0.84 for cola drinks19
and in the HPFS, the correlations were 0.93 for coffee, 0.77 for tea and 0.84 for cola drinks.20
The glaucoma case ascertainment procedure is followed every two years and has three steps. First, in each mailed questionnaire administered every two years to participants, we ask about whether participants received eye exams and whether participants received a diagnosis of “glaucoma” from their eye care provider. In the second step, we follow-up on the participants who stated they received a diagnosis of glaucoma – we seek permission to retrieve their medical records related to their glaucoma diagnosis, then we request the eye care provider to complete a glaucoma questionnaire to provide us information on maximal IOP, information about the status of the filtration apparatus, structural information regarding the optic nerve, prior ophthalmic surgery, and any visual field loss or to send all relevant medical records. In the final step, we evaluated all the provided ophthalmic information from questionnaires / medical records and visual fields in a standardized manner.
All records were reviewed by a glaucoma specialist (LRP), masked to the caffeine consumption patterns of participants, to identify POAG cases according to standardized criteria. Only those appraised as either "definite" or "probable" POAG were included as cases in this analysis. For definite POAG cases, documentation of the following were required: (1) gonioscopy showed that angles were not occludable in either eye, (2) slit lamp biomicroscopy showed no indication in either eye of pigment dispersion syndrome, uveitis, exfoliation syndrome, trauma, or rubeosis, and (3) reproducible visual field (VF) defects were present and consistent with glaucoma (nasal step, nasal depression, paracentral scotoma, arcuate defects, or temporal wedge defects). For probable POAG cases, the slit lamp exam and visual fields criteria were also required, but for determining the angle of the anterior chamber, documentation of pupil dilation without subsequent adverse events was accepted in lieu of gonioscopy. For the cases included in the analysis, >70% met the criteria for “definite POAG”.
For all VF defects, we required that the same defect(s) be present on at least 2 reliable tests. There was no requirement for the type of perimetry performed; however, in 95% of cases, full static threshold testing was documented and only in <1% kinetic visual fields were used. For static threshold or suprathreshold testing, we consider the field reliable if the fixation loss rate was ≤ 33%, the false positive rate was ≤ 20% and the false negative rate was ≤ 20%. For kinetic visual fields, we consider the field reliable unless there is notation by the examiner to the contrary.
During follow-up, 5,809 women and 2,529 men self-reported a glaucoma diagnosis. These were confirmed by eye care providers in 67% of women and 58% of men as follows: POAG with VF loss (29% women; 27% men), only elevated IOP or optic disc cupping (19% women; 20% men) and other types of glaucomas or glaucoma suspects (19% women; 11% men). The remaining 33% of self-reports in women and 42% in men could not be confirmed, as the participants themselves (6% women; 11% men), or their eye care providers (4% women; 5% men) could not be contacted, participants did not give permission to review their records (10% women; 11% men), participants indicated the initial report was in error (11% women; 14% men) or participants’ eye doctors disconfirmed the diagnosis of POAG (2% women; 1% men).
Of the 1,680 women and 695 men confirmed to have POAG with VF loss by their eye care providers, 658 women and 353 men met the criteria for “definite” or “probable” cases of and were included in the analyses.
For the primary exposure, we calculated cumulatively updated caffeine intakes by averaging the intakes from all the available dietary assessments up to the start of each 2-year period at risk. As glaucoma is a slowly developing chronic condition, we chose to study cumulatively averaged caffeine intakes as they best represent long-term intake and average measures have less measurement error than single assessments.21
All caffeine intakes at each questionnaire were total energy adjusted using the residual method.22
Examining caffeinated beverages in addition to caffeine intake is important as results of such analyses strengthen the case for causality for caffeine if similar associations are found with caffeinated beverages that contribute to the caffeine intake. Also, individuals alter their caffeine intake predominately by altering their intake of caffeinated beverages, and thus the net effect of caffeinated beverages on the risk of primary open-angle glaucoma must be evaluated for possible public health recommendations. Thus we examined the risk of POAG in relation to categories of specific beverages: caffeinated coffee, tea, caffeinated soda and decaffeinated coffee. All intakes of specific caffeinated beverages were also cumulatively updated values. In the NHS, for soda and decaffeinated coffee, we used the data starting from 1984 when intakes of these beverages were first asked separately.
We calculated incidence rates of POAG by dividing the incident cases by person-years accrued for each caffeine or beverage intake category. We adjusted for age using 5-year categories, and calculated Mantel-Haenszel age-adjusted incidence rate ratios (RR) and their 95% confidence intervals (CIs). For multivariable analyses, we controlled for potential glaucoma risk factors by including them simultaneously in Cox proportional hazards analysis stratified by age in months and the specific 2-year period at risk.23
We conducted tests for trend by including the midpoint values within each intake category.
Variables considered for inclusion were family history of glaucoma, African-American heritage (yes / no), body mass index (kg / m2), pack years of smoking, physical activity (quartiles of activity intensity / day), cumulatively updated alcohol intake (g /day), report of a physician exam, self reported history (yes / no) of hypertension, diabetes, cataract or age-related macular degeneration diagnoses, and total fluid intake (liters / day). Updated information on covariates was obtained from the biennial questionnaires; cumulatively updated alcohol intake and total fluid intake (based on intake of nearly 30 different types of beverages) was calculated using responses to the FFQs.
We first analyzed the data from each cohort separately and performed tests for heterogeneity of the cohort specific results to check for appropriateness of pooling the results. Then, we pooled the results using meta-analytic methods incorporating random effects.24
Effect modification and secondary analyses
We performed several secondary analyses. First, we examined the influence of timing of exposure by examining caffeine intake only at baseline or at the most recent questionnaire. Second, we evaluated whether detection bias may have influenced the results, especially if caffeine consumption is related to better eye care. For this, we adjusted for other predictors of greater ophthalmic surveillance (i.e. number of eye exams, history of physician exams, diagnoses of other eye diseases, namely cataract and age-related macular degeneration).
We conducted additional analyses where we additionally adjusted for cumulatively updated total fluid intake. Because drinking a large quantity of fluids, particularly in a short period of time, generally causes IOP elevations,25
we conducted this analysis to determine the association with caffeine intake, that was independent of total fluid intake.
Because caffeine increases intraocular pressure, we hypothesized that higher caffeine intake may be more strongly associated with glaucoma that is more likely to be IOP-related optic nerve damage. Thus we separately analyzed the risk of “high-tension” POAG defined as those with maximum IOP ≥ 22 mm Hg before visual field loss (67.5% of all POAG cases).
Also, we conducted an analysis of caffeine intake only among participants who never smoked and who were past or current smokers. Caffeine metabolism is influenced by cigarette smoking; smoking induces the enzyme cytochrome P450 1A2 which is also involved in caffeine metabolism, and thereby reduces the effective dose of caffeine.26
Thus we hypothesized that the effect of the same dose of caffeine may be greater among non-smokers.
Finally, to determine whether the influence of caffeine intake differed by inherent susceptibility to POAG, we examined the associations between caffeine intake and glaucoma separately among those with and without a self-report of family history of glaucoma. The questions on family history of glaucoma were first asked to all NHS and HPFS participants in the 2000 follow-up questionnaires; self-report of family history of glaucoma was defined as a positive answer to history of glaucoma in either of the biologic parents or in any siblings. In this and in other stratified analyses described above, we have statistically tested for effect modifications by testing the significance of pooled results of interaction terms in models.