Coffee and tea drinking in relation to risk of hip fracture in the Singapore Chinese Health Study

Abstract

Meta-analyses of studies conducted among Western populations suggest that coffee consumption does not affect osteoporotic fracture risk. However, experimental studies have shown that the effect of caffeine on bone health may depend on dosage. We examined the associations between consumption of coffee, tea and caffeine and risk of hip fracture in an Asian cohort. In a population-based prospective cohort of 63,257 Chinese men and women aged 45–74 years in Singapore, a validated semi-quantitative food frequency questionnaire was used to assess habitual consumption of coffee and tea at baseline. Cox proportional hazards regression models were used to estimate hazard ratio (HR) and 95% confidence interval (CI) for risk of hip fracture with adjustment for potential confounders. During a mean follow-up of 16.7 years, 2,502 incident hip fracture cases were identified. Compared to coffee drinkers <1 cup/week, those who drank ⩾ 4 cups/day had a statistically significant higher risk to develop hip fractures, the HR (95% CI) was 1.32 (1.07, 1.63) in the whole cohort analysis, 1.46 (1.01, 2.10) for men and 1.33 (1.02, 1.72) for women. Among postmenopausal women, compared to those who drank coffee <1 cup/week, drinking 2–3 cups/day was associated with the lowest risk [HR: 0.88 (0.76, 1.01)] and drinking ≥4 cups/day was associated with the highest risk [HR: 1.31 (1.00, 1.71)]. Similar associations with caffeine intake were found among postmenopausal women. Restricted spline analyses suggested a non-linear association between coffee/caffeine consumption and hip fracture risk in postmenopausal women (p for non-linearity ≤ 0.05). No association was found with tea consumption in either sex. These data suggest that drinking coffee ≥4 cups/day is associated with a higher hip fracture risk, while a moderate intake may alleviate risk in postmenopausal women. Future studies should corroborate these results to determine levels of optimal coffee consumption in relation to bone health.

Introduction

Coffee and tea are among the most widely consumed beverages in the world, and have been studied extensively on health impacts, including osteoporosis and related fractures ^(1,2). However, the effects of coffee and tea on bone health are mixed in experimental and observational studies. This could partially be due to drinking habits and the complexity of chemical constituents in tea and coffee.

Coffee is the major source of caffeine and contains other non-caffeine compounds such as cafestol, chlorogenic acid, caffeic acid and trigonelline ^(1,3). In experimental settings, caffeine at a high dosage has been shown to suppress osteogenesis ⁽⁴⁾, increase osteoblastic apoptosis ⁽⁵⁾, and osteoclastic formation and differentiation ⁽⁶⁾ in vitro. In animal models, high caffeine intake has been shown to inhibit bone growth, decrease bone mineral density ^(5–7) and interrupt bone healing ⁽⁸⁾. By contrast, caffeine at low to moderate doses has been suggested to enhance osteogenesis ⁽⁴⁾ and increase bone mass and strength in oophorectomized rats ⁽⁹⁾. Non-caffeine compounds found in coffee, such as caffeic acid, chlorogenic acid ⁽¹⁰⁾ or trigonelline ⁽³⁾, on the other hand, have been shown to exert unfavorable effects in the skeletons in ovariectomized rats. Hence, it is possible that the effect of coffee consumption on bone health may depend on dosage and the composition of caffeine and non-caffeinated compounds in coffee. Tea is an important source of polyphenols and has a considerably lower level of caffeine than coffee based on volume. Studies have shown that polyphenols from black tea (theaflavin-3, 3′-digallate) and green tea (epigallocatechin-3-gallate) inhibit osteoclastic formation and differentiation ⁽¹¹⁾, reduce bone loss ^(12,13) and improve microarchitecture ⁽¹²⁾ in ovariectomized rats.

In observational studies, no significant results have been found between consumption of coffee or tea and fracture risk, where the pooled relative risk in meta-analyses comparing the highest to the lowest category of coffee intake ranged from 0.94 to 1.30 for hip fractures ^(14,15) and fractures in general ⁽¹⁶⁾. Additionally, the effect of a high intake of caffeine on osteoporosis may not be harmful, particularly among those with sufficient calcium intake ⁽¹⁷⁾. Further, the amount used to define coffee intake levels varied among studies, which may contribute to high heterogeneity (all I² > 56%) ^(14–16) and inconsistent effect estimates. Stratified analysis by gender in these meta-analyses suggested a reduced risk in men but an increased risk in women for the association with coffee consumption ^(14–16). Data in these meta-analyses were further re-analyzed by Poole and colleagues and confirmed the inconsistent associations and a sex-specific difference between coffee consumption and fracture risks ⁽¹⁷⁾. Similarly, inconsistent results were reported for the association with tea ^(2,14).

The results from previous observational studies are primarily conducted in Caucasian populations in Europe or North America ^(16,18). To our knowledge, there is only one Asian study of a case-control study among Japanese men and women, suggesting that those who consumed three or more cups of coffee per day had a higher fracture risk ⁽¹⁹⁾. In the context of a relatively lower consumption of coffee or caffeine, and a rise in hip fracture incidence in Asian populations ⁽¹⁹⁾, we comprehensively examined the longitudinal relationship between consumption of coffee and tea, as well as estimated caffeine intake, and the risk of hip fracture in a large Asian cohort. We further tested for evidence of a potential non-linear relation as suggested in experimental findings. We hypothesized that a high intake of coffee/caffeine might increase the risk of hip fractures while a low intake might reduce the risk.

Methods

Study population

The current study was conducted in the Singapore Chinese Health Study, a population-based prospective cohort of 63,257 men (n=27,959) and women (n=35,298), aged 45–74 years at recruitment between April 1993 and December 1998. The study participants were restricted to two major dialect groups in Singapore, the Hokkiens and the Cantonese, who originated from Fujian and Guangdong provinces in Southern China, respectively. All of our study participants were recruited from residents living in the government housing estates, in which 86% of the Singapore population resided during the time of recruitment. This study was approved by the Institutional Review Board at the National University of Singapore, and all enrolled participants gave written informed consent.

Baseline and dietary assessment

At baseline, an in-person interview was administered to all participants using a structured questionnaire. Baseline information included demographics, weight, height, cigarette smoking, alcohol consumption, dietary intakes, physical activity, medical history such as physician-diagnosed diabetes and stroke, and detailed menstrual and reproductive history (women only).

We used a 165-item validated semi-quantitative food frequency questionnaire (FFQ) to record participants’ habitual dietary intakes during the past year at enrollment. The details regarding the development and validation of the FFQ were reported previously ⁽²⁰⁾. For intake of coffee and tea, participants were asked, on average, how often they consumed one cup of coffee (instant or freshly brewed), one cup of “three-in-one preparation” coffee, one cup of Chinese red tea or Ceylon tea (i.e. black tea), and one cup of green tea, such as jasmine tea. The three-in-one preparation coffee is an instant coffee beverage widely drunk in Singapore and other parts of Asia. It is a mixture of coffee powder, non-dairy creamer, and refined cane sugar packed in individual packs; the contents of each pack are dissolved in hot water to make a cup of coffee. The serving size of one cup of coffee or tea was assigned as 237 mL. Hence, for each participant, the consumption of coffee or tea was the product of 1 cup (tea or coffee) and the frequency of consumption in nine predetermined categories: never or hardly ever, 1–3 cups a month, 1 cup a week, 2–3 cups a week, 4–6 cups a week, 1 cup a day, 2–3 cups a day, 4–5 cups a day, and 6 or more cups a day. Because decaffeinated coffee or tea was rarely consumed in our study population at the time of recruitment, all coffee and tea consumed were assumed to be caffeinated.

Dietary intake of caffeine and other nutrients was estimated from the Singapore Food Composition Database, which listed 96 nutritional and non-nutritional values per 100 grams of the edible foods and beverages specific for this cohort. The foundation of this database relied heavily on the data published by the U.S. Department of Agriculture; and it was supplemented from multiple resources for other foods and components and referenced from published food composition tables from China, Malaysia and Taiwan ⁽²⁰⁾. Coffee and tea accounted for 84% and 12% of total dietary caffeine intake in this cohort, respectively. The remaining minor contribution (4%) came from other caffeinated food sources such as soda (0.7%), cocoa drinks and chocolate-related food items.

Case ascertainment

Hip fracture cases were identified via record linkage with the hospital discharge database of the MediClaim System, which captures inpatient discharge information from all public and private hospitals in Singapore. Discharge diagnoses were coded according to the ninth revision of the International Classification of Diseases (ICD-9) before 2012 and the tenth revision (ICD-10) since 2012. All cases of hip fractures were identified based on ICD-9 820 and ICD-10 S72.0-S72.2, and further verified manually by surgical notes or medical records. As of December 31, 2014, only 47 (0.7%) subjects from this cohort were known to be lost to follow-up due to migration out of Singapore or for other reasons. The causes and dates of death of the cohort participants were ascertained through record linkage with the population-based Singapore Registry of Births and Deaths. After excluding four cases of traumatic fractures from road traffic accidents, and one case due to cancer metastasis in the femur, 2,605 hip fracture cases were identified through the record linkage. We further excluded 103 participants with a history of hip fracture that occurred before the baseline interview.

Statistical analysis

For baseline characteristics, Chi-square test for categorical variables and Student’s t-test for continuous variables were used to compare baseline characteristics between cases and non-cases of incident hip fracture. For each study subject, person-years were counted from the date of baseline interview to the date of hip fracture occurrence, lost to follow-up, death, or 31 December 2014, whichever occurred first. For both coffee and tea drinking, we used multivariable Cox proportional hazard regression to estimate hazard ratio (HR) and the 95% confidence interval (CI) for the relative risk of hip fracture by comparing a higher category to the lowest category of consumption (<1 cup/week, referent group). For coffee, we further categorized our participants into 4 comparator groups: <1 cup/day, 1 cup/day, 2–3 cups/day and ≥4 cups/day. For tea, there were two comparator groups including weekly (1–6 cups/week) and daily drinkers (1 or more cups/day). Since one cup of coffee contains approximately 100 mg caffeine ⁽¹⁾, we studied daily caffeine intake in the categories of 100–<200 mg, 200-<300 mg, or ≥300 mg versus <100 mg (referent group) to estimate relative risk. We did not identify any violation of the proportional hazards assumption or multicollinearity among the covariates that were entered in the models.

The covariates included in the model were age (years), sex (for whole cohort analysis), year at recruitment (1993–1995 and 1996–1998), dialect group (Hokkien, Cantonese), level of education (no formal education, primary school, secondary school or higher), body mass index (BMI) (kg/m²), smoking status (never smokers, ex-smokers, and current smokers), moderate physical activity (none, 0.5–<4 hours weekly, ≥4 hours weekly), total energy intake (kcal/day), dietary calcium (mg/1,000 kcal/day in quartiles), at least weekly use of vitamins/minerals (yes, no), vegetable-fruit-soy dietary pattern scores (in quintiles), use of hormone replacement therapy at recruitment (postmenopausal women only; yes, no), and baseline self-reported physician-diagnosed history of diabetes mellitus (yes, no) and stroke (yes, no). The vegetable-fruit-soy pattern represents diet quality and has been shown to be related to a lower risk of hip fracture in this cohort ⁽²¹⁾. We also adjusted for tea consumption or coffee intake simultaneously for the association with coffee or tea, respectively. In addition, we examined evidence for a non-linear relation between coffee/caffeine intake and risk of hip fracture using restricted cubic spline analysis. Four knots were created at the 5^th, 35^th, 65^th and 95^th percentiles according to the distribution of daily coffee or caffeine intake among all study participants. The results of these analyses are presented as smoothed curves with 95% confidence intervals.

Finally, we performed sensitivity analyses by excluding those who had reported extreme energy intakes (≤ 600 or ≥ 3,000kcals). All statistical analyses were conducted using SAS Version 9.2 (SAS Institute, Inc., Cary, North Carolina) except for the restricted cubic spline analysis, which was performed using Stata version 13 (Stata Corporation, College Station, Texas, USA). A p-value < 0.05 is considered statistically significant and all reported p values are two-sided.

Results

After a mean (standard deviation, SD) follow-up of 16.7 (5.2) years among a total of 63,154 participants, 2,502 incident hip fracture cases were identified through the nationwide hospital linkage. The mean age (SD) at hip fracture was 79.4 (6.9) years. Women accounted for 72.7% of all hip fractures. The age-standardized incidence rate of hip fractures was 294 per 100,000 person-years for women and 157 per 100,000 person-years for men. About 70% of our cohort participants were daily coffee drinkers (≥1 cup/day) and 47% were weekly tea drinkers (≥1 cup/week). Compared to women, men consumed more coffee or caffeine (p<0.001). The mean (SD) for coffee consumption (cups/week) was 9 (8) for women and 11 (9) for men; the mean (SD) for caffeine (mg/day) was 130 (100) for women and 171 (116) for men. Table 1 shows the baseline characteristics between cases and non-cases of hip fracture incidence by sex. In both men and women, hip fracture cases were older and leaner than non-cases (p<0.058). Additionally, those who developed hip fractures were more likely to smoke, consume a less healthy diet and have higher prevalence of diabetes and stroke (p<0.001).

Table 1.

Baseline characteristics [n (%) or mean ±S.D.]¹ of cases versus non-cases of incident hip fracture by gender, the Singapore Chinese Health Study, 1993–2014

Characteristics	Men (n=27,913)		Women (n=35,241)
	Fracture cases	Non-cases	Fracture cases	Non-cases
Number	683	27,230	1,819	33,422
Age at recruitment (years)	62.3±7.21	56.6 ±7.9	63.3±6.8	55.9±7.9
Body mass index (kg/m2)	22.7±2.9	23.0 ±3.2	23.1±3.3	23.2±3.3
Level of education, n (%)
Secondary or higher	157 (23.0)	10,418 (38.3)	169 (9.3)	7,114 (21.3)
Smoking status, n (%)
Never smokers	257 (37.6)	11,452 (42.1)	1,577 (86.7)	30,569 (91.4)
Former smokers	183 (26.8)	5,904 (21.7)	73 (4.0)	823 (2.5)
Current smokers	243 (35.6)	9,874 (36.3)	169 (9.3)	2,030 (6.1)
Moderate physical activity, n (%)
None	493 (72.2)	20,518 (75.4)	1,435 (78.9)	26,735 (80.0)
0.5 – <4 hours/week	96 (14.1)	4,209 (15.5)	244 (13.4)	4,235 (12.7)
≥4 hours/week	94 (13.8)	2,503 (9.2)	140 (7.7)	2,452 (7.3)
At least weekly use of vitamin supplements, n (%)	26 (3.8)	1,320 (4.9)	121 (6.7)	2,537 (7.6)
Calcium (mg/day)	412±216	419±219	383±222	414±239
Lowest quintile of vegetable-fruit-soy pattern score, n (%)	176 (25.8)	6009 (22.1)	444 (24.4)	6001 (18.0)
Diabetes mellitus, n (%)	75 (11.0)	2,340 (8.6)	328 (18.0)	2,925 (8.8)
Stroke, n (%)	30 (4.4)	474 (1.7)	52 (2.9)	380 (1.1)
Postmenopausal women, n (%)			1,734 (95.3)	23,788 (71.2)
Hormone replacement therapy among postmenopausal women only, n (%)			22 (1.3)	1,279 (5.4)

¹Chi-square test was used for categorical variables; Student’s t-test was used for continuous variables.

As expected, there was a high correlation between intake of coffee and caffeine; pairwise Pearson’s partial correlation coefficient, adjusted for sex, was 0.87 (p<0.0001). Comparatively, the pairwise correlation coefficient between caffeine and tea was low at 0.36. Table 2 shows the relationship between coffee intake and risk of hip fracture for all participants, and in men and women separately. Compared with drinking less than one cup of coffee per week, a significantly higher risk of hip fracture was associated with drinking ≥ 4 cups per day by 32% (HR: 1.32; 95% CI: 1.07–1.63) in the whole cohort, by 46% (HR: 1.46; 95% CI: 1.01–2.10) in men and by 33% (HR: 1.33; 95% CI: 1.02–1.72) in women. Among women, particularly in postmenopausal women, there was a suggestive association of a 12% lower risk among those who drank coffee 2–3 cups/day (HR: 0.88; 95% CI: 0.76, 1.01) compared to those who drank less than 1 cup per week.

Table 2.

Consumption of coffee in relation to risk of hip fracture, The Singapore Chinese Health Study, 1993–2014

	Less than weekly	<1 cup3/day	1 cup/day	2–3 cups/day	≥ 4 cups/day
All participants (n=63,154)
Hip fracture cases	532	220	961	677	112
Person-years	213,116	96,382	379,977	322,234	41,157
HR1 (95% CI2)	Reference	1.03 (0.88, 1.20)	1.01 (0.90, 1.12)	0.95 (0.85, 1.07)	1.32 (1.07, 1.63)
Men (n=27,913)
Hip fracture cases	119	70	223	228	43
Person-years	83,500	42,539	136,941	157,740	25,310
HR (95% CI)	Reference	1.29 (0.96, 1.74)	1.21 (0.96, 1.51)	1.14 (0.90, 1.43)	1.46 (1.01, 2.10)
Women (n=35,241)
Hip fracture cases	413	150	738	449	69
Person-years	129,616	53,834	243,036	164,495	15,846
HR (95% CI)	Reference	0.95 (0.79, 1.15)	0.96 (0.84, 1.08)	0.90 (0.79, 1.04)	1.33 (1.02, 1.72)
Postmenopausal women (n=25,522)
Hip fracture cases	396	144	706	422	66
Person-years	89,431	36,274	171,985	116,155	11,410
HR (95% CI)	Reference	0.95 (0.79, 1.15)	0.94 (0.83, 1.07)	0.88 (0.76, 1.01)	1.31 (1.00, 1.71)

¹HRs (95% CI) were adjusted for sex (whole cohort), age at recruitment (years), year at recruitment (1993–1995, 1995–1998), dialect group (Hokkien, Cantonese), and level of education in categories (no formal education, primary school, secondary school or higher), body mass index (kg/m²), smoking status (never smokers, ex-smokers, and current smokers), moderate physical activity (none, 0.5 – <4 hours weekly, ≥4 hours weekly), at least weekly use of vitamins/mineral (yes, no), use of hormone replacement therapy at recruitment (postmenopausal women only; yes, no), total energy intake (kcal), calcium (mg/1,000kcal/day in quartiles), fruit-vegetable-soy dietary pattern (in quintiles), caffeinated tea drinking frequency (<weekly, weekly, daily drinkers), and baseline physician-diagnosed history of diabetes (yes, no) and stroke (yes, no).

²CI, confidence interval.

³1 cup was equivalent to 237 mL

The relationship between tea drinking and hip fracture risk was generally null (Table 3). We did not find associations with hip fracture risk regardless of the type of tea in either men or women. We also did not find significant association between soda intake and risk of hip fracture in either gender (data not shown).

Table 3.

Consumption of tea in relation to risk of hip fracture, The Singapore Chinese Health Study, 1993–2014

	Any tea			Black tea			Green tea
	Less than weekly	Weekly	Daily	Less than weekly	Weekly	Daily	Less than weekly	Weekly	Daily
All participants (n=63,154)
Cases	1,503	545	454	1,934	369	199	1,866	368	268
Person-yrs	559,307	258,394	235,164	752,326	181,736	118,804	742,174	181,584	129,108
HR1 (95% CI)	Reference	1.00 (0.90, 1.10)	0.95 (0.85, 1.06)	Reference	1.10 (0.98, 1.24)	0.97 (0.83, 1.13)	Reference	0.95 (0.85, 1.07)	0.91 (0.80, 1.04)
Men (n=27,913)
Cases	310	175	198	454	133	96	456	115	112
Person-yrs	194,036	120,114	131,880	282,024	92,465	71,542	291,467	86,257	68,306
HR (95% CI)	Reference	1.01 (0.84, 1.22)	1.01 (0.84, 1.22)	Reference	1.08 (0.88, 1.31)	1.02 (0.81, 1.28)	Reference	0.92 (0.74, 1.13)	0.98 (0.79, 1.21)
Women (n=35,241)
Cases	1,193	370	256	1,480	236	103	1410	253	156
Person-yrs	365,271	138,280	103,284	470,302	89,270	47,263	450,707	95,326	60,802
HR (95% CI)	Reference	1.00 (0.89, 1.13)	0.93 (0.81, 1.06)	Reference	1.12 (0.97, 1. 29)	0.95 (0.77, 1.16)	Reference	0.97 (0.84, 1.11)	0.89 (0.75, 1.05)
Postmenopausalwomen (n=25,522)
Cases	1,136	359	239	1,411	230	93	1,340	243	151
Person-yrs	263,461	91,759	70,035	337,658	57,100	30,497	319,227	63,309	42,719
HR (95% CI)	Reference	1.03 (0.91, 1.17)	0.92 (0.80, 1.06)	Reference	1.17 (1.02, 1.35)	0.92 (0.74, 1.13)	Reference	0.98 (0.85, 1.13)	0.91 (0.76, 1.08)

¹HRs (95% CI) were adjusted for sex (whole cohort), age at recruitment (years), year at recruitment (1993–1995, 1995–1998), dialect group (Hokkien, Cantonese), and level of education in categories (no formal education, primary school, secondary school or higher), body mass index (kg/m²), smoking status (never smokers, ex-smokers, and current smokers), moderate physical activity (none, 0.5– <4 hours weekly, ≥4 hours weekly), at least weekly use of vitamins/mineral (yes, no), use of hormone replacement therapy at recruitment (postmenopausal women only; yes, no), total energy intake (kcal), calcium (mg/1,000kcal/day in quartiles), fruit-vegetable-soy dietary pattern (in quintiles), caffeinated coffee drinking frequency (Less than weekly, <1 cup/day, 1 cup/day, 2–3 cups/day, and ≥4 cups/day), and baseline physician-diagnosed history of diabetes (yes, no) and stroke (yes, no).

²CI, confidence interval.

Table 4 shows the association with dietary caffeine intake. The median caffeine consumption in this cohort was 114 mg per day. We did not observe any significant association in all participants or in men. Overall women with daily consumption of ≥300 mg caffeine per were at statistically significant 25% higher risk of hip fracture compared to those with <100 mg caffeine. Among postmenopausal women, the risk estimate for hip fracture was the lowest for daily consumption of 200–<300 mg caffeine (HR 0.89; 95% CI: 0.79, 1.01) and the highest for daily caffeine of ≥300 mg (HR: 1.24; 95% CI: 1.0–1.55). Additional adjustment for dietary vitamin D or calcium did not change the results for the hip fracture risk association with coffee, tea or caffeine (data not shown).

Table 4.

Consumption of caffeine in relation to risk of hip fracture, The Singapore Chinese Health Study, 1993–2014.

	<100 mg/day	100–<200 mg/day	200–<300 mg/day	>300 mg/day
All participants (n=63,154)
Hip fracture cases	1,169	565	584	184
Person-years	442,171	254,669	277,466	78,560
HR1 (95% CI3)	Reference	1.00 (0.91, 1.11)	0.94 (0.85, 1.04)	1.19 (1.01, 1.39)
HR2 (95% CI)	Reference	1.03 (0.93, 1.15)	0.93 (0.84, 1.03)	1.20 (1.01, 1.41)
Men (n=27,913)
Hip fracture cases	234	164	204	81
Person-years	144,312	113,495	138,730	49,494
HR1 (95% CI)	Reference	0.97 (0.80, 1.19)	1.00 (0.83, 1.21)	1.19 (0.92, 1.53)
HR2 (95% CI)	Reference	0.97 (0.79, 1.20)	0.97 (0.80, 1.18)	1.14 (0.87, 1.49)
Women (n=35,241)
Hip fracture cases	935	401	380	103
Person-years	297,859	141,174	138,736	29,066
HR1 (95% CI)	Reference	1.02 (0.91, 1.15)	0.91 (0.81, 1.02)	1.19 (0.97, 1.46)
HR2 (95% CI)	Reference	1.07 (0.94, 1.21)	0.92 (0.82, 1.04)	1.25 (1.01, 1.55)
Postmenopausal women (n=25,522)
Hip fracture cases	898	384	353	99
Person-years	210,058	96,095	98,078	21,024
HR1 (95% CI)	Reference	1.02 (0.91, 1.15)	0.88 (0.78, 1.00)	1.19 (0.96, 1.46)
HR2 (95% CI)	Reference	1.07 (0.94, 1.22)	0.89 (0.79, 1.01)	1.24 (1.00, 1.55)

¹HRs (95% CI) were adjusted for gender (whole cohort), age at recruitment (years), year of recruitment (1993–1995, 1995–1998), dialect group (Hokkien, Cantonese), and level of education in categories (no formal education, primary school, secondary school or higher);

²HRs (95% CI) were further adjusted for body mass index (kg/m²), smoking status (never smokers, ex-smokers, and current smokers), physical activity (none, 0.5– <4 hours weekly, ≥4 hours weekly), at least weekly use of vitamins/mineral (yes, no), total energy intake (kcal), calcium (mg/1,000kcal/day in quartiles), fruit-vegetable-soy dietary pattern (in quintiles), use of hormone replacement therapy at recruitment (postmenopausal women only; yes, no), and baseline physician-diagnosed history of diabetes (yes, no) and stroke (yes, no).

³CI, confidence interval.

Furthermore, we performed restricted cubic spline regression analyses and found a potential non-linear relationship between daily intake of coffee and risk of hip fracture among all participants (p for non-linearity=0.03) and among post-menopausal women (p for non-linearity=0.03), but not among men (p for non-linearity=0.32). Similarly, for the association of caffeine with hip fracture risk, a possible non-linear relation was found among post-menopausal women (p for non-linearity=0.05) but not among men (p for non-linearity=0.15). Figure 1 depicts possible J-shaped risk patterns between consumption of coffee/caffeine and risk of incident hip fracture in postmenopausal women, with the lowest risk at about 2 cups/day of coffee or 220 mg/day of caffeine, and with risk estimates increasing at higher consumption thereafter. Of note, 95% of the women in this cohort who developed hip fracture were postmenopausal at recruitment.

Figure 1.

Association between daily coffee or caffeine intake and risk of hip fracture among postmenopausal women using restricted cubic spline regression (n=25,522)

Hazard ratios and 95% confidence intervals were adjusted for age at recruitment (years), year of recruitment (1993–1995, 1995–1998), dialect group (Hokkien, Cantonese), and level of education in categories (no formal education, primary school, secondary school or higher), body mass index (kg/m²), smoking status (never smokers, ex-smokers, and current smokers), moderate physical activity (none, 0.5 – <4 hours weekly, ≥4 hours weekly), at least weekly use of vitamins/mineral (yes, no), use of hormone replacement therapy at recruitment (postmenopausal women only; yes, no), total energy intake (kcal), calcium (mg/1,000kcal/day in quartiles), fruit-vegetable-soy dietary pattern (in quintiles), and caffeinated tea drinking frequency (for the association of coffee only; <weekly, weekly, daily drinkers), and baseline physician-diagnosed history of diabetes (yes, no) and stroke (yes, no).

In sensitivity analysis, all results remained the same after we excluded those who had reported extreme energy intakes.

Discussion

In this large prospective cohort of middle-aged to elderly Singapore Chinese, compared to those who drank less than 1 cup of coffee per week, we found that consumption of 4 or more cups of coffee per day was associated with a higher risk of hip fracture in both men and women. A possible J-shaped non-linear relationship between coffee or caffeine intake and risk of hip fracture may exist among postmenopausal women, with the lowest risk at drinking 2–3 cups of coffee per day or caffeine intake of 200–<300 mg per day. However, we did not find any association between tea drinking and risk of hip fracture in all cohort participants or in either men or women.

The available epidemiologic evidence on coffee or caffeine intake in relation to hip fracture is inconclusive. Some studies suggested that high coffee or caffeine intake was associated with increased risk of hip fracture ^(19,22–24), while others reported favorable relations ⁽²⁵⁾ or null associations ^(26–28). One reason could be that the different cut-offs were defined for the referent and comparator categories for coffee consumption, which may contribute to the heterogeneous results. For example, in one study, null associations have been reported for all coffee intake categories (1–4 cups/month, 2–7 cups/week, 2–5 cups/day and ≥6 cups/day) relative to women who drank coffee <1 cup/month (p for trend=0.49) ⁽²⁶⁾. If there are different effects between 2–3 cups/day and 4–5 cups/day on hip fractures, such a categorization may not be apparent to discern such differences. Additionally, a recent systematic review of previous observations concluded that there was no significant increased risk of fracture with an intake of 400 mg caffeine per day in healthy adults, particularly those with adequate calcium intake ⁽¹⁷⁾. This may also explain the lack of significant associations from current studies in the literature, which were primarily conducted in Western populations.

In the present study, the cut-off values for coffee or caffeine were based on the distribution of coffee consumption in the study population. Our results are consistent with the previous findings in experimental studies which have suggested bone favorable effects at moderate intake of caffeine and detrimental impacts at higher intake. The possible J-shaped relationship between coffee/caffeine intake and risk of hip fracture among postmenopausal women but not in men implied that the beneficial effect of a moderate dose of coffee or caffeine may be observed only in a state of estrogen deficiency on bone health. At the highest level of coffee or caffeine consumption, we found a consistent result of an increased hip fracture risk in both sexes, suggesting that high dosages of caffeine and possible other non-caffeinated components may have adverse effects on bone health ^(3,10,29). It is possible that, as suggested in previous mechanistic experimental studies, low to moderate dose of caffeine may inhibit bone resorption ⁽³⁰⁾ and increase bone formation ⁽³¹⁾ due to the binding to adenosine receptor isoforms, while a high dosage of caffeine has been found to increase calcium excretion ⁽³²⁾ and possibly affect vitamin D receptor protein expression ⁽²⁹⁾. The higher incidence of hip fracture rate among women compared to men is a global phenomenon, and this difference has been attributed to sex-specific differences in genetic influences on the skeleton, bone geometry and bone gonadal hormone response for osteoporosis ⁽³³⁾. We believe that our observation of a possible differential effect of coffee/caffeine by dosage on hip fracture risk among postmenopausal women, if replicated in future studies, is interesting, and illustrates potential differences in the effects of coffee/caffeine consumption on the pathogenesis of osteoporosis between men and women. Nevertheless, we acknowledge that our results could still occur by chance, especially since the p-values of these associations are of borderline significance, and these results need to be corroborated by future studies. Hence, we have also examined the current literature for experimental evidence to support the biological plausibility of mechanisms underlying the differential effects of moderate and excessive doses of caffeine in animal studies to support our observations.

In the Singapore Chinese cohort, the dietary intake of calcium and vitamin D had previously been noted to be lower relative to that in other Western populations ^(34,35). The increased risk of hip fracture in consumption of at least 4 cups of coffee per day could be attributed to the lower calcium intake in this population, which is in contrast with the lack of significantly increased fracture risk with 400 mg caffeine per day in healthy adults with adequate calcium intake ⁽¹⁷⁾. Nonetheless, adjustment for dietary calcium, overall diet pattern or vitamin D did not change the association between intake of coffee/caffeine and hip fracture risk in our study.

The strengths of this study include the large number of incident hip fracture cases identified from a population-based prospective cohort with a long follow-up time. Our case ascertainment through linkage with the comprehensive, nationwide hospital database can be considered complete. Singapore is a small city-state with a health system for easy access to specialized medical care, and practically all hip fracture cases will seek medical attention immediately and be hospitalized. Finally, we included other established and possible risk factors for hip fracture from our previous study ⁽³⁶⁾ as covariates in our regression-based risk models to minimize the likelihood of spurious associations resulting from insufficient control of confounding.

Inevitably, our study has limitations. One limitation is a possible misclassification of dietary intake of tea or coffee which was recorded from the food frequency questionnaire and only at baseline. However, such misclassifications are likely to be non-differential and thus resulting in an underestimation of the true relative risk. Although individuals may have changed their dietary habit during the follow up of the study, nevertheless, in a follow-up interview that was conducted on 39,258 surviving cohort participants, average 12.7 years after the baseline interview, the majority of them had retained their status as daily or non-daily drinker of coffee (72.3%), black tea (85.6%) and green tea (85.2%). Another limitation is a lack of information on decaffeinated coffee or tea, which precludes our ability to test our speculation on the possible harmful effects from non-caffeine compounds at high dosage of coffee consumption. The lack of measurements on bone mineral density at recruitment precluded us from ascertaining the state of bone health in participants at baseline. Data on history of falls and use of glucocorticoids were not available in this study. Nonetheless, our results should be interpreted cautiously, as we cannot rule out other unmeasured factors or residual confounding in an observational study.

In conclusion, our observations suggest that a high intake of coffee could increase hip fracture risk in both men and women, and that a moderate intake of coffee/caffeine could be associated with a lower risk in postmenopausal women. Future studies in other Asian populations are needed to verify these results, and to determine the optimal amount of coffee/caffeine intake in relation to bone health in men and women.

Highlights.

• Drinking coffee ≥4 cups/d may increase risk of hip fracture in men and women.

• Drinking coffee 2–3 cups/d may lower risk of hip fracture in postmenopausal women.

• Possible J-shaped relation of caffeine to fracture risk in postmenopausal women.