In this cohort the consumption of coffee was high. We observed a decrease in BMD of the proximal femur in men consuming 4 cups of coffee or more daily. In high consumers of coffee, rapid metabolizers had lower BMD values than slow metabolizers of caffeine. A potential risk group more prone to develop osteoporosis might, thus, have been identified.
The observed decrease in BMD in male high consumers of coffee could be estimated to correspond to an approximately 30% increased risk of hip fracture, which would imply a considerable increase in view of public health . This increased risk might have impact on total osteoporotic fracture health economy. This is illustrated by the fact that the number of hip fractures worldwide in the year of 2000 was estimated to 1.6 million . The global cost for hip fractures is rising and by 2050 it has been estimated to be about 132 billion US dollars .
Earlier studies in men [12, 15–21] have not observed any statistically significant relation between consumption of coffee and BMD. It should be noted, however, that some of the studies were small [12, 15, 21]. In most studies [17–21] the exposure was defined as caffeine intake from both coffee and tea. This approach may not be optimal because both beverages contain several other bioactive substances that may modify the effects of caffeine. Furthermore, many of the earlier studies do not clearly state the exposure as amount of coffee or caffeine consumed [15, 17, 18, 20]. When stated, the average intake of coffee/caffeine varied from approximately 200 mg caffeine per day [19, 21] or less than two cups of coffee per day  to 3 cups per day in one study . In our study, compared to most other studies, mean intake of caffeine and consumption of coffee was higher: 367 mg/day and 3.2 cups/day, respectively.
In the majority of the studies of women no relation between consumption of coffee or intake of caffeine and BMD has been detected [6, 7, 13, 14]. Nevertheless, a weak negative relation between coffee or caffeine and BMD has been observed [9, 11, 22, 44–48] but the relation between BMD and intake of coffee/caffeine has been attenuated by adequate intake of milk/calcium [9, 22, 47]. Our results do not support these latter findings but few of our participants had a low calcium intake. In general, the studies in women thus provided limited evidence for the existence of a relation between intake of coffee/caffeine and effects on BMD, which is accordance with our results. It should be noted that as in the studies of men, many of the studies in women were small [9, 49–57]. In addition, no separate analyses of coffee and tea were carried out in the majority of studies of women [8–10, 44, 45, 48–56, 58, 59]. Average intake of coffee or caffeine seems to have been low or modest in some studies [10, 44, 53, 55, 56], i.e., lower than in the present investigation.
There is evidence for females having lower activity of CYP1A2 than men . With a higher CYP1A2 activity in men, caffeine will be more rapidly metabolized and the concentrations of metabolites like paraxanthine will become higher in relation to the concentration of caffeine. The deleterious effect of coffee consumption on bone may be an effect of caffeine metabolites. Consistent with this theory is that we observed lower BMD among rapid compared to slow metabolizers of caffeine with a high coffee consumption. Moreover, we found lower BMD among male high consumers of coffee but not among such women, an observation that may be explained by higher CYP1A2 activity in men . In addition, we did not find any statistical differences between slow and rapid metabolizers with a low consumption of coffee or between slow metabolizers with a high consumption of coffee and rapid metabolizers with a low consumption. Our results may thus indicate that a certain level of metabolites must be reached in order to observe a negative effect on BMD. There are, however, no published data regarding effects of metabolites of caffeine on BMD. Therefore more studies are clearly warranted in order to investigate possible mechanisms of interactions regarding caffeine intake and CYP1A2 genotype in relation to BMD. How caffeine or its metabolites exert effects on bone can theoretically be explained by other mechanisms than by reduced renal calcium conservation. According to some in vitro studies, caffeine may interfere with bone remodelling process. Tsuang et al (2006)  suggested that caffeine may have deleterious effect on the viability of rat osteoblasts, which could enhance the rate of osteoblast apoptosis. In addition, Lu et al (2008)  has demonstrated that cell viability also decreased in human osteoblasts treated with caffeine in a dose-dependent manner mainly due to apoptosis. Zhou et al (2009)  hypothesise, however, that bone marrow-derived mesenchymal stem cells, which are precursor cells of osteoblasts, may be the real target cells of caffeine-induced osteoporosis in vivo. However, it remains to be demonstrated whether a mechanism including direct effects of caffeine or its metabolites on cells involved in the remodelling process could be of importance also in vivo at dosages of relevance to humans.
It has been demonstrated that both the parent compound, caffeine, as well as paraxanthine, might be teratogenic after administration of very high doses in mice with skeletal malformations as a consequence . Caffeine is cleared more quickly than paraxanthine and 8 hours after caffeine intake, plasma concentrations of paraxanthine levels exceed those of caffeine . With long-term exposure of high doses of caffeine there is substantial accumulation of paraxanthine [65, 66]. Paraxanthine has in vitro been found to be a potent suppressor of transforming growth factor beta (TGF-β) , which stimulates bone formation, and TGF-β deficiency may result in osteoporosis . Interestingly, paraxanthine has been found to be the most powerful pharmacological repressor of hepatocellular TGF-β dependent connective tissue growth factor expression among the drug family of methylxanthines, including caffeine . The major caffeine derivatives, including paraxanthine, have common mechanisms of action, i.e. competitive antagonism of the adenosine interaction with A1 and A2 receptors. Deactivation of the adenosine receptors, which are expressed in bone cells, can result in reduced bone formation .
Advantages and limitations of our study
To our knowledge, this is one of few population-based studies investigating possible effects of coffee and tea consumption on BMD in both men and women. In contrast to most other studies, the majority of the participants in our study consumed high amounts of coffee. We had a sufficient number of participants to detect even modest associations. An additional strength is that we did not focus on caffeine intake but on the exposure of coffee and tea separately. This distinction may be important because some studies have indicated that consumption of tea could have a positive influence on BMD, which could counteract the negative influence of coffee. Tea consumption in our study was low and adjusted for in the statistical analyses. The possible modification by genotype for CYP1A2 inducibility has not previously been investigated. We also had the possibility to consider several conceivable covariates in the analysis, including nutrients, physical activity behavior and smoking.
This study nevertheless has several potential limitations. In this study we have measured BMD in the proximal femur only. We refrained from including BMD measurements of the spine since spondylosis is common in elderly individuals, and this condition can confound the relative weak association between BMD and coffee as well as the comparison between sexes. As the measurement of BMD was on average performed 2 years after the dietary investigation, the follow-up time was limited. However, the optimal time between measurements of coffee consumption and BMD is currently not known. Nevertheless, it should be noted that earlier studies on skeletal effects by an exposure that affects calcium metabolism indicate a lag period of 2-3 years before a steady state of bone turnover and BMD is reached [70, 71].
Statistically significant differences in BMD between high consumers who were rapid metabolizers and those who were slow metabolizers of caffeine were generally confined to the whole study group of both men and women, probably because statistical power was too low to attain statistical significance in the groups of each gender. There were, however, clear tendencies of a lower BMD in high consuming males who were rapid metabolizers. In women the same pattern could also be observed.
Because the exposure measurement was based on a single dietary measurement, there may be some degree of error in the measurement. The 7-day dietary recording used in the present study has been found to be valid  and a high reliability of self-reported measures of caffeine consumption has previously been shown . Nonetheless, because no direct measurements of the caffeine content in the consumed coffee and tea were performed, we lack data on the actual intake of caffeine. Still, recall errors of the exposure are known to lead to conservatively biased estimates.
Temporal changes in the consumption pattern of coffee and tea by time might be of importance but this was not assessed since we only determined baseline frequencies of consumption. However, in the Swedish Mammography Cohort  it was found that the consumption of coffee among elderly women during 10 years  was relatively constant (Personal communication with Dr SC Larsson, Division of Nutritional Epidemiology, The National Institute of Environmental Medicine, Karolinska Institute, Stockholm, Sweden)