|
|
| |||||||||||||||||||||||||||||||
Background Smoking is recognized as a risk factor for vertebral, forearm, and hip fractures. Since bone density is an important determinant of bone strength, we conducted a study to ascertain whether a deficit in bone density is associated with tobacco use and, if so, to identify the responsible mechanisms.
Methods We conducted a cross-sectional study of bone density at the lumbar spine and the femoral neck and shaft in 41 pairs of female twins (21 monozygotic pairs), 27 to 73 years of age (mean, 49), who were discordant for at least 5 pack-years of smoking (mean, 23; maximum, 64). Bone density was measured by dual-photon absorptiometry. The difference in bone density between the members of a pair was expressed as a percentage of the mean value for the pair.
Results For every 10 pack-years of smoking, the bone density of the twin who smoked more heavily was 2.0 percent lower at the lumbar spine (P = 0.01), 0.9 percent lower at the femoral neck (P = 0.25), and 1.4 percent lower at the femoral shaft (P = 0.04). These results were not confounded by measured lifestyle factors. In the 20 pairs who were discordant by 20 or more pack-years (mean, 35), the (mean ±SE) within-pair differences in bone density at the three sites were 9.3 ±3.1 percent (P = 0.008), 5.8 ±2.9 percent (P = 0.06), and 6.5 ±3.2 percent (P = 0.05), respectively. Smoking was associated with higher serum concentrations of follicle-stimulating hormone (P = 0.02) and luteinizing hormone (P = 0.03) and lower serum concentrations of parathyroid hormone (P = 0.05). Differences in spinal bone density between members of a pair were associated with differences in the serum concentrations of parathyroid hormone (P = 0.01) and calcium (P = 0.05) and urinary pyridinoline excretion (P = 0.06), a marker of bone resorption.
Conclusions Women who smoke one pack of cigarettes each day throughout adulthood will, by the time of menopause, have an average deficit of 5 to 10 percent in bone density, which is sufficient to increase the risk of fracture.
Smoking has been reported to be a risk factor for vertebral, forearm, and hip fractures4,5,6. Whether smoking increases the risk of fracture by reducing the quantity of bone is unknown. Studies have shown either no association or a small negative association between smoking and bone density7,8,9.
A study of twins of the same sex who are discordant for smoking controls for age, sex, and genetic composition, all of which are major determinants of bone density1,10,11. This matching is important, since the age-adjusted bone density of the axial and appendicular skeleton has a correlation of about 0.8 in monozygotic pairs and about half this value in dizygotic pairs11. We undertook a study to determine whether a deficit in the bone density of the lumbar spine, femoral neck, and femoral shaft is associated with tobacco use in twins, and if so, whether the deficit could be explained by specific lifestyle factors. Demonstration of a deficit that is independent of such factors would support the association between smoking and the risk of fracture at these sites.
Methods
Subjects
Between 1980 and 1982, a questionnaire was sent to adult pairs of twins listed in the Australian National Health and Medical Research Council Twin Registry. A total of 3810 pairs of twins returned the questionnaire, which included questions about a history of smoking12. The 98 female pairs who were most discordant for smoking were invited to visit the Austin Hospital for measurements of bone density; the women were not told that the discordance was the reason for the invitation. Forty-one of the 98 pairs participated in the study; one or both members of 23 pairs were unable to attend, one or both members of 19 pairs declined the invitation, and both members of 15 pairs could not be located. Zygosity was determined by responses to questions that have been shown to be accurate in determining zygosity 95 percent of the time13. Women were excluded from the study if they had any problem that could result in bone loss, such as renal, hepatic, or gastrointestinal disease, primary hyperparathyroidism, Paget's disease, thyroid disease, or osteomalacia, or if they had used corticosteroids, anticonvulsant drugs, antacids, diuretics, fluoride, or vitamin D. The study was approved by the Ethics Committee on Research and the Board of Medical Research of the Royal Melbourne Hospital, and informed consent was given by all 82 women.
Measurement of Bone Density
Bone density was measured at the lumbar spine (second to fourth lumbar vertebrae) and at the neck and mid-shaft of the femur by dual-photon absorptiometry (Novo BMC Lab 22A absorptiometer; Novo Industries, Bagsvaerd, Denmark)14. Bone density (in grams per square centimeter) was calculated by dividing the measured bone-mineral content by the projected area of the region scanned. The coefficient of variation for two to six measurements at each site was 2 percent at the lumbar spine and 3 percent at the two femoral sites in 15 normal young women with no bone disease.
Measurement of Tobacco Use and Other Variables
A questionnaire administered by an interviewer was used to obtain information about age; height; weight; consumption of alcohol, tea, and coffee; use of tobacco; intake of calcium from dairy products; menopausal, reproductive, and fracture history; and use of oral contraceptives or estrogen-replacement therapy. Physical activity (sitting, standing, walking, and lifting loads) associated with the respondent's usual occupation (including housework) was rated on a scale of 1 to 5 (never, seldom, sometimes, often, or always performed).
Lifetime tobacco use was calculated as the total number of years of smoking multiplied by the average number of cigarettes smoked per day, divided by 20, and expressed as pack-years of smoking. The same measure used in the 1980-1982 survey confirmed the relative smoking status of the members of each pair of twins. There was no evidence from urinary cotinine measurements to cast doubt on the claims of twins who said they did not smoke or had given up smoking. The twin who had never smoked or had fewer pack-years of smoking was considered to be the lighter smoker of the pair, and the twin who smoked or had more pack-years of smoking was considered to be the heavier smoker. In all but eight pairs, the lighter smoker had smoked for less than one pack-year. All pairs were discordant by at least five pack-years.
Serum alkaline phosphatase, albumin, calcium, phosphate, and creatinine concentrations and urinary creatinine and calcium concentrations were measured with standard AutoAnalyzer techniques. Serum osteocalcin was measured by radioimmunoassay with the use of antibodies raised in rabbits against bovine osteocalcin15. Serum dehydroepiandrosterone sulfate, testosterone, and 1,25-dihydroxyvitamin D3 were measured by radioimmunoassay. Serum concentrations of sex hormone-binding globulin, follicle-stimulating hormone, luteinizing hormone, prolactin, and parathyroid hormone were measured by immunoradiometric assay. Urinary hydroxyproline excretion was measured by the method of Nobbs et al.16. Twenty-four-hour urinary estrogen excretion was measured as previously described17. Total pyridinoline and deoxypyridinoline were measured in aliquots of urine fractionated on a cellulose column by a modification of the method described by Eyre et al.18. All samples from each twin pair were obtained on the same day after an overnight fast and were analyzed in the same assay.
Statistical Analysis
Let Yi be the bone density of twin i, i = 1,2, and Xli, . . . ,Xqi her covariates, with X1 corresponding to tobacco use. Within each pair, twin 1 was the heavier smoker. If Yi = a0 + aiX1i + . . . aqXqi + Ei, where Ei represents the measurement error and effects specific to a twin, the coefficient a1 represents a linear association with smoking (X1i = pack-years of smoking) or an association with (greater) tobacco use (X11 = 1 and X12 = 0). The difference in bone density between members of a pair was calculated as D = Y1 - Y2 = a1D1 + . . . + aqDq + E, where Dj = Xji - Xj2 and E = E1 - E2. D is a function of the same coefficients a1, . . . ,aq, takes into account the matching, and is independent of the mean bone density of the pair, YM = 1/2(Y1+Y2). The pair difference is expressed as D' = 100D/YM, a percentage of the pair mean, to adjust for age-related and genetic factors that contribute to the variation in YM.
D and D' were regressed through the origin against D1, . . . ,Dq by a least-squares analysis and by a least-absolute-deviations regression with the use of an iteratively reweighted least-squares analysis19. The proportion of pairs in which the heavier smoker had lower bone density was calculated by logistic regression, with the null hypothesis (proportion = 0.5) tested by the likelihood-ratio criterion. Analyses were performed with GLIM20 software and repeated for the most discordant pairs, since the lack of a deficit in bone density in these pairs would argue strongly against a true association. All P values are two-sided.
Results
The mean (±SD) age of the 41 pairs of twins (21 monozygotic and 20 dizygotic) was 49 ±12 years (range, 27 to 73). The mean difference in lifetime tobacco use between members of a pair was 23 ±15 pack-years (range, 5 to 64). For the 20 most discordant pairs (10 monozygotic and 10 dizygotic), the mean age was 51 ±9 years (range, 36 to 73), and the mean difference in lifetime tobacco use was 35 ±15 pack-years (range, 20 to 64); in all but 1 pair, the twin who was the lighter smoker had never smoked regularly.
For the 20 most discordant pairs, the mean bone density (expressed as a percentage of the mean value for the pair) in the twin who was the heavier smoker was lower by a mean (±SE) of 9.3 ±3.1 percent (P = 0.008), 5.8 ±2.9 percent (P = 0.06), and 6.5 ±3.2 percent (P = 0.05) at the lumbar spine, femoral neck, and femoral shaft, respectively (Table 1). The proportion of pairs in which the heavier smoker had lower bone density at the spine, femoral neck, and femoral shaft was 0.85 (P = 0.001), 0.75 (P = 0.02), and 0.65 (P = 0.18), respectively.
|
|
|
Analyses of biochemical measures were conducted in the 26 pairs of twins who were concordant for menopausal status, use of estrogen-replacement therapy, and use of an oral contraceptive. In the 11 pairs who were discordant by 20 or more pack-years, the twin who smoked more heavily had almost 50 percent higher serum concentrations of follicle-stimulating hormone (P = 0.02) and luteinizing hormone (P = 0.03) (Figure 2 and Table 3). For each 10 pack-years of smoking, the difference in the serum parathyroid hormone concentration decreased by 5.0 ±2.5 percent of the mean concentration for the pair (P = 0.05) (Table 3 and Figure 2C). In 17 of the 26 pairs, the serum calcium concentration was higher in the twin who smoked (P = 0.08), although the linear association between differences in serum calcium concentrations and pack-years of smoking was only 0.2 ±0.4 percent per 10 pack-years (P = 0.70) (Figure 2D). For the 11 most discordant pairs, serum phosphate and alkaline phosphatase concentrations were higher in the twin who smoked (P = 0.05 and P = 0.006, respectively) (Table 3).
|
|
The relations between deficits in bone density at all three sites and differences in the number of pack-years of smoking persisted when adjustments were made for biochemical factors. Adjustment for serum follicle-stimulating hormone, luteinizing hormone, phosphate, and alkaline phosphatase had little influence on the association between bone density and smoking. The difference in the serum parathyroid hormone level was negatively associated with the difference in bone density at the lumbar spine (r = -0.60, P = 0.005), femoral neck (r = -0.75, P<0.001), and femoral shaft (r = -0.53, P = 0.01). When adjustments were made for the number of pack-years of smoking and the serum parathyroid hormone concentration, which together were associated with 70 percent of the variation in bone-density differences at the lumbar spine, the deficit in spinal bone density increased with the difference in serum calcium (P = 0.05) and, in the 17 pairs in whom it was measured, with the difference in urinary pyridinoline (P = 0.06).
Discussion
This cross-sectional study provides compelling evidence of an association between smoking and reduced bone density in women. For every 10 pack-years of smoking, the deficit in bone density in the twin who smoked or who smoked more than her twin increased by about 2 percent at the lumbar spine and about 1 percent at the femoral sites. Support for a dose-response relation comes from pairs who were discordant for more than 20 pack-years of smoking (with an average difference of 35 pack-years), in which the twin who smoked more heavily had a deficit of 5 to 10 percent of the mean value for the pair.
Smokers can differ from nonsmokers in weight, caffeine intake, age at menopause, and alcohol consumption,4,5,21,22 all of which can potentially confound an association between smoking and bone density. In our study the association between bone density and smoking was not appreciably modified by these covariates. However, there could have been some uncontrolled confounding, and any errors in the measurement of covariates would weaken the statistical power to detect such confounding.
What are the mechanisms responsible for bone loss associated with smoking? Although there was no evidence of reduced bone formation, since differences in serum osteocalcin concentrations were independent of differences in the number of pack-years of smoking, there was evidence of increased bone resorption. Smoking was positively associated with serum concentrations of follicle-stimulating hormone and luteinizing hormone and negatively associated with the serum concentration of parathyroid hormone. For the lumbar spine, the lower bone density in the twin who smoked more heavily was associated with higher levels of serum calcium and urinary pyridinoline excretion. Differences in serum calcium concentrations were positively associated with differences in markers of bone resorption (urinary hydroxyproline and urinary pyridinoline). These findings support the hypothesis that tobacco use causes a reduction in levels of circulating estrogens, as indicated by increased serum concentrations of follicle-stimulating hormone and luteinizing hormone, and leads to increased bone resorption. The latter results in an increase in the serum calcium concentration, a subsequent reduction in the serum parathyroid hormone concentration, and an increase in urinary hydroxyproline and pyridinoline excretion. There is independent support for an effect of smoking on the production and degradation of estrogens23,24,25. The fact that we found no direct association between measured estrogen levels and tobacco use may have been due to large day-to-day variations in urinary estrogen excretion.
The deficits in bone density that we observed are probably biologically important because of the anticipated increase in bone fragility. In vitro, bone strength decreases threefold with a 10 percent decrease in mineral content2. A 10 percent difference in bone density is equivalent to almost 1 SD for bone density in the general population,1,26 10 years of age-related bone loss,3 and about half the diminution in bone density in postmenopausal women as compared with premenopausal women27. A 10 percent decrease in bone density over a period of 10 years may confer a 44 percent increase in the risk of a hip fracture28.
An association between smoking and the risk of osteoporosis that is not explained by confounding factors has major consequences for public health. In the United States one of every four women is a regular smoker, and the proportion of women who smoke more than 25 cigarettes per day has doubled during the past 20 years. Eighty percent of middle-aged female smokers reportedly started smoking before the age of 2029. In women who smoke an average of one pack per day, a deficit in spinal bone density of 2 percent per decade will result in bone density that is more than half a standard deviation lower than in nonsmokers by the time of menopause. As the population ages, the total number of pack-years of smoking in postmenopausal women will increase. This study suggests that by the early 21st century there will be an increase in fractures attributable to smoking. Recognition of these consequences for osteoporosis may help dissuade women from taking up, or continuing to use, tobacco.
Supported by grants from the Australian National Health and Medical Research Council, the Anti-Cancer Council of Victoria, the Austin Hospital Medical Research Foundation, and the Shepherd Foundation.
We are indebted to the twins who participated in this study and to Professor J. Brown, Professor P. Delmas, Dr. P.R. Ebeling, and Dr. G. Jerums for assistance with the biochemical analyses. The twins were identified through the Australian National Health and Medical Research Council Twin Registry with the permission of Dr. N.G. Martin and Professor J.D. Mathews.
Source Information
From the Faculty of Medicine Epidemiology Unit, University of Melbourne, Carlton (J.L.H.), and the Endocrine Unit, Department of Medicine, Austin Hospital, University of Melbourne, Heidelberg (E.S.) -- both in Australia.
Address reprint requests to Dr. Seeman at the Endocrine Unit, Austin Hospital, Heidelberg, Victoria 3084, Australia.
References
in nonpregnancy urine by spectrophotometry or fluorimetry. In: Breuer H, Hamel D, Kruskemper HL, eds. Methods of hormone analysis. Stuttgart, Germany: Georg Thieme Verlag, 1976:446-62.
| |||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
