Premenopausal Smoking and Bone Density in 2015 Perimenopausal Women



The importance of cigarette smoking in relation to bone mass remains uncertain, especially in younger women. In a recent meta-analysis including 10 studies in premenopausal women no effect was seen in this age group. We used baseline data from a large national cohort study (Danish Osteoporosis Prevention Study [DOPS]) to study the cumulated effect of pre- and perimenopausal smoking on bone mineral density (BMD) measured shortly after the cessation of cyclic bleedings. Baseline observations on 2015 recently menopausal women were available. Eight hundred thirty-two women were current smokers and 285 were exsmokers. Significant negative associations of cigarette smoking coded as current, ex-, or never smoking were seen on bone mass in the lumbar spine (P = 0.012), femoral neck (P < 0.001), and total body (P < 0.001). Quantitatively, the differences between current smokers and never smokers were limited to 1.6, 2.9, and 1.9%, respectively. A statistical interaction was found between smoking and fat mass, indicating that women in the highest tertile of fat mass were unaffected by cigarette smoking. Serum vitamin D levels and osteocalcin were inversely related to the number of cigarettes smoked per day (r = 0.11 and P < 0.001; r = 0.17 and P = 0.04), respectively. Bone alkaline phosphatase (BALP) and urinary hydroxyproline (U-OHP) were unaffected by current smoking. The average cumulated effect of premenopausal smoking on bone is small but biologically significant. Reduced body mass in smokers explains part of the negative effect on the skeleton and a complex interaction between smoking and fat mass on the skeleton is indicated. Serum levels of 25-hydroxyvitamin D (25-OHD) and osteocalcin are lower in smokers, which may effect rate of bone loss.


Reports on the association between cigarette smoking and osteoporosis in women remain conflicting even though several large-scale cohort studies have been performed.(1–6) The variability in study results may relate to differences in study population (number, age, menopausal status, and socioeconomic condition) as well as applied measuring technique, skeletal region of interest, and study endpoint.

A few studies have shown a negative effect of cigarette smoking in young women but most studies conducted in pre- and early postmenopausal women fail to show any association between cigarette smoking and osteopenia or fractures.(3,7,8,12–14) However, several studies have disclosed an association between present and past smoking and low bone mass and fractures in postmenopausal women.(2,4,5,15) A recent meta-analysis including 29 cross-sectional studies and 19 cohort or case-control studies confirmed that smoking has no major effect on premenopausal bone density.(16) The postmenopausal bone loss was 0.2% greater per year in smokers compared with nonsmokers and the risk of hip fractures was 17% higher at the age of 60 years, increasing to a 108% higher risk at the age of 90 years. The apparent premenopausal resistance to the deleterious effect of smoking on bone density may be caused by the estrogen-replete state in combination with a younger skeleton. However, because most studies in premenopausal women are cross-sectional it is not clear whether the effect of smoking before menopause is truly nonexisting.(3,10–12) The time from when peak bone is obtained to the time of menopause is short and one explanation would be that smoking-induced cumulative bone losses during this short period are diminutive and therefore difficult to detect in cross-sectional studies.

Table Table 1.. Criteria for Study Inclusion and Exclusion in the DOPS
Inclusion criteriaExclusion criteria
Intact uterus
 Age 45–58 yearsCurrent estrogen use
 3–24 months past last menstrual bleedingEver treated with glucocorticoids > 6 months
 Perimenopausal symptoms (including menstrual irregularities) and elevated serum follicle-stimulating hormoneChronical disease if newly diagnosed or out of control
 Current or past malignancy
 HysterectomizedCurrent or past alcohol or drug addiction
 Age 45–52 years 
 Elevated serum follicle-stimulating hormone 

Suggested mechanisms by which cigarette smoking exerts it's negative effect on bone are through lowered body mass, through impaired estrogen status, through impaired estrogen action on bone, or through a direct effect on bone cells, but knowledge on these issues is sparse, mainly because of the (obvious) lack of experimental studies in humans.(17)

We used baseline data from a large national cohort study on osteoporosis prevention (Danish Osteoporosis Prevention Study [DOPS]) to investigate the quantitative impact of cigarette smoking on whole body and regional bone mineral density (BMD) in perimenopausal women and to explore potential pathophysiological mechanisms.


Study population

DOPS is a combined observational and intervention study comprising 2015 recently menopausal (within 2 years past last menopause) women in four centers, aged 45–58 years, randomly selected from population lists in four areas: three city areas surrounding the three university clinics and one more suburban and rural district. Criteria for study participation were intended to be as broad as possible, but because of the intervention part of the study, it was necessary to exclude women with, for example, severe hypertension or malregulated diabetes, who should not receive hormone replacement therapy until the condition was well-controlled. The criteria are given in Table 1. Recent menopause was an important criterion for study inclusion and because we wanted women undergoing natural menopause as well as hysterectomized women to participate, separate criteria were made to define recent menopause in these two groups of women. Hysterectomized women were considered recently menopausal if they had postmenopausal serum levels of follicle-stimulating hormone and were under the age 52 years.

The participants were recruited by a mailed questionnaire focused on menopause, health, and the wish to participate in a 20-year follow-up study. The general respondence rate was 60%. Details on the longitudinal study are given elsewhere.(18) The study was approved by the national and four local ethical committees. All participants gave informed consent.

Smoking status and covariates

Information on smoking habits, alcohol consumption, coffee and tea consumption, education, and employment was obtained during a structured interview at study entry. The population was grouped into current smokers (n = 832), exsmokers (n = 285), and never smokers (n = 898). Previous and current smoking was quantitated by recording time for start and, when appropriate, cessation or change in tobacco consumption as well as numbers of cigarettes smoked per day during each period. From these records lifetime consumption (in pack-years), age at start of smoking, and time since cessation were calculated. Alcohol consumption was dichotomized into present daily alcohol consumption, yes or no, and high coffee intake as more than 8 cups per day, yes or no, and more than 1 cup per day, yes or no. Current daily intake of energy and calcium was assessed from a 7-day food record and calculations were performed using the Dankost Database (Danish Catering Service, Copenhagen, Denmark) containing Danish tables of food composition.

Baseline BMD and body composition

Total body and regional bone mineral content (BMC) and BMD were obtained using carefully cross-calibrated QDR 1000/W and 2000 densitometers (Hologic, Inc., Waltham, MA, U.S.A.) in all four centers.(19) The lumbar spine (L2–L4), hip region (femoral neck), forearm (ultradistal and proximal one-third), and total body were scanned. In our hands the in vivo precision errors (BMD) in normal individuals are 1.5% (spine), 2.1% (femoral neck), 1.9% (ultra-distal radius), 1.3% (proximal radius), and 0.7% (total body). Body composition was measured using the same equipment. Precision errors were 0.7% for lean body mass and 2.2% for fat mass.(20) Long-term stability of the equipment was assessed by daily scans of an anthropometric phantom in each center. Changes were <0.2% per year.

A standardized procedure for positioning and data analysis was established and followed for all scans. Intercenter concordance on data analysis was checked during the inclusion period by circulation of unanalyzed scans from all centers. Intercenter calibration differences were checked by scanning a common anthropometric phantom at study start and thereafter once a year.

Biochemical values

All blood samplings were drawn in the fasting state between 7:30 a.m. and 10:30 a.m. to minimize diurnal variation. Serum 25-hydroxyvitamin D (25-OHD) was measured by a radioimmunoassay (RIA) using rachitic rat kidney cytosol as a binding protein.(21) The intra- and interassay CVs were 9.4% and 13.5%, respectively. Serum osteocalcin (bone γ-carboxyglutamic acid-containing protein [S-BGP]) was measured by an in-house RIA.(22) The intra- and interassay CVs were 5% and 10%, respectively. Total alkaline phosphatase (ALP) activity in serum was measured spectrophotometrically with p-nitrophenylphosphate as substrate according to the standard method recommended by the Scandinavian Committee on Enzymes.(23) The intra-assay CV was 2.5% and the interassay CV was 5%. Serum bone isoenzyme alkaline phosphatase activity (BALP) was determined by lectin precipitation.(24) The intra-assay CV was 8% and the interassay CV was 25%. Urine hydroxyproline (U-OHP) was determined spectrophotometrically with p-dimethylaminobenzaldehyde as substrate, according to the manufacturer's direction (Organon Teknika, Boxtel, the Netherlands). It was measured in the second voided morning urine collected after an overnight fast and expressed per excreted millimoles of creatinin in the same sample.

Statistical analysis

Analyses of covariance (ANACOVA) was employed to calculate adjusted BMD values. Values were adjusted for age and body weight. Because recent menopause was part of the inclusion criteria, adjustment for age at natural menopause was obtained simply through the age adjustment. Adjustment for coffee and alcohol intake only slightly affected findings while adjustment for time since menopause, education, employment, and fat mass as well as current intake of energy had no effect. Values were also calculated without weight adjustment to account for the possibility that body weight acts as a mediator of the negative effect of cigarette smoking on BMD as well as a confounder. All values were adjusted to overall cohort means.

The nonlinear difference between the three groups (current, exmoker, or nonsmoker) was tested by analysis of variance (ANOVA). Furthermore, the linear trend between the groups was tested by linear regression analysis. Interactions among independent variables were tested using multiple regression analyses with one dummy variable.(25) Differences between group means were tested using Student's t-test for unpaired data.

Two-tailed P-values are given and statistical significance was defined as P < 0.05, except for analyses of interaction where P < 0.10 was set as level of significance. No correction was done for multiple testing, rather detailed P-values were calculated and whenever appropriate biological meaningful dose-response relations were explored, using linear regression analyses, where β gives the nonstandardized regression coefficient. All calculations were done using the SPSS version 6.01 (SPSS Inc., IL, Chicago).


Eight hundred thirty-two women were current smokers, 285 were exsmokers, and 898 were never smokers. Four current smokers started smoking after menopause and 10 exsmokers quit after menopause. The median time since menopause was 0.5 years, with no difference between smoking groups. Table 2 compares additional baseline characteristics of these groups.

Current smokers were mean 0.8 year younger than never smokers. According to the applied inclusion criteria based on menopausal status, this could be explained by the younger age at natural menopause (mean 0.9 year), which we found in this group. The group of exsmokers was older at natural menopause than the current smokers and younger than the never smokers. The prevalence of being hysterectomized was slightly higher among current smokers compared with never smokers. Furthermore, current smokers had a lower body mass and fat mass and a slightly reduced energy intake compared with never smokers.

Smoking status and BMD

Table 3 gives whole body and regional BMD according to smoking status. BMD values are given as age-adjusted only and adjusted for age and weight. Highly significant differences between groups and a clear linear trend toward lower values among current smokers were seen in the total body and the femoral neck (Table 3). However, quantitatively, the differences between current smokers and never smokers were modest (1.2 and 1.9% for age- and weight-adjusted values, respectively). In the lumbar spine the differences between groups (0.9%) were only marginally significant and disappeared when weight adjustments were performed. Even smaller though still significant differences were found between exsmokers and never smokers in the total body and femoral neck (0.3 and 0.7% for age- and weight-adjusted values, respectively; P < 0.01 in both cases).

No significant differences were observed in the proximal radius (0.2%) or the ultradistal radius (0%). Adjustment for coffee and alcohol intake diminished the difference between current smokers and never smokers by one-third in the femoral neck but did not change the differences in spine and total body BMD. Excluding the 14 women who changed their smoking habits the after menopause status did not change any of the findings.

Table Table 2.. Baseline Characteristics of the Study Population by Smoking Status
 Never smokersExsmokersP-value*Current smokersP-value
  1. Quantitative data are given as mean (SD) and two-tailed P-values are given for differences between exsmokers versus never smokers and current smokers versus never smokers in Student's t-test. Frequencies are given as percentages and p-values given for independence of distribution tested by χ2 square analysis.

  2. * Difference between never smokers and exsmokers.

  3. Difference between never smokers and current smokers.

  4. Difference between exsmokers and current smokers (P = 0.06).

  5. § Skilled workers or less educated.

Pack–years (mean range)11.5 (0.01–93.05)23.5 (0.07–93.05)
Smoking duration (years)16.0 (0.01–38.9)30.3 (1.4–50.4)
Time since quit (years)14.0 (0.6–34.8)
Age (years)51.0 (2.9)50.5 (2.9)0.00550.2 (2.7)<0.001
Age at natural menopause (years)50.6 (2.7)50.1 (2.7)0.00949.7 (2.6)<0.001
Weight (kg)69.0 (12.7)69.0 (11.4)NS66.5 (11.4)<0.001
Fat mass (kg)18.2 (8.2)17.8 (6.7)NS16.7 (7.1)<0.001
Energy intake (kJ/day)7672 (1924)7677 (2000)NS7460 (2125)0.028
Percentage coffee drinkers (≥2 cups/day)71.0%73.7%0.37586.1%<0.001
Percentage heavy coffee drinkers17.5%23.5%0.02142.7%<0.001
Percentage daily alcohol drinkers14.5%22.8%0.00321.4%<0.001
Percentage low educated§20.9%24.2%NS33.4%<0.001
Percentage unemployed13.1%11.1%NS18.3%0.003
Table Table 3.. Adjusted BMD Values (g/cm2) by Smoking Status
 Never smokersExsmokersCurrent smokersDifference PLinear trend P
  1. Difference P is the P-value for difference between groups without trend (see text for details). Linear trend P gives P for linear trend in linear regression analysis.

Total body BMD (n = 2005)
 Age adjusted1.113 (0.08)1.110 (0.08)1.092 (0.08)<0.001<0.001
 Age and weight adjusted1.112 (0.07)1.108 (0.07)1.099 (0.07)0.001<0.001
Spine BMD (n = 2009)
 Age adjusted1.033 (0.14)1.033 (0.14)1.017 (0.14)0.0300.012
 Age and weight adjusted1.034 (0.13)1.033 (0.14)1.025 (0.13)0.3940.188
Femoral neck BMD (n = 1998)
 Age adjusted0.810 (0.11)0.804 (0.12)0.787 (0.11)<0.001<0.001
 Age and weight adjusted0.802 (0.11)0.796 (0.11)0.787 (0.10)0.0170.004
Proximal radius (n = 1999)
 Age adjusted0.643 (0.05)0.646 (0.05)0.642 (0.05)0.4490.805
 Age and weight adjusted0.638 (0.05)0.642 (0.05)0.639 (0.05)0.5560.580
Ultradistal radius (n = 1999)
 Age adjusted0.393 (0.05)0.392 (0.05)0.390 (0.05)0.5490.269
 Age and weight adjusted0.394 (0.05)0.392 (0.05)0.394 (0.05)0.8350.847

Among the current smokers we found no difference in BMD between heavy smokers (≥20 cigarettes/day; n = 734) and light smokers (<20 cigarettes/day; n = 98; data not shown), nor did we observe any linear effect on BMD of lifetime consumption in pack-years.

BMD levels in women who had quit smoking more than 5 years ago (n = 216) were not significantly different from BMD levels in never smokers but no linear relationship was found between time since cessation and BMD levels.

Interaction between fat mass, smoking status, and BMD

Splitting the cohort according to tertiles of fat mass (Fig. 1) revealed that age-adjusted neck and total BMD in women in the highest tertile of fat mass (fat mass > 18.98 kg) were unaffected by smoking status. In the highest tertiles of fat mass no significant effect of smoking was seen on BMD (P = 0.08 and P = 0.44); in the intermediate tertile a significant effect was seen in the total body (P = 0.02) but not in the femoral neck (P = 0.10). A highly significant effect (P < 0.0001) was seen in both regions among women in the lowest tertile of fat mass (having a fat mass below 13.32 kg).

Figure FIG. 1..

Total body (A) and femoral neck BMD (B) in current, exsmokers, and nonsmokers stratified in tertiles of fat mass. Women in the lowest tertile of fat mass had total fat mass < 13.32 kg; women in the highest tertile had a fat mass > 18.98 kg. In the highest tertiles of fat mass no significant effect of smoking was seen on BMD (P = 0.08 and P = 0.44), in the intermediate tertile a significant effect was seen in the total body (P = 0.02) but not in the femoral neck (P = 0.10). A highly significant effect was seen in both regions among women in the lowest tertile of fat mass (P < 0.0001).

The interaction between fat mass and smoking status was highly significant in the total body (P = 0.0003) and femoral neck (P = 0.016), but not in the spine (P = 0.329). In the radius no effect of smoking was seen in any stratum.

Current smoking status and 25-OHD levels

Serum 25-OHD levels were significantly lower in current smokers compared with nonsmokers (Table 4). The reduction was only 6.8%, but serum levels correlated inversely to number of cigarettes currently smoked per day (β = −0.16; P = 0.003), suggesting a true biological association. In univariate models serum 25-OHD correlated positively to serum osteocalcin (Table 5) but not to serum BALP or the renal OHP/creatinine excretion rate. However, serum 25-OH vitamin D levels did not substitute for smoking as an explanatory variable for BMD in multivariate models.

Current smoking status and bone markers

Current smokers had lower serum levels of osteocalcin than non smokers (table 4), while no difference was found in serum bone alkaline phosphatase (BALP) and renal hydroxyproline (OHP) excretion between the groups. The reduction in osteocalcin was modest (8.8%), but a true dose relationship was indicated by a significant negative linear relation between number of cigarettes per day and serum osteocalcin (β = −0.15; P < 0.001).

To explore potential mediators of or confounders to the effect of smoking on osteocalcin, alcohol intake, vitamin D status (serum 25-OHD), bone turnover (renal OHP excretion), and renal function (serum creatinine) were tested in univariate models (Table 5). All five variables were significantly correlated to serum osteocalcin. However, in a combined multivariate model considering all five variables, the significant inverse relation between smoking and osteocalcin remained unchanged (Table 5). Therefore, the effect of smoking on osteocalcin appears not to be mediated primarily through reduced vitamin D status in smokers, through changes in bone turnover rate, or through changes in renal function. Furthermore, we find no indication that the relation between smoking and osteocalcin be merely explained by concomitant alcohol intake.

Among all individuals serum levels of osteocalcin correlated inversely to BMD in the whole body and all regional measurements, accounting for 2.0–7.3% of the total variation in BMD. These negative correlations were found in all skeletal regions and in the subgroups of never smokers and current smokers. Concerning the spine, femoral neck, and total body the regression lines of serum osteocalcin versus BMD, differed significantly between current smokers and never smokers (P = 0.01 to <0.0001), but did not deviate from parallelism (P = 0.11–0.49). Thus in these sites, at any level of serum osteocalcin current smokers had lower BMD values than never smokers. In the radius the lines did not differ (P = 0.49) in accordance with the lack of association between smoking and radius BMD.


In this large and rather homogeneous group of perimenopausal women with a very high prevalence of current smoking we show a quantitative small but statistically highly significant impact of pre- and perimenopausal smoking habits on total body and femoral neck BMD. Furthermore, a causal relationship is corroborated by the linear relation between BMD in current, exsmokers, and never smokers. Like most other studies addressing the same age group we find no effect of smoking on forearm BMD.(26–29)

The effect of current smoking seen in these women is the combined effect of pre- and postmenopausal smoking. However, we do find very small but statistically significant bone mass deficiency also in those exsmokers who quit smoking before menopause and therefore express the effect of premenopausal smoking only. Furthermore, the women in our study are very recently menopausal (they have past median 0.5 years since their last bleeding) and we find no biochemical indications of a markedly increased current rate of bone loss in smokers. Consequently, we believe that the bone mineral values measured at the time of study inclusion mainly reflect the cumulated effect of premenopausal exposure. Essentially, this is in accordance with the recent metaanalysis, which finds the effect of premenopausal smoking statistically and clinically insignificant.(16) The statistical significance seen in our study may reflect the large number of smoking participants (n = 832), which exceeds the combined number of premenopausal smokers (n = 614) in the 10 studies included in the meta-analysis.

Table Table 4.. Biochemical Variables According to Current Smoking Status
 Current smokersCurrent nonsmokers 
 nMean (SD)nMean (SD)P (difference)
S-Osteocalcin (ng/ml)82416.2 (6.1)118017.7 (6.6)<0.001
S-LAP (U/liter)82569.7 (27.2)117869.0 (26.2)NS
U-OHP (μmol/mmol)82121.9 (9.0)116822.1 (8.6)NS
S-25-OHD (ng/ml)82524.1 (12.1)118125.8 (12.3)0.002
S-Creatinin (μmol/liter)82572.9 (9.7)117374.6 (9.4)<0.001
Table Table 5.. Determinants and Correlates of Serum Osteocalcin in Univariate and Combined Models
 Univariate β (P-value)Combined model β (P-value)
  1. The determination coefficient (R2) of the combined model was 0.15.

Cigarettes (number per day)−0.1119 (<0.001)−0.0890 (<0.001)
Alcoholic drinks (number per day)−0.5059 (<0.001)−0.4479 (<0.001)
S-Creatinin (μmol/liter)0.0874 (<0.001)0.0897 (<0.001)
S-25-OHD (ng/ml)0.0429 (<0.001)0.0284 (0.011)
U-OHP (μmol/mmol)0.2374 (<0.001)0.2450 (<0.001)

Our finding is significant from a causal point of view because the mechanisms by which smoking exerts it's effect on the skeleton are not yet fully understood. If the adverse effect of tobacco smoking on the skeleton were truly restricted to the postmenopausal state, we should focus on mechanisms that would be different in pre- and postmenopausal women. From our data, it appears that the difference between pre- and postmenopausal smoking is quantitative rather than qualitative and therefore attention also should be given to mechanisms that would be acting in pre- as well as postmenopausal women.

Because of the large number of participants in the present study and the extensive information collected at baseline, we were able to explore possible mechanisms by which the negative effect of tobacco smoking on BMD could be exerted. The smoking women went through a natural menopause 0.8 year before the nonsmoking women, thus indicating an impaired ovarian function. This has been described in other studies, but the mechanisms are not fully understood.(8,30) The early menopause per se does not explain the findings in the present group of women who are all recently menopausal but may reflect disturbed ovarian function during the premenopausal years.

The lower body weight in smokers compared with non-smokers is found in a majority of studies and could be regarded either as a confounder or as a mediator of the negative effect of smoking on bone mass.(2,6,30,31) The choice between these two views may significantly influence the statistical approach and subsequently the conclusions that are drawn.(17) From our data it appears that current smokers have slightly lower energy intake per day (2.8%) than never smokers. Other studies have shown an increased energy expenditure in smokers compared with nonsmokers.(32,33) Both explanations are in accordance with a causal effect of smoking on changes in body weight and support the view that reduced body weight is not a confounder obscuring the effect of smoking on bone. Hence, smoking could reduce BMD either through changes in body mass or by direct effects on the skeleton. However, in the present study smoking also affected BMD adjusted for body weight in the total body and femoral neck. This finding indicates that the reduction in body weight only explains part of the pathophysiology behind the bone loss in smokers and supports the existence of other mechanisms. The demonstrated interaction between smoking and fat mass on BMD also is in accordance with a more complex mode of action involving at least two mechanisms.

The small but significant negative effect of current smoking on vitamin D status has previously been reported in a subpopulation of this study and also has been shown in elderly men.(34,35) Although we only find minor differences the very clear dose dependence corroborates a causal relationship. Increased activity of liver enzymes induced by cigarette smoking is one potential explanation for the reduced vitamin D level. This mechanism has previously been discussed in relation to estrogen metabolism in smokers and parallels some of the effect of anticonvulsant drugs on vitamin D metabolism.(36,37)

In this group of perimenopausal women we found an inverse linear relation between current cigarette consumption and serum osteocalcin. This is in accordance with two previous Danish studies in premenopausal and postmenopausal women, respectively.(38,39) Vitamin D is a well-known stimulator of osteoblastic activity as expressed by production of osteocalcin.(40) In our study, smoking as well as serum level of 25-OHD were related to serum osteocalcin but did not substitute for each other in a multivariate analysis also including other explanatory variables like bone turnover, alcohol consumption, and renal function.(22,41,42) Therefore, we cannot conclude that the negative effect of smoking on serum osteocalcin is mediated through vitamin D status alone. However, the overall model for determining serum osteocalcin is rather weak (R2 = 0.15), lending little statistical power to test potential biological mechanisms.

Smoking was inversely related to serum osteocalcin. Moreover, smoking and serum osteocalcin were negatively related to BMD. These apparently paradoxical relations were partly resolved in the regression analyses showing that the inverse association between BMD and osteocalcin is maintained in smokers as well as nonsmokers but at different levels. At a given serum osteocalcin value a nonsmoker will have a higher typical BMD value than a smoker and therefore osteocalcin seems not to be a mediator of the effect of smoking on BMD. On the other hand, at any level of BMD smokers revealed lower serum osteocalcin levels than nonsmokers, indicating either a decreased production of osteocalcin or an enhanced degradation. In the multiple regression model the inclusion of renal function as estimated by serum creatinine levels did not influence the inverse association between smoking and osteocalcin. As we stated above this model does not permit too firm conclusions on mechanisms of action and we cannot exclude that the apparently increased renal filtration rate found in smokers is a potential mediator of the reduced osteocalcin level in this group.(43,44) Increased degradation of osteocalcin rather than decreased production would be in accordance with the lack of association between smoking and other biochemical markers of bone formation or degradation (BALP and OHPr) found in this study.

The observed negative relation between serum osteocalcin and BMD could reflect the effect of more recent variations in bone turnover on bone remodeling space and thereby bone loss because a similar relation is found between BALP and OHP versus BMD.(45) However, relations between cumulative long-term alterations in BMD and rather short-term variations in bone markers should be interpreted with great circumspection.

We conclude that the average cumulated effect of premenopausal smoking on bone is small, which may account for the majority of negative findings in previous smaller studies. The reduction in bone mass is not fully explained by the concurrent decrease in body mass, and a more complex relation between smoking, bone mass, and body mass is indicated by the protective effect of high fat mass against the deleterious effect of smoking on the skeleton. Several additional pathophysiological pathways may be involved in the reduction in bone mass. Our cross-sectional study has shown lower levels of vitamin D and osteocalcin. Whether these changes effect rate of bone loss has to be evaluated in longitudinal studies.


Participating centers include the following: Aarhus University Hospital: Professor L. Mosekilde (center leader), Dr. P. Charles, and Dr. A.P. Hermann.

Odense University Hospital: Professor H. Beck-Nielsen (center leader), Dr. J. Gram, Dr. T.B. Hansen, Dr. B. Abrahamsen, and Dr. E.N. Ebbesen.

Copenhagen Municipal Hospital: Dr. O.H. Soerensen (center leader), Dr. L.B. Jensen, and Dr. C. Brot.

Hillerød Hospital: Dr. S.P. Nielsen, (Center Leader) Dr. P. Eiken, and Dr. N. Kolthoff.

This investigation is carried out as a part of the DOPS, supported by the Karen Elise Jensen Foundation.