Risks of cancer among a cohort of 23,935 men and women with osteoporosis
Version of Record online: 11 DEC 2007
Copyright © 2007 Wiley-Liss, Inc.
International Journal of Cancer
Volume 122, Issue 8, pages 1879–1884, 15 April 2008
How to Cite
McGlynn, K. A., Gridley, G., Mellemkjaer, L., Brinton, L. A., Anderson, K. C., Caporaso, N. E., Landgren, O. and Olsen, J. H. (2008), Risks of cancer among a cohort of 23,935 men and women with osteoporosis. Int. J. Cancer, 122: 1879–1884. doi: 10.1002/ijc.23290
- Issue online: 19 FEB 2008
- Version of Record online: 11 DEC 2007
- Manuscript Accepted: 1 OCT 2007
- Manuscript Received: 6 AUG 2007
- breast cancer;
- endometrial cancer;
- prostate cancer;
- colorectal cancer;
Low hormone levels among persons with osteoporosis may decrease risk of some cancers. Other osteoporosis risk factors, such as smoking and alcohol consumption, however, may increase risk. As these deleterious factors are more often associated with osteoporosis diagnosed prior to age 70 years, cancer risk may be higher in these younger persons than in the general population. To examine this hypothesis, a cohort study of 23,935 persons with osteoporosis was conducted in Denmark. Patients hospitalized with osteoporosis between 1978 and 1993 were identified in the Danish Inpatient Register. Linkage to the Danish Cancer Registry identified all cancer outcomes through 2003. Standardized incidence ratios (SIR) and 95% confidence intervals (95%CI) were calculated to compare cancer incidence in the cohort with that in the general population. Persons diagnosed prior to age 70 years were at increased cancer risk (women: SIR = 1.11, 95%CI = 1.04–1.19; men: SIR = 1.31, 95%CI = 1.13–1.50) due, in part, to increased risks of cancers of the buccal cavity, esophagus, liver, pancreas and lung. Persons diagnosed at ages 70 and older were at decreased risk (women: SIR = 0.91, 95%CI = 0.87–0.96; men: SIR = 0.89, 0.77–1.01) due, in part, to decreased risks of breast, endometrial, colon, rectal and brain cancers in women and prostate cancer in men. These results suggest that risk factors associated with earlier onset osteoporosis may be associated with increased risk of cancer. Conversely, factors associated with later onset osteoporosis may be related to a decreased risk of cancer. © 2007 Wiley-Liss, Inc.
Previous studies have reported decreased risks of breast and endometrial cancer among women with preexisting bone pathologies, such as low bone mineral density (BMD), fractures and osteoporosis.1–6 To a lesser extent, risk of prostate cancer in men has also been examined.7, 8 The collective findings suggest that while low levels of estrogen and testosterone increase the risk of bone pathology,9 they may also decrease the risk of hormonally-related cancers.4–8 Hormone levels, however, are not the only determinants of bone density.10 Other factors such as smoking, excessive alcohol consumption, lean body mass, low levels of physical activity and presence of other medical conditions also increase the risk of osteoporosis.11 These factors may be particularly important prior to the age (70 years) when senile osteoporosis becomes common in the population.12 After age 70 years, age-related processes such as lower levels of hormones, vitamin D and calcium are related to senile osteoporosis in a large group of both men and women.12 If earlier onset osteoporosis is more likely to be associated with risk factors such as smoking, drinking and low physical activity, it might be expected that persons with nonsenile osteoporosis would be at increased risk of cancer. In contrast, persons with senile osteoporosis might be expected to have a decreased cancer risk due to lower hormone levels. To examine this hypothesis, a cohort of 23,935 persons with a hospital diagnosis of osteoporosis was identified and traced forward for up to 26 years to determine the risk of developing cancer.
Material and methods
Data resources and patients
Since 1968 all Danish inhabitants have been assigned a unique personal identification number that is registered in the computerized Central Population Register. The number permits linkage of information between different registries. The population-based Danish Inpatient Register has maintained records of ∼99% of all hospitalizations in Denmark since 1977. For each hospitalization discharge, the Register contains the individual's Danish personal identification number, dates of admission/discharge, up to 20 inpatient diagnoses, and surgical procedures performed during hospitalizations. During the study period, diagnoses were classified according to the modified Danish version of the eighth revision of the International Classification of Diseases.
For our study, all patients with a registered diagnosis of osteoporosis (code 723.09) in the Danish Inpatient Register between 1978 and 1993 were identified. The records of the Danish Inpatient Register were linked to the records of the Central Population Register to retrieve information on date of death or date of emigration, where appropriate. To obtain information on subsequent cancer diagnoses, the Inpatient Registry records were linked to the Danish Cancer Registry, which has accurate and updated information on all cancer diagnoses in Denmark since 1943. Cancers are classified according to a modified version of the seventh revision of the International Classification of Diseases.
Our study had no access to risk factor data for the cohort. To estimate whether there were age-related differences in exposure to smoking and alcohol consumption, medical diagnoses were used as surrogates. Alcohol-related diagnoses in the Inpatient Register were used as a crude proxy measure of excessive alcohol consumption. These diagnoses included alcoholic psychosis (ICD-8 code 291), alcoholism (303), alcoholic steatohepatitis (571.10), Laennec's cirrhosis (571.09) and acute toxic over-consumption of alcohol (980.09). Emphysema (492) was used as a crude proxy variable for cigarette smoking. Institutional Review Board approval of our study was obtained from the NIH Office of Human Subjects Research.
All cancer diagnoses among the study cohort that occurred between the date of hospital discharge with a diagnosis of osteoporosis and the date of emigration, date of death or 31 December 2003 were determined. The observed number of cancer cases of each site was compared with the expected number of cases based on the age- and calendar year-specific incidence rates for males and females from the Danish Cancer Registry. Standardized incidence ratios (SIR) (observed number of cases divided by the expected number of cases) were computed with corresponding 95% CI, assuming Poisson distributions of the observed cases. To assure that osteoporosis preceded the diagnosis of cancer, only cancers diagnosed at least 1 year after the date of osteoporosis diagnosis were included in the analysis. All p-values were 2-sided. The analyses were conducted using SAS statistical software package, v. 9.1 (SAS, Cary, NC).
A total of 23,935 individuals (20,880 females, 3,055 males) with a diagnosis of osteoporosis were included in the cohort (Table I). The mean length of follow-up was 6.76 years (range = 1–23 years) among the women and 5.69 years (range = 1–22 years) among the men. Seventy-five percent of the women were diagnosed at age 70 years or older, while only 63.7% of the men were diagnosed at similar ages. The median number of diagnoses recorded for all 4 age and sex groups was 3. Osteoporosis was the primary diagnosis among 47.2% of the younger women and 47.3% of the older women. Among the men, osteoporosis was the primary diagnosis in 39.9% of the younger group and 41.6% of the older group. Diagnoses of emphysema (primary or otherwise) and of alcohol-related conditions (primary or otherwise) were related to age. Among the younger individuals, 6.8% of the women and 11.4% of men had any diagnosis of emphysema. The comparable percentages in the older participants were 2.1% among the women and 9.8% among the men. Alcohol-related diagnoses were also over represented among the younger persons with 5.4% of the women and 17.3% the men having an alcohol-related diagnosis. Among the older persons, only 1.0% of the women and 4.4% of the men had an alcohol-related diagnosis. The top 10 primary diagnoses after osteoporosis are shown in Table II.
|Women (n = 20,880)||Men (n = 3,055)|
|<70 yrs (n = 5,215)||≥70 yrs (n = 15,665)||<70 yrs (n = 1,108)||≥70 yrs (n = 1,947)|
|No. of diagnoses: median, (range)||3||(1–13)||3||(1–16)||3||(1–11)||3||(1–14)|
|Primary1 diagnosis-osteoporosis: n, %||2460||47.2%||7402||47.3%||442||39.9%||809||41.6%|
|Any2 diagnosis-emphysema: n, %||356||6.8%||331||2.1%||126||11.4%||190||9.8%|
|Any2 diagnosis-alcohol-related3: n, %||280||5.4%||159||1.0%||192||17.3%||86||4.4%|
|<70 yrs||≥70 yrs|
|4.8%||Fracture of the vertebral colum||5.5%||Fracture of the vertebral column|
|3.3%||Displacement of intervebral disc||2.0%||Polyarteritis nodosa|
|3.0%||Chronic bronchitis||1.7%||Partial fracture of the proximal femur|
|2.1%||Other nonarticular rheumatism||1.6%||Displacement of intervertebral disc|
|1.5%||Polyarteritis nodosa||1.5%||Other ill-defined and unknown causes|
|1.3%||Asthma||1.4%||Chronic arteriosclerotic heart disease|
|1.2%||Curvature of the vertebral column||1.4%||Non-osteoporosis bone diseases|
|1.1%||Non-osteoporosis bone diseases||1.3%||Other nonarticular rheumatism|
|5.6%||Chronic bronchitis||5.8%||Fracture of the vertebral column|
|5.0%||Fracture of the vertebral column||4.9%||Osteoarthritis|
|2.4%||Rheumatoid arthritis||1.9%||Symptomatic heart disease|
|2.2%||Displacement of intervertebral disc||1.8%||Chronic arteriosclerotic heart disease|
|1.9%||Non-osteoporosis bone diseases||1.5%||Non-osteoporosis bone diseases|
|1.8%||Other nonarticular rheumatism||1.4%||Vertebrogenic pain syndrome|
|1.7%||Alcoholism||1.2%||Acute cerebrovascular disease|
|1.1%||Cirrhosis||1.2%||Generalized ischaemic cerebrovascular disease|
In comparison with the general population, there was no overall significant difference in risk of cancer among women in the cohort (SIR = 0.98, 95% confidence intervals (95%CI) = 0.94–1.02) (Table III). The risks of cancers at certain sites did differ from the general population, however. As anticipated, women in the cohort had significantly decreased risks of breast (SIR = 0.71, 95%CI = 0.63–0.79) and endometrial cancers (SIR = 0.61, 0.46–0.79). In addition, women had significantly decreased risks of brain cancer (SIR = 0.66, 95%CI = 0.46–0.93) and melanoma (SIR = 0.72, 95%CI = 0.50–0.99). Unexpectedly, the women also had significantly decreased risks of colon (SIR = 0.84, 95%CI = 0.74–0.95) and rectal cancers (SIR = 0.60, 0.46–0.76). These decreased risks were counterbalanced by increased risks of cancers of the buccal cavity (SIR = 1.59, 95%CI = 1.19–2.09), liver (SIR = 1.62, 95%CI = 1.12–2.28), lung (SIR = 1.38, 95%CI = 1.22–1.55) and bladder (SIR = 1.26, 95%CI = 1.03–1.52) as well as multiple myeloma (SIR = 1.57, 95%CI = 1.14–2.10).
|All women||Women <70 yrs||Women ≥70 yrs|
|All malignant neoplasms||2,556||0.98||0.94–1.02||932||1.11||1.04–1.19||1,624||0.91||0.87–0.96|
|Buccal cavity and pharynx||52||1.59||1.19–2.09||18||1.79||1.06–2.84||34||1.50||1.04–2.10|
|Lung and trachea||268||1.38||1.22–1.55||152||1.69||1.44–1.99||116||1.10||0.91–1.32|
|Female genital organs||187||0.77||0.66–0.88||72||0.75||0.59–0.95||115||0.77||0.64–0.93|
|Brain and nervous system||33||0.66||0.46–0.93||15||0.74||0.41–1.22||18||0.61||0.36–0.96|
|Primary site unspecified||128||1.22||1.02–1.45||42||1.57||1.13–2.12||86||1.10||0.88–1.36|
The cancer risks stratified by age among women are also shown in Table III. The women diagnosed with osteoporosis prior to age 70 had a significantly increased risk of all cancer (SIR = 1.11, 95%CI = 1.04–1.19). Specific tumor types for which the risks were significantly elevated included cancers of the buccal cavity (SIR = 1.79, 95%CI = 1.06–2.84), esophagus (SIR = 2.13, 95%CI = 1.10–3.71), liver (SIR = 3.08, 95%CI = 1.79–4.93), pancreas (SIR = 1.77, 95%CI = 1.27–2.40), lung (SIR = 1.69, 95%CI = 1.44–1.99) as well as leukemia (SIR = 1.76, 95%CI = 1.18–2.53) and multiple myeloma (SIR = 1.91, 95%CI = 1.11–3.06). Risks of breast and endometrial cancer were significantly reduced (breast: SIR = 0.74, 95%CI = 0.61–0.88; endometrium: SIR = 0.64, 95%CI = 0.42–0.94). In contrast to the younger group, the older women had a significantly reduced risk of cancer overall (SIR = 0.91, 95%CI = 0.87–0.96). The lower risk was due, in part, to significantly lower risks of cancers of the breast (SIR = 0.69, 95%CI = 0.59–0.79), endometrium (SIR = 0.58, 95%CI = 0.39–0.83), colon (SIR = 0.80, 95%CI = 0.68–0.92), rectum (SIR = 0.56, 95%CI = 0.41–0.75) and brain (SIR = 0.61, 95% = 0.36–0.96). The only significantly increased risk was that of buccal cavity cancer (SIR = 1.50, 95%CI = 1.04–2.10).
Among the men, as among the women, the overall risk of cancer did not differ from that of the general population (SIR = 1.05, 95%CI = 0.95–1.16) (Table IV). However, the men did have significantly increased risks of lung cancer (SIR = 1.37, 95%CI = 1.08–1.72) and multiple myeloma (SIR = 2.12, 95%CI = 1.02–3.90), and a significantly decreased risk of prostate cancer (SI = 0.74, 95%CI = 0.54–0.98). Similar to women, men had increased, though nonsignificant, risks of cancers of the buccal cavity and liver.
|All men||Men <70 yrs||Men ≥70 yrs|
|All malignant neoplasms||415||1.05||0.95–1.16||205||1.31||1.13–1.50||210||0.89||0.77–1.01|
|Buccal cavity and pharynx||13||1.62||0.86–2.78||7||1.77||0.71–3.65||6||1.48||0.54–3.22|
|Digestive tract and peritoneum||86||0.94||0.75–1.16||33||0.96||0.66–1.34||53||0.92||0.69–1.21|
|Lung and trachea||76||1.37||1.08–1.72||44||1.69||1.22–2.26||32||1.09||0.75–1.54|
|Male genital organs||47||0.75||0.55–0.99||21||0.97||0.60–1.48||27||0.63||0.42–0.92|
|Brain and nervous system||7||1.20||0.48–2.47||5||1.54||0.50–3.60||2||0.77||0.09–2.79|
|Primary site unspecified||15||1.32||0.74–2.18||9||2.19||1.00–4.16||6||0.83||0.30–1.81|
The younger men were at increased risk of all cancer (SIR = 1.31, 95%CI = 1.13–1.50), while the older men were at nonsignificantly decreased risk (SIR = 0.89, 95%CI = 0.77–1.01) (Table IV). The same tumor sites that were increased in the younger women were increased in the younger men, though most risk estimates did not attain statistical significance. One exception was lung cancer, which was significantly increased (SIR = 1.69, 95%CI = 1.22–2.26). In addition, the younger men had significantly increased risks of nonmelanoma skin cancer (SIR = 1.53, 95% = 1.10–2.08) and Hodgkin disease (SIR = 8.82, 95%CI = 1.77–25.77), though the latter estimate was based on only 3 cases. Among the older men, the only risk that was statistically significant was that of prostate cancer (SIR = 0.62, 95%CI = 0.41–0.91). The reduced risks of colon (SIR = 0.81, 95%CI = 0.47–1.30) and brain (SIR = 0.77, 95%CI = 0.09–2.79) cancers, however, were similar to those of the older women though neither attained statistical significance.
In our study, the largest cancer follow-up ever conducted among persons hospitalized with osteoporosis, cancer risk varied by age at diagnosis of osteoporosis. Though both younger and older persons had decreased risks of hormonally related tumors, the risks of cancer at other sites differed. The overall increased cancer risk in the younger group was largely related to increased risks of alcohol and tobacco-related tumors. The decreased risk in the older individuals was associated with decreased risks of hormonally related tumors, and among women in particular, with decreased risks of tumors of the colon, rectum and brain.
The varying results by age suggest that osteoporosis in younger persons (under age 70 years) is more likely to be associated with deleterious risk factors than is osteoporosis diagnosed after age 70 years. In support of this, our study noted that diagnoses of emphysema and alcohol-related disorders were over represented among the persons less than 70 years of age. In addition, among the younger men, both alcoholism and cirrhosis were among the top 10 primary diagnoses.
Incidence of total cancer in relation to low BMD/fractures/osteoporosis has not been extensively examined. One small cohort study (n = 677) of women with distal forearm fractures4 was previously reported, though the results were not stratified by age. The study population was considerably younger than that of our study, with only 25% of the women being over age 70 years. As with our study, significantly reduced risks of total cancer (SIR = 0.73, 95%CI = 0.57–0.94), and specifically, of breast cancer (SIR = 0.54, 95%CI = 0.27–0.97) were reported. Though not statistically significant, a lower risk of endometrial cancer (SIR = 0.74) was also identified. Unlike our study, the cohort found no increased risk of cancer at any site, though the small number of cancers (n = 66) likely limited the study's power. In addition, it should be noted that although the conditions can be related, osteoporosis and wrist fracture are distinct clinical entities which may have differing risk factors. Particularly among younger women, there may be little overlap between the 2 conditions.
While no other study of total cancer outcomes has been reported, risks of specific tumors, especially breast and endometrial cancers, have been examined in targeted studies. The association between BMD/fractures/osteoporosis and breast cancer has been supported by the majority of these studies. Though 2 early reports13, 14 found no relationship between low BMD/fractures/osteoporosis and breast cancer, subsequent studies have concluded otherwise. Reduced risks of breast cancer have been reported among women with fractures in Sweden4, 5 and France2 and among 3 cohorts of women with low BMD in the U.S.1, 15–17 Similar results have been reported in cancer case-control studies in Australia18 and in the U.S.3 Examination of the U.S. NHANES I follow-up cohort also found a reduced risk of breast cancer in association with low BMD, although the association was not present in women under age 55 years.19 In accord with these findings, our study supported the relationship between breast cancer and low BMD/fractures/osteoporosis. As very few women in our study were under age 55 years, however, the relationship of osteoporosis to breast cancer risk in that age group could not be examined. The current results, among primarily postmenopausal women, are consistent with the view that estrogen levels are a more important determinant of breast cancer risk than is alcohol consumption, as increased alcohol consumption should have increased risk among women with osteoporosis. In fact, the reduced breast cancer risk in the current cohort likely underestimates the reduced risk that may be associated with lower hormone levels if excessive alcohol consumption was not a simultaneous exposure.
The well-described association between estrogen and endometrial cancer suggests that women with osteoporosis should be at lower risk of developing endometrial cancer. In addition, the leaner body mass associated with osteoporosis should decrease risk. As with breast cancer, the relationship between low BMD/fractures/osteoporosis and endometrial cancer has been supported by most studies. Persson et al. reported a lower risk of hip fracture among endometrial cancer survivors,6 which was supported by the authors' subsequent examination of endometrial cancer risk (SIR = 0.87) among women with hip fractures.5 The results of a small Japanese case-control study,20 a larger U.S. case-control study,3 a U.S. prospective study19 and the previously mentioned cohort study of all cancers4 also supported a reduced risk of endometrial cancer associated with low BMD. Previous research has also indicated that cigarette smoking may reduce the risk of endometrial cancer, particularly in postmenopausal women.21, 22 In line with this observation, the reduced risk of endometrial cancer since in our study may be the result of both low BMD and increase smoking exposure.
A possible relationship between osteoporosis and prostate cancer is suggested by several factors. Hypogonadism, characterized by low testosterone levels, is a known risk factor for osteoporosis, as are low calcium levels.23 As high levels of testosterone24 and calcium intake25 have been associated with the development of prostate cancer in some studies, an inverse relationship between osteoporosis and prostate cancer might be anticipated. Previous studies, however, have reported inconsistent results. No association between prostate cancer and BMD was reported by 2 studies; 1 using a case-control design26 and 1 using a cohort design.19 In contrast, 2 studies have reported significant results. Studying the Framingham cohort, Zhang et al. found that white men with high BMD were at greater risk of developing prostate cancer than white men with lower BMD.8 Using a cross-sectional design, Bunker et al.7 found that Afro-Caribbean men with prostate cancer had higher BMD than did control men. Notably, the association was present in men aged 60–79 years, but not in men aged 45–59 years. The stronger association among older men may explain the difference in findings among other studies. Of the 2 null studies, the mean age in one19 was only 49 years and the mean age in the other was unclear.26 In support of the age differences, our study also found the osteoporosis-prostate cancer association was only significant in men of age 70 years and above. As testosterone levels decrease with age, this finding related to age is not entirely surprising. It is also possible that the lower calcium levels in older men with osteoporosis would offer protection against prostate cancer. However, it has been speculated that higher calcium levels increase prostate cancer risk by reducing vitamin D levels.27 If this is the case, it is unlikely to be a relevant mechanism among men with osteoporosis, as osteoporosis is associated with both low calcium and low vitamin D levels.
Osteoporosis and colorectal cancer share several risk factors in common. Increased risks are associated with low vitamin D and calcium levels, low steroid hormone levels and low levels of physical activity, suggesting that the risks of the 2 conditions might be correlated.28 Thus, the finding in our study of significantly lower colorectal cancer risk among the women, particularly the older women, with osteoporosis was unexpected. The present findings are in contrast with 2 other studies that examined colorectal BMD/osteoporosis and colorectal cancer risk.19, 29 Contrary to our study, both previous studies reported decreased risks of colorectal cancer with higher BMD. Reasons for the inconsistency in results may be related to both age and weight of the various study populations. In both prior results, the populations were younger than in our study, and our study did not find a significant relationship among the younger women. In addition, in both prior reports, the mean body mass index (BMI) of the participants was greater than 25; a level considered to be overweight. In our study population, the women are likely to be leaner than the general population in that leanness is a risk factor for osteoporosis. As higher BMI as been reported to increase risk of colorectal cancer,30 older women with osteoporosis may be at decreased risk because of their leaner body mass. Although the relationship was not statistically significant, the risk of colon cancer was also decreased among men and was of the same order of magnitude as the risk among women.
Significantly higher risks of lung cancer in both the younger men and younger women were likely due to higher rates of cigarette smoking among persons with osteoporosis. Buccal cavity cancers, bladder cancer, pancreatic cancer and esophageal cancer are also related to cigarette smoking31 and were significantly increased in the younger women in the cohort. The risks of buccal cavity and esophageal cancers among men were of the same magnitude as the risks among women, suggesting that the risks might have been significant if the number of men had been equal to the number of women.
The increased risk of liver cancer may be related to excessive alcohol consumption.32 In addition, preexisting liver disease is a risk factor for bone loss23 and for the development of liver cancer. Excessive alcohol consumption is also associated with increased risk of buccal cavity cancers.33 The decreased risk of brain cancer, particularly notable in the older individuals, was an unanticipated finding. Previously reported associations between breast cancer and meningiomas, however, suggest that some brain cancers may have a hormonal etiology.34–36
The increased risk of multiple myeloma should be interpreted with caution as multiple myeloma is known to be associated with bone lesions, suggesting that multiple myeloma may have been present, but undetected, when the diagnosis of osteoporosis was made.37 However, there is emerging evidence that bone pathophysiology, as well as clonal myeloma cell growth, is a consequence of a complex interaction between myeloma cells and the bone marrow microenvironment.38, 39 Several cell types in the bone marrow, including stromal cells, osteoblasts, bone marrow leukocytes, endothelial cells, and osteoclasts have been found to interplay bidirectionally with myeloma cells through certain major biological pathways.40–44 In addition, in the related osteoporosis research field there is growing evidence that the same pathways are critical for the development and modulation of osteoporosis.38, 39, 45–48 This commonality of pathways suggests that osteoporosis might be a risk factor for the development of multiple myeloma. In our study, the observation of an increased risk of multiple myeloma fits well within this framework, and further suggests that a pre-existing bone resorption/formation imbalance could have an important role in the development of this cancer. If confirmed, future studies examining underlying mechanisms of the observed findings may provide insights into the etiology and pathogenesis of multiple myeloma.
In the older group of individuals, a conceivable explanation for the decreased risk of some tumors (i.e., brain, colon and rectal cancers) is that elderly persons with osteoporosis might not be worked up as intensively as elderly persons with no prior morbidities. If this were true, the older individuals might have increased risks of cancers of unspecified sites. This was not the case, however, as neither the older women nor the older men had significantly increased risks of unspecified cancers (Tables III and IV).
Our study had several major strengths including its very large size, population-based prospective design and inclusion of both men and women. A total of almost 24,000 participants, 3,000 of whom were male, made the study ∼4 times larger than the next largest investigation.19 In addition, our study examined all cancer outcomes, rather than just focusing on specific ones, had follow-up data for up to 26 years and was able to analyze the data by age. The sources of data were also major strengths in that all diagnoses, both of osteoporosis and cancer, were obtained from registries rather than from participant recall. A primary study limitation was that no risk factor data from the cohort were available and the proxy variables were derived from diagnoses which represented the extremes of the exposures. Because of multiple comparisons, some of the significant findings may represent chance variations. In addition, the osteoporosis diagnoses were drawn from the Danish Hospital Register, thereby only including persons that required hospitalization. In addition, as the cohort was ascertained based on any diagnosis of osteoporosis, it is likely that the participants have more comorbidities than persons with osteoporosis in the general population. Both of these factors make it likely that the study population is biased toward more severe cases of osteoporosis, suggesting that the conclusions should be extrapolated with caution to persons with less severe osteoporosis.
In conclusion, this very large study of hospitalized persons with osteoporosis found that cancer risk varied by age at diagnosis of osteoporosis. Though osteoporosis was associated with decreased risks of breast and endometrial cancer at all ages, those risks were counterbalanced, in the younger persons, by increased risks of smoking and alcohol-related tumors. Thus, persons hospitalized with osteoporosis prior to age 70 may benefit by more regular cancer screening, particularly for tumors associated with smoking and alcohol consumption.
The authors thank Ms. Emily Steplowski of IMS, Inc for invaluable help with programming and statistical analysis. The authors also thank Mr. Danny Carreon of the National Cancer Institute for help in manuscript preparation.
- 28Cancers of the colon and rectum. In: SchottenfeldD,FraumeniJF, eds. Cancer epidemiology and prevention,3rd edn New York: Oxford University Press, 2006. 809–29., .
- 31Tobacco smoke and involuntary smoking. IARC Monogr Eval Carcinog Risks Hum 2004; 83: 1–1438.
- 32Liver cancer. In: SchottenfeldD,FraumeniJF, eds. Cancer epidemiology and prevention,3rd edn. New York: Oxford University Press, 2006. 763–86., .
- 33Cancers of the oral cavity and pharnyx. In: SchottenfeldD,FraumeniJF, eds. Cancer epidemiology and prevention,3rd edn. New York: Oxford University Press, 2006. 674–96., , .