Fracture risk in Danish men with prostate cancer: a nationwide register study

Authors


Bo Abrahamsen, Department of Medicine, Copenhagen University Hospital Gentofte, Denmark. e-mail: b.abrahamsen@physician.dk

Abstract

This section opens with a nationwide Danish study on the risk of fractures in prostate cancer, as well as assessing the impact of exposure to androgen deprivation. The authors found that there was a marked increase in the risk of fractures, especially of the hip.

Authors from the USA address the outcome of cytoreductive nephrectomy for metastatic RCC, finding that it can be predicted by the fraction of tumour volume removed. They give an informative review of the topic and describe their experience in this type of surgery.

Treatment options for hormonerefractory prostate cancer are rather limited and authors from the USA present the results of a phase II trial into the use of gefitinib as a single therapy in patients with non-metastatic refractory disease. They found an absence of PSA response but confirmed the well-established favourable tolerability profile of this agent.

OBJECTIVE

To assess the risk of fracture attributable to prostate cancer, and the impact of exposure to prescribed gonadotrophin-releasing hormone agonists and antiandrogens on this risk in a nationwide, population-based case-control study.

PATIENTS AND METHODS

Data from the Danish National Hospital Discharge Register, the National Bureau of Statistics, and the National Prescriptions Database were merged. The analysis covered 15 716 men aged >50 years presenting with a fracture at any hospital in Denmark in 2000, and 47 149 age-matched control men. A previous diagnosis of prostate cancer had been recorded in 1.3% of controls and 2.5% of those with a fracture.

RESULTS

Prostate cancer was associated with an increased odds ratio (95% confidence interval) for all fractures of 1.8 (1.6–2.1), for hip fractures of 3.7 (3.1–4.4), but no increased risk of vertebral fractures. The increased fracture risk became apparent early after diagnosis and remained pronounced even in long-term survivors. Androgen deprivation therapy (ADT) with an odds ratio of 1.7  (1.2–2.5; P < 0.01) and orchidectomy, at 1.7 (1.2–2.4; P < 0.01) added to the overall fracture risk. In all, 3.1% of hip fractures in Danish men aged >50 years are attributable to prostate cancer.

CONCLUSION

Prostate cancer, orchidectomy and the use of ADT are associated with a markedly greater risk of fractures, especially of the hip. The risk of hip fracture is not confined to the very old, neither is the risk made negligible by the excess mortality in patients with advanced prostate cancer.

Abbreviations
ADT

androgen deprivation therapy

OR

odds ratio

NBS

National Bureau of Statistics.

INTRODUCTION

With a life-time risk of 17% in the USA, prostate cancer is one of the most frequent malignancies in men [1]. In many countries, including Denmark, the mortality rate has increased from 11.4/100 000 in 1943–47 to 38.2/100 000 in 1993–97 (world standard mortality rate) due to an increased incidence [2], and the disease remains a leading cause of death in men in Europe and the USA. Bone metastasis affects 80–90% of the patients who die from the disease [3] and increases morbidity due to fractures, pain and spinal compression. Androgen deprivation therapy (ADT) in the form of orchidectomy, GnRH agonists and antiandrogens have improved the prognosis in patients with advanced disease [4–7]. However, such treatment is associated with an accelerated bone loss within the first 6–12 months [8,9] and continued bone loss during long-term therapy [10,11]. Moreover, studies of prescription-claims databases have suggested that GnRH-agonist treatment is associated with a 1.5-fold greater risk of fracture and a 1.7-fold greater incidence of hospitalization due to fractures [12]. Also, orchidectomy is associated with a similarly increased fracture rate [13,14]. The combination of skeletal malignancy and adverse effects of treatment on bone makes fracture an important concern in these patients. However, it is difficult to estimate the burden of fractures in relation to prostate cancer from existing reports.

Thus we estimated the risk of fracture attributable to prostate cancer in a nationwide case-control study, and analysed the effect of exposure to GnRH agonists and antiandrogens on fracture risk.

PATIENTS AND METHODS

In this nationwide, population-based, epidemiological case-control study we used data from the Danish National Hospital Discharge Register (1977–2000), income and social status data (1999) from the National Bureau of Statistics (NBS) and information on redeemed prescriptions for ADT from the National Prescriptions Database (1995–2000). The analysis covered all men aged >50 years who were registered in 2000 as presenting with a fracture at any hospital in Denmark, and age-matched controls as described below. Key exposure variables were any diagnosis of prostate cancer from 1977 to the time of the index fracture, any fracture before the time of the index fracture, income band, and living alone or not. In the case of patients with prostate cancer, we only included as previous fractures those that occurred before prostate cancer had been diagnosed. The NBS provided the above data in an anonymous form, where social security numbers were replaced by a unique numeric code in a random sequence, ensuring that the identity of the study subjects was unknown to the investigators, while maintaining the unique link between the respective data sources.

Since 1977, the National Hospital Discharge Register has maintained a complete record of diagnoses and social security numbers for all inpatients in Danish hospitals. From 1995 onwards, outpatient diagnoses were also incorporated into the register. All patients who had sustained a fracture between 1 January 2000 and 31 December 2000 (124 655 men and women) were identified and the present analysis conducted for men aged >50 (15 716). The NBS also selected three random age- and sex-matched controls for each patient with fracture, using the incidence-density sampling technique (47 149 men).

In both controls and fracture cases we recorded the presence or absence of a diagnosis of prostate cancer between 1977 and 2000, preceding the date of the index fracture. Tentative diagnoses (i.e. diagnoses appended to an International Classification of Disease-10 Z03 diagnosis) were not counted as exposure to prostate cancer. Fracture diagnoses pre-dating the index fracture were counted as exposure in the form of ‘previous fracture’. Data on personal income and whether patients lived alone or with a spouse and/or children, were obtained for use as covariates in the analysis. Information on all prescriptions redeemed by controls and fracture cases in the period 1995–2000 was retrieved from the National Prescriptions Database for recording exposure to antiandrogens and GnRH agonists. This database contains information on all redeemed prescriptions in the country, but does not cover medications supplied to hospital departments by the hospital pharmacy. Presently, such medications are not traceable by social security number. To assess the proportion of directly issued ADT, we obtained information from the Danish Medicines Agency on the number of defined daily doses sold on prescription vs issued via hospital departments.

We used binary logistic regression analysis to obtain crude and adjusted odds ratios (ORs) for the effect of prostate cancer on fracture risk. An analysis stratified for the duration of prostate cancer was pre-specified to be done if there was a significant association between prostate cancer and overall fracture risk. Unadjusted ORs were used as relative risk estimates in the calculation of attributable risk and impact numbers [15].

The study did not require approval by the ethics committee system but by the Data Protection Authority. Further, the use of highly detailed but anonymous ‘microdata’ from the discharge register and the prescriptions database requires specific permission from the National Board of Health, and data access is subject to approval from the NBS. The study was approved by the appropriate authorities and permission granted to access datasets which had previously been produced for use in epidemiological studies of fracture risk [16].

RESULTS

The analysis comprised 15 716 men with fractures and 47 149 matched controls (Table 1). A previous diagnosis of prostate cancer had been recorded in 1.3% of controls and 2.5% of fracture cases, with a median interval between the first consultation for prostate cancer and the index fracture of 3 years. In 75% of the patients the interval was <5 years. Compared with controls, fracture cases had a significantly higher prevalence of previous prostate cancer and previous fractures, and were more likely to live alone or be in a lower income band (Table 1).

Table 1.  Basic data on fracture cases and controls
VariableFracture casesControlsP
Number15 71647 149 
Mean (sd) age, years66.8 (12.5)66.8 (12.5)0.99
Prostate cancer, % 2.5 1.3<0.001
Mean (sd) time from 1st cancer consultation, years 4.2 (4.5) 4.2 (4.8)0.89
ADT, % 0.5 0.2<0.001
Orchidectomy, % 0.9 0.3<0.001
Previous fracture, %35.712.5<0.001
Single living39.228.0<0.001
Income band (% in interquartile range)  <0.001
1st13.513.8 
2nd41.734.5 
3rd22.923.5 
4th21.926.6 

Prostate cancer was associated with an increased OR (95% CI) for all fractures of 1.8 (1.6–2.1), and in particular for hip fractures, with an OR of 3.7 (3.1–4.4). There was no association with the risk of vertebral fracture (Table 2). For hip fractures, the unadjusted excess risk was particularly pronounced in men aged 50–65 years, where the OR was 9.2 (4.8–17.5), but the risk was increased at all ages including >80 years, where the OR was 1.7 (1.3–2.1). The risk of all fractures was doubled in all age categories. Several demographic covariates influenced the fracture risk, in particular age, previous fracture, size of personal income, and whether the patient lived alone or with spouse/children at the time of fracture. Controlling for these covariates increased the strength of the association between prostate cancer and fracture. By matched analysis within the control group, there was a record of hospital-treated fractures before the diagnosis of prostate cancer (or the same interval in men with no prostate cancer) in 9.5% of patients with prostate cancer in the study and in 10.2% of men with no diagnosis of prostate cancer (chi-square 0.19, P = 0.66). Stratifying the analysis (Table 2) for the time since diagnosis showed a uniformly increased fracture risk irrespective of the duration of diagnosed disease.

Table 2.  An analysis of the influence of prostate cancer on the risk of subsequent fracture; case-control analysis. All men aged >50 years who were diagnosed with fractures (15 716) at any hospital service in Denmark in 2000, and matched controls (47 149). There was no significant effect of time since the prostate cancer diagnosis on the risk of fracture by multivariate analysis controlled for age and previous fracture status (P = 0.94, not shown)
Type of estimateOR (95% CI) with a history of prostate cancer
All fractureHip fracturesSpine fractures
  • *

    All included covariates were significant predictors of overall fracture risk

Number15 71633741411
Analysis by age at the time of fracture, controlled for covariates
Unadjusted for covariates
Age, years:
>50 (total)1.8 (1.6–2.1)3.7 (3.1–4.4)1.0 (0.7–1.5)
50–651.8 (1.1–2.8)9.2 (4.8–17.5)0.6 (0.1–4.4)
65–801.9 (1.6–2.3)2.5 (2.0–3.3)0.9 (0.4–1.6)
>801.8 (1.5–2.2)1.7 (1.3–2.1)1.3 (0.7–2.4)
Adjusted for age
>50 (total)1.8 (1.6–2.1)2.1 (1.8–2.5)1.0 (0.7–1.6)
50–651.8 (1.1–2.8)6.6 (3.4–12.7)0.6 (0.1–4.2)
65–801.9 (1.6–2.3)2.2 (1.7–2.9)0.9 (0.5–1.6)
>801.8 (1.5–2.2)1.7 (1.3–2.1)1.3 (0.7–2.4)
Adjusted for age and previous fracture history
>50 (total)2.1 (1.9–2.5)2.5 (2.1–2.9)1.1 (0.7–1.7)
50–652.1 (1.4–3.3)8.7 (4.8–16.1)0.7 (0.2–1.1)
65–802.1 (1.8–2.5)2.6 (2.1–3.2)0.9 (0.5–1.7)
>802.0 (1.7–2.4)2.0 (1.7–2.5)1.4 (0.8–2.4)
Adjusted for age, prior fracture, single living and income band*
>50 (total)2.2 (1.9–2.5)2.4 (2.0–2.9)1.1 (0.7–1.6)
50–652.2 (1.4–3.4)8.4 (4.6–15.5)0.7 (0.2–4.4)
65–802.2 (1.9–2.7)2.5 (1.7–3.2)0.9 (0.5–1.7)
>802.1 (1.8–2.5)1.9 (1.3–2.4)1.4 (0.9–1.6)
Analysis by duration of disease at the time of fracture
Time since diagnosis, years
≤ 21.6 (1.3–1.9)2.0 (1.6–2.6)0.8 (0.4–1.6)
>2 and ≤42.0 (1.5–2.6)1.9 (1.4–2.8)1.6 (0.8–3.4)
>4 and ≤61.8 (1.4–2.6)1.7 (1.1–2.7)0.6 (0.2–2.5)
>62.0 (1.6–2.5)2.3 (1.7–3.0)1.3 (0.7–2.7)

Of the 1018 men with prostate cancer, 198 (19%) had been treated by orchidectomy, 89 (9%) prescribed ADT and 36 (4%) both orchidectomy and ADT before the time of the index fracture. Patients who had had orchidectomy or had filled at least one prescription for antihormonal medications in the period 1995–2000 up to the date of the index fracture had a greater risk of fracture, which was additive to the risk associated with the disease itself. Thus, when adjusted for prostate cancer, age and previous fracture, both ADT, with an OR of 1.7 (1.2–2.5; P < 0.01) and orchidectomy, with an OR of 1.7 (1.2–2.4; P < 0.01) added to overall fracture risk, and in the case of ADT also to the risk of hip fracture, at 1.9 (1.2–3.0; P < 0.05). By contrast, the risk of spine fractures depended on previous fracture, with an OR of 2.1 (1.2–4.9; P < 0.01), but not on exposure to prostate cancer therapy.

Despite the association in the present study between prescribed ADT and increased fracture risk, only 15% of ADT doses in the country had been dispensed through prescriptions, while the remainder were issued by hospital departments and not traceable at patient level.

To calculate the effect of prostate cancer on the incidence of hip fractures in Denmark, we obtained national age-specific hip-fracture rates and data on the incidence and prevalence of prostate cancer (Table 3). Superimposing the ORs from the present study on the prevalence data implied that 3.1% of hip fractures in Danish men aged >50 years are attributable to prostate cancer, either to the disease itself or its therapy. Annually, 1 in 11 000 men aged >50 years sustained a hip fracture as a result of prostate cancer.

Table 3.  Epidemiological background information
VariableValue
  • *

    Approximated by unadjusted study OR, for hip fracture as rare outcome; men aged >50.

Prevalence of prostate cancer in Danish men aged >50 years, %     0.85
Total hip fractures in Danish men aged >50, per year, n 3 374
Study findings
Estimated relative risk of hip fracture with a history of prostate cancer*     3.7
Impact values
Population attributable risk     9.3 × 10−5
Population impact number10 710
Total hip fractures in Danish men aged >50 years attributable to prostate cancer, per year   106
Fraction of hip fractures in Danish men aged >50 years attributable to prostate cancer, %     3.1

DISCUSSION

The present study shows that the overall risk of fractures was doubled in men with prostate cancer. This increase in risk was apparent early after diagnosis and remained pronounced even in long-term survivors. The risk of hip fracture was increased at all ages, but most strikingly in men aged 50–65 years, where hip fractures were eight times more likely than expected from age and previous fracture history.

By contrast, we found no increase in the risk of vertebral fractures in patients with a history of prostate cancer. This was unexpected, in view of the high prevalence of such fractures in male osteoporosis [17], but not when considering the subclinical presentation of most vertebral fractures, the restrictions in access to radiography in patients with back pain, and the ensuing high prevalence of undiagnosed vertebral deformities in men [17–19]. An increased risk of vertebral fractures after ADT has been reported in one large claims database study [12], while another recent claims study found no such increase [20]. Differences in the threshold for obtaining spinal X-rays might contribute to this apparent discrepancy.

The progression of advanced prostate cancer depends to a great extent on stimulation by androgens; thus ADT remains best current practice in advanced disease. This is accomplished either through orchidectomy or now more commonly through use of androgen antagonists or GnRH agonists. A recent cohort study showed a dose-dependent increase in the risk of fracture in relation to GnRH agonists [12]. In that study, administration of more than nine doses of GnRH agonists during the first year of prostate cancer was associated with an 1.45-fold increase in fracture occurrence [12]. This risk was similarly increased in patients treated with orchidectomy [12]. An increased risk of fracture was also reported in a retrospective study, although there were few participants [21]. However, exposure to ADT is also a marker of more advanced disease. Indeed, the risk of fracture during the previous 12 months before ADT was significantly increased [12]. Nevertheless, osteoporosis rather than skeletal metastases appears to play the major role, with 85% of fractures being osteoporotic and not metastatic [13,22].

Both the present study and others [12,20] indicate that fracture risk is further increased in patients with prostate cancer when they receive ADT or have an orchidectomy. Thus, many fractures might be preventable. Several lines of therapy might attenuate or abolish the ADT-induced bone loss, including bisphosphonates (pamidronate, chlodronate and zolendronic acid) [23–27], oestrogen [28], selective oestrogen-receptor modifiers (raloxifene and toremifene) [10,29], endothelin-1 antagonists, selective matrix metalloproteinase inhibitors [30], and denosumab.

While prostate cancer increases the skeletal morbidity in the individual patients, the impact on society is modest. The present study indicates that the disease is only responsible for 3% of the hip fractures in Danish men aged >50 years. Our analyses provided no evidence of increased fracture risk before the time of the diagnosis of prostate cancer. Clinical trials in male osteoporosis indicate that with appropriate treatment this risk could probably be halved. Our findings indicate that, for the individual patient, the fracture risk is markedly increased after prostate cancer is diagnosed, and the risk remains elevated even in the long term. Therefore, hip- fracture risk is not simply a feature of prostate cancer in the very old, neither is the risk of hip fracture made negligible by the excess mortality in this group of patients.

Our study has an important limitation which is shared by all case-control studies, i.e. that we cannot provide direct estimates of rates of fracture or mortality rates. In effect, the ORs provided reflect the excess risk of fracture minus the excess risk of death. However, the strength of the study design is that it captures all hospital-treated fractures in the country in the year of sampling. This allowed us to estimate the effect of prostate cancer on the total number of fractures in Danish men. Complete information on orchidectomy was available but there were limitations to the pharmaco-epidemiological part of the analysis. We have very comprehensive information on all prescriptions redeemed by cases and controls, but this accounts for only a small proportion of their exposure to ADT, as a very large proportion of such medications were provided directly by hospital departments. Therefore, many patients would have unrecorded iatrogenic hypogonadism. This will dilute the impact and reduce the apparent OR for ADT on fractures, but the overall risk estimates for prostate cancer on fractures stated in this analysis are unaffected by under-capturing the exposure to ADT, as are the risk estimates for orchidectomy. Taken together, these findings indicate a very markedly increased fracture risk in prostate cancer and an additional risk associated with ADT, a risk which is likely to be underestimated in the present analysis. It is logical that a subset of patients with less advanced disease have a much lower risk of fracture, but we have insufficient information to assess the magnitude of this risk. It is likely that the effect of prostate cancer on the total number of fractures will be greater in countries with a higher prevalence of prostate cancer than that seen in Denmark, but differences in diagnostic approach, with detection of disease at a less advanced stage, might increase the prevalence (through extended survival) while at the same time reducing the attributable risk due to smaller effect on bone health.

We conclude that prostate cancer is associated with a 1.8-fold greater risk of any fracture and a 3.7-fold greater risk of hip fractures. When adjusted for previous fracture, prostate cancer and age, any use of ADT was associated with an OR of 1.7 and 1.9 for all fractures and hip fracture, respectively.

CONFLICT OF INTEREST

None declared.

Ancillary