To evaluate the risk of venous thromboembolism (VTE) associated with the use of cyproterone acetate (CPA) amongst men with prostate cancer.
To evaluate the risk of venous thromboembolism (VTE) associated with the use of cyproterone acetate (CPA) amongst men with prostate cancer.
Using data from the General Practice Research Database, cases of VTE were identified amongst men with prostate cancer. Four controls with no evidence of a VTE were selected for each case. The time from diagnosis of prostate cancer to first hormonal treatment, and from first hormonal treatment to VTE, was compared for different treatments. Adjusted risk estimates for VTE were derived from further analysis using a nested case-control study design amongst all men with advanced prostate cancer, qualified by evidence of hormonal treatment.
The time between diagnosis and first treatment was significantly shorter for men first treated with CPA than for men first treated with a luteinizing hormone releasing hormone (LHRH) analogue (adjusted hazard ratio 1.33, 95% confidence interval, CI, 1.06–1.67). When the first treatment was CPA, the treatment-free period after diagnosis was significantly shorter for men who later had a VTE than for those who did not. The case-control study yielded an adjusted risk estimate for VTE amongst CPA users that was significantly higher than amongst men who were prescribed an LHRH analogue or who had had an orchidectomy (adjusted odds ratio 5.23, 95% CI 3.12–8.79).
There was a greater risk of VTE associated with CPA, which might be due to differences in disease severity between users of different products. The clinical significance of this finding is unclear.
European Organization for Research and Treatment of Cancer
incident rate ratio (age-adjusted)
hazard ratio (age-adjusted)
odds ratio (adjusted)
maximum androgen blockade
General Practice Research Database
Cyproterone acetate (CPA) is a steroidal antiandrogen licensed for the treatment of prostate cancer. In doses of 200–300 mg daily it is indicated for long-term palliative monotherapy or for use with LHRH analogues (LHRHa) to inhibit the tumour ‘flare’ that can occur during the first few weeks of treatment. Lower doses of CPA (50–150 mg daily) are recommended for managing the hot flushes associated with LHRHa therapy or orchidectomy. The British National Formulary states that there is a risk of recurrent thromboembolic disease in men prescribed CPA ; we could not find any published data to support a causal association.
Research conducted under the auspices of the European Organization for Research and Treatment of Cancer (EORTC) and based on randomized controlled clinical trials has suggested that the cardiovascular toxicity associated with CPA is much lower than for diethylstilbestrol (DES), medroxyprogesterone acetate or estramustine phosphate [2–5]. The EORTC trial comparing CPA with flutamide revealed that the difference in VTE rates between the groups was not statistically significant (relative risk 2.32, 95% CI 0.61–8.80) .
VTE is a common complication of malignant disease, especially in the first few months after diagnosis and in the presence of distant metastases [7–13]. Therefore investigating the association between CPA and VTE is complicated by the increased risk inherent with the condition for which CPA is being prescribed. Also, any dose-response association would be difficult to interpret because of the different indications for varying doses of CPA.
Guidelines on the management of prostate cancer recommend adjuvant hormone therapy for men with locally advanced disease who are to be treated with radical radiotherapy (RT), and for metastatic disease, in which case orchidectomy or LHRHa should be offered as the first-line treatment . Maximum androgen blockade (MAB, i.e. LHRHa or orchidectomy with continuous antiandrogen therapy) is seldom used, especially not with CPA, in view of the Committee on Safety of Medicine’s warning in 1995 about an increased risk of hepatotoxicity . The indication for CPA was restricted thereafter to short courses for flare prevention, treatment of hot flushes, and for long-term palliative treatment in men who have not responded to or are intolerant of other treatments.
In the present study, we aimed to estimate the risk of VTE associated with CPA in the treatment of prostate cancer. Particularly since 1995 and changes in the way CPA is prescribed, it is possible that patients receiving CPA have more advanced metastatic disease. As advanced disease is independently associated with an increase in VTE risk, one of the challenges was to establish whether any increase in VTE risk associated with CPA was drug- or disease-induced.
We used the General Practice Research Database (GPRD), the largest source of longitudinal data from general practice in the UK. Anonymised data are available for ≈ 40 million patient-years and the population is demographically similar to the general population . The validity and utility of this database for epidemiological research have been reported elsewhere [17–22]. The GPRD started collecting data from contributing practices in 1987, and from 1992 onwards most data are considered to have reached research standard (‘up-to-standard’). Diagnoses and symptoms are recorded using Oxford Medical Information System  or Read codes . Prescription data are almost complete, as virtually all prescriptions issued by the GPs are generated via the computer system . Referrals and admission to hospital are recorded, as are deaths and transfers from the practice. Smoking habits, height, weight, immunization, laboratory results and clinic attendance are recorded with some incompleteness that varies by practice.
The study was restricted to patients with ≥ 6 months of data between the patient registration date (or practice up-to-standard date, whichever was later) and the date of qualification for the study. Only data accrued after 31 December 1991 were used. The initial study population comprised all men with prostate cancer identified by specific diagnostic codes. Amongst all men with prostate cancer, we identified a sub-cohort of men with advanced disease, qualified by evidence of orchidectomy associated with prostate cancer and/or a prescription for one or more hormonal drugs. Exposure to CPA and other treatments was mapped using the date of prescription and dosing information. Whilst it is not possible to confirm full compliance with prescribed medication from the data available, it was possible to determine any individual’s likely drug exposure status at any given point in time.
Men with a diagnostic code for pulmonary embolism or deep vein thrombosis were identified and case status was confirmed by evidence of anticoagulant therapy or death from a cause consistent with VTE. A validation study on the GPRD indicated that using these criteria ensured that ≥ 84% of VTE cases are supported by other evidence . The ‘event date’ was taken as the date upon which the first symptoms were recorded. These symptoms included haemoptysis, shortness of breath, chest pain and swelling or redness of a limb. For each case of VTE, four controls were identified, matched on year of birth and event (index) date.
Age-adjusted incidence rate ratios (IRRadj) for VTE were calculated for CPA vs other current therapies. Amongst all cases and controls with prostate cancer, Kaplan–Meier survival estimates were derived for time from diagnosis of prostate cancer to first hormonal treatment/orchidectomy, and time from first treatment/orchidectomy to VTE or the last date of follow-up. Hazard ratios adjusted for age (HRadj) were calculated to compare the time to first treatment and VTE rates between men receiving different treatments.
A case-control analysis was performed in the sub-cohort of men with advanced disease. For each VTE case with advanced prostate disease, four matched controls were selected as before from the population of men with advanced disease. The records for all cases and controls were reviewed and various characteristics defined including: current use of hormonal therapy (on the index date or within the previous 28 days), history of VTE, smoking, alcohol consumption, body mass index, recent surgery and/or trauma, history of cardiovascular disease, recent use of opioid analgesia, diabetes, recent use of anticoagulants, orchidectomy, history of prednisolone use, the number of prescriptions issued in the 180 days before the index date (as a marker for general debility), the duration of advanced disease, the time between diagnosis of prostate cancer and first hormonal treatment and the time between diagnosis of prostate cancer and the index date. The presence of metastases was used as separate marker for disease severity, as was RT, after careful review of the records to establish whether RT had been given for local or advanced disease (based on the number of RT sessions, the time between sessions, symptoms and the indication for RT where given, e.g. bone pain). We identified men likely to have developed hormone resistance by analysing switches in treatment. Any switch after 3 months (e.g. from continuous LHRHa to continuous CPA) was used as an indicator for hormone resistance. Other alterations in treatment patterns, such as continuous LHRHa supplemented briefly by CPA 50 mg daily, were not considered switches, as this was more likely to signify a problem with hot flushes. Men prescribed oestrogens or prednisolone were considered to have developed hormone resistance. Switches from one antiandrogen to another were assumed to indicate tolerability problems and not hormone resistance. Men starting their treatment with MAB were assumed not to be hormone-resistant; this was assumed to be a difference in preference by the treating physician. Men prescribed CPA for LHRHa flare were classed as using CPA + LHRHa.
The study had a statistical power of 80% to detect a doubling of VTE risk associated with CPA at a confidence level of 95%. Odds ratios (ORs) with 95% CI were calculated, adjusted for covariates and interaction between variables using conditional logistic regression. The Scientific and Ethical Advisory Group of the GPRD gave approval for the study.
We identified 11 199 men with a diagnosis of prostate cancer; 229 had a VTE after the date of diagnosis (9.17 cases/1000 observed man-years); 33 (14%) of the VTE episodes were fatal. The rates of VTE were higher amongst men prescribed CPA (30.79 cases/1000 exposed man-years, EMY) or oestrogen (21.41 cases/1000 EMY) than the rates amongst men prescribed a nonsteroidal antiandrogen (nsAA) (8.89 cases/1000 EMY) or those who had had an orchidectomy or were prescribed an LHRHa (8.67 cases/1000 EMY) (Table 1). Amongst men prescribed CPA who had had an orchidectomy or were also prescribed an LHRHa the rate of VTE was 52.08 cases/1000 EMY.
|Exposure on date of VTE||Age, years||Cases||EMY||Rate/1000 EMY|
|LHRHa/orchidectomy + CPA||55–64||1||37||27.01|
The rates of VTE were significantly higher amongst men whose first treatment was CPA than amongst men whose first treatment was a nsAA (IRRadj 3.46, 95% CI 1.91–6.25), or those who had had an orchidectomy or were prescribed an LHRHa alone (IRRadj 3.35, 2.38–4.72). There was no significant difference between the rates of VTE amongst men first treated with CPA and those first treated with oestrogen (IRRadj 1.25, 0.60–2.60). The rate of VTE amongst men first treated with CPA alone was significantly lower than in those prescribed CPA who had had an orchidectomy or who were also prescribed an LHRHa (IRRadj 0.60, 0.37–0.97).
Amongst the 229 cases and 906 controls with prostate cancer, 831 men were prescribed a hormonal therapy or had an orchidectomy, and 267 (32%) were prescribed their first treatment on the same day as the prostate cancer was first recorded. Compared with men prescribed an LHRHa, those prescribed CPA had a significantly shorter period from diagnosis to first treatment (HRadj 1.33, 95% CI 1.06–1.67; Fig. 1a; Table 2); the time between diagnosis and first treatment did not differ significantly for other treatments.
|First treatment after diagnosis||Users n (%)||Users prescribed treatment > day 0, n (%)||HRadj (95% CI)|
|diagnosis to first treatment||first treatment to VTE|
|CPA*||259 (31.17)||155 (27.48)||1.33 (1.06–1.67)||2.76 (1.66–4.59)|
|nsAA||106 (12.76)||73 (12.94)||1.05 (0.79–1.39)||1.94 (1.02–3.67)|
|LHRHa||187 (22.50)||149 (26.42)||Ref||Ref|
|Oestrogen||17 (2.05)||13 (2.30)||1.15 (0.65–2.04)||3.47 (1.30–9.31)|
|Orchidectomy||55 (6.62)||41 (7.27)||1.12 (0.79–1.58)||2.32 (1.16–4.63)|
|CPA + LHRHa||145 (17.45)||92 (16.31)||1.17 (0.90–1.52)||2.35 (1.34–4.12)|
|LHRHa + nsAA||62 (7.46)||41 (7.27)||1.01 (0.71–1.43)||0.63 (0.21–1.85)|
The time to any hormonal treatment or orchidectomy was significantly shorter for cases than controls (HRadj 1.63, 1.31–2.03; Fig. 1b). This was apparent in men first treated with CPA (HRadj 1.82, 1.27–2.62), orchidectomy (HRadj 2.92, 1.26–6.78) or CPA + LHRHa (HRadj 2.19, 1.32–3.63) but not in men prescribed other treatments. Compared with LHRHa users, the time to VTE was significantly shorter amongst men first treated with CPA (HRadj 2.76, 1.66–4.59), CPA + LHRHa (HRadj 2.35, 1.34–4.12), oestrogen (HRadj 3.47, 1.30–9.31) or orchidectomy (HRadj 2.32, 1.16–4.63) (Fig. 1c; Table 2).
The survival analyses were repeated, stratifying the data by first treatment before or after 1995 to investigate the effects of changes in CPA prescribing practice. Using men first prescribed LHRHa as the comparison group, there was no significant difference in the time to first treatment with CPA before or after 1995 (HRadj 1.09, 0.77–1.55, and 1.26 0.87–1.82, respectively). However, the time to VTE amongst men first treated with CPA was significantly shorter before 1995 (HRadj 5.41, 1.94–15.10) but no different after 1995 (HRadj 1.72, 0.85–3.51).
In the nested case-control study, 7489 men with prostate cancer had evidence of advanced disease; 207 had a VTE (13.38 cases/1000 observed man-years, a rate significantly higher than amongst all men with prostate cancer, IRRadj 1.47, 95% CI 1.22–1.77). Matched unadjusted ORs indicated a significant association between VTE and current use of CPA alone (5.01, 3.14–7.97), CPA + LHRHa/orchidectomy (3.48, 1.96–6.19) or oestrogen (3.76, 1.44–9.83) compared with LHRHa/orchidectomy alone. There was also a significant association with a history of VTE (3.34, 1.88–5.94), recent surgery/trauma (10.94, 5.12–23.36), time since diagnosis of prostate cancer (P < 0.01), time since first treatment of advanced disease (P < 0.01), the number of prescriptions issued in the preceding 6 months (P < 0.01), a suggestion of hormone resistance (2.46, 1.65–3.65), and use of an anticoagulant before the index date (2.02, 1.17–3.48). There were no significant differences between cases and controls for other variables. The medical records of men treated with RT were reviewed but it was seldom possible to distinguish between palliative RT for bone pain and radical RT.
Conditional logistic regression analysis (Table 3) indicated an increased risk for VTE associated with current use of CPA alone (ORadj 5.23, 3.12–8.79), oestrogen (5.67, 1.99–16.18), CPA + LHRHa/orchidectomy (3.35, 1.74–6.45) and other combined therapy (2.07, 1.10–3.89) compared with LHRHa/orchidectomy alone. There was no difference in VTE risk between those exposed to CPA and those exposed to oestrogen (0.92, 0.31–2.78).
|Variable||Cases||Controls||ORadj (95% CI)|
|Exposure within 28 days before index date|
|LHRHa/orchidectomy + CPA||23||39||3.35 (1.74–6.45)|
|Other combined therapy*||17||60||2.07 (1.10–3.89)|
|Previous VTE||24||33||3.96 (2.05–7.68)|
|Recent surgery/trauma||27||13||12.69 (5.54–29.05)|
|Days from first treatment of advanced disease|
|Number of scripts issued in preceding 6 months‡|
Excluding cases and controls from the study population whose inclusion was based solely on their CPA use did not alter the risk estimates. Stratifying by hormone-resistance status rendered unstable data for those who were hormone-resistant; for those with no evidence of hormone resistance, compared with LHRHa exposure, the ORadj for CPA was 5.27 (95% CI 2.96–9.41) and 3.57 (1.54–8.30) for CPA + LHRHa.
We investigated whether the VTE risk associated with CPA changed with dosage or duration of use (Table 4). There was no significant difference in the various point estimates and the CIs were wide. Thus our data did not support the hypothesis that the risk of VTE observed amongst men prescribed CPA could be higher in the first months of use.
|CPA*||ORadj (95% CI)†|
|Daily dose, mg|
|Duration of use, days|
Stratifying by event year gave declining point estimates for CPA vs LHRHa over time, i.e. 7.64 (2.69–21.67) before 1995, 6.20 (2.28–16.90) in 1995 and 1996, and 4.52 (1.79–11.43) from 1997 onwards; none of the differences were statistically significant.
This study showed differences between hormonally treated groups in terms of time from diagnosis of prostate cancer to first treatment and time from first treatment to VTE; men who received CPA other than for flare were treated sooner after diagnosis than men who received an LHRHa alone. The case-control analysis amongst men with advanced prostate cancer indicated a significantly greater risk of VTE with the use of CPA monotherapy, oestrogen or CPA + LHRHa than with the use of an LHRHa alone. The analyses also confirmed an association between VTE and known risk factors. A causal association between CPA and VTE cannot be dismissed given the size of the point estimate (ORadj 5.23, 95% CI 3.12–8.79) and the shorter time to VTE amongst CPA users, especially before 1995. We interrogated the data to try to establish whether the association is causal or whether there is an alternative explanation. We did not find any evidence to support an increasing risk with increasing dosage of CPA, and no association between VTE risk and duration of use. There is some reported evidence that a causal association is biologically plausible  but also contradictory evidence . To our knowledge, no increase in VTE risk with CPA has been found in clinical trials, although the trial populations might have been too small to detect differences in VTE rates . There is some evidence to suggest an association between VTE and low-dose CPA in women [27,28] although concomitant ethinyloestradiol and confounding by indication obscure the association.
Amongst all men, irrespective of case status, the time from diagnosis of prostate cancer to first treatment was significantly shorter when that first treatment was CPA or CPA + LHRHa than with an LHRHa alone. This suggests that men prescribed CPA could be those with more advanced disease at diagnosis or more intolerable symptoms. To investigate this hypothesis, further analyses are required to differentiate between CPA used in MAB, treatment of hot flushes, flare prevention with LHRHa, and treatment of hormone-resistant disease. Interpretation of treatment patterns is further complicated because CPA prescribed to men through hospital pharmacies is not evident from GPRD records. The possibility that hospital prescribing could explain the observed differences in time from diagnosis to first treatment was rejected when we compared the time from diagnosis to first treatment for men prescribed CPA or CPA + LHRHa (mean 159 days) with that for men prescribed an LHRHa alone (mean 309 days).
Cancer is associated with an increased risk of VTE and this risk increases as the disease progresses. The increase in VTE risk found with CPA could be causal or it could arise from the way in which CPA is prescribed. Because CPA appears to be prescribed to men with more advanced disease any association between CPA and VTE might be confounded by disease severity.
Cases started treatment significantly sooner after diagnosis than controls, suggesting that they had more advanced disease. Amongst cases but not controls, time from diagnosis of prostate cancer to first treatment when that first treatment was CPA or an orchidectomy was significantly shorter than when the first treatment was a nsAA or LHRHa.
Our observations support the hypothesis that cases were in more urgent need of hormonal treatment than controls, and that an important difference in the cases (such as more advanced disease) might have contributed to the higher VTE risk associated with CPA than with LHRHa. Interestingly, the time to VTE after starting treatment with an LHRHa alone was significantly longer than the time to VTE after orchidectomy, even after adjusting for age (Table 3). Orchidectomy is biologically analogous to treatment with LHRHa so this difference in risk could be an effect of differences in disease severity between treatments, or could simply highlight the association between VTE and surgery.
Adjusting the VTE risk estimate for time since diagnosis to first hormonal treatment, in an attempt to adjust for disease severity, had no notable impact on any of the ORs. Nonetheless, our observations lead us to question whether CPA users are more ill than those receiving LHRHa and, consequently, whether it is justified to compare VTE risk associated with CPA with that associated with LHRHa. Since the Committee on Safety of Medicines’ warning on hepatotoxicity, the pattern of CPA use might be more akin to that of oestrogen, although this might not be the most appropriate comparator because of its well-established increase in VTE risk. We found no difference in VTE risk between CPA and oestrogen, but there were few oestrogen users.
In conclusion, we detected a greater risk for VTE associated with CPA that could be causal, although we cannot exclude the possibility that the increased risk is affected by residual confounding by disease severity. This might warrant further exploration using a randomized controlled trial. Given that CPA is already restricted to short-term use and long-term treatment in the palliative care of men with advanced disease, should the increase in VTE risk be confirmed independently, this relatively uncommon phenomenon must be weighed against the other adverse effects of oestrogen (in palliative care) and the more frequent side-effects of nsAAs in short-term treatment.
Funding: this work was funded by an unconditional research grant from Schering AG. None of the authors has any conflicts of interest that would prejudice the impartiality of this work.
None declared. Source of funding: see Acknowledgements (above).